Optical characterization of patterned thin ﬁlms
, G. Rattmann
, M. Schellenberger
BAM-Federal Institute for Materials Research and Testing, Unter den Eichen 87, 12200 Berlin, Germany
Research Centre for Natural Sciences, Institute for Technical Physics and Materials Science, Konkoly Thege Rd. 29-33, 1121 Budapest, Hungary
Fraunhofer IISB, Schottkystrasse 10, 91058 Erlangen, Germany
Available online 20 November 2013
Spectroscopic imaging and mapping ellipsometry
Inhomogeneous and patterned thin ﬁlms
The presentstudy investigates the use of imaging and mapping ellipsometry to determinethe properties of non-
ideal and patterned thin ﬁlm samples. Samples which are candidates for future references and standards were
prepared for this purpose. The samples investigated were lithographically patterned SiO
and photoresist layers.
The thickness and the optical constantsof the two materialswere determined usingspectroscopicellipsometry in
the visible spectral range. On a larger lateral scale of several mm lateral resolution, the homogeneity was
investigated using a goniospectral rotating compensator ellipsometer. A nulling imaging ellipsometer was used
to determine the properties of the sample on a smaller scale of 25–150 μm.
© 2013 Elsevier B.V. All rights reserved.
The industry of modern electronic devices relies heavily on thin ﬁlm
coating technology. While the technology of coating dielectrics on clas-
sical semiconductor samples has been investigated also on the analyti-
cal side for decades, emerging technologies like organic electronics,
printed electronics and thin ﬁlm photovoltaics call for analytical
methods forless ideal samples. Maintaining a consistent product quality
is nowadays one of the primary concerns of the electronic industry.
Therefore the development of techniques for thedetection of variations
and/or defects of the devices is essential. Fast, non-destructive and
reliable methods are required to determine thicknesses and optical
properties of thin ﬁlms and surfaces.
Silicon dioxide (SiO
) is one of the most studied materials due to its
technological importance. Its dielectric properties established SiO
passivator and insulator in integrated circuits and other optical and
electronic applications [1–3]. Because of its widespread use, thermal
is still the most used reference or standard sample in optical thin
Ellipsometry is one of the most utilized optical techniques for deter-
mining theproperties of surfaces and thin ﬁlms. In the case of dielectric
materials, ellipsometry is especially useful for its ability to measure ﬁlm
thicknesses independent of and even together with the material
Apart from the classical analysis of thin ﬁlms on unstructured
surfaces, the investigation of micro- and nanostructured devices has
become more and more important. In this work, the use of imaging
and mapping ellipsometry for measuring samples relevant to the elec-
tronic industry is evaluated using the example of patterned layers of
and photoresist material on Si wafers. The results of measurements
on a patterned sample withan imaging ellipsometer and with a rotating
compensator ellipsometer (RCE) on unstructured samples are com-
pared. The respective achievable accuracy is evaluated.
Ellipsometry does not provide direct information about a sample's
properties (e.g. optical constants, thickness, roughness); therefore the
main issue is building a suitable optical model in order to obtain correct
information on several quantities simultaneously. This model has then
to be optimized in an iterative process using a suitable least squares op-
timization algorithm. The accuracy of the quantitative evaluation is
strongly dependent on the quality and completeness of the starting
model. In this work, the inﬂuence of measurements on patterned
samples on the correctness of the ﬁt analysis is investigated [4,5].
2. Experimental details
2.1. Sample preparation
The patterned layers were deposited on single-crystalline silicon
(100)-oriented wafers with diameters of 150 mm. Two different pat-
terns with different lateral resolutions were written on Si wafers by
means of photolithography. Each pattern was then either implemented
as thermal oxide on top of the wafer with a nominal thickness of
300 nm or left as a structured layer of ﬁxated photoresist. The oxide
was etched through the removed parts of the resist layer. As the last
step, the resist was removed, and a silicon wafer with patterned oxide
remained. The photoresist used for our studies is a positive photoresist,
AZ 5214E, and was deposited on the substrate by spin coating.
Thin Solid Films 571 (2014) 601–604
⁎Corresponding author. Tel.: +49 30 8104 3895; fax: +49 30 8104 1827.
E-mail address: Dana-Maria.Rosu@bam.de (D. Rosu).
0040-6090/$ –see front matter © 2013 Elsevier B.V. All rights reserved.
Contents lists available at ScienceDirect
Thin Solid Films
journal homepage: www.elsevier.com/locate/tsf
2.2. Spectroscopic ellipsometry
Ellipsometric measurements were carried out in order to determine
the optical properties and the thickness of the SiO
layers. The samples used are candidates for future reference samples
and standards for the thickness of thin dielectric layers. While the cur-
rent state of theart samples of this kind are ideal dielectric layer systems
(usually stratiﬁed thermal SiO
on Si) [6,7], recent applications in
the electronic industry call for calibration samples with non-ideal
In order toinvestigate the large scale inhomogeneity, measurements
were performed using a M2000DI (J. A. Woollam Co.) rotating compen-
sator ellipsometer in the energetic range 0.7–6 eV under different
angles of incidence (55°–70°). The measurements performed by this in-
strument are traceable to the SI length deﬁnition by calibration with
ﬁlm thickness standards from Physikalisch-Technische Bundesanstalt
(see  for details on these samples).
For investigating the large and the small scale homogeneity of the
samples, an imaging null ellipsometer type EP
-SE (Accurion GmbH)
in the spectral range of 400 nm–1000 nm was employed. Null
ellipsometers require rotating polarizer and analyzer so that the
detected light intensity becomes zero. Ψand Δvalues are then estimat-
ed from the rotation angles of the polarizer and analyzer . Nulling
ellipsometry is often considered a tedious and inaccurate method.
Nevertheless it is still the only measurement scheme of ellipsometry
where the two main “raw”quantities Ψand Δcan be seen as directly
measured and not a result of a ﬁt process. In principle, this should
simplify the quantiﬁcation of the accuracy of the derived quantities.
This method was applied to imaging ellipsometry by using a nulling
ellipsometer setup with different microscope lenses and a CCD camera
as the detector. This setup allows for measuring a two dimensional
plane [9,10,11] with a resolution of 1 million pixels in one measure-
ment. The calibrated lateral resolution of 0.51 μmperpixelisthen
only dependent on the choice of the microscope lens. For the present
measurements, a scheme based on four zone averaging was preferred
as it eliminates intrinsic imperfections in the optical components of
the ellipsometer and/or errors arising due to a misalignment of the
The experimental ellipsometric data were ﬁtted using the
Levenberg–Marquardt algorithm  implemented in the WVASE
software (Woollam) and in the EP4Model software (Accurion). Both
packages enable a free deﬁnition of a surface and layer model within
Fig. 1 (left) shows the patterned layer of SiO
on Si substrate. An
overview of the sample and a fraction of the patterned obtained using
a Polyvar MET metal microscope are presented. The dimension of the
stripes in the picture is between 100 and 5 μm.
For determining the thickness variation along the wafer, points 1 to
7 were measured using EP
-SE and modeled with a model containing
the substrate and one layer. The optical constants for the Si substrate
and the SiO
layer taken into account for the calculations are given in
.Table 1 summarizes the obtained thicknesses on the depicted
spots. A maximum of 1.3% thickness variation was calculated, therefore
the sample can be considered sufﬁciently homogeneous for the purpose
of this work. Additionally, a measurement with the RCE was performed
in the vicinity of measurement spot No. 3. The thickness determined for
this measurement was 297.6 nm and the uncertainty was calculated to
be ~2.4 nm. In the calculation of the uncertainty a possible contamina-
tion of the sample surface was not taken into account. Considering the
Fig. 1. Pictureof the Si wafer with photoresist/SiO2coating and repetitive pattern.The numbers and letters indicatethe measurementsites of the ellipsometric measurement. The centerof
the ﬁgure presents a detail of the pattern.
Thickness across the SiO
covered Si substrate as determined by SE; the measured spots
are depicted in Fig. 1,left.
Thickness of SiO2 layer/nm 299.1 299.3 298.3 300.9 297.3 297.0 300.4
Thickness of the photor esist layer across samp le 2. The position of measu red spots is
indicated in Fig. 1.
Spot ID Thickness/nm
1 1447 145 7
4 1488 152 8
5 1510 155 7
6 1534 157 1
7 1559 156 0
9 1563 150 6
10 1546 1490
11 1519 1466
12 1503 1458
602 D. Rosu et al. / Thin Solid Films 571 (2014) 601–604
different measurement principles and regions of interest of these
measurements, the consistency of the results is very good. Therefore,
it can be concluded that the sample is suitable for studying the homoge-
neity of the thickness determination process. This SiO
will be used in a future study to determine the robustness of a multi-
method referencing process of ﬁlm thickness standards together with
X-ray methods and optical reﬂectometry.
For determining the thickness of the SiO
ﬁlm inside the etched
stripes, measurements using a 20 × objective were performed. The
calculated thickness is close to the native silicon oxide thickness
(1.3 to 1.9 nm). Structures smaller than 25 nm are difﬁcult to measure
and analyze, as any small defect would strongly inﬂuence the measure-
ment and therefore the calculations.
3.2. Photoresist pattern
The right side of Fig. 1 illustrates the measured points across the
patterned layer of photoresist (sample 2). The measurements were
performed using the RCE.
Fig. 2. Thickness proﬁle of the photoresist layer across x and y directions. The correspondent thicknesses are summarized in Table 2.
Fig. 3. Left: experimental andsimulated Δand Ψspectraof the photoresist layer;the measured data are represented by bullets, the simulation by straight lines. Right: determined optical
constants; the optical constants obtained by single spot analysisare presented by continuous lines, the optical constants obtained by multispot analysis are presented by dotted lines.
603D. Rosu et al. / Thin Solid Films 571 (2014) 601–604
For determining the thickness of the resist, a Cauchy dispersion
model  was used up to 3 eV. A multi-sample analysis option allowed
to simultaneously ﬁt 10 measurements (from 10 different spots on the
sample) considering that the optical constants are the same for each
spot and only thethickness of the ﬁlm is changing. The thicknesses sum-
marized in table 2 are obtained using the multi-sample analysis. The
thickness proﬁlealongthexandyaxesispresentedinFig. 2 for a better
understanding. The photoresist layer is much more inhomogeneous
than the SiO
due to its manufacturing process by spin-coating. There-
fore, the photoresist layer is not suitable for studying the consistency
of the ellipsometric measurement depending on the lateral resolution.
Instead, the uncoated parts (Si with only the native oxide layer) were
used. Any blending of the coated areas into this measurement will
immediately generate a large error in the result.
Additionally, the photoresist-coated sample is a good model for a
non-ideal sample: a thick layer of a partially absorbing material on a Si
To obtain the optical constants of the studied layers, the entire
energetic range was extended up to 6 eV and a Gauss oscillator model
was applied. The determined refractive index at 365 nm is 1.7, very
close to the value of 1.69 offered by the manufacturer .Inorderto
conﬁrm the invariability of the optical constants, a multi-spot analysis
(spots 6–11x and 9–12y) and a single spot analysis for the middle
spot on the sample were performed. The results obtained are presented
in Fig. 3. A slight discrepancy in the calculated optical constants can be
noticed towards the higher energy range. That can be attributed to un-
certainties involved in the analysis and therefore exclude a variation of
the optical constants with the thickness. A particular attention was
accorded to the inhomogeneity inside the measured spot (~5 mm).
Any inhomogeneity of the layer is transferred as an inhomogeneity in
the state of the polarization of the reﬂected beam. A thickness inhomo-
geneity of 1.28% which translates into ~20 nm was calculated.
Table 3 lists the results of the native oxide thickness determination
within the patterned structure in the middle of the sample coated
with photoresist. The related ellipsometric raw data and ﬁt model
results are depicted in Fig. 4. All measurements on the native oxide
can be ﬁtted within the measurement accuracy of the 4-zone averaged
nulling ellipsometer data.
Spectroscopic ellipsometry was used to study the inhomogeneity of
ﬁlms and photoresist layers. The homogeneity of the SiO
proved to be maximum 1% (3 nm) across the sample qualifying it as
patterned large-area reference sample for ﬁlm thickness. Further, it
could be shown that ellipsometers with different measurement princi-
ples (RCE, Nulling ellipsometer) yield consistent results when used to
measure these samples.
The local and global inhomogeneities of the photoresist layer were
determined with imaging ellipsometry. The sensitivity of the measure-
ment process to the patterned nature of the sample was determined
by measuring the thickness of the native oxide on the uncoated parts
of the surface between the stripes coated with photoresist. It could be
shown that the accuracy of the thickness determination is not depen-
dent on the type of the ellipsometric measurement scheme or on the
magniﬁcation of the imaging ellipsometer lens. As long as the lateral
patterns of samples can be resolved by an imaging ellipsometer, the
samples can be measured with the typical accuracy and precision of
ellipsometric thickness determination.
This work was funded through the European Metrology Research
Programme (EMRP) Project IND07 Thin Films. The EMRP is jointly
funded by the EMRP participating countries within EURAMET and the
 V. Bhatt, S. Chandra, J. Micromech. Microeng. 17 (2007) 1066.
 Y.P. Li, C.H. Henry, IEE Proc. Optoelectronics 143 (1996) 263.
 J.K. Hong, H.S. Yang, M.H. Jo, H.H. Park, S.Y. Choi, Thin Solid Films 308–309 (1997)
 G.E. Jellison Jr., Thin Solid Films 313–314 (1998) 33.
 G.E. Jellison Jr., Thin Solid Films 234 (1993) 416.
 A. Krumrey, M. Hoffmann, G. Ulm, K. Hasche, P. Thomsen-Schmidt, Thin Solid Films
459 (2004) 241.
 K. Hasche, P. Th omsen-Schmidt, M. Krumrey, G. Ade , G. Ulm, J. Stuempel, S.
Schaedlich, W. Frank, M. Procop, U. Be ck, Proceedings of the Society of
Photo-optical Instrumentation Engineers (SPIE), 5190, 2003, p. 165.
 R.M. Azzam, N. M. Bashara, Ellipsometry and Pol arised Light, No rth Holland,
 G. Jin, R. Jansson, H. Arwin, Rev. Sci. Instrum. 67 (1996) 2930.
 Q. Zahn, J.R. Leger, Appl. Opt. 41 (2002) 4443.
 D. Hönig, D. Möbius, J. Phys. Chem. 95 (1991) 4590.
 R.M. Azzam, N.M. Bashara, J. Opt. Soc. Am. 61 (1971) 1380.
 W.H. Press, S.A. Teulolsky, W.T. Vetterling, B.P. Flannery, Numerical Recipes in C,
The Art of Scient iﬁc Computing, 2nd edition Cambridge University Pr ess,
New York, 1992.
 C.M.Herzinger, B. Johs,W.A. McGahan, J.A. Woollam, J. of App.Phys. 83 (1998) 3323.
 Harland G. Tompkins, Eugene A. Irene, Handbook of Ellipsometry, William Andrew,
Results of the ellipsometric determination of native oxide layer thicknesseson structured
samples (imaging nulling ellipsometer).
Stripe width 150 100 50 25
Thickness of SiO
layer/nm 1.7 1.6 2.1 2.1
Fig. 4. Measurement of the native oxide covered areas in the patterned photoresist layer
with imagingellipsometry (ellipsometricraw data). The symbols represent the measured
quantities, and the straight lines the corresponding simulations.
604 D. Rosu et al. / Thin Solid Films 571 (2014) 601–604