JOURNAL OF THE OPTICAL SOCIETY OF AMERICA VOLUME 64, NUMBER 6 JUNE 1974

Effects of component optical activity in data reduction and calibration of

rotating-analyzer ellipsometers
D. E. Aspnes
Bell Laboratories, Murray Hill, New Jersey 07974
(Received 20 December 1973)

An exact expression for calculating the complex reflectance ratio of a surface, from data obtained with a
rotating-analyzer ellipsometer system using optically active quartz Rochon prisms, shows that optical
activity affects relative values of measured quantities by an amount of the order of 1%. For component
settings near the normal modes of the system, these effects can be much greater. By contrast to null
ellipsometry, there is no surface for which these effects vanish in calibration. Therefore, corrections of the
order of 1% (0.5') are necessary in the calibration of the azimuth scales of these ellipsometer systems, even
in the most-favorable cases.

Index Headings: Ellipsometry; Reflectance; Polarization.

Rochon prisms' of crystal quartz, commonly used as analyzer systems in particular, show considerable
polarizing beam splitters, offer a number of substantial improvement of sensitivity over null instruments, 2'-' 0"
advantages over calcite prisms in high-optical-efficiency such ellipsometers are capable of wavelength scanning
2 5 and appear to be technologically significant. It is
automatic ellipsometers of the rotating-analyzer type. -
In this application, unpolarized flux is incident on the therefore of interest to determine the effects of small
front face of a Rochon prism (acting as a polarizer), and imperfections in these systems. Because a preliminary
emerges as two orthogonally polarized beams, one of treatment of systems of the polarization-modulation
which is undeviated and passes through a suitable type' has been given by O'Handley,6 we shall discuss
aperture to the system, whereas the other is deviated here only effects observed in instruments of the rotating-
and blocked. After passing through the system, which analyzer configuration. In Sec. I, we obtain the neces-
generally- consists of a compensator and a reflecting sary modifications of the linear-mode equations to
surface, the flux is incident on the back face of a Rochon permit the complex reflectance ratio of a reflecting
prism (acting as an analyzer), and emerges as two surface to be calculated exactly in terms of the calibra-
beams, one of which is undeviated and passes through tion azimuths and the optical-activity coefficients of
a suitable aperture to a detector; the other is deviated the prisms and the compensator (if used), from the
and blocked. Quartz Rochon prisms have much higher experimental data. In Sec. II, instrument calibration is
transmittance in the uv than calcite prisms, and discussed in detail, and effects of prism optical activity
generally show less deflection of the nominally un- in calibration are described mathematically. The
deviated beam due to different construction practices. results are discussed in Sec. III. In particular, we find
Moreover, a certain amount of pseudo-depolarization the effect of optical activity to be of order 1% in relative
is always found in the flux incident on the detector for values of experimentally determined quantities, except
beams of finite cross-sectional area, due to variation of for component settings near normal modes, where the
the amount of rotation of the plane of polarization as effect can be much larger. Optical activity influences
different beam elements travel different lengths azimuth reference angles determined in calibration by
parallel to the c axis of the material before emerging about 0.50, an error that cannot be eliminated by
from the analyzer prism. The latter two characteristics proper choice of reflecting surface, in contrast to the
are important in reducing systematic errors that result situation in null ellipsometry. 8 " 2"13
from synchronous motion of the output beam across
I. DATA REDUCTION
regions of varying sensitivity on the detector, and
from any dependence of the detector sensitivity on In this section, we derive the equations necessary to
the polarization state of the detected beam.6 However, calculate the complex reflectance ratio, p- r, /ri, of a
crystal quartz is optically active; consequently the reflecting surface, in terms of the phase and amplitude
normal modes, the waves that propagate unchanged (or normalized Fourier coefficients) of the flux incident
through the material, are elliptically polarized.7 Though on the optical detector, and the system parameters
the minor/major axis ratio is small, of the order of including activity coefficients of the various system
0.001-0.005, the effect of such ellipticity in optical components. Here, r, and rLare the complex-reflectance
components is readily observable in null ellipsometry8; ratios of the surface for the electric-field components
corrections for the standard null equations, which parallel and perpendicular to the plane of incidence,
assume linear polarization modes, have been given to respectively. Since the flux, I, must be evaluated for
first order in the ellipticity parameters.8' 9 any analyzer or polarizer azimuth, A or P, and not
Because photometric systems in general, and rotating- simply for conditions for which IO (except for align-
812
June1974CALIBRATION OF ROTATING-ANALYZER ELLIPSOMETERS
ment purposes to be discussed in Sec. II), a first-order in the Jones-matrix representation.1 6 It follows that
analysis8 91"2-' 4 is precluded and the equations obtained
must necessarily be exact.6 The form of these equations (1 -a') (1 -YA 2
)
is greatly simplified if the rotating analyzer, optical I C 1+ -
detector, and data-reduction components are treated (l+a'yA) 2 +(a+-yA)2
as a separate polarization-state detection system, with X [cos2Q cos2A +sin2Q sin2A] (4a)
the complete ellipsometer separated into the polariza-
tion-detection part, and a polarization-generation part, = 1 +a cos2A +(3 sin2A, (4b)
including the source, the polarizer prism, the compensa-
where a, (3 are the normalized Fourier coefficients
tor (if used), and the reflecting surface. We consider
determined experimentally either by direct computa-
first the detection of the polarization state.
tion4 ,5 or from phase and amplitude measurements.2' 3
A. Polarization-State Detection The mode parameters a, Q therefore, are given in
terms of the experimental variables by
The normal modes of an optically active crystal-
quartz Rochon analyzer prism have the form7
Q=4 tan'1(f/a)+-u(-a) sgn(fl), (5a)
F&= S.(&A+iYAPA), (la) 2
8 11=IS-,,(PA+ iA&A, (lb) 2
- YAD4 (1 -- YA 2) (1 .- 2)1
where 8., 8, are complex amplitudes, YAI<<1 is the a= 21~ (5b)
optical-activity coefficient, and CA, PA are unit vectors. ( +¢) -YA(-)

If these unit vectors form a right-handed orthogonal

set with ZA, which is oriented in the direction of propaga- where u(x)=0 for x<0 and 1 for x>0, sgn(x)= -1 for
tion of the light, and if S. represents the transmitted x<0 and +1 for x>0, and
mode, then &A and PA are perpendicular and parallel, E=
a2+02]1>°. (5c)
respectively, to the optic axis of the entrance half of
the analyzer prism.' The coefficient YA is positive for These equations are exact, with Q being given in the
right-handed and negative for left-handed crystal frame of reference of A. The sign ambiguity in Eq. (5b)
quartz for the time convention e-iwt. ITAI has the cannot be resolved in a rotating-analyzer system, but
measured dispersion 4 " 5 must be determined either by an approximate null
ellipsometric measurement, or by using the fact that
IYAI= 0.0010 (Ihw/eV), (2) a>0 for a bare reflecting surface in polarizer-surface-
where Ico is the photon energy. analyzer (PSA) operation. Equation (5b) shows that
The analyzer azimuth angle, A, has a zero reference, t= 0 if a = 1, proving that optical activity in the
A = 0, which is determined typically by an electrical analyzer prism does not affect its operation as a
synchronization pulse2' 3 or by a table of Fourier coeffi- circular-polarization null detector,' 7 as expected from
cients loaded in the computer memory.4' 5 For con- the symmetry of the circularly polarized mode.
venience, the analyzer prism should be positioned
so that at A=0, the major-axis vector, XA,of the B. Determination of the Complex
transmitted mode is approximately parallel to the plane Reflectance Ratio
of incidence. Exact alignment is not necessary, because
the actual azimuth angle, As, for which XA is parallel In a PSA system, the polarization state (8s,8,) of
to the plane of incidence when a reflecting surface, S, light reflected from a simple reflecting surface is given
is used, can be determined as discussed in Sec. II. In by the Jones-matrix product
the remainder of this section, As is assumed to be
known. Suppose an arbitrary elliptically polarized mode
is incident on the rotating analyzer. This mode can be
represented by its minor/major-axis ratio, a, where
Ia <1, and the azimuth angle, Q, of its major axis, cos(P-Ps)
01 -sin(P-PS)
where IQ <7r/2. This mode results in a transmitted
flux, I, incident on the optical detector. This flux is sin(P-Ps) cos(P-PS) ' 8, (6)
given to within a scaling factor by where yp is the optical-activity coefficient of the
cI (| )=(2 (3a) polarizer prism, and P is the azimuth angle of &p,
the polarizer unit vector equivalent to the analyzer unit
(6 1 )0%%%%%%A% vector XA. Because the reference azimuth P= 0 typically
is determined by the indexing-head scale, Ps is defined
cos(A-Q.) sin(A-Q)\I by analogy with As to represent the value of P for
X which &p is parallel to the plane of incidence. The
-sin(A -Q) cos(A-Q) / \a/ complex-reflectance ratio, p= r, /rL, can be obtained
 D. E. ASPNES Vol. 64

exactly by using the relations connecting the mode These equations are also exact, and together with
projections St, 8, to the experimentally determined Eqs. (4) and (5) determine p in the PCSA system.
[through Eqs. (4)] mode parameters a, Q. These
relations are II. CALIBRATION
A. PSA System
8S= S'[cos(Q-As)-ia sin(Q-As)] (7a)
As indicated by Eqs. (8), calibration of a PSA system
S%=
S'[sin(Q-As)+ia cos(Q-As)], (7b) requires determination of A s and Ps, the azimuth angles
of the plane of incidence defined by the reflecting
where As is the plane-of-incidence azimuth angle in
the frame of reference of the analyzer as defined earlier,
surface S. We suppose, for convenience, that the
prisms are positioned so that A s-Ps-0. Direct
and S' is a scaling factor. Solving for p=r 11 /rL from
Eqs. (6) and (7), we find
phase-dependent calibration methods developed for null
ellipsometers 8 "1'3 require independent adjustment of
p=rII/rL= (tanf,) eil (8a) both analyzer and polarizer azimuths and cannot be
applied to rotating-analyzer systems. Here, an indirect
s)-ia][tan(P-Ps)+i1YP]..(b
Ecot(Q-A phase-dependent method is outlined, which retains
many of the characteristics of the null-calibration
[l+ia cot(Q-As)][1-iyp tan(P-Ps) b techniques. This method depends on the fact that,
in an ideal system, a light beam reflected from the
Equations (4), (5), and (8) represent the main results surface of an absorbing medium will be linearly
of this section; by use of them, p can be calculated polarized if the polarization vector of the incident
exactly, in terms of the experimental parameters a, A linearly polarized beam is parallel or perpendicular to
for any prism. the plane of incidence, in which case the detected flux
For measurements on dielectrics, a compensator is is 100% modulated (ac-component amplitude is equal
necessary to improve sensitivity in rotating-analyzer to the dc background). If the incident-polarization
systems. 10 If the compensator is also constructed of vector does not satisfy this condition, then the reflected
crystal quartz, with plates of opposite handedness, beam is elliptically polarized and the modulation of
properly aligned,"8 then the polarization state (S.,Sy) in the detected flux is less than 100%. Therefore, PS can
the PCSA (polarizer-compensator-surface-analyzer) be found, in principle, by adjusting the polarizer
system is given by' 8 azimuth to the point of 100% modulation, in which
case P= Ps (or P= Ps+7r/2) and A s= tan-' (/3/a)Ip=ps
Sz\ (r11 0 \(cos(Pe-Ps) -sin(PC-Ps) [all angles modulo ir]. Superficially similar alignment
Sy 0 rL sin(Pc-Ps) cos(Pc-Ps)J procedures have been given for null ellipsometry,"'
although these rely on flux measurements rather than
(lyeleil -ioyc(l-eil), polarization-state detection.
In practice, Ps cannot be determined exactly as
iyc(l-e') e '+'yC / outlined in the foregoing, in systems with optically
(cos(P-Pe) -sin(P-Pc)\ 1 active prisms because optical activity shifts the azimuth
of maximum (now not necessarily 100%) modulation
sin(P-Pc) cos(P-Pc) V yp 9 away from Ps. In addition, the modulation level depends
quadratically upon P in the vicinity of Ps or Ps+7r/2,
where 8 is the relative phase shift of the compensator a functional dependence too weak in many cases to
and Pc is the azimuth angle of the major axis of the attain sufficient sensitivity. The first difficulty can be
compensator in the frame of reference of P. The resolved by calculating the relationship between Ps
determination of Pc is discussed in Sec. II. After a and the observed maximum-modulation azimuth.
short calculation, it is found that p is given by The second difficulty is removed by fitting a quadratic
function of P to the modulation amplitudes that are
[cot(Q-As)-ia][tan(Pc-Ps)+z] measured for a series of values of P in the vicinity of
cot(Q-As)][1-z tan(Pc-Ps)]'
[1-E+ia maximum modulation. The specific function of the
modulation amplitude used here is the residual, R(P),
z= ([1+yc 2ei5 ][cos(-PP)-iyp sin(P-Pc)] defined in terms of the experimentally measured
coefficients, a and ,B,as
-ieyc(1-eij)E[sin(P-Pe) +izyp cos(P7--Pc)]} R(P) 1-a2- 3 2. (11)
X -ei 0 )[cos(P-Pc)
{ ic(1iyp sin(P=Pe)] R(P) approaches zero as the modulation approaches
100%.
- +[ei+yC 2][sin(P-Pc)+iyp cos(P-Pc)]}'-- The dependence of R(P) upon P in the vicinity of
(lOb) maximum modulation can be calculated directly from
June1974 CALIBRATION OF ROTATING-ANALYZER ELLIPSOMETERS
the Jones-matrix product of the actual PSA system, and if A2 is defined as
A2= 2 tan-' (i3/a) | P=P2+T/2, (18b)
then
81, 0 0
A s= A 2 + (YP tan&+±yA cosA)/sinA. (18c)
Xcos(A-As) sin(A-As)
\rii 0) Equations (18) relate the azimuth angles As and Ps
-sin(A -As) cos(A-A s) ° r.L to the parameters obtained at the second local minimum
in R(P). Note that, in both cases, the theoretical
/cos(P-Ps) -sin(P-Ps)\/ 1I minimum value of R (P) differs from zero only by terms
sin(P-Ps) of second order in the optical-activity coefficients
cos(P-PS) )ti(12)
'YA and yp.
Using Eq. (8a) and the fact that I o 8. 2, to first The curve-fitting procedure to determine P1 and P2
order in 'YA and 'yp, we find with reasonable sensitivity can be described in the
following way. Let (Rj,Pj), j= 1, 2, ... , N represent a
Icc 1+ax' cos2(A-As)+13' sin2(A-As), (13a) set of values of {Rj} calculated by Eq. (11) from data
a'= [tan2t cos2 (P-PS)-sin 2 (P-Ps)]/D, (13b) obtained for a set of polarizer azimuths {Pj} in a range
about a given local minimum. Then the coefficients co,
,3'= tanit[cosA sin2 (P-Ps)+2-yp sinA]/D, (13c) cl, and C2of the function
D= tan2t COS2 (P-Ps)+sin2(P-Ps) R(P)= co+CIP+C 2 P2, (19)
-274 tanit sinA sin2(P-Ps). (13d) of the form of Eqs. (15) and (17), which minimizes the
Calculation is related to measurement by Eqs. (4b), mean-square deviation with respect to the pairs (Rj,Pj),
(11), and (13a), which show that R(P) is also equal to are given by20
R(P)= 1 _a'2 -,12. (14) co=[ro(p2p4-p3 2
)+rl (p2p3-plp4)
The unknown azimuth angles As and Ps can no-w be w be+r2(pIp3-p2 2)]/d, (20a)
determined. Let JP-Ps3 <«l and expand R(P ) to CI= [rO(p2p3-plp4)+ri(pOp4-p22)
second order in IP-Ps . By Eqs. (13) and (14),
+r2 (pIP2 -p 0p 3)]/d, (20b)
R(P) IPnP 5-4(P-Ps) cot2q sinA C2=[rO(Plp2-pap3)+rl (PlP2-POP3)
X[(P-Ps) sinA-2(-YA tan,&+'yp cosA)], (15) +r2(pop2-p, 2
)]/d, (20c)
which is quadratically dependent on (P-Ps), wi.th a where
linear term proportional to the optical-activity c )effi-3 2p2
cients 'YAand '(p. Define P1 to be that (experimen tally (20d)
d= PoP2P4+2 p1P2P3-P23 -PoP32-P12 P4,
observable) value of P-Ps for which the moduleLtion N
is maximized, i.e., for which R(P) has a local minin num. Pk=N 1
, pjk, (20e)
Then, from Eq. (15), j=1
N
Ps= Pr-('Y tan,+-yp cosA)/sinA lP-p 3 , (16a) rk=N-1 E Rjpjk (20f)
j=1
which relates Ps to Pi through the optical-activity
coefficients and the properties of the reflecting surface. Thus, e.g.,
To obtain As, define N
Po= N'- 1= 1.
Ai= 2 tan-' Wa/<)IP=P,- j=1
(16b)
Then, by Eqs. (4b), (13), and (16a), Then from Eq. (19), and by use of the definitions of
Pi and P 2
A s= A 1 - (yp cotl6+'A cosA)/sinA. (16c)
PI= -cl/(2c2), (Pi) near Ps; (21a)
For the alternate case with P-Ps+7r/2, define
(P2 +r/2) to be the value of P-Ps+7r/2 7r r
at which P2 =-cl/(2c2)--, (Pj) near Ps+- (21b)
R(P) has its second local minimum. Then 2' 2.
R(P) 4(P-Ps) tan2z sinAEX(P-Ps) sinA From Eqs. (16) and (18), alternate expressions to
+2('YA cot/+'YP cosA)], (17) obtain As and Ps in terms of A1 , Pi, A2 , and P2 are
whence A s= (A 1 +A 2 )/2-yp cot2q//sinA, (22a)
Ps3 P 2 + (-YA cots6+'yp cosA)/sinA IPCPS+,r/2, (18a) PS= (P 1 +P 2 )/2+ A cot2#/sinA. (22b)
7
816 D. E. ASPNES Vol. 64

In principle, 'YA and yp can be determined from the weak dependence if the photometric system has residual
remaining linear combinations of Eqs. (16) and (18), nonlinearities. Consequently, in the first calibration
step, both AP, and AP2 are obtained from Eq. (25),
YA= [(AI-A 2 ) cosA- (P1 -P 2 ) cos2i6]/Di, (23a) withP~Pc andP~Pc+7r/2,respectively.
yp=[(P1 -P2 ) COSA-(A1 -A 2 ) csc2i]/Di, (23b) In the second step, the compensator is reinserted
into the system. The PCA Jones-matrix product, to
where first order in yc, the optical-activity coefficient of the
D 1 = 2(cos2A-csc 2 2ip). (23c) compensator, is
However, a better method for determining 'yp employs S. 1 -i A)( cos(A-Ac) sin(A-Ac)
the PCA (polarizer-compensator-analyzer) system.
In summary, the PSA system can be calibrated by S" 0 0 -sin(A-Ac) cos(A-Ac)I
means of the following sequence of steps. (i) For a
1 -i-yC(l -e-i')
series of polarizer azimuths Pi near P= Ps, determine
the residuals Rj according to Eq. (11), where a and j3 (1 x
-ei) eib
are defined by Eq. (14b). (ii) Calculate Pi, the azimuth
of maximum modulation, by least-squares fitting
according to Eqs. (20) and (21a). (iii) Set P=P 1 and
(cos(P-Pc)
-sin(P-Pc) 18
(26)
determine A 1 from Eq. (16b). (iv) Calculate Ps sin(P-Pc) cos(P-Pc) JP/
and AS from Eqs. (16a) and (16c) using Eq. (2) for where 5 is the phase retardation for the axis perpendic-
YA I = IP I, with the signs determined from PCA ular to Pc. Local minima in the residual R(P), defined
measurements described in the next section. Al- by Eq. (11), occur for P-Pc and P~Pc+7r/2. In the
ternatively, steps (i)-(iv) may be performed with second PCSA calibration step, the polarizer azimuths
P~Ps+7r/2, in which case the appropriate quantities P1 and (P 2 +7r/2), at which these minima occur,
are A 2 and P2 and the relevant equations are (21b) are determined experimentally by the quadratic curve-
and (18). Another alternative is to determine all four fitting procedure that utilizes Eqs. (20) and (21) for the
quantities A 1, P1 , A2, and P 2 , and use Eqs. (22) to PSA system. Next, P is set equal to P1, and
find As and Ps, and Eqs. (23) to obtain 'YA and yp.
2 tan7'(#/Cl) IP-PI - A pi (27a)
B. PCSA (PCA) System is calculated from the measured coefficients a and ,B
By Eqs. (10), calibration of a PCSA. system requires and the previously determined azimuth Api. Similarly,
Pc, As, and Ps to be determined, where Pc is the P is set equal to P 2+7r/2, and
azimuth of the fast axis of the compensator in the A2= 2 tan1 (/3/a)IPP 2+r12-AP2 (27b)
polarizer frame of reference. We assume, for simplicity,
that both the surface and the compensator are tempo- is calculated. Starting with Eq. (26), a calculation
rarily demountable and that the analyzer arm can be parallel to that discussed in the preceding section leads
rotated to allow straight-through operation with the to expressions from which PC, YC, 'YA, and yp can be
reflecting surface removed. Then Pc, As, and Ps can calculated from the measured quantities A 1, P1 , A2,
be determined by the following procedure, which is and P2 ,
based on the principle of maximum modulation that
was discussed for the PSA system. Pc=(A +Pi+A 2+P2 )/4, (28a)
First, both compensator and surface are removed 6(
and the analyzer arm is rotated for straight-through IYc= Cot- rA,-P,+A2-P2]/4, (28b)
operation. The resulting PA system allows the analyzer
azimuths to be measured temporarily in the polarizer
frame of reference, a necessary step in determining Pc, '=[(A 1-A 2 ) cot - (P1 -P 2 ) csc&]/2, (28c)
as follows. To first order in and -yp for the PA csc6]/2. (28d)
2) cot3-(AI-A 2)
-YA
yp=[(Pi-P
system, it is easy to show that
Thus Pc is established in the frame of reference of P.
I cc1+cos2[P-(A-Ap)], (24)
We note that these equations are identical to those
where P and A are each measured in their own frames previously derived for PCA calibration in null ellip-
of reference, and Ap is the relative scale shift. By sometry [see, e.g., *Eqs. (21) and (23) of Ref. 8],
Eqs. (4b) and (24), leading to an alternative interpretation of the residual
approach as a null measurement on the elliptical
polarization of the beam transmitted through the PCZ
A p=4 tan'l(f3/a) IP+-u(-a) sgn(B) I p-P. (25) half of the PCA system.
2
In the third calibration step for the PCSA system,
Ap is theoretically independent of P but may show a the reflecting surface is replaced and the correct angle of
June1974 CALIBRATION OF ROTATING-ANALYZER ELLIPSOMETERS 817

incidence established. By temporary removal of the

compensator, A s and Ps are determined by the method
of Sec. III A. Finally, the compensator is replaced and
the system is ready for operation.
In contrast to null ellipsometry, it is not possible to
determine both p for the reflecting surface and a for
the compensator because rotating-analyzer systems N
are physically asymmetric and have no analog to the
two non-equivalent nulls of the other systems. Thus 6
must be calculated for arbitrary wavelengths according
to the phase-dispersion equation 0:

8= 2i (n8 -nf)d/X (29a)

I-rXQ/(2X), (29b)

where XQis the wavelength at which 8=7r/2. Equation

(29b) is an approximation in which the small effects of
dispersion of n8 and nf, the indices of refraction of the
slow and fast axes, respectively, are neglected. If XQ
falls within the tuning range of the light source of the -3 -2 -I 0 1 2 3
ellipsometer, then this quantity can be determined by P (DEGREES)
varying P and X to obtain a=/3= 0, i.e., by elimination
of the ac component in the transmitted flux. It is FIG. 1. Residuals calculated from data, and best-fit quadratic
function for Ni film at X=300 nm. Each point represents an
straightforward to show from Eqs. (26) and (3a) that average over 20 cycles (10 analyzer rotations) at the given
reduction to a=3=0 leads to the conditions, valid to polarizer azimuth. The minimum azimuth PI and the calculated
first order in zyp and yc, azimuth Ps of the plane of incidence are indicated explicitly.

XQ=2XlX2 /(XI+X2 ), (30a) reduction as discussed in this paper can be performed

Pc= (P,+P2 )/2, (30b) without taking optical activity into account. The
question naturally arises, how important is this effect?
Iyc=w/4 - (P2 -P,)/2, (30c) We have found that properly including optical effects
in data reduction is less significant than properly
,yp= rXQ(X1-X 2 )/ (8X1X2), (30d) preparing sample surfaces. Optical-activity corrections
where Pt-Pc-7r/4 and XAare the values required to are typically of the order of one percent of the absolute
obtain the first null, and P2-Pc+ir/4 and X2 are the values of the real and imaginary parts of the complex
values required to obtain the second. Clearly, this dielectric function, except for component settings near
procedure represents an alternate, simpler method of normal modes of the system. The order of magnitude
determining Pc, -yc, and yp if the source wavelength of optical-activity corrections for the PSA system can
is variable. The parameter YA does not appear here, be estimated by inspection of Eq. (8b), which shows
because to obtain a=,B=O, the light incident on the that yp is significant only for tan(P-Ps) very large
rotating analyzer must be completely circularly or very small, and similarly for the PCSA system only
polarized and hence has no preferred azimuth. if tan(P-Pc) is very large or very small. In polariza-
tion-state detection, Eqs. (4a) and (5) show that the
III. DISCUSSION effect of -YAvanishes for a=,6=0, the circularly polar-
ized limit, and is significant, by Eqs. (10), only if
Sections I and II outline data-reduction and calibra- cot(Q-As) is large. Because the latter condition cor-
tion procedures, respectively, for rotating-analyzer responds to operation with P near a normal mode,
ellipsometers, generalized to include the effects of which should be avoided for reasons of precision as
component optical activity. Optical-activity correc- well,10 the effects of 'YA on data reduction should not
tions are obtained exactly for data reduction and to be important in normal operation. This suggests that
first order for calibration. If optical activity is ignored, the conditions on quality are less stringent for the
the appropriate equations are obtained simply by analyzer prism than for the polarizer prism, which
setting YA==Yc= YP= 0- has been already noted for null ellipsometric systems.8
Accurate calibration of analyzer, compensator, and However, Eqs. (16) and (18) show that 'YA and yp
polarizer azimuth scales with respect to the plane of both influence calibration. Changes of the order of one
incidence is clearly necessary and its importance, percent correspond to azimuth errors of the order of
discussed elsewhere for null ellipsometry,'9 will not be 0.50, which may be unacceptable in many situations.
treated here. However, both calibration and data Moreover, Eqs. (16) and (18) show that the magnitude
818 D. E. ASPNES, Vol. 64

TABLE I. Evaluation of Ps and As, the azimuths of plane of incidence in polarizer and analyzer frames of reference, respectively.
The quantities P1 and Al are data, and AP, and AA, represent first-order correction terms calculated from Eqs. (16a) and (16c),
respectively. The columns e (raw) and e (corr) refer to values of the dielectric function calculated without and with optical-activity
corrections, respectively.

X,nm Ps=P,-AP, (deg.) As=A,-AA, (deg.) e (raw) e (corr)

300 0.35=0.25+0.10 15.12= 14.66+0.46 -0.51+i4.93 -0.47+i4.85

400 0.35=0.28+0.07 15.14=14.85+0.29 - 1.85+i5.53 - 1.75+i5.49
500 0.35=0.31+0.04 15.15= 14.95+0.20 -3.76+i7.26 -3.61+i7.27
600 0.35=0.33+0.02 15.14= 14.99+0.15 -5.81+i9.33 -5.55+i9.40

of the correction term is proportional to 1/sinA, optical activity to be nonexistent, we determined the
indicating that the calibration correction will be very dielectric function over the available spectrum. By use
important for low-loss materials where IsinA I 0, and of these values, the correction terms in Eqs. (16a) and
that the calibration procedure will fail completely for (16c) were evaluated and used to obtain Ps and As
dielectrics, for which IsinA I = 0. The latter statement is as a function of X. The results for specific wavelengths
intuitively obvious, because linearly polarized light are summarized in Table I. Finally, the' dielectric
reflected from a dielectric surface remains linearly function was recalculated with the full optical-activity
polarized, and the modulation parabola reduces to a corrections included in both calibration and data
horizontal line, corresponding to 100% modulation, reduction; the results are also given in Table I.
for any value of P. This condition is somewhat analogous Table I shows that the magnitude of the effect of
to the tan6/2 singularity that arises in calibration of optical activity agrees well with the estimates given.
PCA systems in null ellipsometry.5 However, in contrast For instance, the azimuth correction reaches a value of
to null ellipsometry, in which A s and Ps can be nearly half a degree in As at 300 nm. When corrected
determined directly, with no first-order influence of for optical activity, Ps and As are independent of X,
system defects, by use of a dielectric surfaces,8 2" there as expected. The net effect in E is also of the order of
is no possible combination of h1and A, and hence no 1%. Because this change represents contributions from
surface, for which the correction terms vanish in a both data-reduction and calibration errors that result
rotating-analyzer system. Thus, in accurate work, from neglect of optical activity, this suggests that in
these correction terms must always be evaluated many situations full correction for optical activity is
explicitly. not necessary. Measurement of spectra intended for
As a typical example, we consider calibration and numerical differentiation to resolve fine structure would
data reduction of measurements in the PSA configura- be an example of a problem in which optical-activity
tion of the dielectric function of a real surface, an corrections would not generally be needed.
optically thick Ni film evaporated from an Al203-coated
tungsten boat onto a glass slide at 10-1 torr. We show ACKNOWLEDGMENT
in Fig. 1 a least-squares fit of Eq. (19) to the residual,
R(P), calculated from 250 data points taken at X= 300 It is a pleasure to acknowledge the technical assist-
nm at equal intervals from P= -3.00' to P= +3.000. ance of -A. A. Studna in sample preparation and in
The polarizer azimuth was set by means of a stepper measurement of experimental data.
motor operating under computer control, and the
least-squares fit was performed by a separate subroutine. REFERENCES
The relatively large value of R(P) at the minimum is
'E. 0. Ammann and G. A. Massey, J. Opt. Soc. Am. 58, 1427
caused entirely by the attenuation of the ac component (1968).
with respect to the dc component by the low-pass 'B. D. Cahan and R. F. Spanier, Surf. Sci. 16, 166 (1969).
filter used to limit the bandwidth and consequent noise 'R. Greef, Rev. Sci. Instr. 41, 532 (1970).
4
amplitude at the input of the analog-to-digital con- D. E. Aspnes, Optics Commun. 8, 222 (1973).
6P. S. Hauge and F. H. Dill, IBM J. Res. Dev. 17, 472 (1973).
verter. In normal operation, these data are also used
'R. C. O'Handley, J. Opt. Soc. Am. 63, 523 (1973).
to evaluate the attenuation of the filter, because 7
J. F. Nye, Physical Properties of Crystals (Oxford U. P.,
R(Pl)= O to second order in the correction terms. London, 1957), pp. 260ff.
Figure 1 shows that the functional dependence of the 8D. E. Aspnes, J. Opt. Soc. Am. 61, 1077 (1971).
residual near an extremum is accurately parabolic. 9R. M. A. Azzam and N. M. Bashara, J. Opt. Soc. Am. 61,
1236 (1971); J. Opt. Soc. Am. 62, 700 (1972).
The values of P, and Al determined at X=300, 400, 'OD. E. Aspnes, J. Opt. Soc. Am. 64, 639 (1974).
500, and 600 nm are summarized in Table I. The large "S. N. Jasperson and S. E. Schnatterly, Rev. Sci. Instr. 40,
values of A 1 are due to phase shifts in the low-pass 761 (1969); S. N. Jasperson, D. K. Burge, and R. C.
O'Handley, Surf. Sci. 37, 548 (1973).
filter, which are also corrected prior to data reduction. 1R. M. A. Azzam and N. M. Bashara, J. Opt. Soc. Am. 61,
Using as an approximation the values of P1 and Al 1118 (1971).
measured at 300 nm as Ps and As, i.e., assuming "M. Ghezzo, Brit. J. Appl. Phys. 2j 1483 (1969).
Junel974 CALIBRATION OF ROTATI-NG-ANALYZER ELLIPSOMETERS
' 4R. M. A. Azzam and N. H. Bashara, J. Opt. Soc. Am. 61, Optics (Prentice-Hall, Englewood Cliffs, N. J., 1972), pp.
600 (1971); J. Opt. Soc. Am. 61, 773 (1971); J. Opt. Soc. Am. 288ff.
5
61, 1380 (1971). '187C. V.
Kent and J. Lawson, J. Opt. Soc. Am. 27, 117 (1937).
G. Szivessy and C. Munster, Ann. Phys. 20, 703 (1934); D. E.
Aspnes, Appl. Opt. 10, 2545 (1971).
G. Bruhat and P. Grivet, J. Phys. Radium 6, 12 (1935); "209M. J.
Dignam and M. Moskovits, Appl. Opt. 9, 1868 (1970).
C. Munster and G. Szivessy, Phys. Z. 36, 101 (1935). N. Draper and H. Smith, Applied Regression Analysis (Wiley,
6
" J. R. Meyer-Arendt, Introduction to Classical and Modern New York, 1966), p. 129.

Free Registration for Emeritus Members

Emeritus members receive free registration at general (not topical) meetings. They should tell the regis-
tration person that they are emeritus members. Any regular member or fellow who has retired from active
work and whose years of membership in the Society total at least ten (10) is eligible for emeritus status.
All such are invited to request emeritus membership, whether or not they are currently members. All
members are urged to call this to the attention of their retired friends. Emeritus members do not currently
pay. any dues. They do not receive the Journal of the Optical Society of America or Applied Optics but do
receive notices of meetings and ballots. They may vote and hold office. Fellows retain fellow status. See revi-
sion of the Bylaws on p. 258 of the February 1974 issue. Send requests for transfer to Dr. Jarus Quinn,
Executive Director, Optical Society of America, 2100 Pennsylvania Ave., N.W., Washington, D. C. 20037.