PROCEEDINGS OF SPIE

[Link]/conference-proceedings-of-spie

Measurement of rutile TiO2 dielectric

tensor from 0.148 to 33 um using
generalized ellipsometry

Thomas Tiwald, Mathias Schubert

Thomas E. Tiwald, Mathias Schubert, "Measurement of rutile TiO2 dielectric

tensor from 0.148 to 33 um using generalized ellipsometry," Proc. SPIE 4103,
Optical Diagnostic Methods for Inorganic Materials II, (11 October 2000); doi:
10.1117/12.403587

Event: International Symposium on Optical Science and Technology, 2000,

San Diego, CA, United States

Downloaded From: [Link] on 04 Jun 2023 Terms of Use: [Link]

 Measurement of rutile Ti02 dielectric tensor from 0 . 1 48 to 3 3 im using
generalized ellipsometry
Thomas E. Tiwalda and Mathias Schubertb

aj A. Woollam Co., Inc., 645 M Street, Suite 1 02, Lincoln, Nebraska, 68508.

bCenter for Microelectronics and Optical Materials Research, and Department of Electrical Engineering,
University ofNebraska, Lincoln Nebraska 68588-0511

ABSTRACT
We have determined the complex uniaxial dielectric tensor of bulk rutile titanium dioxide (1 10), (100) and (1 1 1) samples
using reflection generalized ellipsometry, which measures both the diagonal and off-diagonal elements of the reflection Jones
matrix. Data were acquired using three commercially available ellipsometers, each covering the following spectral ranges:
0. 148 to 0.292, 0.200 to 1 .7 jim; and 1.7 to 33 tim. Generalized ellipsometry measures three complex ratios involving all
four Jones matrix elements. In principle, this means that the complex dielectric tensor of a uniaxial crystal can be determined
in a single measurement, provided that the sample is oriented such that the off-diagonal components of the Jones matrix are
non-zero. To improve our results, we measure the samples at several rotational orientations around the surface normal. This
insures that the probing electric fields vibrate along substantially different directions with respect to the optic axis. In some
cases, we also varied the angle of incidence. The dielectric tensor was determined at every wavelength directly from a
simultaneous fit to data from all rotational orientations and incident angles. A similar methodology should be applicable to a
wide range of anisotropic optical materials.

Keywords: generalized ellipsometry, anisotropy, titanium dioxide, ruffle, infrared, visible, ultraviolet, optical constants

1. INTRODUCTION
The purpose of this study was two-fold. First, to determine the dielectric response of Ti02 in detail, over the widest possible
spectral range; and second, to develop a methodology to acquire and analyze generalized ellipsometiy (g-SE) data that allows
one to determine the anisotropic dielectric response and optic axis orientation from a minimum set of data.

Precise knowledge of the optical properties of bulk rutile Ti02 across a wide spectral range not only provides insight into the
fundamental properties of the material, but also can be applied to practical applications where Ti02 is used as a substrate
material for optical and electronic devices. The rutile Ti02 dielectric function for various spectral ranges was explored in
previous power reflectivity studies, some of which are listed by Ribarskyl . These studies include Gervais and Piriou2 (1 1.1
—
36.4 tim), Devore3 (0.425 — 1.5 Rm), Cardona and Harbeke4 (0.1 1 — 1.2 tim), Vos and Krusemeyer5 (0.207 — 0.413 tim),
and Cronemeyer6 (2.42 — 8.04 im). Previous ellipsometnc studies of bulk Ti02 include Schubert7 (0.276 —0.354 tim) and
Jellison8 (0.248 — 0.886 jim).

G-SE analysis of anisotropic materials is a promising technique because it combines precision with information-rich spectra.
Each measurement yields six different spectra. Together, those spectra describe the real and imaginary parts of three
complex ratios involving all four elements of the normalized Jones matrix. Reflection-mode g-SE measurements can
determine both the real and imaginary parts of a material' s dielectric tensor, even in spectral regions where the material is
strongly absorbing. Also, the effects of surface roughness, thin oxides or other surface films can be eliminated with
appropriate modeling and numerical curve fitting algorithms. Although there are many instances where it is appropriate to
express the dielectric tensor as a set of parametric equations, the emphasis here is to determine these values on a wavelength-
by-wavelength basis via a direct fit to the data.

Optical Diagnostic Methods for Inorganic Materials II, Leonard M. Hanssen, Editor,
Proceedings of SPIE Vol. 4103 (2000) 2000 SPIE. . 0277-786X/OO/$1 5.00 19

Downloaded From: [Link] on 04 Jun 2023

Terms of Use: [Link]
 2. EXPERIMENTAL
2.1. Samples and instrumentation
Ellipsometric measurements were made on three bulk, single crystal rutile titanium dioxide samples, each approximately 1 xl
cm2 and approximately 350 im thick. The (1 1 1) sample was polished on only one side, while (100) and (1 10) samples were
polished on both sides.

To cover the widest possible spectral range, data were acquired on three commercially available ellipsometers9: an Infrared
Variable Angle Spectroscopic Ellipsometer (JR-VASE®) for the mid-JR wavelengths (2 — 30 jim, 330 — 5000 cm1, 0.41 —
0.62 eV), a VASE® for the NJR-VJS-UV wavelengths (0.200 — 1 .700 tim, 0.729 — 6.5 eV) and a prototype of the VLJV-
VASETM for the VUV wavelengths (0.148 — 0.292 nm, 4.25 — 8.5 eV). All measurements were made at room temperature.

The IRVASE® instrument combines a Fourier-transform-based infrared spectrometer with a rotating-compensator variable
angle ellipsometer. The beam is 8 mm wide at the sample, with a 3° angular spread. For this analysis, data were taken with a
resolution of 2 cm' and at an incident angle t. of 72°, which yielded ellipsometric spectra with reasonable sensitivity and
signal-to-noise ratios.

VUVVASETM is a rotating analyzer ellipsometer equipped with a computer-controlled MgF2 Berek waveplate as a
compensator. The beam-path is purged with dry nitrogen gas to avoid oxygen and water absorption bands. Deuterium and
Xenon lamps provide light for the shorter and longer wavelengths, respectively. This light is passed through a double-
chamber Czemy-Tumer type monochromator to provide wavelength selection and stray-light rejection. To ensure adequate
spectral resolution, the computer-controlled slit width can be adjusted to obtain bandwidths as small as 0.5nm (25 meV at 8
eV). The detector incorporates a PMT for the short wavelength region (146 nm to 250 nm) and a silicon photodiode for
longer wavelengths. The measurements were taken at an incidence angle of 75°.

Additionally, data in the NIR-VJS-UV (200 — 1 700 nm, 0.73 — 6.5 eV) spectral range at 0.025 eV steps were taken with a
VASE® ellipsometer equipped with a precision sample rotator, which can orient the sample around the surface normal at an
angle, &t, relative to some starting position. The rotator was used on the (1 1 1) substrate, which was rotated at & of 450to
90°. Like the VUVVASE®, this instrument also uses a computer-controlled MgF2 Berek waveplate as a compensator. Data
were acquired at incident angles 'a of4O° and 700, at each orientation angle ö4.

Both the computer-controlled Berek waveplate (used on the VASE® and VUV-VASE®) and the rotating compensator (used
on the JR-VASE®) greatly extend the instruments' measurement capabilitieslO'l I . These extended capabilities include the
determination of all elements of the normalized Jones matrix for non-depolarizing samples like those found in this study.

2.2. Uniaxial anisotropy of rutile

To determine the dielectric tensor of rutile, one must also locate the orientation of the c-axis (optic-axis) relative to the
measurement surface of the sample. The dielectric tensor of uniaxial materials like rutile can be expressed as

E0 0 0
â=• 0 E 0 A, (1)
0 0 Ee
The quantity Ee is the material's dielectric response to E-fields parallel to the c-axis of the crystal, while is the response to
E-fields perpendicular to the c-axis (parallel to the a-plane of the crystal). Furthermore, Eeand are generally complex, and
Ee = Cei + lEe2 = (fle + jke)2 and E = + = (n0 + ik)2. The matrix A, the orthogonal rotation matrix, is defined as

cosØ sinçt 0
A = — cos 0 sin 0 cos 0 cos 0 sinO . (2)
sinOsinØ —sinOcosO cosO

20 Proc. SPIE Vol. 4103

Downloaded From: [Link] on 04 Jun 2023

Terms of Use: [Link]
 X-Z plane: plane of incidence

:ci) x

X-Y plane: sample surface

Figure 1 . Definition of Euler angles (4,O), which describe the direction of the optic- (c-axis) of a uniaxial material with respect to the
laboratory frame ofreference. (The third Euler angle, ip, is not defined for a uniaxial material.) The coordinates (x,y,z)
refer to the laboratory coordinates, as defined by the sample surface (x-y plane) and the plane of incidence (x-z plane), with
the reflected light beam traveling in the positive x and z directions.

The Euler angles (4,9) describe the orientation of the optic axis with respect to the laboratory frame of reference as defined
by the intersection of the plane of incidence with the plane ofthe sample surface (see Figure 1). (The third Euler angle, iji,is
not defined for a uniaxial material). The plane of incidence forms the x-z plane, and the sample surface forms the x-y plane,
with the reflected light beam traveling in the positive x and z directions. For a uniaxial material, only 0 (the tilt of the optic-
axis away from the z-axis) and 4 (the counter-clockwise rotation of the optic axis with respect to x) are meaningful.

Rutile Ti02 has a uniaxial optical response, with optic axis parallel to the c-axis. Thus, for the (100) and (110) samples, the
c-axis lies in the sample (or x-y) plane. Obviously, the c-axis lies outside both the incident and sample planes for the (111)
sample.

2.3. Generalized Ellipsometry

The Jones matrix J describes the reflection (or transmission) optical transfer function of a non-depolarizing sample. That is,

E;t = J(EP L (r i;, 1E

where
Et) E") 1E
(2)

etc. are the Fresnel reflection coefficients for p- and s- polarized light, respectively. (for transmission
ellipsometry, they can be replaced with the Fresnel transmission coefficients t,, t,, etc.). Standard ellipsometry measures the
ratio of the diagonal components, since the off-diagonal components are assumed zero (i.e., there is no mode conversion
between the p- and s- polarized waves). Generalized ellipsometry (g-VASE) measures three independent complex ratios that
combine all four components of the Jones matrix. For our study, we will define the following three ratios:

AnE = tan(')e' = PP
(3a)

Aps = tan('55 = (3b)

Asp = tan('Iç. = (3c)

Proc. SPIE Vol. 4103 21

Downloaded From: [Link] on 04 Jun 2023

Terms of Use: [Link]
 For an isotropic sample, Aps = Asp = 0, and AnE reduces to the standard ellipsometric ratio. For transmission, etc.
are replaced with etc.

2.4. Data acquisition strategies

The key to extracting and from a uniaxial material is to have the probe beam's electric fields excite a significant
dielectric response both parallel and perpendicular to the optic axis. For spectroscopic ellipsometry, fulfilling this
requirement can be complicated by the necessity to probe the sample at oblique angles. Dispersion causes the angle of
refraction to vary with wavelength, altering the direction of the electric fields. The angle of refraction can vary dramatically
in regions of strong dispersion, especially near poles and zeros in Re{Ee,o}. Dramatic examples can be found in polar
materials with strong TO-LO phonon splitting (including SiC12 and sapphirel3, as well as Ti02); in those cases, the probe
beam refraction angle changes from essentially parallel to the surface near the LO-phonon energies, when eo 51fl(j (( is the
angle of incidence), to angles nearly normal to the surface as ne,o > 1.

For many materials ne,o > I , particularly in the visible and ultra-violet spectral range. In this case, the probe beam is refracted
toward the surface normal and the electric fields primarily interact with the material's dielectric response parallel to the
surface plane. This is particularly true for high index materials. Thus, for samples whose optic axis is normal to the surface,
reflection measurements (including reflection ellipsometry) cannot determine a material' s extraordinary dielectric response.
Fortunately, this is not the case for the samples used in this study.

Jellison, et a!.8 observe that a single g-VASE measurement is sufficient to determine both £ and provided that a significant
amount of mode conversion occurs (i.e., and 0). This condition is met when the optic axis is not aligned with either
the sample (x-y) or incident (x-z) planes (see Fig. 1 ). Of course, additional measurements taken at different rotational
orientations and incident angles can provide more accurate results.

If an a-plane sample (optic axis parallel to surface) is available, Ee and E0 can also be determined from two standard
ellipsometric measurements. This is accomplished by aligning the optic axis parallel to the plane of incidence (c-axis (x) for
one measurement, then perpendicular to the plane of incidence (c-axis y) for the next'3'l4. By orienting the c-axis along
these two directions, = 0, which eliminates the need for the more complicated generalized ellipsometry analysis (e.g.,
see Azzam and Basharal 5, or Schubertl 6). One can also augment these two standard ellipsometric measurements with a g-
SE measurement with the c-axis oriented anti-parallel to both x and y. Such a procedure was used for the mid-infrared
measurements ofthe (1 10) sample.

Finally, in spectral regions were the sample is transparent, the front and back surface reflections can mix incoherently at the
detector. If the mixing is completely incoherent, the effect can be reasonably approximated using methods similar to those
described by Johsl 7(e.g., see Ref. 1 8). Unfortunately, these sub-mm samples are in a thickness regime where the front and
backside reflected beams mix in a partially coherent manner at mid-infrared wavelengths. Modeling this partially coherent
mixing is problematic. For our purposes, the best solution was to measure samples with roughened backsides, which
effectively eliminates specular backside reflections, or restrict our measurements to the absorbing regions ofthe spectrum.

2.5. Data analysis

We analyzed the g-SE data using the formalism developed by Berremanl9, as modified by Schubert, et al.7"6 In previous
anisotropic studies, Jellison, et al.8'20 have used similar methods. The model parameters were optimized using the same
numerical regression analysis described by Herzinger et al.2 1 The general approach is to vary the optical properties of the
substrate and the thickness of any overlayer until the model data best fit the experimental data (i.e. , a minimum in the mean
squared error between the experimental data and the generated data from the model). For each spectral range, (mid-IR, NIR-
Vis-UV, and VUV) data from all the different sample orientations are simultaneously fit using the "multisample analysis"
technique (see examples in Ref. 1 1 ). The g-SE formalisms, numerical regression and multisample analysis capabilities were
all contained in the software package provided with the elhpsometers9.

Surface roughness can significantly alter the measured optical constants of a bulk material, particularly at shorter
wavelengths8"0'11'20'21. We modeled surface roughness by adding a thin layer consisting of 50% void — 50% substrate (using a
Bruggeman Effective Medium Approximation). Such modifications were unnecessary for the infrared data, since longer
wavelength measurements are less sensitive to surface effects.

22 Proc. SPIE Vol. 4103

Downloaded From: [Link] on 04 Jun 2023

Terms of Use: [Link]
 The analysis proceeded as follows: first, the data in the transparent region (near-JR and visible wavelengths) were fit to a
Cauchy model with a surface roughness layer. This established the thickness of the surface roughness layer, provided good
values for the optical constants in the transparent region, and established the approximate orientation of the optic axis. From
that point, the surface roughness layer thickness was fixed and the complex values of Ce and s, were fit on a wavelength-by-
wavelength basis over the entire wavelength range ofthat particular measurement.

3. Results

3.1. (110) sample

In the (1 1 0) sample, the c-axis is oriented parallel to the sample surface. For the mid-JR spectral range, standard
ellipsometric data were acquired at both c-axis x and c-axis y orientations, for reasons described in Section 2.3 1 . These
measurements were augmented with g-VASE data acquired at 4 anti-parallel to both x and y. All data were acquired at 72°
incident angles. The mid-JR complex ordinary and extraordinary dielectric functions, as well as the Euler angle 4 for anti-
parallel orientation were obtained from the simultaneous analysis of data from all three orientations. The data and fit are
shown in Figs. 2a and 2b.

The NIR-VIS-UV g-SE data were taken at two sample orientations approximately 90° apart. The data from one of the
orientations ( = 26.5°), along with the best-fit results, are shown in Fig. 2c. Again, the complex ordinary and extraordinary
dielectric functions, as well as the Euler angles for both orientations, were obtained from a multisample analysis.

The VUV g-SE data were also taken at three arbitrary orientations, (orientation =48.4° is shown in Fig. 2d), and analyzed
in a manner similar to that described above.

Table I lists 4 for all spectral ranges and orientation angles. Except for the first two orientations of the mid-JR measurements
(which were assumed to be approx. 0° and 90° w.r.t x-ax:is), all 4 were determined from a fit to data. Three of the
orientations were measured at a 72° angle of incidence; the fourth orientation was measured at 30° and 55°.

The real and imaginary parts of C0 Ce determined from a wavelength-by-wavelength fit to data from the (110) sample are
presented in Figs. 3a and 3b. These can be compared to results from the other samples in Figs. 4a —4d.

Table I. Euler angle for the (110) sample. The angle 0 was
assumed to be 90° for all orientations.

(in degrees) for (110) sample

mid-IR UV-visible-NIR VUV
(AOl = 72°) (AOl 40° & 70°) (AOl = 75°)
= 0° 26.5 -1.3
= 90* 129.0 -87.6
18.5 48.4

* Sample was onented approx. (c-axis x) and

(c-axis II y) (See text).

Proc. SPlE Vol. 4103 23

Downloaded From: [Link] on 04 Jun 2023

Terms of Use: [Link]
 50 5 0 (clix) 150

, 40

30
5 90'
(cy)
V .
f '\
''\. /''.'
/ \\
5 90
120

60
90 (cily)

/2' '.\
7
\\/// Y
- 20 (clix) 30 /// 0 (clix)
K' mod&
10 data
I
\ / 4) 90 (cy) 0 _...___._."\)
—--data
..—. 4) 90 (city)
- I I I

5 10 15 20 25 30 35 5 10 15 20 25 30 35
Wavelength (pm) Wavetength (pm)

Figure 2a. Infrared standard SE and L\ spectra for (110) Ti02 sample. The onentation angles are approximate.

nn
50 /\ 4=18.5° /A ExpAnE 72

—
I
200
-

/'\\ a)

I :: g 100 I
,,_/"\V=\KW2td'
M' i' .
.6
20 / /) \ j
'11i
\ I — Model FIt
V --'--ExpAnE 72
- ExpAps72
-
ExpAsp72
0 /I p/
y5
10
4) = 18.50

5 10 15 20 25 30 35 5 10 15 20 25 30 3
Wavelength (pm) Wavelength (pm)

Figure 2b. Infrared g-SE 'P and spectra for (110) Ti02 sample. The orientation angle = 18.5° was determined by
fit to the data.

30
/'\\\\
7
'I'-
—--- Model Fit
. Exp AnE .7O
——-ExpAps-70
150
/
1\ /
a)
I \\, Exp Asp
100 /I \\,7 —.— Model Fit
ExpAnE'70

\
/j
ft A — — -Exp Aps 70
20
i\ I \
- - ExpAsp70
I ° 50
4)26.5°
10 #t_
4=26.5°
\j
\f 0
/-;-:-;-. , -. . -----T

300 600 900 1200 1500 1800 '0 300 600 900 1200 1500 18 )0
Wavelength (nm) Wavelength (nm)

Figure 2c. UV-visible-NIR g-SE 'P and spectra for (110) Ti02 sample. The onentation angle 4 = 26.5° was
determined by fit to the data.

= —P— -__________________________ 30C

4) = 484° ,/ //
/
30 4)=48.4°

a)
S 20
a)
25 7\\ ,5'_"S_S 5 ==_=
— Model Fit
-----ExpAnE 75
a)
a)

, 100
200

I ,-;--',=
V'
' ExpAps 75
Exp Asp 75

15 -'-S ExpAps75 —

? io Exp Asp 75 02 0
0S -?"--='
5
S

Hi
(o/
,...
______________________________________________ -101.
, . ,vS I.'
120 150 180 210 240 270 300 120 150 180 210 240 270 300
Wavelength (nm) Wavelength (nm)

Figure 2d. UV-visible-NIR g-SE 'P and spectra for (110) Ti02 sample. The onentation angle 4 = 484° was
determined by fit to the data,

24 Proc. SPIE Vol. 4103

Downloaded From: [Link] on 04 Jun 2023

Terms of Use: [Link]
 [Link] 50 ! - tio2_ordinary
2C
o2extraordtna to2.e*traordinary
40
!
0 \\ [\\ 0
30
-2C
a) 20
a)
0
a)
-40
gb
E E2 ) \- //\
) 10 20 30 0 10 20 30 40
Wavelength (pm) Wavelength (pm)

Figure 3a. Real and imaginary parts of0 and r versus wavelength for Ti02 (1 10) sample.

20 . . .

. 20 ,.. —tio2ordinary I .. -tio2_ordinary

,, 02_extraordinary . - ---tio2_extraordinary
C '.,
C 15

C
15 ,; r C
0
10 00 10 :;\\ 55 "-S

at : \'
0 — . at
0 5 i \.\ ,,S'

C1 at C2
E
0 2 4 6 8
n
0
I ,)I . t S

10 2 4 6 10
Photon Energy (eV) Photon Energy (eV)
Figure 3b. Real and imaginary parts of E0 and Ce versus photon energy for Ti02 (I I 0) sample.

3.2. (100)sample Table II. Euler angle for the (100) sample. The angle 0
For the (100) sample all data were acquired in the g-SE was assumed to be 90° for all orientations.
mode, without assuming that 4 was aligned with either the 0 (in degrees) for (100) sample
x- or y- axis. Otherwise the approach was identical to that
described for the (110) sample in the previous section. mid-JR UV-visible-NIR VUV
Table II lists 4 for all spectral ranges and orientation angles. (AOl = 72°) (AOl = 40° & 70°) (AOl = 75°)
All 4 were determined from a fit to data. The real and -1.8 87.9 46.3
imaginary parts of c0 and Ee determined from the 37.4 -1.8 68.5
wavelength-by-wavelength fit for this sample are nearly
identical to the (110) sample results, as can be seen in Figs. 88.3
4a—4d.

3.3. (111)sample Table III. Euler angles and 0 for the (111) sample.
There was no choice but to acquire data for the (1 1 1) sample
0 & U (in degrees) for (111) sample
In the g-SE mode, since the optic axis, in general, will not
be aligned with x-, y- or z- axes. This also means that we mid-JR UV-visible-NIR VUV
must extract both 4 and 0 from the data. Again, we (AOl = 72°) (AOl = 400 & 70°) (AOl = 75°)
simultaneously fit data from all orientations. Table III lists U U 0 U 0
0 and 4 for all spectral ranges and orientation angles.
40.2 21.2 43.3 254.3 43.0 29.7

3.4. Comparison of Ee and for all samples 250.4 299.3* 248.4

The dielectric response of all three samples are presented in 294.1 3443* 299.7
Figs. 4a — 4d. Both and Le for the (Ill) sample are 343.0 347.3
noisier than the (110) and (100) sample results. Particularly
note the noise in E for A. > 21 jim of the (111) sample. This * Using a precision rotation stage9, the sample was
suggests that our choice of incident angles and rotational rotated = 450 and & = 90° from starting
orientations were such that the input beam was not orientation of 0 = 254.4°.

Proc. SPIE Vol. 4103 25

Downloaded From: [Link] on 04 Jun 2023

Terms of Use: [Link]
 sufficiently probing the extraordinary dielectric response. Including measurements at additional angles of incidence and
rotational orientations might reduce this uncertainty. This is not a problem for the (100) and (110) samples, since the probe
beam plane of incidence can be freely rotated with respect to the optic axis. This virtually guarantees that both dielectric
responses are probed.

(111)
40 (100)
-—(110)

30
00 0
20
a)
a) a)
0 0 10
a)
a
Ee2
20 0 10 20 30 40
Wavelength (pm) Wavelength (pm)

Figure 4a. Real and imaginary parts of r versus wavelength for all Ti02 samples.

20 -—-(111)
td (100)
(110)
15 ,;r
0
/2 \ 00

,
10

a)
0
.50
a) Eei
,\ a)
a)
0
E
-ç
0 2 4 6 8 1 4 6
Photon Energy (eV) Photon Energy (eV)

Figure 4b. Real and imaginary parts of a versus photon energy for all Ti02 samples.

. 20 - t ——(111) Sc
—(ill)
j\

\
----(100) .
(100)
10
N -(110)
40 (110(

0
-\ 0

(\ c 30
-
-.-
I \ .
10 - - 20

ji
a)
0 0
8a)-20 - E01 , ic Eo2
E
, ,
0 10 20 30 4 ) C 10 20 30 4
Wavelength (pm) Wavelength (pm)
Figure 4c. Real and imaginary parts of r versus wavelength for all Ti02 samples.
10
W 15 . -——(111) (111)
/A\ 12 N
I (100) )\ ----(100)
12 -
/\ ——(110)
I\ (110)

.3
a)
-

.
t--.
- .—
,,/
I
0 9
0
a)
a)
0
6
C I
I
\\ -
E
3 o2 I

I I I 2' ) . ) .

0 2 4 6 8 10 _0 2 4 6 8 10
Photon Energy (eV) Photon Energy (eV)

Figure 4d. Real and imaginary parts of c, and e versus photon energy for all Ti02 samples.

26 Proc. SPIE Vol. 4103

Downloaded From: [Link] on 04 Jun 2023

Terms of Use: [Link]
 3.5. Comparison to results from other studies
The real and imaginary parts of Ee and E0 from this study (sample (1 10)) are shown in Figs. 5a — 5d. Generally, the results
from this study match previous studies3'6 very well in the transparent regions (approx. 0.38 2 6.2 jim, 0.19 co 3.2
eV). In the UV absorbing regions (2c 0.38 or co 3.2 eV), there is some agreement with previous reflectivity studies4'5,
although lower resolution of the previous studies make a comparison difficult. In the JR absorbing regions (A. 0.38 or 0
0.19 eV), these results are in excellent agreement with Gervais2, particularly for e0.

- 20
(0
udy
15
00
10
a,
a,

—-
20 40
Wavelength (pm) Wavelength (pm)
Figure 5a. Real and imaginary parts Of Ce versus wavelength (2 — 33 jim) for sample (110), with data from Refs. 2 and 6.

A
A
0 Ref.3
A Ref.4
25 0 Ref.3
20 Eel A Ref.4
0 Ref.5
A

00
a,
a,
0
a,

10 2 4
Photon Energy (eV) Photon Energy (eV)
Figure 5b. Real and imaginary parts Of Ce versus photon energy (0.7
8.5 —
eV) for sample (110), with data from Refs. 3, 4 and 5.

(" 50
study
g
0
0 30

i
a,
20
0 10

I
10 20 30 40
Wavelength (pm) Wavelength (pm)
Figure 5c. Real and imaginary parts of e0 versus wavelength (2 — 33 jim) for sample (110), with data from Refs. 2 and 6.

0 Ref.3
I,----, ( ,

- 15 1 0 Ref 3
Ei A Ref 4
a Ref 5
— This study
A Ref 4

10 E02
00 00
a,
a,

a,
E

0 2 4 6 8 10 0 2
]'\AAAAA
4 6 8 10
Photon Energy (cv)
Photon Energy (eV)
Figure Sd. Real and imaginary parts ofe0 versus photon energy (0.7 — 8.5 eV), for sample (110) with data from Refs. 3, 4 and S.

Proc. SPIE Vol. 4103 27

Downloaded From: [Link] on 04 Jun 2023

Terms of Use: [Link]
 The (110) sample results are compared at a few data points with previous g-SE studies7'8 along with older reflectivity
studies3'4'5 in Table IV. The (110) results generally match Ref. 8 better than Ref. 7. This is likely because Jellison et al.8
accounted for surface roughness, while Schubert et al.7 did not.
Table II. Comparison of Ti02 rutile dielectric constants at selected wavelength from this and other studies (Dielectric
constants were calculated from refractive index and extinction coefficients obtained by other studies).

Energy (eV) Ref. Re{Ee} Im{r} Re{0} Im{r0}

4.31 This study* 5.894 18.84 4.525 9.937
4 0.0711 25.28 4.697 10.30
7 1.115 14.08 2.785 8.686
8 6.271 19.908 5.289 11.386
3.70 Thisstudy* 19.54 5.21! 14.88 3.436
5 17.19 6.651
7 15.25 10.25 12.53 6.497
8 19.830 4.554 15.46 2.875
1.77 Thisstudy* 7.845 6.459
3 8.01 6.50
8 8.015 6.548

* from sample (1 1 0) data.

4. Conclusions
This study demonstrates the capability ofg-SE to measure the dielectric tensor and Euler angles ofbulk Ti02 continuously
over a wide spectral range (0.148 — 33 jim), thus extending the work of Schubert et al.7 and Jellison et a!.8 on Ti02. The
simultaneous analysis of multiple orientation and angle of incidence measurements improves the results.

Acknowledgements
This work was partially supported by the U.S. National Science Foundation SBIR contract 9901510.

References
I M. W. Ribarsky, "Titanium Dioxide (Ti02) (Rutile)," in Handbook ofOptical Constants ofSolids, edited by E. D. Palik,
pp. 795-804, Academic Press, San Diego, 1985.
2 F. Gervais and B. Pinou, Phys. Rev. B 10, pp. 1642-, 1974.
3 J. R. Devore, J. Opt. Soc. Am. 41, pp. 416-, 1951.
4 M. Cardona and G. Harbeke, Phys. Rev. 137A, pp. 1467-, 1965.
5 K. Vos and H. J. Kmsemey;er, J. Phys. C 10, pp. 3893-, 1977.
6 D. C. Cronemeyer, Phys. Rev. 87, pp. 876-, 1952.
7 M. Schubert, B. Rhein!änder, J. A. Woollam, B. Johs, and C. M. Herzinger, "Extension of rotating-analyzer ellipsometiy
to generalized ellipsometry: determination ofthe dielectric function tensor from uniaxial Ti02," J. Opt. Soc. Am. A 13,
pp. 875-883, 1996.
8 G. E. Jellison, Jr., f. A. Modine, and L. A. Boatner, "Measurement of the optical functions of uniaxial materials by two-
modulator generalized ellipsometry: rutile (Ti02)," Opt. Letters 22, pp. 1808-1810, 1997.
9 Available from J.A. Woollam Co., Inc., 645 M St., Suite 102, Lincoln Nebraska 68508, USA.
10 J. A. Woollam, B. Johs, C. M. Herzinger, J. Hilfiker, R. Synowicki, and C. L Bungay, "Overview of variable angle
spectroscopic ellipsometry (VASE), part I: basic theory and typical applications," in Optical Metrology, edited by G. A.
Al-Jumaily, pp. 3-28, SPIE, 1999.
1 1 B. Johs, J. A. Woollam, C. M. Herzinger, J. Hilfiker, R. Synowicki, and C. L Bungay, " Overview ofvariable angle
spectroscopic ellipsometry (VASE), part II: advanced applications," in Optical Metrology, edited by G. A. Al-Jumaily
(SPIE, 1999), p. 29-56.
12 T.E. Tiwald, J. A. Woollam, S. Zoliner, J. Christiansen, RB. Gregory, T. Wetteroth, and SR. Wilson, and AR. Powell,
"Carrier concentration and lattice absorption in bulk and epitaxial silicon carbide determined using infrared
ellipsometry," Phys. Rev. B, 60, pp. 11464-11474, 1999.

28 Proc. SPIE Vol. 4103

Downloaded From: [Link] on 04 Jun 2023

Terms of Use: [Link]
 t 'vii 'iiqnqog jJ
'pUMtJ 'iouIz1oH paIU1Ju,, tipo[oip XdoJosIuU pu uouoqd spoin jo 'aitqddUssiqj AoJ
H dd'J9 000Z'10Z8L818
j71 I ')IJ1WflH 'I 'UUH UU 'J,\J 'UUOpJU poJiJUJSUOIEJqLAUI i7OEOJSW1 pUE 'iOIV1SW1 Sf1Ic1 1 "U '19 dd 18 L8

ci d
91

Li
'IOZS 000Z

i i" '
'V anzzv pui p j,s 'Enqs1 ,iawosthr puv pzuvoj
'4Joqnqog uopuodop-uoizujod,,
dd

sainpm 8uisç dujnjtj

'L-c9z9661

ij
'qrj
dd 'sc-ot 'puEO-qJOMN ')JJOA LL6I
iodo sJauiBd Jo Auitqi oidoiosiacsnoua8ouioqpiAi 'sinoscs S1lq
SOf )J 'H 'qoUaIj 'U ')ff)J M V 'UqDOJA pUI f \/ 'UI1100M iodO sisuy jo xoiduioj iunv
'sodAj ut pwic4j uaiafauj 'uiwo
11601 'col '91d8 'wEq2u 'YM t66t
qjjg oj
poppa icq •d 'sjoqy dd

81 1 N ')i'JI1H D JAI '1UJZJH 3 j 'un iuoi

i 'uiiooi pui r i 'uujIEoqdO,, uoqzuooiqo jo Didoiosjug
'sotsiTd /vdz,idO3UaAfJdlUJ 'Ui,WO3 '6 vso sou
'souo dd 'Z6106t1odO icopos '[Link]
61 jj 'uuon
UOUPTSM 'DU 8661

ZL6I
sudO,, ut potjuiiis pui oido1osiuE :itpU i X XUW-j 'uoqinuoj f d0 •D0S •LUy '9 dd '-zoc
OZ 0 1I 'UOSiIIf ij pui j 'iouo JoTdO,, suoqunj jo ixun ouz pouiuuop/q pziiou Vioiuosdiiio S1qd
H '8S dd '68c€98cE 8661
ZI 1A1D 'iUiZJH Od 'Jop!u 'SIJOf pui yf 'ujioo i iddV SXq 'U dd i t. i i-c 'iL t c66

OOJd ldS IOA COft 6

Downloaded From: [Link] on 04 Jun 2023
Terms of Use: [Link]