Surface Science 49 (1975) 441-458 0 North-Holland Publishing Company
DETERMINATION
OF THE COMPLEX REFRACTIVE
INDEX PROFILES
IN Pi1 ION IMPLANTED SILICON BY ELLIPSOMETRY J.R. ADAMS* and N.M. BASHARA** Electrical Materials Laboratory, College of Engineering, Lincoln, Nebraska 68508, U.S.A.
Received
29 November
1974; revised manuscript
received
University of Nebraska,
15 January
1975
Ellipsometry was used to determine the complex refractive index profiles in silicon implanted with PJ1 ions with energies of 35, 52,5 and 70 keV. The profiles were determined both by anodization-stripping of the implanted layer and by numerical fitting of multiple-angle-of-incidence ellipsometer data taken on the as-implanted surface, assuming that the implantation would exhibit a Gaussian distribution. Good correlation was obtained between the two types of profiles, indicating that the non-destructive measurements on the as-implanted surface may be useful in process control. Good agreement with published results was also obtained on the increase in depth with energy of both the damage and the implanted species.
1. Introduction
The increasing importance of ion implantation as a method for inducing changes in the electrical and optical properties of semiconductors, glasses and ceramics has motivated extensive study into many aspects of the implantation process [ 11. Because the implantation process introduces a large number of defects into the implanted region, almost as much effort has been devoted to the determination of the nature and the effects of the damage as to the desired effects of the implanted ions [?I. Depending on the ion species, total dose, and implantation energy, the density of defects can be sufficient to drive the surface layer of a crystalline specimen amorphous thereby negating or seriously affecting the properties desired unless the specimen is annealed to a high temperature [2b]. In certain case, such as radiation tolerant transistors, it may be advantageous to retain some of the effects of the damage, provided these effects are defined. Regardless of the end result, it is necessary to characterize the damage region. To accomplish this a number of techniques have been used including helium ion back* Currently a member of the Technical Staff, Sandia Laboratories, Albuquerque, This work was supported by the United States Atomic Energy Commision. ** Work supported by the National Science Foundation.
New Mexico.
442
J.H. 4dams, N.M. Busharu/Conzplex re.fructive index profiles
scattering [ 1,3], electron diffraction [4-6,101 , electron microscopy [7,9], electron spin resonance [lo- 121, electron paramagnetic resonance [ 121 , electron channeling patterns [ 141, visual observation [ 1,101 and optical reflection and transmission spectroscopy [ 10, 15-241. Ellipsometry has also been shown to be a valuable tool for the analysis of refractive index changes in Si [25] and fused siluca [26:27]. To investigate the effects of implantation on the optical properties, visual observation and retlection and transmission spectroscopy leave much to be desired because of the inhomogeneous nature of the implanted region. Consequently, an accurate quantitative characterization cannot be obtained solely from the reflection or transmission spectrum. In contrast multiple-angle-of-incidence ellipsometry [28,29], at a judiciously chosen wavelength [29], coupled with an appropriate optical model and removal of succesive layers of implanted material is well suited to a quantitative analysis of the effects of the ion implantation on the optical properties and a determination of the spatial extent and profile of the implantation. In this paper we report the results of ellipsometric measurements of the complex refractive index profiles produced in high resistivity 5 R cm (111) silicon by implantation of PiI ions at energies of 35, 52.5, and 70 keV, and to doses of 2.0 X lOI and 1.O X 1016 ions/cm2. A comparison o f the results of multi-angle-of-incidence ellipsometer- measurements con the as-implanted surface with the results of ellipsometric measurements and anodization-stripping of the implanted region is given.
2. Experimental
procedure
2. I. Inn irnplan tatio yl Samples for these experiments were cut from (Ill) orientation* high resistivity, z 5 a cm. boron-doped silicon wafers using a YAG laser scriber to 0.250 in. X 0.500 in (6.3 mm X 12.6 mm) dimensions. The samples were then etched in a 1 : 10 solution of 49% HF acid in deionized water, cleaned ultrasonically in trichloroethylene, acetone, and methyl alcohol and jet dried using high purity dry nitrogen. Implantation of the phosphorous ions was carried out at room temperature, -23 “C, in a Lintott ** high-tluence ion accelerator using phosphorous trichloride for the source of P$t ions. The beam (approx. 1 cm dia.) was mass analyzed and slowly scanned over the samples. The background pressure in the implantation chamber was approximately 2 X 1O-6 torr. Samples were implanted at ion energies of 35, 52.5, and 70 keV, to fluences of 2.0 X 1014 ions/cm2 and 1.O X 1016 io%/cm2 and the total dose was controlled to + 1% at a dose rate of approximately 3.5 X 1013 ions/cm2 sec. The samples were misoriented by 7” with respect to the incident beam to minimize channeling of the implanted ions. * This orientations was found to be accurate to within 1” as determined from optical orientation measurements on unimplanted specimens cut from the same wafers according to ASTM Method F26--66. ** Manufactured by Lintott Engineering Co. Ltd., Horsham, England.
J.R. Adami
N.M. GasharalComplex
refractive index profiles
443
2. I. .4nodic oxidation and stripping In order to measure the refractive index profiles induced by the ion implantation, layers of implanted material were removed by anodic oxidation and stripping [30,31]. Samples were mounted in a Teflon fixture which allowed the sample to be anodized, etched, and ellipsometric measurements to be taken without removing the sample from the holder. Electrical contact to the specimen was through a polished stainless steel slug against which the specimen was pressed by the Teflon fixture. Samples were anodized at a constant-current density of 5 mA/cm2 in an electrolyte solution consisting of 1.O g KNO,, 25.0 ml deionized H,O, and 225.0 ml ethylene glycol. The electrolyte was stirred constantly during the anodization and the sample was illuminated with a 60 W incandescent bulb 10 cm from the sample. The cathode for the anodization cell was a 1.O cm X 2.0 cm platinum foil. The anodic oxide was removed in a 1 : 10 solution of 49% HF acid in deionized water. The sample, in its holder, was then rinsed in deionized water, acetone, and methyl alcohol, and dried by a jet of high-purity dry nitrogen. This cleaning procedure was used throughout these experiments. Following each anodization, the thickness and refractive index of the anodic oxide was measured by ellipsometry. In order to obtain an accurate value for the oxide film thickness and refractive index at each step, it was necessary to account for the underlying implanted layer in the calculations because the refractive index of the implanted layer differed from that of the Si substrate [32]. The implanted layer was taken into account by using the effective optical constants for the surface after removal of the anodic axode. These constants were used for the substrate values when computing the oxide thickness and refractive index. To simplify the model, the oxide layer was assumed to be uniform and homogeneous. The average ratio of anodic oxide to phosphorus-implanted Si removed was determined after all of the implanted layer was removed. This ratio was then used to estimate the amount of phosphorus-implanted Si removed by each stripping step. The exact spatial point of onset of the unimplanted substrate could only be determined within the resolution capability of each stripping step. For exact determination of the onset of the unimplanted substrate, an infinite (large) number of stripping steps would be required; a patently impractical procedure. The total amount of phosphorus-implanted Si removed was determined by measuring the depth of the stripped region using multiple-beam interferometry [33]. Knowing the total thickness of the phosphorus-implanted Si removed and the total thickness of anodic oxide removed (thicknesses of oxide at each step multiplied by the number of steps) the ratio of the total amount of phosphorus-implanted Si to total amount of anodic oxide (tSi/t,,) is then known. Furthermore, since the total number of layers stripped is known, the amount of phosphorus-implanted Si removed by each stripping is then determined. The results summarized in table 1 are in good agreement with values published by others [3,31].
___ 100 100 50
___,_ A-l A-2 A-3
___~__._ 8 24 20
Number of anod~zati~~ns
1.406 .i 0.014 1.411 IO.014 1.415 2 Ml27
_I~__._^I___
Average oxide refractive index (computedfa
39.5 +. 1.9 41.5 t 2.3 20.1 * 1.4
Average oxide thi~kness/la~er (computed) (nrn) ___l__ .~ .___...
.~___
15.9 + 1.5 15.8 + 0.6 8.1 + 1.0
0.403 0.380 0.403
thickness Ratiob of Si removed per ‘Siltox anodization Inn0 -____._ ____^_~._.I-_-.. ..__ -..-_-AVenge
,._,., “.--_-.__~^^_^__~-~.~
a In the computer program used to evaluate the refractive index and thickness of the Si0, film, the extinction coefficient could also be computed. The extinction coefficient for these films was found to be zero to at least 4 decimal places and is therefore reported here ta be equal to zero. b Total thickness of talented Si itSi) to total anodic oxide thickness (fox).
,_______
Anodization voltage W)
Sample nunlber
Table 1 A summary of the anodic oxidation and stripping experiments described in this paper _______.-----___*-_.~_._---_--^. --.-~.
% P
44s
2.3. b‘llipsometric measurements The ellipsometcr /34] used for this work was a manual Gaertner L-l 19X conf!gured in the polarizer, compensator. sample, analyzer (PCSA) arrangement with fixed-compensator nulling [32] and a He Ne 632.8 nm laser light source. The compensator was a quartz Babinet Soleil type and the polarizer and analyzer were calcite Glan air-spaced prisms. The ellipsometcr was aligned using the techniques described by Zeidler et al. [35]. Four alignment marks. each more than 2.0 m from the sample, were established during the initial alignment. and these points were used to check the slignment of the sample each time the sample was reinserted following an anodijr.ation and/or stripping step. A clip-in holder f’or the anodization sample holder was designed for the cllipsometer sample table and after the initial alignment it was seldom necessary to adjust the sample table. It was found that the sample could be repositioned to within +O.OlS” as determined from repeated measurements on the same sample surface which was removed from and replaced in the ellipsometer sample fixture. The angle-of-incidence scale was checked periodically by setting the ellipsometer to the straight-through position several times. The absolute error in the angle of incidence, which is dependent on the beam deviation and the azimuth of the polarizer prism [36], was found to be less than approximately 0.06” for our il~s(rument. The relative error in the setting of the angle-(of-incidence sc& was less than 0.01”. The ellipsometer divided circles were calibrated while taking two-zone measurcmenrs on the initial samples using techniques developed in this Laboratory 1371. The uncertainty in the calibration constants was found to be less than approximately 0.06”. (See the Discussion section.) Once the divided circle calibration constants were established, rlleasurert?eI~ts were t:*Set: in only one ellips~)r~letric zone and the ellipsometric parameters Ij and A were computed using the equations given by Azlam and Hashara [38]. 2.4. Data reduction The refractive index profiles were computed from ellipsorllcter rlleasurcillents taken in air at four angles of incidence (57.5”, 60.0’. 62.5”. 65.0’) on the asimplanted surface after etching in 1 : IO t1F and rinsing and drying as described above, and on the sample surface immediately following each anodization-stripping step.
For reduction of the data on the as-implanted surface a Fortran IV computer program was developed based on the nonlinear least-squares numerical inversion of the exact Drude equations for a multiple-filmed surface as described by Schueler [28]. In the program, the ion-implanted surface could be modeled as being composed of one or two independent refractive index profiles. Their parameters were adjusted
SO that the values of Q and A fit the obscrvcd 4 and A taken refractive
at multiple
index
profile
values of the ellipsometer
angles of incidclicc. from
an expression
The program
involving
up to eight
eters. l‘hc profile is nlodclcd as a scrics of ten assumed whi& values of + and A may be computed. For the work
rcportcci
tu the refractive index
change
index caused
refractive
index
deviations
were
squares
by the implanted
adjusted
to oht3iu
of the observed
species.
coefficient values
independent
were used
to that
described
provides a numerical used in dctcrmining
by Schroeder
one applicable to the refractive
The magnitudes
01‘ the peaks in the depth
their
spatial
A which
values
were
of Q and
and Dieselman
[26]
and standard
a nonlinear
refractive
index
Icast-
A at all angles
of in-
were required This approach
. except
that
tit to the experimcrltal the
from
(k).
cidence. Mcasurcmcnts al ;I minimum of four angles of incidence adjust all eight paramctcrs of the two independent distributions. similar
;I
param-
films
and one applicable
of 3 and
(tncasul-cd)
determined
homogeneous
distributions
due to the damage
end extinction
(II)
estimate
two Gausian
here,
change
parameters
explicitly
data and a slightly different profile which is more applicable
to is
the program model was to the
actual physical situation *. It should be pointed out that this method is restricted to samples whwe cxtinstion cocfficicnt (k). even in the implrlntcd layer. is sufficicntly small at the measuring wavelength to allow the lifiit wave to pcnctr:ite the entire inlloriiogcnco\Is region. Also. it is not possible to tinaiiibiquouslS,)llsl}/ determine a coniplex rct‘ractivc index profile 1‘1.oni nmsurcniciits on the surface only, unless a rela. tivcly ;iccuI-ate cstiliiatc‘ of the pt-ofilu (typically within 20%) is known beforch;lr,d.
analysis
Although
could
was used. We knew section
then
of the data obtained
have been done by the method 1. ‘fhc
the avc~agc thickness
avtmgc
be computed
refractive
the optical
from
multiple-aligl~-of
” Schrwdrr
and Diewlman
[ 261
tivc index prot’ilc .ind .illhough
a simpler,
of each composite
and extinction ric measurcmcnts.
at the surt‘aze until the .:--implanted of the rct’ractivc index and extinction calculated
the anodi/.ation-stripping above.
constants
layer
experiments
more
direct
method
as described
in
coefficient of the layer could WC wcrc then able to determine
index
I‘rcmi cllipsomot
the profile in two ways. In the first WC clctcrniincd
from
described
of each successive
layer.
starting
substrate was reached using the averegc cocfficicnt of the composite underlying
incidence
measurements.
used a sum of ten (;auswn
rhcy .~llowed the ret‘rclrtivr
Secondly.
distributions
values layers
we followed
10 compute the refrrlc-
indict\ to he complex.
the mq2nitude
and 111~\jn of the c‘l~~ngcin the c<)mplcx rct’r:lctivc index was fixed lor A yivcn Ckusrrlin hution. yxcics
In our m~dcl. one or tlw IUO (Lussian distribution.
and the other to thr damage distribution.
the chanpec in the refractive distrihutmnz. within
distributions
index of) and extinction
The magnirudes and the signs of
coct’ficicnts tk) bare independent
fhc Jcplh and \tandard deviation of rhc perturbations
LL (;.~ua\~.~r:di\trihution.
dccrcawlg ivhilc maintainins
distri-
WI\ assigned to the implanted
on II and k wrc
for both
the wmc
Our procedure allo\vs )I to increaw at the sxne rime 3s !i is ti (;aussidn
prolile
of the s~n,c depth
and standard deviation.
J.R. Adams, N.M. BasharalComplex
refractive index profiles
441
procedure of Bayley and Townsend [27] in their studies on silica glass by working backwards starting with the unimplanted substrate. The Fortran IV computer program used for this substrate-to-surface analysis involved only a minor modification of the method described by Schueler [28]. The results of the two procedures were nearly identical. The results of the second procedure were used in plotting the data discussed in section 3.
3. Results and discussion 3.1.
Complex refractive index anodized-stripped
profiles
Dose 2.0 X 10’4ions/cm2 Figs. l-4 show the results of the experimental measurement of the refractive index (n) and extinction coefficient (k) profiles in P,,’ ion-implanted Si for incident energies of 35, 52.5 and 70 keV, respectively. The figures clearly show that the peak of the refractive index and extinction coefficient profiles moves deeper into the specimen with increasing ion energy. Within the depth resolution (see table 1 for the values of materials stripped), the initial peaks of both the n and k profiles for both the 35 and 52.5 keV specimens are in good agreement with the position of the damage peaks predicted from the theory of Sigmund and Sanders [39]. The general shape of the profiles is also in good agreement with the damage profiles predicted by Brice 1401. For reference, the location of the peaks of the implanted species distributions as predicted by the LSS [41] theory (Rpr) and the peaks of the damage distributions as predicted by Sigmund and Sanders [39] (RpD) are shown. To illustrate the correlation between the structure in the profiles and the theoretical distributions, the depth of the secondary peak in the refractive index profile (P,) labeled E, F, and G in fig. 2) and the depth of the primary peak in the extinction coefficient (Pk) labeled A in figs. la, 1b and 1c were plotted as a function of incident ion energy. These results are illustrated in fig. 3. The secondary peak in the refractive index was taken as the location of the maximum point immediately preceding a consistent (and usually rapid) downward trend in the refractive index. The solid lines show the depth of the peak of the damage (RpD) and ion (RpI) distributions and the standard deviation of the distributions (ARpD and ARpI) as predicted by Sigmund and Sanders [39] and the LSS [41] theories, respectively. At spatial depths beyong the initial peak there is considerable “tailing” of the complex refractive index profile for all three implantation energies. In each case there is structure particularly in the extinction coefficients profile, which suggests that in unannealed specimens both the damage and the implanted species contribute to the changes in the optical properties of Si. This structure is evident in the second peak in the extinction coefficient labeled B in fig. 1 (particularly in the 52.5 keV specimen fig. lb), the rounding and broadening of the profiles in the vicinity of the LSS peak (RpI), and a third peak labeled C which appears in all three specimens
448
n
-
I
ABSOLUTE
3.9
4.0
la)
DEPTH,nm
128.0
f 3 64.0
F
(b)
DEPTH,
nm
128.0
f
I
192.0
RELATIVE
ABSOLUTE
n
3.E
3.9
4.c
4.1
4.2
4.3
64.0
Cc)
DEPTH,
nm
128.0
I
I RELATIVE
ABSOLUTE
Fig. 2. The refractive index profiles of the specimens of fig. 1. (a). (b) and (c) are for the 35, 52.5 and 70 keV implantation energies, respectively. The points marked E. F and C in the figures approxi~te~y correspond to the peak of the ion distribution predicted by the LSS theory 1411 and are the locations of the “secondary” peak in the refractive profile as discussed in fig. 3. The absolute and rdative uncertainties discussed in section 2 are indicated by the vertical error bars in the upper right of the figures.
64.0
3.7
n
4.1
4.2
4.3
37..
192.0
I RELATIVE
1u--
E
3.e
--+--e--t----
I
3.6-
3%
4.0..
4.1..
4.2..
4.37-
I
450
J.R. Adams, NM Bashava/Complex refractive index profiles
d
ENERGY,
KeV
Fig. 3. Depth of the primary peak in the extinction coefficeint (Q) and the secondary peak in the refrative index (P,) determined from figs. 1 and 2 as a function of incident ion energy. The secondary peak in the refractive index was taken as the location of the maximum point immediately preceeding a consistent (and usually rapid) downward trend in the refractive index. The solid lines show the depth of the peaks of the damage (R& and ion (RPI) distributions and the standard deviation of the distributions (AR~J, and ARpl) as predicted by Sigmund and Sanders [39] and the LSS theory [41].
prominent at 35 keV) near the end of the tail of the distributions. The third peak (C) is probably associated with moderatel~channeled ions which introduce interstitial atoms (atoms which are not incorporated on a lattice site or are associated with vacancy groups) which cause a localized stress and electric field perturbation of the optical constants of the lattice. Also, a stress field at the interface between the disordered and ordered regions might produce similar effects. We are sure these structural features in the profiles are physically significant because the majority of the errors discussed in the section 3 affect only the quantitative values of the complex refractive index at all points in the profile. The qualitative structure in the profiles is inaffected. From an analysis of the possible ellipsornetric errors, discussed previously, and from the results of the computer reduction of the data* we have estimated the absolute and relative errors in the determination of the refractive index profiles caused by the ~In~ertainty in the determinatiol~ of the complex refractive index (both n and k) from point to point within the profile. The magnitudes of these errors are shown by the error bars on the profiles. The absolute error will only shift an entire pr-ofile up or down on the vertical axis, it will not affect the general shape of a profile (271.
(particularly
* The program
computes
the 95% confidence
intervals
for each of the computed
parameters
(n, k).
J. R. Adams, N.M, BasharajComplex
refractive index profiles
451
0.190-
I
Q180--
RELATIVE
0.170-.
0.160--
K 0.150-
-
0.140--
0.130--
0.120
1
64 DEPTH,
128 nm
Fig. 4. Extinction coefficient profile of specimen A-4 implanted at 35 keV, dose = 2.0 X 1Ol4 ions/cm’. Annealed at 650 “C for 2 hr. Note the expanded vertical scale as compared with fig. 1. The relative uncertainty in the extinction coefficient in each layer is shown by the vertical error bar in the upper right of the figure.
The magnitude of the refractive index change (both n and k) for the 35 keV implanted specimen is 50% larger than changes in the 52.5 and 70 keV implanted specimens. This is consistent with the damage theories of Brice [40]. Even after annealing for 2 hr at 650 “C samples implanted at 35 keV exhibit structure in the extinction coefficient profile (see fig. 4). This may correspond to the residual damage observed by Tamura in electronmicroscopy studies [7]. From these results it is obvious that ellipsometry can detect even very small amounts of damage in Si, as was seen by Ibrahim and Bashara [25]. Within experimental uncertainties, no structure was observed in the refractive index (n) profile on the annealed specimen. A brief comment is in order on unaided visual effects. Among the low dose samples, only the specimen implanted at a dose of 2.0 X 1014 ions/cm2 at 35 keV exhibited a milky appearance on the as-implanted surface. This was not evident on the asimplanted surface of the 52.5 or 70 keV specimens. However, all of the high-dose specimens (1 .O X 1016 ions/cm2) exhibited a milky appearance on the as-implanted surface. The milky appearance was definitely not associated with contamination;
J.R. Adams,
452
N.M. Bashara/Complex
refractive
index p-ofiles
many authors have reported it to be associated with disorder. Our observations indicate the disordered region in the 35 keV specimen, (dose = 2.0 X 1014 ions/cm*), was sufficiently close to the surface to be observable. 3.2. Complex refractive index on as-implanted surface In order to evaluate the usefulness of ellipsometric determination of the complex refractive index profile solely from measurements on the as-implanted surface, the data on the as-implanted surfaces of the specimens in figs. 1 and 2 was analyzed using the procedure described in section 2.4.1. Fig. 5 shows the computer-determined profile for the specimen implanted at 35 keV. Similar results were obtained at the higher implantation energies; in general, good agreement was obtained between the profile determined from measurements on the as-implanted surface, and the profile determined from the anodization-stripping experiments*. Several dominant * Since the computer-generated profiles are not perfect duplicates of the true profiles it is not unexpected that the second peaks should be depressed. The major features of the profiles remain in good qualitative
if not quantitative
agreement.
4.5r
0.60
I
II
1
J
140.0
70.0 DEPTH,
r
nm
DEPTH,
“m
(b)
Fig. 5. Refractive index and extinction coefficient profiles - (a) and (b), respectively - in PiI ion-implanted Si as determined by multiple-angle-of-incidence ellipsometry on the as-implanted surface. Implantation energy = 35 keV; dose = 2.0 X lOI ions/cm’.
J.H. Adams. R-M. Baslrara/Complex
wfracrise index profiles
453
features were noted in the computer solutions for the refractive index profile. These are discussed in more detail in the following section.
Ellipsometric measurements on the samples implanted to a dose of I X lO’6 ions/cm* were restricted to four angles of incidence on the as-implanted surface, and the cwnplcx refractive index profile was computed using the techniques described in section 2.4.1. Multiple-angle-of-incidence measurements (at 55.0, 60.0, 65.0 and 70.0”) were made on at least two specimens at each of the three implantation energies. The values of I$ and A at all angles of incidence for different specimens implanted at a given energy were in agreement to within 0.1’. The variation (at ;I given angle of incidence) in $ and A between specimens implanted at the different energies was between lo and 7”. Figs. 6 and 7 show the computer-determined profiles for the specimens implanted at 35 keV and 53. 5 keV, respectively. The 70 keV profile is not shown because is is similar and contributes no new information which is pertinent to the discussion. To find a minimum least-squares solution for the profile. the initial starting guess for ea
4.30r
4.20-
4 io-
n 4.00
3.90.
3.80
3 70 -
1J L-J
L.-.
L _
--
70.0
DEPTH, (a)
140.0
nm
.1400
7To_AL-
D’PTH.
nm
(b)
Fig. 6. Refractive index and extinction coefficient profiles (a) and (b), respectively in I’;, ions-implanted Si as determined by multiple-angle-of-incidence ellipsomctry on the as-implanted surface. Implantation energy = 35 keV; dose = I.0 x IO’” ions/cm’.
J.R. Adams, NM
454
BasharalComplex
refractive index profiles
4.2Or
410-
4.00
-
0.60
-
0.40
-
0.30
-
_
n 310.
k 3.60
-
3.70
-
3.60
I
I
1
70.0 DEPTH,
140.0 nm
(a)
Fig. 7. Refractive index and extinction determined tion energy
70.0
140.0 DEPTH,
nm
@I
coefficeint profiles - (a) and (b), respectively by multiple-angle-of-incidence ellipsometry on the as-implanted surface. = 52.5 keV; dose = 1 .O X lOI ions/cm2.
- as Implanta-
overlapping distributions had to be within approximately 20% of the “true” solution for the program to converge. If the initial starting guess was in error by more than approximately 20% (particularly in the magnitude parameters) the program would diverge, or stop before a reasonably small sum-of-squares had been obtained. On occasion a “reasonable” minimum sum-of-squares was obtained, but one or more of the parameters of the distributions at this “solution” would be in violent disagreement with the known physical situation. (Usually the standard deviation of the refractive index profile would be several hundred nm, giving a very deep, flat profile). This requirement was determined primarily from the data on the specimens implanted to a dose of 2 X 1014 ions/cm. The “best” (minimum sum-of-squares) solution was always obtained by allowing all eight parameters to be adjusted simultaneously, but on occasion it was desirable to hold one or more of the parameters constant in order to obtain a better estimate for the other parameters or to examine the effects of variations in the parameters of the distributions. The “best” solution was used for the profiles of figs. 5-7. In the process of obtaining the computer-generated profiles, several important features of the distributions were noted. First, the magnitude of the peak of the refractive index profile of the Gaussian distribution always fell between 114 and 116% of the substrate refractive index. The most typical value (for 4 out of the
J.R. Adams, N.M. Bashara/Complex
refractive index profiles
455
6 specimens analyzed in this manner), for the peak in the refractive index, was 115% of the substrate refractive index. The magnitude of the refractive index peak of the first Gaussian distribution was the most accurately determined parameter of all the parameters of both distributions. This result is in good agreement with the magnitude of the peak of the refractive index obtained from the anodization-stripping experiments. Second, the profiles determined for the specimens implanted to a total dose of 1 X lOI ions/cm2 were all definitely broader than the profiles of the specimens implanted to a dose of 2 X 1014 ions/cm *. This is consistent with the damage theories of Sigmund and Sanders [39]. Third, in order to obtain a convergent solution, it was necessary to allow the second (deepest) Gaussian distribution to have a negative effect on the refractive index profile. The effect on the overall refractive index profile was to force a relatively narrow dip in the refractive index usually causing y1to go below the refractive index of the substrate. The depth of the minimum in the refractive index profile (not necessarily the depth of the maximum negative effect) always occured at a depth of approximately 70nm. If the positions (depth) of the Gaussian distributions were forced to occur away from the optimum solution (by holding the depth parameters constant at a value 5510 nm from the optimum solution) the magnitudes and standard deviations of the distributions were adjusted by the program to force the negative dip in the overall refractive index profile to occur at the same depth. The depth of this negative peak appeared to move somewhat deeper into the specimen with increasing implantation energy, although within the resolution of these profiles, making a firm statement to this effect is not possible. Finally, the magnitude of the peak of the extinction coefficient profile was the least acc.uately determined parameter of all the parameters of the distributions. The magnitude of the peak of the extinction coefficient profile could be held anywhere between 400% and 600% (300-500% increase) of the substrate extinction coefficient with only a small effect on the sum of the squares. However, the shape of the extinction coefficient profile (depth, standard deviation) remained unchanged when the magnitude of the peak was forced away from the optimum solution. 3.3. Properties of the damage region jhm a simplified model In order to assess the usefulness of ellipsometry for measuring the extent of the residual damage following annealing, the initial multiple-angle-of-incidence ellipsometer measurments of Specimen A-4, both before and after annealing to 650 “C for 2 hr, were used to compute the properties of the surface by assuming simplified one-layer and two-layer models. For the unannealed specimen, neither the one-film model nor the two-film model were adequate enough representations of the refractive index profile to allow the computer program to converge to a satisfactory solution for an average complex refractive index for the damage region; the convergence was extremely slow and the refractive index and extinction coefficeints for the films in
456 Table 2 Kcsults
of nlultiplc-angle-of-incidence
refractive
two ways shown. consistine complex refractive following
complete stripping and k 0.140)
One-film
model
3.920
i 0.204
10.040
*Cl.040
d,.= 77,s
of the phosphorus-implalltcd
1 he
? lI).O’-
wrface
15.3 ? 5.0
model
were
model
film
unreasonably
thicknesses of the
mately
74 nm.
guesses
f01 tl,
74 nm
and
For
the
and
this
the
would
value
by
program.
as the
the found
used
(in
+ d2
of the (The
to be approxiwere
in the
found
models.
two-layer rapidly
Starting
inodel) converge
program
two-
the individual
could to about
adjusted
the
corn-
solutions model
of‘ the in lmtcr
expcrimcnrs.
‘I‘llc
were
provided
found
refractive
index
qrecment
wirh
values
~OI- both
;I slightly
better
and for
the
one-layer
to the
extinction
the values
wmputed
the fit
and
coel‘ficicnr
obtained two
two-
experimental fbr the
by the anodim
models
are shown
in
3.
Although
the
in the
one-film
two-tilm
annealed
model damage
n~~~clcl is in better
of the
damage
Overall.
on
results
;I
was
the
the
profile
;mtl from
cst imate of the
the values
of the film
complcs
refractive
computed
computed
for
the
by
the
thickness
studies.
good
the
3 better thickness
with
unannraled
are in very
computed
Gaussian
provides region.
q,reement
experiments.
which
assuming
region
these
tion-stripping index
about
spccimcn.
were
cl,
sum while
thicknesses
guesses
thicknesses
the
layer(s).
two-film
magnitudes
n~xfcl
lion-stripping
index
oscillate ofrhc
and
the
10.003
76 nm was
computed
starting
-i 0.16(1
that
adjusted
llfll
1.fl + 12.0nm
noted
model
These
initial
and
being
one-layer
model)
00 nm
then index
the
one-layer
it was
were
later.)
10 the
and
the
3.864 to.003
at approximately
the
be discussed
the
nnncalcd
two-layer t:1hlc
for
would
models;
data
indices
computed
ml
refractive
Isyer
refractive
film
(in
50
However.
constant
and
insentitive
be between plex
large.
relatively
3s will
to be relatively
=
cII.1 + d1.2 = 6
remained
thickness
j
near
wbstratc
at the value obtained
region; )I was found to
,111
._ the
Si was modeled in the
in the calculations
Region
layer
on specimen A-4 (the comples
of either one or two layers; the unimplantcd
indcs was held constant
bc 3.863
‘if=
measurements
index and thickness of the I~hosl)h’)rus-iruplantcd
agreement
location
me;lsurements
for the unannealed
with
of the
the
negative
on the
results dip
as-implanted
specimens.
of the
in the
anodiza-
refractive surface
hq
457
4. Conclusions
We have den~onstrated that it is possible to use e~lipsom~try to dctcrrl~inc the complex refractive index profiles of high energy ion-implanted silicon surfaces both by anodization-stripping of the implanted surface and from analysis of multiplcangle-of-incidence measurements on the as-implanted surface assuming the implantation distribution was Gaussian. The complex refractive index profiles generated from the measurements on the as-implanted surface are in good agreernent with those obtained by aItodiZati(~n-stripping. As expected. the depth of the peaks of the profiles increase with increasing energy of the implanted ions. Annealed specimens show effects (evident in the extinction coefficient profiles) of residual damage even after a treatment of 650 “C for 2 hr. The properties of the annealed surface may also be approximated by using a simplified one-layer or two-layer model; the overall depth of the damage region determined from these simplified ~mxlels are in good agreement with results obtained by anodi~atiorl-stripping and/or the use of the more sophisticated Gaussian model. The reasons for the negative dip observed in the refractive index (H) profiles for the unannealed specimens which occur at approximately 70.0 nm below the surface. will be reported elsewhere.
Acknowledgement
The authors are grateful to Dr. W. Beezhold for discussions regarding ion implantation and advice and assistance in fabricating the ion-implanted samples.
References
Ill
See for example: J.W. Mayer, L. Eriksson and J.A. Davies, Ion Impiantalion in Semiconductors (Ac;ldcmic Press, New York, 1970); R.C. Wilson and G.R. Brewer, Ion Beams, with Application to Ion Implantation (Wiley, New York, 1973). (a) J.F. Gibbons, Proc. IEEE 60 (1972) 1062, and refcrcnces therem; tb) J.F. Gibbons, Proc. IEEE 56 (I 968) 295. 131 Y. Akasaka. K. Horie, K. Yoneda, T. Sakurdi, H. Nishi, S. Kawabe and A. Tohi. J. Appl. Phys. 44 f 1973) 220. 141 R.L. Jacobsen and G.K. Wehner, J. Appi. Phyc. 36 (1965) 2674. [51 H.E. Farnsworth and H. Hayek, Surface Sci. 8 (1967) 35. 161 M. Hen&r, Surface Sci. 22 (1970) 12. [‘I M. Tamura, Appl. Phys. Letters 23 (1973) 651: and refercncer therein. Phyc. Rev. Lcttcrc 27 (1971) 1794. 181 T.M. Donovan and K. Hcinemann, 191 D.J. Nelson, R.S. Mazey and R.S. Barns, Phil. Msg. 17 (1968) 1145. 1101 R.L. Crowder, R.S. Title, Y.H. Rrodsky and G.D. Pettit, Appl. Phys. Letters 16 (1970) 205.
458
J.R. Adams. N.M. Bashara/Complex
[ 111 N.N. Gerasimenko, [ 121 [ 131 [14] [l51 [l61 1171 1181 [l91 (201 [2ll [221 ~231 1241 [251 [261 1271 1281 [291
[301 [3ll [321 [331 [341
[351 L361 1371 [381 1391
[401 1411 [421 [431
refractive
index profiles
A.V. Dvurechenskii, S.I. Romanov and L.S. Smirnov, Soviet Phys.Semiconductors 6 (1973) 1692. S. Hasegawa, K. Ichida and T. Shimizu, Japan. J. Appl. Phys. 12 (1973) 1181; and references therein. K.L. Brower and W. Beezhold, J. Appl. Phys. 43 (1972) 3499. S.M. Davidson and G.R. Booker. in: Proc. First Intern. Conf. on Ion Implantation, Eds. L. Chadderton and F. Eisen (Gordon and Breach, New York, 197 1) p. 81. D.J. Bilekno, V.M. Evdokimov, N.P. Kaganova, G.P. Kostyunina, A.A. Kukharshii al:d V.K. Subaskiev, Soviet Phys.-Semiconductors 5 (1972) 1247. L.N. Strel’tsov and LB. Khaibullin, Soviet Phys:Semiconductors 5 (1972) 2083. T.C. McGill, S.L. Kurtin and G.A. Shifrin, J. Appl. Phys. 41 (1970) 246; Appl. Phys. Letters 14 (1969) 223. H.J. Stein, F.L. Vook and J.A. Borders, Appl. Phys. Letters 14 (1969) 328. C.S. Chen and J.C. Corelli, J. Appl. Phys. 44 (1973) 2483. R.R. Hart and O.J. Marsh, Appl. Phys. Letters 14 (1969) 225. V.M. Evdokimov, A.A. Kukharshii, L.N. Strel’tsov, V.B. Titav and LB. Khaibullin, Soviet Phys-Semiconductors 4 (1970) 797. J.C. Dyment, J.C. North and L.A. D’Asaro, J. Appl. Phys. 44 (1973) 207. S.M. Spitzer and J.C. North, J. Appl. Phys. 44 (1973) 214. D.D. Shell and A.U. MacRae, J. Appl. Phys. 41 (1970) 4929. M.M. lbrahim and N.M. Bashara, Surface Sci. 30 (1972) 632.Damage induced by 200~-400 eV ions is discussed. J.B. Schroeder and H.D. Dieselman, J. Appl. Phys. 40 (1969) 2559. Only the real part (n) of refractive index is determined. A.R. Bayley and P.D. Townsend, J. Phys. D (Appl. Phys.) 6 (1973) 1115. D.G. Schueler, Surface Sci. 16 (1969) 104. M.M. Ibrahim and N.M. Bashara, J. Opt. Sot. Am. 61 (1971) 1622. This paper shows that a necessary condition for use of multiple-angle ellipsometry is independence of the optical parameters with respect to an ellipsometric parameters. Because of dispersion effects this constraint is wavelength dependent. K.M. Bunsen and R. Linzey, Trans. Met. Sot. AIME 236 (1966) 306. W. Przyborski, J. Roed, J. Lippert and L. Sarholt-Kristensen, Radiation Effects 1 (1969) 33. J.R. Adams and N.M. Bashara, Surface Sci. 47 (1975) 655. B.J. Stern, Rev. Sci. Instr. 34 (1963) 152. For a general discussion of ellipsometry see for example, N.M. Bashara, A.C. Hall and A.B. Buckman, in: Physical Methods of Chemistry, Vol. I, Part IIIC, Eds. A. Weissberger and B. Rossiter (Wiley, New York, 1972) p. 453. J.R. Zeidler, R.B. Kohles and N.M. Bashara, Appl. Opt. 13 (1974) 1115. J.R. Zeidler, R.B. Kohles and N.M. Bashara, Appl. Opt. 13 (1974) 1591. A. procedure for in-process calibration of the ellipsometer divided circles is being reported elsewhere. R.M.A. Azzam andN.M. Bashdra, J. Opt. Sot. Am. 61 (1971) 600; J. Opt. Sot. Am. 61 (1971) 1118. P. Sigmund and J.B. Sanders, in: Proc. Intern. Conf. on Applications of Ion Beams to Semiconductor Techniques, Grenoble, France, 1967 Ed. P. Glotin (Editions OPHRYS) p. 215. D.K. Brice, Radiation Effects 6 (1970) 77. J. Lindhard, M. Scharff and H. Schiott, Kgl. Danske Videnskab, Selskab, Mat. Fys. Medd. 33 (1963) 1. J.M. Morabito and J.C. Tsai, Surface Sci. 33 (1972) 422. G. Dearnaley, J.H. Freeman, G.A. Gard and M.A. Wilkins, Can. J. Phys. 46 (1968) 587.