Professional Documents
Culture Documents
(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) 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) Taras N. Rudakov (Auth.), Tomaž Apih, Bulat Rameev, Georgy Mozzhukhin, Jamie Barras (Eds.)-Magnetic Resonance Detection of Explos
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
http://www.nato.int/science
http://www.springer.com
http://www.iospress.nl
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
Published by Springer,
P.O. Box 17, 3300 AA Dordrecht, The Netherlands.
www.springer.com
v
vi Preface
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.
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
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
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 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
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.
1 Further Improvement of NQR Technique for Detection of Illicit Substances 7
Note that additional research and more extensive testing may be required for
evaluating and developing STFT processing technique.
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
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
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.
1 Further Improvement of NQR Technique for Detection of Illicit Substances 13
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.
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
1 Further Improvement of NQR Technique for Detection of Illicit Substances 15
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
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
1 Further Improvement of NQR Technique for Detection of Illicit Substances 17
2.1 Introduction
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
T. Apih et al. (eds.), Magnetic Resonance Detection of Explosives and Illicit Materials, 19
NATO Science for Peace and Security Series B: Physics and Biophysics,
DOI 10.1007/978-94-007-7265-6__2, Springer ScienceCBusiness Media Dordrecht 2014
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.
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
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.
X
d
ym .t / D k e k .tCm/ e k jttsp jCi!k .T /t C wm .t /; (2.2)
kD1
x 1010
10
8
NQR frequencies
7
6
Magnitude
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
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
2 An Overview of NQR Signal Detection Algorithms 25
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
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
Probability of false alarm
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.
Probability of detection
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
Probability of false alarm
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
2 An Overview of NQR Signal Detection Algorithms 27
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.
Probability of detection
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].
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
2 An Overview of NQR Signal Detection Algorithms 29
0.9
0.8
0.7
Probability of detection
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
Probability of false alarm
Fig. 2.10 ROC curves for the RESPEQ, SEAQUER, and REMIQS algorithms, for SNR = 20 dB,
ISR = 40 dB
100
90
80
Probability of detection (%)
70
60
50
ESPIRE
40
NLS
30 FSMC
M3L
20 M3L+5
ALS
10
ALS+5
0
0 20 40 60 80 100
Uncertainty level (%),
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
2 An Overview of NQR Signal Detection Algorithms 31
100
90 ESPIRE
NLS
80 FSMC
M3 L
Probability of detection (%)
70 M3L+5
ALS
60 ALS+5
50
40
30
20
10
0
0 20 40 60 80 100
Uncertainty level (%),
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.
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
2 An Overview of NQR Signal Detection Algorithms 33
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
T. Apih et al. (eds.), Magnetic Resonance Detection of Explosives and Illicit Materials, 35
NATO Science for Peace and Security Series B: Physics and Biophysics,
DOI 10.1007/978-94-007-7265-6__3, Springer ScienceCBusiness Media Dordrecht 2014
36 M.D. Rowe et al.
3.2 Experimental
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 .
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.
3 Nuclear Quadrupole Resonance of Pentaerythritol Tetranitrate (PETN). . . 39
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
3 Nuclear Quadrupole Resonance of Pentaerythritol Tetranitrate (PETN). . . 41
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.
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.
3 Nuclear Quadrupole Resonance of Pentaerythritol Tetranitrate (PETN). . . 43
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
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.
T. Apih et al. (eds.), Magnetic Resonance Detection of Explosives and Illicit Materials, 45
NATO Science for Peace and Security Series B: Physics and Biophysics,
DOI 10.1007/978-94-007-7265-6__4, Springer ScienceCBusiness Media Dordrecht 2014
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
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]:
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)
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
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 Cross-Relaxation Enhanced NQR of Ammonium Nitrate in Low Magnetic Field 51
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
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.
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
4 Cross-Relaxation Enhanced NQR of Ammonium Nitrate in Low Magnetic Field 53
1.8
1.6
Relative increase of amplitude
2
1.4
1.2
1.0
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
1.0
0.8
Amplitude (a.u.)
0.6
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
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].
56 G.V. Mozzhukhin et al.
1.0
0.8
Amplitude (a.u.)
0.6
0.4
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.
4 Cross-Relaxation Enhanced NQR of Ammonium Nitrate in Low Magnetic Field 57
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
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
4 Cross-Relaxation Enhanced NQR of Ammonium Nitrate in Low Magnetic Field 59
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
5.1 Introduction
T. Apih et al. (eds.), Magnetic Resonance Detection of Explosives and Illicit Materials, 61
NATO Science for Peace and Security Series B: Physics and Biophysics,
DOI 10.1007/978-94-007-7265-6__5, Springer ScienceCBusiness Media Dordrecht 2014
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 .
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,
64 M.W. Malone and K.L. Sauer
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
5 Investigating Homonuclear Broadening in NQR with Carr-Purcell. . . 65
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
66 M.W. Malone and K.L. Sauer
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
References
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
T. Apih et al. (eds.), Magnetic Resonance Detection of Explosives and Illicit Materials, 69
NATO Science for Peace and Security Series B: Physics and Biophysics,
DOI 10.1007/978-94-007-7265-6__6, Springer ScienceCBusiness Media Dordrecht 2014
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.
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
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).
6 Size Effect in 14 N Nuclear Quadrupole Resonance Spectroscopy 73
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
74 N. Sinyavsky et al.
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
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
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
76 N. Sinyavsky et al.
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
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
Mthodologie RMN (CRM2; UMR 7036, UL-CNRS), Universit de Lorraine, Campus
Aiguillettes, B.P. 70239, 54506 Vanduvre-ls-Nancy (cedex), France
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 ()
Mthodologie RMN (CRM2; UMR 7036, UL-CNRS), Universit de Lorraine, Campus
Aiguillettes, B.P. 70239, 54506 Vanduvre-ls-Nancy (cedex), France
e-mail: daniel.canet@univ-lorraine.fr
T. Apih et al. (eds.), Magnetic Resonance Detection of Explosives and Illicit Materials, 77
NATO Science for Peace and Security Series B: Physics and Biophysics,
DOI 10.1007/978-94-007-7265-6__7, Springer ScienceCBusiness Media Dordrecht 2014
78 J. Jover et al.
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
(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.
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
Spruce
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
7 NQR Detection of Sodium Nitrite Recrystallized in Wood 81
Beech
131072 scans
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.
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
7 NQR Detection of Sodium Nitrite Recrystallized in Wood 83
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
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)
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
T. Apih et al. (eds.), Magnetic Resonance Detection of Explosives and Illicit Materials, 89
NATO Science for Peace and Security Series B: Physics and Biophysics,
DOI 10.1007/978-94-007-7265-6__8, Springer ScienceCBusiness Media Dordrecht 2014
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.
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.
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.
Magnitude
0.6 Water
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.
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.
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
Declaration:
Data Fail to Pass
Processing REPORT ALARM
Automated Scans
Element A
Declaration:
Pass
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.
8 Bottled Liquid Scanner for Security Checkpoints 95
Cardboard box
Results shown in Fig. 8.3
Dark plastic
Colored glass
Clear plastic
Liquid
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
Wine - 100 ml NM
Black plastic - 30 ml NM
Lens solution - 10 fl oz
Black plastic - 10 fl oz
Black plastic - AN
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
25
Discriminator (au) 20
15
10
Non hazardous
5
0
Shampoo
Creams
Rhum
Wine
Perfume
Hydrogen Peroxide
HP & 0.3g/ml coffee
Baby milk
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.
8.7 Conclusions
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
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
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
T. Apih et al. (eds.), Magnetic Resonance Detection of Explosives and Illicit Materials, 99
NATO Science for Peace and Security Series B: Physics and Biophysics,
DOI 10.1007/978-94-007-7265-6__9, Springer ScienceCBusiness Media Dordrecht 2014
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
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 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.
9 MagViz: A Bottled Liquids Scanner Using Ultra-Low Field. . . 103
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.
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).
104 R. Austin et al.
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.
9 MagViz: A Bottled Liquids Scanner Using Ultra-Low Field. . . 105
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
106 R. Austin et al.
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.
9 MagViz: A Bottled Liquids Scanner Using Ultra-Low Field. . . 107
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
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
108 R. Austin et al.
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).
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.
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.
110 R. Austin et al.
References
T. Apih et al. (eds.), Magnetic Resonance Detection of Explosives and Illicit Materials, 111
NATO Science for Peace and Security Series B: Physics and Biophysics,
DOI 10.1007/978-94-007-7265-6__10, Springer ScienceCBusiness Media Dordrecht 2014
112 A.B. Konov et al.
10.1 Introduction
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 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.
10 Multiparameter NMR Identification of Liquid Substances 115
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]
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
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,
10 Multiparameter NMR Identification of Liquid Substances 117
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
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
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 %
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
a
Water
1 Coca Cola
Isopropanol
Ethanol
blend-a-med
T2 [s]
1
T1 [s]
b
Peroxide
Water Coca Cola
Head&Shoulders - shampoo
1
D [10-9 m2/s]
Ethanol Isopropanol
blend-a-med
toothpaste
1
T1 [s]
c
Peroxide
Coca Cola Water
Head&Shoulders - shampoo
1 Ethanol
D [10-9 m2/s]
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
10 Multiparameter NMR Identification of Liquid Substances 121
Acknowledgments This work was supported by NATO Science for Peace and Security Pro-
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
122 A.B. Konov et al.
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
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
T. Apih et al. (eds.), Magnetic Resonance Detection of Explosives and Illicit Materials, 123
NATO Science for Peace and Security Series B: Physics and Biophysics,
DOI 10.1007/978-94-007-7265-6__11, Springer ScienceCBusiness Media Dordrecht 2014
124 A. Gradiek et al.
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.
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
126 A. Gradiek et al.
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
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)
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 NMR-Based Liquid Explosives Detector: Advantages and Disadvantages. . . 127
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
128 A. Gradiek et al.
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
130 A. Gradiek et al.
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
11 NMR-Based Liquid Explosives Detector: Advantages and Disadvantages. . . 131
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
11 NMR-Based Liquid Explosives Detector: Advantages and Disadvantages. . . 133
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
11 NMR-Based Liquid Explosives Detector: Advantages and Disadvantages. . . 135
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.
Springer, Dordrecht, pp 89
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
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
T. Apih et al. (eds.), Magnetic Resonance Detection of Explosives and Illicit Materials, 137
NATO Science for Peace and Security Series B: Physics and Biophysics,
DOI 10.1007/978-94-007-7265-6__12, Springer ScienceCBusiness Media Dordrecht 2014
138 G.S. Kupriyanova et al.
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
01
and
01
pulses are determined by the dependence of the three Euler angles on the parameters
1 j
0
and
0
.
1 1
140 G.S. Kupriyanova et al.
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 Composite Pulses in Inhomogeneous Field NMR 141
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.
12 Composite Pulses in Inhomogeneous Field NMR 143
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).
144 G.S. Kupriyanova et al.
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
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
146 G.S. Kupriyanova et al.
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.
Acknowledgments This work was supported by NATO Science for Peace and Security Pro-
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.
Springer, Dordrecht, pp 193203
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
12 Composite Pulses in Inhomogeneous Field NMR 147
T. Apih et al. (eds.), Magnetic Resonance Detection of Explosives and Illicit Materials, 151
NATO Science for Peace and Security Series B: Physics and Biophysics,
DOI 10.1007/978-94-007-7265-6__13, Springer ScienceCBusiness Media Dordrecht 2014
152 M.L. Chukharkin et al.
13.1 Introduction
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
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].
154 M.L. Chukharkin et al.
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.
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.
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.
13 Novel HTS DC Squid Solutions for NMR Applications 155
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
13 Novel HTS DC Squid Solutions for NMR Applications 157
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
158 M.L. Chukharkin et al.
13.4 Conclusion
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.
13 Novel HTS DC Squid Solutions for NMR Applications 159
References
14.1 Introduction
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
T. Apih et al. (eds.), Magnetic Resonance Detection of Explosives and Illicit Materials, 161
NATO Science for Peace and Security Series B: Physics and Biophysics,
DOI 10.1007/978-94-007-7265-6__14, Springer ScienceCBusiness Media Dordrecht 2014
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.
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
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
164 A. Vertiy and A. Pavlyuchenko
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
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
166 A. Vertiy and A. Pavlyuchenko
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.
14 Passive Sub-THz Imaging 167
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
14.3 Outlook
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