Home Search Collections Journals About Contact us My IOPscience

Depth profile composition studies of thin film CdS:Cu2S solar cells using XPS and AES

This content has been downloaded from IOPscience. Please scroll down to see the full text.

1981 J. Phys. D: Appl. Phys. 14 1647

(http://iopscience.iop.org/0022-3727/14/9/012)

View the table of contents for this issue, or go to the journal homepage for more

Download details:

IP Address: 128.122.253.228
This content was downloaded on 13/02/2015 at 06:56

Please note that terms and conditions apply.

J. Phys. D : AppI. Phys., 14 (1981) 1647-56. Printed in Great Britain

Depth profile composition studies of thin film CdS: Cu,S

solar cells using XPS and AES

VG Bhidet, Saji Salkalachent, A C Rastogit, C N R Raot. and

M S Hegdet.
t Division of Specialised Techniques, National Physical Laboratory, Hillside Road,
New Delhi-110012, India
3 Solid State and Structural Chemistry Unit, Indian Institute of Science, Bangalore-
56qo12, India

Received 21 January 1981, in final form 7 April 1981

Abstract. Surface composition and depth profile studies of chemiplated thin film
CdS:CuzS solar cells have been carried out using x-ray photoelectron spectroscopy
(XPS) and Auger electron spectroscopy (AES) techniques. These studies indicate that the
junction is fairly diffused in the as-prepared cell. However, heat treatment of the cell
at 210°C in air relatively sharpens the junction and improves the cell performance.
Using the Cu(2p3p)/S(2p) ratio as well as the Cu(LVV)/(LMM) Auger intensity ratio,
it can be inferred that the nominal valency of copper in the layers above the junction
is Cut and it is essentially in the CUSSform. Copper signals are observed from layers
deep down in the cell. These seem to appear mostly from the grain boundary region.
From the observed concentration of Cd, Cu and S in these deeper layers and the
Cu(LVV)/(LMM) ratio it appears that the signals from copper essentially originate
partly from copper in CuS and partly from Cu2t trapped in the lattice. It is significant
to note that the nominal valence state of copper changes rather abruptly from Cut to
Cuz+ across the junction.

1. Introduction

Thin film CdS :CuzS solar cells have attracted considerable attention during recent years.
Although various approaches have been followed for the fabrication of these solar cells,
the most popular one uses vacuum evaporated CdS layers overcoated with chemiplated
CUZS(Shiozawa et a1 1969, Boer and Meakin 1975, Schock et a1 1977, Bhide et nl 1979).
Extensive structural (Mukherjee et aZ1977, Shirland 1979a, b, Norian andEdington 1980),
and compositional (Van Aerschodt et a1 1971, Martinuzzi et a1 1972, Hill and Edwards
1977, Iwami et a2 1978, Morimoto 1980) studies of these cells and investigations of the
electrical and optical properties of the constituent CdS and Cud3 layers (Windawi 1975,
Rastogi et aZ 1978, Shewchun et al 1978, Rezig et aZ1979, Hall and Meakin 1979) have
brought out the importanceof junction profile, stoichiometry, thickness and structure of
the CUZSlayer in determining the efficiency of these solar cells. Most of the analytical
techniques employed in these investigations, being insensitive to morphological features,
have yielded information on the average composition and thickness of the CUSS layer
(Caste1 and Vedel 1975, Baron et aZ 1978, Shirland 1979). Since a polycrystalline CdS
layer is riddled with grainboundariesseparating 1-2 pm columnargrainsandalso
various other defects, the composition and thickness of the overcoated CUZSlayer is not

0022-3727/81/091647+
$01.50
11 0 1981 The Institute of Physics 1647
1648 V G Bltide et a1

expected to be uniform all over the cell. In view of this, it appears necessary to study the
composition of CUZS atvarious depths in these cells. In fact, such depth profile studies
will also provide information on the junction profile. Further, these studies are expected
to provide the rationale to the various recipes that have been evolved, such as air heating
of the as-prepared cell at 200°C for a few minutes (Shiozawa et a1 1969, Hall 1975,
Devaney et a1 1979) and depositing a 40-60 A thick copper layer on top of the CUZS
layer (Bogus and Mattes 1972, Pfisterer et a1 1975, Bhide et a1 1979), to optimise the
cell performance. In this paper we report the depthprofile studies of the as-prepared and
heat-treated CdS: CUZSsolar cells using XPS and AES techniques.

2. Experimental details
Polycrystalline cadmium sulphide films (resistivity z 10 !2 cm, thickness z 3 0 pm) are
thermally deposited through suitable masks on to silver-backed electrodes predeposited
on to window glass substrates.CdS films consist of columnar grains approximately
2 pm across. The surface of these films is texturised in hot HCl(1: 1) solution to achieve
enhanced light absorption and increased effective junction area. This procedure results
in sharply defined pyramidal structures with hexagonal bases (Norian and Edington1980)
as seen in figure l(a) (plate). Cuprous sulphide films are topotaxially grown by chemi-
cally plating CdS films in a hot CuCl(1) bath for a few seconds. The CuCl(1) bath was
agitated by continuouslybubblingnitrogenthrough it. This, besides agitatingthe
solution, precludes traces of oxygen being trapped in the film. The plating parameters
such as the pH of the solution (pH=2), concentration of Cu+ ions (5 g CuCl per litre
of water), plating period (5 S ) and temperature of the bath (95 "C) were controlled to
provide optimum composition and growth of rate
theCusS layer. After the formationof the
junction, the cell is thoroughly rinsed in a series of deionised water baths and then dried.
Small area samples (5 mm x 5 mm) were cut from thelarge solar cells (12 cmz) fabricated
as above and used for depth profile studies.
The CdS:CuZS junction samples were mounted on copper specimen holders using
silver paste. The A1 K, x-ray photoelectron spectra and the differentiated Auger electron
spectra were recorded on an ESCA-3 Mark I1 Spectrometer of VG Instruments Ltd,
UK. Thespectra were invariablyrecorded at vacuum levels betterthan Torr.In
order to study the depth profile, the samples were subjected to successive Ar+ ion beam
etches (4 kV, 40 pA) in the preparation chamber at 1 x 10-5 Torr. The etching rate was
determined using CUZSfilms of known thickness and was found to be 100 A min-l.
Examination by scanning electron microscope of the successively etched samples
shows (figure l(b) plate) that the etching is conformal as long as the etching rateis below
200 A min-1. In excess of this rate the etching is not conformal (figure l(c)).
The binding energy of the Au (4f7p) transition 2t 83.7 eV was used to calibrate the
binding energy scale of the spectrometer. The absolute concentrations of the various
elements such as Cd, Cu, S were not determined. The peak heights and areas under the
peaks have been used to measure the relative concentrations.Thebinding energies
reported are within 2 0.2 eV.

3. Results and discussion

In order to facilitate the identification of Cd, Cu and S in CdS: CUZS solarcell samples,
the XPS spectra of model systems such as CdS single crystal, CUZSand CuS powders etc
were initially studied. The binding energies of the valence and core levels of Cd in CdS,
 Depth projiles of CdS:CuZS solnr cells by XPS and AES 1649

Table 1. XPS binding energy (eV) of valence and core level peaks in CdS, CuzS and CuS
systems.

System Level
~ ~~ ~~~ ~

CdS CdS CdS CuzS cus

Single Evaporated Chemically Powder Powder
crystal film etched film
~~

10.9 11 .o 11.0
404.8 405.0 405.2
411.6 411.7 412.0 -
161.4 161.3 161.5 161 .O 161.O
225.4 225.4 225.6 225.3 225.6
2.2 1.7
931.8 931.6
951.8 951.6

Cu in CuaS and in CuS and those of sulphur in CdS, CuaS and in CuS are given in table 1.
It is interesting to note that except for the Cu(3d) valence band, none of the other levels
show any appreciable change in the binding energy. The observed change in binding
energy of the Cu(3d) valence band in CuzS (2.2 eV) and in CuS (1.7 eV) may perhaps
be due to the change in the valence state of copper. It appears, however, that the change
in the binding energy is not a sensitive parameter to determine the change in the valence
state. In contrast, Rao et a1 (1980) have shown that metal Auger line intensity ratios
such as Cu (L~M45M45) toCu(Lz3M23Mm) are quite sensitive to occupancy of valence
band states of the metal in their oxides and sulphides. Our studies on Cu, CuzS and CuS
have shown that this ratio is 5.1,4.7,4.4 ( i-0.1) respectively. In our depthprofile studies
of CdS :CuaS solar cells, we have used the metal Auger line intensity ratio to determine
the charge transfer from the metal ion and in turn the nominal valency of copper in
various layers of Cu2S :CdS solar cells.
Table 2 gives the peak heights of Cd and S in CdS and of Cu and S in CUZS andin
CuS. The intensity of these lines (peak heights) depends upon the intensity of incident
radiation, concentration of the element, transition probability etc. In view of this, the
absolute determination of elemental concentration is not easily possible. However, in a

Table 2. Relative XPS peak heights (IO3 counts S-1) of different levels of Cd, Cu and S
in CdS and CuzS systems.

System Level

CdS CdS CdS cuzs cus

Single Evaporated Chemically Powder Powder
crystal film etched film

51 32.1 26.1
282 186 126
201 114 87 - -
31.5 18.0 14.4 4.8 8.1
18.9 11.4 8.4 3.6 4.8
4.0 6.2
78 109
37 47
1650 V G Bhide et a1

binary compound such as CdS, CuzS or CuS, the ratio of the intensity of two levels,
one of the metal ion and the other of sulphur, serves as a good and reliable measure to
determine the ratio of the concentration of the metal and sulphur ions and in turn the
stoichiometry of the material under investigation. We found that the ratio of the two
levels Cd(3dbp)/S(2p) or Cu(2p3/a)/S(2p) was fairly constant in CdS and CuzS or CuS
respectively and was independent of the operating conditions. Of course, it is possible
to use other ratios such as Cd(3d5p)/S(2s), Cu(2p3/z)/S(2s) etc. It is interesting to note
that Cu(2p3/z)lS(2p) ratio varies from 16.3 to 13.5 as one goes from CUZS toCuS.
These studies on model systems provided us with a means (i) to identify the core and
valence levels of Cd, Cu and S, (ii) to determine the relative concentration of Cd/S and
Cu/S in CdS and in CuzS or CuS respectively and (iii) to determine the nominal valence
state of Cu from the Auger ratios.
The binding energies and the peak heights of various core and valence levels of Cd
and S in single crystal CdS, vacuum evaporated thin films of CdS and etched films of
CdS are shown in tables 1 and 2. In all these specimens, there is no difference in the
binding energies of core and valence levels of Cd. However, the ratio of the intensities
of Cd (3d5,~) and S(2p) varies from 9.0 in CdS single crystal to 10.3 in CdS evaporated
films and then to 8.7 in chemically etched films. These ratios suggest that Cd is in excess
and S is relatively deficient in evaporated films. This result is in conformity with the well
known fact that thermally evaporated films are sulphurdeficient (Hill and Edwards 1977).
A further check on this result is provided by the resistivity of the films. The decrease in
the ratio from 10.3 in evaporated films to 8.7 in chemically etched films indicates that
either the deeper layers of the evaporated thin films of CdS are more stochiometric or
etching reduces the excess of Cd thereby improving its stoichiometry.
A fast scan XPS spectrum of the as-assembled junction is shown in figure 2. Besides
Cu(3d) and S(2p), the XPS spectrum reveals Cd(3d~p),Cd(3d3,z), Cd(4d), O(ls) andCl(2p)
levels. However a mild Ar+ ion etching (about 30 S ) almost entirely removes the Cd, Cl

Blnding
energy lev1 -
Figure 2. Fast scan XPS spectra of chemiplated copper sulphide film; (- - -) as-prepared
and (-) ion etched.
 Depth profiles of CdS: Cues solar cells by XPS and AES 1651

and 0 peaks to give a cleaner CuzS spectrum. This shows that the impurities arelocalised
in the thin surface layers. Perhaps the observation of Cd and Cl peaks are due to CdC12
formed during the topotaxial reaction, which is not washed off completely. This result
brings out the need of thorough washing of the prepared cell, before the front electrode
(grid) is fixed on to it.
The CdS:CuaS junctions were successively sputter etched with an Ar- ionbeam.
After every etch period (30 S ) , the XPS spectra were recorded. Figure 3 shows the results
of a typical depth profile study. It is seen that as the etchingproceeds the copper concen-
tration increases, attains a saturation and then decreases gradually. In contrast, the Cd
concentration is negligible to start with but later increases with etching time. There is a
cross-over of the copper and cadmium concentration at a depth of about 1300 8, below

r __”___

l
1

Ar’ eichlng persod (PIP:

Figure 3. Depth profile of an as-prepared cell using peak heights of Cd(4d), Cu(3d) and
S(2p) as relative elemental concentration: Cu(LVV)/(LMM) Auger intensity ratio is
shown at two regions of the cell by arrows.

the top surface. It is also easy to infer that the heterojunction profile in the as-prepared
cell is fairly diffuse in contrast to a sharper profile in dry-barriered CdS: CUZS junctions
(Morimoto 1980). It is also interesting to note that there is coexistence of Cd and Cu
all through the heterojunction. This is understandable in view of the topotaxial reaction
involved in the formation of CuzS. This reaction proceeds by a continuous exchange of
every Cd ion with two Cu+ ions which get frozen at different sites in the heterojunction
when the reaction is abruptly terminated.
The timescanspectra of the valence bands of Cu(3d) and Cd(4d) on successive
etchingareshownin figure 4. It is observed that the binding energy of theCu(3d)
valence band in the upper CUZSlayer is 2.2 eV, whereas it assumes the value of 1.7 eV
in layers deep in the junction. It is interesting to observe, recalling the binding energy
of the 3d valence band in model systems (Cu2S and CuS) that upperlayers of the junction
correspond to Cu2S whereas in the layers deep down the junction copper appears to be
in the form of CuS. To confirm this we recorded the Cu(LVV)/(LMM) Auger intensity
1652 V G Bhidealet

Blnding energy ( e l ' )

Figure 4. Valence band spectra of Cu and Cd as a function of Art ion etch time.

ratio in the upper and thedeeper layers of the cell. This ratio is about 4.78 at a depth of
300 A, indicating that the valence state of copper is Cu+. In deeper layers 2800 A) this
( W

ratio is seen to be 4.2 indicating that the valence state of copper at this depth is Cu2+.
Figure 5 shows the typical depth profilestudies of the cells heat-treated in airat 210 "C
for 5 min. In contrast to that in the as-prepared cell, the junction in the heat-treated cell
is much sharper. The relative Cu(3d) intensity decreases fairly rapidly after about 600 A.
Judging from this variation of copper as well as the rapid rise in the intensity of Cd(4d)
signal below this depth, itis inferred that air heat-treatmentcauses considerable diffusion
of copperandcadmium species towardstheuppersurface.The diffusion of copper
towards the upper surface is probably due to the presence of oxygen at the top during the
heat-treatment in air. These migrating copper ions fill the copper deficient sites in the
upper layers of copper sulphide and improve its stoichiometry (Burton and Windawi
1976). These studies clearly bring outthe rationale in airheat-treatingthe cells to
improve the cell efficiency. In this particular case the Vac, Zsc and of the as-prepared
andairheat-treated cells are (0.315, 0.505 V), (63.0, 101.0mA)and (1.1 %, 3.97;)
respectively.
 Depth profiles of CdS: CuzS solar cells by XPS and AES 1653

A r etching pertod imin)

Figure 5. Depth profile of an air heated cell (heattreatment at 210°C for 5 min).
Variation of Cu(LVV)/(LMM)ratio shown as a function of depth.Insert indicates
nominal valency of Cu in Cu metal, Cues and CuS model systems.

In the airheat-treated cells also, XPS studies indicate the existence of copper in deeper
layers of the cell. We studied the depth profiles of the Cu(LVV)/(LMM) ratio in these
heat-treated cells. This is also shown in figure 5 . Using the calibration curve relating the
Cu(LVV)/LMM) ratio with the nominal valency of copper, it is easy to infer that in the
upper layers up to 600 A deep, the nominal valency of copper is Cuf whereas in the layers
deeper down the junction, the nominal valency is Cu2+.
The depth profile studies also show that the nominal thickness of CuzS layer formed
during the topotaxial reaction is about 600-700 A. This estimate of the thickness of the
CUZSlayer agrees very well with that inferred earlier fromgrowth kinetic studies
(Salkalachen et a1 1981).
It may be recalled that the C U ~ layer
S is formed topotaxially over a polycrystalline
CdS layer with columnar grains of 2 pm across. This reaction takes place differentially
over the grain and in the grain boundary region. It is well known that diffusion of an
impurity is much faster in the grain boundary than within the grain. It appears more than
likely that copper observed deep in the cell is essentially due to its presence within the
grain boundary region rather than in the grain (Rothwarf 1976, Donaghey et al 1980,
Salkalachen et a1 1981). This inference is strengthened by the scanning-electron-micro-
1654 V G Bhide et a1

il -Grain boundary

i
Figure 6. Schematic cross-section of the cell after a heat treatment showing copper in
the CdS grain boundary (dashed arrow) and a nominal CUZSlayer ofthickness 600-700 A
(thick arrow). A dotted line at a depth of 2800 8, indicates that observed spectra
originate from the CdS grain and grain boundary.

scope studies which have clearly brought out the conformal sputter etching. This is shown
by the dashed arrow in figure 6 which schematically depicts the cross-section of the cell.
The chemical state of copper within the grain boundary region deep in the junction can
be inferred from therelative concentration of Cd, Cu andS as well as from theCu(LVV)/
(LMM) Auger ratios. At a depth of 2800 A below the original top surface one is reason-
ably certain that as far as thegrains are concerned we are in the CdSregime (in view of
the conformaletching, see figure 6). This is also indicatedby the saturationof the Cd(3ds,z)/
S(2p) ratio at about this depth (figure 7). At this depth, the observed concentration of
Cd, Cu and S in arbitrary units is 31.5 x 104, 18.6 x lo4 and 4.0 x lo4 respectively. Using
the Cd(3dbIz)/S(2p) ratio as observed in the model system of evaporated CdS, we find

l
l I l
0 10 20 30 40
Art etching period ( r n ~ n l

Figure 7. Cd(3ds/z)/S(2p) intensity ratio as a function of Ar+ ion etching period.

 Depth projles of CdS: CuzS solar cells by XPS and AES 1655

that the sulphur concentration necessary to account for 31.5 x IO4 units of Cd is about
3.02 x l o 4 units. We are thus left with 0.98 x IO4 units of sulphur to account for the
observed copper.The Auger ratio Cu(LVV)/(LMM) at this depth indicates that the
nominal valency of copper is Cu2+. It appears therefore that the copper observed at
this depth is partly in the form of CuS or in the form of Cu2+ trapped in the CdS lattice.
It is not understood why in the grain boundary region divalent copper is preferred.
Using the observed concentration of Cd, Cu andS in the toplayers, one can reasonably
state that the top layer is essentially Cu2S.

4. Conclusions

(i) The top copper sulphide layer of an as-prepared chemiplated CdS :Cu2S solar cell
contains impurities such as Cd, Cl, 0 and C. These impurities are detrimental to the
efficiencyof the cell and should be completely washed off before the top electrode (grid)
is applied to the cell.
(ii) This impurity layer can be completely removed by mild Ar" ion beam etching
suggesting that it may be desirable to ion clean the cell.
(iii) The junctionprofile in the as-prepared cell is diffuse. This becomes much sharper
on heat treating the cell in air at 210 "C for 5 min. The sharpness of the junction is
caused by the diffusion of copper to the upper surface, improving the stoichiometry of
the Cues layer. Both the sharpening of the junction and improvement of stoichiometry
improves the cell performance.
(iv) The nominal thickness of Cu2Sin operative thin film cellsis approximately
600-700 A.
(v) Vacuum evaporated (CdSFchemiplated (&S), CdS :CuzS solar cells invariably
show the existence of copper deep in the junction. It appears that thisobserved copper is
essentially in the grain boundary region in the form of CuS or C$+ trapped in the
lattice. The nominal valency of copper above the junctionis C u i whereas that below the
junction is Cu2+.

Acknowledgments

We are grateful to the Tata Energy Research Institute, Bombay, forsponsoringthe

project on development of CdS: CuzS solar cells. One of us (SS) is thankful to the
Council of Scientific and Industrial Research, India, for a research studentship.

References

Baron B, Catalan0 A W and Fagen E A 1978 Proc.13thIEEEPhotovoltaicSpecialists Conference,

Washington DC (New York: IEEE) pp 406-1 1
Bhide VG, Rastogi AC, Salkalachen S and Jatar S 1979 Proc. Nat. Solar Energy Convention, Bombay
~~343-8
Boer K W and Meakin J D (Eds) 1975 Proc. Int. Workshop on CdS Solar Cells and other Abrupt Hetero-
junctions, University of'Delaware NSF-RANN AER75-15858
Bogus K and Mattes S 1972Proc. 9th IEEE Photovoltaic Specialists Conference, Siher Springs (New York :
IEEE) pp 106-10
Burton L C and Windawi H M 1976 J . Appl. Phys. 47 4621-6
127
1656 V G Bhide et a1

Caste1 E and Vedel J 1975 Analusis 13 487-90

Devaney W E, Barnett A M, Storti G M and Meakin J D 1979 IEEE Trans. Electron. Dev. ED-26 205-10
Donaghey L F, Duisman J A, Peterson T M and Ryason P R 1980 Proc. 14th IEEE Photovoltaic Specialist
Conference, San Diego (New York: IEEE) pp 728-33
Hall R B 1975 Proc. Int. Workshop on CdS Solar Cells and other Abrupt Heterojunctions, University of
Delaware NSF-RANN AER 75-15858 284-98
Hall R B and Meakin J D 1979 Thin Solid Films 63 203-1 1
Hill R and Edwards I A S 1977 Vacuum 27 277-80
Iwarni M, Kim S , Matsuo I and Hiraki A 1978 J. Cryst. Growth 45 467-70
Martinuzzi S , Cabane-Brouty F and Bretzner J F 1972 Proc. 9th ZEEE Photovoltaic Specialists Confer-
ence, Silver Springs (New York: IEEE) pp1 11-7
Morimoto J 1980 Japan. J. Appl. Phys. 19 L2968
Mukherjee M K, Pfkterer F, Hewig G H, Schock H W and Bloss W H 1977 J. Appl. Phys. 48 153842
Norian K H and Edington J W 1980 Proc. 14th IEEE Photovoltaic Specialists Conference, San Diego
(New York: IEEE) pp 700-5
fisterer F, Hewig G H and Bloss W H 1975 Proc. 11th IEEE Photovoltaic Specialists Conference,
Scottsdale (New York: IEEE) pp46G3
Rao C N R Sarrna
, D D and Hegde M S 1980 Proc. R. Soc. A370 269-80
Rastogi AC, Salkalachen S and Bhide VG 1978 Thin Solid Films 52 1-10
Rezig B, Duchernin S and Guastavino F 1979 Solar Energy Mater. 2 53-67
Rothwarf A 1976 Proc. 12th IEEE Photovoltaic Specialists Conference, Baton Rouge (New York: IEEE)
~~488-94
Salkalachen S,Jatar S , Rastogi A C and Bhide V G 1981 Solar Cells 3 341-53
Schock H W, Bilger G, Bloss W H, Hewig G H and Pfisterer F 1977 Vacuum 27 281-5
Shewchun J, Clarke R, Loferski JJ, Burk D, Vanderwel T, Marton J, Kazandjian A and Wu J 1978
Proc. 13th IEEE Photovoltaic Specialists Conference, Washington DC (New York: IEEE) pp 378-82
Shiozawa LR, Augustine F, Sullivan G A, Smith J M and Cook W K 1969 Aerospace Research Labo-
ratories Report ARL 69-0155 Contract AF 33 (615)-5224
Shirland F A 1979a Solar Cells 1 183-97
-1979b J. Appl. Phys. 50 4714-9
Van Aerschodt AE, Capart J J, David K H, Fabbricotti M, Heffels K H , Loferski J J and Keinhartz K K
1971 IEEE Trans. Electron. Dev. ED-l8 471-82
Windawi H M 1975 Proc. Int. Workshop on CdS Solar Cells and other Abrupt Heterojunctions, University
of Delaware NSF-RANN AER 75-15858 177-90
J. Phys. D : Appl. Phys., 14 (1981)”VG Bhide et al (see pp 1647-56)

Figure 1. (a) Pyramidal surface morphology of a typical chemically etched CdS film.
(b) Pyramidal surface morphology of the crystallites after sputter etching by Ar’ ion
beam showing conformal etching pattern. (c) Destructed pyramidal feature ofcrystallites
after etching with excessive sputter etch rates.

132