This article has been accepted for inclusion in a future issue of this journal.

Content is final as presented, with the exception of pagination.

IEEE JOURNAL OF PHOTOVOLTAICS 1

Inline Characterization of Ultrathin Amorphous

Silicon Stacks in Silicon Heterojunction Solar
Cell Precursors With Differential
Reflectance Spectroscopy
Saravana Kumar , Henri Vahlman , Saed Al-Hajjawi , Christian Diestel , Jonas Haunschild,
Stefan J. Rupitsch , and Stefan Rein

Abstract—In this article, we present a characterization technique photovoltaics. Currently, single-junction crystalline silicon
for thin-film layers on textured surfaces with random pyramids (c-Si) solar cells with carrier-selective passivating contacts
using reflectance spectroscopy and an optical model based on the
(CSPCs) have achieved conversion efficiencies above 26% due
transfer-matrix method and rigorous polarization ray tracing. The
optical model fits the thickness of ultrathin amorphous silicon (a-Si) to high carrier selectivity, and good surface and field-effect
layers from the measured reflectance using spectrophotometry and passivation [1], [2], [3]. Power conversion efficiencies as high
the measured optical constants using spectral ellipsometry. The as 26.81% with industrial-grade M6 size (166 mm × 166 mm)
estimated a-Si layer thickness from the optical model is com- silicon wafers have been achieved recently [4]. Among various
pared with the measured thickness from transmission electron high-efficiency CSPC-based solar cell technologies, the silicon
microscopy (TEM) images. Modeling the absolute reflectance spec-
trum, the a-Si stack thickness is underestimated by 51% mainly due heterojunction (SHJ) technology has the advantage of realizing
to nonidealities such as varying pyramid base angles and scattering CSPCs with a lower thermal budget around 200 °C [1], [2].
effects that are difficult to consider in the optical model. Modeling This has led to the rapid commercialization of SHJ solar cell
alternatively the differential reflectance spectrum, the a-Si stack technology recently.
thickness is determined in accordance with TEM measurements In an SHJ solar cell, carrier selectivity is achieved by deposit-
with relative error as low as 10%. Fitting the relative change in
reflectance before and after a-Si deposition to determine the layer ing an ultrathin intrinsic amorphous silicon (a-Si) layer and a
thickness makes the optical model robust against instrumental doped a-Si layer consecutively by plasma-enhanced chemical
inaccuracies and superposed nonidealities. The on-the-fly nature of vapor deposition (PECVD) [1], [2]. The thickness of the a-Si
the developed optical characterization technique makes it suitable layers ranges between a few to a couple of tens of nanometers
for high-throughput industrial applications. [5]. As the ultrathin a-Si layers in SHJ solar cells are responsible
Index Terms—Amorphous silicon (a-Si), inline characterization, for charge carrier separation and passivation and thus have a
random pyramids, ray tracing, reflectance spectroscopy, silicon strong impact on cell performance, a fast and inline-compatible
heterojunction (SHJ) solar cells, textured surfaces, thin films. characterization technique is required to identify lateral inho-
mogeneities within a wafer and production fluctuations in an
I. INTRODUCTION
industrial environment.
AMPING-UP the transition to a carbon-neutral energy
R sector requires high-efficiency solar cell concepts in
Conventionally, spectral ellipsometry (SE) is applied to char-
acterize thin-film layers. As textured surfaces with random
pyramids are the typical surface morphology of a solar cell,
Manuscript received 17 April 2023; revised 12 June 2023; accepted 25 July
using SE to analyze thin-film layers on these surfaces demands
2023. This work was supported by the German Federal Ministry for Economic careful alignment of the sample with respect to the polarizer and
Affairs and Climate Action with the Project SALSA under Grant 03EE1096A. analyzer of the ellipsometer. Additionally, the integration time of
(Corresponding author: Saravana Kumar.)
Saravana Kumar, Saed Al-Hajjawi, Christian Diestel, Jonas Haunschild,
SE measurements influences the quality of the signal measured
and Stefan Rein are with the Fraunhofer Institute for Solar Energy Sys- at the detector. These factors challenge SE to be used for inline
tems, D-79110 Freiburg im Breisgau, Germany (e-mail: saravana.senthil.kumar inspection in solar cell production. In contrast, with reflectance
@ise.fraunhofer.de; [email protected]; christian.diestel@ise.
fraunhofer.de; [email protected]; [email protected]
spectroscopy, high-quality reflectance spectra can be achieved
.de). with an integration time of a few milliseconds. A spectrometer
Henri Vahlman was with the Fraunhofer Institute for Solar Energy with an integrating sphere can be easily implemented in an inline
Systems, D-79110 Freiburg im Breisgau, Germany (e-mail: henri.vahlman
@ise.fraunhofer.de).
environment and does not require any alignment of the textured
Stefan J. Rupitsch is with the Department of Microsystems Engineering, surfaces with respect to the integrating sphere [6]. The measured
University of Freiburg, D-79110 Freiburg im Breisgau, Germany (e-mail: ste- reflectance spectra can be fitted to estimate the thicknesses
[email protected]).
Color versions of one or more figures in this article are available at
of the thin films deposited on a textured surface by means
https://doi.org/10.1109/JPHOTOV.2023.3301132. of an optical model and the measured optical constants from
Digital Object Identifier 10.1109/JPHOTOV.2023.3301132 SE [6].
2156-3381 © 2023 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.

Authorized licensed use limited to: Tsinghua University. Downloaded on August 14,2023 at 13:07:06 UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

2 IEEE JOURNAL OF PHOTOVOLTAICS

The thickness of the thin a-Si layers can be determined either Scanning electron microscopy (SEM) was performed to study
(approach 1) by modeling the absolute reflectance spectrum the topology of the textured surfaces and to align the sample for
measured after a-Si deposition or (approach 2) by modeling the a lamella extraction. A lamella of size 10 µm × 8 µm × 0.15 µm
relative change in reflectance before and after a-Si deposition, was cut out by using a focused ion beam [10] after local deposi-
which is known as differential reflectance spectroscopy [7], tion of a protective carbon film (C film) to secure the area under
[8], [9]. This technique has been used widely as an in situ investigation from ablation. The extracted lamella was analyzed
characterization tool in studying organic thin-film layers and with TEM to measure the actual a-Si stack thickness at different
their formation mechanisms [7], [8]. points as reference values. The a-Si stack thicknesses from
Approach 1 has the advantage of a single reflectance measure- the TEM images were measured using the image processing
ment that allows easy implementation in a production line, but software ImageJ with proper scaling. The SEM and TEM images
is sensitive to superposed effects, such as oxidation of the sub- were performed at Fraunhofer CSP.
strate, instrumental variations, calibration-related inaccuracies,
and so on. If applied inline instead of in situ, approach 2 faces B. Construction of the Optical Model
the challenge that measurements in different production steps
(before and after layer deposition) have to be combined and thus The fraction of the reflected light from a surface consid-
assigned wafer-specifically. Nevertheless, the above-mentioned ering interference due to thin-film structures is calculated by
disturbing superposed effects and variations in the texture of exploiting the transfer-matrix method (TMM) based on Fresnel
the substrate are eliminated because of the difference-forming equations [11]. To model the random pyramid structures with
nature of the method [9]. In this study, we assume similar different pyramid base elevations under normal incidence of
texture morphology for all the samples and hence the reflectance light, rigorous polarization ray tracing technique is used [12].
before a-Si layer deposition is measured from a single reference The base angle of the pyramid is set to 54.74°, which is the angle
wafer and used to calculate the differential reflectance of all the between (100) and (111) planes in monocrystalline silicon [13].
samples with the reflectance measured after a-Si deposition. Verily, a random textured silicon substrate shows a distribution
As the suitability of differential reflectance spectroscopy to in the base angle and it is not considered within the optical model.
characterize thin films on pyramidal structures has not yet The change in polarization during successive reflection of light
been explored in literature, adapted optical models for both at pyramid facets is included in the model via polarization ray
approaches are developed in this work and compared in terms of tracing matrices, as developed by Yun et al. [14]. The effective
accuracy and reliability in studying textured silicon wafers with optical constants of the a-Si stack determined from SE measure-
a-Si layers of intentionally varied thickness using reflectance ment and the measured reflectance are input data for the optical
data from an industrial inline tool. To determine the ground-truth model. The a-Si stack is described by effective optical constants
values of the layer thickness, transmission electron microscopy to account for the variations in porosity and doping within the
(TEM) on cross section [10] is used as a reference technique. a-Si stack. Details on the significance of such variations for
modeling accuracy are published elsewhere [6]. The measured
II. METHODS reflectance is fitted with the a-Si stack thickness and the planar
fraction of the textured substrate as fit parameters by using
A. Experimental Procedures the Levenberg–Marquardt least-squares fitting algorithm [15].
The samples investigated in this study were prepared by saw- Planar fraction is the parameter that accounts for areas in the
damage etching (SDE) of monocrystalline n-type Cz-Si wafers sample without pyramids, chemically smoothened pyramid tips,
followed by alkaline texturing, resulting in random pyramid- and incomplete etching, as studied by Birmann et al. [16]. Thus,
like structures, and cleaning with ozone and hydrofluoric acid. the developed optical model allows the thickness of the a-Si
Subsequently, intrinsic and n-doped a-Si layers were deposited stack to be determined from the measured reflectance spectra.
on the wafers by PECVD. From here on, we collectively refer
to the a-Si layers on the c-Si substrate as an a-Si stack. Different III. RESULTS AND DISCUSSION
a-Si stack thicknesses were achieved by varying the deposition
time. A. Differential Reflectance Over Absolute Reflectance
The reflectance measurements were performed using a Zeiss Fig. 1 shows the reflectance spectrum of a typical a-Si stack
OFR 104 spectrophotometer with an integrating sphere in an on a c-Si substrate of an SHJ solar cell precursor and the
inline configuration. The calibration of the tool was performed corresponding fit curve from the optical model. The optical
by a certified 99% reflective white standard made of spectralon. model estimates an a-Si stack thickness of 3.3 nm. This thickness
We used the spectral range from 390 to 950 nm in this study. value determined by fitting the absolute reflectance spectrum
The integration time used to measure the reflectance spectra is is suspected to be unrealistic for the used deposition parame-
24 ms. The spectrometer was installed in an inline front-end ters and the expected thickness is around 7 nm. To determine
wafer inspection system. The belt speed for inline reflectance the exact thickness of the a-Si stack on the pyramidal struc-
measurements was set at 0.275 m/s. An M-2000 SE from tures, TEM measurements were performed on the same sample
J. A. Woolam was used to measure the effective optical constants across four adjacent pyramid facets. Fig. 2(a) and (b) shows the
(input data for the optical model) and thicknesses of the a-Si exemplary TEM cross-sectional images taken across the pyra-
stack on planar reference samples. mid facet closer to the apex and base of the pyramid, respectively.

Authorized licensed use limited to: Tsinghua University. Downloaded on August 14,2023 at 13:07:06 UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

KUMAR et al.: INLINE CHARACTERIZATION OF ULTRATHIN AMORPHOUS SILICON STACKS 3

fraction of the sample). Scattering effects have been considered

in optical simulations by various research groups by imple-
menting Phong scattering to study the reflection, transmission,
and absorption of textured solar cell structures. For instance,
Fung et al. [13] arrive at a perfect reflectance fit by modifying
the Phong exponent and the base angle of the pyramid by trial
and error. The Phong exponent is the parameter that governs
the decay of diffuse light intensities as a function of the angle
between diffuse and specular reflection.
As a second reason, we speculate that any calibration and
instrumental-related inaccuracies could cause errors in the mea-
sured reflectance spectrum. In Fig. 1, the green dotted curve
is the simulated reflectance of a 7-nm-thick a-Si stack on an
Si substrate. The average difference between the simulated
reflectance and the measured absolute reflectance (black solid
curve) is 0.84%abs . This value is within the photometric ac-
curacy of the spectrometer, which is ≤ 1%. However, such a
very slight inaccuracy in the measured reflectance spectrum
Fig. 1. Absolute reflectance spectrum of an a-Si stack on a silicon substrate
(black solid line) and the corresponding fit curve (blue dashed line) from the could lead to a significant error in the estimated thickness of the
optical model. The green dotted line shows the simulated reflectance of a 7-nm- ultrathin a-Si stack from the optical model. Including scattering
thick a-Si stack on the textured silicon substrate with the optical model. in the optical model will reduce the error in thickness estimation
but not elimate the errors due to measurement or instrumental
inaccuracies.
It is seen that the a-Si thin-film stack is conformal across the That is why we suggest a new method to circumvent the effects
textured surface with some variation in thickness ranging from of scattering on the pyramid facets and the errors arising in
7.7 nm close to the pyramid apex to 6.0 nm close to the pyramid the measured reflectance. By fitting the so-called differential
base. The measurement lines in blue are defined with proper reflectance spectrum (DRS) with an adaption of the developed
scaling from the point where the crystalline lattice ends to the optical model, an improvement in the estimated a-Si stack thick-
point where there is a visual change in the gray scale value at the ness can be achieved. We define differential reflectance as the
a-Si/C film interface. The average thickness of the a-Si stack has relative change in the reflectance of the textured silicon substrate
been determined from measurements at one hundred different after the deposition of the a-Si stack. The DRS is a function of
positions and amounts to 6.7 ± 0.9 nm. This indicates that the wavelength (λ) and thickness (d) and is calculated as
developed optical model underestimates the thickness of the
RSi/a−Si (λ, d) − RSi (λ, 0)
ultrathin a-Si stack by a relative error of 51% when the absolute DRS (λ, d) = (1)
reflectance of the sample is used to fit the a-Si stack thickness. RSi (λ, 0)
Using the absolute reflectance approach, the optical model has where RSi (λ,0) is the reflectance of the textured silicon substrate
demonstrated relative errors of around 10% for indium tin oxide and RSi/a-Si (λ,d) is the reflectance after the deposition of the
layers thicker than 50 nm [6]. a-Si stack of thickness d. Fig. 3 depicts the DRS of the same
One reason for such a systematic underestimation by the sample as in Fig. 1 and the corresponding fit curve. The estimated
model can be explained by the SEM image in Fig. 2(c). In this a-Si stack thickness in this case is 6 nm, which is in good
image, we can clearly see rough pyramid facets with wrinkle-like agreement with the reference value determined from the TEM
structures (pointed with red arrow heads) that appear during the images and, henceforth, reduces the relative error significantly to
texturing of the silicon substrate. These rough topologies of the 10%. The simulated differential reflectance curve (green dotted
pyramid facets persist even after the deposition of the a-Si stack, curve) for a-Si stack thickness (3.3 nm) as determined with
which shows that the thin film is conformal over the rough facets the absolute reflectance approach differs significantly from the
of the pyramids. On the contrary, the developed optical model measured differential reflectance data (black solid curve) below
considers these facets to be perfect planes and, consequently, 600 nm. It is to be noted that a slight change in the a-Si stack
assumes that they lead to specular reflection of the incident thickness affects the DRS significantly compared to the absolute
light. This assumption is used in ray tracing for defining the reflectance. In other words, DRS is more sensitive to thickness
propagation direction and the successive angle of incidence of variations compared to the absolute reflectance leading to better
the reflected light on the adjacent pyramid facet [12]. fit thickness values.
The pyramid facets exhibiting a surface roughness of their The improvement of the fitted thickness value can be ex-
own could lead to scattering of the incident light, resulting in plained by the fact that the scattering effect arising from the
diffuse reflectance from the pyramidal facets. Assuming the rough pyramid facets and other measurement inaccuracies are
pyramid facets with zero roughness in the optical model will lead subtracted out by calculating the relative change in reflectance
to an inaccurate reflectance spectrum, which in turn gives rise to after the a-Si stack deposition that enhances the fit quality of
an error in the fit parameters (thickness of the layer and planar the optical model substantially. Moreover, in the view of an

Authorized licensed use limited to: Tsinghua University. Downloaded on August 14,2023 at 13:07:06 UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

4 IEEE JOURNAL OF PHOTOVOLTAICS

Fig. 2. Microscopic images of samples with a-Si stack on c-Si substrate. (a) TEM cross section across the pyramid facet closer to the apex and (b) closer to
the base with local a-Si stack thickness measurements. The thickness of the a-Si stack closer to the apex and base is determined on average as 7.7 and 6.0 nm,
respectively. (c) SEM top view of the silicon substrate with a-Si stack showing pyramids with rough facets and indicating exemplary the positions the cross-sectional
images in (a) and (b) are taken from. The red arrow heads point toward the wrinkle-like structures making the pyramid facets rough.

a-Si stack thicknesses, with each group containing one planar

and eight textured samples. The planar samples are produced
by SDE, skipping the texturing process. The thicknesses of the
a-Si stack on the three planar samples are measured by means
of SE and found to be 6.8, 10.5, and 15 nm. Performing the
three deposition processes on samples with textured morphology
instead of planar surface reduces the thickness of the a-Si layers
due to the increase in surface area of the silicon substrate upon
texturing, which can be considered by an empirical factor. The
empirical factor is calculated as ratio dSE,planar /dTEM,tex with
dSE,planar being the thickness determined by SE on a similarly
processed planar sample and dTEM,tex being the average thick-
ness measured from the TEM image on a textured sample [see
Fig. 2(a) and (b)]. The empirical factor is found to be 1.6 ± 0.2 in
this study, meaning the thickness of the a-Si on a planar surface
will be about 1.6 times larger than on a similarly processed
textured sample. This factor is calculated for a sample from the
Fig. 3. Differential reflectance spectrum of a-Si stack on the silicon substrate middle group and assumed to be the same for samples from
(black solid line) and the corresponding fit curve (red dashed line) from the the other groups. Such an empirical factor is generally used to
optical model. The green dotted curve is the simulated DRS of a 3.3 nm (as estimate the thickness of a thin film on a textured surface from
estimated by using absolute reflectance in Fig. 1) thick a-Si stack on the textured
silicon substrate with the optical model. the thickness on a planar surface and the determined empirical
factor is in accordance with the values found in literature [18],
[19].
industrial application, the geometry of the pyramidal structures
Fig. 4 shows the a-Si stack thickness on textured samples
depends on the nature of the texturing solution used and, hence,
estimated by the optical model using both absolute and differ-
varies from manufacturer to manufacturer [13], [17]. Thus, by
ential reflectance measured in our inline set-up, as a function
using the relative change in reflectance instead of the absolute
of the thickness values measured by SE on planar reference
reflectance for optical modeling, the results for the fitted layer
samples. The solid line reflects the expected layer thickness on
thickness become robust towards variations in the pyramid ge-
the textured surface and is calculated using the surface geometry
ometry.
factor that has been determined empirically (see above). The
error of the empirical factor translates into an error of the
B. Inline Compatibility of the Technique expected thickness, which is shown as the shaded band around
To demonstrate the inline compatibility of the devel- the solid line. As can be seen from Fig. 4, the thickness values
oped method to characterize ultrathin layers in an industrial determined by modeling the differential reflectance (red circles)
environment, we installed the spectrometer in our automated are close to the reference line, whereas those determined by
inline front-end wafer inspection system. By varying the a-Si modeling the absolute reflectance (blue triangles) underestimate
stack deposition time and keeping the other deposition parame- the expected thickness roughly by a factor 2. Moreover, for
ters fixed, we processed three groups of samples with different samples with higher layer thickness, the dispersion of the fitted

Authorized licensed use limited to: Tsinghua University. Downloaded on August 14,2023 at 13:07:06 UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

KUMAR et al.: INLINE CHARACTERIZATION OF ULTRATHIN AMORPHOUS SILICON STACKS 5

a-Si stack deposition cancels out any effect of instrumental

inaccuracies and nonidealities that are not considered within
the optical model.

REFERENCES
[1] Y. Liu et al., “High-efficiency silicon heterojunction solar cells: Materials,
devices and applications,” Mater. Sci. Eng., R: Rep., vol. 142, Oct. 2020,
Art. no. 100579, doi: 10.1016/j.mser.2020.100579.
[2] K. Yoshikawa et al., “Silicon heterojunction solar cell with interdigitated
back contacts for a photoconversion efficiency over 26%,” Nature Energy,
vol. 2, Mar. 2017, Art. no. 17032, doi: 10.1038/nenergy.2017.32.
[3] R. V. K. Chavali, S. De Wolf, and M. A. Alam, “Device physics underlying
silicon heterojunction and passivating-contact solar cells: A topical re-
view,” Prog. Photovolt., Res. Appl., vol. 26, no. 4, pp. 241–260, Oct. 2017,
doi: 10.1002/pip.2959.
[4] H. Lin et al., “Silicon heterojunction solar cells with up to 26.81% effi-
ciency achieved by electrically optimized nanocrystalline-silicon hole con-
tact layers,” Nature Energy, May 2023, doi: 10.1038/s41560-023-01255-2.
[5] P. Procel et al., “The role of heterointerfaces and subgap energy
states on transport mechanisms in silicon heterojunction solar cells,”
Prog. Photovolt., Res. Appl., vol. 28, no. 9, pp. 935–945, May 2020,
Fig. 4. Thicknesses of a-Si stacks estimated by the optical model with absolute doi: 10.1002/pip.3300.
reflectance (blue triangle) and differential reflectance (red circle) on a textured [6] S. Kumar et al., “Characterization of thin-film structures of silicon hetero-
substrate with respect to a-Si stack thickness measured by SE on a planar junction solar cells with inline reflectance spectroscopy,” AIP Conf. Proc.,
reference substrate. The black line represents the reference line that accounts vol. 2826, Jun. 2023, Art. no. 030005, doi: 10.1063/5.0141006.
for an empirical factor of 1.6, being the ratio between the thickness of a-Si [7] J. McIntyre and D. Aspnes, “Differential reflection spectroscopy of very
layers processed similarly on planar and textured silicon substrate. The gray thin surface films,” Surf. Sci., vol. 24, no. 2, pp. 417–434, Feb. 1971,
band represents an error in the empirical factor. doi: 10.1016/0039-6028(71)90272-X.
[8] R. Forker, M. Gruenewald, and T. Fritz, “Optical differential reflectance
spectroscopy on thin molecular films,” Annu. Rep. Prog. Chem., vol. 108,
pp. 34–68, May 2012, doi: 10.1039/C2PC90002E.
thickness data is reduced significantly by using the differential [9] R. Hummel and T. Dubroca, “Differential reflectance spectroscopy
reflectance data instead of the absolute reflectance data. Note in analysis of surfaces,” in Encyclopedia of Analytical Chem-
that the differential reflectance has been calculated in this study istry. Hoboken, NJ, USA: Wiley, Jan. 2014, pp. 9047–9071,
doi: 10.1002/9780470027318.a2504.pub2.
by measuring the reflectance of the textured surface before a-Si [10] N. N. M. Rashid, N. H. A. Junaidi, and S. R. Aid., “Interface damage of pro-
stack deposition, RSi (λ,0), only on one reference sample, as all tective layer in TEM lamella preparation for highly doped Ge substrate,”
samples have been subjected to the same texturing process. That IOP Conf. Ser., Mater. Sci. Eng., vol. 522, Jun. 2019, Art. no. 012003,
doi: 10.1088/1757-899X/522/1/012003.
is, possible variations in the texture quality between samples and, [11] S. J. Byrnes, “Multilayer optical calculations,” Charles Stark Draper Lab.,
thus, in the initial reflectance spectrum have not been taken into Cambridge, MA, USA, Jan. 2021. [Online]. Available: https://arxiv.org/
account. This may explain the remaining dispersion of differen- pdf/1603.02720.pdf
[12] S. Baker-Finch and K. R. McIntosh, “Reflection of normally incident light
tial thickness values observed in Fig. 4. Although the accuracy from silicon solarcells with pyramidal texture,” Prog. Photovolt., Res.
of the developed optical model for fitting differential reflectance Appl., vol. 19, no. 4, pp. 406–416, Jun. 2011, doi: 10.1002/pip.1050.
curves is already high (> 90%), further improvement is expected [13] T. H. Fung et al., “Improved ray tracing on random pyramid texture
via application of Phong scattering,” IEEE J. Photovolt., vol. 9, no. 3,
by measuring the initial reflectance of each individual sample, pp. 591–600, May 2019.
RSi (λ,0), and using these wafer-specific data for the calculation [14] G. Yun, K. Crabtree, and R. A. Chipman, “Properties of the polarization
of DRS. ray tracing matrix,” Proc. SPIE, vol. 6682, Sep. 2007, Art. no. 66820Z,
doi: 10.1117/12.734315.
[15] S. J. Rupitsch, “Characterization of sensor and actuator materials,” in
IV. CONCLUSION Piezoelectric Sensors and Actuators: Fundamentals and Applications.
Berlin, Germany: Springer, Jul. 2018, pp. 143–144.
A nondestructive inline-compatible technique for thin-film [16] K. Birmann, M. Demant, and S. Rein, “Optical characterization of ran-
characterization on textured surfaces has been developed and dom pyramid texturization,” in Proc. Eur. PV Sol. Energy Conf. Ex-
hib., Hamburg, Germany, Sep. 2011, pp. 1454–1458, doi: 10.4229/26thE-
successfully demonstrated by constructing an optical model UPVSEC2011-2BV.2.18.
based on the TMM and the rigorous polarization ray-tracing [17] H. Mäckel, H. Holst, M. Löhmann, E. Wefringhaus, and P. P. Altermatt,
technique. Using the absolute reflectance spectrum to fit the “Detailed analysis of random pyramid surfaces with ray tracing and image
processing,” IEEE J. Photovolt., vol. 6, no. 6, pp. 1456–1465, Nov. 2016.
a-Si stack thickness led to an underestimation by up to a factor [18] D. Pysch, M. Bivour, M. Hermle, and S. Glunz, “Amorphous silicon
2 compared to the measured thickness using TEM. This error carbide heterojunction solar cells on p-type substrates,” Thin Solid Films,
in a-Si stack thickness estimates is significantly reduced to 10% vol. 519, pp. 2550–2554, Feb. 2011, doi: 10.1016/j.tsf.2010.12.028.
[19] S. Olibet et al., “ Textured silicon heterojunction solar cells with over 700
if the differential reflectance spectrum is applied to the optical mV open-circuit voltage studied by transmission electron microscopy,” in
model. Using the relative change in the spectrum due to the Proc. 23rd Eur. PV Sol. Energy Conf. Exhib., Valencia, Spain, Sep. 2008,
pp. 1140–1144, doi: 10.4229/23rdEUPVSEC2008-2DO.1.2.

Authorized licensed use limited to: Tsinghua University. Downloaded on August 14,2023 at 13:07:06 UTC from IEEE Xplore. Restrictions apply.