Fluorescence lifetime sensor using

optical fiber and optical signal processing

William Heuywon Park

A thesis submitted in confonnity with the requirements for the Degree of

Master of AppIied Science, Graduate Department of Aerospace Science and
Engineering, in the University of Toronto

@Copyright by William Heuywon Park 1998

 Bibliothéque .nationale
1+1 National Library
ofCanada du Canada
Acquisitions and Acquisitions et
Bibliographie Services services bibliographiques
395 Wellington Street 395, rue Wellington
OttawaON K 1 A W Ottawa ON K I A ON4
Canada Canada

The author has granted a non-
exclusive licence alîowing the
National Library of Canada to
reproduce, loan, distnbute or sell
copies of this thesis in microform,
paper or electronic formats.
L'auteur a accordé une licence non
exclusive permettant a la
Bibliothèque nationale du Canada de
reproduire, prêter, distribuer ou
vendre des copies de cette thèse sous
la forme de microfiche/film, de
reproduction sur papier ou sur format
électronique.

The author retains ownership of the L'auteur conserve la propriété du

copyright in this thesis. Neither the
thesis nor substantial extracts from it
may be printed or otherwise
reproduced without the author's
permission.
droit d'auteur qui protège cette thèse.
Ni la thèse ni des extraits substantiels
de celle-ci ne doivent être imprimés
ou autrement reproduits sans son
autorisation.
Abstract
I propose in theory and demonstrate in experiment a fibre optic sensor which
measures fluorescent Iifetime and fiber length using an all optical signal process-
h g scheme. The scheme uses a double modulation homodyne method to produce
auto or cross correlation in fÎequency domain. Because î t ehninates the neeà
for high bandwidth optical detection and electronic processing, it has the p c ~
tentiai of superior performance with less complexity than conventionai time or
frequency domain fluorometers. To demonstrate, fiber length of 50.28 f 0.27m
and fluorescent lifetimes of 28.3 f 3.5ns and 360 f 28ns are rneasured with only
4MHz bandwidth. Furthemore, the fluorescence collection a c i e n c y of multi-
mode fiber is derived for single and double fiber configurations. And the effect
of optical fiber on excitation and fluorescence is analyzed- Aho, possible appli-
cation in the cure monitoring of epoxy resin is addressed.

Fluorescence lifetime sensor using

optical fiber and optical signal processing
by
William Heuywon Park
Master of Appiied Science
Aerospace Science and Engineering
University of Toronto
1998
Acknowledgements

This thesis is dedicated to my family who stood by and endured with me through
Iong iliness and even longer recovery. 1 used to take health and family for
granted; but I found that they are the most important things in the world.
Experiment was performed in Ontario Laser and Lightwave Center at the
University of Toronto, and 1 wish to arknowledge the fnendly people there, Dr.
Ad Hubert, Ed Kammermayer, and Victor Isbrucker. A h , 1 am grateful to Steve
Brown of Chemical Sensor Group in Erindale campus a t U of T for collecting
the fluorescence spectrum and decay.
Finaily, 1 would iike to thank my supervisor, Dr. R M Measures, for giving
me a chance to explore the topics of fibre optic fluorescence sensor which, 1
believe, has many interesting applications in bioengineering and medical fields,
even though it has Little to do with our lab's smart structure program.
Abbreviat ions
A0 acousto-optic
AOD driver for AOM
AOM acousto-optic moduiator
BS beamsplitter
BW bandwidth
DFT Discrete Fourier Transform
EOM electr-p tic moddator
FT Fourier Transfonu
IF iaterrnediate kequency
IOM integrated optic moddator
LO local osciUator
MMF multimode fiber
NA numericaI aperture
OFDR optical frequency dornain reflectometer
OTDR optical tirne dornain reflectometer
QY quantum yield
RF radio fiequency
RU ruthenium tris- bipyridy1 dichloride, hexahydrate
SMF singlemode fiber
SPA N-(3-sulfopropyl) acridinium, inner sait
Contents

1 Introduction 7

2 Multimode Fiber 9
2.1 Mode Propagation in Steady-State . . . . . . . . . . . . . . . . . 9
2.2 Dispersions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.1 Modal Dispersion . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.2 Material Dispersion . . . . . . . . . . . . . . . . . . . . . 11
2.2.3 Waveguide Dispersion . . . . . . . . . . . . . . . . . . . . 11
2.3 Rayleigh Backscattering . . . . . . . . . . . . . . . . . . . . . . . 11
2.4 Fkesnel Backreflection . . . . . . . . . . . . . . . . . . . . . . . . 12
2.5 Fluorescence Collection by Direct Coupling . . . . . . . . . . . . 12
2.5.1 Liouville's Theorem . . . . . . . . . . . . . . . . . . . . . 12
2.5.2 Collection Efficiency . . . . . . . . . . . . . . . . . . . . . 13
2.5.3 Coilection Efficiency of Single Fiber . . . . . . . . . . . . 15
2.5.4 Collection Efficiency of Double Fibers . . . . . . . . . . . 15
2.6 Fluorescence Collection by Evanescent Coupling . . . . . . . . . 16
2.6.1 Evanescent Excitation . . . . . . . . . . . . . . . . . . . . 16
2.6.2 Collection Efficiency . . . . . . . . . . . . . . . . . . . . . 17

3 Fluorescence System 18
3.1 Fluorescent Response . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.2 Determining Lifetime in T i e Domain . . . . . . . . . . . . . . . 19
3.3 Determining Lifetime in Frequency Domain . . . . . . . . . . . . 19
4 Acousto-Optic Modulator 22
4.1 Acoustic Perturbation . . . . . . . . . . . . . . . . . . . . . . . . 22
4.2 DiffractionEfficiency . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.2.1 Raman-Nath AOM . . . . . . . . . . . . . . . . . . . . . . 24
4.2.2 Bragg AOM . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.3 Modulation Bandwidth . . . . . . . . . . . . . . . . . . . . . . . 25

Signal Processing 27
5.1 Signal Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
5.2 Auto and Cross Correlations . . . . . . . . . . . . . . . . . . . . 29
5.3 Time Domain Responses . . . . . . . . . . . . . . . . . . . . . . . 30
5.4 Estimation of Parameters . . . . . . . . . . . . . . . . . . . . . . 34
5.5 Estimation of Error . . . . . . . . . . . . . . . . . . . . . . . . . . 34
6 Experiment 36
6.1 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . 36
6.2 Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
6.2.1 Laser . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
 6.2.2 AOM and AOD . . . . . . . . . . . . . . . . . . . . . . . . 39
6.2.3 MMF . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
6.2.4 SPAandRU . . . . . . . . . . . . . . . . . . . . . . . . . 43
6.2.5 LO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
6.2.6 PIN Detector . . . . . . . . . . . . . . . . . . . . . . . . . 46
6.3 Jhperirnental Procedures . . . . . . . . . . . . . . . . . . . . . . 46
6.4 Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
7 Conclusions 56

A Formula and Tables 58

A.l Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
A.2 Impulse Function . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
X.3 Solid Angles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
A.4 Table of Fourier 'handorrn Pairs . . . . . . . . . . . . . . . . . . 59
A.5 Decomposition into Fourier Series . . . . . . . . . . . . . . . . . . 60
A.6 rmswidth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
A.? Fiber-twFiber Coupling . . . . . . . . . . . . . . . . . . . . . . . 61
B FIuorescence Application in Cure Monitoring 62
B.l Epoxy Resin and Composite . . . . . . . . . . . . . . . . . . . . . 62
B.2 Fluorescence during Cure . . . . . . . . . . . . . . . . . . . . . . 63
B.3 Cure Monitoring ........................... 64

Bibliography 65
List of Figures
Fiber tip with an angled cleave . . . . . . . . . . . . . . . . . . . 12
Excitation and collection of fluorescence using single fiber . . . . 14
Excitation and collection of fluorescence using double fibers . . . 14
Graph of F ( L ) and G(L.s) . . . . . . . . . . . . . . . . . . . . . 16

3-level mode1 of fluorescence system . . . . . . . . . . . . . . . . 18

Survey of fiequency domain fluorometers . . . . . . . . . . . . . . 21

Diffraction at Bragg angle . . . . . . . . .............. 24

Diffraction at non-Bragg angle . . . . . . . . . . . . . . . . . . . 24
Signal flow diagram . . . . . . . . . . . . . . . . . . . . . . . . . . 28
Square-wave excitation and fluorescent response . . . . . . . . . . 28
Simulation of Resnel backreflection fiom 50m fiber . . . . . . . . 31
Simulation of SPA fluorescence ming 50m fiber . . . . . . . . . . 32
Simulation of RU fluorescence using 50m fiber . . . . . . . . . . . 33

Schematic of experimentai setup . . . . . . . . . . . . . . . . . . 37

Pictures of optical (top) and electrical (bottom) equipments . . . 38
Picture of inside AOM . . . . . . . . . . . . . . . . . . . . . . . . 40
Transfer function of AOM/AOD . . . . . . . . . . . . . . . . . . 41
Chernical formula for SPA (left) and RU (nght) . . . . . . . . . . 43
Absorption and emission spectrum for SPA (top) and RU (bottom) 44
Calibration data for LO's voltage-to-frequency conversion . . . . 45
DC-coupled lowpass detector . . . . . . . . . . . . . . . . . . . . 47
c o u & ideteetor . . . . . . . . . . . . . . .
~ h ~ l e r ~ ~ ~ - bandpass 47
6-10 Aligning AOM for first p a s (top) and second pass (bottom) . . . 47
6.11 Pictures of SPA (top) and RU (bottom) fluorescence . . . . . . . 49
6.12 Raw data for Fresnel backreflection from 50m fiber . . . . . . . . 50
6.13 Raw data for SPA fluorescence using 50m fiber . . . . . . . . . . 51
6.14 Raw data for RU fluorescence using 50m fiber . . . . . . . . . . . 52
6.15 Raw data for Fkesnei backreflection h m 49m fiber . . . . . . . . 53
6.16 Fluorescence decay of SPA and Fluorescein . . . . . . . . . . . . 55
6.17 Fluorescence decay of RU and Fluorescein . . . . . . . . . . . . . 55

B.l epoxy resins DGEBA and TGDDM . . . . . . . . . . . . . . . . . 63

B.2 curing agents DDS and DDM . . . . . . . . . . . . . . . . . . . . 63
Chapter 1

Introduction

Many have used optical fiber sensor and fluorescent lifetime to measure such
parameters as temperature, pH, O2 concentration, etc. The optical fiber allows
one to excite and coilect the fluorescent signal from a remote medium that may
otherwise be ciifficuit to access. The lifetime of fluorescence decay is chosen
because the decay mechanism is weU understood, is sensitive to its molecujar
environment, and is mostly independent of the intensity and the wavelength
of both excitation and fluorescence. In combining opticd fiber and fluorescent
lifetime, the sensors represent a significant development not o d y in fluorescence
field, but also in other sensing applications where the parameter of interest can
be measured from an optical signal whose profiie and mechanism are simiiar to
the fluorescence decay.
Unfortunately, al1 the sensor implernentations, so far, suffers hom 3 major
problems:
1. The c o u p h g between the optical fiber and the fluorescence has not been
studied anaiytically. The fluorescent coupiing is important in determining
the amount of fluorescence that is collected by fiber, and in evaluating the
performance and potential of optical fiber sensor.
2. The waveguide effect of optical fiber on the excitation and the fluorescence
signais has not been quantified. Unlike the hee-space which it replaces, the
optical fiber generates Rayleigh scattering, Ftesnel reflection, dispersions,
modal interference, wavelength dependent attenuation, coupling loss, in-
cidental luminescence, etc.
3. Most of signal processing for the lifetime is done electronically in either
time or frequency domain, using expensive components and complicated
arrangements. The main advantage of electronic signal processing is that
amplification, filtering, and other signai conditioning can be done. As the
Metirne gets shorter, it requires faster detection and higher bandwidth
signal processing. This becomes more difficult to accomplish with the
equipments that are cmently anilable.
1 address these problems in my thesis with the emphasis on understanding the
fundamentah, rather than on reciting a list of possible applications or on per-
forming characterization of actual materials.
In my thesis, 1propose in theory and demonstrate in experiment a fibre optic
sensor which measures fluorescent lifetbe and fiber length using an aii optical
signal processing scheme. A continuous laser beam is intensity modulated by an
acousto-optic modulator and is launched into a multimode fiber. At the far-end,
the optical fiber coUects either fluorescent response fiom which to measure the
lifetime or fiesne1 backreflection from which to measure the length. The return-
ing signal is intensity modulateci again by the modulator and is detected using
a lowpass detector. Although the fluorescent lifetime is the main parameter of
interest, the fiber length must be aiso measured because the use of optical fiber
inevitably introduces propagation delay.
The first and second problems are addresseci by deriving the fluorescence
collection efficiency for single and double fiber configurations, and by a n a i m g
the effect of optical fiber on the propagating signais. The results are obtained
rnainly for multimode fiber, but they can be useful for singiemode fiber as well.
To address the thid problem, the signal processing scheme uses a double
optical modulation homodyne method to produce auto or cross correlation in
frequency domain. The length and Metirne parameters are estimated by data fit-
ting auto and cross correlation modei directly in frequency domain. My scheme
eiiminates the need for wide bandwidth opticai detection and high frequency
electronic processing, whde at the same tirne, reducing the component count and
simplifying the experimentd setup. It has the potential of superior performance
with less complexity than conventional time or frequency domain fluorometers.
Chapter 2

Multirnode Fiber
In this chapter, 1 will present some basic expressions which are helpful in under-
standing the workings of multimode fiber (MMF) and occasionally singlemode
fiber (SMF).Consider a MMF with core radius a and index profile

where Ml = ni sin BM = ni = is knom as numericd aperture,

and eEt4is half-angle of acceptance aperture inside the fiber. 2 common types
of fiber are step-index fiber (g = w) in which the core index stays constant nl
and pambolàc-index fiber (g = 2) in which the core index decreases parabolically
btom nl to na.

2.1 Mode Propagation in Steady-State

At steady-state, optical power can be propagated only in guided mode with
kon2 < P < b n l , where B is the propagation constant along the fiber axis.
[Marcuse, Ungar, Cheo,Ghatak]. The total number of guided modes is

N=--= v2/2, step

g+2 2 v214, parabolic
where V = koaM. Since diffuse source, such as LED and fluorescence, wiii
launch equal power into each mode, fibers with equd number of modes wiil
camy equd amount of power. If Pm is the propagation constant of mth-order
mode, then the number of modes with Dm c /3 < konl is

which can be r e h t t e n for the propagation constant as

Since each mode propagates with dEerent velocity v, = &/d&,, the time
that a mode takes to travel distance L is [Cheo,Marcuse, Ghatak, Cowar]
where C = (~,/lV)g/(g+*), and small A is assumed in Eqn 2.4. Ignoring higher
order terms, the difference between the fastest time (( = 0) and the slowest
t h e ('t= 1) is
Ln
a2/2, parabolic (2.6)
= c
One can ais0 describe propagation conditions in terms of geornetnc optics.
By cornparhg Eqn 2.4 and B, = kmlcos& where 8, is the angle between ray
vector and fiber axis, one obtains [Ungar, Barnoski, Cheo]

which explains concisely much of the experimentaily observeci phenomena in

MhfF'. Furthemore, outside the opticd fiber, the far-field angular (8) depen-
dence and the near-field radial (r)dependence of normalized power are (Barnoski]

where is haif-angIe of fiber aperture in air. As expected, the power distri-

butions for stepindex fiber (g = oo) are uniform; that is, P(8) = P ( r ) = 1,

2.2 Dispersions
In this section, 1review only the broadening of rms width (a)due to 3 types of
dispersion associateci with optical fiber: [Cheo,GhatakJ
Modal dispersion ( a l ) is caused by the fact that Merent mode travels at
different speed.
Material dispersion (oz)is caused by the fact that refractive index n(X)
varies wit h wavelength.
Waveguide dispersion (03) is caused by the fact that waveguide parameter
V depends on A.
For MMF, 01 is usuaüy dominant, but 0 2 makes contribution in fluorescence
measurement where source spectral width is large. For SMF,o d y 02 and a3
are applicable. The dispersions are uncorrelated with one another. And if the
fiber is operating iinearly, the rms width of total dispersion becomes [Gowar]

2.2.1 Modal Dispersion

Fortunately for stepindex and parabolic-index MMF, the modai impulse re-
sponse tums out to be a square pulse [Marcuse]; that is, when an impulse exci-
tation is launched into the fiber, the propagating modes are distributed evenly
between the fastest mode and the slowest mode. Then, using Eqns 2.6 and A.53,
the rms width of modal dispersion becomes

01=-=-
step
Since modal coupling tends to reduce pulse broadening in exchange for increased
attenuation, the modal dispersion exhibits square-root dependence with distance
after travelling mtical length of L, = 1/47,. [Cheo, Gowar]

where a, is the rms width at L,,and 7, is steady-state attenuation c oefficient.

2.2.2 Materid Dispersion

The rms width of materiai dispersion is [Cheo, Ghatak]

where n(X) is the core index and as is the spectral rms width of source. An
empirid acpression of n(X) and n"(X) for pure silica are [Ghatak]

where p = X x 106 is wavelength in hm], n''(A) = n"(li)/p2, and

2.2.3 Waveguide Dispersion

Waveguide dispersion is relevant only for SMF. An empirical expression for
step-index SMF is [Ghatak]

2.3 Rayleigh Backscattering

Consider an impulse excitation of unit energy travelling down an optical fiber.
The spatial profile of excitation is sirnply E ( z ) = e-Tz, where 7 is the totai at-
tenuation coefficient due to Rayleigh scattering (r,a and absorption (y,).
+
In [z, z dz] interval, y.E(z)dz is lost to scattering, of which only S-y, E(z)dz
is coilected back into fiber. Assuming Rayleigh scattering is isotropic at l e s t
over the acceptance aperture of fiber, the collection efficiency is given as

S= 3g A, for MMF [Kg, K33,Ungar]

4(9 + 1)
- 3a2
-- for SMF [KlO,K11,K33]
y2,,9 A'

where w is the spot size of Gaussian profile that approximates SMF's fundamen-
ta1 mode field. SpeQfically, S = 3A/4 for stepindex MMF,and S = 3A In V/V2
for stepindex SMF using Eqn 2.36.
S i c e the attenuation is same for backward and forward propagation, the
+
backscatterecl energy arrîving at z = O fkom the interval [t,z &] is dE, =
S7.e-2Tz&. The impulse response h(t) = dEs/dt is simply the rate of scattered
 Figure 2.1: Fiber tip with an angleci cleave

energy returning back to the launch point. Then, using 22 = d / n for 2-way
propagation in an optical fiber of length L,the impulse and frequency responses
of Rayleigh backscattering are [KlO, K13, K14,K32]

2.4 Fresnel Backreflection

At a cleaved end of MME' as iuustrated in Fig 2.1, the cleave angle 8 deter-
mines how much of Ftesnel ref'iection, which is normdy produced due to the
index difierence between nl and no, is coiiected back into the fiber. Using Eqn
A.63 with small angle approximation (@,OEui < 1) or from Ref [K12], the total
backrefiection coefficient can be writ ten directly as

if RQ and RL are the reflection coefficients respectively at near-end (z =

0) and far-end ( 2 = L), then the impulse and frequency responses of Fresnel
backreflection are [K13,K14, K32]

2.5 Fluorescence Collection by Direct Coupling

Coupling between diEuse source and optical fiber has been studied using Li-
ouville's theorem and ray optic assumption, mainly in the context of taunching
LED into MMF [Hl].By extending the results of LED-tefiber coupling, 1derive
the efEciency of collecting fluorescence through MMF, whether the excitation is
canied by the sarne fiber or by another. Others have studied the case of single
fiber p6,H7, Hg]and double fibers [H3,H4, H8, HlO], all utiiizing brute force
integration with complicated geometry. My approach yields simpler yet more
complete solutions.

2.5.1 Liouville's Theorem

Liouville's theorem uses canonicd conjugate variables of Hamiltonian mechan-
ics. For ray optics, it begins with Format's principle on optical path length,
[Hl, Marcuse]
E
y , =)& = L(Z. y, I , = extremum (2.23)

where 2 = &/&, y1 = dyldz, and 8 s = &z + <Py + P z . This variational

problem has well-known solution in terms of Lagrangian L ( x ,y, x', y ' ) or in terms
of Hamihonian H ( z , y,p=,p,) = p,d +puy1 - L,where pz = a L / W = n&/ds
and p, = aL/ayt = ndylds.
The generaiized coordinats (2,y ) and generalized rnomentwns @,,p,) con-
stitute 4-Dphase space, in which point (2, y, pz, p,) represents an individual ray,
density p ( z , y , pz, pu) represents the L o d distribution of ray flux, and phase vol-
ume r(x,y , pz, p,) represents the domain of ray ensemble. For a ray distribution
propagating dong Z &, the phase volume becomes [Hl]

where A is normal area and 0 is planar solid angle as defineci in Eqn A.14.
Liouville's theorem States t hat , as rays propagate through a lossless opticai
system according to the equations of motion,
a local density ( p ) in the neighbourhood of any moving point does not
change, and
local phase volume (ï)surrounding a fixeci ensemble of points does not
change even though its b o u n d q and shape may change.
This means that once the ray distribution is established by source, no optical
system c m decrease the phase volume without excluding some ray points, since
by doing so the local density in some region would necessarily increase in vi-
olation of Liouville's theorem. The use of lem and other optical cornponents
merely changes the shape of phase volume. For uniform n, Liouville's theorem
gives
I: = n 2 ~ S =
1 constant (2.25)
It is instructive to note that phase volume î is proportional to the total number
of modes N as defineci in Eqn 2.2.

2.5.2 Collection Efficiency

Out of the total rays emitted by a source, an optical fiber can only collect those
ray flux which fa11 within its aperture. If î, and r are the phase volume of
source and fiber, respectively, the collection efficiency of optical fiber is a simple
ratio of ray flux,
rn

assuming that source density p, and refractive index n are uniform which is
usuaily the case for diffuse source such as LED and fluorescence.
LED Because the emitting area is smaUer than fiber core, LED is usudy butt-
coupled to MMF [Gowar]. Using RI = ssin2Bnu for the fiber's aperture
and R, = a for source emission into 1hemisphere, the collection efkiency
is
 Figure 2.2: Excitation and co11ection of fluorescence using single fiber

Figure 2.3: Excitation and collection of fluorescence using double

+ excitation

t2u+2s+
Fluorescence As shown in Fig 2.2, the phase volume of fluorescent source at
z = z emitting into 2 hemispheres is.'I = GASR,, where '4. = na2(l +
) ~ R, = 2a. The phase volume of fiber at z = O is Tt = n:AfQ
~ / b and
where Af = ?ru2and RI = ir sin2BEU Then, the collection efficiency is

2.5.3 Collection Efficiency of Single Fiber

Consider a LW illuminating a conical volume with haü-angle 0Et4as shown
in Fig 2.2. lui an absorbing medium, the excitation profile dong 2 axis is
I ( z ) = Ioe-Q', where a is bu& absorption coefficient and Io is the initial inten-
sity over the core area 4. In [z,z + dz] interval over cross-section area A(%),
aI(z)dt is absorbed and Q Y d ( r ) & is emitted as fluorescence, of which ody
qc QYckI(z)dz is coiiected by the fiber. In travelling back to z = 0, fluorescence
suffers from various attenuations which are lumped into another coefficient fl
(usudy negligible) .
Using Eqn 2.28, the o v e r d collection efinciency for single fiber becomes

where x = (a+ P )z and L = (a +P )b are normalized distances. Asymptotically,

-
F ( L ) - L a s L + 0, and F ( L ) 1 - 2 / L a s L -+ W. Sirnilar result has been
obtained for backscattering from buik medium [H2,Culshaw].

2.5.4 Collection Efficiency of Double Fibers

Consider 2 identical and parallei MMFs whose cores me separated by 2s distance
as shown in Fig 2.3. Excitation is carried by one fiber and fluorescence is
collected by the other. The derivation for double fibers is same as for single
fiber, except that the collecting fiber does not "sees" the entire cross-section
A(z) but o d y the overlap area B ( z ) beginning at = s/ tanom = bsla. This
reduces rl, in proportion by

Then, the overaii collection efficiency for double fibers becomes

G(L,s) = lr (-1
ër L + x
L # - sin t$
R &

+
where x = (a B)z. L = (a B)b, + x0 = Ls/a are normalized distances, and
+ +
cos($/2) = (L x o ) / ( L 2).
Figure 2.4: Graph of F ( L ) and G(Lya) with s = O and s = 1.440- F(L)is for
single fiber. G(L,O) is for double fibers with no cladding so that the fiber cores
touch eôch other. G(L,1.44~)is for double fibers with reaiistic dimensions of
62.5pm cladding radius and 2 5 . 6 core~ radius, just We the MMF used in my
experiment.

F (L)and G(L,s) are e d u a t e d using Romberg numericd integration [Press]

and plotted in Fig 2.4. Since F ( L ) > G(L,s) for aii L > O, as expected,
single fiber coliects more fluorescence than double fibers. The assumption of
zero separation (s = O), even though unrealistic because cladding is present in
most optical fibers, establishes the upper limit of collection efficiency for double
fiber configuration. Whea a real cladding dimension of MMF (s = 1.44~)is
considered, the efficiency decreases by an order of magnitude.

2.6 Fluorescence Collection by Evanescent Cou-

pling
In order to maximize evanescent couphg in MMF [I9], the cladding is removed
so that the core is in direct contact with sample (no)which acts as the rnissing
"cladding" . Similady in order to rn-e evanescent coupiing in SMF [I11], the
fiber is heated and stretched so that the core for ail practical purpose disappears,
and a new optical waveguide is formed by the cladding (na)acting as "coren
and sample (no)acting as "cladding". Since typicai sample index is not equal
to the dadding index, there is coupiing Ioss [Il&115, 1161 due to AH mismatch
a t the beginning and at the end of sensing region.

2.6.1 Evanescent Excitation

The sample absorbs emescent power just as it does in the case of bulk excita-
tion. But, because the sample experiences only p fraction of the total power, the
absorption coacient decreases to 7 = a p and the d t a t i o n profile becornes
I ( z ) = l ~ e - ~ 'For
. a SMF or MMF with weakly guiding stepindex profile, the
cladding fraction of mode power is [Cheo, Ghatak]

where u2 = k$p: - f12, 3 = f12 - k&& u2a2+ w2a2= V2, and Bessel function
Kl() represents the radial field of mode order I in the dadding.
By stimming over all N = V2/2modes in stepindex MMF, the cladding
fraction of total power propagating in MMF becomes p8, 19, after minor cor-
rectionl

In the case of SMF, there is oniy one mode to consider, nameIy 1 = O. However,
it is much easier to work with a Gaussian approximation than with the exact
Bessel function. For stepindex SMF,the radial field of fundamental mode can
be approximated by [Snyder]

where w is the spot size of Gaussian profile and a is the core radius. Then, the
cladding fraction of mode power in SMF becomes a simple ratio of integrais,
Pl1
P=
/na $2(r)rdr
- 1
- e-2a21w2 = -
v 2

2.6.2 Collection Efficiency

The collecting fluorescence through evanescent coupling can be thought of as
the inverse of evanescent absorption; the amount that "tunnels" back into a
guiding mode is proportionai to the evanescent field intensity of that mode.
Hence, MMF is expected to have higher collection efficiency than SMF because
of the simple fact that MMF supports many modes and SMF supports only one.
Using Eqn 2.36, the collection &ciency for stepindex SMF is given as [Ill]

This expression was derived with 2 assumptions - first, the indexes of fiber and
sample are similar (nl = n2 a no);second, the wavelengths of excitation and
fluorescence are close enough (A, x X p ) so that their modal fields are similar.
Unfortunately, there is no comparable analytical expression for MMF. The
collection efficiency of weakly guiding stepindex MMF [IIO] and general s t e p
index MMF p12, 1131 has been studied using numerical simulations. For bulk
source distributeci uniformly throughout the sample, fl, a V since oniy high-
order modes have appreciabie penetration into the cladding. For surface source
located at the fiber-sample interface, II, cc V2 since every mode has a reasonable
power at the interface. But, the exponentiaily decaying source profile which
occurs in evanescent excitation has not been studied so far.
Chapter 3

Fluorescence System

3.1 Fluorescent Response

A single lifetime fluorescence system ean be modelled as 3 level system as de-
picted in Fig 3.1. When exciteci, the system moves from ground state (1) to
absorption state 13), then quiddy settles to emission state 12) via non-radiative
relaxations. The system drops down to ground state through radiative (A) and
non-radiative ( k )decays. When the excitation is t h e varying E ( t ) ,the emission
state 12) will be populated at a rate proportiond to the excitation. The rate
equation descnbing the emission state population &(t) is [Lakowia, Birks]

dN2(t) - - ( A + k)N2( t )+ E ( t )
--
dt
which has weil-known solution N2( t ) = e-llT * E ( t ) ,where r = 1/(A + k ) is the
iifetime of fluorescent decay.

- -
Since the fluorescent intensity that is actually observed is ( h c / X p ) M 2(t),
the solution can be rewritten in terms of input z(t) E (t) and output y (t)
N2( t )of linear, tirne-invariant, and stable system

which is entirely characterized by the impulse response h(t) = e-'1' and the
frequency response H(f) = r / ( 1 + j27rf r ) . Since optical signais cannot be

Figure 3.1: 3-levei mode1 of fluorescence system

absorption state
13)
mission state

11) ground state

negative, the input and output must be biased above O. This is important for
sinusoidd input and output
x(t) = 1 + a cos(2lrf t), la1 5 i (3-3)

=l+a
cos (2sf t - tan-' (2rf s ) )
(3.5)

where the constant term has been normalized to 1.

3.2 Determining Lifetime in Time Domain

The fluorescent Iifetime of sample can be measured in time domain using pulse
excitation. Because of finite pulse width, finite detector bandwidth, A-bias, and
other problems, it is a standard practice to measue fluorescence fcom both the
sample and a reference using the same setup. This arrangement is outlined in

where 1(t) is laser excitation, d(t) is detector response, h(t) = e-'1' is sample re-
sponse with lifetime r, &(t) = edtlTr ci refereace response with lifetime r,, z(t)
is measured reference £iuorescence, and y(t) is measured sample fluorescence.
Mathematicdy, it can be described as

This problem is solved by iterative convolution in which the measured data

{xi,Yi ) is fitted directly to the convolution models. m e r convolving h, (t) with
Eqn 3.7 and h(t) with Eqn 3.6, the residual fiinction

is minimized for the least-square estimator of r. Since the reference Iifetime

T, is a fixeci constant, this is a 1-D minimization probtem which can be solved
by Brent's method press], for exampIe. Although one convolution needs to
be calculateci for every new s point, the convergence for 1-D problem is u s u d y
fast. Because of its simplicity and flexibility, this method is preferred over direct
deconvolution metho&.

3.3 Determining Lifetime in Requency Domain

The fluorescent lifetime of sample can be measured in frequency domain using si-
nusoidal or harmonic excitation. Typically, sinusoidal excitation is generated by
rnodulating a continuous wave source; and harmonic excitation is derived fiom
a train of pulses which decomposes into hamonic series. 3 signals are involved
in general - excitation z ( t ) with fundamental kequency f 1 and harmonics n f i ,
fluorescent response y ( t ) ,and LO signal z ( t ) a t frequency f,.

where, from Eqns 3.3 and 3.5,

If the frequency f 1 is low enough, the amplitude factor (ml) and phase shift
( p i) are measured directly from Eqns 3.9 and 3.10-But if the frequency is hi&,
heterodyrre or homodyne method is used, where LO signal z ( t ) is mixed with RF
signals z(t)and y(t), either inside or outside photodetector. Subsequent Iowpass
filter produces "down-convertetin LF signals X ( t ) and y ( t ) at a sufliciently low
fiequency so that ml and ipt can be measured using a lockin amplifier. In
heterodyne method, fo * f I and the IF signals are at frequency 4f = f1 - fo ;
and in homodyne method, f, = f i and the IF signals are DC:

UsuaUy, the data are sampled at more than one frequency, either by changing
f, to near any of the harmonics or by changing f1 itself.
Once { m ,cp) data set has been obtained, the Metirne T is determineci either
through linear regression of [P4]
tan ip = laf r, l / m 2 = 1 + (2afT)* (3.13)
against the frequency f, or through minimization of residual function

where mi and vi are measured at frequency fi,and m(fi) and cp( fi)are cal-
cdated from Eqn 3.12. For an example, m and cp that will be produced a t
f = lMHz are

Fig 3.2 lists papers, in the order of increasing bandwidth, which use sinu-
soidal or harmonic excitation to measure fluorescent lifetime in the fiequency
domain. 1should point out that since fluorescencesystem is treated as linear and
time-invariant, a Network Analyzer (some with 20GHz bandwidth) can perform
all the signal processing and can in principle replace most of the equipments de-
scnbed in the papers. But, Network Analyzer is very expensive, and the papers
basically talk about cheaper alternative of measuring the sinusoidai response.
Main advantages of frequency domain method over time domain method are
that high detection bandwidth and deconvolution of sample response are not
requird.
 Figure 3.2: Survey of frequency domain fluorometers
Refi 1 maxkeq 1 source 1 data
Sinusoidai Excitation
EOM
LED
LED
AOM
EOM
D2 Lamp
EOM
EOM
Quartz AOM

Harmonie E x i tation
synchrotron radiation
mode-locked laser + AOM
mode noise of CW laser
mode-locked laser
pulsed laser diode
mode-locked laser
mode-lockeù laser
Chapter 4

Acousto-Optic Modulator

1 used PbMo04 acousto-optic modulator (AOM) in m y experiment because it

was cheap and available. Others [N9, N10, N12, N15, N16, N17, NI81 have used
different optical modulators to implement various schemes in which the double
modulation of optical signal is the key feature just as in my own correlation
scheme. The choice depends on factors like modulation bandwidth, operating
wavelength, compatibility with optical fiber, and, most of dl, cost and availabil-
ity. In this section, 1 review o d y those aspects of AOM relevant to amplitude
modulation, since that is how 1 use it.

4.1 Acoustic Perturbation

Acousto-optic (AO) interaction can be modelled as inelastic scattering between
incident optical photon (w, k) and acoustic phonon (0, K) , producing a diffracted
photon that is "upshifted" (w+, k+) if the phonon is absorbed or "down-shiftedn
(w-, k-) if an identicai phonon is emitted. [Korpel, Magdich, Yarivj

The phase-matching conditions determine diffraction angle, polarizat ion, fie-

quency shift, and propagation constant. But the modulation of optical signai is
achieved through the difEaction efficiency.
Outside the scattering site, incident wave E and difEracted wave E' propagate
as separate and uncoupleci plane waves. Inside the scattering site, however, E
and El waves are coupled through acoustic perturbation which is produced
according to the weU-known strain-optic relation, [Ghatak, Magdich]

where p is strain-optic constant and S is strain field set up by the acoustic wave.
Because of syrnmetry in the tensor components, matrix notation [.II[ and vector
notation ( - ) i can be converted from one to the other, dependhg on the context.
Then, the coupled wave equation becornes [Magdich, Ghatak, Yariv]

where
 The degree of coupling between the inadent and dïfEacted waves is given
by scaIar pagdich, Ghatak, Yariv]

where e and e' are unit vectors of E and Et respectively. Usually, the acoustic
and opticai waves are aligned wit h AOM's axb, so that (e'lA[e] le) d e c t s only
one element from AM. Using appropriate scalar parameters in e = c6n2 and
A(l/n2)= pS, the change in refractive index due to acoustic perturbation is
pagdich, Ghatak, Yariv]

where M2 = n 6 $ / p : is a figure of merit for AOM, A is area of transducer, and

fi = ( A U . ) ( ~ J ~is~ acoustic
) power. Fkom experimental point of view, A n is
the most important parameter because it influences ail optical manifestations
of A 0 interaction.
For example, consider longitudinal acoustic wave dong t in PbMoO* which
is an uniaxial crystai with 4/m point symmetry:

Then, the acoustic perturbation as given by Eqn 4.4 is

Because A[€] is diagonal, the diffracteci and incident waves have the same po-
larization; otherwise, there wodd be no A0 coupling. The scalar Ae and An
that are selected depend on the orientation of E and K. Therefore, from Eqns
4.5 and 4.6,

4.2 Diffraction Efficiency

AOM can be either Bmgg or Raman-Nath type depending on [Ghatak, Korpel,
N25]
K2L
Q=-=- 27rAL Q < 1, for Raman-Nath
k nA2 Q » 1, for Bragg
where K is acoustic wavevector, k is optical wavevector, and L is transducer
length.
Figure 4.1: Diffraction at Bragg angle for upshift case (top) and down-shift case
(bottom)

Figure 4.2: Diffraction at non-Bragg angle for upshift case (top) and down-shift
case (bottom)

4.2.1 Raman-Nath AOM

In Raman-Nath AOM, multiple scattering takes place because the acoustic di-
vergence is greater than the diffraction angle. If eo is the angle of incidence
with respect to the acoustic nomd, the diffraction efficiency ( q ) and angle (8)
of mth-order beam are [Ghatak, Korpel, N25, Yariv]

where J,() is standard Bessel function, and Alp = kpAr~L/cos8~ is optical

phase shift eaused by An over the interaction length.
When O0 = O for example, 100% of 0th-order beam is difiacted out at
Aip = 2.40; maximum 34% diffraction into 1st-order occurs at Aip = 1.84; and
maximum 2496 diffraction into 2nd-orderoccurs at A p = 3.05. Naturally, higher
orders exhibit progressively lower efficiencies. Due to its low efficiencyand low
bandwidth, Raman-Nath AOM appears o d y in specialized application (N24J.
4.2.2 Bragg AOM
In Bragg AOM, only single scattering ocnrrs because the acoustic divergence is
smaller than the difkaction angle. As depicted in Fig 4.1, the Bragg diffraction
produces a single lst-order beam when the angle of incidence (Bo) is equal to
the Bragg angle

2nA
The dinraction efficiency of ûth and 1st order beams are [Ghatak, Korpel, N25,
Yarivl

where 49 = koAnL/ cos OB is optical phase shift caused by An over the inter-
action length, Bragg AOM can reach 100% 1st-order dif&action at Acp = 3.14-
As shown in Fig 4.2, if the angle of incidence increases by A8 fiom the
Bragg angle, then the lst-order angle decreases by the same amount, so that
the propagation constants become = k cos(dB + A@) and fl1 = kcos(BB -
A8). The resulting phase mismatch reduces the difFraction efficiency of 1st-order
beam to [Ghatak, KorpeI, N25, Yariv]

where 4~2= K4dLI cos e0 is acoustic phase shift caused by the anguiar change
48, and Acp = ko4nL/ cos go is optical phase shift caused by An. The phase
mismatch usually produces higher orders, as weli.

4.3 Modulation Bandwidth

Intensity modulation in AOM is accomplished by moddating the amplitude of
acoustic carrier wave, which in turn modulates the diffraction efficiency. If f,
and f, are modulation and carrier fiequacies respectively, amplitude modula-
tion (AM) creates 2 sidebands at f,*f,. Experimentally, maximum moduiation
frequency ,f is reached when modulation wavelength A, is equal to optical
beam width D, a t which point there is zero net modulation. [Korpel]

where T = D / v is transit t h e for the acoustic wave to cross the optical beam,
and v = Actc = A, f, is the acoustic speed. As the beam width is reduced,
the modulation bandwidth increases, the optical divergence increases, and the
ciiffiaction efficiency decreases because greater proportion of incident beam is
phase mismatched.
+
When acoustic wave with sinusoidal envelope s ( z ) = I a cos(2?rz/A,)
passes through optical beam with spatial profile p(z), the diffracted signai
q(t)lZzvt = p(z) * s(z) is also sinusoidal. This situation resembles the sinu-
soidal excitation of a hear system as deihed in Eqns 3.3 and 3.4, where s(~)
acts as %putn, p(z) acts as "system response", and q(t) acts as "outputn. If
the optical beam has unifom profile with D width, p ( z ) = fl(z/D);and if it
has Gowsian profile with D full width at ë2point, p ( z ) = e-8ia/D'. Then,
the difEraction output becomes pagdich, Goutzoulis]

= 1 + a sinc(fdf,,) c o s ( 2 ~ f , t ) , uniform beam (4.18)

- 1 + oe-.'f;/~f2-cos(2nfmt), Gaussianbeam (4.19)
where P(f) is the spatial FT of optical beam profile p(z). The 3dB point or
50%modulation depth occurs at

0.60fm,, uniform beam

f 3 d =
~
0.75fm,, Gaussian beam
Chapter 5

Signal Processing

5.1 Signal Flow

Fig 5.1 depicts the block diagram of my aii optical signal processing scheme:
1. LO signal x(t) is rnodulated onto DC laser intensity by AOM.
2. 1-way propagation in optical fiber introduces At/2 delay.
3. At the far-end, fiesne1 reflection x(t - At/2) or fluorescent response y(t -
A t / 2 ) couples back into the fiber.
4. On the return path, the fiber introduces another A t / 2 propagation delay.

5. Second optical modulation by AOM mixes the LO signal z ( t ) with the

-
returning signal, giving z ( t ) z ( t At) or x(t)y(t- At).
6. The mixed signai is integrated by lowpass detector, producing auto corre-
lation &,(-At) or cross correlation &,(-At).
The goal is to determine the fiber length (L) by analyzing the auto correlation
which cornes from Fresnel backreflection with impulse response h(t) = 6 ( t ) , and
to determine the fluorescent lifetime ( 7 )by andyzing the cross correIation which
cornes nom SPA or RU fluorescence with impulse response h ( t ) = e - t / T .
AU signals are periodic, either sinusoidal or square-wave, with frequency
f = 1/T. Since the propagation delay At = 2Lnlc is fixed by the fiber Iength,
and since it is the modulation frequency that is being varied in my experiment,
the cordations turn out to be functions of f:

This represents a fundamental change of domain which does not easily lend itseif
to the conventional techniques of (stationary ) random variable. Ait hough time
domain responses can be inferreci from inverse FT of correlations, it is simpler
to estimate L and T directly in iiequency domain.
Figure 5.1: Block diagram of signal flow in my experiment. Auto correlation
(top) is produced when measuring Ftesnel backreflection, and cross correlation
(bottom) is produced when measuring SPA or RU fluorescence.

. - +)
-z(t
fiber 4
4t)
laser
* At/2
.x(t -F) l ( t - At)
z(t)z(t - At)

&.(-At) 4 1
detector
+*Z&

* - A j
detector
;,-:& /

Figure 5.2: Square-wave excitation x(t) and fluorescent response y ( t ) of single

lifetime system
5.2 Auto and Cross Correlations
Square-wave excitation is used extensiveiy in my experiment because the simple
"on/off States are easier to generate, e s p d y with nonfinear moduiator like
AOM.But its analysis is based on sinusoidal case. The sinusoidal input z(t) and
the fluorescent response y(t) are &en by Eqas 3.3 and 3.5. Now, consider an
ideai (equal on/off) square-wave excitation at fiequency f = 1/T, as iliustrated
in Fig 5.2. The waveform is decornposed into more convenient Fourier series
using Eqns A.48 and A.49.

-
By superposing Eqn 3.5, the fiuorescent response to the square-wave excitation
is
00 4(- cos ( 2 mf t -tan-l(~mfr))
~ ( t ) = l + x (5.6)
odd rn dl+( 2 r mf T ) ~
which, as shown in Fig 5.2, rises and falls exponentiaiiy between the maximum
B = 1+ tanh(T/4r) and the minimum A = 1- tanh(T/4r).
Then, the correlations defined by Eqns 5.1 and 5.2 becorne, for sinusoidal
case

Rdf)= 1 +
a2 cos (27rf At + tan-' (27rf 7))
dl + (Zrf r ) 2
and, for square-wave case

8 cm (2amf ~t + t a K 1( 2 mf r))
00

&,cf) = l + 1-
~*m2
odd m dl+( 2 mf T)? (5.10)

The effect of finite bandwidth on the shape of correlation cannot be ignoreci.

First , the AOM hquency response is sinc(0.6f / f s d ~ )fkom Eqns 4.18 and 4.20,
where beam profile for both the laser beam and the MMF output are assumeci
sufiiciently uniform. After 2 passes thmugh AOM,the fkequency signai is atten-
uated by sine2(0.6f / f 3ds). Second, the cosine series is truncated after 36MHz
because very little contribution cornes from higher fiequencies due to bandwidth
and lifetime attenuations. Therefore, the auto and cross correlations must be
corrected to, for sinusoidai case
a2
&, (f) = 1 t 2 sinc2(w ) cos(v)
a2 cos(v + tan- (u))
a,cn = i + -
2
s~~c~(w)
,/m
l
and, for square-wave case

%.(f) = 1+ C O
sine2(mw) codmv)
odd nt
mf =36MHz
8 cos (mu + tan-l (mu))
(5.14)
odd m
,/-
where w = 0.6f / f s d ~ , u = 27r f At, u = 27rf r,and At = 2Ln/c.
Using parameter values which are dose to the actual experimental condi-
tions - f s =~8MHz, L = 50m, T = 30ns (SPA), and T = 35011s (RU) -
both square-wave and sinusoidal excitation have b e n simulateci. The AC corn-
ponent of &,(f) for Fresnel backreflection is plotted in Fig 5.3; and the AC
component of Ru( f ) for SPA and RU fluorescence are plotted in Figs 5.4 and
5.5, respectively.

5.3 Time Domain Responses

Ignoring the bandwidth attenuation and using Eqns A.43 and -4.44, inverse FT
of correlations are, for sinusoidal case
a2 a2
= 6(t) + -6(t
4
- At) + -6(t
4
+ At) (5.15)

do
4
rl{&(f 11 = 6(t) + { ! ! - ( ] t , - m ~ t ) / m,r Itl < rnAt (5.18)
Itf>mAt
odd m

Therefore, at least for t > O (sinusoidal) or for O < t < 3At (square-wave),
the inverse M: correctly represents the impulse response of Fresnel reflection
6(t - At) and fluorescent sample e - ( t - A t ) / T , with propagation delay At and
lifetime T.
The inverse FT was done analyticaliy because &, (f) and &, (f) are known
for aii -00 < f < oo. But in real world, the inverse FT must be done by DFT
because the correlations are known only at discrete points over -f, <f <
f, ,. That is, N data points are sampled at 6f interval in the positive frequency
range, and then fiipped onto the negative frequency range by taking advantage
of the fact that the correlations are even functions. This gives us 2N data points
for DFT whose fundamental relation between tirne domain resolution (6t) and
fiequency domain resolution (6f ) is

Since f, = 4MHz in my experiment, the time and spatiai resolutions are

limiteci to
bt = 125ns, 61; = (c/2n)dt = 13m (5.20)
Clearly, these resolutions are inadequate for fiber Iength of 50m and fluorescent
lifetimes of 30x1sand 350ns, if 1 were to foUow the traditional DFT method
Figure 5.3: S i d a t e d AC component of &,( f ) for Fresnel backreflection under
square-wave excitation (top) and sinusoidal excitation (bottom), with L = 50m
and f s d ~= 8MHz
Figure 5.4: Simulated AC component of &,(f) for SPA fluorescence under
L = 50m,
square-wave excitation (top) and sinusoida1 excitation (bottom), wit h
fJdB = 8MHz, and r = 30ns
Figure 5.5: Simulated AC component of &(, f ) for RU fluorescence under
square-waveexcitation (top) and sinusoidal excitation (bottom),with L = 5ûm,
fsds = SiMHz, and T = 350ns
which estimates the parameters kom time domain responses. In my correla-
tion scheme, however, the parameters are gtimated by data fitting a,
(f) and
&.(fi directly in the hequeney domain, thereby circumventing the resolution
constraint of DFT formalism.

5.4 Estimation of Parameters

The auto correlation of Eqns 5.11 and 5.13 and the cross correiation of Eqns 5.12
and 5.14 can be modelled as y = &= (f ;L) and y = R,(f;L,r ) , where fiber
length L and £luorescent Iifetime T are unknown parameters. Fortunatdy, the
2 parameters can be determined sequentially. First, the least-square estimator
of L is found by minimizing residual function

where Yi is the actual measurement of auto correlation due to Resnel backre-

Election at frequency fi, and &=(fi) is calculated fiom the analytical model.
Second, the least-square estimator of r is found by mhimkhg residuai function

where y; is the measurement of cross correiation due to SPA or RU fluorescence

at frequency fi, and &,(f;) is calcdated from the model.
In each case, the data fitting reduces to 1-D minimization problem which
W solved by Brent's method [Press]. When evaluating the residual in Eqns
5.21 and 5.22, the measured and the calculated data are standardized, £ùst by
subtracting linear regression iine from the data, and then by normalizing the
data to zero mean and unit standard deviation.

5.5 Estimation of Error

The accuracy and the confidence intervai of estimation can oniy be established
through ensemble average. Conceptually, the '%ruenparameter Uttue is statisti-
cally reaüzed, dong with other random processes or errors, as an experimental
data set D lfiom which an estimation a1 is obtained by data fitting to a model.
However, because of the random components, V 1is not an unique realization
of ut,, . There are infinitely many other redizations V2, V s, - from which
estimations a2,a3, are obtained. After sufficient number of data sets, the
ensemble of ai would describe some probabiiity distribution about utruc in the
parameter space. [Press]
I f { a i ) is a sample of parameter estimations, then the true parameter is
approximated by ensemble average: [Gutt man]

where tN-1;a/2is a point in Student-t distribution with N - 1degree of fieedorn

and 100(1 - a)% confidence interval, and a and sa are the usuai mean and
standard deviation,
 -&O, if (b*) is a sample of estimations for another parameter b, the Merence
a - b is approximated by [Guttman]

where
Chapter 6

Experiment

6.1 Experimental Setup

The experimental setup of my correlation scheme is outlined in Fig 6.1; and
the optical and electrical equipments are s h o m in Fig 6.2. The purpose of
this experiment is to demonstrate a novel optical signal processing scheme for
measuring the lifetime of fluorescence decay. Because the optical fiber introduces
round-trip propagation delay, the fiber length is also measured using F'resnel
backreflection fiom the far-end.
-4 continuous Ar laser beam passes through halfwave rotator (X/2) and polar-
king beamsplitter (BS), and is focused by a Iens onto acousto-optic modulator
(AOM) at its Bragg angle. AOM modulates the intensity of laser beam in si-
nusoidal or square-wave whose frequency is ramped in time by local oscillator
(LO). The resulting 0th-order beam is spatiaiiy filtered by adjustable iris and
launched into multimode optical fiber (MMF)by a standard fibre optic coupler
equipped with 10X objective.
At the far-end, the optical fiber collects either Fresnel backrefiection or ffu-
orescence from SPA or RU solution, and guides it back tu AOM. The returning
signai is focused by the fibre optic coupler onto AOM again at its old Bragg
angle. The 0th-order beam of this second intensity modulation is collimated by
the lem, filterd by uis and interference filter, and finally detected by a lowpass
PIN diode detector. A digital oscilioscope coiiects the voltage input to LO as
"inputn data and the voltage output from PIN detector as "output" data, and
uploads to computer for data fitting.
When m e a s m g Fiesnel backrefiection as the auto correlation, my scheme
resembles a hybrid of modulation OFDR and correlation OTDR,because the
correlation is measured as function of modulation frequency. When measuring
fluorescent response as the cmss correlation, my scheme resembles frequency
domain fiuorometer, because the Metirne is determineci through its sinusoidal
response.

6.2 Components
6.2.1 Laser
Argon laser was pre-configured for single-he and single-fiequency operation at
wavelength X = 45?.9nm, with bandwidth Au = 3 M h , and coherence length
Lcor = c/Au = 100m. Its output was linearly polarized and was stable at
l5mW f 0.5% without optical feedback problem. The laser polarization was
Figure 6.1: Schematic diagram of experimental setup. Ali components are from
Ontario Laser and Lightwave Research Center (OLLRC)or Department of mec-
tricd Engineering at U of T.

I I
SPA
RU
,amp
scope ,
* oscill* I

10X
ucornputer

10X objective fibre optic coupler (Newport)

AOD 140MHz quartz o s d a t o r driver (NEC OD-8802)
AOM Lead Molybdate (PbMo04) acousto-optic moduiator
(NEC OD-8813)
Ar laser Argon laser (Spectra-Physics 2020-05)
BS Broadband polarization beamsplitter cube (Newport
lOFC16/PB.3)
detector Low-noise J-FETinput Quad OpAmp (TL 074)
filter Interference bandpass ûiters (Melles Griot 460 f IOnm,
550 Jr ZOnm, 600 f 20nm)
X/2 Broadband polarization rotator (Newport PR-550)
lens Bi-convex o p t i d glas lem (Newport KBX 076)
LO 4MHz Pulse/Fbction Generator (Wavetek 187)
MMF 51-2pm diameter core multimode fiber (Corning 1517)
oscilloscope lOOMHz digital storage oscilioscope (Tektronics 2230)
PIN Blue enhanced PIN diode (Silicon Detector Corp. SD
200-22-12-041)
Programmable Ramp Generator (Burleigh RC-43)
Rut henium tris-bipyridyl dichloride, hexahydrat e
(Molecular Probes R- 1498)
SPA N-(3-sulfopropyl) acridinium, inner salt (Molecular
Probes S-460)
Figure 6.2: Pictures of optical (top) and electrical (bottom) equipments
rotated by hahave (X/2) plate for maximum transmission through beamsplitter
(W-
Because of very long coherence Iength, Fresnel backrdection coming out
of MMF contains speckle pattern which is local fluctuation of intensity due to
interference amongst modes. In addition, 6082 "humming" vibration fiom the
laser cooiing mechanism contributes to the fluctuation by inducing an acoustic
perturbation of fiber index. But, since there is no mode dependent c o u p h g
anywhere in the signal path, the total intensity over entire optical beam should
be constant despite the local phase noise. Nonetheles, some intensity noise may
be introduced because, as the fiber optic launcher expaxiences the vibration, its
focused spots move about over the fiber end-face and over the AOM aperture.

6.2.2 AOM and AOD

.4cousto-optic modulator (AOM)and its matching driver (AOD) were salvaged
from old LWClaser fax machine. AOM, shown in Fig 6.3, consists of Lead
Molybdate (PbMo04) crystai with 633nm anti-refiection coating and lOmm
Lithium Niobate piezoelectric transducer. AOD is 140MHz quartz oscillator
which is amplitude modulated by a R F mixer. Since the acoustic speed is va =
3.63km/s dong f axis (B12,CRC, N251, 140MHz carrier frequency produces
acoustic wavelength of A = 25.9pm. The AOM is Bragg type because Q =
17 > 1.
PbMoOs is tetragonal(4/m point group) system with strain-optic coefficient
at 633nm p 1 2 , CRC, Yariv]

It is interesthg to note that nfpi3 = nJm3holds true for PbMoOc throughout

its optical spectrurn [B12]. This means that, from Eqn 4.10, the Bragg diffrac-
tion is isotropie and independent of polarization between E and K. Since BS
refiects (TE) wave polarized aloag 1 and transmits (TM) wave polarized dong
2, n3 acts on the laser excitation on the first p a s , and nl acts on the retuming
signal that actually reaches PIN detector. The relevant refiactive indexes (in
bold) and their Bragg angles are [Bass]

The transfer function of AOM/AOD is plotted in Fig 6.4. In order to obtain

a linear response fiom GOM, the electrical input to AOD must be biased at
the linear portion of traasfer function. The best sinusoidal output with 25dB
harmonic suppression was observed with 2.2 f 1.OV input to AOD. For square-
wave modulation, biasing does not matter since it works on simple "on/off
principle. Few parameters of Oth-order beam that have been measured at 458nm
are
Figure 6.3: Picture of inside AOM
Figure 6.4: Normaiized output inteosity of Othorder beam kom AOM in re-
sponse to input voltage to AOD

normalized
output
fkom AOM

input to AOD [VI

The Bragg condition is not M y satis6ed in m y experiment mainly due to

optical divergence and Mependence. Especially, the A-dependence of s a c -
tion efficiency becomes important when measuring SPA and RU fluorescence
at 550nm and 600nm respectively, since AOM was aiigned at 458nm for both
the laser excitation and the returning signals. The efficiency decreases not
oniy because of the longer wavelengths, but also because the old incident an-
gle (O0 = 0.214") no longer matches the new Bragg angles. Using small angle
approximation in Eqn 4.15, I)I decreases from 90.5% at 458nm to
550nm (SPA) 600nm (RU)
4 a = KA8L 1.50 2.62

The optical divergence of returning signai will reduce the diffraction effi-
ciency even further, because AOM difnacts oniy those portion of the focused
beam that satisfies Bragg condition- in addition, since the half-angle diver-
gence (0.86") of 10X objective is greater than the Bragg angle (0.51°), 0th and
1st order beams of returning signal overlaps. This reduces the signal modula-
tion depth and contributes to intensity noise, because the PIN detector is not
integrating over the entire 0th-order beam.
The double p a s through AOM is compiicated by
a dinerent Bragg angle and efficiency for excitation (at 458nm)and fluores-
cence signal (at 550nm or 600nm),
Mixent focal Iength for 10X objective (1.48cm) and lens (20cm), and
different beam profile for first pass (nearly p a r d e l laser beam) and second
pass (11.5O MMF aperture).
These asymrnetries mean that optimal coupliig condition for one way does not
automaticdy Iead to optimal coupling for the other way. A compromise is
unavoidable.

6.2.3 MMF
1used 50m long step-index MMF with 25.6pm core radius and 62.5prn cladding
radius. Some of material data at 458nm are
The attenuation coefficient a t 458nm was dculated from 2.3dB/km given at
850nm, using the fact that most of the propagation loss occurs through Rayleigh
scattering where 7 oc R o m Eqn 2.18, the impulse response of Rayleigh
backscattering is an exponential decay 6 t h "Lifetime" of 0 . 7 7 ~ Although
. the
critical length in Eqn 2-11 is 40m, it is convenient to take L, = Born which is
equal to the fiber length.
F'rom Eqns 2.10 and 2.11, 1-way propagation in MMF produces modal dis-
persion of ol= 0.66ns. and 2-way modal dispersion increases t o clfi = 0 . 9 3 ~
This limits the spatial resolution to

Also, the use of 550 f 20nm and 600 f 20nm interference fiiters t o isolate SP.4
and RU fluorescence is equivalent to having 40nm spectral source which produces
a substantial materiai dispersion according to Eqn 2.12. Then, using Eqn 2.9,
the total dispersion for fluorescent signai adds up to d ? ~ : .
55Onm (SPA) 600nm (RU)
0.6611s 0.661~

Fortunately, the total dispersion of SPA and RU signais are insignificant com-
pared to their lifetimes of about 30ns and 350ns.
Maximum scattering takes place when the excitation is continuous and con-
stant, because the entire length of fiber acts as scattering source. And maximum
reflection occurs when the fiber end-face is clean and perpendicular. If the ex-
citation input z(t) = 1 is launched into MMF, then the steady-state output
coming back is y(t) = h(t)* x(t) = H(0). Using Eqns 2.19 and 2.22, the Fkesnei
and Rayleigh components of the output are

When the fiber tip is held in air (no = 1.0), kesnel refiection is about lOdS
above Rayleigh scattering. When the fiber tip is imrnersed in aqueous solution
(no = l.33), Fresnel reffection falls below scattering; but, this is irrelevant
because fluorescence is measured a t different wavelength. Fresnel reflection from
the far-end and Rayleigh scattering are depolarized for al1 practical purpose
and are split 50/50 by BS. However, Ftesnel reflection from the near-end is stili
polarized and passes straight through BS.
Because the laser excitation has shorter penetration in RU solution than in
SPA, the RU fluorescence is "closer" and has better coupling t o the fiber end-
face. Nonetheless, QY must be included when calculating the total amount of
fluorescence that is collecteci. Using no = 1.33 and b = a/tanom = 168pm in
water, the total collection efficiencies for SPA and RU fluorescence become
 Figure 6.5: Chernical formuia for SPA (Mt) and RU (right)

I'here is about 5dB clifference between SPA and RU signais. Considering that
fiber attenuation, diffraction efIiciency, AOM attenuation, flter attenuation,
and PIN diode responsitivity are ail A-dependent, mostly in RU'S favour, the
final signals shouid be of similar magnitude. Fhrthermore, had 1 used double
fiber configuration in my experïment, the fluorescent signais would be 9-13dB
lower since G(L,1.44~)= 0.0023 for SPA and G(L,1.44a) = 0.0078 for RU.
Comparing the collection efficiency of fiesne1 refiection (RL)and fluores-
cence (QYrl,)at the far-end, the Fkesnel backreflection is 23dB higher than SP.4
fluorescence and 27dB higher than RU fluorescence. This is consistent with ex-
perimental observation of few p W for F'resnel signal and few n W for both SPA
and RU signals.
Yeliow-orange luminescence pl61 is obsemed from the fiber's plastic buffer.
Fortunately, even though this luminescence can be seen easily by naked eyes?
its contribution at 550nm and 600nm is below the noise flmr of PIN detector.
This is due to very poor evanescent coupling between the cladding source and
the propagating modes in the core.

6.2.4 SPA and RU

SPA and RU compounds were selected based on their single exponential decay,
long lifetime, visible absorption and emission spectnun, large X shift, high QY,
stable photochernistry, easy preparation, and low cost. While SPA is a reguiar
fluorescent compound, RU is fundamentally a phosphorescent compound whose
emission is greatly enhanced by spin-orbit coupling due to heavy-atom pertur-
bation. Structurally, RU is &-transition metal cornplex with 2,2'-bipyridine
ligand in octahedral configuration. Pertinent transitions in the visible spectrum
are met al-teligand charge-transfer states (MLCT) which are formed when e-
is excited from metal's d-orbitai to ligand's n'-antibonding orbital. Referring
to Fig 3.1, the absorption state is l (MLCT) and the emission state is the spin-
forbidden (MLCT). (B131
Chernicd formula for SPA and RU are shown in Fig 6.5, and their absorption
and &ion spectnun are included in Fig 6.6. The SPA and RU solutions are
prepared as air-saturated aqueous solutions at room temperature, and in rather
heavy concentration to maximize the fluorescence. Some pertinent material
data are [Probes]
Figure 6.6: Absorption and emission spectnim for SPA (top) and RU (bottom)
 Figure 6.7: Calibration data for LO's voltage-to-frequency conversion

molecular mass
concentration c
E , a at 458nm

QY
7

LO is basically a voltage controlled oscillator, converthg input voltage (V)

to output frequency (f). The calibration data for this conversion is recorded
rnanually using lodcin amplifier' and fiequency counter2, and is plotted in Fig
6.7. Linear regression gives [Guttrnan]

with f in p z ] and V in (VI. A stable and accurate frequency generation is

essentid to my correlation scheme; regrettably, LO exhibited some thermai drift
which Limits the degree of accuracy that can be achieved in m y experhent.
Since LO frequency is swept M M H z every second, the modulation frequency
that the laser excitation experiences in the first pass is different fiom that the
returning signal experiences in the second pass due to propagation delay At =
2Lnlc. For 50m fiber, the difference is A f = (4MHz/ls)0.5ps= 2Hz. This beat
frequency is ciifficuit to deal with because it appears as slow swing of the entire
signal baseline.
'Stanford R e s e a d System SR530
2Ftuke 1900A
6.2.6 PIN Detector
Fig 6.8 shows the schematic diagram of detector that 1 used in the experiment.
Essentially, it is DGcoupled lowpans transimpedance amplifier with the PIN
diode in photovoltaic mode (OV bias) so that diode m e n t (1)is proportional
to optical power (P),[JPnkins]

where RA = 0.2-0.3A/W is the respoasitivity of diode for 45û-ôOOnm. The

ha1 output signal becomes

=
By trial and error, the upper limit of fo, f3 160Hz seemed sufficient to filter
out noise without excessive smoothing. Bowever, the circuit in Fig 6.8 is "home-
maden; and, as such, the effects of improper grounding and stray capacitance
are unavoidable even under the best of circumstances.
Unfortunately, after al1 the experiments were done, 1 realized that detector
in Fig 6.9 wouid have been h p l e r and better. It is AC-coupled bandpass circuit
with the final output signal

Since the correlation output is repeated every ramp intervai, this ramp frequency
constitutes the lowest kequency component of output signal. Also, the output
sits on top of large DC bias. The bandpass detector, with low cut-off (fi) set
to the ramp frequency and with high cut-off (fo) set sufficiently high but not
too high, would filter out the large DC bias, slow drifts, and fluctuations which
plagued my expriment.

6.3 Experimental Procedures

Preparations involving sample solutions, fiber, laser, LO, and AOM alignments
are summarized as foUows:
SPA/water (458pM)and RU/water (468pM)solutions are prepared using
laboratory grade water.
Both ends of approximately 50m long MMF are stripped of plastic buffer
with P o l y s t r i p p ~cleaned
, with acetone, and cleaved with a fiber cleaver.
Since fiber coupling depends critically on the quality of end-face, it is
important to obtain a cut that is flat and perpendicular at l e s t across
the fiber core.
The Ar laser output is set to 15mW at 458nm.
Ramp generator is set to ramp W V at every 1 second interval, and LO
converts this to a correspondhg frequency ramp 0-4- every second.
Fig 6.10 (top) illustrates the aligning of AOM for the f i s t p a s . The Ar
laser beam is focused by lem so that the spot size a t AOM is within its
active aperture. The fiber coupler is adjusteci for maximum coupling into
MMF by monitoring the optical power corning out of the far-end.
Figure 6.8: DC-coupled Iowpass detector, using blue enhanced PIN diode and
low-noise J-FETinput Quad Op-Amp

1-teV converter Amplifier DC remover Lowpass filter

vo = -ROI Vz
1 + if lfo 1 +if l f 3

Figure 6.9: Simpler AC-coupled bandpass detector

1-to-V converter 1 AC coupling 1 Amplifier

Figure 6.10: Aligning AOM for first pass (top) and second p a s (bottom)

+ laser
 Fig 6.10 (bottorn) illustrates the aligning of AOM for the second pass. .Ar
laser is Iaunched from the far-end to simulate and make visible the path
of retuming signal. The fiber coupler is adjusted until the focused spot
a t AOM is within the same aperture that the first pass went through,
thereby maintaining the same Bragg angle. Basicaiiy, AOM is geometri-
cally aiigned for optimum performance at 458nm;the subsequent decrease
in performance at 550nrn and 600nm is accepted as a compromise in return
for easier alignment.
The experimental procedures of a c t u d y measuring Fkesnel backrefiection
(for fiber Iength) and fluorescent response (for SPA and RU lifetimes) are as
follows:
1. Decide on the type of measurement. When measuring Fkesnel reflection,
the fiber tip is held in air. When measuring fluorescent response, the
fiber tip is immersed in SPA or RU solution as shown in Fig 6.11. This
maxirnizes the collection efEciency, and represents the simplest configura-
tion because there are no external components to aiign. Of course, when
changing hom one measurement to another, the fiber tip must be washed
t horoughly with distiUed water-
2- To reduce contamination of signal, 460 f lOnm interference filter is placed
in front of PIN detector when measuring Fkesnel reflection, 550 f 20am
filter when measuring SPA fluorescence, and 600 f 20nm filter when mea-
suring RU fluorescence.
3. The gain, reference bias, and bandwidth of PIN detector, desaibed in Fig
6.8, are detennined by trial and error. Generdy, F'resnel signal was few
pW,and SPA and RU signals were both about few nW.
4. Fiilally, both the ramp input and the correlation output are sampled with
digitking oscilloscope using built-in exponential averaging algorithm. The
data set is then uploaded to computer for data anaiysis and ensemble
averaging. Raw data for Fresnel reflection, SPA fluorescence, and RU
ffuorescence are shown in Figs 6.12, 6.13, and 6.14, respectively.
5. So f a t the opticai fiber is nomindy 50m long. In order to assess the
resolution in fiber Iength measurement, 1.01m is cleaved off the far-end,
making the fiber 49m long nominally. F'resnel reflection is then measured
by repeating steps 1-4. A raw data for this is shown in Fig 6.15.
1 tried to collect as many sets of data as possible within the time allocated
by OLLRC lab. Unfortunately, 1 managed to coiiect only 35 data sets for 50m
fiber, 16 data sets for SPA fluorescence, 13 data sets for RU fluorescence, and 10
data sets for 49m fiber. Collecting more data sets will help, of course. But a fun-
damentai improvement in the accuracy and confidence of parameter estimation
can only occur when better quality equipments are used.

6.4 Data Analysis

When the bottom graphs in Figs 6.12, 6.13, and 6.14 are compared to the top
graphs in Figs 5.3, 5.4, and 5.5, one can see that there is very good corre-
spondence between the experimental measurements and the calculated models.
Using Student-t values for 95% confidence interval (a = 0.05), the length of
Figure 6.11: Pictures of SPA (top) and RU (bottom) fluorescence
Figure 6.12: Raw data for Resnel backrefiection from 50m fiber. The voltage
input (top) to LO and the correlation output (bottom) from PIN detector are
shown for one complete ramp period. The excitation waveform was square-wave-
Figure 6.13: Raw data for SPA fluorescence using 50m fiber. The voltage input
(top) to LO and the correlation output (bottom) from PIN detector are shown
for one complete ramp period. The excitation waveforrn was square-wave.
Figure 6.14: Raw data for RU fluorescence using 50x11fiber. The voltage input
(top) to LO and the correlation output (bottom) from PIN detector are shown
for one complete ramp period. The excitation waveform was square-wave.
Figure 6.15: Raw data for Fresnel backreflection fkom 49m fiber. The voltage
input (top) to LO and the correlation output (bottom) from PIN detector are
shown for one complete ramp period. The excitation waveform was square-wave.
50x11 and 49m optical fibers are estimated to be

The 0.27m uncertainty for LS0 is close to (rms)modal dispersion limit of 0 . 1 b

in Eqn 6.2,and represents almost 2 orders of magnitude improvement over the
DFT resolution of 13m in Eqn 5.20. Unfortunatdy, the tme merence of 1.01m
between 50m and 49m fibers lies outside the confidence interval.
The lifetime of SPA and RU fluorescence are estimated to be

-
6.64
(SPA)r = 28.3 f 3.5ns t- (tls:o.ozs)
rn (6.10)

-
46.7
(RU)r = 360 f 28ns t (ti2;o.ozs)
JT3
(6. i l )

The SPA and RU lifetimes were also determined in time domain, using a con-
ventional pulse excitation setup in Chernical Sensor Group at Erindale campus
of University of Toronto. The excitation source was N p laser with 10x1spulse
width at 337m. Fluorescein was used as reference because it exhibits a stable
single lifetime of 3.8-4.3ns [A7, A81 over the a i o n spectrum of interest. The
fluorescent decays are shown in Fig 6.16 for SPA-Fluorescein pair and in Fig
6.17 for RU-Fluorescein pair. Setting the reference lifetime at 4.0ns in Eqn 3.8,
the estimated Metimes are
(SPA)r = 2 9 . 2 ~f~0.1%
(RU)r = 379ns 90.1%
which are consistent with Eqns 6.10 and 6.11,as weH as, with the d u e s cited
from Iiterature in Section 6.2.4.
Figure 6.16: Fluorescence decay of SPA sample y ( t ) and Fluorescein reference

O tirne index i [0.2ns/step) 997

Figure 6.17: Fluorescence decay of RU sample y ( t ) and Fluorescein reference

tirne index i [2.0ns/step]

Chapter 7

Conclusions

In my thesis, I demonstrate a fibre optic sensor for measuring fluorescence life-

t h e , in which the main signal processing is performed entirely by optical de-
vices. It has severai distinct characteristics and advantages when cornpared to
the conventional time dornain and frequency domain fluorometers:
A single optical fiber is used to c q the laser excitation and to coUect
fluorescent response or Fresnel backrefiection. This gives remote mea-
surement capability which is important for the development of fibre optic
sesors.
The optical signal passes twice through AOM which acts as an optical
bidirectional mixer. Since the double modulation occurs at almost same
frequency, the signai processing is basically a homodyne scheme. Be-
cause AOM is the only active component, system bandwidth can be easily
upgraded by replacing it with another opticai modulator of higher band-
width.
The DC output from the PIN detector becomes auto or cross correlation.
The detector 's bandwidth is independent of modulator 's bandwidt h due
to somewhat unusuai change of domain brought about by the correlation
operation. Because only DC component from AOM mixing is measured,
the PIN detector does not have problems associated with wide bandwidth
opticai detection and high frequency electronic signal processing.
Since the modulation frequency of AOM is rampeci, the input and output
data are repeated a t every ramp period and are measureù as functions of
frequency.
Continuous wave excitation has lower peak power than pulse excitation,
so that there is l e s likeiihood of damaging a sample.
0 Although the experiment uses buik components, the system is simple
enough to be integrated into a single "in-line" device.
hthermore, I address theoretical problems involved in using optical fiber to
excite and coilect fluorescence from remote medium. The fluorescence coupling
aciency of multimode and singlemode fiber is derived, and the waveguide d-
fect of opticai fiber on the propagating signals is anaiyzed in detail. 1 hope
that the foundation laid in this thesis heips reader to understand better the
workings of optical fiber sensor, whether it is for measuring temperature, pH,
O2 concentration, or even epoxy resin cure [See Appendix BI.
 Fkom experiment, 1 have successrully dete&ed the fluorescent lifetime
of N-(3-sulfopropyl) acridinium (SPA) and rutheniwn tris-bipyridyl dichloride
(RU), and the fiber length of mdtimode optical fiber sensor. SPA and RU com-
pounds were chosen because they have a stable single Metime, visible spectrum,
and relatively high quantum yield; and multimode fiber was chosen because
it has large core which facilitates better coupling. The main results of data
analysis are as follows:

My correlation scheme gives an order of magnitude improvement over the con-

ventionai DFT method. The uncertainties represent the usual 95% confidence
intervai for ensemble average. UtimateIy, the resolutions are limited by optical
fiber dispersions; o d y modal dispersion is present in fiber length measurement,
but both modaI and materiai dispersions are present in fluorescent iifetime mea-
surement.
The lifetimes are consistent with literature values and with the values decon-
voluted from pulse excitation method. Unfortunately, however, the fiber length
shows poor intemal consistency, since the actual 1.01m that is cleaved off the
far-end is estimated as 1.75 f O.Mm a t 95% confidence. This inconsistency is
attributed to the drift and fluctuation of modulation frequency coming from
LO.
The coupling efficiency of F'resnel backreflection is a t least lOdB greater than
Rayleigh backscattering, and 23-27dB greater than SPA and RU fluorescence.
This is consistent with experimental observation of few pW for Fresnel signal and
few nW for both SP.4 and RU signals. In coliecting SPA and RU fluorescence,
as expected, the single fiber configuration is 9-13dB more efficient than double
fiber configuration.
AOM is the key component in my correlation scheme, because it performs the
modulation of laser excitation in the k t pass and the opticai mixing necessary
for auto and cross correlations in the second p a s . Using one AOM for bath
functions allows the signai processing to be done as much as possible in opticai
domain with minimum extemal components. Diffraction efficiency and Bragg
angle depend on the input wavelength. Due to phase-mismatch, the Bragg
diffraction decreases fkom 90.5% efficiency for laser excitation (458nm)to 61%
efficiency for SPA fluorescence (550am) and to 35% efficiency for RU fluorescence
(600nmj.The efficiency decreases even more if optical divergence is considered.
The merits of my fibre optic fluorescence lifetime sensor become obvious
when one realizes that an experirnentai setup with only 4MHz bandwidth is Ca-
pable of measuring fluorescent lifetime of 28.3 f 3.5ns and 360 f 28ns. Shorter
lifetimes can be measured by simply upgrading LO and AOM to a higher band-
width, with no changes to the other components, experimentai Iayout, or data
analysis. Although the 4MHz bandwidth is rather low for practicd applica-
tions, it is suficient for the purpose of my thesis which is to demonstrate a new
concept in optical fiber sensor, fluorescence lifetime measurement, and optical
signal processing.
Appendix A

Formula and Tables

A. 1 Definitions

A.2 Impulse F'unction

The impulse hinction d(t) has the foliowing properties: [Brigham]

d ( t ) = 0, t#O
L' d(t) = 1
lald(at) = d(t)
h(t)d(t- a ) = h(a)d(t- a)
d(t - a ) * d ( t - 6) = d ( t - ( a + b ) )

A.3 Solid Angles

Consider a conical aperture dong i axis with ha-angle 8,so that dS is diner-
entiai sphtxicai area normal to the radius vector ilO is angle between i and i,
and dS cos 8 is dinerential plana area normal to i. The sphe7ifol solid angle of
aperture is

and the planor solid angle is defineci as [Wyatt]

Ir
cos 8
n = ~ d-S=
r2 Je
@=O
J
4=0
(rd8)(rshfld#)cae=nSin28
r2
(A.14)

A.4 Table of Fourier Transform Pairs

The foiiowing Fourier Traasform pairs

have been selected fkom Refç prigham, Bracewell]. If a function is d&ed

with conditional domain, then it is zero outside the domain and is equai to the
midpoint at the discontinuity.

(A. 16)
(A.17)
(A. 18)
(A.19)
(A.20)
(A.21)
(A-22)
(A.23)
(A.24)
(A.25)
(A.26)
(A.27)
(A.28)
+
(even) j ( 0 )
(od4 + j(O) (A.29)
+
(0) j(even)
(0) + j ( o W
(A.30)

(A.31)

(A.32)

(A.33)

(A-34
 n(t) A sinc(f) (A.35)
~ ( t )A sinc2(f) (A.36)
cos(2irat) A J(f - 4+6(f +4 (A.37)
2
sin(2irat) & d(f - a) - 4 f + 4 (A.38)
22
7
t >O
ët/'. A 1+j2rfr
( A.39)

e-ltl/r & 27 (A.40)

1 + ( 2 s fr ) 2
e-rt2 e-zrf2 (A.41)
e-t2/az afie-r2a2 f2 (A.42)
6(t - A t ) + d(t + At)
2
& cos(2sfat) (A.43)
1
-e-(ltl-At)/rl ~tl> ~t COS ( 2 sf At + tan-' ( 2 sf T))
(A.44)
2r JI+ ( 2 sf T ) ~

A.5 Decomposition into Fourier Series

Consider a waveforrn z ( t ) which consists of ( h ( t ) ,Itl < T / 2 ) repeated at every T
interval or at frequency 1/T. The time domain expression and the corresponding
complex Fourier series are [Brigham]

Applying FT to both sides and solving for on,

If h ( t ) is real and even, H(f ) is &O reai and even. Then, the complex series
becomes a cosine series,

r ( t )= 0 0
00

+ 12 0 , cos
n= 1
(F)
A.6 rms Width
bot-mean-square (rms)width of function h(t) is definecl as [Bracewell]
-
c2 = (t2) (t)* (A.50)
where

(t)= =- W O )
/* h(t)dt 27rjH (O)
 (F)=
1-_ - -

Wt)dt
= - H"(0)
$' h(t)dt 4n2H (O)

The rms width of few cornmon functions are

A.7 Fiber-to-Fiber Coupling

Coupling &ciency ( q ) fiom MMFl to MMF2 is important not only for splices
but also for many fiber-tip sensor configurations. Using subscrïpt I and 2 to
denote the parameta of interest from respective fibers, (Ungar, Ettenj

diameter 2a:
.=-.{$ l}
aperture M:

index profile g:
.=min
1 + 2/92 '
core index n:

axial gap z:
a2
ifg=oo
'= ( a + r t a n e n u ) Z p
radiai shift 2r:
anguias tilt 28:
#-sin@
rl= A
, ifg=oo
_4- -4sin4
- - +-sin24 ifg = 2
R 3n 67r '
where
COS
d =-
-
2
r
a
or -sin 0
sin gW
Appendix B

Fluorescence Application in
Cure Monitoring

B.l Epoxy Resin and Composite

4
The term epozy refers to molecule containing -&c- ring structure. The cure
process breaks open the epoxy rings and converts epoxy monomers into network
of crossIinked polymers which is characterized [Ency] by high adhesive strength,
low sbrinkage, good mechanical and thermal properties, high chemical and cor-
rosion resistance, and good electrical insulation. To bring about some optimum
characteristics, most commercial epoxy reçins contain proprietary mixture of
epoxies, curing agents (hardeners), accelerators, diluents, Mers, flexibilizers,
and impurities (dyes, solvents, etc.) [Ency, Lubin]. The generic chemical for-
mula of 2 popular epoxies and curhg agents are shown in Figs B. 1 and B.2, for
which the ring opening reaction occurs mainiy between the tenninal glycidyl
group and the primary amine group as foilows:

One important application of epoxy resin is in composite [Ency] where strong

and stifF fibers (usualiy glass, boron-tungsten, graphite, or Kevlarl) are digned
in epoxy rein rnatrix. Most composites are manufacturecl by baking in an oven
under pressure. Temperature provides heat needed to initiate and maintain the
epoxy cure reactions, and pressure compacts the fiber matrix by squeezing out
excess resin and gas pockets. Typically, the cure cycle follows a &ed time-tem-
perature-pressure profile [08], even in the presence of chemical variation in resin
mixture, unequal distribution in composite laminates, ciifference in preparation
or storage, and other practical situations that corne up in the real world.
The fixed cycle is simple and cheap to implement, but the quality of com-
posite varies from batch to batch. Obviously, the composite daes not attain its
optimum properties when the epoxy resin is not M y cured; fihermore, m e -
acted epoxy resins serve as sites [O131for crack initiation and propagation. The
quaiity can be improved by using some feedback control which requires monitor-
ing of parameters (temperature, pressure, viscosity, capacitance, conductance,
'Kevlar, also k n m a P P T poly(pphenyleneterephthalamide),
~ ~ ia trademark of
Dupont
Figure B.1: Generic formula of 2 epoxy reins: (top) DGEBA or digIyudyl
ether of bisphenol A, and (bottom) TGDDM or tetraglycidyl diamino diphenyl
methane

Figure B.2: Generic formula of 2 curing agents: (top) DDS or diamino diphenyl
sulfone, and (bottom) DDM or diamino diphenyl methane

acoustic impedance, refractive index, IR absorption, fluorescence, etc.) that af-

fect or indicate the state of cure.

B.2 Fluorescence during Cure

If A and k are, respectively, rates of radiative and non-radiative decays fiom
an emitting state of molecule, the quantum yield (QY) and the lifetime (r)of
fluorescence are defined as

The radiative decay remains fairly constant, but the non-radiative decays, such
as rotational and vibrationai relaxations, are idluencecl by the molecular en-
Wonment. A t the beginning, an uncureù epoxy resin exhibits high rnolecular
mobility which means fast k, low QY,and short T. After full crosslinking poly-
merization, a cured epoxy rein exhibits low moIecuIar mobility which teads to
slow k, high QY,and long T .
If the conversion from uncureci to cured state occurs hornogeneously through-
out the rein, then QY and T would increase smoothly during cure. However,
£rom phosphorescence studies [O10, 011, 0121, ESR spectroscopy [O121, and
electron and optical microscopy [013],the initial formation of network and the
subsequent growth in crosslinking are found to be nucleation process, in wbich
the cure proces occurs heterogeneouslp or locally with distinct boundary be-
tween the 2 states. Therefore, QY increases because the cured domain grows
at the expense of the uncured domain, but T remains constant because the
underlying nature of A and k within each domain do not change.
B.3 Cure Monitoring
Usually, one ad& a dye or probe to an epoxy resin mixture, as long as it does not
interfere with the cure reactions and does not overiap with other fluorescence
aiready present. In some cases, one can take advantage of intrinsic fluorescence
from epoxy resin itseif- So far, the foiiowing mechanisms or parameters have
been used for cure monitoring:
rn increase in QY which depends on local viscosity [08,031, 034, 0391
a çhift in emission spectrum [039, 0561
a increase in QY of reactive probe that has similar structure and reactivity
as the curing agent 108, 031, 0551
rn decrease in excimer fluorescence which depends on local "fret? volumen
[08,0311
rn increase in fluorescence polarization which depends on rotational diffusion
[O311
a longer recovery tirne after photobleaching due to slow diffusion [O541
An accurate interpretation of data is diEcult because of impurities and complex
chernical reactions. But, fluorescence method is uniquely suited for fibre optic
application [039,055, 0561, since typicai fluorescent spectrum of near-W to
near-IR wavelength is easily supported by many optical fibers.
Even though the fluorescent lifetime mechanism does not change during
epoxy resin cure, if r is measured using a fluorescent Lifetime sensor that assumes
single lifetime, as 1 have done in demonstrating my experiment, then T wiU
appear to increase during the cure, because the numerical algorithm cannot
Werentiate signals from 2 different domains and simply follows the growing
(shrinking) contribution from the cured (uncureci) domain. Consequently, the
fluorescent lifetime can be used for cure monitoring of epoxy resin.
A more precise analysis requires distributed lifetime mode1 where one as-
sumes that varying molecular environments produce a distribution of fluorescent
Lifetimes. As the molecular compositions changes with cure, the profile of Me-
time distribution changes accordingly. However, this requires more cornplicated
data analysis than the one 1 used in my experiment.
Bibliography

[Barnoski] Er1 K Barnoski (ed),Fundamentah of Optical Fiber Communications,

1981, chapter 1, D B Keck, "Optical fiber waveguide"
[Bass] M Bass, E W van Stryland, D R Williams, W L WoIfe, Handbook of
Optics (2nd ed), 1995, vol 2, chapter 33
pirks] J B Birks, Photophysics of Ammatic Molecules, 1970, chapter 4
[Bracewell] R N Bracewell, The Fou* 2"knsfonn and Its Applications, 1985,
chapter 8
[Brigham] E O Brigham, The Fwt FouRer Thnsfonn and Its Applications,
1988, chapters 2-5, 14, appendix A
[Cheo] P K Cheo, Fiber Optics: Devices and Systems, 1985, chapters 4, 5
[CRC] M J Weber (ed), CRC Handbook of Laser Science and Technology, 1982,
vol IV, p 328
[Cuishaw] B Culshaw, Optical Fibre Sensing and Signai Pmcessing, 1984, sec-
tion 8.2
[Ency] H F Mark, N M Bikales, C G Overgergev, G Menges, J 1 Kroschwitz,
Encyclopedia of Polymer Science and Engineering, 1985; "Compositen vol
3, pp 776-820; "Epoxy Resin" vol 6, pp 322-382
[Etten] W van Etten, J van der Plaats, Fundamentais of Optical Fiber Commu-
nications, 1991, chapter 11
[Ghatak] A K Ghatak, K Thyagarajan, Optical Electmnics, 1989, chapters 13,
16-19, appendix F
[Goutzoulis] A P Goutzoulis, D R Pape, (ed), Design and Fabrication of
Awwto-Optic Deuices, 1994, chapter 3, R V Johnson, "Design of acousto-
optic modulators"
[GowarJ J Gowar, Optical Communication Systems, 1984, chapters 2, 6
[Guttrnan] 1 Guttman, S S Wilks, J Stuart Kunter, Introductory Engineering
Statistics, 1982, chapters 9, 15
[Jenkins] T E Jenkins, O p t i d Sensing Techniques and Signol Pmcessing, 1987,
chapters 1, 2
[Korpel] A Korpel, Awusto-Optics, 1988, chapters 2, 3, 6
[Lakowia] J R Lakowin, Pnnciples of Flvomcmce Spectroswp y, 1983, chapter
3
[Lubin] G Lubin (ed), Handbook of Composites, 1982, chapter 5
(Adagdich] L N Magdich, V Y Molchanov, Acowto Optics Deuices and Th&
Applications, 1989, chapters 1, 2
parcuse] D Marcuse, Light Bunmnission Optics, 1982, chapters 3 - 8 , 11, 12

press] W K Press, B P Flannery, S A Teukolsky, W T Vetterling, Numen-col

Recipes in Ci me Art of Scientafic Computing, 1988, sections 4.3, 4.4,
10.2, 14.5
probes] Molecular Probes Inc., Handbook of Fluorescent Probes and Research
Cherniculs, 1989-1991
[Snyder] .A W Snyder, J D Love, O p t i d Waueguzde Tlieory, 1983, chapter 15
[Ungarj S Ungar, Fiber Opticsr Theory and Applications, 1990, chapters 2, 5, 7
Wyatt] C L Wyatt, Radiometric System Design, 1987, chapter 3
[yariv] A Yariv, P Yeh, Optical Waves in Crystals, 1984, chapters 9, 10
F V Bright, "A new fiber-optic-based rnultifkquency phase-modulat ion
fluororneter", Applied Spectroscopy, vol 42, no 8, pp 1531-1537, 1985
J R Lakowia, B P Maliwal, "Construction and performance of a variable-
fiequency phase-modulation fluorometer", Biophysical Chemistry, vol 21,
pp 61-78, 1985
R B Thompson, E Gratton, "Phase fluorometric method for determina-
tion of standard lifetimesn, Andytad Chemistry, vol 60, no 7, pp 670-
674, 1 Apr 88
R Krapf, N P IUsley, H C Tseng, S Verkman, "Structure-activity rela-
tionships of chloride-sensitive fluorescent indicators for biological appli-
cation", Analytical Biochemistry, vol 169, pp 142-150, 1988
S W Snyder, J N Demas, B A DegraE, "Single photon counting lifetime
measurements of weak long-lived sampiesn, Andytical Chemistry, vol 61,
no 24, pp 2704-2707, 15 Dec 89
J N Demas, R A Keller, "Enhancement of lriminescence and raman spec-
troscopy by phase-resolved background suppressionn, Andyticul Chem-
istry, voI 57, no 2, pp 538-545, Feb 85
M J Cook, A P Lewis, G S G McAuliffe, V Skarda, A J Thomson,
"Luminescent metal cornpiexes. Part 1. ?kis-chelates of substituted 2,2'-
bipyridyls with ruthenium(1I) as dyes for luminescent solar collectorsn,
Journal of Chernical Society. Perkin Runsaction II, ppp 1293-1301, 1984
G A Coquin, D A Pinnow, A W Warner, "Physical properties of Lead
Molybdate relevant to acousto-optic device applications", Journal of Ap-
plied Physics, vol 42, no 6, pp 2162-2168, May 71
J N Demas, B A DeGrafI, "Design and applications of highly lumines-
cent transition metal complexes", Analytid Chemistry, vol 63, no 17, pp
829A-837A, 1 Sep 91
M K Corroll, R V Bright, F M =&je, "Luminescence originating in an
optical fiber", Applied Spectroscupy, vol 43, no 1, pp 176-178, 1989
[Hl] M C Hudson, uCalculation of the maximum o p t i d coupling a à e n c y
into muitimode optical waveguides", Applied Optics, vol 13, no 5, pp
1029-1033, May 74

[II21 R B Dyott, "The fibr~opticDoppler ammorneter", Micmwaves, Optiw

and Acowtics, vol 2, no 1, pp 13-18, Jan 78

[H3] S D Schwab, R L Mccreery, "Versatile, efficient rama sampling with fiber

opticsn, Analytid Chemistry, vol 56, no 12, pp 2199-2204, Oct 94

[H4] P Plaza, N Q Dao, M Jouan, H Fevrier, H Saisse, "Simulation et optimi-

sation des capteurs à fibres optiques adjacentesn, Applied Optics, vol 25,
no 19, pp 3448-3454,l Oct 86
[H6J C Komives, J S Schultz, "Fiber-optic fluorometer signal enhancement and
application to biosensor designn, TcJonta, vol 39, no 4, pp 424-441, 1992

EH71 Z Y Zhu, M C Yappert, "Determination of effective depth and equivalent

pathlength for a singlefiber fluorometric sensor", Applied Spectmscopy,
vol 46, no 6, pp 912-918,1992

[HS] Z Y Zhu, M C Yappert, "Determination of the effective depth for double-

fiber fluorometric sensors", Appfied Spectmscopy, vol 46, no 6, pp 919-924,
1992

[Hg] Z Y Zhu, M C Yappert, "Sensitivity enhancement in capillary/fiber-optic

fluorometric sensors", Andytical Chemistry, vol 66, no 5, pp 761-764, 1
Mar 94
[Hl01 E H RatalafF, R G Efarfmann, S R Crouch, "Absorption-corrected fiber
optic fluorometer", A n a l y t i d Chemistry, vol 56, no 3, pp 342-347, Mar
84
D Gloge, "Weakly guiding fibersn, Applied Optics, vol 10, no 10, pp 2252-
2258, Oct 71
V Ruddy, B D ~MacCraith,J A Murphy, "Evanescent wave absorption
spectroscopy using multimode fibersn, Journal of Applied Physics, vol
67, no 10, pp 6070-6074,15 May 90
D Marcuse, "Launching light into fiber cores fiom sources tocated in the
cladding" , Journal of Lightwave Technology, vol 6, no 8, pp 1273-1279,
Aug 88
Z M Hale, F P Payne, "Fluorescent sensors baseci on tapered single-mode
optical fibresn, Sensors and Actuators B, vol 17, pp 233-240, 1994
C O Egdon, R S Rogowski, "Theoretical mode1 for a thin cylindrical film
optical fiber fluorosensorn, O p t i d Engineering, vol 31, no 4, pp 237-244,
Feb 92
C O Egalon, R S Rogowski, "Efficiency of core light injection £rom sources
in the cladding: bulk distribution", Opt i d Engineering, vol 3 1, no 4, pp
846-851, Apr 92
R B Thompson, L Kondracki, "Sensitivity enhancement for evanescent
waveexcited fiber optic fluorescencesensors", SPIE 1204: Time-Resolved
Laser Spectmscopy in Biochemistry II, pp 35-41,1990
G P Anderson, J P Golden, F S Ligler, "An evanescent wave biosensor -
Part 1: Fluorescent signal acquisition from stepetched fiber optic probes",
IEEE 7hnsaction on Biomedid Engineering, vol 41, no 6, pp 578-584,
Jun 94
J P Golden, G P Anderson, F S Ligler, "An evanescent wave biosensor -
Part II: Fluorescent signal acquisition kom tapered fiber optic probes",
IEEE ïhnsaction on Biomedid Engineering, voI 41, no 6, pp 585-591,
Jun 94
E G Neumann, "Analysis of the backscattering method for testing optical
fiber cablesn, AEÜ: Electmnicits and Communimtions, vol 34, no 4, pp
157-160, Apr 80
A H Hartog, M P Gold, "On the theory of backscattering in single-mode
optical fibersn, Journal of Lightwaue Technology, vol 2, no 2, pp 76-82,
Apr 84
E Brinkmeeyer, "Analysis of the backscattering method for single-mode
optical fibersn, Journal of the Optical Society of America, vol 70, no 8,
pp lOl(t1012, Aug 80
Y Ueno, M Shimizu, "Opticai fiber fault location method", Applied Op-
tics, vol 15, no 6, pp 1385-1388, Jun 76
F P Kapron, B P Adams, E A Thomas, J W Peters, "Fiber-optic re-
flection measurements using OCWR and OTDR techniques", Journal of
Lightwave Technology, vol 7, no 8, pp 1234-1241, Aug 89
H Ghafoori-Shiraz, T Okoshi, "Opticai frequency-domain reflectometry" ,
Optical and Quantum Electronics, vol 18, pp 265-272, 1986
M Shadaram, W L Kuriger, "Using the opticai fiequency domain tech-
nique for the analysis of discrete and distributed reflections in an optical
fibern, Applied Optics, vol 23, no 7, pp 1092-1095 Apr 84
B L Danielson, "Optical tirne-domain reflectorneter specifications and
performance testing" , Applied Optics, vol 24, no 15, pp 2313-2322,l Aug
85
1 Fujima, K Seta, H Matsumoto, T Ohishi, "GHz traveling-wave opticai
modulator for precision distance measurement" , SPIE 889: Airborne and
Spaceborne Lasers for Temstnal Gwphysical Sensing, pp 86-90, 1988
S Nakayama, "Development of a microwave electro-optic modulator for
a laser distance-measuring instrumentn, Japanese Journal of Applied
Physics, vol 26, no 12, pp 2102-2106, Dec 87
L Thevenaz, J-P Peliaux, "Group delay measurement in singlemode
fibers with true picosecond resolution using double optical modulationn,
Journal of Lightwave Technology, vol 6, no 10, pp 147û-1475, Oct 88
D W Dolfe, M Nazarathy, "Optical frequency domain reflectometry with
high sensitivity and resolution using optical synchonous detection with
coded modulators", Electronics Letters, vol 25, no 2, pp 16&162, 19 Jan
p l 6 1 M Izutni, H Haga, T Sueta, "Picosecond signal s a m p h g and multiplica-
tion by using integrated tandem light moduiatorsn, Journul of Lightwoue
Technology, vol 1, no 1, pp 285-289, Mar 83
p l 7 1 H Nakatsuka, Y Katashima, K Inouye, R Yano, "Tncoherent light gated
optical kerr shutter for fluorescence lifetime measurement" , Optics Com-
municutions, vol 69, no 2, p 169, 15 Dec 1988
p l 8 1 J Nakatsuka, Y Katsahima, K houye, R Yano, S Uemura, "Fluorescence
Wetime measurement of semiconductor doped glass by using an incoher-
ent light gated optical k e n shutter", Optics Communications, vol 74, no
3,4, p 219, 15 Dec 1989
p24] D W Piston, F Marriott, T Radivoyevich, R M Clegg, T M Jovin, E
Gratton, "Wide-band acousto-optic Iight modulator for frequency domain
fluorometry and phosphorixnetry", Reuiew of Scientific Instruments, vol
60, no 8, pp 2596-2600, Aug 89
[N25] U Naoya, at el, "Acoustwptic deflection materiais and techniquesn Pm-
ceedings of IEEE, vol 61, no 8, p 1073, Aug 73. Also reproduced in A
Korpel (ed), Selected Papers on Acousto-Optics, SPIE Milestone Series,
vol MS16, p 530
[O81 G A George, "New approaches to the characterization of the cure of epoxy
resins for advances composite materiais", Matmial Fomm, vol 9, no 4, pp
224-236, 1986
[O101 E W Meijer, "Polyepoxides: Formation and properties of their network
structure", Phalips Technology Rern'ew, vol 44, no 4, pp 110-121, Sep 88
[O111 E W Meijer, J M Zwiers, "Photoinitiated cationic polymerization: A
study to the polymerization and molecular mobility of polyepoxide net-
works using phosphorescence spectroscopyn, B Sedlacek, Kahovec, (ed),
Crosslinked Epoxies, 1987, pp 27-40
[O121 E W Meijer, D M De Leeuw, F J A M Greidanus, R J M Zwiers, "Molec-
ular mobility studies in polyepoxides obtained by photoinitiated cationic
polymerization", Polymer Communications, vol 26, pp 45-47, Feb 85
[O131 R J Morgan, "Structure-property relation of epoxies used as composite
matricesn, Advances in Polymer Science, vol 72, pp 32-33, 1985
[O311 H Itagaki, et al., "Luminescent probe studies of solid polymersn, Pm~ress
in Polymer Science, vol 15, no 3, pp 408-414, 1990
[O341 R L Levy, D P Ames, "Monitoring epoxy cure kinetics with a viscosity-
dependent fluorescent proben, L H Lee (ed), Adhesive Chemistry: Devel-
opments and ïknds, 1984, pp 245-256
[O391 R L Levy, S D Schwab, "Performance characteristics of the fluorescence
optrode cure sensorn, R Dickie, S Labana, R Bauer, (ed), Cross-Linked
Polymers: Chernistry, Pmperties, and Applications, no 367, 1988, pp
113-121
[O541 F W Wang, E S Wu, "Cure monitoring of epoxy resins by fluorescence
recovery after photobleaching", Polymer Communications, vol 28, pp 73-
75, Mar 87
 W Dang, W H Sung, "In-situ cure monitoring of diamine cured epoxy
by fiberoptic fluorirnetry using extrinsic reactive fluorophdre", Polyner
'Engineering and Scàence, vol 34, no 9, pp 707-715, May 94
[O561 H-J Paik, N-H Sung, "Fiberoptic intrinsic fluorescence for in-situ cure
monitoring of amine cured epoxy and composites", Polymer Engineering
and Science, vol 34, no 12, pp 1025-1032, June 94

11 G Ide, Y Engelborghs, A Persoons, ''Fluorescence M e t h e resolution with

phase fluorometryn, Reuàev of Scientific Iwtnrments, vol 54, no 7, pp
841-844, Jd 83

p 2 ] E Gratton, 1 M Limkeman, "A continuously variable frequency cross-

correlation phase fluorometer with picosecond resolution" , Biophysical
Journai, vol 44, pp 315-324, Dec 83
[P3] G Mitchell, K Swift, "The 48000 MHFm, a dual-domain Fourier
Tkansform fluorescence iifetime spectrofluorometer", SPIE 1204: Time-
Resolved Laser Spectmswpy in Biochemistry il, pp 27&274,1990

[P4] K Berndt, H Durr, D Palme, "Picosecond phase fluorometry by mode-

locked cw lasersn, Optics Communications, vol 42, no 6, pp 419-422, 15
Aug 82

[P5] F V Bright, C A Monnig, G M Hieftje, "Rapid frequency-scanned fiber-

optic fluorometer capable of subnanosecond Metirne determinationsn, An-
alyticd Chemàstry, vol 58, no 14, pp 3139-3144, Dec 86

[P7] E Gratton, D M Jarneson, N Rodato, G Weber, "Multifrequency cross-

correlation phase fluorometer using synchrotron radiation", Review of
Scientific Instruments, vol 55, no 4, pp 486-494, Apr 84
es] J R Alcala, E Gratton, "A multifrequency phase fluorometer using the
harmonic content of a mode-locked lasern, Andyticai Instrumentation,
vol 14, no 3,4, pp 225-249, 1985
[Pl01 D A Wilson, G H Vickers, G H Hieftje, "Novel techniques for the deter-
mination of fluorescence lifetimesn, Andytical Instrumentation, vol 14,
no 3,4, pp 483-501, 1985
[Pll] K W Berndt, 1 Gryczynski, J R Lakowicz, "Phase-modulation fluorometry
using a frequency-doubled pulsed laser diode iight sourcen, Rem'ew of
Scientific Instruments, vol 61, no 7, pp 1816-1820, Jul 90
[Pl41 J R Lakowicz, G Laczko, 1 Gryczynski, "2-GHzfrequency-domain fluo-
rometer" , Review of Scientr'fc instruments, vol 57, no 10, pp 2499-2506,
Oct 86
[Pl51 G Laczko, 1 Gryaynski, Z Gryaynski, W Wiak, H Malak, J R Lakowia,
"A 10-GHzfrequency-domain fluorometern, Reuim of Scientific Instru-
ments, vol 61, no 9, pp 2331-2337, Sep 90

[Pl71 A T Augousti, K T V Grattan, A W Palmer, "A laser-pumped temper-

ature sensor using the fluorescent decay time of alexandriten, Journal of
Lightwave Technology, vol 5, no 6, pp 759-762, Jun 87
pl81 T VeDinh, T Nolan, Y F Cheng, M 3 Sepaniak, 3 P Marie, "rime-
resolved fiber-optics fluoroimmunosensor" , Applied Spectmswpy, vol 44,
no 1, pp 128-132,1990
pi91 C G Morgan, Y Hua, A C Mitchell, J G Murray, A D Boardman, "A
compact frequency dornain fluororneter with a directly modulated deu-
terium light sourcen, Reuàew of Scientific Instruments, vol 67, no 1, pp
4147, Jan 96
[P20]J H Marsh, G Dickson, P A CIaxton, "rime-resalved photoIuminescence
system with subnanosecond resolution at wavelengths up to 1.65&,
Electmnics Letters, vol 24, no 12, pp 744-746,9 Jun 88

[P21] M E Lippitsch, J Pusterhofer, M J P Leiner, O S Wolfbeis, "Fiber-optic

oxygen sensor with the fluorescence decay time as the information car-
rier", Andytica Charnica Acta, vol 205, pp 1-6, 1988
[PZ21 V Vadde, V Srinivas, "A closed loop scheme for phase-sensitive fluorom-
etryn, ReWev of Scientific Instruments, vol 66, no 7, pp 375Ck3754, Jul
95
 IMAGE EVALUATION
TEST TARGET (QA-3)

APPLIED - IMAGE .I n c
--
--
-
-- = 1653 Easî Main Street

-
--
--
-
,
-

- - Rochester. NY 14609 USA

Phone: 71ôi4û2-0300
F a 71W208-5989