Professional Documents
Culture Documents
A comparative analysis is made of the errors in deriving the optical parameters ~n, refractive index; k,
absorption coefficient; d, film thickness! of thin films from spectrophotometric measurements at normal
light incidence. The errors in determining n, k, and d by the ~TRf Rb!, ~TRf Rm!, ~TRb Rm!, ~TRf !, ~TRm!,
and T~k 5 0! methods are compared. It is shown that they are applicable to optical constants of thin
films in the n . 1.5, k , 4.5, and dyl 5 ~0.02–0.3! range, and their combinations make possible the
determination of n and k to an accuracy of better than 64%. To derive the optical constants in a wide
spectral range with high accuracy and isolate the correct physical solutions reliably, one should apply all
methods, using the relevant solutions with the lowest errors, as shown in this research, when determin-
ing the optical constants of As2S3 and Sb2Se3 films. © 1998 Optical Society of America
OCIS codes: 120.0120, 120.4530, 120.6200, 310.0310, 310.6860.
UU U U U U U U
nm km
]n ]n ]n ]n
Dn 5 DT 1 DR 1 DRm 1 Dns
]T ]R ]Rm ]ns where the coefficients Eij are given in the error
U U U U
matrix @Eq. ~6!#, similar to Hansen’s matrix7:
]n ]n
5 6
S D SD F G
1 Dnm 1 Dkm,
]nm ]km T ]n T ]k T ]~dyl!
Dk 5 UU
]k
DT 1U U
]k
DR 1
]k
U U
DRm 1
]k
U U
Dns
n ]T k ]T ~dyl! ]T
1
]T
U ]k
]R
U U
Dnm 1
]k
]Rm
U
Dkm,
]ns
S
R ]n
n ]R
D SD R ]k
k ]R
R
~dyl!F]~dyl!
]R G
Dd 5
]d
UU
]nm
DT 1
]d
U U
]km
DR 1
]d
U U
DRm 1
]d
Dns U U S
Rm ]n
n ]Rm
D S D Rm ]k
k ]Rm F
Rm ]~dyl!
~dyl! ]Rm G
S D S D F G
]T ]R ]Rm ]ns E5 . (6)
ns ]n ns ]k ns ]~dyl!
1U ]d
U U
Dnm 1
]d
U
Dkm. (1) n ]ns k ]ns ~dyl! ]ns
S D S D F G
]nm ]km
nm ]n nm ]k nm ]~dyl!
Since the partial derivatives cannot be evaluated n ]nm k ]nm ~dyl! ]nm
S D S D F G
numerically, they must be expressed by derivatives
that can be calculated. For this purpose, simple lin- km ]n km ]k km ]~dyl!
ear expressions are used, which for the TRRm meth- n ]km k ]km ~dyl! ]km
ods have the form
DT 5 SD SD
]T
]n
Dn 1
]T
]k
Dk 1
]T
]d
SD S D
Dd 1
]T
]ns
Dns,
In a similar way, Dkyk and Ddyd are calculated by
the corresponding columns of error matrix ~6!. We
mentioned above that Dl is negligibly small, so the
DR 5 SD SD
]R
]n
Dn 1
]R
]k
Dk 1
]R
]d
SD S D
Dd 1
]R
]ns
Dns,
relative error Ddyd is equal to D~dyl!y~dyl!.
Similarly, for the T~k 5 0! method,
DRm 5 S D S D
]Rm
]n
Dn 1
]Rm
]k
Dk 1 S D
]Rm
]d
Dd DT 5 SD SD S D
]T
]n
Dn 1
]T
]d
Dd 1
]T
]ns
Dns. (7)
1S D S
]Rm
]nm
Dnm 1
]Rm
]km
D
Dkm. (2) Hence Dn is
YS D S DYS D
Using a more compact matrix expression, we write
Eqs. ~2! as ]T ]T ]T
E1 5 1 E2 5
S D SD S D
, ,
]n ]d ]n
DT Dn Dns
DR 5 M1 p Dk 1 M2 p Dnm .
DRm Dd Dkm
(3) E3 5 S DYS D
]T
]ns
]T
]n
. (8)
U U U U U U
SD S D S D
Dn T DT d Dd ns Dns
Dn DT Dns 5 E1 1 E2 1 E3 . (9)
n n T n d n ns
Dk 5 u~M1!21u p DR 1 u~M1!21 p ~M2!u p Dnm .
Dd DRm Dkm The values of T and R of the thin film are calculated
(4) from T and R, the transmission and the reflection of
Rf 5 Rf 2
T Rs 2
1 1 Rb Rs 2 2Rs
, Rs 5 S D
ns 2 1
ns 1 1
,
2
U U U U U U
the most suitable method in a given spectral range,
]T ]T ]T depending on the accuracy of the measured photo-
DT 5 DT 1 DRb 1 DRs,
]T ]Rb ]Rs metric quantities. The practical approach is to eval-
U U U U U U U U
uate the error in each set of the obtained n, k, and d
]Rf ]Rf ]Rf ]Rf of the investigated thin film and to compare directly
DRf 5 DT 1 DRf 1 DRb 1 DRs,
]T ]Rf ]Rb ]Rs the results from the various methods. This is illus-
U U U U
trated by the following examples.
]Rb ]Rb
DRb 5 DRb 1 DRs. (11)
]Rb ]Rs A. Maximum Relative Errors in TRR Methods
In TRR methods three different combinations be-
3. Maximum Errors in Photometric Methods for tween the photometric quantities are possible—
Normal Light Incidence ~TRf Rb!, ~TRf Rm!, ~TRb Rm!. The first combination
In the following considerations we present a sche- is the least applicable, since it is unusable for
matic picture of the errors in the optical constants the films in the spectral range around and beyond
and the thin-film thickness as dependent on the val- their absorption edge,1 where Rf 5 Rb. Besides, the
ues of n, k, and dyl. The experimental inaccuracies analysis performed showed that the errors in this
assumed in the spectrophotometric quantities are combination are substantially higher than those in
DT 5 60.001 and DR 5 60.005, and the errors in the two other combinations. Because in the two
the refractive indices of the two substrates are combinations with Rm the errors differ negligibly, in
Dns 5 60.015, Dnm 5 60.03, and Dkm 5 60.02. the following we present the results on the ~TRf Rm!
Corresponding inaccuracies DT and DR are calcu- method only.
lated by Eqs. ~11!. The values of T, Rf, and Rm are 1. Assessment of the Error Matrix Terms
calculated by Fresnel equations ~Ref. 8, pp. 6 – 8!, and
the maximum errors are calculated by the equations According to Hansen,7 the method for determining
in Section 2. Because nonabsorbing substrate BK7 the optical constants is excellent if the error matrix
glass is assumed, ns is calculated by the dispersion terms @Eq. ~6!# are near unity or less. A value of
formula of Sellmeier ~Ref. 8, p. 134!: unity for each element of the matrix @Eq. ~6!# means
that optical parameters n, k, and d are determined to
4. Application of the Error Analysis and Rm of the samples of As2S3 and Sb2Se3 thin films,
Next we illustrate the application of the error anal- deposited on BK7 and Si substrates, are given in
ysis to the method selection and to the most accurate steps of 50 nm in Table 1.
determination of the thickness and the optical con-
stants of thin films.
B. Determination of the Thickness and the Optical
A. Sample Preparation and Spectrophotometric Constants
Measurements T, Rf, and Rb of the thin films were evaluated from
The thin films of As2S3 and Sb2Se3 were produced by the photometric quantities by Eq. ~10!; n, k, and d
vacuum thermal evaporation of the substances11,12 on were determined by the ~TRf Rm!, ~TRf !, and ~TRm!
planetary moving substrates, which ensures excel- methods. Figure 5 displays all solutions for thick-
lent uniformity of the thin film. Pieces of Si wafers, ness d derived from the TRf Rm method and the ac-
whose surface was cleaned9 before film evaporation, curacy for the correct solutions. As seen, the only
were fixed to the preliminary cleaned glass sub- solutions with one and the same film thickness in the
strates ~BK7! with high surface flatness. whole spectral range are 58 6 1.6 nm for As2S3 and
The absolute reflectance and the transmittance of 78 6 0.9 nm for Sb2Se3. The indicated values and
the samples were measured with a spectrophotome- the maximum errors are taken from the spectral
ter Cary 5E at normal light incidence in the spectral range, corresponding to the most accurate determi-
range of 400 –1000 nm. The measured values of the nation of d, marked with arrows in Figs. 5~b! and
transmittance T and the absolute reflectance Rf, Rb, 5~d!.
Obviously, the correct physical solutions for n and
k correspond to these thicknesses, and they are given
in Figs. 6 and 7 with the corresponding absolute er-
rors. In the same figures we also display the values
of n and k evaluated by the ~TRf ! and ~TRm! methods,
using the iteration Newton–Raphson’s method13 and
the corresponding errors calculated at the aforemen-
tioned accuracy of d. Only values corresponding to
the true solutions from the ~TRf Rm! methods were
accepted.
Figure 8 presents the solutions obtained for n by
the T~k 5 0! method for the spectral range in which
k 5 0.
Figures 6 – 8 reveal that the methods used provide
Fig. 4. T~k 5 0! method: solid curves, contours of the relative solutions with different accuracy in the spectral
error in n with the indicated values, plotted in the n–dyl plane; range of 400 –1000 nm. Obviously, for n of the As2S3
dashed curves, contours plotted for ~]Ty]n! 5 0, showing the areas film the most accurate solutions are from 400 to
of uncertainty of the method in extracting n. 550 nm by the ~TRf Rm! method and from 550 to 1000
Fig. 5. Solutions for film thickness d by the ~TRf Rm! method and absolute error in d for the correct values @d~l! ' const#: ~a!, ~b!, As2S3
films, d 5 58 6 1.6 nm; and ~c!, ~d!, Sb2Se3 films, d 5 78 6 0.9 nm.
Fig. 6. Dispersion curves of n and k and their absolute errors for As2S3 ~d 5 58 6 1.6 nm!, obtained by the three designated methods.