OPTICAL LOW-COHERENCE REFLECTOMETRY FOR
NONDESTRUCTIVE PROCESS MEASUREMENTS
Summer Lockerbie Randall, Anatol M. Brodsky, Lloyd W. Burgess, and Robert L. Green
Center for Process Analytical Chemistry
University of Washington
Box 351700
Seattle, WA 98195-1700
ABSTRACT. Optical Low-Coherence Reflectometry (OLCR) is a white-light interference technique
which can be used as a sensor in a number of processing applications. Research at the Center for
Process Analytical Chemistry focuses on analysis of both transparent and highly-scattering materials.
OLCR has been applied to monitor the thickness of polymer films and clear coatings on scattering
matrices. Additional process applications include fermentation monitoring and non-invasive thickness
determination of highly scattering coatings on both conducting and non-conducting substrates.
INTRODUCTION
Optical Low-Coherence Reflectometry (OLCR) is a white-light interference
technique originally developed for applications in the optoelectronics test and measurement
arena [1,2]. The use of reflectometry has since diversified greatly, including the
development of the 2D medical imaging technique Optical Coherence Tomography [3],
Research at the Center for Process Analytical Chemistry focuses on process applications
for both transparent and highly-scattering materials. The reflectometer is a fiber-optic
based instrument in which 180° reflected light is collected as incident light passes through a
heterogeneous medium. The instrumental design is that of a modified Michelson
interferometer to detect interference fringe position and intensity, using a broadband LED
source and with the sample replacing the second reference mirror. The detection of only
coherent light allows us to extract characteristic information about the sample, such as
particle size and concentration.
By taking advantage of advances in scattering theory and applying appropriate
signal processing, OLCR can be used as a sensor in a number of processing applications.
In reflective or weakly scattering systems, OLCR can be applied for thickness monitoring
of single and multiple layer polymer films, as well as transparent coatings upon complex,
scattering substrates. In dense, highly scattering systems, process applications may include
the on-line measurement of mean particle size determination in attrition milling, the state of
mixing, monitoring of paint curing rates, non-invasive thickness determination of highly
scattering coatings on both conducting and non-conducting substrates, and fermentation
monitoring.
BASICS OF REFLECTOMETRY
The technique is based upon an adapted Michelson interferometer, which utilizes a
broadband (low coherence) light source and substitutes the sample under test for the
CP657, Review of Quantitative Nondestructive Evaluation Vol. 22, ed. by D. O. Thompson and D. E. Chimenti
© 2003 American Institute of Physics 0-7354-0117-9/03/S20.00
1713
stationary mirror. A schematic diagram of the reflectometer used in these studies is shown
in Figure 1. Photons travel through standard, single mode optical fiber and are split into
reference and sample arms. The reference arm is composed of a reflective mirror moving
at a constant velocity, which imposes a 27 kHz Doppler shift on the wave packets, and
modulates them. In the sample arm, wave packets are reflected at 180° or backscattered, in
phase, from the sample and are collected by the same optical fiber. Wave packets from the
two arms are recombined by the beamsplitter, now acting as a wavelength independent
coupler.
Constructive interference occurs only when the path difference traveled by both
beams of light is within the coherence length of the source, creating a beat frequency. The
polarization of the two beams must also be parallel for full intensity of the interferogram to
occur.
The recombined light then hits a photodetector which is connected to a heterodyne
receiver. The heterodyne receiver is tuned to the Doppler shift frequency and is
responsible for the large dynamic range of the reflectometer (80-90 dB, or 8-9 orders of
magnitude). Subsequently, an envelope detector processes the beat pattern by passing only
the positive envelope of the signal, as shown in Figure 2.
The reflectometer profiles are in the form of reflectance vs. photon dwell distance
(time spent within the matrix). As described in Equation 1, reflectance is related to the
proportion of light which returned in phase to the optical fiber.
Reflectance (dB) = 10 * logio (Reflected Light / Incident Light).
(1)
The OLCR profile depends upon particle size, shape, and concentration, change in
refractive index, sample thickness, sample pH, probe geometry, and source coherence
length. Two representative reflectometer profile types are shown in Figure 3 a and b.
Current studies are focused on variations in particle size and concentration.
Sample
FIGURE 1. Instrumental diagram of the HP 8504A High Precision Reflectometer.
FIGURE 2. The positive envelope over the interferogram.
1714
Transparent matrices generate OLCR profiles with discrete reflection peaks. A
single, discrete peak is formed from each interface, including the probe-air and air-sample
interfaces, and the distance between two peaks is the optical thickness of that layer.
Provided the change in refractive index between two materials is greater than 0.001 group
index units and the separation between features (thickness of each layer) is greater than the
source coherence length, then excellent distinction can be found between the layers [4].
Once the peaks have fulfilled the minimum thickness requirement and are
resolvable, changes in thickness much smaller than the coherence length can be measured
(± 2.5 um for a 1 mm scan). The physical thickness of a transparent material can be easily
calculated if its refractive index is known, from the following equation:
Refractive Index = Optical Thickness / Physical Thickness.
(2)
Whereas transparent matrices produce a series of sharp peaks, a tailing decay
profile is indicative of a highly scattering sample. During any scattering event, wave
packets of light may change their direction of propagation as well as the phase relations
they have with the wave packets of the incident light. OLCR is a unique light scattering
technique in that it is sensitive to the very small portion of light that retains its coherence
despite multiple scattering events. Thus, information regarding the phase properties of
scattered light and the morphology of the matrix is contained, albeit in a convoluted way,
in the OLCR scattering signal.
In general, the decay profile can be divided into three regions based on correlation
length and density of fluctuations of particle coordinates. The first section is the leading
edge of the profile, which is related to the initial interaction of the wave packets with the
sample. This region of the profile can provide information regarding interface roughness.
The second region is an exponential decay, where the quantity of wave packets returning to
the probe in phase is directly related to the distance traveled, or time spent within the
sample. This can be viewed as though the wave packets scatter off multiple particles, and
then must re-trace their path exactly to return to the probe in-phase. Only a very small
portion of the wave packets, around 10"8 or 10"9 times the incident light, will actually
return. The third portion of the decay profile is a double exponential. We interpret this
region as due to strongly localized photons, otherwise known as Anderson localization
[5,6]. Photons become 'trapped' within secondary structures of the heterogeneous matrix,
spending additional time in the trap before returning to the probe. This produces additional
photon dwell distance, while wave packets retain their coherent characteristics. We have
proposed that the size of these traps or morphological heterogeneities must be on the order
of the coherence length of the source for them to be observed.
-45 -
-60
55-
8 -70
|-75
I -80
&
-85 -I
-95
-90
0
0.25
0.5
0.75
(a) Photon Dwell Distance (mm)
0
0.25
0.5
0.75
1.0
(b) Photon Dwell Distance (mm)
FIGURE 3. (a) Discrete peaks are produced from each interface of a transparent matrix, (b) Tailing decay
profiles are produced by highly scattering samples.
1715
EXPERIMENTAL
The commercially available Hewlett-Packard 8504A High Precision Reflectometer
was used to investigate both transparent and scattering samples. Nine urn core, 125 um
cladding standard single mode optical fibers were used for both the reference and sample
arms. The internal broadband light source was an edge emitting light emitting diode
(EELED) centered around 1300 nm with a coherence length of 10 microns. The 1310 nm
option of the multiple LED external HP 83437A source, with a 20 micron coherence
length, was also used. The 1300 nm/1310 nm sources were chosen because there are no
major absorption peaks in that region of the Near IR; therefore, any loss of coherence is
due to the properties of the material (RI change or scattering) and not absorption. For
highly scattering liquid samples, the fiber endface of the probe was polished at an 8° angle
to prevent collection of photons reflected from the probe/medium interface, which may
overlap with the sample profile. A flat-ended probe was used for transparent and stand-off
measurements. For the sake of clarity, the discrete peak measured from the probe end-air
interface is not shown.
Polystyrene nanospheres standards were purchased from Duke Scientific, at 1%
concentration and with particle sizes ranging from 20 to 100 nm. Industrial process
samples were contributed by member sponsors of the Center for Process Analytical
Chemistry (CPAC). Fermentation samples were acquired commercially and through the
University of Washington.
RESULTS AND DISCUSSION
Transparent Samples
Packaging is a very important parameter for shelf life determination of
pharmaceutical tablets. Multiple polymer layers provide protection from degradation due
to oxidation and moisture [7]. Often the formation process of tablet blister packs will thin
the polymer layers, but not necessarily uniformly. The reflectometer can be used to
measure the thickness of several layers individually and the pack-making process can be
-45-
-45-
§ 55
3
|-65
-75-
-85-
-95
-95
0.2
0.4
0.6
0.8
0
(a) Photon Dwell Distance (mm)
0.2
0.4
0.6
0.8
(b) Photon Dwell Distance (mm)
FIGURE 4. (a) OLCR measurement of a multi-layer polymer film, non-stretched. The layer thicknesses,
respectively, are 112.5 um and 402.5 um. (b) Measurement from the center of a blister, where both polymer
layers have been stretched and thinned. The layer thicknesses, respectively, are 32.5 um and 105.0 urn.
1716
adjusted or re-designed to provide increased uniformity and thus longer shelf life. Figure 4
illustrates the changing thickness of the layers depending upon location on the blister.
Additionally, measurements from the lower corners of the blisters consistently showed one
or more additional peaks in the profiles, suggesting the existence of delamination of the
two layers or air bubbles caught between the layers.
OLCR can be used to measure different clear coating thicknesses upon the same
scattering substrate. This type of analysis is useful for the pharmaceutical industry, in the
measurement of clear coating thickness on a powder tablet. Coatings act as oxidative and
moisture barriers and may determine time-release characteristics of the tablet. Single
profiles measured from several locations on a pharmaceutical tablet with differences in
coating thickness clearly distinguishable are shown as Figure 5. The scattering profiles
vary by location as well, depending on physical heterogeneities within the sample.
0
02
0.4
0.6
0
0.2
0.4
0.6
0.8
0
0.2
0.4
0.6
0.8
(b) Photon Dwell Distance (mm)
(a) Tablet Measurement Locations
0.8
(c) Photon Dwell Distance (mm)
(d) Photon Dwell Distance (mm)
FIGURE 5. (a) Clear coating thickness was measured at multiple locations on the pharmaceutical tablet as
the optical distance between the discrete peak and the tailing decay profile. Coating optical thicknesses were
calculated to be (b) 100 urn, (c) 142 um, and (d) 120 urn.
0
Coating
Thickness
-10
u
-30
U
lgjUya.^ ^^.Jfe^J
TST^^^lPs^F5 "Sp
-50 c—————•————-
1
2
3
4
Photon Dwell Distance (mm)
FIGURE 6. OLCR determination of two different primer thicknesses underneath a corrosion-inhibiting
compound coating of equal thickness.
1717
Similarly, OLCR can also detect a change in the scattering matrix underneath a
clear coating. An example from the airplane industry is shown in Figure 6, where not only
can the clear corrosion inhibiting compound (CIC) thickness be measured, but a change in
the opaque primer thickness underneath can also be seen. The wave packets do not
penetrate enough to reach the aluminum coupon underneath the primer. Paint and primer
thickness affect not only an airplane's aesthetics, but also fuel consumption due to
increased overall weight.
Highly Scattering Samples
The decay pattern shown on the right hand side of Figure 3 is typical of nonuniform
scattering samples. We can extract quantitative information regarding the sample from the
height, shape, and fluctuations of the decay. The height of the initial leading edge is
related to the particle size and concentration through the relationship [6]:
Log (Smax - Sbase) = Log (pr3) = Log p + Log T3
(3)
where Smax is the reflectance of the leading edge, Sbase is the average reflectance of the
baseline in non-sample regions, p is the volume fraction of the sample, and r is the mean
particle radius. When comparing standards of the same concentration and with mean
particle radius smaller than the wavelength, Smax increases with an increase in particle size.
The shape of the reflectometer profiles remains comparable for all particle sizes at a
constant concentration. Profiles from standard polystyrene nanospheres are shown in
Figure 7a, and the corresponding leading edge analysis is Figure 7b. Measurements were
performed in triplicate.
1
I""
1
& l'2~
.1
0.8-
0
0.2 0.4 0.6 0.8 1.0
1.2 1.4
(a) Photon Dwell Distance (mm)
0.6
2.75
Log S = 0.31372 * (Log R3) - 0.14057
R2 = 0.98141
3.25
3.75
4.25
4.75
5.25
(b) Log Radius Cubed
FIGURE 7. (a) OLCR profiles from monodispersed polystyrene nanospheres at 1% concentration, with
particle diameters ranging from 20 to 100 nm. (b) The direct relationship between the leading edge height
and the cubed particle radius is shown, as follows from Equation 3.
1718
1.0
1.5
2.0
Photon Dwell Distance (mm)
FIGURE 8. OLCR profiles from monodispersed polystyrene nanospheres of 308 nm particle diameter and
concentrations ranging from 0.25 to 10%. Figure reprinted with permission from [8].
When the particle size of the polystyrene standards is held constant and the
concentration is varied, a different relationship between OLCR profiles is observed. Figure
8 shows that while the leading edge height changes to some extent, a dramatic shift occurs
in profile shape [8]. The slope of the second region of the profile changes dramatically
with concentration. This is consistent with the idea that as more particles are available to
scatter light in a small region, the probability of the wave packet returning in phase to the
probe decreases.
In addition, other portions of the profile can be analyzed to extract particle size
information. Particle radius is related to the signal in the first linear portion of the profile
through the relationship:
Log(S-Sbase) = A-
(4)
where S is the OLCR signal, L is the photon dwell distance in microns, -|3 is inversely
proportional to the square of the particle radius, and A is a constant. Particle radius is
directly proportional to the derivative of the plot of OLCR signal vs. photon dwell distance,
-74.5
0
0.1
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
(a) Photon Dwell Distance (mm)
0.2
0.3
0.4
(b) Photon Dwell Distance (mm)
FIGURE 9. (a) OLCR profiles, from top to bottom, of fermentation broth at 3.5 x 109 cells/mL and triplicate
dilutions of 50%, 25%, 10%, and 1%. (b) Low concentration ale profiles from yeast pitching to 50 hours.
1719
S - Sbase = C + (XL
(5)
where a is the derivative and C is a constant. Information may also be deconvoluted from
the fluctuation of an individual signal from the mean throughout the entire OLCR profile.
These fluctuations are not noise-based and do not average out with multiple scans in the
same location. Rather, they are due to set heterogeneities within the system and contain
information regarding particle clustering. Further data analysis techniques are currently
being developed.
Data deconvolution methods are very important when observing more complex
systems, such as fermentation processes. Industrial fermentations often occur at much
higher concentrations and volumes than lab fermentations, which may limit the availability
of measurement tools. The reflectometer was used to measure particle concentration of
several fermentation systems, through dilution of a high concentration yeast broth and as a
dynamic ale system from pitching to final product. The former system was of a known
concentration, 3.5 x 109 cells/mL, which corresponds to an absorbance, or optical density,
at 600 nm of 113. The dynamic system occurred at much lower concentrations, but
changes in signal were easily seen over a 50-hour time period. The two profile sets are
shown in Figure 9a and b.
CONCLUSIONS
Optical Low-Coherence Reflectometry has been shown to be a potentially effective
tool for process measurements of both polymer films and different highly scattering
samples. Further work will include theoretical development of, and experimental studies
on, multiple scattering systems. Data deconvolution techniques used with monodispersed
standards will be expanded further to include both standard and industrial polydispersed
systems.
ACKNOWLEDGEMENTS
This work was supported by the Center for Process Analytical Chemistry (CPAC)
and the Department of Energy EMSP Grant #81964. Samples were provided by Merck
Research Laboratories, The Boeing Company, and other CPAC sponsors. We thank
Robert L. Green (Merck Research Laboratory) for technical assistance and Michelle A.
Pressler (Dow Chemical).
REFERENCES
1. Takada, K., J. Noda, and S. Nakajima. 4th Conference on Optical Fiber Sensors
Proceedings, 23 (1986).
2. Youngquist, R. C., S. Carr, and D. N. Davies. Optics Letters, 12, 158 (1987).
3. Huang, D., et al. Science, 254, 1178 (1991).
4. Shelley P. and Burgess, L. At-Process (IFPAC proceedings), 71 (1996).
5. Brodsky, A., S. R. Thurber, and L. W. Burgess. J. Opt. Soc. Am. A, 17, 2024 (2000).
6. Thurber, S. R., L. W. Burgess, A. Brodsky, P. H. Shelley. J. Opt. Soc. Am. A, 17, 2034
(2000).
7. Griffin, R. C. Jr., and S. Sacharow. Drug and Cosmetic Packaging. Noyes Data
Corporation, New Jersey, 1975, pp. 124-173.
8. Thurber, S. R., A. M. Brodsky, and L. W. Burgess. AppliedSpectroscopy, 54, 1506
(2000).
1720