Thin Solid Films 520 (2012) 4242–4248

Contents lists available at SciVerse ScienceDirect

Thin Solid Films

journal homepage: www.elsevier.com/locate/tsf

Carbon coatings on silica glass optical fibers studied by reflectance Fourier-transform

infrared spectroscopy and focused ion beam scanning electron microscopy
Andrei A. Stolov a,⁎, Jeffrey J. Lombardo b, Brian E. Slyman a, Jie Li a, Wilson K.S. Chiu b
a
OFS, Specialty Photonics Division, 55 Darling Drive, Avon, CT 06001, USA
b
Department of Mechanical Engineering, University of Connecticut, Storrs, CT 06269, USA

a r t i c l e i n f o a b s t r a c t

Article history: Carbon coatings applied on optical fibers via chemical vapor deposition were characterized by a resistance
Received 26 August 2011 technique, focused ion beam/scanning electron microscopy (FIB/SEM), and reflectance Fourier-transform
Received in revised form 10 January 2012 infrared spectroscopy (FTIR). The resistance technique measures the thickness of carbon film by measuring
Accepted 22 January 2012
the resistance over a section of optical fiber, and backing out the film thickness. The FIB/SEM system was
Available online 27 January 2012
used to remove a cross section of the optical fiber and carbon coating and using a scanning transmission elec-
Keywords:
tron detector the thickness was measured. The FTIR approach is based on the fact that the wavelength of the
Optical fibers light in the mid-infrared region (~ 10 μm) is significantly larger than the typical thickness of the carbon coat-
Carbon coatings ings (b 0.1 μm) which makes the coating “semi-transparent” to the infrared light. Carbon coating deposition
Hermetic coatings results in significant transformations of the band profiles of silica in the reflectance spectra that were found to
Reflectance Fourier-transform infrared correlate with the carbon coating thickness for films ranging from 0.7 nm to 54.6 nm. The observed transfor-
spectroscopy mations of the reflectance spectra were explained within the framework of Fresnel reflection of light from a
Silica dual-layer sample. The advantage of this approach is a much higher spatial resolution in comparison with
Focused ion beam
many other known methods and can be performed more quickly than many direct measurement techniques.
Scanning electron microscopy
© 2012 Elsevier B.V. All rights reserved.
Scanning transmission electron microscopy

1. Introduction precursor gas is introduced to the fiber at an elevated temperature

(≈800–1000 °C) and thermal decomposition of the gas leads to the
Hermetic carbon coatings are widely used on specialty optical fibers deposition of carbon on the surface of the glass. Material properties
[1–3]. Those coatings have two major functions. First, they improve the of the resulting carbon coating depend on the processing conditions,
fiber fatigue behavior by preventing strength degradation caused by including precursor gas, reactor configuration, gas flow rates, fiber
moisture attack on the fiber surface. Second, they reduce the diffusion draw speed and temperature. The applied carbon layer is opaque,
of hydrogen into the fiber core which may degrade the performance electrically conductive, and highly adherent to the glass surface via
of the optical fibers. These two features become crucial when the fiber chemical bonding [5]. Hermetic properties of the carbon layer are
is used in harsh environments and/or under mechanical stresses [2,3]. related to its molecular structure and the coating thickness [6–8].
Application areas of carbon-coated optical fibers include oil wells, Thick carbon coatings are known to have a much stronger hermetic
aerospace, undersea and many others. seal of the fiber [9].
A definite advantage of carbon coatings is their small thickness, Several off-line and on-line methods of measuring the carbon
which is negligible in comparison with other parts of the fiber: typi- coating thickness are known. The off-line methods include scanning
cally a few tens of nanometers. With such a small thickness they can- electron microscopy (SEM) [10–12] and measurements of electrical
not create substantial mechanical stress in the fiber and therefore do resistance. A stand alone SEM is one model-independent way for
not contribute to microbending significantly. Since carbon coatings assessing the coating thickness, however there are drawbacks that
are brittle, optical fibers commonly contain additional coating layers limit the usefulness of this technique for examining sub 50 nm
over the carbon, e.g., acrylates, silicones or polyimides [4]. films with high precision. First a sample must undergo sample prep-
Carbon coatings are applied on bare glass fiber via chemical vapor aration to obtain a specimen with a visible cross section of the carbon
deposition (CVD) during the fiber draw process. A hydrocarbon coating without damaging or altering the carbon coating. Second, one
has to deal with the electron charging of the specimen which strongly
reduces the spatial resolution and can make imaging insulating sam-
⁎ Corresponding author. Tel.: + 1 860 678 6629; fax: + 1 860 674 8818. ples extremely difficult. Despite these difficulties, relatively thick car-
E-mail address: [email protected] (A.A. Stolov). bon coatings 100–2600 nm were evaluated by SEM [13–24], while

0040-6090/$ – see front matter © 2012 Elsevier B.V. All rights reserved.
doi:10.1016/j.tsf.2012.01.032
 A.A. Stolov et al. / Thin Solid Films 520 (2012) 4242–4248 4243

thinner coatings of a practical size for optical fibers, i.e. 20–50 nm coating may affect the spectrum of the substrate material, and such
[10,25] have proven more difficult to measure. effects (if any) can be observed by using reflectance FTIR spectra.
Electrical resistance measurements need less time and do not In this work we utilize FTIR reflectance microscopy for character-
require sophisticated equipment. In this approach, the coating thick- izing carbon coatings on optical fibers. Our approach is based on the
ness is determined from the following equation: fact that typical carbon coating thicknesses (20–50 nm) are much
smaller than the wavelength in the mid infrared region (2–20 μm).
ρ Coating thickness is validated to FIB/SEM measurements. This
dc ¼ ð1Þ
πDRΩ makes the carbon coating “semi-transparent” for an IR beam, so that
we are able to “look through” it. Commercial FTIR microscopes can
where RΩ (Ω/cm) is the electrical resistance of 1 cm of the fiber, ρ focus IR beam on a small area (a few tens of microns in diameter),
(Ω·m) is the resistivity of the carbon layer, D is the fiber diameter which is comparable with dimensions of optical fibers (typical clad-
and dc is the carbon layer thickness. A shortcoming of this approach ding diameter is 125 μm).
is the fact that the coating resistivity ρ may vary, depending on the
molecular structure of the carbon material. Coatings made by differ- 2. Experimental details
ent precursors and applied under different conditions may result in
different values of the resistivity. Generally speaking, the resistivity The studied samples were optical fibers drawn by OFS. The glass
of carbon may vary between that of bulk amorphous carbon, cladding was pure silica, and the nominal cladding diameter was
ρ = 60 × 10 − 6 Ω·m and crystalline graphite 0.4 × 10 − 6 Ω·m [9]. In 125 μm. Carbon coatings were applied in-line by chemical vapor
different studies of the carbon coatings, the following resistivity values deposition, using the same precursor hydrocarbon. To obtain coatings
were observed: 11.5–40× 10− 6 Ω·m [25], 17.2–19.6 × 10 − 6 Ω·m [26], with different thickness, we intentionally varied the fiber draw speed
17.6 × 10− 6 Ω·m [27], 11.5–15.1 × 10− 6 Ω·m [22], 16.3–21.8 × and the precursor gas flow rate. Pieces of fiber with no carbon coating
10− 6 Ω·m [18], 19.6 × 10− 6 Ω·m [9]. It follows from the above that were also collected as the reference.
the electrical resistance approach requires an independent benchmark- Some of the fibers were drawn with only a carbon coating and no
ing (e.g., SEM), so that it can be calibrated for certain precursors and polymer coating layers over it. The others were drawn with a stan-
coating application conditions. dard dual acrylate coating applied above the carbon layer to a size
The past two decades has seen a drastic increase in the use of of 250 μm. The total number of fibers studied in this work was 28.
focused ion beam/scanning electron microscopy (FIB/SEM) for sam- For another reference, we used a plane fused quartz slide (product
ple preparation [28–32], especially for transmission electron micros- of Chemglass, Inc, part number CGQ-0640-01). Before use, the slide
copy (TEM) sample preparation, because of the system's ability to surface was cleaned by washing with warm soapy water, followed
prepare electron transparent samples from selected sample locations. by continuous flushing with acetone combined with mechanical
Additionally many FIB/SEM systems are equipped with high resolu- wiping. No carbon coating was applied on the slide.
tion equipment such as field emission electron sources and scanning The electrical resistance of the carbon coatings was measured
transmission electron (STEM) detectors which can provide electron using an Omega HHM31 multimeter. Before each measurement,
system resolution of 0.8 nm [33] in STEM mode. With sufficient sam- whenever necessary, the acrylate coating was mechanically stripped
ple preparation the FIB/SEM technique should be able to measure the off the fiber using a heatable (80 °C) strip tool. Then the carbon sur-
film thickness of carbon films that are thinner than the stand alone face was gently wiped using a Kimwipe paper rinsed with isopropyl
SEM systems that have been previously discussed. alcohol. For each fiber, six specimens were prepared, 5 cm long
The on-line methods include electrical and optical approaches. It each. The electrical resistance was averaged over the results obtained
has been suggested to insert the moving fiber inside an inductive for those six specimens.
coil that is involved in an RLC circuit [34] and then relating a circuit FIB/SEM analysis was performed using a Strata 400 DualBeam sys-
Q-factor with the resistance of the inserted fiber, which can then be tem with STEM detector. To reduce charging during the FIB milling
recalculated into the coating thickness. In a similar approach [35], process and to protect the carbon film from ion damage, a layer of
the authors scanned the frequency and investigated the intensity Au/Pd was deposited using a Polaron E5100 SEM coating unit at
of the output signal of the coil. The coating thickness was then calcu- 20 mA for 10 min. The Au/Pd layer did not form a uniform layer on
lated from the electrical resistance, which is obtainable from the the carbon, but rather it beaded on the surface causing further protec-
frequency scan measurements. As above, the in-line electrical ap- tion of the carbon film to be necessary. To further protect the film, a
proaches are based on using Eq. (1), so the comparison with another layer of Pt was deposited using the electron beam in the FIB/SEM sys-
method is required to ensure the correct resistivity values are being tem to fill in the gaps left by the Au/Pd coating, and add additional
used for the carbon films being studied. protective material. Following the electron beam Pt deposition, an
The known optical methods are based on partial light transmis- ion beam deposition of Pt was done until the Pt layer was near
sion of the carbon coatings [26,36]. The carbon thickness is estimated 0.5 μm thick. Initial ion milling was done at 21 nA at 30.00 kV and
from the attenuation of the visible light passing radially through the was eventually stepped down to 2.8 nA for thinning the sample to a
carbon-coated fiber. The challenges for in-line optical approaches 1 μm thickness. A micromanipulator was then brought in and at-
are the fiber oscillations during draw and poor optical transmissivity tached to the surface of the thinned samples using Pt. The thinned
of carbon in the visible range. These measurements also must be samples were then cut off from the bulk samples and mounted on a
calibrated by an independent method. TEM sample holder. Additional thinning of the samples was then
Since none of the aforementioned approaches is free of shortcom- done using the ion beam stepping down to 93 pA at 30.00 kV for
ings, it is of interest to explore other possible ways of determining the the final surface cleaning. Once the sample was made electron trans-
thickness of carbon coatings. Thus far, Fourier-transform infrared parent, a built in STEM detector was brought in and the carbon film
(FTIR) spectroscopy has been used for characterizing carbon coatings' was aligned perpendicular to the electron beam. STEM imaging was
chemistry [37]. This way, different types of residual functional groups, done at 30.00 kV for different thicknesses of the carbon film.
such as C\H can be detected. However, if carbon is the only chemical Raman spectra were taken using a Renishaw Ramascope Micro-
element of the coating material, FTIR spectroscopy becomes ineffec- Raman with 514 nm laser excitation. Detailed spectra were taken
tive since all the chemical bonds of the material have a zero dipole over the range 800 cm − 1–2100 cm − 1 and the peaks at 1360 and
transition moment. Thus, no peaks due to carbon–carbon vibrations 1590 cm − 1 were observed. Those peaks belong to the “disordered”
can be detected in the FTIR spectra. On the other hand, the deposited (D) and “graphite-type” (G) phases of carbon, respectively [38] and
4244 A.A. Stolov et al. / Thin Solid Films 520 (2012) 4242–4248

can be used to obtain information about the underlying structure while the one with the thickest coating exhibited the resistance
[39]. D and G peak intensities were determined with respect to the of 8.2 KΩ·cm − 1. As it was mentioned above, the carbon coating
baseline drawn between 2000 and 900 cm − 1, and the D/G intensity thickness can be calculated from the electrical resistance only if
ratios were calculated. the coating resistivity is known, so a carbon resistivity value of
Reflectance FTIR spectra, were collected in the range 4000– 17.6 × 10 − 6 Ω·m [27] was used unless otherwise noted. With this
600 cm − 1 using a Thermo-Fisher Nexus 670 FTIR spectrometer com- assumption, the coating thicknesses for the collected fibers fall in
bined with a Centaurus microscope. This microscope has a Triton™ the range between 0.7 nm (thinnest) and 54.6 nm (thickest). The
objective with a numerical aperture of 0.71 and is equipped with an validity of the above assumption will be analyzed in the sections
LN2-cooled mercury cadmium telluride detector. The interferograms below.
were averaged over 256 scans, Happ-Genzel apodized, and Fourier
transformed with a 2-level zero filling to yield spectra at a resolution 3.2. Microscopy
of 2 cm − 1 and a data spacing of 0.24 cm − 1. The wavenumber calibra-
tion was performed using a standard polystyrene film (SRM 1921a) STEM electron micrographs were obtained for two carbon films
provided by NIST. Reflectance of an aluminum mirror was used as with resistance calculated thickness of 29.9 nm and 51.3 nm and are
background. As was reported by the instrument manufacturer, the shown in Fig. 1. The top of both images display the protective Pt coat-
average reflection angle was 35° with respect to the normal. The ing, followed by the Au/Pd layer. The Au/Pd protective layer was
spectra were obtained with a completely open full-filled aperture. beaded up on the carbon film surface, and did not form a uniform
At this condition, the IR beam focuses on a spot of approximately layer. In cross section as shown in Fig. 1 this Au/Pd layer can be
140 μm in diameter. The fiber specimens were aligned horizontally seen in both images as the darker color of the protective coatings
so that the reflectance spectra were taken from their carbon-coated (near black), with the Pt layer showing up as the lighter color
cylindrical surfaces. The specimen X–Y position was adjusted to (gray) filling in the gaps in the Au/Pd layer as well as on top of the
maximize the reflected intensity. The sample preparation was the Au/Pd layer. Below these coatings is the carbon layer which has the
same as for the electrical resistance measurements. lightest color (white), appearing as white due to its low density com-
The actual intensity of the reflected beam is due to the reflectivity pared to the other materials. Finally the bulk of the optical fiber is
of the specimen surface and also its size and the shape. To obtain the on the bottom of the micrographs and shows as a grayscale color
“true reflectance” from the experimental spectra, it is possible to use between the carbon film and the protective coatings. During the
the following approach: In the spectra of pure silica, the strongest following discussion these two films will be referred to by their
reflection peak is observed at ~ 1120 cm − 1. Using a plane quartz resistance-calculated thickness.
slide as a sample, we observed that the 1120 cm − 1 peak intensity To obtain the thickness of the carbon films from the STEM micro-
was 0.73 (i.e., 73%). By definition, the spectrum of the slide can be graphs, ImageJ software was used to create a plot of the color values
considered as “pure reflectance” of silica, since the slide and the along a line beginning in the Au/Pd layer, passing through the carbon
aluminum mirror used for the background have the same planar film and ending in the optical fiber. Next the position of the edges of
geometry. When taking a spectrum of an uncoated silica fiber from the carbon film was determined using a threshold color value to
its cylindrical surface, we obtained the 1120 cm − 1 peak intensity of denote the transition between the carbon film and the substrate
0.12 only. To reconstruct the “pure reflectance” from the obtained material and the protective coating. Due to the presence of a transi-
spectrum of the fiber surface, we multiplied it by 0.73/0.12 ≈ 6.0. tional layer between the carbon film and the neighboring materials,
The value of 6.0 can be called the “shape factor” of the system, likely due to the film being tilted in the beam or having a non uniform
comprising the fiber sample and the experimental setup. Below, all thickness through the depth of the sample, it was also necessary
experimental spectra were scaled by the shape factor of 6.0 to obtain to establish a “film edge” region in the measurement. This calculated
the pure reflectance magnitudes. thickness of the “film edge” region, which was also determined using
All the data processing was performed using Omnic 6.0 software a color value threshold, was divided in half and added to the carbon
supplied with the spectrometer. film thickness. The thickness of the “29.9” nm film was measured at
three spots along the cross section and was 30 nm with a standard
3. Results and discussion deviation of 0.6 nm which is at or below the measurement accuracy
for this method. The “51.3 nm” film measured to be 41 nm with a
3.1. Electrical resistance measurements standard deviation of 1.2 nm.
From the STEM measurements it appears that resistance technique
For all the collected fibers the electrical resistance of the coating validates very well to electron microscopy data in the “29.9 nm” case
(RΩ) was determined. Among the collected fibers, the one with the while the “51.3 nm” resistance measurement is 25% higher than the
thinnest carbon coating displayed the resistance of 690 KΩ·cm − 1, electron microscopy technique. The reason for this difference in one

Au/Pd
with Pt

Carbon
Film

Bulk
Fiber

Fig. 1. STEM images of a two carbon films with resistance calculated thicknesses of “29.9 nm” (left) and “51.3 nm” (right).
 A.A. Stolov et al. / Thin Solid Films 520 (2012) 4242–4248 4245

trial but not the other trial is not readily explained. However there are 6.0 to produce the “true reflectance” spectra of the uncoated and
possible explanations for this discrepancy. First, electron microscopy carbon-coated silica. The spectrum of the uncoated fiber displays
is a highly localized technique and the values obtained utilizing the the most intense reflection peak at ~1120 cm − 1, the intensity being
micrographs only represent a snapshot of the film at one location around 0.73. A shoulder at ~ 1250 cm − 1 is partly related to a longitu-
along the fiber. If any change in film thickness exists along the fiber dinal optic (LO) mode of the stretching asymmetric Si–O–Si vibra-
length past the scale of a few microns, a way to observe the changes tions of the silica. This LO shoulder is known to become x stronger
would be to make multiple samples at different intervals. This is at oblique angles of incidence of the IR beam [45]. The reflectance
potentially very time consuming and without knowledge of the peri- goes through a minimum at ~ 1360 cm − 1. At wavenumbers above
odicity of the spatial variation of the thickness, should any variation 2500 cm − 1 the reflectance changes slowly with the frequency and
exist at all, this approach would not be practical. has a magnitude of about 0.04 at 3500 cm − 1.
The morphology of the carbon film itself could also change during The deposition of the carbon coating changes the spectrum signif-
the different growth conditions needed to obtain different film thick- icantly (Fig. 2). Nevertheless, the vibrational modes of silica hold in
ness. This hypothesis is supported by Raman data which shows that all the spectra, indicating that the coatings are “semi transparent”.
the “29.9 nm” film has a D/G ratio of 0.81 ± 0.01 while for the No C\H, C_O or any other functional groups are observed in the
“51.3 nm” film, the ratio was 0.88 ± 0.01. The observed D/G ratios spectra of the coated fibers. We see that adding a carbon coating
indicate that there is a structural difference between the studied results in the following changes in the spectra: (i) the reflectance
films which may be responsible for different resistivity magnitudes. maximum and minimum shift to lower frequencies; (ii) the reflec-
So far, we assumed ρ = 17.6 × 10 − 6 Ω·m for the films. At the same tance peak intensity decreases, while the reflectance minimum
time, assuming that the true thickness of the “51.3 nm” film is value increases; (iii) at wavenumbers above 2500 cm − 1, where silica
41 nm, one can calculate that the carbon resistivity value for this has no vibrational modes, the reflectance increases considerably with
film is ρ = 14.1 × 10 − 6 Ω·m. This value is quite reasonable since it the carbon coating thickness; (iv) the 800 cm − 1 peak shifts slightly to
falls within the magnitudes documented for carbon coatings [22,25]. lower frequencies when the carbon coating is added.
Determination of the exact cause for the inconsistency between the It was of interest to study correlations between the observed spec-
resistance and SEM data requires a more detailed understanding of tral transformations and the carbon coating thickness. As a measure
the exact nature of the carbon film. of the coating thickness, we utilized the reciprocal electrical resis-
tance of the carbon fibers (1/RΩ). Three different features of the spec-
3.3. Analysis of FTIR spectra tra were selected for quantitative characterization of the coated fiber
samples. The first one was the ~1100 cm − 1 reflectance peak frequen-
According to the literature, [40–44] only two vibrational modes of cy, νMax. The second selected feature was the intensity of this peak.
amorphous silica fall in the range 4000–600 cm − 1. The strongest Third, we trialed the reflectance in the spectral range free of absorp-
band observable in infrared spectra at ~1100 cm − 1, represents a tion bands of silica. For this, we selected an arbitrary frequency of
transverse optical (TO) asymmetric Si–O–Si stretching mode, while 3500 cm − 1 and determined the reflectance intensities at this particu-
a weaker absorption band at ~ 800 cm − 1 belongs to a TO symmetric lar wavenumber.
stretching mode. Depending on the measurement technique, the Prior to obtaining the numerical data, the spectra were smoothed
band positions, shapes and intensities may vary significantly. Reflec- in the region 1500–700 cm − 1 via Savitsky–Golay algorithm. The de-
tance spectra of pure silica obtained at different angles of incidence gree of smoothing was 25 points, which is equivalent to 6 cm − 1.
have been studied in detail [42,45]. A great deal of the information This is much smaller than the width of the peak under study
has been obtained from the band at 1120 cm − 1 since it is very strong (>100 cm − 1), so it did not affect the band shape significantly. Then
and also because it was proven that its peak frequency is correlated the peak frequencies and the reflectance intensities were determined
with fictive temperature, density and mechanical stresses in amor- using Omnic 6.0 software supplied with the spectrometer. For each
phous silica [46–51]. specimen, 2–4 spectra were collected, processed separately and
Fig. 2 shows FTIR reflectance spectra obtained from one uncoated their results were averaged.
and several carbon-coated optical fibers near the 1120 cm − 1 reflec- Fig. 3 shows the obtained correlation between 1/RΩ and the reflec-
tion peak. For the reasons described in the Experimental section, the tance peak frequency. The data points can be fitted reasonably well
“as obtained” reflection magnitudes were scaled by a shape factor of using a quadratic function:

ν Max ¼ −1904:1⋅R−2 −1
Ω −159:56⋅RΩ þ 1119:7; r 2 ¼ 0:997 ð2Þ
0.8

0.7 Extended range No carbon coating

to 3600 cm-1 12.5 nm
0.6 26.3 nm
34.6 nm
54.6 nm
Reflectance

0.5
3500 3000 2500 2000 1500 1000

0.4

0.3

0.2

0.1

0
1500 1400 1300 1200 1100 1000 900 800 700
Wavenumbers (cm-1)

Fig. 2. Reflectance FTIR spectra of carbon-coated optical fibers. The carbon coating
thicknesses displayed in the legend were estimated from the electrical resistance
assuming ρ = 17.6 × 10− 6 Ω·m. The curves were obtained by scaling the experimental Fig. 3. A correlation observed between the ~ 1100 cm− 1 peak frequency in the reflec-
data by a “shape factor” of 6.0. Inset shows extended range to 3600 cm− 1. tance spectra and the carbon coating resistance.
4246 A.A. Stolov et al. / Thin Solid Films 520 (2012) 4242–4248

In the above expression and also below, the dimensions of νMax

−1
and RΩ are cm − 1 and KΩ − 1 cm, respectively, and r 2 is the squared
correlation coefficient.
−1
Note that that RΩ ~ 0.12 KΩ − 1 cm corresponds to the film thick-
ness of approximately 50 nm. From Fig. 3, it can be deduced that the
sensitivity of the peak frequency to the carbon coating thickness is
about 1 cm − 1/nm. Since the peak maximum in FTIR can be deter-
mined with a very high accuracy (better that ± 0.1 cm − 1) this pro-
vides a promising opportunity for accurate measurements of the
carbon coating thickness. A definite advantage of the reflectance
FTIR approach is its high spatial resolution. Modern FTIR spectrome-
ters can focus the IR beam on spots of a few microns in diameter.
Thus, in this work the IR beam is focused on a ~ 140 μm spot, and if
needed, FTIR spectra can be obtained from significantly smaller
areas [52]. It seems possible to use the reflectance FTIR for mapping
the coating thickness profile for inhomogeneously applied coatings.
Another possibility to assess the coating thickness is by studying Fig. 5. Reflectance intensity from carbon-coated fibers at 3500 cm− 1 vs. reciprocal
the intensity of the peaks of silica observable through the carbon electrical resistance.
coating. The Si–O–Si vibrational mode develops a “sigmoid” shape
with a clear maximum and the minimum. From Fig. 2 one can see
that upon increasing the coating thickness, the peak maximum the magnitude of reflectance will reach a plateau, which a priori is
value (RMax) drops while the minimum value (RMin) increases. unknown. Therefore, we fitted the data by the following equation:
Thus, for evaluating the peak intensity we followed an empirical



approach described in Ref [53]. It has been found that the “contrast R3500 ¼ 0:04 þ 0:594⋅ 1− exp −R−1 r2 ¼ 0:971
Ω =0:208 ; ð4Þ
amplitude”, i.e., the difference between RMax and RMin, can be utilized
for semi-quantitative analysis of the reflectance spectra. For each of
the spectra, we determined the maximum and the minimum reflec- where (0.04 + 0.594) is the reflectivity of the “infinitely thick” carbon.
tivity in the range 1400–1050 cm − 1 and used them for calculating As in the above case, relatively large scatter of the data points should
(RMax–RMin). A correlation between (RMax–RMin) and 1/RΩ is displayed be attributed to the effects of the specimen's alignment in the FTIR
in Fig. 4. It seems obvious that the in the limiting cases of “no carbon” microscope.
and “infinitely thick carbon” the (RMax–RMin) magnitudes should be Of the obtained three correlations, the one described by Eq. (2)
0.73 (maximum reflectivity of pure silica glass) and 0, respectively. seems the most promising for the characterization and mapping of
Hence, we fitted the data using an exponential curve: carbon coatings deposited on silica. It usually takes just a few minutes
to collect and process an IR spectrum. Thus, this approach can be

 almost as fast as measurements of the coating resistance. With the
RMax −RMin ¼ 0:73⋅ exp −R−1
Ω =0:118 ; r 2 ¼ 0:967 ð3Þ noted above advantage of a high spatial resolution, the IR method
seems especially effective for troubleshooting and mapping, when
the carbon coating thickness varies from one spot of the sample to an-
The correlation is not as strong as the one described by Eq. (2).
other. The accuracy of predicting RΩ from the peak frequency is the
The weaker correlation should be attributed to the fact that measure-
highest for relatively thin coatings (≤50 nm). For larger thicknesses,
ments of the reflected intensities are affected by the specimen's
the silica bands become less prominent and the peak maximum gets
alignment in the FTIR microscope. In contrast, the peak frequencies
“flatter” which makes harder to determine the peak position. It
are almost insensitive to the sample alignment.
must be also reminded that the actual coating thickness magnitudes
Fig. 5 shows the reflection intensity at 3500 cm − 1 (R3500) plotted
can be determined only if the carbon resistivity is known. Hence,
against the reciprocal electrical resistance. For uncoated fibers, the re-
this approach needs to be calibrated by another one, like electron
flectance is about 0.04. It is obvious that at large coating thicknesses,
microscopy.

3.4. Modeling of FTIR spectra transformations

We attempted to model the spectral transformations shown in

Fig. 2 by applying Fresnel equations. The studied system includes
three media: air, a carbon layer and the bulk of silica. Those media
hereafter will be denoted by subscript indices 0, 1 and 2, respectively.
Each of the media is characterized by its own complex refractive
index na (a = 0, 1 or 2). An IR beam falls on the air/coating (0/1) inter-
face at an angle of θ0 and is partly transmitted into the media 1 and 2,
where its direction is characterized by complex angles θ1 and θ2. From
the Snell's law:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
na−1
sin θa ¼ sin θa−1 ; cos θa ¼ 1− sin2 θa ð6Þ
na

In our experiments, the IR beam is not polarized, while the Fresnel

equations were derived for the s- and p-polarized light. Therefore we
Fig. 4. A correlation between ~ 1100 cm− 1 peak “contrast” (RMax–RMin) and the carbon performed the calculations for s- and p-polarizations separately and
coating resistance. then averaged the results. For the beam traveling in medium a and
 A.A. Stolov et al. / Thin Solid Films 520 (2012) 4242–4248 4247

hitting the a/b interface, the Fresnel amplitude reflection coefficients

are:

na cos θa −nb cos θb n cos θb −nb cos θa

r Sab ¼ ;r ¼ a ð7Þ
na cos θa þ nb cos θb Pab na cos θb þ nb cos θa

and the transmission coefficients are:

2na cos θa 2na cos θa

t Sab ¼ ;t ¼ ð8Þ
na cos θa þ nb cos θb Pab na cos θb þ nb cos θa

The reflection constants from the whole carbon-coated silica sam-

ple can be calculated by the following equations [54]:

t S01 t S10 r S12 expð−2iφÞ t t r expð−2iφÞ

r S ¼ r S01 þ ; r ¼ r P01 þ P01 P10 P12
1−rS10 r S12 expð−2iφÞ P 1−r P10 r P12 expð−2iφÞ
ð9Þ
Fig. 6. Experimental and modeled reflectance FTIR spectra of silica coated by a 34.6 nm
where carbon film. The modeled spectra are obtained at various magnitudes of the real and
imaginary parts of n1.
dc
φ ¼ 2π ð10Þ
λη1 cosðθ1 Þ
where real part of the refractive index was varied between 4 and 6
and the imaginary was varied between 3 and 5. The best match was
In the above equation, ϕ is the complex phase coefficient for the
observed for n1 ≈ 5–4i. Still, the model somewhat overestimates the
carbon coating layer, dc is the carbon coating thickness (same as in
reflectance at the higher-frequency side of the band and somewhat
Eq. (1)) and λ is the wavelength in vacuum.
underestimates it at the lower-frequency side. It should be reminded
Finally, from the reflection constants obtained for s- and p-
that in our calculations we assume n1 to be frequency independent,
polarization, the theoretical average reflectance can be calculated:
which is a very rough assumption. Moreover, it is well known that

     the real and imaginary parts of refractive indices of electric conduc-
 2  2
Rtheor ¼ r S  þr P  =2 ð11Þ tors (metals) significantly increase with the wavelength [55]. This
fact may explain the difference between the n1 = 2.52–1.94i deter-
This value has the same physical meaning as the reflectance mined the for the visible region (λ ~ 0.5 μm) and the our best fit
measured by FTIR spectra. data for λ ~ 10 μm. It seems obvious that a much better fit of the ex-
For utilizing the model, one should know the optical parameters perimental spectra can be achieved if one allows the components of
of the media under study. The refractive index values for silica in n1 to vary with the wavelength. However, modeling at such a deep
the frequency range 2000–400 cm − 1 (both the real and the imagi- level is outside the scope of our current work.
nary components) were taken from Trasferetti and Davanzo [42]. It is worth noting, that even with the frequency-independent
Review of relevant literature reveals little information on the com- n1, the model is able to predict all the main features of the reflec-
plex refractive index of graphite-like carbon in the mid-infrared tion band, including the minimum at ~ 1250 cm − 1, a shoulder at
region. However, these values were determined for the visible light ~ 1010 cm − 1, etc. Since the only adjustable parameter of the
region [54]: n1 ≈ 2.52–1.94i. We used those numbers as a starting model was the complex refractive index of the coating, it should
point and then adjusted them by fitting the experimental data. It be concluded that reflectance FTIR spectra can provide a way of
was assumed in our calculations that both the real and the imaginary evaluating the electro-optical parameters of electrically conductive
parts of n1 are frequency-independent in the region 2000–400 cm − 1. materials in the mid-infrared spectral range.
As it was reported by the microscope manufacturer, the average
angle of incidence θ0 should be 35° with respect to the normal. This
statement can be checked by comparing a spectrum of an uncoated
silica fiber with the model predictions. At this step of the modeling,
the electro-optical parameters of the carbon coating are not involved,
and the calculations are based purely on the parameters of silica
which are well known [42]. Varying the magnitude of θ0 between
0 and 35°, we obtained a series of theoretical spectral curves. The
1300–1150 cm − 1 region is the one most sensitive to the angle of
incidence since it is related to the development of the LO mode in
the spectra [45]. It was found that the best match is observed for
the angle of incidence of ~ 25°. Thus, in our further calculations we
assumed θ0 = 25° instead of 35° as reported by manufacturer.
Next, the electro-optical parameters of the carbon layer were ad-
justed for the spectral range of interest. For this, we selected a fiber
with a typical carbon coating thickness: dc = 34.6 nm (as above, the
thickness was estimated from the resistance measurements and as-
suming ρ = 17.6 × 10 − 6 Ω·m [27]) and tried to adjust the modeled
spectrum to the experimental one. Using n1 = 2.52–1.94i gave very
Fig. 7. Experimental and modeled reflectance FTIR spectra of uncoated and carbon-
poor agreement with the experimental data. Much better results coated silica. The coating thicknesses of the experimental spectra are determined
were obtained using higher values for both real and imaginary parts from the resistance measurements assuming ρ = 17.6 × 10− 6 Ω·m. The theoretical
of the coating refractive index. Fig. 6 shows a few modeled spectra spectra assume θ0 = 25° and n1 ≈ 5–4i.
4248 A.A. Stolov et al. / Thin Solid Films 520 (2012) 4242–4248

Fig. 7 illustrates the ability of the model in reproducing the reflec- [9] S.L. Semjonov, A.F. Kosolapov, I.V. Nikolin, R. Ramos, V. Vaynshteyn, A. Hartog,
Proc. SPIE 6193 (2006) 61930 N.
tance spectra of samples with different coating thicknesses. Note [10] H. Aikawa, M. Nakamura, H. Ishikawa, I. Fujita, T. Danzuka, Proc. Internat. Cable &
that all the calculations were performed using the same magnitudes Wire Symp., 1993, p. 374.
for n1. It appears that the model predicts the trends described by [11] K.H. Kwok, W.K.S. Chiu, Carbon 41 (2003) 2295.
[12] H.-C. Lin, S.-T. Shiue, Y.-M. Chou, J. Chin. Inst. Eng. 31 (2008) 253.
Eqs. (2)–(4). The model can even predict with a reasonable accuracy [13] C.A. Taylor, M.F. Wayne, W.K.S. Chiu, Surf. Coat. Technol. 182 (2004) 131.
the magnitudes of νMax and (RMax–RMin) if 1/RΩ is known, or vice [14] C.A. Taylor, M.F. Wayne, W.K.S. Chiu, Carbon 41 (2003) 1867.
versa. The model is not so adequate in predicting the reflectance at [15] C.A. Taylor, M.F. Wayne, W.K.S. Chiu, Thin Solid Films 429 (2003) 190.
[16] K. Kwok, W.K.S. Chiu, Opt. Eng. 44 (2005) 073801.
3500 cm − 1 (Eq. (4)) since this spectral region (λ ~ 3 μm) should be [17] K. Kwok, W.K.S. Chiu, Carbon 41 (2003) 673.
characterized by a different magnitude of n1. [18] P.-Y. Chen, S.-T. Shiue, H.-Y. Lin, Phys. Status Solidi A 206 (2009) 1537.
[19] H.-C. Lin, S.-T. Shiue, Y.-M. Chou, J. Chin. Inst. Eng. 31 (2008) 253.
[20] S.-S. Chen, S.-T. Shiue, W.-C. Tang, Opt. Eng. 46 (2007) 035008.
4. Conclusions
[21] S.-T. Shiue, J.-L. He, L.-Y. Pan, S.-T. Huang, Thin Solid Films 406 (2002) 210.
[22] S.-S. Chen, S.-T. Shiue, Y.-H. Wu, K.-J. Chen, Surf. Coat. Technol. 202 (2007) 798.
The thickness of hermetic carbon films on optical fibers has been [23] R.-S. Chu, S.-T. Shiue, T.-J. Yang, T.-C. Wu, H.-Y. Li, J. Chin. Inst. Eng. 30 (2007) 503.
measured using a resistance technique, FIB/SEM with a STEM detec- [24] S.-S. Chen, S.-T. Shiue, T.-J. Yan, K.-J. Cheng, P.-Y. Chen, H.-Y. Lin, Chin. Inst. Eng. 32
(2009) 711.
tor, and reflectance FTIR spectroscopy. It has been found that the [25] E.S.R. Sikora, J.V. Wright, S.J. Pycock, M.J. Yates, Proc. Int. Wire & Cable Symp.,
FIB/SEM approach can be used for evaluating the coating thicknesses 1991, p. 663.
in the range 30–50 nm. A comparison of the resistance and SEM data [26] M.R. Tuzzolo, A.E. Allegretto, E.H. Urriti, Proc. Int. Wire & Cable Symp., 1993, p. 381.
[27] E.A. Lindholm, J. Li, A.S. Hokansson, J. Abramczyk, Proc. Int. Wire & Cable Symp.,
suggests that coatings with different thicknesses may have different 1999, p. 672.
morphology, which is confirmed by Raman spectroscopy. Spectral [28] R.J. Young, P.D. Carleson, X. Da, T. Hunt, J.F. Walker, Proc. Int. Symp. Testing &
transformations observed in reflectance FTIR correlate well with the Failure Anal., 1998, p. 329.
[29] L.A. Giannuzzi, F.A. Stevie, Micron 30 (1999) 197.
carbon coating resistance. The observed spectral correlations can be [30] F.A. Stevie, C.B. Vartuli, L.A. Giannuzzi, T.L. Shofner, S.R. Brown, B. Rossie, F. Hillion,
utilized for quick and precise measurements of the carbon coating R.H. Mills, M. Antonell, R.B. Irwin, B.M. Purcell, Surf. Interface Anal. 31 (2001) 345.
thickness provided that the method was preliminary calibrated by [31] G.J. Nelson, W.M. Harris, J.J. Lombardo, J.R. Izzo Jr., W.K.S. Chiu, P. Tanasini, M.
Cantoni, J. Van herle, C. Comninellis, J.C. Andrews, Y. Liu, P. Pianetta, Y.S. Chu,
SEM and that the structure of the carbon stays consistent. The advan- Electrochem. Commun. 13 (2011) 586.
tage of the FTIR approach is its high spatial resolution, which makes [32] J. Li, T. Malis, S. Dionne, Mater. Charact. 57 (2006) 64.
its potentially useful for mapping and troubleshooting of nano-thick [33] FEI Company (2006) Strata 400 Workhorse DualBeam for Semiconductor Laboratories
[Brochure].
conductive coatings deposited on dielectric substrates. Theoretical
[34] P.I. Kingsbury, US Patent 5,142,228 (1992).
modeling of the reflectance FTIR enables determination of the com- [35] R.M. Atkins, E.E. Peterson, R.D. Tuminaro, US Patent 5,057,781 (1991).
plex refractive index of the coating material in the mid-infrared [36] R.E. Frazee, D.H. Smithgall, US Patent 4,973,343 (1990).
region. [37] K.A. Hamouda, A. Ababou, M. Ouchabane, N. Gabouze, S. Belhousse, H. Menari, K.
Beldjilali, Vacuum 81 (2007) 1472.
[38] F. Tuinstra, J.L. Koenig, J. Chem. Phys. 53 (1970) 1126.
Acknowledgments [39] K. Sato, R. Saito, Y. Oyama, J. Jiang, L.G. Cancado, M.A. Pimenta, A. Jorio, Ge.G.
Samsonidze, G. Dresselhaus, M.S. Dresselhaus, Chem. Phys. Lett. 427 (2006) 117.
[40] C.T. Kirk, Phys. Rev. B 38 (1988) 1255.
The authors acknowledge Dr. Roger Ristau for assistance with [41] P. Lange, J. Appl. Phys. 66 (1989) 201.
STEM measurements using the FIB/SEM system. JJL and WKSC [42] B.C. Trasferetti, C.U. Davanzo, Appl. Spectrosc. 54 (2000) 502.
acknowledge financial support from National Science Foundation [43] P. Innocenzi, J. Non-Cryst. Solids 316 (2003) 309.
[44] A.A. Stolov, D.A. Simoff, Appl. Spectrosc. 62 (2008) 624.
under award number CBET-0651687 and OFS Specialty Photonics [45] R. Almeida, Phys. Rev. B: Condens. Matter 45 (1992) 161.
Division. [46] S.-R. Ryu, M. Tomozawa, J. Non-Cryst. Solids 352 (2006) 3929.
[47] T. Hirose, K. Saito, A. Ikushima, J. Non-Cryst. Solids 352 (2006) 2198.
[48] C. Martinet, V. Martinez, C. Coussa, B. Champagnon, M. Tomozawa, J. Appl. Phys.
References 103 (2008) 083506.
[49] M. Lancry, E. Regnier, B. Poumellec, in: C. Lethien (Ed.), Optical Fibre, New Devel-
[1] P.G. Lemaire, E.A. Lindholm, in: A. Mendez, T.F. Morse (Eds.), Specialty Optical
opments, In-Tech, Rijeka, 2009, p. 125.
Fibers Handbook, Elsevier, Amsterdam, 2007, p. 453.
[50] A.A. Stolov, D.A. Simoff, J. Lightwave Technol. 29 (2011) 1046.
[2] I. Severin, R.E. Abdi, M. Poulain, Opt. Laser Technol. 39 (2007) 435.
[51] M. Tomozawa, R. Hepburn, J. Non-Cryst. Solids 345&346 (2004) 449.
[3] T. Reinsch, J. Henninges, Meas. Sci. Technol. 21 (2010) 094022.
[52] D.-L. Kim, M. Tomozawa, J. Non-Cryst. Solids 286 (2001) 132.
[4] S.R. Schmid, A.F. Toussaint, in: A. Mendez, T.F. Morse (Eds.), Specialty Optical
[53] G. Moore, A. Chizmeshya, P.F. McMillan, Geochim. Cosmochim. Acta 64 (2000) 3571.
Fibers Handbook, Elsevier, Amsterdam, 2007, p. 95.
[54] H.S. Skulason, P.E. Gaskel, T. Szkopek, Nanotechnology 21 (2010) 295609.
[5] F.V. Dimarcello, R.G. Huff, P.J. Lemaire, K.L. Walker, US Patent 5,000,541, 1991.
[55] E.D. Palik (Ed.), Handbook of Optical Constants of Solids, Academic Press, Boston, 1985.
[6] N. Yoshizawa, Y. Katsuyama, Electron. Lett. 25 (1989) 1429.
[7] N. Yoshizawa, Y. Katsuyama, J. Lightwave Technol. 9 (1991) 417.
[8] P.J. Lemaire, Opt. Eng. 30 (1991) 780.