(NATO Science for Peace and Security Series B_ Physics and Biophysics) Taras N. Rudakov (Auth.), Tomaž Apih, Bulat Rameev, Georgy Mozzhukhin, Jamie Barras (Eds.)-Magnetic Resonance Detection of Explos

NATO Science for Peace and Security Series - B:
Physics and Biophysics
Magnetic Resonance
Detection of Explosives
and Illicit Materials
Edited by
Toma Apih
Bulat Rameev
Georgy Mozzhukhin
Jamie Barras
AB 3
Magnetic Resonance Detection of Explosives
and Illicit Materials
NATO Science for Peace and Security Series
This Series presents the results of scientific meetings supported under the NATO
Programme: Science for Peace and Security (SPS).
The NATO SPS Programme supports meetings in the following Key Priority areas:
(1) Defence Against Terrorism; (2) Countering other Threats to Security and (3) NATO,
Partner and Mediterranean Dialogue Country Priorities. The types of meeting supported
are generally Advanced Study Institutes and Advanced Research Workshops. The
NATO SPS Series collects together the results of these meetings. The meetings are
co-organized by scientists from NATO countries and scientists from NATOs Partner or
Mediterranean Dialogue countries. The observations and recommendations made at the
meetings, as well as the contents of the volumes in the Series, reflect those of participants
and contributors only; they should not necessarily be regarded as reflecting NATO views
or policy.
Advanced Study Institutes (ASI) are high-level tutorial courses to convey the latest
developments in a subject to an advanced-level audience
Advanced Research Workshops (ARW) are expert meetings where an intense but
informal exchange of views at the frontiers of a subject aims at identifying directions for
future action
Following a transformation of the programme in 2006 the Series has been re-named and
re-organised. Recent volumes on topics not related to security, which result from meetings
supported under the programme earlier, may be found in the NATO Science Series.
The Series is published by IOS Press, Amsterdam, and Springer, Dordrecht, in conjunction
with the NATO Emerging Security Challenges Division.
Sub-Series
A. Chemistry and Biology Springer

B. Physics and Biophysics Springer
C. Environmental Security Springer
D. Information and Communication Security IOS Press
E. Human and Societal Dynamics IOS Press
http://www.nato.int/science
http://www.springer.com
http://www.iospress.nl
Series B: Physics and Biophysics

Magnetic Resonance Detection
of Explosives and Illicit Materials
edited by
Toma Apih
Joef Stefan Institute
Ljubljana, Slovenia
Bulat Rameev
Gebze Institute of Technology
Gebze, Turkey
Georgy Mozzhukhin
Kazan State Power Engineering University
Kazan, Russia
and
Jamie Barras
Kings College London
London, UK
123
Published in Cooperation with NATO Emerging Security Challenges Division
Proceedings of the NATO Advanced Research Workshop on
Magnetic Resonance Detection of Explosives and Illicit Materials
Izmir, Turkey
26 September 2012
Library of Congress Control Number: 2013949690
ISBN 978-94-007-7267-0 (PB)

ISBN 978-94-007-7264-9 (HB)
ISBN 978-94-007-7265-6 (e-book)
DOI 10.1007/978-94-007-7265-6
Published by Springer,
P.O. Box 17, 3300 AA Dordrecht, The Netherlands.
www.springer.com
Printed on acid-free paper
All Rights Reserved

Springer Science+Business Media Dordrecht 2014
This work is subject to copyright. All rights are reserved by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation,
reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms
or in any other physical way, and transmission or information storage and retrieval,
electronic adaptation, computer software, or by similar or dissimilar methodology
now known or hereafter developed. Exempted from this legal reservation are brief
excerpts in connection with reviews or scholarly analysis or material supplied
specifically for the purpose of being entered and executed on a computer system,
for exclusive use by the purchaser of the work. Duplication of this publication or parts
thereof is permitted only under the provisions of the Copyright Law of the Publishers
location, in its current version, and permission for use must always be obtained from
Springer. Permissions for use may be obtained through RightsLink at the Copyright
Clearance Center. Violations are liable to prosecution under the respective Copyright
Law.
The use of general descriptive names, registered names, trademarks, service
marks, etc. in this publication does not imply, even in the absence of a specific
statement, that such names are exempt from the relevant protective laws and
regulations and therefore free for general use.
While the advice and information in this book are believed to be true and accurate
at the date of publication, neither the authors nor the editors nor the publisher
can accept any legal responsibility for any errors or omissions that may be made.
The publisher makes no warranty, express or implied, with respect to the material
contained herein.
Preface
In their preface to a previous volume in this series (Explosive Detection using

Magnetic and Nuclear Resonance Techniques, 2009), editors Profs. Fraissard and
Lapina laid out the potential advantages and recognised challenges of the application
of magnetic resonance techniques including Nuclear Magnetic Resonance (NMR)
and Nuclear Quadrupole Resonance (NQR) spectroscopies to the detection of
explosives [1, 2] in a world where individuals and groups were employing an ever-
widening range of homemade explosive materials for use in so-called improvised
explosive devices (IEDs). The key advantages of magnetic resonance methods are
that they employ radiofrequency waves that are chemically-specific, non-invasive
and non-destructive and that are able to pass through many materials including
woods, plastics, fabrics and soils. In the case of NQR, the key challenge is the low
sensitivity low signal-to-noise ratio making the highly-specific signals hard to
pick out amongst the background radiofrequency noise, particularly in unshielded
environments. In the case of NMR, the challenges are the need to incorporate a
magnet in the detection system, with inherent issues with weight and the effect of
magnetic fields on electronic equipment; and the need to differentiate threat from
benign liquids.
In the years since the publication of that volume, the number of attacks and
diversity of materials and methods employed have, alas, only increased. There has
been continuing violence in places such as Iraq, Afghanistan and the Caucasus;
and prominent isolated attacks and attempted attacks elsewhere in the world
including the underwear bomber (explosive: PETN; December 2009), the Times
Square car bomb (explosives: black powder and urea; May 2010), the printer
bombs (explosive: PETN; October 2010), the Domodedovo Airport suicide attack
(explosive: TNT; January 2011), right up to the Boston bombings of this year
employing pressure cookers (explosive: black powder; April 2013). At the same
time, other developments, such as the ban on air passengers carrying liquids
containers larger than 100 ml in their carry-on luggage being due to be lifted in
the next year, have focussed the attention of equipment makers and legislators both
on the need to deploy non-invasive technologies for this, specific application.
v
vi Preface
So how have magnetic resonance methods progressed towards meeting the

challenges laid out by Profs. Fraissard and Lapina in their preface to the 2009
volume in the face of the still-diversifying threat and other developments? This
current volume presents a snapshot of progress in the application of NQR to
the detection of solid explosives and the application of NMR to the detection of
liquids explosives; with papers covering other techniques, for example Terahertz
spectroscopy (THz), that illustrate parallel efforts in similar fields.
The NQR papers deal directly with the issue of signal-to-noise and can roughly
be divided into two types: those concerned with increasing the strength of the signals
generated (Rudakov and Mozzhukhin et al.) and those concerned with increasing the
specificity of the technique through a better understanding of the factors that affect
the quadrupolar parameters of specific formulations of explosives (Rowe et al.;
Malone and Sauer; Sinyavsky et al.; Jover et al.). There is in addition an important
paper on the application of signal processing techniques that make use of these
specific quadrupolar parameters to aid in situations where signal-to-noise is poor
(Butt et al.). This focus serves to illustrate the recognition in the NQR community of
the paramount importance of improving signal-to-noise and discrimination, with the
latter of particular importance as the range of threat materials grows the spectrum
of NQR responses from explosives is a much more crowded place in 2013 than it
was just 10 years ago.
The NMR papers are united in their focus on the issue of differentiating threat
liquids from benign liquids, although the approaches taken are diverse, ranging from
a method that uses a multi-element NMR (1 H, 14 N) approach (Prado) to those that
focus on characterising different time constants of the 1 H signal alone (Austin et al.;
Konov et al.) and consider different configurations for capturing these different 1 H
time constants (Gradiek et al.). These papers particularly the Prado and Austin
et al. papers incidentally also illustrate a trend for the development of more
compact systems that has overtaken NMR for the detection of explosives in recent
years; a recognition that, any system that hopes to be deployed in an environment as
crowded as an airport security lane needs to have as small a footprint as possible.
This volume is based on NATO Advanced Research Workshop Magnetic
resonance detection of explosives and illicit materials MRDE-2012 and Editors
acknowledge the support of NATO by ARW grant 984375. Additionally, many of the
authors would like to acknowledge the NATO support by Science for Peace grants
978007 A Quadrupole Resonance Instrument for the Clearance of Abandoned
Minefields, and 982836 Highly Sensitive NQR/NMR Technique for Explosive
Detection.
While working on the current volume, the editors received the sad news that
Prof. John A S Smith had passed away. John Smith was a leading light in the
NQR community for nearly 50 years, forming his first NQR research group at
the University of Warwick, UK, in 1965. There, he co-invented (with his then
research student, David Tong) a Nuclear Quadrupole Resonance Spectrometer,
which was manufactured, and exported, by Decca Radar Instruments. In 1972,
shortly after moving to Queen Elizabeth College (QEC), London, John organised the
first International symposium on NQR. In 1985, following a merger, QEC became
Preface vii
part of Kings College London, and it was there that John spent the whole of the
rest of his career. In the course of that career, he was the co-author of more than
200 papers and reviews in the scientific and technical literature, and joint holder of
16 patents. He was also, for many years, the editor of the Advances in Quadrupole
Resonance series published by Wiley & Sons. He maintained a keen interest in mine
and IED detection technology and airport security scanning technology, as can be
evidenced from his publication in this current volume, the last that he completed
before his death. John was a true gentleman as many people who met him can attest.
All his life, he was dedicated to Chemical Physics and at leisure, to his wife and
daughter, sharing his love of opera and art culture with them.
The 2013 Magnetic Resonance for the Detection of Explosives Workshop
(MRDE2013), held at Kings College London, 812 July 2013, was dedicated to
Johns memory.
London, UK Jamie Barras

Kazan, Russia Georgy Mozzhukhin
Gebze, Turkey Bulat Rameev
Ljubljana, Slovenia Toma Apih
References
1. Gradiek A, Apih T (2010) NMR-based liquids explosives detector. Appl Magn Reson 38:
485493
2. Osn TM, Cerioni LMC, Forguez J, Olle JM, Pusiol DJ (2007) NQR: from imaging to explosives
and drugs. Physica B 389:4550
Contents
Part I Nuclear Quadrupole Resonance Detection of Solids
1 Further Improvement of NQR Technique for Detection

of Illicit Substances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Taras N. Rudakov
2 An Overview of NQR Signal Detection Algorithms . . . . . . . . . . . . . . . . . . . . 19
Naveed R. Butt, Erik Gudmundson, and Andreas Jakobsson
3 Nuclear Quadrupole Resonance of Pentaerythritol
Tetranitrate (PETN) in Different Compositions . . . . . . . . . . . . . . . . . . . . . . . . 35
Michael D. Rowe, John A.S. Smith, Jamie Barras,
and Kaspar Althoefer
4 Cross-Relaxation Enhanced NQR of Ammonium Nitrate
in Low Magnetic Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
Georgy V. Mozzhukhin, Bulat Z. Rameev,
Galina S. Kupriyanova, and Bekir Aktas
5 Investigating Homonuclear Broadening in NQR with
Carr-Purcell Meiboom-Gill Performed on p-Chloroaniline . . . . . . . . . . . 61
Michael W. Malone and Karen L. Sauer
6 Size Effect in 14 N Nuclear Quadrupole Resonance Spectroscopy . . . . . 69
Nikolay Sinyavsky, Georgy V. Mozzhukhin,
and Philip Dolinenkov
7 NQR Detection of Sodium Nitrite Recrystallized in Wood. . . . . . . . . . . . . 77
Jrmy Jover, Sarra Aissani, Laous Guendouz,
Andr Thomas, and Daniel Canet
Part II Nuclear Magnetic Resonance Detection of Liquids
8 Bottled Liquid Scanner for Security Checkpoints . . . . . . . . . . . . . . . . . . . . . . 89

Pablo J. Prado
ix
x Contents
9 MagViz: A Bottled Liquids Scanner Using Ultra-Low

Field NMR Relaxometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
Robert Austin, Michelle Espy, Andrei Matlashov,
Henrik Sandin, Larry Schultz, Algis Urbaitis,
and Petr Volegov
10 Multiparameter NMR Identification of Liquid Substances. . . . . . . . . . . . 111
Andrey B. Konov, Kev M. Salikhov, Evgeniya L. Vavilova,
and Bulat Z. Rameev
11 NMR-Based Liquid Explosives Detector: Advantages and
Disadvantages of Different Configurations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
Anton Gradiek, Joe Luzar, Janez Lunik, and Toma Apih
12 Composite Pulses in Inhomogeneous Field NMR . . . . . . . . . . . . . . . . . . . . . . . 137
Galina S. Kupriyanova, Vitaliy V. Molchanov,
Evgeniy A. Severin, and Ivan G. Mershiev
Part III Other Techniques
13 Novel HTS DC Squid Solutions for NMR Applications . . . . . . . . . . . . . . . . 151

Maxim L. Chukharkin, Alexey S. Kalabukhov,
Justin F. Schneiderman, Fredrik isjen, Magnus Jnsson,
Minshu Xie, Oleg V. Snigirev, and Dag Winkler
14 Passive Sub-THz Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
Alexei Vertiy and Andrei Pavlyuchenko
Part I
Nuclear Quadrupole Resonance
Detection of Solids
Chapter 1
Further Improvement of NQR Technique
for Detection of Illicit Substances
Taras N. Rudakov
Abstract The nuclear quadrupole resonance (NQR) method has been used for the
detection of explosives and other illicit materials in a number of applications, includ-
ing baggage screening and landmine detection. This method provides automated
detection of explosives types in configurations missed by the imaging X-ray method.
This paper describes the continuing efforts to improve the function of the NQR
baggage scanner by increasing detection sensitivity, and the use of better methods
of reducing false alarm events generated from the RF excitation and various types
of interference. Based on the current research findings and a review of the previous
NQR based scanners, the key design objectives have been identified and discussed.
1.1 Introduction
One of the hardest to detect threats at todays checkpoints is explosives, particularly

plastic explosives. The most common screening technique based on using X-ray
method provides a high probability of detection for a range of both explosive
and non-explosive threats. However certain configurations of explosives cannot
be detected with the required probability by this kind of technique. The nuclear
quadrupole resonance (NQR) method has been evaluated for integration into
security equipment for the detection of concealed explosives [1, 2]. NQR technology
is non-disruptive to normal screening operations. NQR is complimentary with other
technologies such as X-ray and trace detection as it can increase the accuracy of
automated detection. The NQR method provides automated detection of explosives
types in configurations missed by the imaging X-ray method. Explosives can be
automatically detected regardless of size, shape, weight or bag clutter.
T.N. Rudakov ()

Lynx Engineering Ltd., 30 Brown Street, East Perth, WA 6004, Australia
e-mail: trudakov@iinet.net.au
T. Apih et al. (eds.), Magnetic Resonance Detection of Explosives and Illicit Materials, 3
NATO Science for Peace and Security Series B: Physics and Biophysics,
DOI 10.1007/978-94-007-7265-6__1, Springer ScienceCBusiness Media Dordrecht 2014
4 T.N. Rudakov
The technology of NQR detection is very similar to that normally used in

nuclear magnetic resonance (NMR). However in NQR a strong magnetic field is not
required to create magnetization because the electric quadrupole moments of the
nuclei interact with the internal electric field gradients of the surrounding charges
in molecules. Therefore the NQR equipment is relatively simple, more compact
and less costly than NMR. The detection process is automated so no reliance on
operator vigilance or skill is required. Besides the NQR technology is quite safe and
successfully trialled by authorities worldwide.
A relatively poor signal-to-noise ratio (SNR) is often an issue for explosive
detection by using NQR method [15]. Most explosives contain quadrupolar
nuclei (nitrogen-14) the spectral lines of which are usually located at low RF
frequencies. The signal intensity in this frequency range is quite low because the low
energy difference between excitation levels. Another issue is external and internal
interference and spurious signals. All these problems are solved by using optimised
NQR detection technique, equipment and signal processing. A choice of the optimal
detection technique is to a great extent determined by the NQR parameters of a
substance such as excitation frequencies and relaxation times [69].
This paper describes the continuing efforts to improve the function of the NQR
scanner through increased detection sensitivity, and the use of better methods of
reducing false alarm events generated from the RF excitation and interference.
Based on the current research findings and a review of the previous NQR based
scanners, the main design objectives are discussed.
1.2 Short Time Fourier Transform (STFT) Method
The first dominant source of false alarms in NQR baggage scanning is magneto-
acoustic (MA) ringing generated by some items. These items, when irradiated with
strong RF pulses, can become sources of coherent spurious signals. Once the applied
magnetic field has been removed, the magnetic domains in these materials continue
to oscillate for a time period in the order of a few milliseconds, until they dissipate
the applied energy. Typical sources of MA ringing include nickel plated objects,
magnets and umbrellas.
As these signals are coherent with RF pulses, they can be considerably reduced
by using a special multi-pulse technique [1]. The most common way to cancel
unwanted coherent noise is to use special multi-block pulse sequences which can
accumulate the total detected signal and the signal without NQR contribution and
subtract the latter from the former [10]. However, despite cancelling much of the
ringing signal, MA induced false alarms can still occur. This is due to the fact that
MA ringing signals have characteristics that vary over the duration of the pulse
sequence causing some residual MA ringing signal to remain after the subtraction
process. This, coupled with the fact that MA signals can be very large compared to
the NQR signal can result in a false alarm. MA induced false alarms predominately
occur at lower frequencies (<4 MHz) and are generally not observed above 5 MHz.
1 Further Improvement of NQR Technique for Detection of Illicit Substances 5
Fig. 1.1 FFT and STFT comparison
The second dominant source of false alarm is battery powered electronics

causing RF interferences. Active electronic items in the scanned baggage can cause
interference signals by re-radiating energy absorbed from the applied RF field of the
NQR scan or by radiating RF energy through their normal operation. The scanned
item may radiate RF energy intentionally, or the radiated RF interference may be due
to poor EMC (Electromagnetic Compatibility) design of internal electronics. Active
electronics interference signals, either induced or self-generated, can be sometimes
mitigated by turning off electronics before they are scanned.
As was observed, some interference signals from both MA and active electronic
sources only occur for a very short time in the time domain signal, whereas the NQR
signal usually extends across the full time window. This difference can be exploited
by the Short Time Fourier Transform (STFT) method to improve the false alarm rate.
The STFT permits the time-varying frequency content of a signal to be analysed,
that is it describes local changes in the frequency content within a short time
window as the window slides along the entire time record of the signal. Such
information cannot be determined by the Fast Fourier Transform (FFT) because it
analyses the whole data record. Comparison between conventional FFT and STFT
is shown in Fig. 1.1.
The STFT for a continuous-time signal x(t) is given by
Z1
X .f; / D x.t /w.t /e j 2f t dt; (1.1)
1
where w(t) is a real-valued window function. The STFT in (1.1) can be interpreted
as a conventional Fourier transform with a sliding-window. For discrete-time data,
x(nT), the STFT can be expressed as
X
R1
k n2
X .k2f; mT / D x ..n m/ T /w .nT / e j N ; (1.2)
nD0
6 T.N. Rudakov
Fig. 1.2 STFT processing of scan data for: (a) the bag was empty; (b) the bag contained 280 g
PETN
where R is the length of the window function, N is the length of data record x(nT),
T is the sampling period and f is given by
1
f D (1.3)
NT
As mentioned above, the STFT permits the discrimination between benign and
explosive signals based on the time-varying spectral properties of the scan data.
Figure 1.2 shows the corresponding STFT surfaces for two scan records obtained
from the empty bag and the same bag containing explosives. In this experiment the
signal has been detected at the intermediate frequency which is shown in this figure
as an off-set frequency. Note from the left-hand surface that the largest peak only
extends for a short proportion of the total time record while the explosive signal
(right-hand plot) has high peak amplitude that extends for almost the whole time
record. By taking into account the time-extent of the spectral characteristics of the
scan signals, the STFT has been able to discriminate between the benign signal and
the signal that contained an NQR response.
Application of the STFT to NQR detection of explosives has shown that the false
alarm rate (FAR) in carry-on baggage screening can be lowered by a substantial
amount. Application of a hybrid algorithm which combines the STFT and the
standard FFT processing method, results in an even greater reduction in the false
alarm rate. Measurements have shown that this reduction can lie between 27 and
58 % over current processing. Figure 1.3 demonstrates the example of Receiver
Operating Characteristics (ROC) for the STFT and standard FFT obtained for PETN
based explosives. In this example the total number of scans with and without
explosives was 116 and 455.
It can be seen that the standard STFT improves the FAR by 37 % over standard
processing for this particular dataset.
Fig. 1.3 Comparison

between normal FFT based
and STFT based processing
Note that additional research and more extensive testing may be required for
evaluating and developing STFT processing technique.
1.3 Pulse Sequence Optimisation for Detection of RDX

Based Explosives
NQR is a very effective method for detection and identification of RDX based
explosive materials in any configuration [1]. The spinlattice relaxation time T1
of this material is relatively short so that the Steady-State Free Precession (SSFP)
multi-pulse sequences are normally used for their detection [1116]. The main
element of this method is the creation of the continuous steady-state (SS) signals
which recur for as long as the pulse sequence is applied [17]. This is very convenient
for the fast coherent accumulation of the NQR signals. In order to achieve a strong
SS signal the pulse repetition time in the sequence should be less or close to the
signal decay time T2 * (line shape parameter) which is inversely proportional to the
spectral line width. Therefore the effectiveness of SSFP sequence strongly depends
on the material line width. As was found from our research different RDX forms
may have different value of the line width. This is clearly demonstrated in Table 1.1
for seven types of RDX from different manufacturers.
The SSFP pulse sequence optimised for detection of RDX with a narrow NQR
line may not be so effective for detection of RDX with much broader line. In this
case a good solution is to use the sequence optimised for detection of the RDX
with the broadest line expected, which here is 405 kHz. This sequence will be good
enough for other RDX forms. Otherwise additional SSFP pulse sequences optimised
for several groups of RDX materials may be applied.
8 T.N. Rudakov
Table 1.1 Some parameters of different RDX explosive materials

Source RDX type Quality Line width (Hz) Std deviation (%)
ADI Grade A RS-RDX 106 6
Royal Ordnance (RO) Type I RS-RDX 118 7
Eurenco IRDX RS-RDX 134 3
DYNO RS-RDX RS-RDX 200 5
BAE OSI Type II Non RS-RDX 305 3
DYNO Type II Non RS-RDX 405 4
Eurenco MI-RDX Non RS-RDX 298 2
1.4 NQR Baggage Scanners: Current and Future Design
In a scanning application, a NQR scanner irradiates an item of interest positioned

within the scanning volume with a series of RF pulses at a particular radio frequency.
The RF signal interacts with the quadrupolar nuclei contained within the scanned
volume, resulting in a small RF signal which can be detected by a high Q resonant
circuit.
Figure 1.4 shows a block diagram of the main functional components of a
NQR based explosive scanner. The RF synthesizer, pulse sequencer and pulse gate
together produce a series of low amplitude RF pulses at a frequency close to
the required NQR resonant frequency. The phases, on-times and off-times of the
pulses in the sequence are tightly controlled by the pulse sequencer to optimize the
measurement of a NQR response signal.
The RF pulses are amplified and then applied to the input of the probe which
is an LC resonant circuit. The probes inductor is an air-core coil that encircles the
scanned volume into which the item to be scanned is placed. The probe circuit is
tuned to have a resonant peak at approximately the required NQR frequency such
that a uniform RF magnetic field is generated within the coil. This field excites the
quadrupolar nuclei. In the period of time between pulses in the sequence, a low-
noise amplifier boosts the very small NQR response signals that are induced in the
coil. The output RF signal from the low-noise amplifier is down-converted to base-
band or to an intermediate frequency, f0 , and sampled for further signal processing
on a digital computer or dedicated digital signal processor (DSP).
One of most advanced baggage scanners T-3-03 was designed and developed by
the Australian Company QRSciences Limited (Fig. 1.5). The general block diagram
of this scanner was described previously [1]. The machine provides both NQR and
metal detection, operates at a speed about 200 bags per hour and can automatically
detect sheet explosives.
The main disadvantages of the scanner are the relatively low throughput,
relatively high FAR, large size, high manufacturing cost, and manual calibration
without self calibrating capability. In order to make the NQR based machines much
more successful in the market all these problems should be solved in new generation
of baggage scanners.
Fig. 1.4 Functional block-diagram of a NQR based scanner
Based on the current research findings and a review of the previous NQR based
scanners, the following key design objectives have been identified:
Enforce a tighter timing relationship between the transmitter and receiver
sections of the scanner and reduce the impact of externally and internally
generated interference by realizing both transmitter and receiver sections in the
same hardware module.
Boost the signal processing capabilities of the signal acquisition hardware in
terms of processing performance and flexibility.
Increase the capture memory available to the acquisition hardware.
Implement direct digital down-conversion to base-band of the sampled NQR
probe signal.
10 T.N. Rudakov
Fig. 1.5 NQR baggage scanner T-3-03
Implement automatic calibration and self-test capabilities.

Increase the number of digital acquisition channels.
Reduce the overall cost of manufacture.
These key design objectives for a new generation of NQR scanner will help to
realise the full potential of the signal processing techniques developed in the earlier
stages of the research and development process and increase the commercial value
of the final product.
Figure 1.6 demonstrates a block diagram of proposed new NQR scanner. In
the new scanner design the control, receiver and signal acquisition modules of the
previous scanner are integrated in a single hardware module. This has a number of
advantages, including enforcing a tighter timing relationship between the transmitter
and receiver sections of the scanner and reducing the chances of coupling of
externally and internally generated noise into the measurement. Direct digital down-
conversion of the sampled NQR probe signal to base-band is implemented in the
new scanner. This will reduce errors associated with analogue quadrature mixing
and down-conversion such as phase offset errors in the mixer, gain imbalance in
the I and Q analogue channels and I/Q DC offsets. The size and access speed
of the capture memory available to the acquisition hardware should be increased
significantly. This will allow using much more advanced detection technique and
signal processing. In order to reduce time consuming factory commissioning and
eliminate the need to calibrate the scanner using test bag sets automatic calibration
and self-test capabilities should be implemented in the new design.
Fig. 1.6 Block diagram of proposed new NQR baggage scanner
1.5 Coil Design Optimisation
The probe of a pulsed NQR detection system provides the interaction between the
RF field of the RF power amplifier with the target substance, as well as capturing
the RF field response from the target substance for detection by the receiving part
of the NQR scanner. Strong RF pulses, typically with the power of hundreds of
watts are used. In practical NQR scanners, when detecting specific substances such
12 T.N. Rudakov
Fig. 1.7 Single turn coil: (a) Coil design, (b) calculated RF magnetic field amplitude profile,
normalised to the field amplitude Bm in the centre of the coil, along the central axis of the coil
as explosives and narcotics, the RF pulse power can reach several kW. The probe
contains a resonant tank circuit with adjustable matching and tuning. The main
component of the resonant tank circuit is a coil.
A practical NQR scanner of specific substances must be able to detect at least
several types of such substances. Each substance has its own NQR frequency, with
the frequencies of different substances being within the range from 400 kHz to
5.5 MHz. For example RDX and HMX have NQR frequencies around 5 and 3 MHz,
PETN around 890 and 500 kHz, Black Powder around 660 and 560 kHz and
ammonium nitrate around 500 and 420 kHz, TNT has 12 frequencies within the
range of 700 to 900 kHz. Therefore an NQR detection system needs to operate over
a wide frequency range.
The typical coil of the NQR baggage scanner is a distributed RF coil sheet which
is a single turn solenoid (Fig. 1.7). Single turn solenoidal coils are widely used in
magnetic resonance [18] because of high homogeneity of the RF field and the high
Q factor. Besides, they have very low inductance, which is also an advantage if they
are used at high frequencies. However when using such coils at low frequencies
to tune them to resonance a high value of capacitance must be connected to tune
them to resonance, which is technically difficult. It becomes particularly important
for small or medium size coils of up to several tens of litres that are used for the
control of mail, mailbags or small baggage. As the coil inductance depends on the
coil volume, smaller coils would have very low inductance. This will require very
high capacitance. For example tuning to resonance a single turn solenoid coil with
the volume of about 25 L requires approximately 430,000 pF of capacitance for
tuning at about 890 kHz in order to detect the PETN based explosives. To shift the
resonant tank circuit from the frequency of one substance to the frequency of another
substance is achieved by switching a series of fixed value capacitors switched by
equal number of vacuum relays. For example to retune a single turn solenoid coil
with the volume of about 25 L from the frequency of about 3.4 MHz to the frequency
of about 890 kHz it is necessary to switch approximately 400,000 pF of capacitance.
Fig. 1.8 Modified two turn coil: (a) Coil design, (b) calculated RF magnetic field amplitude
profile, normalised to the field amplitude Bm in the centre of the coil, along the central axis of
the coil
A modified coil with bigger inductance and better RF field profile than the single
turn coil is shown in Fig. 1.8. The coil is made of a material of high electrical
conductivity (ribbon copper, for example) and has two turns, with one half of each
turn being much wider than the other half. Parts of turns with a smaller width are
situated at the ends of the coil, and parts of the larger width section constitute the
central and main area of the coil. Neither turn of the coil has direct contact on
the sides with the large width section. In this area the turns are separated with
longitudinal gap. A multiplicity of fixed value capacitors (or/and variable value
capacitors) are connected between the wide ends of the turns and located along
the gap.
The double turn coil presented in Fig. 1.8 not only preserves the uniformity
of the RF field in the central area of the studied volume, typical for a single turn
solenoidal coil, but also increases the uniformity of the RF field at the edges of the
coil. Furthermore, this double turn coil has a greater inductance as compared with
a single turn solenoidal coil of the same dimensions, which makes it more suitable
for use in a low-frequency range as well as for the detection of target substances in
small or medium volumes.
1.6 Probe Tuning System
The probe tuning system is a very important part of the baggage scanner. The
resonant coil is tuned to the appropriate frequency for irradiation and detection of
a particular explosive. Normally the tuning process is broken up into two parts
course tuning and fine tuning. Course tuning is used to select the type of explosive
and fine tuning is used to compensate for change in the tuning frequency caused by
baggage items and temperature.
14 T.N. Rudakov
Fig. 1.9 Probe tuning system of NQR baggage scanner (old version)
The typical tuning system contains a big number of fixed capacitors and special
high voltage RF relays. One of old versions of the tuning system used in baggage
scanners with single turn coils is shown in Fig. 1.9. As noted in Sect. 1.5, the
capacitors are connected between the ends of the coil. The RF relays are mounted
directly on the copper sheet connected to the coil.
The more advanced design of the probe tuning system is shown in Fig. 1.10. As
one can see from the picture the RF relays and capacitors are mounted on a special
Printed Circuit Board (PCB). This board is a four layer PCB with high voltage traces
Fig. 1.10 Probe tuning system of NQR baggage scanner (new version)
on the top and bottom layers and control lines sandwiched between two ground
planes. The whole assembly is split into two sections a left and a right section as
shown in Fig. 1.10.
The main enhancements of this new system design are the high reliability, low
noise and very low overall cost of manufacture and maintenance.
1.7 Conclusion
Research into the practical implementation of new advanced techniques and signal
processing algorithms has identified a number of areas within the NQR scanner
where architectural changes will provide substantial improvements in the perfor-
mance and capabilities of future scanner systems. These areas include the following:
Reducing the influence of externally and internally generated interference on the
detection rate and false alarm rate of the scanner.
Increasing the performance of the signal acquisition and storage system to permit
more flexible partitioning of the raw acquisitions into multiple data blocks.
16 T.N. Rudakov
Enabling fast, real-time, low-level signal processing of the raw acquisitions as

they are captured.
Increasing the signal processing capability of the scanner to allow advanced
signal processing and detection methods such as cross-polarization and multi-
frequency technique.
Improving the signal acquisition, capture and logging facilities for the develop-
ment of more advanced processing and detection algorithms.
In addition to these performance enhancements, the new architecture should be
aimed at:
Reducing production and maintenance costs.
Implementing more advanced self test and self calibration features.
Improving manufacturability.
Producing a flexible and expandable hardware platform that has the capacity to
form the foundation of NQR and other detector applications outside airport carry-
on luggage scanning.
We believe that the new generation of NQR based explosives detection systems
will be designed to operate much more effectively in the real world conditions.
It is also important to point out that the new generation of NQR technology
will not only be applicable to NQR baggage scanning systems, but also to other
NQR systems such as personnel screening (whether body or foot), scanning mail
packages, narcotic detection, landmine detection and the many other areas where
NQR can be used.
References
1. Rudakov TN (2009) Detection of explosives by NQR method: main aspects for transport
security. In: Fraissard J, Lapina O (eds) Explosives detection using magnetic and nuclear
resonance techniques, NATO science for peace and security series B. Springer, Dordrecht,
pp 111138
2. Mozzhukhin GV, Molchanov SV, Kupriyanova GS, Bodnya AV, Fedotov VV, Guoxin H, Yanbo
J, Tianliang R, Guojin Z (2009) The detection of industrial explosives by the quadrupole
resonance method: some aspects of the detection of ammonium nitrate and trinitrotoluene.
In: Fraissard J, Lapina O (eds) Explosives detection using magnetic and nuclear resonance
techniques, NATO science for peace and security series B. Springer, Dordrecht, pp 231234
3. Mozzhukhin GV, Rameev B, Dogan N, Aktas B (2009) The two-frequency multipulse sequence
in nuclear quadrupole resonance of N-14 nuclei. In: Fraissard J, Lapina O (eds) Explosives
detection using magnetic and nuclear resonance techniques, NATO science for peace and
security series B. Springer, Dordrecht, pp 205230
4. Mozjoukhine GV (2002) The frequency offset effects of NQR of spin I D 1 for remote
detection. Z Naturforsch 57:297303
5. Belyakov AV, Mozjoukhine GV, Rudakov TN (1993) A remote transducer of 14 N NQR.
Instrum Exp Tech 36:112113
6. Hirshfeld T, Klainer SM (1980) Short range remote NQR measurements. J Mol Struct 58:6377
7. Rudakov TN (1999) Magnetic resonance: nuclear quadrupole resonance, instrumentation. In:

Lindon JC, Tranter GE, Holmes JL (eds) Encyclopedia of spectroscopy and spectrometry, vol 2.
Academic, London, pp 16631671
8. Marino RA, Klainer SM (1977) Multiple spin echoes in pure quadrupole resonance. J Chem
Phys 67:33883389
9. Ostroff ED, Waugh JS (1966) Multiple spin echoes and spin locking in solids. Phys Rev Lett
16:1097
10. Gerothanassis IP (1987) Methods of avoiding the effects of acoustic ringing in pulsed Fourier
transform nuclear magnetic resonance spectroscopy. Prog NMR Spectrosc 19:267329
11. Rudakov TN, Mikhaltsevich VT, Selchikhin OP (1997) The use of multi-pulse nuclear
quadrupole resonance techniques for the detection of explosives containing RDX. J Phys D:
Appl Phys 30:13771382
12. Klainer SM, Hirschfeld TB, Marino RA (1982) Fourier transform NQR spectroscopy. In:
Marshal AG (ed) Fourier, Hadamard and Hilbert transforms in chemistry. Plenum, New York,
pp 147181
13. Buess ML, Garroway AN, Miller JB (1991) NQR detection using a meanderline surface coil.
J Magn Reson 92:348362
14. Rudakov TN, Belyakov AV (1998) Modifications of the steady-state free- precession sequences
for the detection of pure nuclear quadrupole resonance. J Phys D: Appl Phys 31:12511256
15. Rudakov TN (2006) Modification of SSFP technique for the effective detection of NQR
signals. Phys Lett A 358:322326
16. Ernst RR, Bodenhausen G, Wokaun A (1987) Principles of nuclear magnetic resonance in one
and two dimensions. Clarendon, Oxford
17. Rudakov TN, Mikhaltsevitch VT, Flexman JH, Hayes PA, Chisholm WP (2004) Modified
multipulse technique for the effective detection of pure nuclear quadrupole resonance. Appl
Magn Reson 25:467474
18. Hornak JP, Szumowski J, Bryant RG (1987) Elementary single turn solenoids used as the
transmitter and receiver in magnetic resonance imaging. J Magn Reson Imaging 5:233237
Chapter 2
An Overview of NQR Signal Detection
Algorithms
Naveed R. Butt, Erik Gudmundson, and Andreas Jakobsson
Abstract Nuclear quadrupole resonance (NQR) is a solid-state radio frequency

spectroscopic technique that can be used to detect the presence of quadrupolar
nuclei, that are prevalent in many narcotics, drugs, and explosive materials. Similar
to other modern spectroscopic techniques, such as nuclear magnetic resonance,
and Raman spectroscopy, NQR also relies heavily on statistical signal processing
systems for decision making and information extraction. This chapter provides
an overview of the current state-of-the-art algorithms for detection, estimation,
and classification of NQR signals. More specifically, the problem of NQR-based
detection of illicit materials is considered in detail. Several single- and multi-
sensor algorithms are reviewed that possess many features of practical importance,
including (a) robustness to uncertainties in the assumed spectral amplitudes, (b)
exploitation of the polymorphous nature of relevant compounds to improve detec-
tion, (c) ability to quantify mixtures, and (d) efficient estimation and cancellation of
background noise and radio frequency interference.
2.1 Introduction
Nuclear quadrupole resonance (NQR) is a solid-state radio frequency (RF) spectro-

scopic technique that can be used to detect the presence of quadrupolar nuclei, for
example 14 N, an element contained in many high explosives [15]. Furthermore,
as quadrupolar nuclei are prevalent in many narcotics and drugs, NQR can also
This work was supported in part by the European Research Council (ERC Grant Agreement n.
261670), the Swedish Research Council, and Carl Tryggers foundation.
N.R. Butt E. Gudmundson A. Jakobsson ()
Mathematical Statistics, Centre for Mathematical Sciences, Lund University, Lund, Sweden,
e-mail: naveed@maths.lth.se; erikg@maths.lth.se; aj@maths.lth.se
20 N.R. Butt et al.
be used for drug detection and in pharmaceutical applications [6]. Recently, the
technique has also been discussed in the area of oil drilling and geothermal heat
drilling. NQR is related to both nuclear magnetic resonance (NMR) and magnetic
resonance imaging (MRI), but does not require a large static magnetic field to split
the energy levels of the nuclei. This makes it attractive as a non-invasive technique
that can be used for detection of counterfeit medicines, land mines and unexploded
ordnances, or for screening baggage for explosives and narcotics at airports. For
instance, the current commonly used counterfeit medicine detection technologies
generally require varying degrees of sample pre-treatment or removal of pills from
packages [7, 8]. In contrast, portable NQR-based sensors can be developed that
custom officers and other agents of law enforcement can use without having to re-
move the medicines from their packaging. In case of explosive detection, contrary to
metal detectors and, for instance, ground penetrating radar (GPR), NQR detects the
explosive itself and its signature is unique; the NQR signal depends on the chemical
structure of the molecule. Hence, in the case of land mine detection, NQR will detect
the 14 N of the explosive, without suffering interference from, e.g., any fertilizer in
the soil. Furthermore, metal detectors will have problems in magnetic soil and with
mines containing very little metal,1 GPRs in clay or wet soils and with shallow
mines. The NQR technique, on the other hand, suffers mainly from its inherently
low signal-to-noise ratio (SNR), RF interference (RFI), and spurious signals such
as piezoelectric and magnetoacoustic responses, see, e.g., [1, 3]. The low SNR can
be remedied by repeating measurements, as NQR signals can be added coherently
(indeed, an NQR detection system can clear its own false alarms). However, the
time needed to guarantee accurate detection can be prohibitively long, especially
for the case of the common explosive trinitrotoluen (TNT). RFI, on the other hand,
can be alleviated using proper shielding, which, unfortunately, is only possible in
laboratory environments and not when used in practice. Radio transmissions are
extremely problematic for NQR signals if they lie at or near the expected locations of
the NQR resonance frequencies. This is the case for TNT as it has its resonances in
the radio AM band, often causing the AM signal to effectively mask the weak NQR
signal. The remainder of this chapter focuses on the recent advances on solutions to
the aforementioned problems. We discuss different data acquisition techniques and
summarize detector and interference cancellation algorithms.
2.2 Signal Models and Data Acquisition
Historically, the NQR signal has been measured as the free induction decay (FID),
which is the response after a single excitation pulse. The FIDs can then be added
coherently to improve the SNR, indicating that an NQR detection system is able to
1
Data from the Cambodian Mine Action Centre, taken from March 1992 until October 1998, shows
that for every mine found, there was more than 2,200 false alarms, mainly due to scrap items in the
ground.
2 An Overview of NQR Signal Detection Algorithms 21
Intensity
2 2 2

Echo
Echo
t0 Echo
xxxxxxxxx
p r r r r
Time
Fig. 2.1 Illustration of a PSL sequence
clear its own false alarms. However, measuring FIDs may not be the best strategy
for compounds with very long spin-lattice relaxation time, T1 , as one needs to
let the system fully relax before acquiring another FID. A delay time of 5T1 is
normally required between two excitation pulses, which could be as much as 30 s
for substances such as TNT. To improve the SNR per time unit, several multiple
pulse techniques have been proposed, of which the main techniques for detection
and quantitative applications are based on steady-state free precession (SSFP) and
pulsed spin locking (PSL)2 sequences. An example of the former sequence is the
strong off-resonant comb (SORC) [9]. Other SSFP-type sequences have been used
for the detection of cocaine base [6] and the explosive RDX [10]. In the interest
of brevity, we will here not further consider the SSFP techniques, merely noting
that the development for PSL sequences can be paralleled for SSFP sequences. The
signal obtained by PSL sequences consists of echoes that are measured between a
string of pulses [1, 2], see Fig. 2.1. The sequence consists of a preparatory pulse,
followed by a train of refocusing pulses (i.e., pulses which refocus the transverse
magnetization to produce an echo), written as
p
p tsp r r tsp M ; (2.1)
where p and r denote the flip angles of the preparatory and refocusing pulses,
respectively, while p and r denote their associated RF phases. Moreover, M is
the number of refocusing pulses, or, equivalently, the number of echoes, and tsp
is the time (normalized with respect to the dwell time) between the center of the
preparatory pulse and the center of the first refocusing pulse. This generates a train
of echoes, see Fig. 2.2, where each individual echo can be well modeled as a sum
2
The PSL sequence is sometimes referred to as the spin-locking spin-echo (SLSE) sequence.
22 N.R. Butt et al.
Fig. 2.2 Illustration of the real part of a typical echo train
of d exponentially damped sinusoids. In [11], the authors proposed the following

model for the mth echo in the echo train:
X
d
ym .t / D k e k .tCm/ e k jttsp jCi!k .T /t C wm .t /; (2.2)
kD1
where m D 0; : : : ; M 1 is the echo number; t D t0 ; : : : ; tN 1 denotes the sampling

time, measured with respect to the center of the refocusing pulse and typically
starting at t0 0 to allow for the deadtime between the pulse and the first measured
sample; k , k , k , and !k .T / denote the complex amplitude, the sinusoidal
damping constant, the echo-train damping constant, and the temperature dependent
frequency shifting function of the kth sinusoid, respectively. Moreover, wm .t /
denotes an additive colored noise, which often can be modeled using a low order
autoregressive model [11, 12]. It is important to note that the number of sinusoids,
d , and the frequency shifting function, !k .T /, can generally be assumed to be
known. For many compounds, such as TNT and RDX, !k .T / is a linear function of
the temperature T at likely temperatures [13]. In Fig. 2.3, a periodogram spectrum
of an averaged NQR signal from a shielded TNT sample is displayed. The above
mentioned acquisition techniques, which we term conventional, or classical, NQR
(cNQR), use powerful coherent RF modulated pulses to interrogate the sample.
Alternatively, one can use stochastic excitation, where the excitation sequence
consists of trains of low power coherent pulses whose phases or amplitudes are
randomized [14, 15]. This technique is in the following termed stochastic NQR
x 1010
10
8
NQR frequencies
7
6
Magnitude
Interference and noise

5
0
700 750 800 850 900 950
Frequency (kHz)
Fig. 2.3 Illustration of the periodogram spectrum of an NQR signal from a TNT sample
(sNQR). Provided the pulses are sufficiently weak, the sNQR system can be treated
as linear and time invariant. Hence, cross-correlation of the observed time domain
signal with the pseudo-white input sequence will produce an FID which can be well
modeled as [15]
X
d
y.t / D k e k Ci!k .T /t C w.t /; (2.3)
kD1
where t D t0 ; : : : ; tN 1 . As it is not possible to acquire the signal when shooting

a pulse, the FID obtained using sNQR will contain gaps and the signal will consist
of blocks of regularly sampled data. Furthermore, the time between the first sample
of each block is often not an integer multiple of the inter-sampling time within
the blocks. In [16], the authors proposed, for NMR, to fill the gaps by repeating
the measurements with different experimental settings so that the gaps occur at
different times. The different signals can then be stitched together. This technique is
slow and is therefore not recommended. As sNQR uses low power pulses, it has the
advantage, as compared to cNQR, that it can be used to interrogate samples hidden
on people and that it simplifies the construction of light-weighted, man-portable
detectors for use in, e.g., land mine detection. Another advantage with sNQR is that
the problem of waiting 5T1 between the measurements that is needed in cNQR is
alleviated, and the data can, in principle, be acquired continuously. Furthermore, due
24 N.R. Butt et al.
to the cross-correlation, sNQR measurements are less affected by RFI and spurious
signals as compared to cNQR. The advantage of cNQR over sNQR is primarily
the higher SNR. Several compounds of interest appears in different crystalline
structures, or polymorphs. For example, the explosive TNT exists in orthorhombic
and monoclinic polymorphs, and the proportions are often not known [4]. Searching
for monoclinic TNT when the explosive contains a mixture of both can severely
deteriorate the detection peformance [17, 18]. Sometimes the explosive is a mixture
of several explosives, e.g., TNT and RDX [19]. In [17,18], the authors proposed the
following signal model for the mth echo of PSL data from a mixture of different
explosives or polymorphs, or both:
X
P
ym .t / D .p/
p y m .t / C wm .t /; (2.4)
pD1
.p/
where ym .t / is defined as in (2.2) with the addition that the model parameters
depend on the pth polymorph, and where p denotes the proportion of the pth
polymorph. We also note that in pharmaceutical applications, it is sometimes
important to know the amount of each polymorph [20, 21].
2.3 Detectors
During the last 10 years, several NQR detectors have been proposed; however,
most of them do not fully exploit the richness of the NQR model. For example,
the demodulated approach (DMA) detects only one single resonance frequency.
Recently, more effective detectors have been proposed, exploiting more features
in the NQR model. In [22], the authors proposed using the echo train model (2.2)
together with a matched filter; in [11, 15, 17, 23, 24], generalized likelihood ratio
tests were used in combination with the models presented in Sect. 2.2. Commonly,
the amplitudes of the NQR signal were considered known to a multiplicative factor;
however, in practice, this would not be the case in most realistic scenarios as the
field at the sample will vary, causing variations in the NQR signal amplitudes. In
[25], this was remedied by allowing for uncertainties in the amplitudes, introducing
the FRETAML detector. Figures 2.42.6 display the performance of some of the
current state-of-the-art cNQR detectors, applied on partially shielded measured
data. The FETAML and ETAML detectors are both derived using model (2.2),
whereas FSAML and AML do not fully exploit the echo train structure. All four
algorithms assume the amplitudes to be fully known, as compared to FRETAML.
Furthermore, FETAML and FSAML are frequency selective. In order to utilize the
possibly polymorphic structure of compounds, an extension of FETAML, termed
FHETAML, was developed in [17]. FHETAML utilizes the polymorphic model
in (2.4). In [24], FHETAML was generalized to allow for uncertainties in the
assumed signal amplitudes, leading to the REMIQS detector. Typical comparative
Fig. 2.4 ROC curves 1

comparing state-of-the-art
cNQR detectors, using 0.9
partially shielded measured 0.8
TNT data
Probability of detection
0.7
0.6
0.5
0.4
0.3
0.2 FETAML
ETAML
0.1 FSAML
AML
0
0 0.2 0.4 0.6 0.8 1
Probability of false alarm

cNQR detectors, using 0.9
partially shielded measured
0.8
TNT data
0.7
0.6
0.5
0.4
0.3
0.2
DMAp
0.1 DMAr
DMAs
0
0 0.2 0.4 0.6 0.8 1
performance of the FHETAML and REMIQS detectors is shown in Fig. 2.7, where
LS-FHETAML is a variant of FHETAML that does not assume any prior knowledge
of the relative signal amplitudes. Common analysis and detection algorithms for
echo-train data, including the ones discussed above, require some initial estimates
of the expected echo decay within each echo, as well as the overall echo-train
decay. Also, the number of frequency components needs to be specified. One way
to retrieve initial estimates, and at the same time estimate the number of frequency
components, is to use non-parametric data-adaptive estimators, such as the ones
based on the Capon, APES, and IAA algorithms [26, 27]. The downside with these
26 N.R. Butt et al.

cNQR detectors, using
partially shielded measured 0.95
0.8
TNT data
0.9
0.6 0.85
0.8
0.4 0.75
0.7
0 0.1 0.2 0.3 0.4
0.2
FRETAML
FLSETAML
FETAML
0
0 0.2 0.4 0.6 0.8 1
100
FHETAML
95 REMIQS
LSFHETAML
90
Probability of detection (%)
85
80
75
70
65
60
55
50
0 20 40 60 80 100
Uncertainty level (%),
Fig. 2.7 Plots illustrating probability of detection as a function of the amplitude uncertainty
level,
, for probability of false alarm = 2 %
algorithms is that they are not able to estimate the finer structure of the echo train.
To alleviate this problem, the so-called ET-CAPA algorithm was recently introduced
in [28], which takes the whole echo train structure into account and estimates the
damping constants for every component present in the signal. NQR measurements
are often highly contaminated by powerful interference, and depending on the power
and frequency of the interference, detection may be very difficult. ET-CAPA is
Fig. 2.8 The amplitude landscape for an echo-averaged experimentally realistic metham-
phetamine NQR signal corrupted by sinusoidal interferences
more resilient to interference and manages to visualize both the interference and
the signal of interest. This is depicted in Fig. 2.8 for an experimentally realistic
methamphetamine NQR signal corrupted by several sinusoidal interferences. The
signal of interest is located in the middle of the plot at a damping of 0.01. Even
though the interference is 40 dB stronger than the signal of interest, the plot shows
that the amplitudes of the interference and the signal of interest are of almost
the same magnitude. This is due to the interference cancellation power of the
ET-CAPA estimator, making the signal easier to locate. The algorithm is also
suitable when faced with a new or an unknown substance. In a typical NQR
detection setup, one may first use ET-CAPA to get initial estimates to limit the
search space, and then use previously discussed parametric methods to obtain more
precise results. An alternative way to form the initial estimates of the expected
echo decay and the overall echo-train decay is to use the parametric ET-ESPRIT
estimator [29]. This is a computationally and statistically efficient estimator that
assumes that the measured signal can be well modeled using (2.2), and that the
additive noise may be approximated as being white, although it has been found
that the estimator finds reasonably accurate estimates even in cases when these
assumptions are somewhat violated. In the same work, the theoretically lower limit
on the parameters estimation variance is also presented, which may be useful in,
for example, determining an appropriate SNR to achieve a desired estimation or
classification accuracy. An example of how this can be done was discussed in
[30], where the possible classification of different manufacturers of paracetamol
28 N.R. Butt et al.
Fig. 2.9 Probability 1

of detection as a function
of ISR, for a probability 0.9
of false alarm of 0.05, using
0.8
simulated data with
SNR D 28 dB 0.7
0.6 SEAQUER
RTDAML
0.5 FRETAML
0.4 RETAML
0.3
0.2
0.1
0
0 20 40 60 80 100
Interference to signal ratio (dB)
was examined using such a theoretical expression. As for sNQR systems, there are
notably fewer detector algorithms published and the most efficient seems to be the
method published in [15]. This detector, as well as the ones shown in Figs. 2.42.6
and 2.9, are CFAR, i.e., they have constant false alarm rate with respect to the power
of the additive white noise. An alternative sNQR detector is also the below discussed
interference-resilient REWEAL detector, which was presented in [31].
2.4 Interference Rejection
One of the major concerns with NQR is the interference, both from RFI and from
piezoelectric and magnetoacoustic responses caused by, e.g., sand or by metal. To
remedy this, one idea is to use frequency selective algorithms, such as the one intro-
duced in [25]. As the frequency shifting functions are known and the temperature is
approximately known, the idea is to operate only on a subset of possible frequency
grid points. This not only makes it possible to omit frequencies where interference
signals are located, but also substantially reduces the computational complexity. In
this section, we consider single-sensor detection algorithms, proceeding in the next
section to also consider detectors based on spatial diversity. In [15, 32], the authors
proposed highly efficient projection algorithms to remove interference signals, using
the idea that secondary data, i.e., signal-of-interest (SOI) free data, can easily
be acquired without additional hardware. In sNQR, only a very small amount of
the data contains the FID, the rest can be considered secondary data; in cNQR,
secondary data can be acquired by continuing the measurement after the pulsing has
ceased. This information is used to construct an interference subspace, to which the
signal is then projected orthogonally, removing the RFI components. The detectors
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2 RESPEQ
SEAQUER(m)
0.1 SEAQUER(o)
REMIQS
0
0 0.2 0.4 0.6 0.8 1
Fig. 2.10 ROC curves for the RESPEQ, SEAQUER, and REMIQS algorithms, for SNR = 20 dB,
ISR = 40 dB
based on these principles have shown extremely good interference cancellation

properties, and simulations show that interference-to-signal ratios (ISRs) of 60 dB
are easily cancelled [15], without losing much detection performance; see Fig. 2.9,
where SEAQUER denotes the projection algorithm, RTDAML is a detector where
the data was prewhitened using a covariance matrix estimated from the secondary
data, and RETAML is the non-frequency selective version of FRETAML. Note
that the SNR is defined as the ratio of the energy of the noise-free signal and the
energy of the noise. The main limitation of SEAQUER is that it does not work
well when the SOI contains mixed NQR signals (e.g., from different polymorphic
forms of the same substance). In order to remove this limitation, a generalization
of SEAQUER, termed RESPEQ, was presented in [33]. The performance of the
RESPEQ algorithm is shown in Fig. 2.10, where it is also compared to other state-
of-the-art detection algorithms described previously. REMIQS, discussed in the
previous section, takes the full polymorphic structure of the signal into account,
but does not have support for dealing with the strong RFI. The suffixes (m) and
(o) refer to detection of the monoclinic and the orthorhombic parts of the signal,
respectively. As is seen from the figure, RESPEQ offers the best performance as
it can take all polymorphs into account as well as have an effective interference
rejection support. In both SEAQUER and RESPEQ, it is assumed that the SOI and
the RFI reside in low dimensional signal and interference subspaces. For various
reasons, such as deviations in the assumed SOI and uncertainty in the estimated
interference subspace due to finite sample effects, the SOI and RFI may, however,
deviate from their assumed subspaces. These errors may be in the measurements or
30 N.R. Butt et al.
in estimations of the subspaces, but they can alternatively be viewed as uncertainties

in the subspaces. In [31], in order to compensate for these errors, the authors allowed
for a small part of the energy in the signals to be outside the subspaces, leading to
the REWEAL detector. REWEAL carries out interference rejection by constructing
hyper-cones around the subspaces and searching for the SOI and RFI in these cones
instead of in the subspaces, and is particularly useful when only small amounts of
secondary data are available.
2.5 Multi-channel Detectors
Using multiple antennas for efficient interference rejection in NQR-based detection

has been proposed in several papers, e.g., [19, 24, 3437], and has shown good RFI
mitigation properties. Typically, one antenna is used to acquire the NQR signal
and the others measure the background interference and noise. This information
is then used to improve the detection. Among these, the NLS detector and the
frequency-selective FSMC detector developed in [38], are based on the structured
NQR signal model (2.2), and exploit the fact that the shifts of the spectral lines
depend in a known way on temperature. In [24], FSMC is generalized to include
polymorphism and to also allow for amplitude uncertainties, leading to the ESPIRE
detector. Figures 2.11 and 2.12 compare the different multi-channel detectors,
including the alternating least squares (ALS) and the model-mismatched maximum
likelihood (M3 L) detectors, both presented in [35], and the NLS, FSMC, and
100
90
80
70
60
50
ESPIRE
40
NLS
30 FSMC
M3L
20 M3L+5
ALS
10
ALS+5
0
0 20 40 60 80 100
Fig. 2.11 Plots illustrating probability of detection as a function of the uncertainty level,
,
for probability of false alarm = 8 %, SNR = 36 dB, with no RFI
100
90 ESPIRE
NLS
80 FSMC
M3 L
70 M3L+5
ALS
60 ALS+5
50
40
30
20
10
0
0 20 40 60 80 100
Fig. 2.12 Plots illustrating probability of detection as a function of the uncertainty level,
, for
probability of false alarm = 8 %, SNR = 36 dB, ISR = 42 dB
ESPIRE detectors. The NLS, FSMC and ESPIRE algorithms allow for a temperature
uncertainty region of 10 K around the true temperature, as well as large search
regions over the damping and echo damping constants. On the other hand, the ALS
and M3 L detectors assume perfect knowledge of the nonlinear parameters. To mimic
a more realistic scenario, however, the figures also include results for ALS and
M3 L for a 5 degrees (K) offset. In a more recent work, [39] have introduced the
NORRDIQ detector, that extends and improves ESPIRE by exploiting secondary
data to estimate the interference subspace.
2.6 Concluding Remarks
In this chapter, we have discussed recent advances in the detection, classification,

and identification of explosives, narcotics, and counterfeit medicines using NQR,
giving an overview of the data acquisition techniques and their mathematical
models. Furthermore, we have overviewed a variety of different detection and
interference cancellation algorithms and compared them on both measured and
simulated data.
Acknowledgements The authors are grateful to their long collaboration with Prof. John Smith,
Dr Jamie Barras, Dr Michael Rowe, Dr Iain Poplett, Dr Samuel Somasundaram, and Ms Georgia
Kyriakidou of the NQR group at Kings College London, who have been actively involved in
posing and examine many of the examined problems, as well as also kindly provided all the
measurements.
32 N.R. Butt et al.
References
1. Smith JAS (1995) Nitrogen-14 quadrupole resonance detection of RDX and HMX based
explosives. In: European convention on security and detection, Brighton, vol 408, pp 288292
2. Rowe MD, Smith JAS (1996) Mine detection by nuclear quadrupole resonance. In: Proceedings
of the EUREL international conference on the detection of Abandoned land mines, Edin-
bourgh, pp 6266
3. Garroway AN, Buess ML, Miller JB, Suits BH, Hibbs AD, Barrall AG, Matthews R, Burnett
LJ (2001) Remote sensing by nuclear quadrupole resonance. IEEE Trans Geosci Remote Sens
39(6):11081118
4. Deas RM, Gaskell MJ, Long K, Peirson NF, Rowe MD, Smith JAS (2004) An NQR study of
the crystalline structure of TNT. In: SPIE defense and security symposium, Orlando
5. Barrall GA, Arakawa M, Barabash LS, Bobroff S, Chepin JF, Derby KA, Drew AJ, Ermolaev
KV, Huo S, Lathrop DK, Steiger MJ, Stewart SH, Turner PJ (2005) Advances in the
engineering of quadrupole resonance landmine detection systems. In: Holloway JH et al (eds)
Detection and remediation technologies for mines and minelike targets X. Proceedings of the
SPIE, vol 5794. SPIE, Bellingham
6. Yesinowski JP, Buess ML, Garroway AN (1995) Detection of 14 N and 35 Cl in Cocaine base
and Hydrochloride using NQR, NMR and SQUID techniques. Anal Chem 67(13):22562263
7. Lopes MB, Wolff J-C, Bioucas-Dias JM, Figueiredo MAT (2009) Determination of counterfeit
HeptodinTM tablets by near-infrared chemical imaging and least squares estimation. Anal
Chim Acta 641:4651
8. Ricci C, Nyadong L, Yang F, Fernandez FM, Brown CD, Newton PN, Kazarian SG (2008)
Assessment of hand-held Raman instrumentation for in situ screening of potentially coun-
terfeit artesunate antimalarial tablets by FT-Raman spectroscopy and direct ionisation mass
spectrometry. Anal Chim Acta 623:178186
9. Klainer SM, Hirschfield TB, Marino RA (1982) Fourier transform NQR spectroscopy. In:
Marshall AG (ed) Hadamard and Hilbert transforms in chemistry. Plenum, New York, pp 147
182
10. Rudakov TN, Belyakov AV (1998) Modifications of the steady-state free-procession sequence
for the detection of pure nuclear quadrupole resonance. J Phys D Appl Phys 31:12511256
11. Jakobsson A, Mossberg M, Rowe M, Smith JAS (2006) Exploiting temperature dependency in
the detection of NQR signals. IEEE Trans Signal Process 54(5):16101616
12. Tan Y, Tantum SL, Collins LM (2004) Cramr-Rao lower bound for estimating quadrupole
resonance signals in non-gaussian noise. IEEE Signal Process Lett 11(5):490493
13. Smith JAS, Rowe MD, Deas RM, Gaskell MJ (2003) Nuclear quadrupole resonance detection
of landmines. In: Sahli H, Bottoms AM, Cornelis J (eds) International conference on
requirements and technologies for the detection, removal and neutralization of landmines and
UXO, Brussels, 1518 Sept 2003, vol 2, pp 715721
14. Garroway AN, Miller JB, Zax DB, Liao M-Y (1997) Method and apparatus for detecting
target species having quadrupolar nuclei by stochastic nuclear quadrupole resonance. US Patent
US5,608,321
15. Somasundaram SD, Jakobsson A, Rowe MD, Smith JAS, Butt NR, Althoefer K (2008)
Detecting stochastic nuclear quadrupole resonance signals in the presence of strong radio
frequency interference. In: Proceedings of the 33rd IEEE international conference on acoustics,
speech and signal processing (ICASSP), Las Vegas, Mar 304 Apr 2008
16. Yang D-K, Zax DB (1998) Bandwidth extension in noise spectroscopy. J Magn Reson
135:267270
17. Somasundaram SD, Jakobsson A, Smith JAS (2008) Analysis of nuclear quadrupole resonance
signals from mixtures. Signal Process 88(1):146157
18. Butt NR, Somasundaram SD, Jakobsson A, Smith JAS (2008) Frequency-selective robust
detection and estimation of polymorphic QR signals. Signal Process 88(4):834843
19. Xiong H, Li J, Barrall GA (2007) Joint TNT and RDX detection via quadrupole resonance.
IEEE Trans Aerosp Electron Syst 43(4):12821293
20. Balchin E, Malcolme-Lawes DJ, Poplett IJF, Rowe MD, Smith JAS, Pearce GES, Wren SAC
(2005) Potential of nuclear quadrupole resonance in pharmaceutical analysis. Anal Chem
77:39253930
21. Prez SC, Cerioni L, Wolfenson AE, Faudone S, Cuffini SL (2005) Utilisation of pure
nuclear quadrupole resonance spectroscopy for the study of pharmaceutical forms. Int J Pharm
298:143152
22. Gregorovic A, Apih T (2009) TNT detection with 14N NQR: multipulse sequences and
matched filter. J Magn Res 198(2):215221
23. Somasundaram SD, Jakobsson A, Smith JAS, Althoefer K (2007) Exploiting spin echo decay
in the detection of nuclear quadrupole resonance signals. IEEE Trans Geosci Remote Sens
45(4):925933
24. Butt NR, Jakobsson A, Somasundaram SD, Smith JAS (2008) Robust multichannel detection
of mixtures using nuclear quadrupole resonance. IEEE Trans Signal Process 56(10):50425050
25. Somasundaram SD, Jakobsson A, Gudmundson E (2008) Robust nuclear quadrupole resonance
signal detection allowing for amplitude uncertainties. IEEE Trans Signal Process 56(3):887
894
26. Stoica P, Sundin T (2001) Nonparametric NMR spectroscopy. J Magn Reson 152:5769
27. Gudmundson E, Stoica P, Li J, Jakobsson A, Rowe MD, Smith JAS, Ling J (2010) Spectral
estimation of irregularly sampled exponentially decaying signals with applications to RF
spectroscopy. J Magn Reson 203(1):167176
28. Kronvall T, Swrd J, Jakobsson A (2013) Non-parametric data-dependent estimation of
spectroscopic echo-train signals. In: Proceedings of the 38th IEEE international conference
on acoustics, speech and signal processing (ICASSP), Vancouver, Canada
29. Gudmundson E, Wirflt P, Jakobsson A, Jansson M (2012) An ESPRIT-based parameter
estimator for spectroscopic data. In: Proceedings of the IEEE statistical signal processing
workshop (SSP12), Ann Arbor, 58 Aug 2012
30. Topa H (2011) Detection of counterfeit medicines. M.S. thesis, Lund University
31. Svensson A, Jakobsson A (2011) Adaptive detection of a partly known signal corrupted by
strong interference. IEEE Signal Process Lett 18(12):729732
32. Somasundaram SD, Jakobsson A, Butt NR (2009) Countering radio frequency interference in
single-sensor quadrupole resonance. IEEE Geosci Remote Sens Lett 6(1):6266
33. Rudberg T, Jakobsson A (2011) Robust detection of nuclear quadrupole resonance signals in a
non-shielded environment. In: Proceedings of the 19th European signal processing conference
(EUSIPCO), Barcelona
34. Tan Y, Tantum SL, Collins LM (2005) Kalman filtering for enhanced landmine detection using
nuclear quadrupole resonance. IEEE Trans Geosci Remote Sens 43(7):15071516
35. Jiang Y, Stoica P, Li J (2004) Array signal processing in the known waveform and steering
vector case. IEEE Trans Signal Process 52(1):2335
36. Liu G, Jiang Y, Xiong H, Li J, Barrall G (2006) Radio frequency interference suppression for
landmine detection by quadrupole resonance. EURASIP J Appl Signal Process 2006(29890).
doi:10.1155/ASP/2006/29890
37. Stoica P, Xiong H, Xu L, Li J (2007) Adaptive beamforming for quadrupole resonance. Digit
Signal Process 17:634651
38. Jakobsson A, Mossberg M (2007) Using spatial diversity to detect narcotics and explosives
using NQR signals. IEEE Trans Signal Process 55(9):47214726
39. Butt NR, Jakobsson A (2011) Efficient removal of noise and interference in multichannel
quadrupole resonance. In: Proceedings of the 45th asilomar conference on signals, systems
and computers, Pacific Grove, pp 10721076
Chapter 3
Nuclear Quadrupole Resonance
of Pentaerythritol Tetranitrate (PETN)
in Different Compositions
Michael D. Rowe, John A.S. Smith, Jamie Barras, and Kaspar Althoefer
Abstract Pulsed NQR signals from the 14 N 495 kHz spectral line of pentaery-
thritol tetranitrate (PETN) for the two PETN containing high explosives, Semtex
H and Detasheet are compared with those observed from a commercial sample of
20 % PETN/Lactose as used in vasodilator drugs. The results show that the latter has
a narrower spectral line and a longer T2e decay time for PSL multiple echo trains.
3.1 Introduction
Pentaerythritol tetranitrate (PETN) C(CH2 -O-NO2 )4 is a well known explosive

found in compositions such as the various forms of Semtex and Detasheet. It
is also the active component of a number of vasodilator drug formulations such
as Dilcoran-80 and Pentral-80; for such medicinal use it is usually mixed with
lactose which acts as a desensitiser. To investigate the feasibility of using NQR
to distinguish between various PETN compositions this paper compares the NQR
parameters of PETN in Semtex H and Detasheet to those in a commercially available
20 % PETN/Lactose. The 495 kHz NQR line of PETN [1] was used for this
study as it is close to the two NQR lines near 502 kHz of RDX, also present in
Semtex, enabling us to estimate the relative proportions of the RDX and PETN in
this explosive. The 495 kHz PETN line is a singlet which is consistent with the X-ray
crystal structure [2] and we assume all samples contain the common polymorphic
form PETN I [2].
M.D. Rowe () J.A.S. Smith J. Barras K. Althoefer

Department of Informatics, Kings College London, Strand, London WC2R 2LS, UK
e-mail: michael.d.rowe@kcl.ac.uk; jamie.barras@kcl.ac.uk
36 M.D. Rowe et al.
3.2 Experimental
We were supplied with two samples of Semtex H of 29.75 and 31.00 g in 5 cm

diameter, 2 cm long acrylic cylinders with disc end seals. Both Semtex H samples
were in the NQR detection coil for this study so the total Semtex H weight was
60.75 g. From measurements of the integrated spectral line intensities of the two
RDX lines near 502 kHz and the 495 kHz PETN line we calculated a PETN/RDX
ratio of 0.75 in this particular sample of Semtex H. Assuming the sample to be
16.9 % binder, this would give a composition of 35.6 % PETN and 47.5 % RDX.
So we estimated that the two Semtex H samples together contain 21.6 g of PETN.
A 117.5 g Detasheet sample, which contained 63 % PETN, was supplied as small
cut pieces of the sheet material contained in a Pyrex glass bottle of about 5 cm
diameter and 6 cm length. Since Detasheet is 63 % PETN this sample contained
75 g of PETN. The 50 g sample of 20 % PETN/Lactose powder (Fluka Chemika
76653), obtained from Sigma-Aldrich, was lightly compressed into a 5 cm diameter
plastic bottle giving a sample length close to 5 cm.
The 14 N NQR measurements were carried out at room temperature (295 K)
using a Tecmag Libra pulsed RF spectrometer and a NQR probe with a Q of 80
90 fitted with an active Q damping system. The RF coil was a 6 cm long solenoid,
wound on an ABS cylinder of 5.8 cm outside diameter, giving a sample volume of
about 150 cm3 . While in each case, the sample was placed centrally in the detection
coil, the coil volume occupied by the sample varied due to the different volumes
and nature of the samples. To enhance the signal-to-noise ratio (SNR) trains of echo
signals were acquired using the pulsed spin locking (PSL) [3] sequence:-
P1 .PhP1/ P2 .PhP2/ 2n
where P1 and P2 are 119 flip angle RF pulses (i.e. 90 eff) and PhP1 and PhP2
are the phases of P1 and P2 which differ by 90 . The RF pulse phase cycling
was PhP1 D 0 ,180 ,0 ,180 and PhP2 D 90 ,270 ,270 ,90 . The echo signals,
refocused at the centre of the 2 spacing between the P2 pulses, were recorded using
the signal data acquisition phase cycling PhD D 0 ,180 ,0 ,180 . This phase cycling
ensures that only echo signals are acquired and spurious responses are cancelled.
The train of echo signals decays exponentially with the effective decay time T2e .
3.3 Results and Discussion
Figure 3.1 shows the 495 kHz NQR line of PETN for the Semtex H and 20 %
PETN/Lactose samples together with Lorentzian fits to the line shapes. The spectra
were obtained from 70 summed PSL echo signals obtained using a 2 of 5.36 ms
with excitation at 497 kHz. With this 2 spacing the data acquisition time is long
enough to obtain the full echo envelope ensuring good spectral line characterization
3 Nuclear Quadrupole Resonance of Pentaerythritol Tetranitrate (PETN). . . 37
Fig. 3.1 The 495 kHz NQR line of PETN for Semtex H (red) and 20 % PETN/Lactose (blue)
at about 295 K shown on the same plot for comparison. A Lorentzian fit to each line is also shown
(black)
without artifacts. For Semtex H, a signal is also detected from the RDX component,
but since the T2e is much shorter (11 ms) than for PETN this RDX signal was
eliminated by rejecting the first six echoes. As seen in Fig. 3.1 the PETN line of the
20 % PETN/Lactose is narrower than for the Semtex H.
For a Lorentzian spectral line the width 1/2 (FWHH, full width at half height)
and time domain signal decay times T2 * are related through 1/2 D (T2 * )1 so
both parameters can be obtained from either the frequency or time data. In the
frequency domain we used a Lorentzian fit to the line profile (as in Fig. 3.1) to
obtain 1/2 from which T2 * was calculated. Working directly on the exponentially
decaying time domain signals, we have used the Matrix Pencil (MP) technique [4]
to extract T2 * from which 1/2 was calculated. Figure 3.2 shows the summed PSL
echo signal for the Semtex H sample together with the signal modelled using the
parameters extracted with the MP technique. Table 3.1 lists T2 * and 1/2 for all
three PETN compositions along with the method used to obtain the parameters.
The results in Table 3.1 show that the two explosive samples have very similar
line width values and that for both the NQR line is broader than observed for
the 20 % PETN/Lactose sample. In general NQR line widths are dominated by
inhomogeneous broadening so variation is to be expected with the nature of the
sample and depends on factors such as crystallite size, defects and purity. For PETN
there is evidence [5] that the principal defects are dislocation slip planes with f110g
38 M.D. Rowe et al.
Fig. 3.2 The summed echo signal obtained for the 495 kHz NQR line of PETN in Semtex H
using a PSL sequence with 2 D 5.36 ms (the real, imaginary and magnitude signal components
are shown in red, green and blue respectively). The modelled signal components obtained from the
parameters extracted using the Matrix Pencil technique are shown in black
Table 3.1 The PETN 495 kHz NQR line-width and signal decay parameters for the
three samples at about 295 K
Sample Semtex H 20 % PETN/Lactose Detasheet
Method MP Lorentzian MP Lorentzian MP Lorentzian
T2 * [ms] 0.50(2) 0.50(0) 0.62(6) 0.62(5) 0.50(0) 0.50(4)
1/2 [kHz] 0.63(4) 0.63(7) 0.50(8) 0.50(9) 0.63(7) 0.63(1)
The method used to extract the parameters as discussed in the text is indicated by MP
and Lorentzian
being the most active as they are the most closely-packed. The line broadening is
most likely to be due to variations in the frequency of modes involving torsions of
the NO2 group at the slip planes, as we have previously proposed in the explosive
RDX [6]. For PETN below 300 K these modes have been assigned a frequency
close to 95 cm1 from the temperature variation of the NQR C line frequency [1].
Analysis of the terahertz spectrum suggests that this mode may actually correspond
to the peak near 101 cm1 although this is assigned to a CONO2 twisting mode
about the CC bond [7]. The same mode may have been observed as a broad
feature in the single crystal spectrum above 90 cm1 . However, a normal mode
analysis [8] gives frequencies of 50 and 56 cm1 (irreducible representations A and
E respectively) for librational modes that involve NO2 torsions.
Fig. 3.3 Integrated real line intensity for the 495 kHz NQR line of PETN in Semtex H plotted
versus 2n for each echo in a train of 2,048 with 2 D 2.84 ms. A fit to Aexp(2n/T2e ) C B is
also shown with T2e D 843 ms, A D 73,200 and B D 5,166
We have measured T1 for the PETN 495 kHz NQR line of the Semtex H
and 20 % PETN/Lactose samples to be 27.5 4 and 28.5 4 s respectively. The
measurement was carried out using a steady-state saturation recovery method [9]
in which the repetition time of a PSL detection sequence was varied. In this case,
since it was not necessary to acquire the entire echo signal, a shorter 2 of 2.84 ms
was used allowing more echoes (190 here) to be summed for signal enhancement
as T2e is longer under these conditions. The integrated spectral line intensity was
used to plot the recovery curves and T1 was obtained from a fit assuming a single
exponential component.
The use of PSL type sequences for signal enhancement by summing echoes is
often crucially important in detection applications particularly when T1 is long but
it is often difficult in such a case to obtain accurate values for T2e . Long signal
averaging times are required to acquire the individual weak echoes with sufficient
SNR to define the full decay envelope of the echo train accurately. Figure 3.3 shows
an example of such data obtained here for the 495 kHz PETN line of the Semtex H
sample. In this case, 2048 sequential echoes were acquired using a PSL 2 spacing
of 2.84 ms and the data points are the integrated intensities of the real component
of each spectral line obtained from the echo FT of the individual decaying echoes
in the train. Even though this data took about 16 h to acquire the plot shows a high
degree of scatter. The fit in Fig. 3.3 was obtained using Aexp(2n/T2e ) C B where
40 M.D. Rowe et al.
Fig. 3.4 Integrated real line intensity for the 495 kHz NQR line of PETN in 20 %
PETN/Lactose plotted versus 2n for each echo in a train of 2,048 with 2 D 2.84 ms. A fit to
Aexp(2n/T2e ) C B is also shown with T2e D 988 ms, A D 54,930 and B D 4,973
A is the initial amplitude, n is the number of the echo in the train and B is a non
zero baseline term. In general the latter can be due to a steady state signal level that
may be produced [10] by PSL sequences from some materials. For the fit in Fig. 3.3
T2e D 843 ms, A D 73,200 and B D 5,166.
Figure 3.4 shows the equivalent measurement for the 20 % PETN/Lactose
sample, here the fit has T2e D 988 ms, A D 54,930 and B D 4,973.
It is possible to reduce the scatter in such plots by taking the mean of adjacent
points and re-spacing the data, so for example the mean of points 1 and 2 is plotted at
1.5(2) and that of points 3 and 4 at 3.5(2) etc thus the new spacing becomes 2(2).
This process can be repeated a number of times and Fig. 3.5 shows the Semtex H
data of Fig. 3.3 averaged this way five times reducing the 2,048 points to 64. For the
fit here T2e D 851 ms, A D 73,000 and B D 5,109 which is in good agreement with
the fit of Fig. 3.3 adding credence to the reliability of the measured T2e values.
Apart from the first Hahn (or direct echo) of a PSL echo train the subsequent
echoes are a mixture of Hahn and stimulated echoes and the region of exponential
decay usually takes several echoes to evolve. After the first weakest Hahn echo,
the next few echoes often show an initial growth in intensity or an alternation in
intensity depending on the pulse sequence and off-resonance conditions. In phase
1 of PETN, the dipolar coupling can be estimated by combining structural details
from the X-ray analysis [2] with the assumption that the orientation of the principal
Fig. 3.5 The Semtex H data of Fig. 3.3 with adjacent points averaged and re-spaced until 2,048
points are reduced to 64. Here the fit to Aexp(2n/T2e ) C B is with T2e D 851 ms, A D 73,000
and B D 5,109
axes is the same as in nitric acid, viz. qxx is perpendicular to the ONO plane and
qzz makes an angle of 24 with the NOH bond [11]. The intra and intermolecular
contribution to the second moment can then be calculated [12] to be approximately
75 and 33 Hz2 respectively. The first order N H dipolar contribution is negligible
in zero magnetic field; the second order is calculated to be 3 Hz2 [12] based on an
approximate estimate of the N H distance of 262 pm assuming CH2 hydrogen
coordinates based on a tetrahedral geometry for this group. Adding all terms gives
111 Hz2 which for a Gaussian line shape gives a predicted value for the width 1/2
of 12 Hz and a decay time due to the dipolar coupling Td of 13 ms. Figure 3.6
shows the first 100 ms of a sequential plot of the individual NQR spectra used to
produce the data of Fig. 3.4 for 20 % PETN/Lactose. While the random scatter in
the PETN spectral line intensities make it difficult to estimate any possible dipolar
contributions to the decay the behavior below about 50 ms can be tentatively fitted
to a Gaussian function with a decay time of 18 ms. However, such initial behaviour
is negligible compared to the overall echo train decay and the T2e values given here
have been obtained from single exponential fits as shown above.
Our measured values for T2e including those obtained for the PSL sequence with
2 D 5.36 ms are given in Table 3.2 together with our values for T1 .
Within experimental error, the values of T1 are the same for both samples and
agree well with the T1 of 30 s we measured in previous work for the same 495 kHz
NQR line of PETN in Detasheet at room temperature. These T1 values are very
42 M.D. Rowe et al.
Fig. 3.6 The first 100 ms of

a sequential plot of the
individual NQR spectra used
to produce the data of Fig. 3.4
for 20 % PETN/Lactose
Table 3.2 Values of T2e and Sample Semtex H 20 % PETN/Lactose

T1 for the PETN 495 kHz
NQR line of the Semtex H T2e [ms] (2 D 2.84 ms) 843 988
and 20 % PETN/Lactose T2e [ms] (2 D 5.36 ms) 294 376
samples at about 295 K T1 [s] 27.5 4 28.5 4
close to the value of T1 for the 890 kHz C NQR line of PETN in Detasheet which
is about 34 s at 20 C [1]. Interestingly this C spectral line is much narrower
(1/2 D 0.35 kHz and T2 * D 0.9 ms) than the line studied here.
The values of T2e in Table 3.2 for Semtex H are reduced compared to those
for 20 % PETN/Lactose by factors of 0.85 and 0.78 for 2 of 2.84 and 5.36 ms
respectively. This is very similar to the factor of 0.8 reduction in T2 * for Semtex H
compared to that for 20 % PETN/Lactose seen in Table 3.1. These differences may
be due higher purity and/or crystallinity of the PETN/Lactose sample.
3.4 Conclusions
A comparison of the signals for the 14 N NQR 495 kHz line of pentaerythritol
tetranitrate (PETN) in the high explosives Semtex H and Detasheet with those
observed from 20 % PETN/Lactose as used in vasodilator drugs, has shown that
the latter has a narrower spectral line and a longer T2e decay time for PSL multiple
echo trains. These parameters may be of use in distinguishing different PETN
compositions.
Acknowledgements We thank the UK Home Office, Centre for Applied Science and Technology
(CAST) and the Defence Science and Technology Laboratory (DSTL) at Fort Halstead for support
of this work.
References
1. Smith JAS, Rayner TJ, Rowe MD, Barras J, Peirson NF, Stevens AD, Althoefer K (2010) Mag-
netic field-cycling NMR and 14 N, 17 O quadrupole resonance in the explosive pentaerythritol
tetranitrate (PETN). J Magn Reson 204:139144
2. Trotter J (1963) Bond lengths and angles in penterythritol tetranitrate. Acta Crystallogr 16:698
699; Cady HH, Larsen AC (1975) Pentaerythritol Tetranitrate II: its crystal structure and
transformation to PETN I; an algorithm for refinement of the crystal structure with poor data.
Acta Crystallogr B31:18641869
3. Marino RA, Klainer JM (1977) Multiple spin echoes in pure quadrupole resonance. J Chem
Phys 67:33883389
4. Lin Y-Y, Hodgkinson P, Ernst M, Pines A (1997) A novel detection-estimation scheme for
noisy NMR signals: applications to delayed acquisition data. J Magn Reson 128:3041
5. Gallagher HG, Halfpenny PJ, Miller JC, Sherwood JN (1992) Dislocation slip systems in
pentaerythritol tetranitrate (PETN) and cyclotrimethylene trinitramine (RDX). Philos Trans
R Soc Lond A339:293303
6. Smith JAS, Blanz M, Rayner TJ, Rowe MD, Bedford S, Althoefer K (2011) 14 N quadrupole
resonance and 1 H T1 dispersion in the explosive RDX. J Magn Reson 213:99108
7. Allis DG, Korter DM (2006) Theoretical analysis of the terahertz spectrum of the high
explosive PETN. Chemphyschem 7:23982408
8. Grudzkov YA, Gupta YM (2001) Vibrational properties and structure of pentaerythritol
tetranitrate. J Phys Chem A105:61976202
9. Fukushima E, Roeder SBW (1981) Experimental pulse NMR: a nuts and bolts approach.
Addison-Wesley, Reading
10. Ya Osokin D, Ermakov VL, Kurbanov RH, Shagalov VA (1992) The quasi-stationary states in
multipulse NQR. Z Naturforsch 47a:439445
11. Millen DJ, Morton JR (1960) The microwave spectrum of nitric acid. J Chem Soc 15231528
12. Vega S (1973) Second and fourth moments in NQR spectroscopy of spins with I D 1. Adv
Magn Reson Waugh JS (ed) 6:259302
Chapter 4
Cross-Relaxation Enhanced NQR of Ammonium
Nitrate in Low Magnetic Field
Georgy V. Mozzhukhin, Bulat Z. Rameev, Galina S. Kupriyanova,

and Bekir Aktas
Abstract Nuclear quadrupole resonance (NQR) with use of pulses of low magnetic
field has been studied. The technique is based on the matching of proton frequency
(
L ) to one of the NQR frequencies (
0 ,
C or
) for the period of application
of pulse magnetic field. Theoretical approach to analyse the NQR experiments in
pulse magnetic fields is outlined. In this work the NQR on ammonium nitrate (AN)
sample for specific case of
L D
0 have been studied. It has been shown that the
technique provides essential shortening the effective spin-lattice relaxation time and
can be applied for the detection of explosive materials.
G.V. Mozzhukhin ()

Department of Physics, Gebze Institute of Technology, Cayrova Campus, 41400 Gebze-Kocaeli,
Turkey
Department of Physics, Kazan State Power Engineering University, Krasnoselskay 51, 420066
Kazan, Russia
e-mail: mgeorge@yandex.ru
B.Z. Rameev
Department of Physics, Gebze Institute of Technology Cayrova Campus, 41400 Gebze-Kocaeli,
Turkey
Laboratory of Radiation Physics, Kazan Physical-Technical Institute (KPhTI), Sibirsky Tract
10/7, 420029 Kazan, Russia
G.S. Kupriyanova
Department of Radiophysics and Information Security, Institute of Physics and Technology,
Immanuel Kant Baltic Federal University, 14 Nevskogo St, 236041 Kaliningrad, Russia
B. Aktas
Department of Physics, Gebze Institute of Technology, Cayrova Campus, 41400 Gebze-Kocaeli,
Turkey
46 G.V. Mozzhukhin et al.
4.1 Introduction
During last 20 years essential efforts has been concentrated on Nuclear Quadrupole
Resonance (NQR) as the method for explosive detection [1, 2]. The prototypes of
NQR devices for luggage scanners and landmine detectors have been designed.
However, there are still some issues preventing a wide practical application of
this technology for explosive detection. These are: (1) a low signal-to-noise ratio
(SNR) for some important explosives that makes the detection of these explo-
sives very slow; (2) a strong influence on SNR of radiofrequency interference
(RFI) due to external noise sources (such as AM radio); and (3) an effect of
spurious signals due to various metal or piezoelectric materials. The explosive
list, which can be easily detected by NQR, includes e.g. RDX (C3 H6 N6 O6 ),
HMX (C4 H8 N8 O8 ), tetryl (C7 H5 N5 O8 ), and their mixtures. However, explosives
such as trinitrotoluene (TNT, C7 H5 N3 O6 ), ammonium nitrate (AN), pentaerythritol
tetranitrate PETN (C5 H8 N5 O12 ) have much smaller signal-to-noise ratios because
of lower NQR frequencies (<1 MHz) and longer spin-lattice relaxation times. The
effect of radiofrequency interference can be reduced by shielding, advanced signal
processing [3, 4] and various methods of active suppression of RFI noise [5, 6].
However, it is obvious that for the practical application of NQR for explosive
detection, the development of novel physical methods increasing the detected signal
intensity is highly desirable.
In this work we studied the detection of AN as one of the substances with
very weak NQR signal. It is well known that multipulse sequences are commonly
used in NQR to speed up the detection [7]. Further gain may be obtained by use
of nitrogen-proton cross-polarization methods. There are two main modifications of
these double resonance approaches. In the first one (direct method) the transfer of
polarization from proton system results in the increase of intensity of NQR signal on
nitrogen nuclei (see, e.g., [8]), while in the second one (indirect method) the nitrogen
nuclei are detected through the proton NMR signal [911]. The gain in SNR depends
on the strength of the static magnetic field. However, the indirect NMR method
and cross-polarization NQR, which rely on use of rather high magnetic fields,
are not feasible for implementation in commercial devices, such as the luggage
scanners and mine detectors. There is an additional possibility to increase SNR
by the cross-relaxation interactions between quadrupole and proton systems. As
was already mentioned above, a very long spin-lattice relaxation time T1 of some
explosives (PETN, AN and TNT) is another obstacle in 14 N NQR detection that
prevents a fast accumulation of quadrupole signal. For example, the longitudinal
relaxation time for TNT lines is between 4.6 and 9 s, while such substance as
hexamethylenetetramine, which is well-known as explosive precursor, relaxes for
much shorter time of 23 ms. A long relaxation time T1 makes the process of signal
accumulation (averaging) in NQR slower due to a need to wait for a time of about
5T1 between successive pulse sequences before the nitrogen nuclei magnetization
recovers. The cross-relaxation effects between 14 N and 1 H spin systems may be
applied to reduce this time of recovering of the quadrupole magnetization [8, 12
14]. Two provisions are needed for application of these methods: (1) a substance
4 Cross-Relaxation Enhanced NQR of Ammonium Nitrate in Low Magnetic Field 47
contains both nitrogen and hydrogen nuclei, and (2) the contact (cross-relaxation
interactions) between the proton and quadrupole spin-systems is created. For the
second provision, so-called level crossings should be made as result of application
of a static or quasistatic (i.e. slowly-changed) magnetic field. In the work [14] it
has been shown that shortening up to 3.4 (4.4) times in the relaxation times of C
( ) NQR lines of AN sample is obtained as result of the cross-relaxation contact.
The cited work reveals the potential prospects of this technique for practical use.
However, the applied experimental protocols were based either on preparatory RF
saturation of the transition under study or on the slowly sweeping the magnetic
field from level crossing at the upper (detected) transition through other (lower
frequency) double-resonance conditions of AN.
In our work we study in details those double-resonance protocols, which
theoretically do not provide the largest gain, but are more feasible to apply in
practice. In our scheme, we apply pulses of small magnetic field to match the proton
resonance frequency
L to the lowest quadrupole resonance frequency
0 . Then we
detect the spin-echo NQR signal on the
transition which is characterized by the
smallest temperature shift of the resonance frequency. It has been revealed that this
technique provides at least twofold shortening of the effective spin-lattice relaxation
time and therefore can be applied in the detection of explosive materials.
4.2 Theory
The theoretical treatment of cross-relaxation effects in double NMR/NQR has been

proposed in the Ref. [10]. In this approach one starts with consideration of two
kinds of spins, which interact via dipole-dipole interactions: proton spins and 14 N
quadrupole nuclei. The relaxation of quadrupole subsystem is defined by spin-
lattice relaxation rates W0,,C . For the proton subsystem P in the contact with
quadrupole subsystem Q, the quadrupole subsystem may also relax through the
interaction with protons with the cross-relaxation time (TCR ). It is the time that
required for the P and Q spin subsystems to come in equilibrium with a common
spin temperature (if other relaxation channels are switched off). In the double
resonance there is another important parameter the time for polarization of
proton subsystem. This time increases with value of the static magnetic field. Long
polarization time of proton system will increase the total time of double resonance
experiment [15]. Below we will discuss the case of
L D
0 . Similar to work [10] the
populations of three-level quadrupole system are considered. In Ref. [10] the effect
of cross-relaxation on the proton spin-lattice relaxation and the proton polarization
has been calculated, while in our case we consider the backward effect on the
magnetization of quadrupole system. According to the results of the work [14] the
rate of the proton-nitrogen cross-relaxation is much faster than the rates of both
the proton relaxation and the nitrogen spin-lattice relaxation. It means that TCR is
very small and in the limiting case we can neglect it. We used the high temperature
approximation with the same direct and reverse transition rates for both proton and
Fig. 4.1 The energy levels and populations of the quadrupole (at left) and proton (at right)
systems
nitrogen systems. We suppose that the ratio of protons to nitrogen nuclei D NN /NH
is smaller than 1. The populations of energetic levels are defined by following rate
equations [10]:
d.N1 C n1 /=dt D WH .N1 N2 / W0 .n1 n2 / WC .n1 n3 /
d.N2 C n2 /=dt D WH .N1 N2 / C W0 .n1 n2 / W .n2 n3 /
d n3 =dt D WC .n3 n1 / W .n3 n2 /; (4.1)
where n1 , n2 , n3 are the populations of energy levels of the quadrupole system

(Fig. 4.1); and N1 , N2 are the populations of proton systems. WH D 1/(2T1 ) is the
proton spin-lattice relaxation rate at the magnetic field
L D
0 (without the cross-
relaxation between the proton and quadrupole systems). W0 , WC and W are the
nitrogen relaxation rates for the transitions
0 ,
C and
, respectively.
The energy of the 14 N levels with spin I D 1 may be presented as follows [11]:
E1 D A .1 C / ; E2 D A .1 / ; and E2 D 2A; (4.2)
2
where A D e Qq 4
zz
and e 2 Qqzz is quadrupole coupling constant, is asymmetry
parameter of electric field gradient. Therefore, the transition frequencies of 14 N are:
2
C D A.3 C /; 2
D A.3 /; and 2
0 D 2A (4.3)
In the steady state the population differences may be presented as following

!0
n21 D n2 n1 D n2 1 exp
kT

!
n32 D n3 n2 D n3 1 exp (4.4)
kT
Similar to work [10] we write the populations as follows:
n21 D n2 .1 exp .z// n3 .1 C z 1::/ nz
n31 D n3 .1 exp .y// n3 .1 C y 1::/ ny (4.5)
Then we used following relations:

N.N / A .1 C /
n1 D exp
3 kT

N.N / A .1 C / 2A 2A 2A 2A
1 C C D
3 kT kT kT kT kT

N.N / A .3 / 2A 2A N.N /
D 1 C .1 y C x z/
3 kT kT kT 3

N.N / A .1 / N.N / A .1 / 2A 2A
n2 D exp D 1 C D
3 kT 3 kT kT kT
N.N /
D .1 y C x/
3

N.N / 2A N.N / 2A N.N /
n3 D exp 1C D .1 C x/ (4.6)
3 kT 3 kT 3
For proton system we have

N.H / 1 !0 N.H /
N1 D exp ! N1 D .1 z/
2 2 kT 2

N.H / 1 !0 N.H /
N2 D exp ! N2 D .1 C z/ (4.7)
2 2 kT 2
A similar approach was already used in the work [10]. We only partially changed
the meaning of x, y, z in according our task.
Here is proportional to deviation of the population of lowest level n3 , y is
proportional to the magnetization on the transition with frequency
, and z is
proportional to the proton magnetization
0 . The new parameters of x, y and z are
introduced instead n1 , n2 , n3 , N1 , N2 .
We assume that in the quasi-steady approximation we can still use the same
parameters of x, y and z to characterize the population differences of the spin levels.
Thus we arrive to the following equations:

1 " d z " dy " 2 "
C D WH C W0 C "WC z C .2WC C W / y (4.8)
2 3 dt 3 dt 3 3 3
1 d z " dy " " "

D WH C W0 WC z C .WC C 2W / y (4.9)
2 dt 3 dt 3 3 3
Finally, the following approximate solution for these equations is derived:
z.t / D c1 exp.k1 t / C c2 exp.k2 t /; (4.10)
with the parameters k1 and k2 as follows:
1
k1 D 2" .WC C W C W0 / C 9 .WC C W / (4.11)
2 ." C 3/
1
k2 D " .WC C W C W0 / C 6WH (4.12)
." C 3/
The solution has been obtained by symbolic calculations using the Maple 10
software packet. Comparing our solution and the solutions obtained in [10] we see
that in our approximation k1 coincide with value of W1 in [10] for case << 1, while
the value of k2 does not match exactly, but it is still very close to the given in Ref.
[10] (the numerical difference is only 10 %).
The relation between the quadrupole polarization y(t) and the proton polarization
z(t) is obtained subtracting Eq. 4.8 from Eq. 4.9:
"
1C
yD" 3 d z .a C b/ z (4.13)
"
.d c/ dt .d c/
3 3
with use of the following notations:

2 1
a D WH C "WC C "W0 ; d D 2WC C W
3 3

1 1
b D WH "WC C "W0 ; c D WC C 2W : (4.14)
3 3
Thus the solution for y(t) similar to z(t) is the two-exponential decay with relative
weights of the terms, which depend on the relaxation parameters as well as on the
initial state of the system in beginning of the magnetic field pulse. Our estimations
reveals that the term with exp(k2 t) will govern the relaxation of the nitrogen spins
if the proton relaxation rate WH is faster than WC , W or W0 . For the relaxation rates
of AN we use the values given in the Ref. [14]: W0 D 0.024 s1 , WC D 0.033 s1 ,
W D 0.015 s1 , and D 0.25. An estimation of the proton autorelaxation time
WH of about 0.32 s1 is given in Ref. [14]. The term with exp(k2 t) represents
the slow NQR spin-lattice relaxation, while the term with exp(k2 t) describes the
faster relaxation channel through the protons mediated by cross-relaxation between
the nitrogen and proton spins. Therefore, it is expected that application of pulse
magnetic field results in an essential shortening of the effective relaxation time.
4.3 Experimental Details
The sample consists of 120 g powder of 99.5 % pure ammonium nitrate NH4 NO3
(AN). In general, five different crystalline phases of AN are known. However, at
typical outdoor temperatures only three of them are relevant to the detection of AN.
The form V (tetragonal phase) is stable below a temperature of 16.8 . The phase
IV (-rhombic) is stable from 16.8 to C32.3 and the phase III (-rhombic) is
stable from C32.3 to C84.2 . AN is a widely used component of explosives
for industrial and military applications (ANFO, ammonite, amatol, ammonal, etc.).
During experiments our sample of AN is at room temperature is in the IV phase.
The most suitable frequency for the detection of AN is
D 424 kHz
(at room temperature). This transition has a smallest temperature coefficient,

/ D C100 Hz/K [16]. For the transition with a frequency of
C D 496 kHz
(room temperature), the temperature coefficient is higher: 300 Hz/K. The spin
lattice relaxation parameter T1 is 13 and 16 s for the NQR frequencies of
C and
, respectively. The spin-spin relaxation parameter T2 is 8 and 6 ms for

C and
,
respectively. Our experiments have been performed at the frequency D 424 kHz.
Tecmag Apollo NQR/NMR console (0.1100 MHz) with two-channel transmitter
and one-channel receiver modules, equipped by two Tomco power amplifiers with
output power of up to 500 W have been used. For multifrequency NQR experiments
the detector unit consists of the system of two orthogonal coils, each of which is
connected to variable capacitor to form a serial/parallel resonance circuit tuned
to its own resonance frequency. For single frequency measurements the probe
consisting of solenoidal coil connected to a variable capacitor has been used. A low-
noise single-channel preamplifier Miteq and auxiliary electronic circuits have been
connected between the probe and Tecmag NMR/NQR Console. A special magnet
system consisting of Helmholtz coil set and pulse current source has been designed.
A set of Helmholtz coils provides a uniform magnetic field (2 % in the sphere with
D D 4 cm) at the sample. The current source for creation of the pulse magnetic field
was home-made. The control pulses were supplied by Apollo console. The output
current of this modulator is in the range 100 mA to 30 A. KEPCO-BOP-20-50MG
power supply has been used as a current source to the modulator. We tested magnet
in a pulse mode up to 30 A which corresponds to the magnetic field of 11.88 mT for
our Helmholtz coil system. The current stability was about of 2 % is for the pulse
duration of 2 s. The NQR radiofrequency coil has been tested in two relative posi-
tions in our experiments: parallel to the pulse magnetic field and perpendicular to the
pulse magnetic field. The magnetic field broadening of nitrogen resonance line at
0
is approximately ( N / H )2
0 [10]. The
0 linewidth for AN in zero magnetic field is
near 70 Hz, while in the magnetic field of 1.7 mT it is about of 370 Hz. It means that
the peak value of the magnetic field pulse should be settled with accuracy of 0.5 %.
Taking into account low current stability in our home made modulator and not very
high uniformity, it is expected that use of the magnetic field which is oscillating or
swept through the resonance could provide the better results. We also used a spe-
cially designed Q-switch to suppress ringing in the resonance circuit of NQR probe.
The spin-lock spin-echo (SLSE) sequence [7] was applied in our NQR ex-
periments with AN sample. To measure the effect of the cross-relaxation in low
magnetic field we change the time interval between sequences. Since the pulse
magnetic field is switched off during the observation of NQR signal, the line
broadening effect of the magnetic field on the effective relaxation time T2eff , which
defines the decay of the signal during the SLSE sequence, is excluded.
4.4 Experiments and Discussion
For matching the energy levels of the proton and nitrogen spin systems we used two
approaches. First one is an application of a single rectangular pulse of the magnetic
field (Fig. 4.2a). Second one is use of series of m short pulses of magnetic field
(Fig. 4.2b, c). In both cases the value of magnetic field B was settled to correspond
to the double resonance conditions for
0 transition (Fig. 4.1). After switching off
the pulse(s) of magnetic field, a multipulse spin-locked spin echo (SLSE) sequence
(Fig. 4.2) was applied to detect an NQR signal at transition. The main goal of
the application of static magnetic pulse B is a decrease of T1 by cross-relaxation
contact between the proton and nitrogen spins. On the other hand, the protons in
the zero magnetic field have equal populations of the C1/2 and 1/2 energy levels
before an application of magnetic field B. This state corresponds to saturation state
of protons system. A pulse of the magnetic field transfers of the protons system
to the double resonance condition. As a result of the matching of energy level
Fig. 4.2 The pulse sequence for cross level in AN with use pulse magnetic field
1.8
1.6
Relative increase of amplitude
2
1.4
Relative increase of amplitude
1.2
1.0
1E-3 0.01 0.1 1

Time (s)
0.8
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Time (s)
Fig. 4.3 The dependence of the signal increase on the duration of magnetic pulse t1 in comparison
with usual signal. The time space between sequences is 10 s. The solid line is only guide for eyes.
Inset is a same plot but given in the log log scale
differences between the proton and nitrogen systems, the

0 transition of quadrupole
spin system is saturated too. It is known that the saturation of the
0 transition
increases the gain on the signal on the observable frequency
[17].
Firstly we studied an effect of the duration of a pulse of the magnetic field on
the signal intensity in the scheme shown in Fig. 4.2a. The time space between the
successive SLSE sequences is 10 s, that is much shorter than the usually used delay
times of (3 5)T1 . The cross-relaxation time according to work [14] is an order
of 10 ms. It is seen in Fig. 4.3 that the magnetic field pulse length less than 1 s is
enough see an essential increase of the SLSE signal intensity. Figure 4.3 shows the
ratio of the SLSE signal with magnetic field pulses to that without the magnetic field
as a function of the magnetic pulse length t1 . It is seen that there is rapid increase
in the signal intensity for t1 < 0.1 s (the gain of 1.6 times is at t1 D 15 ms), while
with further increasing the pulse length at t1 < 0.5 s we have a gradual decrease of
the signal. It is obvious that the region of the fast increase of the signal intensity
(10 ms) agrees well with the cross-relaxation time estimated in the Ref. [14]. The
signal amplitude increases due to shortening of the effective spin-lattice relaxation
time T1 for the detected transition as well as saturation of the transition 0 by
polarization transfer from nitrogen spins to the non-polarized proton system. The
first effect, related the cross-relaxation contact, has been already discussed in
Introduction, while the second contribution should be considered in more details. It
1.0
0.8
Amplitude (a.u.)
0.6
0.4 Without magnetic field

DC field perpendicular to RF field
DC field perpendicular to RF field
exponential fits by 1-exp(-t/T1)
0.2
0.0
0 20 40 60 80
time (s)
Fig. 4.4 The dependence of the NQR signal at transition on the time interval between the
SLSE sequences. The duration of magnetic pulse is 1 s. The solid lines are results of modelling.
Both limiting orientations (perpendicular and parallel) of DC magnetic field with respect to RF
field are given
is explained as a result of the polarization exchange between the proton system and
the transition with frequency 0 . The protons in the zero magnetic field have equal
populations of the C1/2 and 1/2 energy levels. A pulse of the magnetic field means
an adiabatic transfer of the system to the double resonance condition. As result of the
matching of energy level differences between the proton and nitrogen systems, the
0 transition of quadrupole spin system is saturated. Very simple estimations (see,

e.g., Fig. 4.5 in Ref. [13] or an appropriate formula in the Table 1 of Ref. [18]) shows
that the later effect could contribute to an increase at most for 8.6 % (as a factor of
3 /3, D 0.24 for AN). It is clear that the effect of the polarization transfer which
equalizes the populations of nitrogen spin levels at 0 should vanish for longer pulse
duration of the DC magnetic field because of polarizing the proton spins. Therefore
the contribution of this effect vanishes at time of about 1/WH . Taking into account
the estimation of the proton autorelaxation time 1/WH D 3.1 s given in Ref. [14],
we see that the gradual decrease of the signal at time interval of 0.5 3 s in Fig. 4.3
agrees well with the spin-lattice relaxation time for protons. A slightly higher value
(1012 %) of this decrease could be attributed to a very small temperature shift
of the sample temperature induced heating of Helmholtz coils for longer pulses (i.e.
higher duty cycle) of the DC magnetic field.
Secondly we measured the NQR signals as function of the time interval between
the multipulse series for two mutual orientations of RF and Helmholtz coils: parallel
to each other (Fig. 4.4a) and perpendicular to each other (Fig. 4.4b). Depending
on the time interval between the multipulse series we observed an increase of the
signal amplitude up to 3040 %. As it was already discussed above, the main

contribution is due to shortening of the effective spin-lattice relaxation time T1 for
the detected transition . Modelling of the experimental results, obtained in the
scheme shown in Fig. 4.2a, result in the following values of the effective spin-lattice
relaxation time: (a) in the parallel configuration, T1 is equal to 10.0 0.5 s; (b) for
the perpendicular one, T1 is equal to 8.8 0.8 s (see Fig. 4.4).
In very recent work of Prescott et al. [19] it has been shown that the effect
of dipolar interactions between proton and quadrupole systems depends on the
orientations of laboratory frame and principal axes of electric field gradient
(EFG) on nitrogen nuclei with respect to displacement vectors connecting them to
neighbouring protons. Furthermore, even in powder samples the cross-relaxation
rates may be dependent on the mutual orientation of RF excitation and magnetic
field axes.1 Theoretical calculations of the cross-relaxation rates, which require the
structural parameters and EFG axes to be accurately known, are beyond the scope
of our paper. More detailed double-resonance experiments with single crystalline
samples are desirable for experimental verification of the dependence of dipolar
fields on orientations of the RF and static magnetic fields. In this work, we limit
ourselves by making experiments with a powder sample for two possible extreme
orientations of the RF field with respect to the static magnetic field: (1) the RF field
is parallel to the static magnetic field and (2) the RF field is perpendicular to the
static magnetic field.
Better results are obtained in the experimental scheme shown in Fig. 4.2b. We
applied pulses with length of 20 ms and separation of 20 ms. Total number of the
magnetic field pulses was 8. In this case we however applied the pulse magnetic
field B with a peak value slightly larger than optimal one for the static cross
relaxation scheme in Fig. 4.2a. The results of the experiments are presented in
Fig. 4.5 for both parallel and perpendicular geometries. The following values of
T1 have been obtained: (a) for the parallel geometry T1 is equal to 8.6 0.8 s; (b)
for the perpendicular geometry T1 is equal to 7.9 0.6 s. We performed also the
NQR experiment with use of the same spin-locked spin echo (SLSE) sequence to
detect an NQR signal at transition without application of the pulse magnetic
field. The SLSE NQR signal as function of the time interval between the multipulse
series for the same two mutual orientations of RF and Helmholtz coils (parallel
and perpendicular) has been obtained (they are shown by red lines in Fig. 4.4).
Modelling of our experiment results provide the T1 values with the average between
two geometries equal to 16 1 s, which is well corresponds to T1 D 16.9 1 s
cited in the work [16]. Thus, in the scheme shown in Fig. 4.2b at least twofold
shortening the effective spin-lattice relaxation parameter for the transition has
been obtained.
1
Combining of various types of double resonance experiments for all NQR transitions, allows,
however, excluding the influence of a specific orientation of laboratory frame, thereby making
possible to find experimentally the principal axes of EFG even in powder sample [19].
1.0
0.8
Amplitude (a.u.)
0.6
0.4
Without magnetic field

0.2 DC field perpendicular to RF field
DC field parallel to RF field
exponential fits by 1-exp(-t/T1)
0.0
0 10 20 30 40 50 60
time (s)
Fig. 4.5 The dependence of the signal in multipulse SLSE sequence with magnetic pulse sequence
given in Fig. 4.2b on the interval between sequences. A comb consisting of eight pulses with
individual duration of 20 ms and separated by 20 ms has been applied. The solid lines are results
of modelling. Both limiting orientations (perpendicular and parallel) of DC magnetic field with
respect to RF field are given
However, we established that the application magnetic field after the SLSE
sequence is even more effective (Fig. 4.2c). Modelling experimental results for the
parallel geometry gives the T1 value of 3.9 0.4 s (Fig. 4.6). Nearly the same result
is obtained in the perpendicular geometry (not shown). Thus we established that
the protocol, where the magnetic field in the form of the comb of pulses applied just
after the NQR SLSE sequence, provides the most effective shortening of spin-lattice
relaxation time. The more pronounced effect of the protocol in Fig. 4.2c is not clear
yet, because in the limit of large number of repetitions both protocols are expected
to give similar results (for the same number of repetitions).
As regards the results the difference between the results of the protocol in
Fig. 4.2ac we suppose that the application of the comb of magnetic pulses is more
effective due to insufficient current stability in our home made modulator and not
very high uniformity. Therefore, use of the magnetic field which is oscillating or
swept through the resonance, could be more practical. In the case of the comb
of the magnetic field pulses the proton-nitrogen spin systems move through the
double resonance conditions many times. Owing to very fast cross-relaxation time
(10 ms), the cross-relaxation works even in alternating (fast-switching) magnetic
fields. We plan to study further an effect of switching frequency of the pulse
magnetic fields, which in turn could provide information on the cross-relaxation
rates of the proton-nitrogen spin systems.
1.1
1.0
0.9
0.8
0.7
Scaled (a.u.)
0.6
0.5
0.4
with pulse of DC field
0.3 without pulse of DC field
0.2 modelling by 1-exp(-t/T1)
0.1
0.0
0 10 20 30 40 50 60
Time (s)
Fig. 4.6 The dependence of the NQR signal amplitude on the time interval between the successive
sequences in the experimental scheme given in Fig. 4.2c. The DC magnetic field is applied
after the SLSE NQR sequence as the comb consisting of m pulses. The DC magnetic field is
aligned in parallel to the axis of RF coil. The solid lines are results of modelling. The value of
T1 D 3.9 0.4 s is obtained for the pulse of DC magnetic field applied
4.5 Conclusions
As was mentioned in the introduction, there are two possible double-resonance

methods of the increasing the signal to noise ratio in the direct NQR detection: (1)
pre-polarization of proton system with use of a high magnetic field with following
level crossing to polarize the nitrogen spin system by cross-relaxation; (2) the cross-
relaxation method in lower magnetic field to shorten the spin-lattice relaxation time
of 14 N system. In the first case, the sample is needed to be pre-polarized with typical
polarization time of about 30 s before decreasing the field down to zero with the
following detection of NQR signal. Although the gain is quite large (an order of
ten times), the long duration of the pre-polarizing magnetic field and need in the
system providing a strong magnetic field prevent using this method in the luggage
detection devices. In contrast, the second approach looks easier to realize in practice.
In this work we studied an application of small magnetic fields (1.42.2 mT) of very
short total duration (1 s). Although in our case the gain is not as high as in the
pre-polarized double resonance, the method is more appropriate for practical use.
The main advantage of the proposed approach is shortening of T1 . It is well known
that the time delay between the multipulse series is needed to be (3 5)T1 . The
decrease of T1 allows faster accumulation that make the detection time shorter. In
our experiment we obtained the fourfold shortening of the effective T1 . Using rather
low magnetic fields makes easier to implement this protocol in the luggage scanners
for explosives with technical point of view as well as with respect of the safety
regulations restricting the use of the magnetic field in such devices.
Thus, our cross-relaxation experiments revealed that it is possible to use a pulse
or pulse series of the DC magnetic field of short duration both for increasing the
amplitude of the NQR signal in the SLSE sequence and for shortening the spin-
lattice relaxation time T1 . The later, in turn, results in increasing the signal intensity
further due to increased rate of NQR signal averaging.
Acknowledgments This work was supported by NATO Science for Peace and Security Pro-
gramme under Science for Peace project No. 982836, and by the TBITAK/RFBR joint project
programme, grant No. 212T131/13-03-91372_CT_a. Authors also acknowledge the referees of this
paper for their very valuable suggestions and comments.
References
1. Miller JB, Barrall GA (2005) Explosives detection with nuclear quadrupole resonance. Am Sci
93(1):5057
2. Garroway AN, Buess ML, Miller JB, Suits BH, Hibbs AD, Barral GA, Matthews R, Burnett
LJ (2001) Remote sensing by nuclear quadrupole resonance. IEEE Trans Geosci Remote
39(6):11081118
3. Svensson A, Jakobsson A (2011) Adaptive detection of a partly known signal corrupted by
strong interference. IEEE Signal Process Lett 18(12):729732
4. Jinping Niu, Tao Su, Xuehui He, Kairan Zhu, Huiyang Wu (2010) Weak NQR signal detection
based on generalized matched filter. Procedia Eng 7:377382
5. Suits BH, Garroway AN, Miller JB (1998) Noise-immune coil for unshielded magnetic
resonance measurements. J Magn Reson 131:154158
6. Mozzhukhin GV, Efremov AV, Bodnya AV, Fedotov VV (2005) A two-spiral flat coil for
detecting 14N NQR signals. Russ Phys J 48(9):978983
7. Hirshfeld T, Klainer SM (1980) Short range remote NQR measurements. J Mol Struct 58:
6377
8. Rudakov TN, Hayes PA (2006) Cross-polarisation method for improvement of 14N NQR
signal detectability. J Magn Reson 183(1):96101
9. Seliger J, Zagar V (2009) Double resonance experiments in low magnetic field: dynamic
polarization of protons by 14N and measurement of low NQR frequencies. J Magn Reson
199(2):199207
10. Seliger J, Zagar V, Blinc R (1994) New highly sensitive 1 H-14 N nuclear-quandrupole double-
resonance technique. J Magn Reson A 106(2):214222
11. Grechishkin VS, Anferov VP (1980) Two-frequency method and double nuclear quadrupole
resonance. In: Smith JAS (ed) Advances in nuclear quadrupole resonance, vol 4. Heyden,
London, pp 71113
12. Blinc R, Apih T, Seliger J (2004) Nuclear quadrupole double resonance techniques for the
detection of explosives and drugs. Appl Magn Reson 25:523534
13. Thurber KR, Sauer KL, Buess ML, Klug CA (2005) Increasing 14N NQR signal by 1H14N
level crossing with small magnetic fields. J Magn Reson 177:118128
14. Prescott DW, Olmedo O, Soon S, Sauer KL (2007) Low-field approach to double resonance in
nuclear quadrupole resonance of spin-1 nuclei. J Chem Phys 126:204504
15. Rudakov TN (2012) Some aspects of the effective detection of ammonium nitrate-based
explosives by pulsed NQR method. Appl Magn Reson 43(4):557566
16. Barras J, Gaskell MJ, Hunt N, Jenkinson RI, Mann KR, Pedder DAG, Shilstone GN, Smith
JAS (2004) Detection of ammonium nitrate inside vehicles by nuclear quadrupole resonance.
Appl Magn Reson 25:411437
17. Mozzhukhin GV, Rameev BZ, Dogan N, Aktas B (2011) The application of the two frequency
composite pulses for NQR detection of nitrogen-based compounds. J Supercond Nov Magn
24:653658
18. Mozzhukhin GV, Rameev BZ, Dogan N, Aktas B (2009) In: Fraissard J, Lapina O (eds)
Explosives detection using magnetic and nuclear resonance techniques, NATO science for
peace and security series B: physics and biophysics. Springer, Dordrecht, pp 205230
19. Prescott DW, Malone MW, Douglass SP, Sauer KL (2012) Rabi and Larmor nuclear quadrupole
double resonance of spin-1 nuclei. J Chem Phys 137:214201, 112
Chapter 5
Investigating Homonuclear Broadening in NQR
with Carr-Purcell Meiboom-Gill Performed on
p-Chloroaniline
Michael W. Malone and Karen L. Sauer
Abstract We investigate the possibility of obtaining a strong homonuclear response

in the NQR signal from a sample expected to have stronger broadening due to
heteronuclear dipolar coupling than homonuclear coupling. Previous work, also
with a powder sample, revealed a strong response from homonuclear coupling in
a substance without a strong heteronuclear broadening component. This new work
could expand the number of samples for which those conclusions are applicable,
creating more opportunities to use NQR to measure homonuclear dipolar coupling
for both material characterization and determining efficient detection procedures.
5.1 Introduction
Despite the benefits of unique frequency signatures and a simple experimental

apparatus, the adoption of nuclear quadrupole resonance (NQR) spectroscopy for
substance detection has been impeded due to its low signal to noise ratio (SNR).
In addition to the typically low transition frequencies, particularly for nitrogen
nuclei, and the difficulty of optimally exciting powder samples, various NQR line-
broadening mechanisms can cause the net signal to dephase and produce a rapidly
decaying signal in time.
An optimal radio frequency (rf) excitation, or =2 pulse, produces a free
induction decay (FID) signal that decays with a time constant T2 . A series of N
refocusing pulses of duration tp separated by time 2 can refocus selective broad-
ening mechanisms, depending on the nature of the pulses [16]. These multi-pulse
sequences can therefore extend the signal in time with a new time constant T2e ,
potentially increasing SNR by an order of magnitude or more compared to the
M.W. Malone K.L. Sauer ()

George Mason University, Fairfax, VA, USA
e-mail: ksauer1@gmu.edu
62 M.W. Malone and K.L. Sauer
FIDs signal. Our previous work [6] focused on how the broadening mechanisms
of electric field gradient (EFG) inhomogeneity and homonuclear dipolar coupling
respond to the standard Carr-Purcell Meiboom-Gill [7] (CPMG) .=2/x C C
y C N and Spin-Lock Spin-Echo [8] (SLSE) .=2/x C C .=2/y C N multi-
pulse experiments, where the subscripts x and y correspond to the phase of the
pulse. For spin 1, it was shown that the CPMG refocused dephasing due to EFG
inhomogeneity, while the SLSE did not. Conversely, the SLSE refocused homonu-
clear coupling while the CPMG did not. This selective refocusing is significant for
substance detection because the sizes of the two broadening mechanisms determine
which sequence is better for detection. Additionally, we described the specific
experimental conditions under which a strong response due to the unrefocused
homonuclear coupling could be triggered, even for a powder sample. This allows
for a measurement of the dipolar coupling, which is a useful tool for material
characterization.
The previous work was performed on the nitrogen in powder samples of sodium
nitrite NaNO2 due to the samples relatively strong signal size and well known
structure. One broadening mechanism that sodium nitrite lacked, however, was
strong heteronuclear dipolar coupling: a second order [9] broadening mechanism
which should be common in hydrogen-rich explosives and narcotics. To see if the
homonuclear response found in the previous work could still be found in a substance
with a large heteronuclear component, we began investigating the nitrogen in p-
chloroaniline ClC6 H4 NH2 . While sodium nitrite should have some heteronuclear
coupling between the nitrogen and sodium, it is expected to be much weaker than the
coupling between the pair of hydrogen nuclei and the nitrogen in p-chloroaniline.
This follows from the second order broadening h ! 2 i due to heteronuclear coupling
being proportional to the fourth power of the gyromagnetic ratio of the non-nitrogen
nuclei, while also decreasing faster than the distance separating the nuclei raised
to the sixth power. Since hydrogen has a gyromagnetic ratio approximately four
times the value of sodium [10], and the nitrogen-sodium bond length in sodium
nitrite [11] is approximately 2.5 times the nitrogen-hydrogen bond lengths in the
NH2 group in p-chloroaniline [12], p-chloroaniline should have much stronger
heteronuclear coupling. The theoretical heteronuclear linewidth broadening based
on
q the second moment calculation by Vega [9] gives a value on the order of
h ! 2 iNH =2 D 1 kHz, which is more than an order of magnitude larger than
the theoretical
q broadening due to homonuclear coupling calculated by Sauer and
Klug [5] of h ! 2 iNN =2 D 17 Hz at !x .
5.2 First Observations
We performed room temperature experiments on two 100 g samples of 98 % purity

p-chloroaniline from Sigma-Aldrich. We used a home built probe with a quality
factor of 30 and dimensions large enough to hold the original sample container and
ensure a uniform field across the samples. The experiments required an rf input
5 Investigating Homonuclear Broadening in NQR with Carr-Purcell. . . 63
Fig. 5.1 The average echo

size as a function of j f j 0
reveals an asymmetry around
j f j D 1=.4 0 /, with the
signal at f D 1=.4 0 /
being 75 % the value at
f D 1=.4 0 /. Data here
and elsewhere has been
normalized to the initial peak
signal strength from an FID
sequence. The value of
0 D C tp =2 for this
experiment was 0.882 ms,
with tp D 110 s. Each data
point is the average signal of
35 echos
power of 1,000 W or less for pulses of length tp D 110 s, and were repeated over
at least 160 scans. To remove spurious signals, the excitation pulse was varied by
180 in phase between scans with alternate scans subtracted. Since the nitrogen
nuclei only share the x-axes of the principal axes frame of their EFGs, only the
2 fNQR D !x transition was investigated.
The full width at half maximum (FWHM) linewidth, obtained from a series of
FIDs, for both powder samples was 110 6 Hz, with the line shapes comparable
to that observed by Ambrosetti for a single crystal [13]. By fitting the lineshape
to
p a Voigt profile, the second moment of the Gaussian contribution estimates
h ! 2 i to be 50 12 Hz, which is much larger than the expected homonuclear
broadening of 17 Hz, but much less than expected from the second moment of
heteronuclear coupling from Vega. In an attempt to distinguish broadening due
to EFG inhomogeneity from homonuclear coupling, one sample was melted and
allowed to recrystallize. While cooling, the recrystallized sample formed smaller
dendritic crystals, resulting in a larger FWHM of 130 6 Hz. The unaltered sample
is referred to as the narrow-sample, and the recrystallized sample the broad-sample.
All results refer to the narrow-sample unless explicitly mentioned.
The homonuclear response for sodium nitrite manifested itself in time as a strong
Gaussian signal with a decay rate determined by the homonuclear dipolar coupling
[6]. Experimentally this response was achieved using both a sufficiently small ratio
of tp = and a CPMG operated at an off-resonance f D frf fNQR D 1=.4 0 /,
where 0 D C tp =2. To see if we could trigger p-chloroaniline to provide the
Gaussian response, multiple CPMGs were performed across a frequency range of
j f j 0 2. The results, shown in Fig. 5.1, showed a strong asymmetry in the
average echo data around j f j D 1=.4 0 /. This was a welcome sign because
the asymmetry had been observed before and explained as a response of the
homonuclear coupling refocusing to the shape of the pulse.
As confirmation that homonuclear coupling was driving the asymmetry, the
actual pulse shapes were measured using a small sniffer coil located several
centimeters from the main coil. The representative pulse shape, shown in Fig. 5.2,
a b
Fig. 5.2 (a) The shape of the 110 s refocusing pulse of the CPMG sequences with the probe
tuned to !x . The presence of the non-zero imaginary component explains the consistent presence
in the average echo data of the asymmetry around j f j D 1=.4 0 /, such as shown in Fig. 5.1.
(b) Using the observed pulse shape, the percent of the signal experiencing a strong homonuclear
response can be predicted, pt h . f /, as explained in our previous work [6]. The ratio of this
prediction for the two off resonance conditions, pt h .1=4 0 /=pt h .1=4 0 /, is found to correlate
with the ratio of the average echo size, avg. echo.1=4 0 /=avg. echo.1=4 0 /, confirming that it
is homonuclear dipolar coupling that drives the asymmetry
contains a non-zero imaginary component. This is what had previously been shown
to produce an asymmetry in the average echo data by selectively triggering the
homonuclear response [6]. As shown in Fig. 5.2, a predicted asymmetry in the size
of the Gaussian response, calculated using the actual pulse shape, neatly tracked the
observed asymmetry in the average echo size across multiple values of , even while
ignoring the contribution of EFG to the linewidth.
Focusing on the off-resonance conditions f D 1=.4 0 / and 0 it was found
that the echo trains generally did not fit well to a single exponential function.
Instead, as shown in Fig. 5.3 they fit well to a function composed of the addition
of a Gaussian and exponential functions:
" 2 # !
1 t t
S.t / D A pg exp C .1 pg / exp ; (5.1)
2 Td 0 T2e
where A is the initial amplitude, pg the percent of the signal due to a Gaussian
component, Td the width of the Gaussian component, and T2e the long term decay
constant. The factor of = 0 is added to the Gaussian term since effectively no
dipolar evolution occurs during the pulse [6]. These fits showed that, in addition to a
faster T2e , the asymmetry was correlated with a larger Gaussian component, further
strengthening the case that the rapid decay was due to the homonuclear coupling.
These experiments were repeated across multiple values of , and the net results
in Fig. 5.4 show a generally stronger Gaussian response with increased . This is
consistent with the theoretical understanding that as increases the homonuclear
Fig. 5.3 Corresponding to the data in Fig. 5.1, the individual signal for each echo in the CPMG
echo train is plotted against time for three resonance conditions. A stronger Gaussian component
appears in the echo train below resonance ( f D 1=4 0 down-facing triangles) than for the
resonant ( f D 0 squares), or above resonance ( f D 1=4 0 upward-facing triangles) cases.
This is consistent with the asymmetry observed in Fig. 5.1. Fits of the data (lines) were found using
Eq. 5.1. Echo data here and elsewhere was acquired stroboscopically at time after each refocusing
pulse
a b
Fig. 5.4 The CPMG echo train fit parameters pg (a) and Td (b) were found for multiple values of
for the samples and off resonant conditions given in the legend of (b). The increasing value of
pg with is consistent with the behavior observed for sodium nitrite. The values of Td approach
a maximum value above the prediction for a single crystal (16.2 ms) and below the value expected
for a powder (18.8 ms). Data points with pg consistent with zero were not included in the graphs
response becomes less sensitive to the finite length of tp . As expected for the
measurement of this intrinsic parameter, the average of Td for both the narrow
16:0 0:5 ms and broad samples 15:5 0:4 ms are in agreement. These were
calculated from sequences with j f j D 1=.4 0 / and for all values of above
Fig. 5.5 The long time decay

constant T2e as a function of
for the off-resonance values
f D 0; 1=.4 0 / is on the
order of the expected Td for a
powder sample of 18.8 ms,
and for short approaches
the observed T1
the theoretical
p minimum needed to observe a strong homonuclear response [6]:
min D tp =DC C D 1:2 ms, where DC C D 2 12 Hz for p-chloroaniline [5].
As shown in Fig. 5.5, T2e varied from two to three times the expected Td as the
various experimental parameters were changed. This is in contrast to sodium nitrite
where the T2e s were consistently a factor of 10 or more greater than Td . That the
homonuclear response can still be seen against this other comparable time constant
is indicative of the robustness of the measurement process.
5.3 Conclusion
We observe a Gaussian response in p-chloroaniline consistent with the behavior

of the homonuclear response previously observed in sodium nitrite. This occurs
in spite of both the relatively weak homonuclear contribution to the linewidth and
the heteronuclear broadening that this new sample is expected to have. That we
are observing the homonuclear coupling in p-chloroaniline is supported by the
observation of a significant Gaussian component in the CPMG echo train with a
width close to the expected value, under the experimental parameters where it is
expected to appear: j f j D 1=.4 0 / and small values of tp = , recognizing that there
is sensitivity to the shape of the pulses. While additional work is needed to under-
stand the theoretical aspects, this result suggests that our procedure for measuring
homonuclear dipolar coupling is useful even for samples with heteronuclear dipolar
broadening. For the purpose of material characterization, this would substantially
increase the number of samples that could be targeted. Additionally, by providing a
tool to measure the size of this broadening mechanism, more efficient detection of
NQR substances is possible.
Acknowledgements This work was supported by NSF Award Number 0547987.

References
1. Osokin D Ya (1982) Coherent multipulse sequences in nitrogen-14 NQR. J Mol Struct

83:243252
2. Osokin D Ya (1980) Pulsed line narrowing in nitrogen-14 NQR. Phys Status Solidi B
102(2):681686
3. Zueva OS, Kessel AR (1982) Dynamic line narrowing of nuclear quadrupole resonance. J Mol
Struct 83:383386
4. Mikhaltsevitch VT, Rudakov TN (2004) Study of quasistationary and stationary states in
the short-repetition-time sequences in the NQR of nitrogen. Solid State Nucl Magn Reson
25(13):99111. Thirty-first congress Ampere, magnetic resonance and related phenomena
5. Sauer KL, Klug CA (2006) Spin dynamics in the pulsed spin locking of nuclear quadrupole
resonance. Phys Rev B 74(17):174410
6. Malone MW, McGillvray M, Sauer KL (2011) Revealing dipolar coupling with NQR off-
resonant pulsed spin locking. Phys Rev B 84:214430
7. Meiboom S, Gill D (1958) Modified spin-echo method for measuring nuclear relaxation times.
Rev Sci Instrum 29(8):688691
8. Marino RA, Klainer SM (1977) Multiple spin echoes in pure quadrupole resonance. J Chem
Phys 67(7):33883389
9. Vega S (1973) Advances in magnetic resonance, vol 6. Academic Press, New York
10. Sheriff RE, Williams D (1951) Nuclear gyromagnetic ratios. III. Phys Rev 82:651655
11. Carpenter GB (1952) The crystal structure of sodium nitrite. Acta Crystallogr 5(1):132135
12. Trotter J, Whitlow SH, Zobel T (1966) The crystal and molecular structure of p-chloroaniline.
J Chem Soc A 353356
13. Ambrosetti R, Colligiani A, Grigolini P, Salvetti F (1974) Conformation of the -NH2 group in
a single crystal of p-chloroaniline as determined by combined Zeeman effects on the 35 Cl and
14
N nuclear quadrupole resonances. J Chem Phys 60(2):459467
Chapter 6
Size Effect in 14 N Nuclear Quadrupole
Resonance Spectroscopy
Nikolay Sinyavsky, Georgy V. Mozzhukhin, and Philip Dolinenkov
Abstract The influence of the size effect of the crystallites in powders on the form
and width of spectral lines, on the spin-spin and spin-lattice relaxation parameters
of the nuclear quadrupole resonance (NQR) of 14 N nuclei in sodium nitrite was
studied. It was established that a decrease of the average crystallite size produces the
widening of the NQR lines and the shortening of the relaxation times in the direct
method of NQR detection. It was supposed that these are the results of the spin-spin
diffusion process. A multi-exponential inversion of the decays of the longitudinal
and transverse components of the nuclear magnetization was used to obtain the
distribution of relaxation times.
6.1 Introduction
The main identification parameters for the nuclear quadrupole resonance (NQR)
explosives detection are NQR frequency, a resonance line width and relaxation
parameters like spin lattice relaxation time constant T1 and spin-spin relaxation
time constant T2 . The knowledge of the temperature dependence of the quadrupole
N. Sinyavsky ()
Department of Physics, Baltic State Academy, Molodiozhnaya 6, 236029 Kaliningrad, Russia
e-mail: n_sinyavsky@mail.ru
G.V. Mozzhukhin
Department of Physics, Kazan State Power Engineering University, ul. Krasnoselskaja 51, 420066
Kazan, Russia
Department of Physics, Gebze Institute of Technology, P.K.141, 41400 Gebze-Kocaeli, Turkey
e-mail: mgeorge@yandex.ru
P. Dolinenkov
Department of Radiophysics and Information Safety, Immanuel Kant Baltic Federal University,
A. Nevsky St. 14, 236041 Kaliningrad, Russia
70 N. Sinyavsky et al.
frequency, T1 and T2 for NQR data are also needed. A research of the influence of
the size effects on various physical properties of materials is important for NQR
applications. The properties of materials in porous matrices depend on the size and
geometry of the pores, on the state of the internal pore surfaces, on the length of the
molecular chains of polymer molecules and etc. [1, 2]. However the results of the
NQR detection should be free from the significant influence of the size of powder
grains and the disposition of the explosives in porous materials. Thus, the size effects
should be considered for NQR detectors.
The relaxation times reflect the molecular dynamics, and also, to some extent,
provide information about the microstructure of the material. The decay of NQR
signals in homogeneous samples is defined by one exponential function. In this
case, one relaxation time parameter can be accurately determined by exponen-
tial approximation. However, there are other samples as molecular crystals with
impurities, micro and nano powders, porous substances. The distribution of the
relaxation parameters for these samples is continuous due to different intermolecular
interactions. The obtaining of the distribution of relaxation parameters requires a
multi-exponential inversion. In recent years, there were proposed a lot of algorithms
and applications of the inversion [36]. However, their effectiveness is not high due
to their instability.
The longitudinal and transverse relaxation time constants for the molecules of
liquids in pores and in small particles are reduced in comparison to the relaxation
parameters in bulk homogeneous samples. This is typical for the nuclear relaxation
of water in biological cells, fluids in porous samples and powders consisting of small
solid particles. In all these cases the main physical model of process is the following:
(i) the precessing magnetization diffuses to the surface; (ii) the magnetization has
decay in a very short time due to the strong coupling with the lattice (for the
longitudinal magnetization) and due to the gradient of the internal field (for the
transverse component of magnetization).
Spin diffusion transfers redundant nuclear spin polarization in the direction
of smaller polarization (or spin temperature) for leveling-off of spin temperature
in the sample. The usual diffusion is associated with the transport of the mass,
but the spin diffusion is only transfer of the spin excitation without the movement
of the particles with the magnetic moments. The measurement of the rate of spin
diffusion, depending on the direct dipole-dipole coupling, provides information
about the distances between nuclei in the sample that can be used for the study
of inhomogenuities.
In work [7] the spin-lattice relaxation constant T1 in powder samples was
analyzed on the assumption that the spin diffusion in solids is similar to molecular
diffusion in liquids. It is shown that the decay of the nuclear magnetization of the
solid state particles has a multi exponential dependence. It was concluded that the
relaxation curve may be used to determine the spin diffusion coefficient, surface
relaxation and the relative size of the particles.
The double nuclear quadrupole resonance of 14 N, 2 D and 23 Na nuclei [8] was
used for the research of the influence of grinding of the particles in powders on
spin-lattice relaxation time constants. The mitigation of the spin-lattice relaxation
time for smaller size of particles was detected. An indirect method for the detection
6 Size Effect in 14 N Nuclear Quadrupole Resonance Spectroscopy 71
of NQR didnt allow the authors of this research to detect changes in the line width.
The authors propose to reduce the relaxation time for samples by means of grinding
the powder sample, because the inclusion of paramagnetic impurities in various
materials is not always possible. Moreover, the inclusion of paramagnetic impurities
is sometimes ineffective. Additionally, paramagnetic impurities decrease the purity
of the sample and can change the structure of the material.
The aim of this work is the experimental study of the size effect in the direct
NQR method of the 14 N nuclei in powder, the research of the distribution of
relaxation parameters by means of the procedure of inverse Laplace transform, and
the explanation of the changes in line width and relaxation parameters.
6.2 Experimental Results and Discussion
The experiments were performed on the NQR spectrometer Tecmag Apollo with
software NTNMR. A sequence of 90 1 90 pulses and a pulse sequence
of inversion-recovery: 180 1 90 were applied to measure the spin-lattice
relaxation time T1 . The sequence of Carr Purcell Meiboom Gill (CPMG)
was used to measure the spin-spin relaxation time of T2 . The accuracy of the
frequency measurements and width of the 14 N NQR lines in sodium nitrite NaNO2
and hexamine C6 H12 N4 is 100 Hz.
We used 99 % chemically pure commercial samples of sodium nitrite NaNo2
and hexamine C6 H12 N4 for our measurements. The sizes of powder particles were
changed by grinding in a mortar and ball mill and further refined by use of a set of
sieves. The sample was obtained also by means of crystallization from an aqueous
solution. A microscope with an ocular micrometer was used to measure the size
of the granules. The accuracy of these measurements of the mean diameter of the
granules is 5 m.
To solve the problem of detecting the inversion of the distribution functions of
relaxation times T1 and T2 we used the following expressions:
X
m
1
y.1 / D xj 1 2 exp. / ;
j D1
T1j
X
m
2
y.2 / D xj exp ;
j D1
T2j
where y is relative amplitude of the signal, xj is an amplitude for j signal with

constant T1j and T2j , m is the number of components.
Additionally, for non-linear fitting of data there were used the least squares
method (solve nonlinear data-fitting problem by means of least-squares method)
with the procedure lsqcurvefit (in the software of MATLAB 7.0.1).
The effects of the influence of crystallite size in powder on the shape and width
of the 14 N NQR lines are demonstrated in Fig. 6.1a, b, where <d> the average
Fig. 6.1 The changes of line widths in the hexamine (a) <d> D 100 m (1) and <d> D 40 m
(2) and in sodium nitrite (b) <d> D 42 m (1) and <d> D 14 m (2); T D 300 K
size (diameter) of the crystallite in the powder. The widths of the Gaussian lines
for hexamine as shown in Fig. 6.1a, are equal to 571 and 900 Hz for samples
1 (<d> D 100 m) and 2 (<d> D 40 m), correspondently. The width of the
NQR line of 14 N nuclei for NQR frequencies C D 4.6367 MHz in NaNO2 , as
shown in Fig. 6.1b, increases from 293 Hz for <d> D 42 m (1) to 587 Hz for
<d> D 14 m (2).
The results of inversion of the Laplace transform by applying of the function
lsqcurvefit to analyze the envelope of echo signals in CPMG sequence for the
sodium nitrite are shown in Fig. 6.2. It is clear that to decrease the size of the powder
particles of NaNO2 the distribution curve of T2 relaxation time constant is shifted
to small transverse relaxation time constants. From the distributions in Fig. 6.2
most probable values of the spin-spin relaxation time constant T2 for the sample
1 (<d> D 42 m) and for the sample 2 (<d> D 14 m) are equal to 20.0 0.1 ms
and 18.9 0.1 ms, correspondently. The distributions of the relaxation time constant
T2 in Fig. 6.2 show that there are also two peaks at 3.5 0.1 ms (<d> D 42 m)
and 2.5 0.1 ms (<d> D 14 m). We suppose that these peaks can be related to
the contribution of surface atoms with short times relaxation.
The relaxation time constants T1 in a polycrystalline sample C6 H12 N4 at the
resonant frequency 3.306 MHz at room temperature were obtained by means
of the sequence of 180 t1 90 pulses. The distributions of the spin-lattice
relaxation time constants T1 was obtained by means of the inversion of the
longitudinal component of magnetization decay and it is shown in Fig. 6.3. The
obtained values of the spin-lattice relaxation time constants are the following:
T1 D 38.0 0.1 ms for sample 1 (<d> D 100 m) and T1 D 22.0 0.1 ms for
sample 2 (<d> D 40 m).
Fig. 6.2 The distribution of spin-spin relaxation time in sodium nitrite: sample 1 (<d> D 42 m)
and sample 2 (<d> D 14 m)
Fig. 6.3 The distribution of spin-lattice relaxation time constant T1 in the hexamine for sample 1
(<d> D 100 m) and sample 2 (<d> D 40 m)
The grinding of the powder has an influence on the surface of the granules and,
in some extent, on the microstructure under surface. Similar measurements were
carried out on the samples with different sizes of the particles obtained by sieving.
A certain amount of particles of different sizes was obtained by crystallization from
aqueous solution. However, the influence of the sizes of the granules on the line
width and the relaxation are not changing.
In the result of the performed experiments, we can conclude that the reduction
of the relaxation constants was caused by an increase in the ratio of quantity of
atoms on the surface of the powder particles to the quantity of atoms inside of the
particles. It is clear that fluctuations of the atoms on the surface of the crystallites
in the powder have larger amplitude, especially along the normal to the surface, in
comparison to that of the atoms inside the volume. The high mobility of atoms on
the surface produces reduced relaxation constants for spins on the surface. However,
the small relaxation time constants of the spins on the surface cannot be a single
factor in the reduction of the relaxation time constants in the powder samples. Let
us say, for very small size of crystallites in powders the quantities of the atoms
on the surface are only a few percent of all atoms in the crystallite. Thus, another
mechanism in the process of reduction of the relaxation time constants should be
involved.
Bloembergen in work [9] suggested that the processes of spin-spin diffusion are
responsible for the heat transfer to paramagnetic impurities, those transfer the heat
to the lattice. In our case, the same mechanism is responsible for the transfer of
heat to the surface spins. The spin-lattice relaxation time constant of the spins at the
surface is very small. Therefore, these spins easily transfer the heat from the major
part of spins to the lattice.
A simple model proposed in the work of Rabbani and Edmonds [8] may be used
for the explanation of the process of spin-lattice relaxation for powder samples.
This model is based on the account of the spin diffusion. For the experiment with
the sequence of 180 t1 90 pulses, the longitudinal magnetization corresponds
to the following formula [8]:
( " X1
#)
Tt
T1 T1s 6 T T
1 T 1s 1 n2 22 Dt
M D M0 1 2e 1 e T1 C 2 1e 1 e b ; (6.1)
nD1
n2
where M0 is the equilibrium magnetization, T1s is the spin-lattice relaxation on the

surface of the powder particles, D is the spin diffusion coefficient, b is the radius of
the powder particles. The time required to transfer magnetization M from the center
of the spherical powder particles to its surface, is approximately equal to b2 /D.
The result of the approximation of the signal for the recovery of the
longitudinal magnetization for hexamine (6.1) for the number of averages
n D 100 is shown in Fig. 6.4. We obtained the following data for coarse powders
(<d> D 100 m),: D/b2 D 1.7107 s1 , T1 D 37.9 ms, T1s D 22.4 ms. We have for
the (<d> D 40 m): D/b2 D 2.4105 s1 , T1 D 23.9 ms, T1s D 2.5 ms. These results
produce the spin diffusion coefficient D D 4.21012 cm2 /s for coarse powders, and
D D 9.61011 cm2 /s for ground powder. It is clear that the decrease of the grain
size leads to the increase of the rate of spin diffusion. Usually the estimation of
spin diffusion coefficients are calculated for the model obtained for a cubic crystal:
a2
D 30T 2
where a distance between the nearest nuclear spins [8]. The reduction
Fig. 6.4 Simulation of the recovery of longitudinal magnetization for C6 H12 N4 by means of
function (1); curve 1: <d> D 100 m, D/b2 D 1.7107 s1 , T1 D 37.9 ms. T1s D 22.4 ms, curve
2: <d> D 40 m, D/b2 D 2.4105 s1 , T1 D 23.9 ms. T1s D 2.5 ms
of the transverse relaxation time T2 as a result of the reduction in powder granules

size also increases the rate of spin diffusion, although this estimation produces a
lower value of diffusion coefficient D [10].
However, we have some mismatches with the theoretical model described in [8]
and by Eq. (6.1). After grinding the powder particles have not a spherical form
and are not uniform in size. Moreover, the model assumes that only the surface
of the sphere has a fixed temperature. In reality, a few outer layers of each grain
produce very small relaxation time constants. The grains have dislocations, which
correspond to additional surfaces. Thus, the actual average diameter of grains is
bigger in comparison to the uniform surface of grains.
A detailed discussion about the roles of impurities, dislocations, lattice defects
and the temperature dependence of the NQR parameters is beyond the scope of this
paper and requires a further study of size effects, which we plan to continue.
6.3 Conclusion
We investigate the influence of the crystal size in the powder on the form and
the width of the NQR line 14 N, on the spin-spin and spin-lattice relaxation time
constants. It was shown that the width of the NQR line increases and the relaxation
time constants decrease with the decrease of the average crystallite size. We explain
these results by an increase in the relative number of surface atoms and the surface
condition of the crystallites and the diffusion of the magnetization to the surface of
the granules. The constants of NQR spin-lattice relaxation on the surface and inside
the volume of the crystallite were obtained too. Additionally, we suggested using the
inversion of NQR signals in time domain for the distribution of relaxation time
constants in powder microparticles. We suppose that these results may be useful
for the development of NQR explosives detectors, for the research of the diffusion
of nuclear magnetization and the surface condition of micro-nano-sized crystals of
solids.
Acknowledgements One of us (NS) thanks the Russian Foundation for Basic Research (RFBR,
grant 11-03-00124 a) for financial support. One of us (GM) was supported by NATO Science
for Peace and Security Program, under Science for Peace project No. 982836.
References
1. Milia F, Fardis M, Papavassiliou G, Leventis A (1998) NMR in porous materials. Magn Reson
Imaging 16(56):677678
2. Ramesh KP (2010) NMR studies of disorder in condensed matter systems. Annu Rep NMR
Spectrosc 71:139175
3. Hornemann JA, Codd SL, Romanenko KV, Seymour JD (2009) T2T2 exchange in biofouled
porous media. Diffus Fundam 10:1.11.3
4. Borgia GC, Brown RJS, Fantazzini P (2000) Uniform-penalty inversion of multiexponential
decay data. J Magn Reson 147:273285
5. Song YQ, Venkataramanan L, Hurlimann MD, Flaum M, Frulla P, Straley C (2002) T1T2
correlation spectra obtained using a fast two-dimensional Laplace inversion. J Magn Reson
154:261268
6. Sternin E (2007) Use of inverse theory algorithms in the analysis of biomembrane NMR data.
Methods Mol Biol 400:103125
7. Rabbani SR, Mendonc C, Mamania JB, Cervantes HR (2006) Analysis of nuclear relaxation in
granular systems. Braz J Phys 36(1A):2833
8. Rabbani SR, Edmonds DT (1994) Nuclear spin-lattice relaxation-time reduction in small
particles. Phys Rev 50(9):61846188
9. Bloembergen N (1949) On the interaction of nuclear spins in a crystalline lattice. Physica
15:386426
10. Khutsishvili GR (1965) Spin diffusion. Uspekhi Fizicheskikh Nauk(rus) 87(2):211254
Chapter 7
NQR Detection of Sodium Nitrite Recrystallized
in Wood
Jrmy Jover, Sarra Aissani, Laous Guendouz, Andr Thomas,

and Daniel Canet
Abstract Cylindrical pieces of wood which can be placed in 10 mm o.d. NMR

sample tubes have been impregnated by an aqueous solution of sodium nitrite
(NaNO2). They were subsequently dried and examined by 14 N quadrupole reso-
nance. To our surprise, we were able to observe a signal at the right frequency
(4.64 MHz, the highest frequency NaNO2 line). This implies that the material has
been properly embedded but also that sodium nitrite was properly recrystallized.
This latter point was more than speculative. Anyway, we could observe that it
works for spruce, beech, ash, maritime pine, but not for oak (which is known to be
difficult to impregnate). Instrumental parameters have been optimized for reducing
J. Jover
Mthodologie RMN (CRM2; UMR 7036, UL-CNRS), Universit de Lorraine, Campus
Aiguillettes, B.P. 70239, 54506 Vanduvre-ls-Nancy (cedex), France
CRAN (UMR 7039, UL-CNRS), Universit de Lorraine, Campus Aiguillettes, B.P. 70239,
54506 Vanduvre-ls-Nancy (cedex), France
S. Aissani
Mesures et architectures lectroniques (IJL, UMR 7198, UL-CNRS), Universit de Lorraine,
Campus Aiguillettes, B.P. 70239, 54506 Vanduvre-ls-Nancy (cedex), France
L. Guendouz
Mesures et architectures lectroniques (IJL, UMR 7198, UL-CNRS), Universit de Lorraine,
Campus Aiguillettes, B.P. 70239, 54506 Vanduvre-ls-Nancy (cedex), France
A. Thomas
CRAN (UMR 7039, UL-CNRS), Universit de Lorraine, Campus Aiguillettes, B.P. 70239,
54506 Vanduvre-ls-Nancy (cedex), France
D. Canet ()
e-mail: daniel.canet@univ-lorraine.fr
78 J. Jover et al.
the duration of the experiment. If the first measurements required around 4 h, we

obtain at the present time acceptable results in about 2 min or less. In addition, we
have designed a simple NLLS (Non Linear Least Squares) algorithm by which the
spectral parameters of the NQR signal can be retrieved even if the peak is hardly
visible in the frequency domain. Extensive measurements have been performed on
NaNO2 recrystallized in spruce. They show, among other things, that the totality of
NaNO2 which has penetrated the material has actually recrystallized and seemingly
forever. Various assays have been carried out as a function of impregnation and
drying conditions. The evolution of the width of the NaNO2 line reflects defects in
the crystal lattice of NaNO2 due probably to alteration of the wood structure upon
drying.
7.1 Introduction
This work represents an attempt for detecting by NQR a molecule which has
been embedded in wood, the far-off objective being material labeling. In fact, our
objective is rather to demonstrate the feasibility of this approach. Moreover, because
we are familiar with 14 N quadrupole resonance, we have focused on this quadrupolar
nucleus in spite of its low sensitivity. Indeed, we are facing several challenges:
how to insert a substance within the material? If wood is impregnated, what is the
amount of the substance which has recrystallized (if any)? Will it be possible to use
NQR as a sensor? This paper is devoted to these questions. For simplicity, we have
chosen a molecule often used as a model in Nitrogen-14 Quadrupole Resonance,
namely sodium nitrite (NaNO2 ). It has the further advantage of being soluble in
water and can be considered as bio-compatible at certain doses (especially for food
preservation).
7.2 Experimental
7.2.1 Impregnation
Small cylindrical pieces of wood of 70 mm height and 8 mm diameter were

machined so that they could fit into 10 mm o.d. tubes (employed for NMR
measurements). They were impregnated with a saturated aqueous solution of sodium
nitrite with the help of a specially constructed device (Fig. 7.1). Impregnation can
be possibly performed under high pressure (up to 15 bar) [1], but also under partial
vacuum. We mean by partial vacuum soft pumping of the wood rod immersed
in the aqueous solution of sodium nitrite. This has the virtue of removing air
thus facilitating the impregnation process. In the case of impregnation under high
pressure, pumping is first applied to the wood rod (again for removing air) and
then the rod is immersed in the pressurized aqueous solution. Only raw material
7 NQR Detection of Sodium Nitrite Recrystallized in Wood 79
Fig. 7.1 Block diagram of the impregnation device
(without any preparation prior to impregnation) led to significant results (see below).
Wood samples were impregnated for 48 h. They were subsequently dried either in
the open-air (ambient temperature and normal moisture) or in a dry kiln at 103 C
(highest limit before wood degradation [2]) for a stabilization period (controlled by
weighing at regular intervals) which generally did not exceed 48 h.
7.2.2 Nitrogen-14 Quadrupole Resonance
NQR experiments were carried out at 4.64 MHz (the highest resonance frequency
of NaNO2 ) and at ambient temperature with a home-made spectrometer described
previously [3]. For the most recent measurements, a new probe was constructed:
still with a solenoidal coil surrounding the sample but with a cylindrical shield
at a distance of 8 cm from the solenoid center. This prevents electromagnetic
reflections thus parasitic capacitors which exist in the previous probes. Also, tuning
and matching of this probe involve equilibrated circuits. All measurements reported
here have been performed with a simple low power pulse of 100 s duration [4]
(corresponding to the first maximum of the nutation curve). A dead time (between
the end of the pulse and the beginning of data acquisition) of 300 s along with
low power pulses avoids baseline distortions. A repetition delay of 0.5 s was
80 J. Jover et al.
allowed between two consecutive acquisitions. Typically, with this new probe, 4,096
transients (scans) were accumulated but, as shown later, much less accumulation can
be envisioned.
7.3 Results
It was almost a surprise to observe a NQR signal in most impregnated wood

samples. This implies that, within this material, sodium nitrite has recrystallized
upon drying. Moreover, relatively sharp lines indicate few (if any) defects in the
lattice. Such experiments proved to be successful for almost all species that we have
tried. Spectra are reported in Fig. 7.2. The first obvious observation concerns the
ability of the different species to be impregnated by aqueous solutions of sodium
nitrite. This can be quantified by the peak area in either of the two series, thus
by the amount of recrystallized NaNO2 . This impregnation ability is in the order
spruce > Sylvester pine > beech > ash > oak (for the latter, we just observe the spike
Spruce
16384 scans. A = 46699 16384 scans. A = 58376

LW = 41310 Hz LW = 75030 Hz
Sylvester pine
16384 scans. A = 27813 16384 scans. A = 42160

LW = 44510 Hz LW = 49940 Hz
Fig. 7.2 4.64 MHz spectra of recrystallized NaNO2 in different wood species (impregnation under
partial vacuum) with the two drying modes described in the text. Left column: open-air drying; right
column: dry kiln. fftSexp stands for the Fourier transform of the experimental data. Continuous
curves (in red, denoted fftScalc) correspond to recalculated spectra (see below). A stands for the
peak area normalized to 4,096 scans whereas LW is the line-width at half-height
Beech
16384 scans. A = 21181 16384 scans. A = 34626

LW = 46540 Hz LW = 48540 Hz
Ash
65536 scans. A = 8534 65536 scans. A = 14202

LW = 90040 Hz LW = 67030 Hz
Oak
131072 scans
Fig. 7.2 (continued)
at zero frequency and it is indeed well known that oak is difficult to impregnate [5]).
Two other properties concern the drying mode: less NaNO2 is recrystallized in the
case open-air drying while line-widths are smaller than in the case dry kiln. The
interpretation of the first feature is immediate: upon heating at higher temperature,
more water is eliminated. The second feature can be explained by the fact that taking
the material at a relatively elevated temperature modifies its structure and this may
entail some defects in the crystal lattice of NaNO2 [6] (due to possible interactions
of NaNO2 with the material).
Note that these early experiments have been performed with the old probe. In
fact, the above trends have been confirmed by accurate measurements carried out
with the new probe on spruce samples. Relevant spectra are shown in Fig. 7.3.
82 J. Jover et al.
512 scans. A= 2319450. LW=48015 Hz
4096 scans. A = 456477 4096 scans. A = 461949

LW = 3924Hz LW = 67013Hz
Fig. 7.3 4.64 MHz spectra of NaNO2 obtained with the new probe. Top: reference spectrum of a
NaNO2 powder sample of the same size as the spruce samples (bottom). Impregnation was carried
out under partial vacuum. Left: open-air drying; right: dry kiln. fftSexp stands for the Fourier
transform of the experimental data. Continuous curves (in red, denoted fftScalc) correspond to
recalculated spectra (see below). A stands for the peak area normalized to 4,096 scans whereas LW
is the line-width at half height
It can be seen that the amount of recrystallized NaNO2 is hardly larger when the
sample has been subjected to the dry kiln. This was already recognized in Fig. 7.2
but seems to be specific of spruce (for other species, less NaNO2 is recrystallized
when drying in open-air). This amount, as determined by peak areas (the quantities
A), is five times larger in the powder than in wood. From the weight of NaNO2 in
the powder sample, we arrive at the absolute weight of NaNO2 in the spruce sample
(650 mg). It turns out that this value coincides almost perfectly with weighing data
(before impregnation and after drying). The next issue is evidently the line-width
of the recrystallized NaNO2 NQR signal. In the case of spruce, there is an order of
magnitude when going from crystallized powder to NaNO2 recrystallized in wood
(with open-air drying). As already mentioned, this can be reasonably attributed to
defects in the crystalline lattice which arise from interactions with the material.
These defects are still more important (line-width increases by a factor of two) when
drying is carried out in a kiln. As assumed above, this is very likely due to structural
modifications within the material and thus to modifications of the interactions with
NaNO2 .
In order to increase the amount of NaNO2 in wood, impregnation was achieved
under a pressure of 15 bars [7] (see Fig. 7.1). The spectra obtained after drying
are displayed in Fig. 7.4. They are quite different from those corresponding
4096 scans. A = 529148 4096 scans. A = 1134670

LW = 2574Hz LW = 156045Hz
Fig. 7.4 4.64 MHz spectra of NaNO2 in spruce. Impregnation was carried out under a pressure
of 15 bars. Left: open-air drying; right: drying in the kiln. fftSexp stands for the Fourier transform
of the experimental data. Continuous curves (in red, denoted fftScalc) correspond to recalculated
spectra (see below). A stands for the peak area normalized to 4,096 scans whereas LW is the line-
width at half height
to impregnation under partial vacuum. In the case of drying in open-air, we

observe a sharper line and roughly the same amount of recrystallized NaNO2 .
Conversely, when the sample has been dried in the kiln, the line-width increases
considerably and the amount of recrystallized NaNO2 is roughly twice that of
the sample impregnated under partial vacuum. On the other hand, weighing the
sample before impregnation and after drying in the kiln indicates that the quasi
totality of sodium nitrite corresponds to the A value of the right spectrum of
Fig. 7.4. This is twice the amount of sodium nitrite when the sample has dried
in open-air. Again, these observations can be interpreted by invoking the location
of NaNO2 and its interactions with wood. When the sample has dried in open-air,
only a part of water is taken off (probably the part near the surface as for samples
which have been impregnated under partial vacuum), leaving a limited quantity of
recrystallized NaNO2 . It is observed that the line-width is even weaker than in
the case of impregnation under partial vacuum. This means that the interactions
of recrystallized NaNO2 with the material are weakened due to the impregnation
mode. Conversely, as all water has been eliminated by the drying process in the
kiln (see above), and owing to the large line-width (right spectrum of Fig. 7.4),
we can infer that the spruce sample was impregnated to the core and that high
pressure has contributed to alter the material and, as a consequence, to enhance
the interactions undergone by sodium nitrite. Indeed, alteration of the material is
especially deleterious when, prior to impregnation, the piece of wood is dried in
the kiln. This procedure was attempted, again with the hope to increase the amount
of recrystallized sodium nitrite [8]. It turns out that, in these conditions, we totally
failed to observe any NQR signal (with impregnation under partial vacuum or under
high pressure). This could mean that water, which is naturally present in wood, is
necessary for keeping a structure prone to recrystallization of embedded sodium
nitrite.
Finally, we have checked that, when sodium nitrite recrystallizes in wood, its
NQR spectrum remains the same for months and probably longer.
84 J. Jover et al.
7.4 Data Analysis
All results, in terms of peak area (A) and line-width at half-height (LW), which have
been given until now, have been obtained through the procedure described below.
As seen above, obtaining a spectrum with a suitable signal-to-noise ratio (S/N)
requires the accumulation of a large number of transients. Let us recall that the
spectrum (frequency domain) results from the Fourier transform of accumulated
free induction decays (fids; time domain). As a consequence, when the NQR line
is relatively broad, the S/N is evidently lowered. Hence, a better strategy is to look
at the time domain data since the parameter of interest is the amplitude A of the fid
(which is also the peak area in the frequency domain) while the fid damping factor
is related to the line-width (LW) in the frequency domain. It is also important to
realize that, in our case, the spectrum involves a single line, this point facilitating
any fitting procedure.
In Fig. 7.5 are gathered a series of spectra (frequency domain) as a function of the
number of scans (or, equivalently, as a function of the measuring time). Of course,
with 4,096 scans (half an hour of measuring time) the peak is perfectly defined with
an excellent S/N. However, for future applications, this measuring time is evidently
too long. With 1 min measuring time, we can guess the presence of a peak but this
is no longer true for a measuring time of 30 s.
Now, we are going to show that a simple non-linear least squares (NLLS) fitting
procedure applied to the time domain data (fid) is able not only to reveal the presence
of a peak but also to provide reliably the parameters of interest, namely A and
LW. Such least-squares procedures are relatively common in multiline NMR spectra
[9, 10] and make use of the following mathematical model for the fid (in the case of
a single line and assuming that the fid is devoid of any dc component)
S.t / D A exp.2i
0 t / exp.i'/ exp.LW /t (7.1)
The parameters A and LW have been defined earlier;

0 is the resonance
frequency with respect to the carrier frequency and ' the phase factor. Alternatively
the damping function exp.LW /t can be replaced by a Gaussian function.
A graphical user interface has been created under MATLAB. It requires starting
values for these four parameters and provides refined values along with the
reconstructed spectrum in the frequency domain. The starting values can be derived
from the experimental spectrum if it is sufficiently legible. Otherwise, as the spectral
parameters of the considered NQR signal can be relatively easily guessed, they can
be used manually as input parameters. As shown in the right column of Fig. 7.5,
reconstructed spectra are perfectly consistent with spectra obtained from Fourier
transform of the experimental fids. It was even possible to detect, thanks to the fitting
of time domain data, a signal which is invisible in the frequency domain (bottom
spectra of Fig. 7.5). The A values are, as expected, approximately proportional to
the number of scans while LW values remain of the same order of magnitude.
4096 scans (half an hour). A = 390851. LW = 4254Hz
128 scans (1 minute). A = 19450. LW = 39212Hz
64 scans (30 seconds). A = 6483. LW = 26080Hz
Fig. 7.5 4.64 MHz spectra of NaNO2 in spruce (impregnation under partial vacuum; drying
out in open-air). Left column: raw spectra (fftSexp). Right column; the same raw spectra with
the reconstructed spectra (fftScalc, continuous curves in red) with parameters derived from time
domain data fitting
7.5 Conclusion
When this project started, it was considered as very challenging and rather utopian.
The first result was looked at with some mistrust because it required extensive time
averaging (several hours) and involved some artefacts. In fact, with some tenacity,
some instrumental improvements and with appropriate data processing, we were
able to obtain very consistent and reproducible results in a very reasonable mea-
suring time (lower limit: 30 s). Extensive studies of spruce impregnated by sodium
nitrite as a function of impregnation and drying conditions have shown that heating
always modify the wood structure, as reflected by the NaNO2 signal line-width.
86 J. Jover et al.
As demonstrated by Sauer and collaborators [11, 12], this line-width increase cannot
be attributed to homonuclear dipolar coupling but rather to electric field gradient
inhomogeneity. Thus, this latter feature is truly indicative of interactions between
the NaNO2 crystal lattice and the material. In particular, it was noticed that drying
wood at elevated temperature prior to impregnation totally precludes the formation
of NaNO2 crystals. Otherwise, it was shown that in soft conditions (without going to
elevated temperatures), a large amount of NaNO2 could be recrystallized in spruce
wood. This opens the way of material labelling by nitrogen containing compounds
and its further detection by nitrogen-14 Quadrupole Resonance.
Acknowledgements We are grateful to the referees for their constructive remarks and
suggestions.
References
1. Malkov S, Tikka P, Gullischen J (2001) Towards complete impregnation of wood chips with
aqueous solutions. Pap Timber 83:468
2. Norme NF B51-004, Bois Dtermination de lhumidit. AFNOR, Paris, 1985
3. Hiblot N, Cordier B, Ferrari M, Retournard A, Grandclaude D, Bedet J, Leclerc S, Canet D
(2008) A fully homemade 14 N quadrupole resonance spectrometer. C R Chimie 11:568
4. Guendouz L, Robert A, Retournard A, Leclerc S, Aissani S, Canet D (2012) Off-resonance
effects and selectivity profiles in pulsed nitrogen-14 nuclear quadrupole. Solid State Nucl Magn
Reson 4748:39
5. Humar M, Zlindra D, Pohleven F (2007) Influence of wood species, treatment method
and biocides concentration on leaching of copper-ethanolamine preservatives. Build Environ
42:578
6. Buess ML, Caulder SM (2004) Factor affecting the NQR line width in nitramine explosives.
Appl Magn Reson 25:383
7. Siau JF, Shaw JS (1971) The treatability of refractory softwoods. Wood Fiber 3:1
8. Matsumura J, Booker R, Ridoutt B, Donaldson N, Mikajiri N, Matsunaga H, Oda K (1999)
Impregnation of radiate pine wood by vacuum treatment: effect of pre-steaming on wood
structure and resin content. J Wood Sci 45:456
9. Barkhuijsen H, de Beer R, Bove WMMJ, van Ormondt D (1985) Retrieval of frequencies,
amplitudes, damping factors, and phases from time-domain signals using a linear least-squares
procedure. J Magn Reson 61:465
10. Montigny F, Brondeau J, Canet D (1990) Analysis of time domain NMR data by standard
non-linear least-squares. Chem Phys Lett 170:175
11. Malone MW, McGillvray M, Sauer KL (2011) Revealing dipolar coupling with NQR off-
resonant pulsed spin locking. Phys Rev B 84:214430
12. Malone MW, Sauer KL (2012) Homonuclear Dipolar Coupling and CPMG Spin-Echoes in
NQR. Appl Magn Reson 43:541
Part II
Nuclear Magnetic Resonance
Detection of Liquids
Chapter 8
Bottled Liquid Scanner for Security Checkpoints
Pablo J. Prado
Abstract Sealed bottles of any opacity are automatically inspected for hazardous
liquids in seconds using a novel Nuclear Magnetic Resonance method. Liquid
explosives and explosive precursors such as hydrogen peroxide are detected using
a multi-element Nuclear Magnetic Resonance protocol, solving the shortcomings
of optical techniques. The Bottled Liquid Scanner is capable of inspecting multiple
bottles in a single scan with unprecedented low false alarm rates, ensuring minimal
or no disruption at security checkpoints.
8.1 Introduction
There has been an increased threat of terrorist attacks involving plastic and liquid
explosives, including homemade or Improvised Explosive Devices. Because of this
increased threat, security concerns are now greater and extend to airports, seaports,
rail stations, prisons, embassies, and many other secured and unsecured facilities.
In the U.S. and Europe, the threat of airplane-bound explosives in particular
has highlighted the need for security systems that provide effective detection and
warning. These systems have to be able to detect a wider array of materials
and objects than ever before. A particular focus area is the screening of liquid
explosives and explosive precursors at airport and secured facility checkpoints. The
U.S. Transportation Security Administration and the European Commission jointly
put into place restrictions on traveling with bottled liquids to counter threats to
international civil aviation. These restrictions were intended to be temporary and
eventually to be replaced by technological solutions able to deliver a high level of
P.J. Prado ()

One Resonance Sensors, LLC, 2878 Camino del Rio South, Suite 115, San Diego,
CA 92104, USA
e-mail: pablo.prado@detect-ors.com
90 P.J. Prado
security. Then, again to allow passengers the convenience of carrying bottled liquids
of all types on board. The removal of existing restrictions will result in a need for
effective Bottled Liquid Scanner (BLS) technologies.
The Nuclear Magnetic Resonance (NMR) screening approach presented here
offers a BLS solution with high performance, while being cost effective for
wide use by airport and secured facility checkpoint operators all over the world.
NMR outperforms optical detection and dielectrometry in the range of containers
inspected, passenger throughput, and detection efficiency, which are the three most
important metrics in operational aviation security.
8.2 Bottled Liquid Explosive Detection
Liquid explosives may be manufactured from chemicals that are used for legitimate
purposes and thus legally obtainable and readily available. Explosives are typically
mixtures of an oxidizer, which supplies oxygen to a chemical reaction, and a fuel,
which supplies the element that reacts with oxygen.
In general, liquid explosives are chemically easier to combine and simpler to
ignite than solid explosives. Even though explosives in liquid form are highly
sensitive to shock, they may be transported in sealed containers by preparing
solutions that keep the compound more stable or by transporting the components
in separate containers.
Energetic materials -typically containing nitrogen- and peroxide-based com-
pounds are used to create homemade explosives. Hydrogen peroxide is of particular
interest for security checkpoints as it may be used as an explosive precursor. Contact
between hydrogen peroxide and combustible materials such as wood, paper, or oil,
causes spontaneous ignition or combustion. When mixed with materials such as
sugars, alcohols, or acetone, the result is a powerful explosive. Hydrogen peroxide
chemically resembles water, so it is a good candidate for a smuggled explosive
precursor.
Vapor and trace detectors may be used to screen containers for explosives, but
these require direct contact with the material inside the containers. In addition,
emulsion and slurry-based compounds such as ammonium nitrate have low vapor
pressure, making them more difficult to detect with vapor and trace detectors.
Optical detection technologies such as infrared [6] and Raman [3] spectroscopy
have been demonstrated to be effective for detecting some hazardous liquids, but
rely on illumination of the sample and thus are limited by the opacity of the
containers walls. Recently, Spatially Offset Raman Spectroscopy (SORS) [9] has
shown success with some level of opacity for the bottle wall, but still fails on a
series of plastics and on paper-wrapped bottles. These spectroscopic techniques rely
on matching the spectrum of the interrogated liquid to one found in the instrument
library of pre-selected hazardous substances.
NMR has been demonstrated to be an effective method of discriminating between
the expected parameters (e.g. T1 and T2 ) of a known liquid and an altered liquid in
8 Bottled Liquid Scanner for Security Checkpoints 91
an unopened container, that is, ensuring that the liquid specified on the bottle label
matches the bottle content [1, 7, 8, 10]. This method, however, is based on content
verification, which requires an extensive knowledge base of the NMR responses
for all compounds of interest. Hence, the implementation of such a method has
been abandoned because it is not practical as the knowledge database is extremely
extensive and constantly changing. Low field NMR has also been proposed to
discriminate hazardous and non-hazardous liquids [4, 13]. This approach offers the
potential to screen large bags, but suffers from low sensitivity and lack of sufficient
specificity.
Accordingly, a need exists for a method of detecting liquid explosives quickly,
efficiently, and inexpensively, with a high degree of accuracy and a low incidence
of false alarms. The method described below addresses the shortcoming of optical
techniques and previous approaches using NMR, providing a high performance and
rapid BLS solution for security checkpoints.
8.3 Multi-element NMR Screening Modality
The method described below uses safe, Magnetic Resonance Imaging (MRI)-
based technology. The detection protocol employs a unique combination of NMR
parameters that help identify the presence of liquid explosives or other hazards
in less than 5 s. These parameters are clearly differentiated from those of non-
hazardous liquids, providing exceptionally low false alarm levels.
The NMR approach works with bottles made of glass, plastic, and any other non-
metallic material as well as with paper and cardboard-wrapped bottles.
The detection works with all bottles that fit in the detection cavity, regardless
of their shape. The technique used addresses several of the weaknesses of current
bottle screening methods. Importantly, NMR is effective even with container walls
that are dark or opaque, which present a critical limitation for optical-based
modalities.
Parameters driving the alarm decision matrix include nitrogen and proton NMR
characteristic constants. The discrimination is based on parameters such as signal
amplitude, relaxation times, and the diffusion constant.
8.3.1 Liquid Explosives
Most liquid explosives are organic compounds with nitrogen-containing groups

such as NO2 , ONO2 , and NHNO2 . When detonated, these materials release
abundant energy as strong-bond gases are formed.
Effective detection of nitrogen-containing liquid explosives is achieved by tuning
the NMR instrument to the specific Larmor frequency of 14 N.
92 P.J. Prado
Fig. 8.1 Shows the detection 1.0

of an explosive precursor
(3 % hydrogen peroxide) by
0.8
proton NMR relaxation time
Magnitude
0.6 Water
0.4 Hydrogen Peroxide
0.2
0.0
0 100 200 300 400
Time (a.u.)
14
N is one of two stable (non-radioactive) isotopes of the chemical element
nitrogen and makes up approximately 99.63 % of natural nitrogen. The NMR
resonance frequency for nitrogen is 7.2 % that of hydrogen. 14 N is quadrupolar,
with nuclear spin I D 1, which results in NMR line broadening and an associated
fast signal decay. The broadening is moderate because the quadrupole coupling is
low, Q D 0.017 1028 m2 .
Multi RF pulse detection results in high detection sensitivity and therefore low
false alarms. A series of results are shown below.
8.3.2 Explosive Precursors
Hazardous liquids explosive precursors (e.g., hydrogen peroxide) are identified

by proton NMR by way of their distinctive NMR response. Identifying hydrogen
containing liquids is achieved without the need to have any information about the
liquids to be scanned.
Parameters used to identify hydrogen-containing explosive precursors include
the signal amplitude, spin-spin relaxation times, spinlattice relaxation times, and
the diffusion constant. A single parameter is not sufficient to identify explosive
precursors.
The decision matrix may contain two or three of these parameters, combined to
generate a discriminator. As an example, Fig. 8.1 shows the relative signal decay
during a Carr, Purcell, Meiboom, Gill (CPMG) pulse sequence [2, 11] of pharmacy-
grade hydrogen peroxide.
The diffusion constant (D) is another parameter used to interrogate liquids and
gels [5, 12]. In the presence of field gradients, the diffusion constant is calculated
using Eq. (8.1) below.
.1=T2eff / D 1=T2 C .1=3/ 2 G2 DTE 2 ; (8.1)

which may be defined as

1=T2eff D 1=T2 C 1=T2G I with T2G D 3= 2 G2 DTE 2 (8.2)
The inhomogeneous field contribution (1/T2G ) is determined by the static magnetic

field gradient (G), the diffusion coefficient of the inspected liquid (D), and the
gyromagnetic ratio (/2 D 4.3 kHz/G for protons), where TE is the inter-echo
duration during a CPMG pulse sequence. Therefore
D D 3=. 2 G2 / .1=T2effA 1=T2effB / =.TEA 2 TEB 2 /: (8.3)
Where the subscripts A and B indicate two measurements with different echo times.
The selection of the above parameters is performed with the goal of optimizing
the detection of hydrogen in the samples during the screening process. A hydrogen-
containing hazardous liquid is detected by evaluating its responses in the multi-
dimensional NMR parameter space.
8.4 Detection Workflow
The screening of bottles at checkpoints is performed by placing the bottle in the

NMR instrument and automatically running a multi-element detection protocol.
Figure 8.2 shows an example workflow for a dual-element NMR BLS. The initial
step is the sorting of metallic and non-metallic bottles. If the bottle to be inspected is
metallic, an alternative method should be used or an alarm is declared. For example,
metallic bottles may be screened by verifying the density of the contained liquid.
The non-metallic bottles are automatically screened by using a multi-element
protocol, e.g. nitrogen and proton NMR. The instrument scans at two or more
frequencies and interrogates the liquid content using a proprietary discrimination
algorithm. The process may be sequential (as shown in Fig. 8.2) or may be combined
to run both elements in an interleaved way.
8.5 Results and Performance
The effectiveness of the NMR BLS has been demonstrated by testing over 30
explosives and precursors at an independent laboratory, the Energetic Materials
Research and Testing Center (EMRTC) in Socorro, New Mexico. Performance with
high explosive materials was carried out using sub-milliliter samples in a compact
NMR device, using 5 mm diameter sample tubes. Tests with stable materials were
performed with a wide-bore NMR instrument using a range of 30 ml to 1 l bottles,
including wine, medicine, water, juice, and shampoo bottles.
94 P.J. Prado
Non Metallic Metallic

Bottle Sorting
Bottle Placement Complementary

in NMR Scanner Method
Element A Scan Option: high

sensitivity scan
Declaration:
Data Fail to Pass
Processing REPORT ALARM
Automated Scans
Element A
Declaration:
Pass
Element B Scan Option: high

sensitivity scan
Declaration:
Data Fail to Pass
Processing REPORT ALARM
Element B
Declaration:
Pass
REPORT CLEAR
Fig. 8.2 Example dual-element NMR workflow for hazardous material screening
Table 8.1 shows the characteristics of the inspected bottles. The test included
measurements with the original content and with Nitromethane and an aqueous
solution of Ammonium Nitrate.
Bottles in their original boxes and black plastic bottles present a limitation
for Raman or Infrared methods. These do not present issues for the NMR-based
inspection.
Figure 8.3 shows the sensitivity of the NMR method in detecting Nitromethane
and Ammonium Nitrate in several configurations.
An unambiguous Clear and Alarm declaration is achieved by selecting a
threshold amplitude above the Discriminator level for non-hazardous materials.
Table 8.1 Description of the

bottle types and wrapping Bottle type
used for the screening tests.
Cardboard box
Results shown in Fig. 8.3
Dark plastic
Colored glass
Clear plastic
Liquid
Concealed liquid explosives

Nitromethane
Ammonium nitrate aqueous solution
Original content
Soda - 7.5 oz
Shampoo, Tresseme 15 oz
Contact lens solution 5 oz
Cough syrup 415 oz
Mineral water 750 ml
Red wine 750 ml
80 Hazardous
70
Discriminator (au)
60
50
40
30
20 Non hazardous
10
0
Orange Juice - 20 fl oz
Gass bottle - 30 ml NM
Medicine Syrup - 8 fl oz
Orange bottle - 100 ml NM

Wine - Water 750 ml
Medicine Syrup - 100 ml NM

Orange Juice - Empty
Black plastic - Empty
Lens solution - Empty
Wine - 100 ml NM
Black plastic - 30 ml NM
Lens solution - 10 fl oz
Black plastic - 10 fl oz
Black plastic - AN
Lens solution - 100 ml NM

Medicine Syrup - Empty
Wine - Empty
Wine - Full 750 ml
Black plastic - 100 ml NM
Bottle Type and Content
Fig. 8.3 Liquid explosive detection performance. Two target materials are shown: Aqueous
solution of Ammonium Nitrate (AN) and Nitromethane (NM). Test with a wide-bore NMR probe
96 P.J. Prado
30 Hydrogen Peroxide Mixture
25
Discriminator (au) 20
15
10
Non hazardous
5
0
Shampoo
Creams
Rhum
Wine
Perfume
Hydrogen Peroxide
HP & 0.3g/ml coffee
HP & 0.02g/ml cumin

HP & 0.02g/ml nuts
HP & 0.33ml/ml ethanol
HP & 0.33ml/ml acetone
HP & sawdust
Water
Baby milk
HP & 0.4g/ml sugar

HP & 0.4g/ml OJ powder
HP & 0.02g/ml flour
HP & 0.02g/ml pepper
Liquid Type
Fig. 8.4 Detection of precursor materials. Discriminator amplitude for Hydrogen Peroxide (HP)
and non-hazardous targets. Measurements with 5 mm test tubes at EMRTC, Socorro, NM
Figure 8.4 shows the detection performance for precursor materials using proton
NMR. Mixtures of hydrogen peroxide with various additives were used for the
performance evaluation.
8.6 Checkpoint BLS
A compact benchtop instrument can be configured for security checkpoints. The

rapid detection protocol of the NMR scan results in minimal or no disruption to
passenger traffic. Unopened bottles are placed inside the detection cavity and then an
automated or single touch command triggers a scan. A Clear or Alarm declaration
is presented to the operator (see Fig. 8.2).
Figure 8.5 shows a block diagram of the NMR BLS apparatus.
Figure 8.6 represents an example configuration for the BLS instrument. All
bottles that fit in the cavity are screened.
8.7 Conclusions
The multi-element NMR technique provides an effective modality to inspect sealed

bottles of any opacity, including those wrapped in paper or inside their original box.
User Interface
Alarm/Clear
Data Processing
PC/Controller
Multi-element
RF Controller
Transmitter -
Signal Amplifier
Power Amplifier
Multi-tune Circuit
RF Switch
Multi-tune Antenna
or Separate Antennas
Electro Magnet or
Bottle Permanent Magnet
Fig. 8.5 Diagram of the Bottled Liquid Scanner instrument
1
2
4 1 Detection Cavity
2 Bottle Holder
3
3 User Interface
4 Back Panel
Fig. 8.6 Design for a checkpoint Bottled Liquid Scanner. 48 cm (1900 ) width, 35 cm (1400 ) depth
The method was tested with a series of liquid explosives and explosive precursors
demonstrating high detection performance and unprecedented low false alarm
levels. For these reasons, this NMR approach has clear advantages over optical
techniques and previously proposed NMR solutions.
98 P.J. Prado
Acknowledgments The author would like to thank Robert Lown for his valuable help designing
and building a compact magnet to screen full-sized bottles, to James Chepin and Nikolay Rusakov
for their critical contribution to the development of the NMR system and to the implementation
of measurement protocols, to Shouqin Huo for his help with data collection and analysis, and to
Sankaran Kumar and Lowell Burnett for valuable discussions and their continuous encouragement.
The author would also like to thank the organizing committee of the MRDE 2013 meeting in
Izmir, Turley, particularly to Tomaz Apih and Bulat Rameev.
References
1. Burnett LJ (1993) Liquid explosives detection. Proc SPIE 2092:208217

2. Carr HY, Purcell EM (1954) Effects of diffusion on free precession in nuclear magnetic
resonance experiments. Phys Rev 94:630638
3. Eliasson C, Macleod NA, Matousek P (2008) Non-invasive detection of powders concealed
within diffusely scattering plastic containers. Vib Spectrosc 48:811
4. Espy M, Baguisa S, Dunkerley D, Magnelind PE, Matlashov AN, Owens T, Sandin HH,
Savukov IM, Schultz LJ, Urbaitis AV, Volegov PL (2011) Progress on detection of liquid
explosives using ultra-low field MRI. IEEE Trans Appl Supercond 21:530533
5. Gudmundson E, Jakobsson A, Poplett I, Smith J (2009) Detection and classification of
liquid explosives using NMR. In: Proceedings of the 2009 IEEE international conference on
acoustics, speech, and signal processing, Taipei, pp 30533056
6. Itozaki H, Miyamura R, Sato-Akaba H (2012) Detection of bottled liquid explosives by near
infrared. Proc SPIE 8546:85460E
7. Kumar S, McMichael WC, Kim Y-W, Sheldon A, Magnuson EE, Ficke L, Chhoa TK-L,
Moeller CR, Barrall GA, Burnett LJ, Czipott PV, Pence JS, Skvoretz DC (1997) Screening
sealed bottles for liquid explosives. In: Proceedings of the SPIE conference on security systems
and nonlethal technologies for law enforcement, Boston, vol 2934, pp 126137
8. Kumar S, McMichael WC, Magnuson EE, Lee YK, Moeller CR, Czipott PV, Rayner TJ,
Newman DE, Wroblewski D (2001) Liquid contents verification for explosives, chemical
agents, and dissolved narcotics. In: Proceedings of the SPIE conference on enabling technolo-
gies for law enforcement and security, Boston, vol 4232, pp 206216
9. Loeffen PW, Maskall G, Bonthron S, Bloomfield M, Tombling C, Matousek P (2011) Spatially
Offset Raman Spectroscopy (SORS) for liquid screening. Proc SPIE 8018:80181E-1
10. Mauler J, Danieli E, Casanova F, Blumich B (2009) Identification of liquids encountered in
carry-on-luggage by mobile NMR. In: Fraissard J, Lapina O (eds) Proceedings of the NATO
advanced research workshop on explosives detection using magnetic and nuclear resonance
techniques, St. Petersburg, Russia, 79 July 2008, NATO science for peace and security series
B: physics and biophysics. Springer, Dordrecht
11. Meiboom S, Gill D (1958) Modified spin-echo method for measuring nuclear relaxation times.
Rev Sci Instrum 29:688691
12. Prado PJ, Mastikhin I, Karlsson MT (2012) Rapid method to screen unopened bottles to detect
concealed drugs. J Appl Nucl Magn Reson 43(4):531540
13. Sato-Akaba H, Itozaki H (2012) Development of the earths field NMR spectrometer for liquid
screening. Appl Magn Reson 43(4):579589
Chapter 9
MagViz: A Bottled Liquids Scanner Using
Ultra-Low Field NMR Relaxometry
Robert Austin, Michelle Espy, Andrei Matlashov, Henrik Sandin,

Larry Schultz, Algis Urbaitis, and Petr Volegov
Abstract Field Forensics, Inc. (FFI) has built a bottled liquids scanner utilizing
ultralow field NMR relaxometry. This device, called MagViz, is based upon
a prototype developed at the Los Alamos National Laboratory (LANL) (Espy
et al. Appl Supercond IEEE Trans 21(3):530, 2011; Espy et al. Supercond Sci
Technol 23:034023. doi:10.1088/0953-2048/23/3/034023, 2010) [1, 2]. Despite
using conventional Faraday detection coils in lieu of SQUIDs, MagViz, has
demonstrated sufficient sensitivity to identify a number of threat liquids of interest to
the Department of Homeland Security (Matlashov et al. Appl Supercond IEEE Trans
21(3):465468, 2011) [3]. By accurate measurement of T1 and T2 , liquids contained
in opaque bottles and even non-ferromagnetic metal containers can be reliably
identified. Protons are aligned using a 50 mT pre-polarizing field. T1 is determined
in the pre-polarizing field, and T2 relaxation time is typically measured at 2,048 Hz
in a 48 T field. The coil assembly is contained within a table-top 0.79 m tall
magnetically shielded enclosure. Although primarily intended for commercial and
security applications, MagViz, works at Larmor frequencies that correspond to
timescales that are characteristic of a host of interesting, slow, molecular dynamic
processes like diffusion and intramolecular motion as well as biological processes
such as protein folding, catalysis, and ligand binding and could conceivably serve
as a COTS research instrument for fundamental studies in these areas.
R. Austin ()
Field Forensics, Inc., Saint Petersburg, FL 33701, USA
e-mail: raustin@fieldforensics.com
M. Espy A. Matlashov H. Sandin L. Schultz A. Urbaitis P. Volegov
Los Alamos National Laboratory, Physics Division, Los Alamos, NM 87545, USA
100 R. Austin et al.
9.1 Introduction
The need to screen passenger carry-on items for concealed hazards has evolved in
response to rapidly changing threats. The most ubiquitous method in use remains
X-ray. However, this method in its simplest form is looking for changes in density
or shape (e.g. a gun or detonator wires hidden amongst clothes in a suitcase). To
detect explosives, which often can be disguised as items that appear benign to
this sort of X-ray, requires additional approaches. Ion mobility mass spectroscopy
in conjunction with swabbing is now routinely used as a secondary method to
determine the presence of residual material from bomb making. However, as
a thwarted terrorist plot involving liquid explosives indicated in 2006 [4], the
problem of reliably detecting liquid explosives in carry-on luggage (without opening
containers) is especially challenging. Most threat and benign liquids appear the same
to traditional X-ray approaches. At the time of this writing, owing to lack of another
screening technology, airport rules require that all of a passengers carry-on liquids
be restricted to 100 ml (3-oz) bottles and that all of the bottles be placed in a
single one-quart, clear bag. This is known in the United States as the 3-1-1 rule.
This approach limits the volume of liquids available to safeguard against the threat
of liquid explosives.
Recently very sophisticated X-ray systems have been able to distinguish liquids
from solids [5], and there is some evidence that this approach might be able to
identify at least some threat liquids. These methods rely on density and effective
atomic number. However, no approach for screening liquids inside carry-on bags
is presently proven. The technology focus at the time of this writing has largely
been towards deployment of secondary bottle screening methods that can enable
exceptions to the 100 ml limitations, for example to allow medicine or baby
formula.
Various techniques are presently being implemented for screening liquids, and
each presents advantages and draw-backs. Mass spectrometry is a gold standard
for chemical identification but has the drawback of requiring a physical sample,
usually obtained by swabbing. Sensitive vapor detection based on amplifying
fluorescent polymers [6] has also been demonstrated, but again, this requires at
least a small trace of the sample be available. Raman spectroscopy is a promising
approach that can measure through certain types of packaging. Raman relies on
measurement of the vibrational transitions in a sample through collection and
analysis of scattered photons after a sample has been excited by a laser [7].
The acquired spectrum is then compared to a threat library. Systems based on
Raman scattering have been demonstrated to be portable and suitable for some
liquid screening applications [8]. One potential source of confusion is background
fluorescence from packaging, or the material itself. One approach to mitigate the
effects of packaging is spatially offset Raman spectroscopy (SORS), which can
reduce these effects and has proven effective at colored glass and plastic opaque
packaging [9]. Electromagnetic methods based on microwaves [10] are also in
use. However, none of these methods appear capable of screening through the
9 MagViz: A Bottled Liquids Scanner Using Ultra-Low Field. . . 101
most challenging packaging (such as aluminized cans or bottles). It should also be

mentioned that methods using more advanced multi-spectral X-ray signatures from
liquids also continue to be advanced [11].
Methods based on nuclear magnetic resonance (NMR) have also been proposed
(see for example [12] and therein), primarily because NMR is excellent at distin-
guishing between liquids and gels based on chemical properties. An example of
this is clearly illustrated by magnetic resonance imaging (MRI), which is spatially
encoded NMR: MRI is an excellent medical diagnostic for differences in soft tissues
while X-ray based methods are not. Furthermore, because NMR can be spatially
encoded as MRI, there is potential to screen multiple bottles simultaneously. Classic
NMR spectroscopy based on chemical shift is challenging for a security screening
situation, because the magnetic fields generally have to be very high (>1 T) and
uniform (ppm) which is challenging to achieve in the presence of bottles that might
be of any material or shape. However, numerous NMR- based methods relying on
one or more MR relaxation parameters [1, 12], or NMR relaxation parameters with
diffusion [13] have proven effective at distinguishing explosive from benign liquids
in laboratory tests. Because of the complexity of the liquid detection problem,
NMR methods continue to be explored as a promising approach for bottled liquid
scanning.
In this paper we report on our efforts to implement NMR relaxometry and
dispersion (the change in relaxation parameters with magnetic field) at ultra- low
(T to mT) magnetic fields. This approach (like X-ray) is not a direct chemical
signature. But ULF NMR enables unique contrast for a wide variety of liquids, the
use of magnetic field dispersion, and the ability to detect through opaque or even thin
metal packaging, which is not possible with any of the optical or electromagnetic
methods presently available. Additionally, NMR-based methods hold the promise of
robust screening of multiple bottles via MRI (e.g. screening entire 3-1-1 baggies).
Here we present recent implementation of a single bottle screener, as well as results
obtained during R&D evaluation at the Transportation Security Laboratory.
9.2 ULF NMR Relaxometry
The NMR signal arises from the manipulation of spin polarized nuclei in an object.
Initially a magnetic field is applied along some direction to recruit some fraction of
the nuclear spins. In our discussion we refer to this as the pre-polarization step. The
equilibrium magnetization of the spins is given by
N 2 2 I.I C 1/Bp
Meq D ; (9.1)
3kB T
where kB D 1.3811023 J/K and D 1.0551034 J-sec, and is in the Rad/T-sec

form (267.513106 for the spin I D protons), Bp is the pre-polarization field, and
T is the temperature.
In reaching this equilibrium, the magnetization develops over time as

t

M.t / D Meq 1 e T1 ; (9.2)
where T1 is the spinlattice relaxation time, which describes how long it takes for
the magnetization to develop. In Eq. (9.2) T1 refers to the value in the Bp field.
The measurable NMR signal is derived from the precession of the magnetization
when spins are tipped such that there is a component of magnetization transverse to
the applied magnetic field. This process is described by the Bloch equations:

@M.t / M.t / Meq M.t / M.t /
D .M.t / Bm .t // zO C xO C yO ; (9.3)
@t T1 T2 T2
Here we adopt the convention that the spins are precessing about a measurement
magnetic field that is aligned along the z-axis. T2 is the spin-spin relaxation term.
The first term describes the precession about any transverse magnetization and the
second describes the effects of spin relaxation. We also note that while in Eq. 9.1
the field is the pre-polarization field, here B is the measurement field (oriented in
the z-direction). We also note that in Eq. (9.3) T1 and T2 refer to values in the Bm
field. The fields Bp and Bm may or may not be equal in strength (typically Bp is
10100 mT while Bm is 10100 T) and are likely not generated by the same
magnet.
The principle behind NMR is the resonance equation, which shows that the
Larmor (resonance) frequency of a spin is proportional to the magnetic field,
Bz , it is experiencing:
! D Bz ; (9.4)
where is the gyromagnetic ratio. In the case of MagViz, we focus on the hydrogen
in liquids, thus measuring the proton resonance. In MagViz, the polarization field,
which is used to achieve equilibrium magnetization (see Eq. 9.1), is distinct from
the measurement field that appears in the Bloch equations. The use of a larger Bp
field enables achieving a higher signal, while the use of a lower measurement field
enables the benefits of a low Larmor frequency, enabling the signal to penetrate
through metal.
The classic NMR technique to make a chemical determination is chemical
shift spectroscopy, which measures small changes in precession frequency due to
variations in the local magnetic field arising from the chemical environment. While
chemical shift is a direct effect of chemical structure, it is typically on the order of
a few ppm, and the associated frequency shifts scale with the measurement field. At
ULF these shifts cannot be resolved. Our approach to ULF spectroscopy relies on
measurement of the relaxation parameters, T1 and T2 , to provide information on the
local magnetic field surrounding the precessing nuclei and hence the physical and
chemical properties of the sample, an approach known as relaxometry.
Local magnetic field fluctuations, e.g., due to particle motion, matching the
spin Larmor frequency cause relaxation. The longitudinal or spinlattice relaxation,
T1 , involves redistributing the populations of the nuclear spin states to reach
the thermal equilibrium distribution by energy exchange with their surroundings.
The transverse or spinspin relaxation, T2 , corresponds to decoherence of the
transverse nuclear spin magnetization caused by random fluctuations of the local
magnetic field [14]. The T1 relaxation is generally strongly dependent on the
magnetic field strength. This is relevant for a ULF relaxometer approach where
the polarization and measurement fields are independent in strength and orientation,
and the measurement field can be easily varied to provide additional information
based on dispersion [15].
Relaxometry at low fields, >0.1 T is frequently used in industrial applications.
Some of these relaxometers also utilize different magnetic fields for polarization,
spin evolution, and measurement, a technique known as Field-Cycling NMR.
MagViz operates as a field-cycling relaxometer for chemical identification, oper-
ating in a regime of ultra low T magnetic fields where in the effect of changing
field strength on T1 is often the most dramatic. In the MagViz approach T1 is
determined in the Bp field, while T2 is determined in the measurement field. In
early implementations of MagViz, the superconducting quantum interference device
(SQUID) was used as the sensor of choice [1, 2], because of its unprecedented
sensitivity as a magnetic field detector at low frequencies. However more recently,
and in the version of MagViz we describe below, induction coils are used. We used
induction coils formed from a few thousand honeycomb-wound turns of AWG24
copper wire, 80 mm in diameter and 10 mm in depth. Further details on the
coil design process may be found in ref. [3]. Though the SNR of the coil system
was about 1/3 that of a SQUID based version, there is still sufficient signal to
discriminate materials of interest, and the complexity of requiring cryogens to cool
the SQUIDs is removed.
The MagViz instrument also has the advantages of very simple magnetic field
generation hardware, low magnetic fields, low Larmor frequencies (at which signals
can penetrate through metal cans, pipes, foil packaging), and dispersion as described
above.
9.3 The FFI Commercial Prototype
FFI has built a prototype commercial version of MagViz, based closely upon the
original LANL device. Modifications of the LANL design were few: primarily less
expensive material choices for coil support structures, simplified electronic coil
driving circuitry, and a much streamlined and simplified graphical user interface
(GUI).
Fig. 9.1 The image above shows three views of the commercial MagViz prototype. The leftmost
image shows the side and front, where the liquid sample bottle is inserted into the instrument.
The middle image shows the removable back panel that is attached by six latches, permitting easy
access to the coil support assembly. The rightmost picture shows the coil support assembly after
removing the back panel
9.3.1 Mechanical
Many of the parts for the coil support system of the original LANL experimental
unit were machined out of G10 glass-reinforced epoxy laminate. In the FFI device,
Delrin was used wherever convenient to reduce cost and weight. The mechanical
design of the LANL unit had many features to facilitate easy modification of coil
orientation and sample placement within the magnetic field. Since most of these pa-
rameters had been optimized by the time FFI began construction of its own version
of MagViz, many of these features were removed. Two, large pre-polarization coils,
and four Faraday detection coils were wound to LANL specifications by Panservic
of Scotts Valley, CA, USA [16]. Excitation, and measurement field coils (which
generate the static, 48 T field) were wound in house at the FFI facility.
The coils and signal preamplifiers are located within a magnetically shielded
housing, having walls consisting 12.7 mm thick aluminum honeycomb, sandwiched
between two layers of 1.27 mm thick mu-metal shielding. This magnetic shielding
attenuates stray external magnetic fields by a factor of approximately 16. The
aluminum honeycomb walls serve to shield the sensitive internal electronics from
external RF fields and also serve to dissipate eddy currents induced by the pre-
polarization coils that lie within the shielded housing (Fig. 9.1).
The pre-polarization coils draw approximately 15 amps of current, and therefore,
overheating has been a problem with MagViz. For example, a typical pulse sequence
might include successive polarizations from 0.5 s to as long as 8 s with subsequent
readout periods of 1 s. Thus the duty cycle of pre-polarization can be quite high.
To reduce this problem, we added aluminum ducts to channel air drawn in from
fans over the coils. This allows more frequent measurements without exceeding the
85 C temperature threshold on the pre-polarization coils.
Fig. 9.2 Raw MagViz signals, obtained from the four detection coils, showing the pulse sequence.
Please refer to the text for a description
The pulse sequence used by MagViz is illustrated in Fig. 9.2, above. Following
a 50 mT pre-polarization pulse, the 48 T measurement field turns on, and a
2 kHz 90 pulse starts the nuclei precessing. Analysis is performed on the spin-
echos that occur in the 100 ms long intervals between a series of 24 180 pulses.
The measurement field is switched off after the last echo is measured. This process
is performed repeatedly for the varying durations of pre-polarization pulses. The
decay of the spin-echo pulse amplitudes provides a measure of T2 and the sequence
of first echo amplitudes following the pre-polarization pulses provides a measure of
T1 (Fig. 9.3).
9.3.2 Electronics
The signals from the four Faraday detection coils are amplified 25,000 times by a
preamplifier provided by LANL. Data acquisition is performed by a SDAS (Syn-
chronization and Data Acquisition System) 6412 that was provided by Research
Electronics Development, Inc. [17]. The SDAS 6412 serves as an interface between
the control computer and the coil system in the relaxometer. Digital commands from
the computer are converted by the SDAS to analog signals, which are then further
Fig. 9.3 Plots showing how spin-echo amplitudes are used to measure T1 and T2 . The plot on
the left shows spin-echo amplitudes graphed against time measured from the end of the pre-
polarization pulse; the decay rate of spin-echo amplitudes give a measure of T2 . The plot on the
right shows first spin-echo amplitude measured after each pre-polarization pulse, plotted against
the duration of the pre-polarization pulse; an exponential fit to these points gives a measure of T1
Fig. 9.4 At left is a schematic of MagViz. Digital instructions from the computer are converted to
analog by the SDAS unit, amplified, and then sent to the Pre-polarization (Bp ), measurement field
(Bm ), and excitation coils (Be ). Detection coil signals converted to digital words by the SDAS and
sent to the computer for analysis. The coils within the coil assembly are labeled at right
amplified and sent to the pre-polarization, excitation, and measurement field coils.
The signals from the detection coils are converted via the SDASs 24-bit ADC to
digital words that are analyzed by the computer (Fig. 9.4). To avoid noise from
stray electromagnetic fields, optical cables provided communication between the PC
which controlled MagViz and the SDAS unit. An AE Techtron 7224 amplifier [18]
provides the high current pulse to the pre-polarization coils. Custom-built amplifiers
for the excitation and measurement field coils were provided by Analog-S, Inc. [19].
To suppress noise during the measurement interval, high current automotive relays
disconnect the pre-polarization coils from the Techtron amplifier, while much faster
reed relays disconnect the excitation coil from its amplifier.
Fig. 9.5 Plot showing the temperature obtained from Eq. 9.5 with input data from an IR and
thermocouple measurements of water inside MagViz vs. the known temperature obtained from a
thermocouple immersed in the water sample
Temperature T1 and T2 vary with temperature of the sample. An approximate

temperature of the sample is obtained using an Omega model OS136 infrared sensor
mounted approximately 20 cm above the sample. Since the temperature measured
by the infrared sensor is affected by the temperature of other parts of the instrument
within the sensors field of view, an emissivity correction based on the ambient
temperature measured by a thermocouple mounted next to the IR sensor is then
applied (Eq. 9.5).
s
4
4
Tsensor .1 "/Tbackground
4
Tt arg et D (9.5)
"
Here, Ttarget is the temperature of the liquid in the bottle. Tsensor is the temperature
measure by the IR sensor, is the emissivity of the bottle a value of 0.73 was
used, and Tbackground is the ambient temperature measured by a thermocouple located
adjacent to the IR sensor. Over the range of temperatures explored, this indirect
temperature measurement agrees pretty well with the temperature obtained by direct
immersion of a thermocouple in the liquid (i.e. the measured temperature never
deviates by more than 2 C from the known temperature of the liquid) (Fig. 9.5).
9.3.3 Software
A simple graphical user interface (GUI) was written in the LabView program-
ming environment. The analysis engine was written using MatLab. LabView and
Fig. 9.6 (a) MagViz can be controlled wirelessly via a GUI displayed on an IPAD. (b) The user
has the option of displaying more detailed information obtained from the analysis
Matlab generated executables obviate the need for MatLab and LabView licenses,
and they allow easy installation under the Windows operating system. The interface
is mirrored on an IPAD over a wireless network to allow remote control of the
system (Fig. 9.6).
For ease of use, the interface is stripped down to the barest essentials. A RUN
button on the upper left side of the screen initiates the test sequence. A READY in-
dicator on the upper right indicates whether the system is ready, running, analyzing,
or if a threat has been detected. Indicator lights on the lower right simply indicate
if the liquid is benign (Okay) or a threat. To initiate a measurement sequence, the
operator simply presses the RUN button. The measurements constituting a sequence
cannot be changed through the GUI, but they can be easily changed by editing a
setup file. The currently employed sequence proceeds as follows: a pre-polarizing
field is run for 0.5, 1, 2, 4, and 8 s for the T1 measurement. Each polarization pulse
is followed by a sequence of excitation pulses at the Larmor frequency. The decay
time of the spin-echo pulses during an excitation pulse sequence gives a measure
of T2 . A complete record of a run, indicating temperature, liquid identification,
temperature, T1 and T2 values, etc. is saved to a text file. Graphical displays,
providing information for the expert, can be displayed by setting the appropriate
parameter in a setup file (Fig. 9.6).
9.4 Recent Results
In August 2011 a version of MagViz was tested at the Transportation Security

Laboratory (TSL) in Atlantic City, NJ. The device was tested against some 500
benign items from streams of commerce in original packaging and 12 threat items.
The test protocol involved pre-polarization at 50 mT for a time tp , followed by
a multiple spin-echo sequence at 100 T in which T2 was determined. T1 was
determined by repeating the pulse sequence at tp D 0.5, 1, 2, 4, 8 s.
Fig. 9.7 R1 (1/T1 ) vs. R2 (1/T2 ) for 500 benign items from streams of commerce in original
packaging (black crosses). (a) shows data for eight threat materials (large spots) that cluster in
regions of the plot that are more easily distinguished from benigns. (b) shows data for four threat
materials (large spots) that cluster in regions of the plot that would produce a high false-positive
As shown in Fig. 9.7a, for eight of these threats discrimination from benign
materials based on relaxation rates, R1 (1/T1 ) and R2 (1/T2 ), is possible with a very
low false positive rate. However, as Fig. 9.7b shows, for four of those materials it
would be very difficult to distinguish them from benign items.
9.5 Conclusion and Future Directions
Based on the preliminary data acquired at TSL, we believe that for many threats
NMR relaxation alone is adequate for robust discrimination. However there are
other threat liquids where there may be the need for an orthogonal parameter. One
such candidate might be X-ray attenuation. In addition, we have not fully exploited
the information enabled by NMR alone, such as making full use of the amplitude
of the NMR signal (which indicates the proton content of a liquid). A full data
taking/analysis sequence takes about 1 min and 10 s, largely due to the long times
required to obtain T1 . This can also be reduced by using an orthogonal parameter
which would eliminate the need to make multiple measurements for different pre-
polarization times.
Acknowledgements The work presented here accomplished under the support of Department of
Homeland Security, Science and Technology Directorate under agreement HSHQPM12X00166.
The authors would also like to personally thank Stephen Surko for his guidance at DHS, Valerie
Lively and Paul Ruwaldt at TSL for helping with data collection, the support of LANL colleagues
John Gomez, Shaun Newman, Mark Peters, Robert Sedillo, our FFI colleagues Al Guim, Mark
Tesone, and special thanks to Lloyd Bastian for his expert electronics help.
LA-UR-12-24380Approved for public release; distribution is unlimited.
References
1. Espy M, Baguisa S, Dunkerley D, Magnelind P, Matlashov A, Owens T, Sandin H, Savukov

I, Schultz L, Urbaitis A, Volegov P (2011) Progress on detection of liquid explosives using
ultra-low field MRI. Appl Supercond IEEE Trans 21(3):530
2. Espy M, Flynn M, Gomez J, Hanson C, Kraus R, Magnelind P, Maskaly K, Matlashov A,
Newman S, Owens T, Peters M, Sandin H, Savukov I, Schultz L, Urbaitis A, Volegov P, Zotev
V (2010) Ultra-low-field MRI for the detection of liquid explosives. Supercond Sci Technol
23:034023. doi:10.1088/0953-2048/23/3/034023
3. Matlashov A, Schultz L, Espy M, Kraus R, Savukov I et al (2011) SQUIDs vs. induction coils
for ultra-low field nuclear magnetic resonance: experimental and simulation comparison. Appl
Supercond IEEE Trans 21(3):465468
4. 2006 transatlantic aircraft plot. (n.d.) http://en.wikipedia.org/wiki/2006_transatlantic_aircraft_
plot, Accessed January 2013
5. Smiths HI-SCAN 6040aTiX. 64 Clarendon Road, Watford, Herts WD17 1DA, UK
6. Thomas S, Joly G, Swager T (2007) Chemical sensors based on amplifying fluorescent
conjugated polymers. Chem Rev 107:13391386
7. Caygill J et al (2012) Current trends in explosive detection techniques. Talanta 88:1429
8. Moore D, Scharff R (2009) Portable Raman explosives detection. Anal Bioanal Chem
393:15711578
9. Hargreaves M, Matousek P (2009) Threat detection of liquid explosive precursor mixtures
by Spatially Offset Raman Spectroscopy (SORS). In: Lewis C (ed) Optics and photonics
for counterterrorism and crime fighting V, vol 7486, Proceedings of the SPIE. SPIE, Berlin,
p 74860B
10. CEIA USA Ltd (n.d.) EMa series bottle and liquid scanner. http://www.ceia.net/security/pdf/
EMAbrochureUS.pdf, Accessed January 2013
11. Joyce D, Gibson G, Radley I, Senior M Method for the identification of materials in a container.
http://www.freepatentsonline.com/y2012/0116691.html, Accessed January 2013
12. Fraissard J, Lapina O (eds) (2009) Explosives detection using magnetic and nuclear resonance
techniques, NATO science for peace and security series B: physics and biophysics. Springer,
Dordrecht
13. Gudmundson E, Jakobsson A, Poplett I, Smith J (2009) In: IEEE International Conference on
Acoustics, Speech and Signal Processing, Taipei, 1924 April 2009
14. Kimmic R, Anoardo E (2004) Field-cycling NMR relaxometry. Prog Nucl Magn Reson
Spectrosc 44:257320
15. Bene G (1980) Nuclear magnetism of liquid systems in the earth field-range. Phys Rep
58:213267
16. Panservic, 847 Pinecone Drive, Scotts Valley, CA 95066
17. Research Electronics Development, Inc. 20 Maple Avenue, East Setauket, NY 11733, USA
18. AE Techron, Inc., 2507 Warren St., Elkhart, IN 46516 USA
19. Analog-S, 608 Roxboro Rd, Oxford, NC 26565
Chapter 10
Multiparameter NMR Identification
of Liquid Substances
Andrey B. Konov, Kev M. Salikhov, Evgeniya L. Vavilova,

and Bulat Z. Rameev
Abstract In this paper we study the method of distinguishing the substances

by measuring their nuclear paramagnetic longitudinal and transverse relaxation
times and the diffusion coefficient of molecules. Experiments performed using a
commercial high magnetic field NMR spectrometer show the possibility to use
this method for reliable identification of liquids. Observables in these experiments
rather often cannot be described by a single exponential function. In the article we
discuss how to utilize the non-single exponential experimental dependences for a
quantitative processing of the NMR experimental results.
A.B. Konov ()

Laboratory of Spin Physics and Spin Chemistry, Zavoisky Physical -Technical Institute,
Kazan Scientific Center of RAS, Sibirsky tract, 10/7, 420029 Kazan, Russia
K.M. Salikhov
Laboratory of Quantum Dynamics, Zavoisky Physical-Technical Institute, Kazan Scientific
Center of RAS, Sibirsky tract, 10/7, 420029 Kazan, Russia
e-mail: salikhov@kfti.knc.ru
E.L. Vavilova
Laboratory of Physics of Perspective Materials, Zavoisky Physical-Technical Institute,
Kazan Scientific Center of RAS, Sibirsky tract, 10/7, 420029 Kazan, Russia
e-mail: jenia.vavilova@gmail.com
B.Z. Rameev
Laboratory of Radiation Physics, Zavoisky Physical-Technical Institute, Kazan Scientific Center
of RAS, Sibirsky tract, 10/7, 420029 Kazan, Russia
Gebze Institute of Technology, 41400 Gebze-Kocaeli, Turkey
e-mail: rameev@gyte.edu.tr
112 A.B. Konov et al.
10.1 Introduction
Identification of illicit liquids in security check-points and the creation of reliable

and fast detectors for liquids are very actual problems. This is due to the need to
ensure security in public places and transportation. Starting from 2006 a very strict
regulations on liquids allowed on board of aircraft have been introduced in airports.
This causes a considerable inconvenience for airliners passengers. Moreover there
are cases where there is a vital need to take liquids with a passenger on the plane
board. For example, in the case of some medicines, baby food for infants and others.
The method of nuclear magnetic resonance (NMR) can be successfully applied for
the liquid substances identification.
It is well known that the NMR method is one of the most powerful tools
for the study of micro-and nano-scale structure and dynamics of both the liquid
and the solid states. With the help of NMR it is possible to study the structure
even of large and complex biological molecules. For this kind of experiments one
needs the expensive high-resolution spectrometers operating at high magnetic fields.
Unfortunately, it seems unpractical to use huge high magnetic field high resolution
NMR spectrometers for rapid screening of substances carried by passengers on
airports, etc. For the last purpose the strength of the magnetic field has to be
relatively small. In these fields the NMR spectral lines of different substances
overlap and the spectra contain minor information which can be used to discriminate
substances. But fortunately, the spectral lines (their positions) are not the exclusive
parameters of different molecules. There are also other magnetic resonance and
molecular kinetics parameters which can help to discriminate substances: molecular
diffusion coefficient, D, longitudinal relaxation time T1 of magnetic nuclei and
their transverse relaxation time T2 . Note, that in organic substances the major
magnetic nuclei are protons (nuclei of hydrogen atoms). It is well known that all
these parameters, T1 ,T2 , D, can be measured using methodology of pulse NMR
spectroscopy and pulse magnetic field gradients [1]. In contrast to the position of
the NMR lines, the molecular parameters D, T1 and T2 can differ considerably
also at the low magnetic fields and, as a consequence, these parameters can be
exploited for identification of substances in airports and etc. This approach to
identification of liquids was proposed by L. Burnett et al [2] and S. Kumar
et al [3]. They suggested to measure the NMR relaxation times T1 and T2 . These
works were followed by publications [4, 68]. It has been stressed in the work
[4] that using of only the T1 and T2 relaxation times does not provide a reliable
identification when several liquids are to be discriminated, so that there is a need
to use additional parameters to discriminate substances. Some possible parameters,
which have already been proposed to use as additional ones, are the integral intensity
of NMR signals [5], relaxation parameters of other than protons magnetic nuclei
(e.g. 14 N) [6] or dielectric characteristics of the substances [7]. Recently, time-
domain NMR approach involving the measurements of the self-diffusion coefficient
of the molecules as additional parameter has also been proposed [8]. It is worth to
10 Multiparameter NMR Identification of Liquid Substances 113
note that the molecular diffusion coefficient is expected to be independent on the

magnetic field strength while the paramagnetic relaxation times are depending on a
magnetic field strength.
In our work we study potentials of discrimination of different substances in
liquids by measuring three parameters T1 , T2 , D. We suggest visualizing of data
as the 3-dimensional picture-3D presentation.
In this paper, we present the results of our experiments for NMR identification
of liquids by measuring longitudinal as well as transverse relaxation times, and
molecular diffusion coefficients in liquids molecules. We applied a commercial
high-resolution NMR spectrometer to obtain the high magnetic field relaxation
parameters T1 and T2 and diffusion constant D of various liquids. The multi-
exponential dependence of experimental parameters and its practical implication
for development of NMR identification of liquids has been discussed.
10.2 Experimental Results
Measurements of parameters fT1 , T2 , Dg were performed on high-resolution

spectrometer Bruker Avance 400 with BRUKERs extensively used multinuclear
tuneable broad band NMR probe (BBO probe), equipped with magnetic field
gradient; the gradient magnitude equals to 53.5 G/cm. The experiments were
performed on nuclei 1 H. The resonance frequency for protons is 400 MHz. This
spectrometer uses a very high magnetic field (9.4 Tesla) with high homogeneity,
produced by the superconducting magnet. The molecular diffusion measurements
were performed using the pulse sequence based on the stimulated echo with bipolar
pulsed magnetic field gradients with sine pulse shape. In most experiments, the
diffusion time was 50 ms. Duration of the gradient pulses, depending on the sample,
was varied from a few milliseconds to tens of milliseconds. In each separate
experiment the magnetic field gradient strength varied from minimum to maximum.
The NMR signal amplitude decreases with increasing the gradient strength. The rate
of decrease of the NMR signal depends on the diffusion coefficient, and the signal
decreases faster when the molecular diffusion coefficient increases.
For diffusion measurements we used stimulated echo pulse sequence with two
gradients pulses: RF90 gp RF90 tau RF90 gp echo, where RF90 is 90
radio frequency pulse, gp is the gradient pulse, tau is the diffusion time. According
to the theory in experiments of measuring the molecular diffusion coefficient D of
the molecules in a liquid containing only one type of molecule, using stimulated
echo pulse sequence with two magnetic field gradient pulses, the NMR signal is
given as follows [9]:

I D I0 exp c D g 2 ; (10.1)
where I0 and D is the intensity of the NMR signal and the diffusion coefficient of the
molecule, g is the gradient strength of the magnetic field, c is a constant parameter
for each experiment, it depends on the settings of the pulse sequence (gradient pulse
duration, diffusion time).
Longitudinal relaxation time T1 was measured using a pulse sequence inversion-
recovery. In each experiment, the time between 180 and 90 pulse was varied.
In this experiment, for the liquid containing the same type of relaxing molecules,
dependence of the signal on the time is given by:

t
I D I0 C P exp ; (10.2)
T1
where I0 , P is the intensities of the NMR signal (usually P D 2 I0 , but P also is a

complicated function of diffusion coefficient and also depends from radiofrequency
pulse imperfection; that is why usually P is treated as an additional variable) [8] and
T1 is the longitudinal relaxation time of the nuclear magnetic moments of the nuclei.
Transverse relaxation time T2 was measured using a Carr-Purcell-Meiboom-Gill
(CPMG) pulse sequence. The delay time t between pulses (between 90 and 180
pulses) was 1 ms. In these experiments the observable depends from time gap t
according to the formula [10]:

t
I D I0 exp ; (10.3)
T2
where I0 and T2 are the intensity of the first echo signal and the transverse relaxation
time, respectively.
The equipment used in our experiments allows us to obtain high-resolution
spectra. That is the parameters fT1 , T2 , Dg may be determined separately for each
resonance line in the spectrum. In the experiments we excite non-selectively all
protons so that we detect a total contribution of all molecules to observables.
The following set of liquids have been used as the test substances: ethanol,
water, an aqueous solution of hydrogen peroxide, isopropanol, Coca Cola, shaving
foam Nivea for men: sensitive skin, shampoo for head Head & Shoulders: ocean
energy, toothpaste blend-a-med 3D white fresh: cool fresh.
The results obtained demonstrate that for some liquids the parameters fT1 , T2 ,
Dg can be obtained from the experimental data by fitting the experimental results
using the formulas given above. But for other liquids the decay kinetics of the
observed signals due to T1 , T2 relaxation and molecular diffusion processes is not
so simple, a real life appears to be more complex. This fact can be considered
both as a disadvantage and an advantage. Disadvantage arises because of more
complicated processing procedure for not simple exponential dependencies, while
advantage appears because the nonexponential behaviour can be considered as one
more degree of freedom, which can be utilized to distinguish liquids from each
other.
Fig. 10.1 NMR signal decay

with increasing time in the
pulse sequence of CPMG, the
1
NMR signal is from all the
Integral Intensity
protons contained in the
shaving cream Nivea for
Men; straight line is the
approximation using single
exponential decay
h function
i
I D I0 Exp Tt2 with the
least squares method
0,1
0,00 0,01 0,02 0,03 0,04 0,05 0,06 0,07
t [s]
Many of the liquids are mixtures of several substances. Various molecules of

the liquid, generally speaking, have their own set of parameters fT1 , T2 , Dg. And
the values of these parameters also depend from type and concentration of other
substances. Furthermore, the relaxation of the protons of different functional groups
of even one molecule can occur with different characteristic relaxation times. The
presence of several relaxation times and molecular diffusion coefficients leads to
the fact that the amplitude of the observed signals is to be described by the sum of
exponential functions
Xh i
I D .I0 /i exp c Di g 2 ; (10.4)
i
X
t

I D .I0 /i C Pi exp ; (10.5)
i
.T1 /i
X
t

I D .I0 /i exp (10.6)
i
.T2 /i
The experimental data presented on Fig. 10.1 demonstrate these features of

observables.
For these experimental data, in general, it is not clear how to measure the
relaxation time or molecular diffusion coefficient. Below are shortly considered
some possible strategies.
An evident suggestion is the first moment (weighted average value) approach: a
single exponential approximation for the initial part of experimental data that are
the sum of exponents. In this case the derived parameter is a statistical average
relaxation rate or a statistical average molecular diffusion coefficient. The logic
of this method is roughly as follows. Sum of exponentials for small values of
the argument (time t for relaxation or gradient strength g for diffusion coefficient
measurements) is a linear function with the averaged relaxation rate and molecular
diffusion coefficient:
X X
I.x/ D fIi exp ki xg D fIi Ii ki x C : : :g D
i i
X X X
D fIi g fIi ki xg C f: : :g D
i i i
X X X
D fIi g x fIi ki g C f: : :g D
i i i
! 0 P P 1
X fIi ki g f: : :g
fIi g @1 P A
i i
D xC P
i
fIi g fIi g
i i
! 0 P 1
X fIi ki g
@ i
fIi g exp P xA (10.7)
i
fIi g
i
Here ki is the inverse relaxation time or molecular diffusion coefficient for the
i-th component, Ii is the absolute intensity of the NMR observed signal for given
component, x is a varied parameter: the time interval (for a relaxation experiment)
or the square of the amplitude of the magnetic field gradient (for diffusion
experiment). P
fIi ki g
The first moment k D iP
fIi g
of a distribution of k values is fully determined
i
by the first derivative of the corresponding curve I.x/ at x D 0, i.e., I 0 .0/ (see
Eq. 10.7). However, this method has the disadvantage. It can give the same effective
parameter for mixtures of different composition.
For example, consider the case where there are two types of molecules with
D1 and D2 diffusion coefficients, and I1 and I2 are absolute intensities of the
NMR signal for corresponding components. Then the effective molecular diffusion
coefficient of the molecules is given by the expression:
D D I1 DI11 CI2 D2
CI2
. For example, in case of I1 D 12 I0 , I2 D 12 I0 ,
D1 D 2 D, D2 D 32 D, we can get
1
1
Deff D 12 I0 12 D C 12 I0 32 D 2
I0 C 12 I0 D D. If we have another
intensities and diffusion coefficients I1 D 13 I0 , I2 D 23 I0 , D1 D 2 D,
D2 D 32 D, we can get the same effective diffusion coefficient:

1 2 1 1 2
Deff D I0 2D C I0 D I0 C I0 D D:
3 3 2 3 3
The approach outlined gives the first moment of the distribution either of the
paramagnetic relaxation times or of molecular diffusion coefficients. Of course,
these first order moments do not allow to fully determine the distribution function
of those parameters and to discriminate mixtures of substances. For that one needs
more information. But the known first moments can serve as useful constraints for
testing of liquid composition and possible values of parameters which characterize
the constituents of a liquid.
In the experiments under consideration the observed signals decay when the
time interval increases (see, e.g., Eqs. 10.2 and 10.3) or the magnetic field gradient
value (see, e.g., Eq. 10.1). When this decay is described by a simple exponential
function exp .a x/ it is straightforward to introduce a characteristic parameter a1
of the decay. When experimentally decay I(x) cannot be described by the simple
exponent one can introduce an overall averaged parameter which in some cumulate
way characterized the decay of observable. In the case of the NMR relaxation
experiments the overall averaged relaxation time is determined as
Z1
1
.Tk /av: D I.t /dt; (10.8)
I0
0
where k D 1 or 2 depending on experiment. In the case of the pulse gradient NMR

experiments for measuring the diffusion coefficient, the overall averaged molecular
diffusion coefficient is determined as (compare with Eq. 10.1)
2 31
Z1
2 2
.D/av: D I0 4 I cg d cg 5 : (10.9)
0
These overall averaged parameters do not allow identifying constituents of

liquids. But they can serve as useful constraints for testing of liquid composition
and possible values of parameters which characterize the constituents of a liquid.
There is one rather simple realization of the determining the overall averaged pa-
rameter. This can be found from a slope of a straight line attempting to approximate
the real experimental data. An example of an approximation of experimental data
by a single exponential at the entire range of time intervals between the pulses in the
pulse sequence for the measurement of the relaxation times and the entire range of
the square of the gradient of the magnetic field in the measurement of the coefficient
of molecular diffusion coefficient is shown on Fig. 10.1. The results obtained using
this approximation are shown in Table 10.1 and in Fig. 10.2.
Approaches outlined above cannot lead to the definite identification of con-
stituents of mixtures. In that respect is more promising to present experimentally
detected decay curves as a sum of exponents. This strategy to use multiexponential
probe functions to describe experimental data arises from the fact in a mixture each
constituent molecules have their characteristic nuclear magnetic relaxation times
(rates) and molecular diffusion coefficients. In the multiexponential presentation of
experimental data a number of exponents is dictated by a number of constituent
molecules as well as by the number of magnetically non-equivalent protons which
Table 10.1 Relaxation times T1 , T2 and the SDC for the investigated
compounds
Diffusion coefficient,
T1 , [s] [109 * m2 /s] T2 , [s]
Ethanol 2.29 0.925 0.3
Water 1.25 2 1.02
Water solution of Peroxide 3.27 2.28 0.01
Isopropanol 3.28 0.57 0.47
Coca Cola 1.93 1.76 0.637
Nivea shaving cream 1.97 0.078 0.037
Head&Shoulders shampoo 1.83 1.27 0.09
blend-a-med toothpaste 0.391 0.72 0.077
Water Coca Cola

Isopropanol
Ethanol
1,0
0,8
blend-a-med
toothpaste
0,6
T2 [s]
2,5 0,4
D [10-9 m /s]
2,0
2
1,5 0,2
1,0
0,5
0,0 0,0
3,5
1,0 1,5 2,0 2,5 3,0
0,0 0,5
Head&Shoulders T1 [s] Peroxide
shampoo Nivea shaving cream
Fig. 10.2 Set of data fT1 , D, T2 g for the investigated compounds. Measurement error does not
exceed 10 %
have different paramagnetic relaxation times. Sometimes the NMR observables

decay is not given by a simple exponent even for liquids containing only one kind
of molecules. For example, in the case of restricted diffusion the NMR observables
demonstrate non-exponential decay [11]. Bringing number of exponents to infinity,
we obtain the inverse Laplace transform: decomposition of the initial experimental
data in a series in which the basis functions are the exponents. Recently, Laplace
transform started to be very popular algorithm used to describe the NMR decay in
heterogeneous systems (see e.g. [1214]). However, the inverse Laplace transform
(ILT) is ill-posed problem due to the fact that the basis set, consisting of the
exponents, isnt an orthogonal basis. For this reason, the expansion using exponents
is not unique: the transform can produce several different results for the expansion.
Especially the effect of noise in ILT is very important issue. For that, various
regularization (e.g. Tikhonov regularization) approaches are used to stabilize ILT.
The application of ILT to the NMR identification is quite challenging issue and we
plan to study in details various modifications of ILT for the NMR identification of
molecules.
All samples under investigation showed one exponential dynamic of NMR signal
time dependence in T1 inversion-recovery experiments according to formula
(10.2). In diffusion measurements pure liquids (ethanol, isopropanol, peroxide,
water, cola) showed one-exponential dynamic while mixtures (toothpaste, shaving
cream and shampoo) exhibited non-exponential dynamic of the signal. Almost all
samples showed exponential signal decay in CPMG T2 -measurements. Isopropanol
and ethanol samples showed the most non-exponential dynamics in T2 -measuring
experiments.
The experimental results show that the use of three parameters allows identifying
constituents of a liquid. Some authors proposed the use of two parameters to
distinguish between substances: using relaxation times fT1 , T2 g [15] and using
the transverse relaxation time and the molecular diffusion coefficient fT2 , Dg
[16]. However, the use of only two parameters makes the method less reliable.
This is evident in Fig. 10.3. It shows that in these projections the values of
parameters fT1 , T2 , Dg for some substances have similar overlapping values For
example, Coca Cola and shaving cream have similar values of the longitudinal
relaxation time T1 . However, the substances can be distinguished using the other two
parameters (T2 and D) (Fig. 10.3c). Another example is Coca Cola and shampoo.
Parameters T1 and D for Coca Cola and shampoo have close values (Fig. 10.3b).
For Coca Cola we have T1 D 1.93 s, D D 1.76 * 109 m2 /s, while for shampoo
T1 D 1.83 s, D D 1.27 * 109 m2 /s. However, the transverse relaxation time T2
differs significantly for these two liquids (Fig. 10.3a). It does remove the degeneracy
(the ambiguity in identifying the substance): for Coca Cola T2 D 0.64 s and for
shampoo T2 D 0.09 s, that is the transverse relaxation time differs significantly from
each other. Thus, using of the third parameter helps one to identify uniquely the
substances in a mixture.
10.3 Outlook
Our main goal is to develop an efficient NMR scanner to identify substances,

identify liquids. In real prototype of the NMR scanner for identification of liquids
we plan to use relatively low magnetic fields when observables are a sum of
contributions from different substances. As a consequence, a behaviour of ob-
servables can be given not by a single exponential dependences but by complex
functions. Under these conditions it is not trivial task to get out an information
about individual components, individual liquids. One needs to develop algorithms
to analyze experimental data when there are mixtures of liquids or when there are
several vessels (bottles) with different liquid substances.
a
Water
1 Coca Cola
Isopropanol
Ethanol
blend-a-med
T2 [s]
0,1 toothpaste Head&Shoulders

shampoo
Nivea - shaving cream

Peroxide
0,01
1
T1 [s]
b
Peroxide
Water Coca Cola
Head&Shoulders - shampoo
1
D [10-9 m2/s]
Ethanol Isopropanol
blend-a-med
toothpaste
0,1 Nivea - shaving cream
1
T1 [s]
c
Peroxide
Coca Cola Water
Head&Shoulders - shampoo
1 Ethanol
D [10-9 m2/s]
blend-a-med toothpaste Isopropanol
0,1
Nivea - shaving creame
0,01 0,1 1
T2 [s]
Fig. 10.3 Two-dimensional projections of three-dimensional data fT1 , T2 , Dg: (a) T2 T1 ; (b) D
T1 ; (c) DT2
The problems of identification of substances on a base of paramagnetic relaxation

times and molecular diffusion coefficient exist irrespective of the magnetic field
strength. Here we have used high magnetic field NMR unit and investigated a
possibility of using the NMR relaxometry and diffusiometry to distinguish liquids.
For a set of substances, which contain both benign liquids and chemicals illicit to
transportation, we measured paramagnetic relaxation times and molecular diffusion
coefficient. The experimental results showed that the method can be successfully
applied in order to distinguish one kind of liquid from another. At the same time,
we show that it is non-trivial task to choose a routine that provides the most robust
and effective processing of NMR data with aid of discrimination between various
heterogeneous liquids. This issue will be the subject of our further research.
gramme, under Science for Peace project No. 982836. B.Z.R. also acknowledges the support of
TBITAK by grant No. 212T321.
References
1. Abragam A (1961) The principles of nuclear magnetism. Oxford University Press, Oxford
2. Burnett LJ (1994) Liquid explosives detection. Proc SPIE 2092:208217
3. Kumar S, McMichael WC, Kim Y-W, Sheldon A, Magnuson EE, Ficke L, Chhoa TK-L,
Moeller CR, Barrall GA, Burnett LJ, Czipott PV, Pence JS, Skvoretz DC (1997) Screening
sealed bottles for liquid explosives. Proc SPIE 2934:126137
4. Mauler J, Danieli E, Casanova F, Blmich B (2009) In: Fraissard J, Lapina O (eds) Explosives
detection using magnetic and nuclear resonance techniques. Springer, Dordrecht, pp 193203
5. Burnett LJ, McKay DR (1993) In: Connelly JM, Cheung SM (eds) Applications of signal and
image processing in explosives detection systems, Proceedings of the SPIE, vol 1824, pp 212
222. doi:10.1117/12.142899
6. Prado PJ, Lown R, Chepin J (2012) Compact bottle screening MR probe, NATO ARW on
magnetic resonance detection of explosives and illicit materials, Izmir, 26 September 2012
7. Divin Y, Lyatti M, Poppe U, Urban K (2012) Identification of liquids by high-Tc Josephson
THz detectors. Phys Procedia 36:2934
8. Gradiek A, Apih T (2010) NMR-based liquid explosives detector. Appl Magn Reson
38:485493
9. Stejskal EO, Tanner JE (1965) Spin diffusion measurements: spin echoes in the presence of a
time-dependent field gradient. J Chem Phys 42:288292
10. Slichter C (1996) Principles of magnetic resonance, 3rd edn. Springer, Berlin/Heidelberg/New
York
11. Tanner JE, Stejskal EO (1968) Restricted self-diffusion of protons in colloidal systems by the
pulsed-gradient, spin-echo method. J Chem Phys 49:17681777
12. Borgia GC, Brown RJS, Fantazzini P (1998) Uniform-penalty inversion of multiexponential
decay data. J Magn Reson 132:6577
13. Blumich B, Anferova S, Pechnig R, Pape H, Arnold J, Clauser C (2004) Mobile NMR for
porosity analysis of drill core sections. J Geophys Eng 1:177180
14. Callaghan PT, Arns CH, Galvosas P, Hunter MW, Qiao Y, Washburn KE (2007) Recent Fourier
and Laplace perspectives for multidimensional NMR in porous media. Magn Reson Imaging
25:441444
15. Kumar S (2004) Liquid-contents verification for explosives, other hazards, and contraband by
magnetic resonance. Appl Magn Reson 25(34):585597
16. Gudmundson E, Jakobsson A, Poplett IJF, Smith JAS (2009) Detection and classification
of liquid explosives using NMR. In: Proceedings of the IEEE international conference on
acoustics, speech, and signal processing. IEEE Computer Society, Los Alamitos, pp 30533056
Chapter 11
NMR-Based Liquid Explosives Detector:
Advantages and Disadvantages of Different
Configurations
Anton Gradiek, Joe Luzar, Janez Lunik, and Toma Apih
Abstract We present three different setups that can be used to discriminate between
different liquids and gels, based on various parameters obtainable by means of
nuclear magnetic resonance (NMR). Such parameters include signal amplitude,
self-diffusion constant, and spin-lattice and spin-spin relaxation times in various
magnetic fields. The presented configurations include an electromagnet with a
fast switching rate (fast field-cycling relaxometer), a permanent magnet with a
strong magnetic field gradient, and a permanent magnet where the sample is placed
in a homogeneous field. Measurements of a series of liquid samples that can
be encountered in hand luggage are presented. We discuss the advantages and
disadvantages of different detector configurations.
11.1 Introduction
Since the foiled terrorist plot to detonate liquid explosives on several commercial
airplanes in 2006, the amount of liquids passengers are allowed to bring onboard
in hand luggage has been considerably limited in several countries worldwide.
The regulation restricts liquids to a couple of containers with a maximum volume
of 100 ml each, with gels and lotions counting as liquids. A detector that could
discriminate between various liquids would speed up security checks, increase
chances of identifying potential threats and possibly allow the regulators to ease
the post-2006 restrictions.
A. Gradiek ()
Joef Stefan Institute, Jamova 39, SI-1000 Ljubljana, Slovenia
Korea Basic Science Institute, 52 Yeoeun, Yuseong, Daejeon 305-333, Republic of Korea,
e-mail: anton.gradisek@ijs.si
J. Luzar J. Lunik T. Apih
Joef Stefan Institute, Jamova 39, SI-1000 Ljubljana, Slovenia
124 A. Gradiek et al.
Nuclear magnetic resonance (NMR) is a powerful non-contact and non-invasive

technique and has been previously used to measure different parameters on which
one can base discrimination of different liquid samples. These parameters include
spin-lattice (T1 ) and spin-spin (T2 ) relaxation times, self-diffusion constant (D),
spin-spin coupling constant (J ), and NMR signal amplitude that is proportional
to the density of the investigated nuclear species in the sample [1]. The values
obtained are compared to the values from a previously collected database of a vast
variety of possible samples. The use of several parameters is advantageous because
of enhanced resolution, as opposed to the use of a single parameter.
Experiments typically employ proton NMR as most of the samples of interest
contain water or other proton-rich substances. Although providing information
about chemical composition of the samples (chemical shift spectroscopy), high
resolution NMR techniques are not appropriate for fast screening, as high and highly
homogeneous magnetic fields required are typically not achievable for large samples
or out-of-magnet measurements. Additionally, the setup is large and expensive and
the sample preparation and handling is time-extensive.
Low fields are advantageous in this regard, as the detector becomes simpler and
cheaper. Various magnet setups have been used to construct NMR-based liquid
discriminators in the past years. They include an electromagnet-based bottle scanner
that scans the bar code on the product and compares the measured values of T1
and T2 with those from the database [2], permanent magnet with a surface coil
[3], permanent magnet with the sample positioned in huge field gradient [4], the
Earths field NMR [5], ultra-low-field MRI, aided by a superconducting quantum
interference device (SQUID) sensor [6], and ultra-low field NMR using induction
coils [7]. In the last three cases, sample polarization by an additional external coil
was used in order to improve the signal-to-noise ratio. Additionally, relaxation and
diffusion measurements were applied to detect the presence of illicit substances,
such as narcotics (cocaine), dissolved in alcoholic beverages as such additives
alter the properties of the liquids [2, 8]. A theoretical analysis of processing the
Carr-Purcell-Meiboom-Gill (CPMG) decays to extract the (T2 , D) pair for liquid
discrimination was discussed in Ref. [9]. As an alternative approach, 14 N NMR was
attempted in order to identify potential explosive substances, as they almost always
contain nitrogen, whereas most of the liquids found in luggage do not [10].
The motivation for our work consisted of two parts. The first one was to construct
a relatively simple setup that allows us to measure various NMR parameters or to
use an existing setup, intended for some other application, and to use it for liquid
screening without major modifications. The focus was on different configuration of
magnets and not on the improvement of the sensors this remains a possible task in
the future. The second objective was to investigate the advantages and disadvantages
of different setups on a relatively large set of realistic samples, that are likely to
be encountered by an operating detector. Here, we present a comparison between
three types of setups. First one uses an electromagnet with a fast switching rate
(fast field-cycling), where it is possible to measure proton T1 over a broad range of
Larmor frequencies. The other two setups are both based on permanent magnets.
In one case, the sample was positioned out of the center of the magnet in order to
11 NMR-Based Liquid Explosives Detector: Advantages and Disadvantages. . . 125
provide a stabile field gradient while in the other case the sample was placed in a
homogeneous (non high-resolution) low field. Several liquids and gels were tested
in order to build a database that can be later used for sample identification.
11.2 Experimental Setups
For fast field-cycling measurements, Stelar SPINMASTER FFC-2000 relaxometer

[11] was used, in order to measure proton spin-lattice relaxation time (T1 ) in the
frequency range from 18 MHz to 5 kHz. For frequencies higher than 6 MHz, a non-
prepolarized pulse sequence (NPS) was used, while below that frequency relaxation
times were obtained using a prepolarized pulse sequence (PPS). The prepolarization
and acquisition frequencies were 18 and 9.25 MHz, respectively.
The static magnetic field gradient setup used a cylindrical Halbach magnet with a
40 mm bore with 1.1 T (50.2 MHz) magnetic field in the lateral direction (AMT&C,
Russia [12]). The sample was positioned 13 mm off the magnet center, where the
proton Larmor frequency was 48.7 MHz and the field gradient was 4.8 T/m. A
homemade probe and a homemade spectrometer were used. This setup allowed for
measuring proton T1 with the standard inversion recovery technique while T2 and
D were determined by using a series of CPMG echo sequences [4], as described in
the following.
The setup with a homogeneous field was a commercial system, originally used
for the analysis of hydrocarbon reservoir rock cores, Magritek 2 MHz Rock Core
Analyzer [13]. The sample of maximal dimensions l D 62 mm, D 39 mm
was placed in magnetic field with the proton Larmor frequency of 2 MHz. For T1
measurements, the inversion recovery technique was used. T2 was obtained from a
CPMG echo train.
In all three cases, the samples were measured at room temperature. The 2 MHz
Rock Core Analyzer uses a heater to keep the magnet temperature at 30 C for the
field stabilization, however, the sample chamber is isolated from the magnet.
11.3 Theoretical Background
In the analysis of experimental data, we use the relaxation theory that holds for
simple liquids and is a reasonable first approximation for the realistic samples
studied.
In a homogeneous field, the amplitude of spin echoes in a CPMG echo train
decays exponentially with a decay time T2 . The amplitude of the n-th echo can be
expressed as [14]

n2
M.n2/ D M0 exp ; (11.1)
T2
where 2 is the spacing between the pulses. In a presence of a magnetic field

gradient g, the decay of the echo amplitude is additionally affected by molecular
self-diffusion. In this case, the amplitude will decay as [14]

1 n2
M.n2 / D M0 exp . g/2 D.n2/ 2 exp ; (11.2)
3 T2
where is the gyromagnetic ratio and D is the self-diffusion constant. Since n2 D
t , this equation can be simplified as

1 2 2 2 1
M.t / D M0 exp . g D / t: (11.3)
3 T2
Magnetization therefore decays with a characteristic time T20 , where
1 1 1
. / D C D. g/2 : (11.4)
T20 T2 3
If we plot a graph 1/T20 . 2 /, the slope of the line will be proportional to D while the
intersection with the y-axis will correspond to 1/T2 .
Inversion recovery sequence consists of an inverting pulse, relaxation delay
t , and a reading 2 pulse. In a homogenous magnetic field or if only a weak field
gradient is present, the magnetization recovers exponentially as
Tt
M.t / D M0 2M0 e 1 : (11.5)
However, it was demonstrated in [4] that in a presence of a strong magnetic field

gradient, the measured magnetization recovery can be strongly affected by diffusion
for systems with fast diffusion rates and long relaxation times. The measured
recovery then behaves as [4]

a=2 t
M.t / D M0 1 p e T1 ; (11.6)
6Dt
where is the error function and a is the thickness of the slice of the sample that we
observe in the experiment. This relation allows us to estimate the value of T1 if we
know the values of D (measured in a separate experiment) and a, that is related to
the field gradient and the length of the inverting pulse. For systems with a very long
T1 , the magnetization recovery process is completely driven by diffusion, therefore
the value of T1 cannot be determined by this method.
11.4 Experimental Results and Discussion
11.4.1 Fast Field-Cycling
Proton spin-lattice relaxation depends on the magnetic field strength. By means of

fast field-cycling relaxometry one can measure T1 over a broad Larmor frequency
range and the T1 values at high and low fields can be used as the discrimination
parameters. Figure 11.1 shows relaxation dispersions of a representative set of
samples with significantly different field dependencies and an example of a 2-
dimensional data representation where the parameters are the T1 values at 10 kHz
and 18 MHz.
While field cycling relaxometry is not the most practical method for fast
screenings due to its complicated (non-portable) hardware, including a power supply
and a cooling system, it can provide useful information regarding the choice of
magnetic field when constructing a detector with a permanent magnet. As we can
see, the relaxation times of pure substances, such as acetone or peroxide, do not
vary significantly over the measured frequency range. On the other hand, relaxation
times of mixtures and high viscosity samples can change even over one order of
magnitude from high to low field.
a b
Fig. 11.1 (a) Proton spin-lattice relaxation times as a function of Larmor frequency for a series
of liquid and soft samples, measured at fast field-cycling relaxometer. (b) A two-dimensional
representation of relaxation times for this particular set of samples, T1 at 10 kHz vs. T1 at 18 MHz.
Experimental error is comparable with the symbol sizes. The diagonal line indicates the samples
with field-independent T1 values
11.4.2 Permanent Magnet with a Strong Field Gradient
An experimental setup, based on a permanent magnet, that allows us to measure

T1 , T2 , D, and signal amplitude, is presented in details in Ref. [4]. The NMR
probe is designed to take standard 10 mm sample tubes, with the amount of sample
typically around 1 cm3 . High magnetic field at the position of the sample (48.7 MHz)
results in a good signal-to-noise ratio and the static field gradient allows us to
measure diffusion without additional gradient coils. On the other hand, strong field
gradient affects the T1 measurements, as discussed above. Another advantage of
this approach is the fact that only a thin slice is sensitive to the radiofrequency (RF)
pulses, resulting in a well-defined volume being scanned. Therefore, information
about the proton density in the sample can be obtained, as opposed to approaches
using a homogeneous field. There, the signal strength is influenced by the shape
and position of the sample in the coil and by the filling factor of the coil. However,
while the proton density is possible to measure in the controlled environment (in
the experimental setup the sample always filled the test tube), many problems can
be expected when dealing with realistic containers in a working application it
is not always possible to ensure that the sample occupies all of the investigated
volume.
Figure 11.2 shows a D vs. T1 plot for a series of samples. Although the values
of either D or T1 often overlap within the experimental error, the samples are well
separated on a 2-dimensional plot. Additional separation can be obtained using a
3rd parameter (T2 ).
11.4.3 Low Field Permanent Magnet
To perform the T1 and T2 measurements in a homogeneous field at low field strength

(2 MHz), a commercial setup, originally intended for well-logging applications, was
used. An advantage of a homogeneous field is a bigger sensitive volume, as opposed
to the gradient configuration, where only a tiny slice is excited by the RF pulses.
In this study, a larger set of samples was analyzed, in order to construct a model
database of relaxation values. The T1 measurements were performed in two steps
for each of the samples. In the first step, the estimate of the T1 value was made,
using six delays, logarithmically spaced between 10 ms and 4 s, the range expected
to cover most of the relaxation times for investigated sample. The waiting time
between consecutive experiments was 3 s and a single repetition was used. Such
measurement takes around 20 s. In the second step, the minimum and maximum
delays and the repetition time were optimized according to the estimated T1 value
and a more detailed measurement was carried out using 16 logarithmically spaced
delays with 4 repetitions. For samples with relaxation times around 1 s or less, the
values from the estimating run typically did not differ more than 20 % from the
values, obtained by detailed measurements. The values obtained from the first run
Fig. 11.2 D vs. T1 plot for various samples, measured in the setup with a permanent magnet at
proton Larmor frequency of 48.7 MHz with a strong field gradient of 4.8 T/m. The samples are
as follows: 1 acetone, 2 peroxide, 3 milk, 4 toluene, 5 methanol, 6 Coca Cola, 7 white wine, 8
tetrahydrofuran, 9 hexane, 10 ethanol, 11 whiskey, 12 banana, 13 isopropanol, 14 acetonitrile, 15
Irish cream. The horizontal and vertical symbol dimensions correspond to the experimental error.
The horizontal line represents diffusion coefficient of water (D D 2:2 109 m2 /s). The samples
are divided into three groups, as discussed in the following: empty symbols indicate laboratory
chemicals, light grey symbols indicate beverages and dark grey symbol indicates other samples.
Full set of experimental values is listed in Ref. [4]
for the samples with T1 > 2 s were typically shorter than the actual values. T2
measurements were performed in a similar two-step procedure. The estimating run
consisted of a CPMG echo train, recorded over 1 s with 15 s spacing between the
echoes, with 12 repetitions and waiting time between repetitions of 3 s. In the second
step, the parameters were again optimized according to the estimated T2 value. In
this case, the values from the estimating run mostly matched the values from the
detailed measurement, within the experimental error. The motivation in this case was
to build a database and not to focus on the speed of each experiment. On the other
hand, a realistic detector would need to perform relaxation measurements much
faster than by the above procedure, therefore some optimization of the algorithm
and/or pulse sequences will be required in future steps.
Figure 11.3 shows a T2 vs. T1 plot of investigated samples. For convenience
reasons, the samples are divided into several groups, similar to Fig. 11.2. One group
includes laboratory chemicals, such as ethanol, methanol, toluene, peroxide, etc.
These substances typically have relaxation times longer than 1 s and the values of T1
and T2 are close together. The second group contains various drinks, such as milk,
Coca Cola, and several alcoholic beverages. This group overlaps to some degree
with the first one, however, some samples, such as milk, have considerable shorter
relaxation times. Next group are three types of petrol, 95 and 100 octane, and diesel.
Within this group, the relaxation times differ significantly between the samples, with
T1 for 100 octane petrol 2.64 s, for 95 octane 2.35 s, and for diesel 0.7 s. The fourth
Fig. 11.3 T2 vs. T1 plot for a series of samples, measured at the setup with a permanent magnet
with a homogeneous field at proton Larmor frequency of 2 MHz. Samples are divided into groups
according to their types: soft and highly viscous samples, laboratory chemicals, alcoholic and non-
alcoholic drinks, different types of petrols, and other samples. The diagonal line T1 D T2 is shown
for comparison. Experimental error (5 %) is omitted from the plot for clarity
group includes soft or highly viscous edible samples. Among them are jam, honey,
Nutella, several fruits, butter, mayonnaise, mustard, and several vegetable oils. The
common property of samples from this group is that the spin-lattice and spin-spin
relaxation is typically not monoexponential, but can be better described either as a
sum of two or as a continuous distribution of exponential functions. This feature can
be attributed mostly to the inhomogeneity of these samples, containing both various
molecular species and microscopic structures of different compositions. The values,
presented on Fig. 11.3 are the values obtained by assuming a monoexponential
relaxation and therefore depend strongly on the range and number of experimental
points for each experiment. Still, relaxation in samples from this group is mostly
considerably faster than the relaxation in samples from other groups, of the order of
magnitude of 100 ms or less.
Since the number of possible liquids found in hand luggage can be vast, we used
a model system to analyze one group of possible samples. We tried to simulate
alcoholic beverages by mixing two of their main ingredients, ethanol and water
(distilled water was used in order to exclude paramagnetic ions present in tap
water). Figure 11.4 shows T1 and T2 for model system as a function of (volumetric)
ethanol concentrations, accompanied with the values for real drinks. It is interesting
to observe a minimum in spin-lattice relaxation for the water-ethanol around
the 60 % ethanol concentration. This phenomenon has been investigated before
(although at a different magnetic field) and has been associated with the formation
of ethanol molecular clusters due to hydrogen bonds [15]. It can be observed that
both T1 and T2 values of real drinks are typically smaller than the values for the
model system what can be explained by the presence of other molecular species
Fig. 11.4 Proton spin-lattice a

relaxation time T1 (a) and
spin-spin relaxation time T2
(b) of model water-ethanol
mixture as a function of
ethanol concentration at
proton Larmor frequency of
2 MHz. The dotted lines are
guides to an eye and were
obtained by fitting the data
with a third-degree
polynomial curve. For b
comparison, the T1 and T2
values for realistic drinks are
shown in the corresponding
plots
(contributing to higher viscosity of the liquid and consequentially slower dynamics)

and paramagnetic nuclei in some of the drinks.
Liquid samples are often packed in metallic containers, such as aluminum cans
what presents a problem for NMR measurements. RF pulses induce eddy currents
in the metal and these currents generate magnetic screening fields (the skin effect).
This effect attenuates both the RF pulses form transmitter and the NMR signal
from the sample. The RF field inqmetal decreases exponentially, with the skin depth
2
usually approximated as D !
, where is the resistivity, ! is the angular
frequency, and is the magnetic permeability of the metal. At low fields, the skin
depth is larger than at high fields. In the MHz frequency range, the skin depth for
aluminum is of order of magnitude of 0.1 mm (what is the range of thickness of alu-
minum cans for drinks). To test the effect of metallic shielding on the measurements
at the 2 MHz system, we encased the sample (40 ml PET bottle of water with added
CuSO4 ) in the aluminum foil. The proton signal from empty bottle is negligible.
Figure 11.5 shows the T1 and T2 measurements of the unshielded sample and of the
sample, wrapped in one layer of foil (d D 30 m). Although the signal from the
wrapped sample is more than two times smaller than the signal from the unshielded
sample, it is still possible to determine both T1 and T2 values with sufficient accu-
racy. However, the presence of metal changes the inductance of the coil, therefore
the resonant circuit has to be tuned to new conditions. In our study, two layers of foil
already changed the resonant circuit to the degree where it was not possible to tune
it because of the limited range of capacitors. On the other hand, the tuning problem
could in principle be avoided with the use of untuned or broadband sensors.
Fig. 11.5 (a) Magnetization a

recovery after the inversion
recovery pulse sequence at
2 MHz for water with added
CuSO4 in a PET bottle and
for the same sample, wrapped
in one layer of aluminum foil.
Solid lines represent fit to the
equation M.t / D M1
t
M 0 e T1 , where M 0 and
M1 are related to the sample
magnetization at short and
long delays, respectively.
(b) CPMG echo integral as a
function of time at 2 MHz for b
water with added CuSO4 in a
PET bottle and for the same
sample, wrapped in one layer
of aluminum foil
11.5 Conclusions
In this work, we present three different setups for measuring different NMR
parameters of liquids and soft materials, that can provide insight in constructing an
NMR-based liquid explosives detector. Each setup has some advantages and some
drawbacks.
Fast field-cycling
Advantages:
The system allows measurements at any chosen field within its range
(10 kHz18 MHz), taking advantage of the fact that different samples
exhibit different frequency dependencies of relaxation times. That way the
strength of the magnetic field, where the differences are the largest, can be
chosen.
Disadvantages:
The complexity of the setup makes this configuration unpractical for fast
screening applications
Small sample volume
Permanent magnet with a strong field gradient

Advantages:
Diffusion measurements are possible
Spatial resolution
Planar configuration allows measurements using large samples
Higher signal-to-noise ratio at high fields than at low fields
Disadvantages:
Small sensitive volume
Diffusion influences T1 measurements if using a static gradient
Low field permanent magnet
Advantages:
Simple T1 and T2 measurements
Large sample volume possible
Disadvantages:
Diffusion measurements are not possible without additional modifications
(gradient coils)
No spatial resolution
Cannot determine proton density
A similar technique, not explored here, is the detection in ultra-low magnetic

field, even down to Earths magnetic field. There, prepolarization coils can be used
to improve the signal-to-noise ratio [5, 7].
Closer to practical implementation are the two configurations, based on per-
manent magnets. While a homogeneous field configuration allows larger sample
volumes, resulting in better signal-to-noise ratio, field gradients are required in
order to measure diffusion. The gradient can be obtained either by using gradient
coils or by placing the sample in the gradient field of the magnet. The gradient
configuration also allows spatial resolution, to some degree, that homogeneous field
configuration does not. However, it is not very realistic to expect useful information
regarding proton density even in the gradient configuration, as it is difficult to
assure that the sensitive volume will be completely filled with the sample during
the scan.
In our setup, a cylindrical Halbach magnet was used, with the sample
placed inside the bore. Alternatively, other groups have been developing planar
configurations of permanent magnets, combined with a surface sensor, such as
NMR-MOUSE [16] and NMR-MOLE [17]. Such configurations are probably even
more appropriate for the actual realization of the detector since the dimension of
the sample is not limited by the bore diameter.
When building a sample database, one has to take special attention to the samples
that exhibit non-monoexponential relaxation (present in many soft and viscous
samples investigated within this study). If assuming a simple exponential decay, the
obtained T1 and T2 values strongly depend on the number and range of experimental
delays measured.
RF pulses at low frequencies can penetrate thin metallic shielding and relaxation
measurements are still possible. However, tuning of the resonant circuit can present
a problem since the inductance of the coil changes with the presence of metallic
object. This issue can be possibly avoided with the use of untuned or broadband
sensors.
The existing prototype detectors mainly focus on T1 and T2 measurements.
It is clearly advantageous to add diffusion measurements to the detector, as D
represents an additional dimension and makes the discrimination easier. Magnetic
field gradients, required for such measurements, can be obtained either with gradient
coils or with a permanent magnet. The advantages of the gradient coils include the
fact that the gradient can be switched off during the relaxation measurements. On
the other hand, higher and stable field gradients can be obtained using a permanent
magnet. For comparison, the full-body MRI systems operate with pulse gradients
of around 50 mT/m while the gradient in the presented setup was almost 5 T/m.
The disadvantage of the permanent magnet setup includes interference with the
relaxation measurements for samples with longer T1 and fast diffusion.
The optimal detector would include measurements of all three important param-
eters T1 , T2 , and D. The practical configuration of hardware depends on particular
demands of the end user. We also highlighted some of the obstacles likely to be
encountered during the real operation.
References
1. Burnett LJ, McKay DR (1993) Applications of signal and image processing in explosives
detection systems. doi:10.1117/12.142899
2. Kumar S (2004) Liquid-contents verification for explosives, other hazards, and contraband by
magnetic resonance. Appl Magn Res 25:585
3. Mauler J, Danieli E, Casanova F, Blmich B (2009) Explosives detection using magnetic and
nuclear resonance techniques. Springer, Dordrecht, pp 193203
4. Gradiek A, Apih T (2010) NMR-Based Liquid Explosives Detector. Appl Magn Res 38:485
5. Sato-Akaba H, Itozaki H (2012) Development of the Earths Field NMR Spectrometer for
Liquid Screening. Appl Magn Res 43:579
6. Espy M, Flynn M, Gomez J, Hanson C, Kraus R, Magnelind P, Maskaly K, Matlashov A,
Newman S, Owens T, Peters M, Sandin H, Savukov I, Schultz L, Urbaitis A, Volegov P,
Zotev V (2010) Ultra-low-field MRI for the detection of liquid explosives. Supercond Sci
Technol 23:034023
7. Austin R, Espy M, Matlashov A, Sandin H, Schultz L, Urbaitis A, Volegov P (2013) MagViz:
a bottled liquids scanner using ultra-low field NMR relaxometry. In: Apih T, Rameev B,
Mozzhukhin G, Barras J (eds) Magnetic resonance detection of explosives and illicit materials.
Springer, Dordrecht, pp 99
8. Prado PJ, Mastikhin I, Karlsson MT (2012) Appl Magn Res 43:531
9. Gudmundson E, Jakobsson A, Poplett IJF, Smith JAS (2009) IEEE international conference
on acoustics, speech, and signal processing, Taipei. IEEE Computer Society, Los Alamitos,
pp 30533056
10. Prado PJ (2013) Bottled liquid scanner for security checkpoints. In: Apih T, Rameev B,
Mozzhukhin G, Barras J (eds) Magnetic resonance detection of explosives and illicit materials.
11. http://www.stelar.it/
12. http://www.amtc.ru/en/
13. http://www.magritek.com/
14. Slichter C (1996) Principles of magnetic resonance. Springer series in solid-state sciences, 3rd
edn. Springer, Berlin, p 369
15. Takamuku T, Kumai T, Yoshida K, Otomo T, Yamaguchi T (2005) Structure and Dynamics
of HalogenoethanolWater Mixtures Studied by Large-Angle X-ray Scattering, Small-Angle
Neutron Scattering, and NMR Relaxation. J Phys Chem A 109:7667
16. Blmich B, Blmler P, Eidmann G, Guthausen A, Haken R, Schmitz U, Saito K, Zimmer
G (1998) The NMR-mouse: construction, excitation, and applications. Magn Reson Imaging
16:479484
17. Manz B, Coy A, Dykstra R, Eccles CD, Hunter MW, Parkinson BJ, Callaghan PT (2006) A
mobile one-sided NMR sensor with a homogeneous magnetic field: The NMR-MOLE. J Magn
Reson 183:2531
Chapter 12
Composite Pulses in Inhomogeneous Field NMR
Galina S. Kupriyanova, Vitaliy V. Molchanov, Evgeniy A. Severin,

and Ivan G. Mershiev
Abstract This work presents the result of a series of experiments for the study
of composite pulses to compensate the inhomogeneity of a magnetic field in
measurements of relaxation parameters in nuclear magnetic resonance (NMR).
The possibilities of the identification are based on the relaxation characteristics
of the NMR signals. An experimental study of 18 types of composite pulses for
the excitation of the induction signal and spin echo without phase distortion was
made. The most effective types of composite pulses were proposed for relaxation
measurements in low inhomogeneous fields.
12.1 Introduction
In recent years, a tendency to minimize the equipment with compact permanent

magnets in low field nuclear magnetic resonance (NMR) has been developed
[13]. Although high resolution NMR in a high field has excellent possibilities for
chemical analysis of liquids, the mobile applications are based on the low field NMR
relaxometry. Low field NMR may be used to identify liquid explosives, to establish
the structure of synthesized compounds, to identify illicit drugs and counterfeit
products [3, 4]. New NMR detectors for industrial applications were proposed [1, 4].
To minimize NMR detectors the use of small permanent magnets based on new
composite materials was developed too. The introduction of permanent magnets
leads to a number of problems associated with the low intensity of signals in a
low magnetic field, the need to eliminate distortions caused by static magnetic
field inhomogeneity and heterogeneous sample excitation RF field, the lack of
stability in frequency due to the effect of temperature drift in a magnet and a
G.S. Kupriyanova () V.V. Molchanov E.A. Severin I.G. Mershiev

Department of Radiophysics and Information Security, Institute of Physics and Technology,
Immanuel Kant Baltic Federal University, 14 Nevskogo St, 236041 Kaliningrad, Russia
e-mail: galkupr@yandex.ru
138 G.S. Kupriyanova et al.
spectrometer probe. Additionally the frequency shift produces phase distortions

caused by off-resonant irradiation of a sample. One way to solve these problems
is the application of composite pulses [316].
The low field NMR spectroscopy has a small signal to noise ratio (SNR). Thus,
it is necessary to use the multipulse series for accumulation of the signals to
increase SNR. To achieve the optimal values of the signal/noise ratio, the refocusing
cycles are additionally used. It is results to eliminate the phase distortions for the
accumulation of the signals. The most used sequence is Carr-Purcell-Meiboom-
Gill (CPMG): 90x ( 180y )n . Ideal 180 refocusing cycle, with the
axis of rotation lying completely in the transverse plane, fully eliminates the loss
of coherence in the transverse magnetization associated with frequency dispersion
and heterogeneity of radio frequency and static magnetic fields. However, using of
traditional refocusing pulse sequences [1416] does not always lead to the desired
results in small-sized facilities with permanent magnets and outside magnet [1].
The magnetic field gradients may be changed during long-time measurements due
to absence of the magnetic field stabilization. The using of 180 pulses in the
refocusing cycle with RF power limits results in an error caused by the deviation of
the effective axis of refocusing sequence from the axis of the magnetization rotation.
The transients, which occur in these cases, lead to a modulation of the echo that is
the source of great uncertainty in the short relaxation time measurements.
In this work we studied the set of composite pulses to solve the problem of
the refocusing phase distortion and to create the best composite pulse to invert
magnetization. The part of studied composite pulses was successfully used in some
applications of NMR spectroscopy. However, it was also noted that most of the so-
called composite pulses with a constant-rotation were not ideal for use in spin-echo
experiments [1719].
In this paper, the effectiveness of the different composite pulses for free induction
and echo signals excitation in a small setup of nuclear magnetic resonance with the
permanent magnet B D 328 mT have been investigated.
The main purpose of our study was to select the best set of composite pulses
to compensate the phase shift for 90 and 180 magnetization rotation when using
of limited the rf power with resonance frequency offset and in the presence of a
magnetic field gradient, which may change due to temperature drift. In this paper
the possibility of the improvement of the sensitivity and correct measurement of
relaxation times has been analyzed.
12.2 Theory
A composite pulse consists of a set of RF pulse sequences with different phases and
durations [13]. The action of the composite pulse can be equivalent to the usual 90 -
or 180 - pulses. An additional rotation of the magnetization vector with the varying
phase of pulses can compensate for the effect of the magnetic field inhomogeneity
in the sample volume or the effect of the resonance offsets.
12 Composite Pulses in Inhomogeneous Field NMR 139
The action of the composite pulse, consisting of a set of ideal pulses can be
j
described by propagator b
U [8]:
j j j
b
U DbU n :::b
U1 (12.1)
Here
b n
U j p D exp ijpb
IJ njp
o
describes the result of spin rotation b

j
IJ under the influence of individual pulse p, p
j
is the rotation angle, np is the rotation axis. For an ideal rectangular pulse with a
j
nominal flip angle p D 0p and the phase P the rotation axis is given by:
njp D ex cos p C ey sin p (12.2)

j
In the presence of the resonance frequency offset j D 0 1 or in the
amplitude deviation of the pulse, the rotation angle is given by:
v
u
u 2
0 t .1 / C . /
j 2
p D p
j
0 2 (12.3)
1
Here the following notations are introduced:

j
0 is the Larmor frequency of the j-nuclear spin with magnetogyric ratio j in
j
static magnetic field B0 , 0 D j B0 . 1 is the transition frequency. It is close to
the Larmor frequency. 1 D j B1 , B1 is a radio frequency field amplitude at the
registration of signal, 01 is the nominal precession frequency that is given by the
relationship 01 0p D 2 . The rotation axis is determined by the equation:
njp D ez cos j C ey sin j cos p C ex sin j sin p (12.4)
where j is derived from:

1
tan j D
j
According to Euler, any rotation may be presented as the series of three
consequential rotations around axis Z, around axis Y and around new axis Z again.
Then the propagator of the composite pulse may be seen as:
j n o n o n o
b Ijz exp iN jb
U D exp iN jb Ijy exp iN jb
Ijz (12.5)

where N j ; N j ; N j are Euler angles. Besides, the values of Euler angles depend on
and are different for each set of spins OIj . The properties of composite
1 j
01
and
01
pulses are determined by the dependence of the three Euler angles on the parameters
1 j
0
and
0
.
1 1
One of the first applications of composite pulses in NMR was the compensation
for the rotation angle errors in spin echo experiments [13]. In Levitts and Freemans
paper the application of composite pulses in Carr-Levitt series have been studied
for the longitudinal relaxation time measurement. As a refocusing pulse, the first
proposed composite pulse 900 18090 900 was used. The results of the experiments
with the magnetization vector inversion depend on the residual component of the
magnetization vector lying in XY plane. To reduce its influence and to increase
the accuracy of relaxation times measurements, a phase cycling i.e. a sequential
shift of phases of influential pulses, may be used. Thus, the phase shifts caused by
the residual transverse magnetization will be mutually compensated in each second
accumulation cycle. The signal amplitude equals (1 cos)/2 of the maximum, with
as the effective rotation angle.
According to Levitt, the composite pulse trajectory of magnetization vector is
not symmetrical. For the Carr-Parcell-Meiboom-Gill sequence pulse the imperfect
90 pulse will result in a residual longitudinal magnetization component, which
will be compensated in even and odd echo signals. For reasons of asymmetry, the
application of such compound pulse in a spin-echo experiment results in a strong
dependence of the signal phase on the frequency offset. The further development
of the theme was the creation of such composite pulses which does not lead to
dependency on the phase of the signal frequency offset and thus does not introduce
phase distortions in an echo signal [15, 16].
Applying the Magnus dispersion, R. Tycko introduced extra conditions for the
resultant rotation axis for the construction of 180 pulses and as well as for the result
of rotation angle for 90 pulses [1416]. As a result, the sequences of extremely
low compound of phase shift have been developed, where the zero and first order
Magnus expansion were almost reduced to zero with maintaining a small number of
pulses in the sequence.
These observations were further developed in papers [17, 18]. Relying on the
initial terms which allowed to avoid phase shifts, the general set of composite
180 pulses was proposed, for which the residual Magnus terms of the expansion
would be zero both for inhomogeneous fields and for cases of a frequency offset.
The proposed composite pulses are very wideband and highly effective due to
the fact that the first two terms in the Magnus expansion are equal to zero. The
only disadvantage is the relatively long length of the sequence. The minimal
antisymmetrical sequence length is five pulses.
It was demonstrated in the recent article [18], that in general the antisymmetrical
composite pulse consists of 713 180 pulses. Their phases can be presented the
following [18]
Xk1
0k D .1/kC1 k C .1/j C1 2j :
j D1
In this case the pulse structure depends on the phase relations between its parts and
the start phase may be arbitrary.
12.2.1 Experimental Results and Discussion
We experimentally investigated the conditions of effective excitation signal through

composite pulses in NMR installation created on the basis of the NMR/NQR
Apollo console (TECMAG inc.) operating in the frequency range from 0.5 to
500 MHz. Nuclear magnetic resonance of the protons was observed at a frequency
of 13.99 MHz. We used a permanent magnet without temperature stabilization with
homogeneity of the magnetic field near B/B 104 .
Since in this setup the phase distortion can be caused by various factors and their
totality, the focus of our attention is on the search for such composite 90 and 180
pulses, which would compensate some effects directly caused by the inhomogeneity
of the magnetic field, the inhomogeneous excitation of the sample, the resonance
offsets.
The special attention was paid to the search for such combination of the
composite 90 and 180 pulses that would give consistent results with long-term
measurements of the relaxation times, and immediately make up for a number of
effects caused by non-uniform excitation of the sample, the resonance offset and the
thermal fluctuations of the magnetic field.
Alcohol, motor oil, Coca-Cola, glycerin and a set of polymers (kestamid pro-
duced by Polikim Corp, Turkey) were used as the samples. Preliminary experiments
on the relaxation times measurements of alcohol by inversion-recovery method
and spin-echo using the conventional /2 and pulses of different durations have
been performed. For each experiment, the /2 pulse duration is determined by
the maximum induction signal by means of appropriate of rf power selection. The
bandwidth of 20 kHz filter is installed. Spectral line width of this sample reached
129 Hz. The analysis of experimental data showed that the relaxation times are
dependent on the duration of the 90 pulse and offset resonance frequency j .
The nonlinear dependencies of longitudinal T1 () and transverse T2 () relaxation
times are observed for short durations of 90 pulse (from 2 to 8 s) (Fig. 12.1). The
measurement accuracy of relaxation times decreases dramatically when measuring
viscous samples and solids with a broad linewidth. Therefore, the use of data on
the relaxation times as a diagnostic sign of the samples identification becomes
problematic.
In order to select the best set of composite pulses for effective excitation NMR
signal and relaxation measurements in samples of different origin the NMR signals
have excited by composite pulses, equivalent to 90 and 180 pulses. In order to
investigate the most effective compensation phase the intensity and the line width
of the NMR signal were studied as a function of the type of composite pulse,
the residual magnetization after inversion under 180 pulse are determined with
the use of conventional 180 , and composite 180 pulses with offset resonance.
We compared the results of experiments carried out at the beginning of the
spectrometers work, and after its long-term work.
To estimate the effectiveness of composite pulses versus equivalent conventional
pulses, we have taken into consideration several factors which can affect the NMR
signal, such as inhomogeneous excitation of the sample, inhomogeneity of magnetic
Fig. 12.1 The dependence of relaxation times of aqueous solution of alcohol on the 90 pulse
length
field, resonance offset. For this purpose we introduced some dimensionless param-
eters. The seriousness of the resonance offset may be quantified by the relative
j
resonance offset A D ! 01
which equal to the ratio of the resonance offset of
received signal to the nominal nutation frequency determined from the condition
01 D 2 . Here is the 90 pulse duration.
B is equal to the ratio of NMR signals linewidth after excitation to conventional
90 (or 180 ) pulse to the linewidth after excitation with composite pulse sequence,
and describes refocusing performance of the sequence.
C is equal to ratio of signal-to-noise ratios of signals, obtained with composite
pulse sequences and with conventional ones. Beside this, a stability of parameters
was tested due to the temperature and equipment parameters drift.
The composite pulse sequences from Tables 12.1 and 12.2 were tested.
We have studied the effect of broadband composite pulses at various rf field
power and offset values under which it is possible to carry out the rotation by angle
(90 or 180 ), characterized by 10 from the desired ideal angle. Measurements
were taken at different rf power. The 90 pulse duration was 9 s at B D 35 dB and
5 s at D 32 dB. The amplitude and line width signal of different sampled were
studied. The results of experiments are given in Table 12.1. In Table 12.1, the first
j
column shows the published data [68, 21, 22] for A D ! 0
at a deviation angle
1
of 10 to 0p at the nominal rf field [8]. The obtained results have shown that the
efficiency of the composite pulses for samples of different nature is different, other
things being equal.
Table 12.1 The relative value of the line width B obtained for the effects of 90 composite
pulse and the relative value of signal-to-noise ratios C, obtained with composite pulse sequences
and to conventional ones
Alcohol Oil Kestamid
a
! j
Composite pulse AD 01
B b
C B b
C Bb C
90 0.92, 0.92 [8] 1 1 1 1 1 1
900 9090 0.1, 0.1 [19] 0.92 0.99 0.85 1.04 1.0 1.14
900 180120 0.13, 0.13 [20] 1.65 1.06 0.82 0.96 0.79 1.41
1800 360180 1800 270180 9090 1.24, 1.24 [21] 1.64 0.88 0.9 1.19 0.98 1.46
240 152180 3460 152180 240 0.4, 0.4 [6] 1.02 0.8 1.17 0.95 1.09 1.34
3850 320180 250 0.3, 0.3 [22] 1.06 0.6 1.14 0.99 0.66 1.71
18097.2 360291.5 18097.2 900 0.88, 0.36 [7] 1.39 0.3 1.08 0.38 0.98 1.08
113180 3160 113180 0.20, 0.20 [6] 1.01 0.9 0.9 1.01 2.14 1.09
a
The data are taken from [8]. According to [8] the range of offset value for which the angle 0p
is between 80 and 100 at nominal rf field
b
The line width measurement was made in the relative frequency range from A D 0.1 to 0.03
with 90 pulse length equal to 9 s. The relative measurement error was estimated to be 12 %
Table 12.2 The relative value of the line width B obtained for the effects of 180 composite
pulse and the relative value of signal-to-noise ratios C, obtained with composite pulse
sequences and to conventional ones
Oil Kestamid

! j
Composite pulse AD 01
B C B C
180 0.08, 0.08 [8] a
1 1 1 1
9090 1800 9090 0.09, 0.09 [8]a 0.75 1 0.8 0.92
270 99180 1800 211180 3860 211180 1800 99180 270 0.85, 0.85 [6] 0.9 0.8 1.2 0.87
ASBO-9(7A) 0.428 [18]b 1.5 1.1 0.6 0.52
360162.4 180313.4 180284.5 1800 18075.5 18046.6 360197.6
ASBO-9(7C) 0.332 [18]b 0.8 0.8 0.6 0.41
18046.6 18075.5 360255.5 1800 360104.5 180284.5 180313.4
180256 18052 1800 180128 1800 180232 1800 180308 180104 0.224 [18]b 0.9 1.2 0.9 0.46
a
According to [8] the range of offset value for which the angle 0p is less than 10 at nominal rf
field
b
According to [18] 0.428 means that the spin-echo amplitude has >99 % of its full amplitude
between A D 0.428 and C0.428
For non-viscous liquids the effect of the 90 composite pulses (except for the
900 ,9090 -pulse) led to a narrowing of the line width of 180 Hz (at 90 pulse)
to 120146 Hz. The strongest effect was observed for the 90 composite pulse
900 180120 , which were initially designed to compensate for distortions caused by
phase shifts. However, this pulse is effective in a very narrow frequency offsets
range. In a wider offset range the maximum signal narrowing in inhomogeneous
field is obtained with 1800 360180 1800 270180 9090 composite pulse. The relative
change in the line width when using composite pulses was 2037 % (Table 12.1).
For viscous compounds, like oil and glycerin, the maximum narrowing was
obtained with 900 180120 , 1800 360180 1800 270180 9090 and 240 152180 3460 152180 240
composite pulses.
The relative change in the linewidth, measured at the same resonance offset, by
using composite pulses is insignificant. During prolonged signal measurements with
90 pulse, it was found that in the absence of stabilizing the frequency shift of the
resonance frequency increases, and the line width increases to 146245 Hz after 4 h
measurements. The linewidth changed accordingly from 161 to 214 Hz when using
of composite pulse.
For solid samples the best results for 90 composed pulses are obtained with the
use of 1800 360180 1800 270180 9090 , 240 152180 3460 152180 240 , and 113180 3160 113180
sequences. However, 1800 360180 1800 270180 9090 sequence allows to obtain better
results in the relative offsets range which is wider than 900 180120 .
The dependence of the residual magnetization value on the resonance offset
at various power of radiofrequency field was studied after the magnetization
inversion by 180 composite pulse. Experiments were performed on inversion
of magnetization using 12 different composed pulses. The special interest
is to test composite pulses such as ASBO-9 (7C), ASBO-9 (7A), ASBO-9
(B1), which are dual-compensated for B1 inhomogeneity and offset [18].
The results were compared to the effect of simple 180 pulses at the
corresponding offset value. The results for some effective pulse are given
in Table 12.2. The analysis showed that 9090 1800 9090 , 180120 180240 180120
and 270 99180 1800 211180 3860 211180 1800 99180 270 composite pulses invert the
magnetization in the most effective way. Figure 12.2 demonstrates the dependence
of the relative signal value on the resonance frequency offset A. Figure 12.2 it is
shows that the experimental curve for composite pulse 9090 1800 9090 is similar to the
theoretical curve described in [12] and 270 99180 1800 211180 3860 211180 1800 99180 270
composite pulse inverts magnetization at the wider offset range and in a more
uniform way. As it can be seen from the Table 12.2 and Fig. 12.2 the composite
pulse sequences ASBO-9(7C) and ASBO-9(7A) cause more efficient magnetization
refocusing and the line width narrows down, while the amount of residual
signal decreases. It should be pointed out that during long-time experiments the
frequency offset increases, and the composite pulses ASBO-9(7C) and ASBO-
9(7A) demonstrate stable results of magnetization inversion, thus compensating the
signal broadening due to the increase of offset and extra heterogeneity of magnetic
field caused by temperature effect.
12.3 Conclusion
We investigated the possibilities of compensating disadvantages of small permanent

magnets with the homogeneity of a permanent magnetic field near B/B 104
105 by the use of composite pulses. The analysis of the experimental data showed
that the impact of composite pulses in an inhomogeneous magnetic field depends
Fig. 12.2 The dependence of the residual magnetization level after the single 180 pulse and
equivalent 180 composite pulse, depending on the resonance offset (a) the duration of 180 pulse
was 18 s, (b) the duration of 180 pulse was 10 s
on the frequency offset from the resonance and is different in liquids and solids.
So the equivalent 90 composite pulses lead to a linewidth narrowing in liquids
(Table 12.1), but do not yield any advantage when subjected to solid-state sample.
It should be noted that for the signal registration in a solid sample it is necessary
to accumulate signals to achieve a greater signal to noise ratio. Due to the lack of the
resonance condition stabilization, this leads to an additional linewidth broadening.
However, we observed an increase in parameter C, which characterizes a relative
change in the signal/noise ratio of the same number of accumulation in all cases.
Thus the following equivalent 90 composite pulses: 900 180120 , 9090 1800 9090 ,
900 180120 , 1800 360180 1800 270180 900 may be recommended for experiments with
liquid samples. The composite pulses: 1800 360180 1800 270180 9090 , 240 152180 3460
152180 240 , and 113180 3160 113180 give the better result for solids. However,
1800 360180 1800 270180 9090 sequence allows obtain better results in the relative
offsets range which is wider than 900 180120 . In relaxation experiments, it is
important to eliminate phase distortions which induce residual signals after
magnetization inversion. The investigation of the residual magnetization magnitude
and the shape echo after the action of a composite pulse have showed that ASBO-9
composite pulses give the best results, particularly ASBO-9 (7C). Moreover, the
replicable results were obtained during long-term measurements. These sequences
will be used in further relaxation experiments.
gramme, under Science for Peace project No. 982836.
References
1. Blmich B, Casanova F, Appelt S (2009) NMR at low magnetic fields. Chem Phys Lett
477:231240
2. Blumich B, Guthausen A, Zimmer G, Blumler P (1998) Analysis of polymer materials by
surface NMR via the MOUSE. J Magn Reson 130:17
3. Mauler J, Danieli E, Casanova F, Blumich B (2009) Identification of liquids encountered in
carr-on-luggage by mobile nmr. In: Fraissard J, Lapina O (eds) Explosives detection using
magnetic and nuclear resonance techniques, NATO science for peace and security series B.
4. Kumar S, Prado P (2009) Detection of concealed liquid explosives and illicit drugs in unopened
bottles. In: Fraissard J, Lapina O (eds) Explosives detection using magnetic and nuclear
resonance techniques, NATO science for peace and security series B. Springer, Dordrecht,
p 73
5. Shaka AJ, Freeman R (1983) Composite pulses with dual compensation. J Magn Reson
55:487493
6. Shaka AJ, Pines A (1987) Symmetric phase-alternating composite pulses. J Magn Reson
71:495503
7. Wimperis S (1994) Broadband, narrowband and passband composite pulses for use in advanced
NMR experiments. J Magn Reson A 109:221231
8. Levitt MH (1996) Composite pulses. In: Grant DM, Harris RK (eds) Encyclopedia of nuclear
magnetic resonance. Wiley, Chichester, pp 26942711
9. Waugh JS (1982) Systematic procedure for constructing broadband decoupling sequences.

10. Shaka AJ, Keeler J, Frenkiel T, Freeman R (1983) An improved sequence for broadband
decoupling: WALTZ-16. J Magn Reson 52:335338
11. Srensen OW, Eich GW, Levitt MH, Bodenhausen G, Ernst RR (1983) Product operator
formalism for the description of NMR pulse experiments. Prog NMR Spectrosc 16:163192
12. Bodenhausen G, Kogler H, Ernst RR (1984) Selection of coherence-transfer pathways in NMR
pulse experiments. J Magn Reson 58:370388
13. Levitt M, Freeman R (1981) Compensation for pulse imperfections in NMR spin-echo
experiments. J Magn Reson 43:6580; (1979) J Magn Reson 33:473
14. Tycko R (1983) Broadband population inversion. Phys Rev Lett 51:775777
15. Tycko R, Pines A, Guckenheimer J (1985) Fixed point theory of iterative excitation schemes
in NMR. J Chem Phys 83:27752802
16. Tycko R (1985) Composite pulses without phase distortion. J Magn Reson 61:90101
17. Odedra S, Wimperis S (2012) Use of composite refocusing pulses to form spin echoes. J Magn
Reson 214:6875
18. Odedra S, Thrippleton MJ, Wimperis S (2012) Dual-compensated antisymmetric composite
refocusing pulses for NMR. J Magn Reson 225:8192
19. Freeman R, Kempsell SP, Levitt MH (1980) Radiofrequency pulse sequences which compen-
sate their own imperfections. J Magn Reson 38:453479
20. Levitt MH (1982) Symmetrical composite pulse sequences for NMR population inversion.
I. Compensation of radiofrequency field inhomogeneity. J Magn Reson 48:234264
21. Levitt MH, Ernst RR (1983) Composite pulses constructed by a recursive expansion procedure.
22. Tycko R, Schneider E, Pines A (1984) Broadband population inversion in solid state NMR.
J Chem Phys 81:680689
Part III
Other Techniques
Chapter 13
Novel HTS DC Squid Solutions
for NMR Applications
Maxim L. Chukharkin, Alexey S. Kalabukhov, Justin F. Schneiderman,

Fredrik isjen, Magnus Jnsson, Minshu Xie, Oleg V. Snigirev,
and Dag Winkler
Abstract We have developed a multilayer flux-transformer-based high-TC SQUID

(flip-chip) magnetometer that improves signal-to-noise-ratios (SNR) in ultra-low
field magnetic resonance (ulf-MR) recordings of protons in water. Direct ulf-MR-
based benchmarking of the flip-chip versus a standard planar high-TC SQUID
magnetometer resulted in improvement of the SNR by a factor of 2. This gain
is attributable to the improved transformation coefficient (1.9 vs 5.3 nT/0 ) that
increased the signal available to the flip-chip sensor and to the lower noise at the
measurement frequency (15 vs 25 fT/Hz1/2 at 4 kHz). The improved SNR can lead
to better spectroscopic resolution, lower imaging times, and higher resolution in
M.L. Chukharkin ()

Department of Microtechnology and NanoscienceMC2, Chalmers University of Technology,
Gteborg SE-412 96, Sweden
Faculty of Physics, Moscow State University, Moscow, Russian Federation
e-mail: maxim.chukharkin@gmail.com
A.S. Kalabukhov
Skobeltsyn Institute of Nuclear Physics, Moscow State University, Moscow, Russian Federation
J.F. Schneiderman
MedTech West and the Institute of Neuroscience and Physiology, Sahlgrenska Academy
and the University of Gothenburg, Gteborg, Sweden
F. isjen M. Jnsson M. Xie D. Winkler
O.V. Snigirev
Faculty of Physics, Moscow State University, Moscow, Russian Federation
152 M.L. Chukharkin et al.
ulf-MR imaging systems based on high-TC SQUID technology. The experimental

details of the sensors, calibration, and ulf-MR benchmarking are presented in this
report.
13.1 Introduction
Ultra-low-field (ulf) NMR/MRI is a new spectroscopic/imaging modality that can

potentially lead to cheaper, simpler, and, for some applications, including MR
detection of explosives and illicit materials, better magnetic resonance capabilities.
Because the measurement field (B0 ) is on the same order of magnitude as that
of the earths magnetic field (50 T), ulf-MR systems do not require bulky
and expensive superconducting coils, thereby reducing up-front and running costs.
Perhaps most importantly, T1 contrast at ultra-low fields is significantly enhanced
and allows, for example, contrast-free imaging of prostate cancer [1].
However, recording at ultra-low fields poses several challenges, the most sig-
nificant of which is a low signal-to-noise ratio (SNR). Standard MR systems tend
to employ high magnitude measurement fields (several Tesla) and coil-based RF
antennas as the sensing element because the signal from such a coil is proportional
to the measurement field squared. As that field is reduced below the milliTesla
range, SQUIDs become advantageous because the signal from a SQUID is directly
proportional to the measurement field [2, 3].
There are presently two approaches to ulf-MR with SQUID sensors: low-
and high-TC technology. While high-TC SQUIDs benefit from relaxed cooling
requirements, their sensitivity are significantly lower than their low-TC counterparts.
Our group developed an ulf-MR system based on planar direct-coupled high-
TC SQUID magnetometers (Fig. 13.1, left) with the aim of combining it with our
MEG capability for simultaneous structural and functional brain imaging [4]. While
we have achieved moderate spectroscopic resolution in NMR recordings (see later
discussion on Fig. 13.4a) and have generated 2D images [5, 6], the present SNR
level makes our long-term goal of ulf-MR imaging of the human brain unfeasible
because e.g. acquisition times would be on the order of tens of hours.
The typical sensitivity of our planar direct-coupled high-TC SQUID magnetome-
ters is 40 fT/Hz1/2 at 4 kHz [4]. The main limitation of the direct-coupling scheme
is an inductance mismatch between the pickup loop and the SQUID. A candidate
solution to improve this mismatch and thus increase the magnetic field sensitivity of
high-TC SQUID magnetometers is to use a high-TC superconducting multiturn flux
transformer [7].
To this end, we have designed and fabricated high-TC multilayer flux transform-
ers on 10 mm 10 mm SrTiO3 (STO) substrates. We chose a configuration wherein
the flux transformers and SQUIDs are fabricated on separate substrates that are
pressed together in order to form flip-chip SQUID magnetometers. We compared the
performance of these sensors with that of our planar high-TC SQUID magnetometers
in ulf-MR experiments and have achieved an improvement in SNR by a factor of
more than 2.
13 Novel HTS DC Squid Solutions for NMR Applications 153
Fig. 13.1 Schematic diagrams of the sensors compared in the ulf-MR system. Top row: The
planar SQUID magnetometer (left) and the flux transformer (right) fabricated on the same size
10 mm 10 mm substrates. Bottom row: zoomed-up view of the (redundant pair of) SQUIDs of
the planar sensor (left) and the input coil of the flux transformer (right)
13.2 Methods
13.2.1 Fabrication of SQUID Magnetometers
Our recently developed chemical mechanical polishing (CMP) technique was used
to produce multilayer superconducting flux transformers [8]. The transformers were
fabricated on double-side polished STO substrates and had 9 mm 9 mm pickup
loops. The multiturn spiral input coils had a square shape with outer dimensions of
0.6 mm 0.6 mm and consisted of a 10-m wide YBa2 Cu3 O7x (YBCO) line with
5-m pitch (see Fig. 13.1). The 11 turns of the input coil were chosen to match
inductance with that of the pickup loop. Detailed information about the flip-chip
magnetometer design is described in reference [8].
Pulsed laser deposition (PLD) was used for the YBCO-electrode and RF mag-
netron sputtering for the insulator-layer deposition. The thicknesses of the YBCO
films and insulator layers were 300 and 330 nm, respectively. All electrodes were
patterned using UV photolithography and ArC ion-beam etching. Two polishing
procedures followed deposition of the bottom YBCO electrode and insulator trilayer
patterning. Detailed information about the fabrication process can be found in
reference [8].
Table 13.1 ulf-MR system coil characteristics

Number of turns/
Coil description/Field Dimensions Inductance Field strength
Cancellation coil x-axis/BCX 2.93 m 2.93 m 20/10 mH 10 T/A
Cancellation coil y-axis/BCY 2.93 m 2.37 m 20/10 mH 10 T/A
Cancellation coil z-axis/BCZ 2.93 m 2.37 m 20/10 mH 10 T/A
Measurement coil/B0 R D 0.8 m 100/80 mH 120 T/A
Prepolarization coil/BP Inner R D 0.012 m 1,600/30 mH 20 mT/A
Outer R D 0.036 m
Height h D 0.078 m
The planar SQUID magnetometers as well as the SQUIDs for the flip-chip mag-
netometers were fabricated via PLD of a single YBCO layer that was patterned with
UV photolithography and ArC ion-beam etching. Planar SQUID magnetometers
(see Fig. 13.1) were fabricated on 10 mm 10 mm bicrystal STO substrates and
had pickup loop of 8 mm 8 mm directly coupled to the hairpin SQUID [4]. The
flip-chip SQUIDs were made on 5 mm 5 mm bicrystal STO substrates and had
1.2 mm 1.2 mm square washers matched to the dimensions of the input coil. The
chips of the flux transformers and square-washer SQUIDs were pressed together
to form flip-chip magnetometers. The gap between two chips was estimated to be
3 m with an in-plane misalignment of less than 2 m.
13.2.2 Magnetometer Characterization
Both planar and flip-chip SQUID magnetometers were calibrated using a Helmholtz
coil. The flux-to-field transformation coefficient Aeff 1 was estimated at 5.3 nT/0
for the planar and 1.9 nT/0 for the flip-chip SQUID sensors. The estimated
coupling coefficient between the input coil of the flux transformer and the SQUID
washer of the flip-chip magnetometer was 0.51.
An SR780 Dynamic signal analyzer (Stanford Research Systems Inc.) was used
to investigate the noise characteristics of the SQUID magnetometers. We used
Magnicon SEL-1 SQUID electronics to control the bias current (AC- and DC- bias
modes) and operate the SQUIDs in a flux-locked loop [9]. The sensors were housed
in a superconducting shield during cooling and measurements.
13.2.3 Ultra-Low Field NMR System
We have developed an ulf-MR system based on our high-TC SQUID technology [5,
6]. The characteristics of the coils we used for the ulf-NMR sensor benchmarking
experiments are listed in Table 13.1. The NMR setup is shown in Fig. 13.2.
Fig. 13.2 The NMR setup.

The measurement (B0 ) field is
in the z-direction. The 10-ml
water sample was surrounded
by the sensing coil of the
copper flux transformer/
resonator, both of which were
placed inside the bore of the
prepolarization (BP ) solenoid
coil. After prepolarization (in
the x-direction), switches
connecting the sensing coil to
a coupling coil were closed.
The coupling coil inductively
coupling the MR signal to the
SQUID magnetometers
(flip-chip magnetometer
pictured) inside the 77 K
liquid nitrogen bath
The earth-field cancellation system consisted of three orthogonal Helmholtz-like

rectangular coils. Each coil was driven by a DC power supply (BK Precision 1745A)
and low-pass filtered at 2 Hz. These coils were attached along the inner walls of
the RF shielded room that housed the ulf-MR setup. The measurement (B0 ) field
coil consisted of a large Helmholtz pair (R D 0.8 m) that produced a homogenous
measurement field over the sample volume. The current for the measurement coil
was also supplied by DC power source (HP 6030A) and low-pass filtered at 1 Hz.
A 1,600-turn solenoid coil produced the prepolarization (BP ) field.
In order to protect the SQUID sensor during MR pulse sequences, we constructed
a copper flux transformer that inductively coupled the MR sample to the sensor [10].
The transformer had a 400-turn sensing coil ( 0.5 mm copper wire) that fit inside
the BP solenoid. The coil had an inner diameter of 22 mm and a length of 60 mm.
The sensing coil was connected in series to a 200-turn coupling coil ( 0.15 mm
copper wire). The total inductance of the copper flux transformer was 1 mH that,
when connected to a 1.47 F capacitor, had a resonant frequency of 4,150 Hz with
Q-factor of about 8. A computer-controlled switch circuit was used to disconnect
the transformer line from the coupling coil during prepolarization pulse. A 10-ml
water sample was placed inside the sensing coil as shown in Fig. 13.2.
We employed a simple free-induction decay (FID) ulf-MR pulse sequence for
comparing the SNR of the planar and flip-chip SQUID magnetometers. After
switch-off the prepolarization pulse, the proton magnetization starts to precess in the
x-y plane, i.e. around the measurement field B0 applied in z-direction, decaying with
spin-lattice relaxation time T1 [2]. The SNR of this FID signal is the benchmark
figure of merit for comparison of our sensors.
13.2.4 Benchmarking Activities
To consistently benchmark the sensor technologies, we made no effort to select

the best of either. Upon selecting the given flip-chip and planar SQUID sensors,
we performed NMR measurements with both the same day and a few days later
recorded noise spectra with the same sensors (again on both, the same day).
The measurement setups were furthermore kept identical for the two technologies
(similar geometries simplified this aspect).
13.3 Results
Noise spectra of the flip-chip and planar SQUID sensors are presented in Fig. 13.3.
The planar SQUID demonstrated a magnetic field sensitivity of 35 and 25 fT/Hz1/2
at 10 Hz and 4 kHz, respectively. The flip-chip SQUID magnetometer had a higher
low-frequency noise but a better white-noise level of 15 fT/Hz1/2 (e.g. at 4 kHz). We
presume that the excess 1/f noise was caused by a-oriented grains in the top YBCO
electrode [8]. The sensors characterized here were employed in the benchmarking
NMR experiment below.
Fig. 13.3 Noise spectra of the bicrystal planar SQUID magnetometer (dotted line) and the
flip-chip SQUID magnetometer (solid). The planar SQUID demonstrates low 1/f noise at low
frequencies with AC bias. The flip-chip SQUID shows a better white noise level when compared
to the planar SQUID, but suffered from 1/f noise that was not improved in AC bias mode. The
dash-dot line indicates the 4 kHz frequency at which NMR was performed and the flip-chip
magnetometer outperforms the planar sensor
Fig. 13.4 (a) A single-shot

NMR peak of tap water
obtained with a single-layer
SQUID sensor. B0 was 90 T
and BP was 50 mT. (b)
Average of 50 NMR peaks
directly comparing the SNR
of our flip-chip (solid) and
planar (dotted line) SQUID
magnetometers
(B0 D 102 T, BP D 18 mT).
In this case, the SNR for the
flip-chip sensor was 7 and
that of the planar was 3. The
lower overall SNR in these
spectra is likely due to
fluctuations in the
measurement field throughout
the measurement time and
limitations in the
experimental hardware.
However, the lower noise
level and higher NMR peak
magnitude with the flip-chip
sensor demonstrates the
viability of the approach: the
flip-chip sensor enables a
gain in SNR of 2C over the
equivalent planar SQUID
Magnetic field sensitivity of the full sensor setup (flip-chip SQUID C copper flux
transformer) was also estimated. The level of the magnetic field noise increased
when we added copper transformer and was measured about 80 fT/Hz1/2 at
resonance frequency. The reason of such high noise level is the large resistance of
the copper coils of the transformer and, possibly, it can be reduced by cryocooling
of the pick-up coil of the copper transformer.
Figure 13.4a presents a single-shot water NMR peak obtained with a planar
SQUID sensor under optimized system parameters. The peak yielded an SNR of
90 and was obtained with a measurement (B0 ) field of 90 T and a prepolarization
pulse of 50 mT. As a proof-of-principle demonstration, we performed direct
benchmarking in which the flip-chip SQUID magnetometer was compared to a
planar one. This experiment was performed with a measurement field of 102 T
and prepolarization pulse of 18 mT. We present the average of 50 NMR peaks
obtained under these experimental conditions with both the flip-chip and planar
SQUID sensors in Fig. 13.4b.
The overall SNRs in the benchmarking experiment presented in Fig. 13.4b
were lower than that of Fig. 13.4a for several reasons. First and foremost, the
prepolarization pulse was lower in magnitude because the prepolarization coil was
modified for the benchmarking experiments. Furthermore, low-frequency magnetic-

field fluctuations affected the NMR frequency over the course of tens of seconds.
We therefore did not use single-shot recordings for the benchmarking because, for
example, it is difficult to be certain the fluctuations in the magnetic field would be the
same for the two conditions under such a short time window (5 s per shot). Only
by averaging many periods of the fluctuating field (5 min total measurement time)
can we ensure the fluctuations are averaged out in the same way for both sensor
technologies. Nevertheless, the measurements demonstrate a more than twofold
improvement in the SNR for the flip-chip sensor.
13.4 Conclusion
We have demonstrated improved SNR ratios in NMR recordings of water samples

with our multilayer-based flip-chip high-TC SQUID sensors. The developed mul-
tilayer flux transformer consisted of a square spiral input coil and pickup coil of
9 mm 9 mm. When coupled to a bicrystal SQUID, we achieved a magnetic field
sensitivity of 15 fT/Hz1/2 at 4 kHz.
Of note is the slight difference between geometries pickup coils in the flip-
chip (9 mm 9 mm) and planar (8 mm 8 mm) magnetometers. Larger area
pickup loops are more easily realized with the flip-chip approach because it enables
independent fabrication of the pickup coil and the junctions for the SQUID sensor.
Flux transformers with much more large pickup loops (up to 50 mm 50 mm)
can be fabricated using developed fabrication process that can further improve the
sensitivity of our flip-chip magnetometers.
The ultra-low-field MR system served as a standard with which we compared
our flip-chip and planar SQUID sensors. Ongoing efforts include increasing the
strength of the pre-polarization field for improved MR signals, developing the
cryocooled gradient coil and pulse sequences needed for extending to 3D imaging,
and enlarging the imaging space so that imaging can be performed on human
subjects.
These MR experiments demonstrate the improvement in the SNR when using a
flip-chip as compared directly to a planar SQUID magnetometer. The demonstrated
gain in SNR indicates the new multilayer structures are a promising technology for
high-TC SQUID-based ulf-MR systems.
Acknowledgments We thank the European Union for support of the FP7 project MEGMRI,
contract number 200859, the Swedish Research Council, Knut and Alice Wallenberg foundation,
the European Union via Tillvxtverket and Regionala Utvecklingsfonden for MedTech West, and
Kristina Stenborgs stiftelse for financial support. Authors from Russian Federation thank the
Russian Ministry of Education and Science for financial support under grant 14.513.11.0009.
References
1. Busch S, Hatridge M, Mssle M, Myers W, Wong T, Muck M et al (2012) Measurements of

T1 -relaxation in ex vivo prostate tissue at 132 T. Magn Reson Med 67:11381145
2. Clarke J, Hatridge M, Mssle M (2007) SQUID-detected magnetic resonance imaging in
microtesla fields. Annu Rev Biomed Eng 9:389413
3. Chen H-H, Yang H-C, Horng H-E, Liao S-H, Yueh S, Wang L-M (2011) A compact
SQUID-detected magnetic resonance imaging system under microtesla fields in a magnetically
unshielded environment. J Appl Phys 110:093903
4. isjen F, Schneiderman JF, Figueras GA, Chukharkin ML, Kalabukhov A, Hedstrm A et al
(2012) High-Tc superconducting quantum interference device recordings of spontaneous brain
activity: towards high-Tc magnetoencephalography. Appl Phys Lett 100:132601
5. isjen F (2011) High-TC SQUIDs for biomedical applications: immunoassays, MEG, and
ULF-MRI. PhD dissertation, Chalmers University of Technology, Gteborg, unpublished
6. Jnsson M (2011) Ultra-low field nuclear magnetic resonance using high-TC SQUIDs.
Masters thesis, Chalmers University of Technology, Gteborg, unpublished
7. Ludwig F, Dantsker E, Koelle D, Kleiner R, Miklich AH, Clarke J (1995) Multilayer
magnetometers based on high-TC SQUIDs. Appl Supercond 3:383398
8. Chukharkin ML, Kalabukhov A, isjen F, Schneiderman JF, Snigirev O, Winkler D (2012)
Noise properties of high-TC superconducting flux transformers fabricated using chemical-
mechanical polishing. Appl Phys Lett 101:042602
9. Drung D (2003) High-TC and low-TC dc SQUID electronics. Supercond Sci Technol 6:
13201336
10. Enpuku K, Hirakawa S, Momotomi R, Matsuo M, Yoshida T (2011) Performance of HTS
SQUID using resonant coupling of cooled Cu pickup coil. Physica C 471:12341237
Chapter 14
Passive Sub-THz Imaging
Alexei Vertiy and Andrei Pavlyuchenko
Abstract A prototype of a multipurpose scanning radiometric imaging system

of extremely high frequency range of 97104 GHz with sensitivity of T D
0:0031 K. D 1s/ for 1 ms 100 ms is presented in the work. Investigations of
transmission factors of different materials are carried out for through-wall detection
of different objects as well as for detection of weapons concealed under the clothing.
Results of radiometric measurements of objects located on the sea surface at
distances in between 10 m and 2.5 km are also given.
14.1 Introduction
Passive millimeter-wave imaging systems are very promising for detection of

dangerous objects (weapons, explosives, drugs, etc.) concealed under persons
clothing, without revealing itself. Because of the terroristic threats in the last decade,
the high-resolution radiometric systems for fast imaging at short distances are
essential for personal weapon control in public places. Owing to ability to penetrate
poor weather for surveillance in the long-range regime, the passive radiometric
imaging holds great promise as a means to aid aviation in low visibility conditions.
Millimeter-wave imaging systems have shortages of low speed of operation and
small spatial resolution in comparison with optical systems [1, 2]. The first problem
is usually tried to be overcame by using multi channel scheme of the imaging sensor
module. Such array of sensors is capable of scanning 2-D scenes in real time.
A. Vertiy
International Laboratory for High Technology, TUBITAK-MRC, MI, Gebze-Kocaeli, Turkey
A. Pavlyuchenko ()
Department of Industry and Enterprise Development of Kyiv local state administration, State
Research Center Iceberg, Av. L. Kurbasa, 2B, 03148 Kiev, Ukraine
e-mail: andy_pavluchenko@ukr.net
162 A. Vertiy and A. Pavlyuchenko
Unfortunately this high cost design has disadvantages of the channel identity and
stability. Spatial resolution constrained by the Rayleigh criterion is determined first
of all by the receiver sensitivity and image reproduction quality of antenna. Low-
noise amplifiers and low aberration antenna systems are intended for improving the
spatial resolution.
In the work a prototype of a passive scanning radiometric system operating in
frequency range of 97104 GHz and employing high sensitive receiver created on
super-heterodyne scheme is presented. Polyethylene lens antenna with the aperture
size 500 mm was used for carrying out radiometric experiments for distances
to objects from 10 m to 2.5 km. Investigations of reflection factors of different
construction materials are carried out for through-wall detection of different objects
as well as for detection of weapons concealed under persons clothing. The receiver
sensitivity achieved allowed us to conduct radiometric scanning of low contrast
objects such as ships at different elevation angles for distances from 1 to 2.5 km.
The aim of the work was investigation of potential possibilities of the system and
areas of applications of this high sensitive radiometric system.
14.2 Experimental Results
A modified 2-D scanning system presented in Fig. 14.1 was used as a tool for
conducting radiometric experiments and receiving images of different objects. The
scanning system consists of a focusing lens [3, 4] manufactured from polyethylene,
radiometric receiver with a horn antenna, two-coordinate device for mechanical
Fig. 14.1 A photo of 2-D

passive radiometric scanning
system
14 Passive Sub-THz Imaging 163
Fig. 14.2 The photo front-end part of radiometric receiver
moving of the receiver [5], analog-to-digital converter and PC for data storage,
visualization of the information received and system managing.
The superheterodyne radiometric receiver for extremely high frequency range
(3-mm band) was developed and created in ILHT. The radiometric receiver is
created as a waveguide superheterodyne scheme with a low-noise amplifier at
the input. The front-end part of radiometric receiver is presented in Fig. 14.2.
Employment of the low-noise amplifier at the receiver input allows to decrease of
noise factor and to increase its sensitivity [6]. The receiver heterodyne consists of a
master oscillator based on Gunn diode and on low-noise buffer amplifier. A mixer
is made by balance scheme using AsGa Schottky diodes. Conversion losses of the
mixer did not exceed 6.5 dB in working frequency band. Intermediate frequency
section consists of three low-noise amplifiers and a band filter. Amplification factor
of the intermediate frequency section was more than 50 dB and noise factor was
1.1 dB. The detector was done by voltage doubling scheme on Schottky diodes.
Working frequency range of the receiver was from 97 to 104 GHz. The receiver
sensitivity measured and adjusted to 1 s was T D 0:0031 K. The radiometric
receivers are characterized by the amplifier factor from 60 to 120 dB, thus obtaining
of minimum fluctuations of the amplifier factor is a very complicated problem and
actual problem in the radiometry [7]. In the case under consideration this problem
is solved by using double supply voltage stabilization of all the units in combination
with temperature stabilization of the whole receiver. Temperature inside the receiver
is supported at the level of 45 C with accuracy no less than 0.1 C. As a result
of achieving the given temperature regime of the receiver, deviations of the output
fluctuations do not exceed 10 mV (for integration constant value of 100 ms). p
Value of radiometric gain [8] obtaining as the following formula q D f ,
shows a signal which q times smaller than intrinsic noises of the receiver and it
may be registered by radiometer. For achieving different values of the radiometric
gain when conducting experiments, different filters with D 1 ms, D 10 ms
D 100 ms were manufactured. The outputs of receiver signals for filters with
Fig. 14.3 Output of receiver signals: (a) D 1 ms; (b) D 10 ms; (c) D 100 ms
different integration time are presented in Fig. 14.3. Coaxial input and output of
the filters gave the opportunity to change the value of the radiometric gain of the
receiver q when carrying out the experiments during very short time.
It is known [9], that when energy with power of P incidents on an arbitrary
material than part of this power P is reflected, the other part of the power P is
absorbed, and the third part of the incident power P passes through the material.
Reflection factor, absorption factor and transmission coefficient are connected with
each other by the following equation:
P P P
CCD C C D 1;
P P P
and usually 0, so C 1. In literature you may find that 0, but

practically 0, thus, part of the power passes through the material and this part
may be registered. In his work [10] the author gives only results of investigations of
the reflection factor of some materials at frequencies of 94 GHz.
In our work investigations of reflection factor for different construction materials
were carried out in the first series of experiments. For our experiments we used the
high sensitive radiometric scanning system, and irradiation of the cold sky which
was supplied by a metal plate located at the distance of 10 m from the receiving lens
at the angle of 45 . This irradiation from the cold sky was employed as non-coherent
illuminating radiation. The materials under investigation were located at the distance
of 2 m from the lens and they totally overlapped its aperture. The results of the
experimental investigations of the transmission factors of different construction
materials for the frequency range of 97104 GHz are given in Table 14.1.
In the next series of experiments a pattern of letter A made from absorbing
material was fixed on the metal reflector retranslating irradiation of the cold sky.
Scanning of the object, letter A, was conducted in two regimes: the first regime
with using of different construction materials placed on the way of radiometric
signal propagation and the other regime without such materials. Photo and the
radiometric image obtained for letter A when scanning without obstacles and
through plasterboard are presented in Fig. 14.4, correspondingly.
The experiments carried out showed us possibility of obtaining radiometric
images of objects located behind obstacles if transmission coefficient of materials no
less than 0.005. One should notice good quality of the obtained radiometric images
for contrast objects located at distances from 10 to 150 m.
Table 14.1 The transmission

factor for different Type of materials Transmission factor,
construction materials Concrete 0.000174
Brick 0.000133
Marble 0.000522
Gypsum 0.00148
Pine (dry) 0.000601
Plasterboard 0.0365
Fiberglass 0.003
Cellular plastic 0.00458
Beech galley 0.000893
Window pane 0.08125
Plywood 0.0458
Paper board 0.151
Polyethylene (film) 0.1717
Board-on-edge floor 0.0032
Fig. 14.4 Radiometric scanning at distance of 10 m: () photo of the scanning system and photo
of letter ; (b) radiometric image of an object without wall; () radiometric image of the object
behind the wall made from plasterboard
The paper [11] showed a possibility of detection of weapon and other different
dangerous objects concealed under persons clothing. However, it should be noticed
that in the most of the experiments described light summer clothing with reflection
factor less than 0.2 was used. In our experiments we investigated possibilities
of detection of weapons and other objects concealed under the winter clothing
(clothing with high reflection factor 0:4), and the person was behind the wall
( 0:005) at distance of 10 m. One of the results obtained is presented in Fig. 14.5.
Radiometric image of dangerous objects (weapons) with reflection factor of 1
has sufficient dependence on the angle of view. For one-to-one detection of objects
concealed under the clothing, it is necessary to obtain scanned images at different
angles. It may be illustrated by Fig. 14.5. In Fig. 14.5c a gun concealed at chest of
a person has a smaller radiometric contrast than when scanning it at angle of 45
(Fig. 14.5d).
During the experiments, we changed value of radiometric gain and tried to
find such its value when concealed weapon and other objects may be detected at
Fig. 14.5 Photo, radiometric image and levels of the output signal when scanning weapons
concealed under persons clothing: (, c) frontal scanning; (b, d) scanning at random angle. The
outline of the gun is marked by the dotted line
Fig. 14.6 Photo and radiometric image when scanning weapon concealed under clothing at
distance of 10 m, building at distance of 150 m and clouds at the background. The outline of
the gun is marked by the dotted line
distances of 10 m and also objects located at distances of 150 m at the same scene
and clouds in the sky. As it was mentioned above, the lens antenna system was
used in the scanning system that did not allow obtaining in-focus image of objects
located at different distances. Results of the experiments are given in Fig. 14.6. At
the foreground of the radiometric image one can see a gun concealed under winter
clothing at distance of 10 m, at the right behind there is a building at distance of
150 m, and at the background at the top there are floating clouds.
Landscape where the experiments were carried out allowed to scan objects at the
sea surface and to estimate maximum operational range of the system. Maps of this
area, its photo and radiometric image are represented in Fig. 14.7. In the images
one can see low-contrast sea targets such as ships located at distance of 1,000 m
(Fig. 14.7a, b), at distance of 2,500 m (Fig. 14.7c, d) and also an opposite side of a
bay located at distance of 10 km.
Fig. 14.7 Radiometric measurements at sea: (a) photo and (b) ship image at distance of 1,000 m;
(c) photo and (d) ship image at distance of 2,500 m; (e, f) maps of scanning area for (a) and (c)
cases correspondingly. The outline of the ship is marked by the dotted line
Difference in radiometric images in Fig. 14.7b, d is image inversion. It may be

explained by different elevation angles for scanning. In the first case, Fig. 14.7b, the
scanning system was fixed at the level of 10 m above the sea level, and in the second
case, Fig. 14.7d, the level was approximately 300 m above the sea level.
14.3 Outlook
In the work the possibility of creating of multipurpose scanning radiometric

system operating in frequency range of 97104 GHz with sensitivity of T D
0:0031 K . D 1 s/ for 1 ms 100 ms was demonstrated. The system
provides obtaining of radiometric images as for weapon concealed under persons
clothing at distance of 10 m as well as low-contrast sea targets at distance of
2,500 m. Investigations of reflection factors of different construction materials were
conducted, possibility of detecting of objects and people behind walls as well as
weapon concealed under persons clothing behind walls at distance of 10 m were
showed.
References
1. Pergande A (2007) New steps for passive millimeter imaging. In: Proceedings of SPIE in
passive millimeter-wave imaging technology X, Orlando, vol 6548, pp 654802-1654802-4
2. Born M, Wolf E (1999) Principles of optics. Cambridge University Press, Cambridge
3. Vertiy A, Tekbas M, Kizilhan A, Panin S, Ozbek S (2010) Sub-terahertz radiometric imaging
system for concealed weapon detection. In: PIERS-2010, Cambridge
4. Vertiy A, Ozbek S, Pavlyuchenko A, Panin S, Tekbas M, Kizilhan A, Cetinkaya H, Unal A
(2011) Short- and long-range passive imaging in millimeter-wave-band. In: URSI GASS,
Istanbul
5. Vertiy A, Ozbek S, Pavlyuchenko A, Panin S, Tekbas M, Kizilhan A, Cetinkaya H, Unal A
(2011) Passive radiometric imaging systems in millimeter wavelength range. In: 2011 IEEE
AP-S/URSI, Washington
6. Turk AS, Hocaogly AK, Vertiy AA (2011) Subsurface sensing. Wiley, Hoboken, p 890
7. Skou N (1986) Microwave radiometer systems: design and analysis. Artech House, Boston,
p 222
8. Korolkov ADV (1973) Radioteleskopy and radiometry. Nauka, Moscow
9. Finckelshtein MI (1973) Theory of radiolocation. Sovereign Radio, Moscow
10. Appleby R (2003) Mechanically scanned real time passive millimeter wave imaging at 94 GHz.
Proc SPIE 5077:16
11. Appleby R, Anderton N (2007) Millimeter-wave and submillimeter-wave imaging for security
and surveillance. Proc IEEE 95(8):16831690

(NATO Science for Peace and Security Series B_ Physics and Biophysics) Taras N. Rudakov (Auth.), Tomaž Apih, Bulat Rameev, Georgy Mozzhukhin, Jamie Barras (Eds.)-Magnetic Resonance Detection of Explos

Uploaded by

Copyright:

Available Formats

You might also like

(NATO Science for Peace and Security Series B_ Physics and Biophysics) Taras N. Rudakov (Auth.), Tomaž Apih, Bulat Rameev, Georgy Mozzhukhin, Jamie Barras (Eds.)-Magnetic Resonance Detection of Explos

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(NATO Science for Peace and Security Series B_ Physics and Biophysics) Taras N. Rudakov (Auth.), Tomaž Apih, Bulat Rameev, Georgy Mozzhukhin, Jamie Barras (Eds.)-Magnetic Resonance Detection of Explos

Uploaded by

Copyright:

Available Formats

NATO Science for Peace and Security Series - B:

Physics and Biophysics

A. Chemistry and Biology Springer

Series B: Physics and Biophysics

Library of Congress Control Number: 2013949690

ISBN 978-94-007-7267-0 (PB)

Printed on acid-free paper

All Rights Reserved

In their preface to a previous volume in this series (Explosive Detection using

So how have magnetic resonance methods progressed towards meeting the

London, UK Jamie Barras

Part I Nuclear Quadrupole Resonance Detection of Solids

1 Further Improvement of NQR Technique for Detection

Part II Nuclear Magnetic Resonance Detection of Liquids

8 Bottled Liquid Scanner for Security Checkpoints . . . . . . . . . . . . . . . . . . . . . . 89

9 MagViz: A Bottled Liquids Scanner Using Ultra-Low

Part III Other Techniques

13 Novel HTS DC Squid Solutions for NMR Applications . . . . . . . . . . . . . . . . 151

One of the hardest to detect threats at todays checkpoints is explosives, particularly

T.N. Rudakov ()

The technology of NQR detection is very similar to that normally used in

1.2 Short Time Fourier Transform (STFT) Method

Fig. 1.1 FFT and STFT comparison

The second dominant source of false alarm is battery powered electronics

Fig. 1.3 Comparison

1.3 Pulse Sequence Optimisation for Detection of RDX

Table 1.1 Some parameters of different RDX explosive materials

1.4 NQR Baggage Scanners: Current and Future Design

In a scanning application, a NQR scanner irradiates an item of interest positioned

Fig. 1.4 Functional block-diagram of a NQR based scanner

Fig. 1.5 NQR baggage scanner T-3-03

Implement automatic calibration and self-test capabilities.

Fig. 1.6 Block diagram of proposed new NQR baggage scanner

1.5 Coil Design Optimisation

1.6 Probe Tuning System

Enabling fast, real-time, low-level signal processing of the raw acquisitions as

7. Rudakov TN (1999) Magnetic resonance: nuclear quadrupole resonance, instrumentation. In:

Naveed R. Butt, Erik Gudmundson, and Andreas Jakobsson

Abstract Nuclear quadrupole resonance (NQR) is a solid-state radio frequency

Nuclear quadrupole resonance (NQR) is a solid-state radio frequency (RF) spectro-

2.2 Signal Models and Data Acquisition

Fig. 2.1 Illustration of a PSL sequence

Fig. 2.2 Illustration of the real part of a typical echo train

of d exponentially damped sinusoids. In [11], the authors proposed the following

where m D 0; : : : ; M 1 is the echo number; t D t0 ; : : : ; tN 1 denotes the sampling

Interference and noise

where t D t0 ; : : : ; tN 1 . As it is not possible to acquire the signal when shooting

Fig. 2.4 ROC curves 1

Fig. 2.5 ROC curves 1

Fig. 2.6 ROC curves 1

Fig. 2.9 Probability 1

2.4 Interference Rejection

based on these principles have shown extremely good interference cancellation

in estimations of the subspaces, but they can alternatively be viewed as uncertainties

2.5 Multi-channel Detectors

Using multiple antennas for efficient interference rejection in NQR-based detection

2.6 Concluding Remarks

In this chapter, we have discussed recent advances in the detection, classification,

Pentaerythritol tetranitrate (PETN) C(CH2 -O-NO2 )4 is a well known explosive

P1 .PhP1/ P2 .PhP2/ 2n

E1 D A .1 C / ; E2 D A .1 / ; and E2 D 2A; (4.2)

1 d z " dy " " "