surface science Surface Science 372 (1997) 289-299 ELSEVIER . ! Surface properties of indium-doped Cd2SnO 4 ceramics studied by EELS and photoemission spectroscopy Y. D o u , R.G. Egdell * Inorganic chemistry Laboratory, South Parks Road, Oxford OX1 3QR, UK Received 6 M a y 1996; accepted for publication 3 September 1996 Abstract Ceramic samples of indium-doped dicadmium stannate (Cd2SnO4) have been studied over the doping range 0.5-2.0 at% In by electron energy loss spectroscopy (EELS) and ultraviolet and X-ray photoemission spectroscopy (UPS and XPS). The samples were prepared from SnO2 and Cdl_~InxO, which was itself made from CdO and In203. The maximum surface plasmon energy is around 0.450 eV. The carder concentration derived from the plasmon energy is always lower than that estimated from the bulk dopant concentration. The atomic percentages of both cadmium and tin on the surface before annealing, determined from XPS, are lower than those in undoped Cd2SnO4, giving evidence for the substitution of In 3+ ions on both Cd 2+ and Sn 4+ sites. XPS shows a pronounced decrease in the surface concentration of segregated indium after annealing under UHV conditions. He(I) UPS shows a well defined conduction band feature with a higher intensity than that for tmdoped samples. It is evident that the lower edge of the valence band shifts towards higher binding energy as the doping level increases. Keywords: Cadmium oxides; Ceramics; Electron energy loss spectroscopy; Surface segregation; Tin oxides; UV photoelectron spectroscopy; X-ray photoelectron spectroscopy 1. Introduction The binary post transition metal oxides SnO2, and CdO are extrinsic n-type semiconductors. Both SnO2 and In203 have bandgaps in the near UV (Egap=3.62eV for S n O 2 [ 1 ] ; E g a p = 3.75 eV for In203 [-2]). When doped, these n-type materials show a unique combination of electrical and optical properties: they are electrically conductive, but at the same time they are transparent to visible radiation and reflective to infrared radiation [3]. The infrared reflectivity is dependent upon the carrier concentration: carriers can be introduced by doping the parent compound either by In203 * Corresponding author. E-mail: egdell@vax.ox.ac.uk creating oxygen vacancies or by substituting pentavalent Sb 5+ for Sn4+ ions or quadrivalent Sn4+ for In 3+ ions. The carrier concentration in turn determines the plasmon energy. Maximum plasmon energies of 0.59 eV [-4] and 0.63 eV [5] have been achieved in ceramic samples of Sb-doped SnO 2 and Sn-doped In203, respectively. Recently, we found [6] that the maximum plasmon energy in ceramic CdO doped with trivalent In 3+ ions is 0.72 eV and exceeds that for Sb-doped SnO2 and Sn-doped In203. However, CdO is not useful as a transparent conducting material because of its relatively small bandgap (Eg,p=0.55 eV for CdO at 300K [7]) which results in excessive light absorption from about 0.5#m down to the ultraviolet. 0039-6028/97/$17.00 Copyright © 1997 Elsevier Science B.V. All fights reserved PII S0039-6028 ( 9 6 ) 01141-7 290 I:. Dou. R.G. Egdell/ Surface Science 372 (1997) 289-299 The ternary oxides, Cd2SnO4, CdSnO 3 and CdIn20 4 also have attractive electrical and optical properties. They are also extrinsic semiconductors with relatively high electrical conductivity, and their bandgaps (Egap=2.06 eV for Cd2SnO 4 1,8], Eg~p=2.23 eV for Cdln20 4 [9] and Egap=2.3 eV for CdSnO3 [10]) are bigger than for CdO itself. The carrier concentration in these materials can also be tuned either, by creating oxygen deficiencies or by chemical substitution. These ternary oxides, therefore, stand alongside SnO2 and In20 3 as potential transparent conducting materials and they have received extensive study over the past few years. Dicadmium stannate, Cd2SnO4, is the most intensively investigated material among these ternary oxides 1'11 ]. It has a relatively high carrier mobility and its visible absorption edge moves to much higher energy with n-type doping owing to the exceptionally large Moss-Burnstein shift associated with the low carrier effective mass [8,12,13]. This results in an increasing window of transparency in the U.Vregion and increasing reflectance in the IR region with: increasing electrical conductivity. Polycrystalline Cd2SnO4 adopts an orthorhombic structure of the SI:2PbO 4 type 1,~14], ila which SnO6 octahedra form chains by: sharing edges along the 1-001] direction and these linear chains are held together by CdO 7 polyhedra (Fig. 1), The Sn 4+ cation keeps roughly the same coordination environment as in the tin oxide SnO2. However, the Cd 2+ site geometry represents a deviation from the octahedral symmetry which it holds in cadmium oxide CdO itself. By analogy with the binary systems discussed above, it is to be expected that substitution of In into Cd sites or Sb into Sn sites in Cd2SnO 4 will result in n-type doping. The effect of n-type doping on the electrical properties of Cd2SnO 4 thin films 1'15,16] and thick films 1,17]~ have been extensively studied. There have also been previous studies of antimony-doped Cd2SnO4 (Cd2Snl_~SbxO4) ceramics by NMR, E P R a n d Mrssbauer spectroscopies [18] and by conductivity measurements~ [ 19]. Recently, we reported a detailed study of Sb doping in Cd2SnO 4 ceramics by electron energy loss spectroscopy (EELS) and X-ray and ultraviolet photoelectron spectroscopy (XPS and UPS) [20]. It was shown b ..~ C a • 0 Cd 0 Fig. 1. The structure of Cd2SnO4, showingthe chains of edgesharing SnO 6 octahedra, O (open circles) and Cd ions (solid circles). that EELS combined with UPS and XPS provided direct information on the effect of doping on the surface electronic structure. By analogy with In doping in CdO, substitution of In into Cd sites in Cd2_~In=SnO4 is also expected to produce n-type doping and indeed this has been demonstrated experimentally in thin film material [21]. However, In in Sn sites will function as a compensating acceptor. The present paper reports the application of EELS, XPS and UPS to study changes in electronic structure of CdzSnO 4 produced by In doping. This in turn allows us to appraise themerits of Sb and In as n-type dopants. The analysis of EELS data shows that carrier concentrations introduced by indium doping are always lower than those expected from nominal dopant concentrations. With XPS it is found that Y. Dou, R.G. Egdell/Surface Science 372 (1997)289-299 the atomic percentages of cadmium and tin on the surface are lower than those found experimentally in undoped Cd2SnO4, suggesting that when doped into Cd2SnO4 the In 3+ ions are distributed between Cd 2+ and Sn4+ cation sites. In a + is a donor in the former, whereas it is an acceptor in the latter. It is also been demonstrated that there is a dramatic change in the surface segregation of indium after annealing under UHV conditions. Finally In-doped Cd2SnO 4 is compared with Sb-doped Cd2SnO4 and with In-doped CdO. 2. Experimental Polycrystalline samples of C d 2 S n O 4 containing 0.5, 1.0, 1,5 and 2.0 at% In were prepared by firing a well-blended mixture of the required quantities of Cdl_xInxO (0<x<0.02) and SnO2 at 1050°C for 6 h in air in a recrystallised alumina boat. The product was slowly cooled to room temperature. The samples had a yellow colour with a green shade that deepened slightly as the dopant concentration increased from 0.5 to 2.0 at%. The phase compositions of the products were monitored by X-ray powder diffraction using a Philips movingarm diffractometer. The samples were all single phase with an orthorhombic structure [ 14]. Fig. 2 shows the variation of the orthorhombic unit cell ? E o 176.8 ! Q Tq) 176.4 E 0 > 176.0 l) O ._~ 175,6 0 O 175.2 o 0.0 ~I I I I 0.5 1.0 1.5 2.0 bulk In content / 2.5 otom Fig. 2. Orthorhombic unit cell volumes of indium-doped Cd2SnO4 as a function of bulk doping level. 291 volume with doping level in indium-doped An increase in theunit cell volume with x indicates substitutional incorporation of the In dopant into the C d 2 a n O 4 host lattice. The Cdl_xInxO was itself synthesized by heating a thoroughly blended mixture of stoichiometric amounts of CdO and In203 at 880°C for 7 days in air. Phase pure polycrystalline material with the rocksalt structure of CdO was produced for doping levels up to 2.3 at%, beyond which small peaks due to CdIn204 were observed. The samples were pressed into 10 mm diameter pellets at 5.5 MPa, and then sintered in air at 900°C for 8 h. XPS and UPS were measured in an ESCALAB spectrometer with facilities for excitation of photoelectron spectra using unmonochromated A1 Ka or MgK~ X-rays or UV radiation from a noble-gas discharge lamp. Samples were mounted on Pt stubs and held in position with Pt wires. Sample cleaning was achieved by annealing pellets for 3 h at around 400°C(stub temperature) in the spectrometer preparation chamber (base pressure<10-9mbar) with the aid of a watercooled copper workcoil coupled to a 400kHz radiofrequency generator. After transfer to the main chamber (base pressure 2 x 10-lo mbar) XPS were found to be completely free ,of signals due to carbon or other contamination. The O ls core peak showed a single component with no evidence of high binding energy shoulders due to water or hydroxide contamination. Similarly, He(II) photoemission spectra contained no adsorbate-related peaks on the high binding energy side, of the main O 2p valence band, although a strong Cd shallow core level peak was evident in all spectra. The nominal analyser resolution~ was set at 100 meV for He(I) photoemission measurements, and at 400meV for He(II) photoemission and XPS. Structure due to HeIfi and HeI], lines in He(I) spectra, and HeIIfi and HeIIy~lines in He(II) spectra was removed from spectra using an interactive stripping routine which ensured optimal removal of satellite intensity subject m the constraint that there should be no negative dips in,the stripped spectra. The position of the Fermi energy in the spectrometer was established from measurements on a sample of clean polycrystalline Ag foil. Fermi levels of degenerate conducting stannate C d 2 a n O 4. 292 Y. Dou, R.G. Egdell/ Surfaee Science 372 (1997) 289-299 samples should equalise with those of the Ag and indeed a weak Fermi-Dirac-like onset was found for In-doped Cd2SnO4 samples at the same position as the Ag onset. Binding energies are referred to this Fermi-Dirac onset. X-ray photoelectron spectra were stripped of satellites due to A1K~x3,4 radiation again using in-house software. Energyloss spectra were measured in the same spectrometer using an incident electron beam derived from an electron gun usually used for low-energy electron-diffraction studies. The analyser resolution was set at the low value of 20 meV for EELS measurements, mainly to protect the channeltron from high incident electron fluxes in the elastic peak. The experimental resolution in EELS was, however, limited by the thermal spread of electrons from the electron gun. Typically, the elastic peak displayed the asymmetry to high kinetic energy (negative energy loss) expected for electrons thermionically emitted from a hot filament with a fullwidth at half-maximum height of less than 400 meV. ~ m C I I I I 0 3. Results and discussion 3.1. Electron energy loss spectroscopy Electron energy loss spectra were recorded using a 200 eV exciting electron source. Fig. 3 shows spectra for samples containing 0.5, 1.0 and 2.0 at% In taken after UHV annealing. The strongest peak in each spectrum corresponds to electrons scattered elastically, with no energy loss. The rather intense peak next to the elastic peak is the surface plasmon peak which results from a longitudinal surface excitation of the conduction electron gas and provides valuable information about the concentration of conduction electrons close to the surface. The shoulder to the right of the plasmon peak, which is clearly visible in the expanded scale, is owing to multiple loss processes. It can be seen from Fig. 3 that the plasmon peak moves tohigher energy loss as the doping level increases. Measured plasmon loss energies for the complete series of doped samples are given in Table 1. Clearly, after reaching 0.42 eV at a bulk content of 1.0 at% In, EELS shows only small subsequent increases in plasmon 1 2 loss energy / ,.3 4 eV Fig. 3. EELS spectra of (A) 0.5, (B) 1.0 and (C) 2.0 at% In-doped Cd2SnO 4 excited with a 200 eV electron beam after annealing under UHV conditions. Note the resolved plasmon loss peak in each case. energy with bulk In content. This is owing both to the fact that the carrier concentration falls increasingly below the nominal In doping level as the latter increases, and to the increasing effective mass with increasing occupation of the conduction band (see below). In order to study the effect of annealing under UHV conditions on the plasmon energy we also recorded the EELS of 2 at% In-doped Cd2SnO 4 before annealing. Fig. 4 compares loss spectra before and after annealing. It is seen that the plasmon peak shifts as a result of UHV annealing. The plasmon peak before annealing appears as a shoulder and is estimated to be at a loss energy of around 0.33_+0.05 eV. After UHV annealing the loss energy increases to 0.43 eV. At the low beam energy of the present experiments, surface loss will predominate over bulk loss. 293 Y. Dou, R.G. Egdell/Surfaee Science 372 (1997) 289-299 Table 1 Surface plasmon frequencies, carrier concentrations and effective mass in indium-doped Cd2SnO4 In content (at%) Plasmon frequency (eV) Carrier conc. (10 TM electrons/cm3) Doping level (1018 atoms/cm a) Effective mass ratio (m*/me) 0.5 1.0 1.5 2.0 0.375 0.420 0.430 0.450 31 51 56 68 40 79 119 158 0.068 0.084 0.088 0.092 ! out by Nozik [8] and Miyata et al. [23] e(oe) was found to be 4.0___0.1. Empirically we have found that six values of m* at different carrier concentrations tabulated in the earlier work are best described in the terms of a quadratic variation with the Fermi wavevector k F, i.e. m* varies linearly with n 2/a : ! A m* = m* + an 2/3, - I I I ! 0 I 2 3 loss energy / eV Fig. 4. EELS spectra of 2 at% In-doped Cd2SnO4 (A) before and (B) after annealing under UHV conditions. The surface excitation condition is given by R e [ e ( c o ) ] = - i instead of Re[e(co)]=0 [22] for the bulk excitation condition. The surface plasmon frequency, o)sp is given by: o~2p = ne2/[eo {[e(oo) + 1}re*I, (1) where n is the concentration of conduction electrons, e the electronic charge, eo the permittivity of free space, e(oe) the high-frequency dielectric constant and m* the effective mass of conduction electrons at the Fermi level. From the studies of optical properties of C d 2 S n O 4 thin films carried (2) where m* = 0.02765 me (me being the rest mass) and the constant a is such that a/me = 0.41 x 10-14 cm 2. Combining Eqs. (1) and (2) it is possible to use the measured values of the surface plasmon loss energy to define the carrier concentration. The values obtained from the analysis of energy loss spectra are given in Table 1 and compared with the carrier concentration derived from the bulk In content in the initial mixtures. It can be seen from Table 1 that the carrier concentrations determined from EELS are always lower than bulk doping levels. This could be caused by the compensation of In a + substitution for C d 2 + by cation vacancies. A more plausible explanation is that In a + ions are substitutionally incorporated into both C d 2+ and S n 4+ sites. Each In a+ contributes an electron to the conduction band on the former whereas it accepts an electron on the latter. Substitution of In 3 + ions for both C d 2 + and S n 4 + cations produces a material with a chemical formula C d 2 _ x I n x + y S n l _ y O 4. When x > y there should be net n-type doping, but with a lower carrier concentration than that when only C d 2+ sites are doped as in Cd2_xInxSnO 4. Preparation of samples from C d l , x I n x O a n d S I l O 2 w a s intended to avoid the problem of anti-site substitution and replacement of Sn by In possibly implies the presence of extraneous phases below the limit of X-ray detectability. 294 ]7. Dou, R.G. Egdell/ Surface Science 372 (1997) 289-299 We are unable to establish the m o d e of substitution of In 3 + from consideration of the increase of crystal unit cell volume. This is because replacing either Cd 2+ or Sn 4+ ions with In 3+ could cause an expansion of the cell. For the former it is owing to the occupation of antibonding conduction band states with the free electrons introduced by chemical doping [20], and for the latter it is the consequence of the bigger size of In 3÷ (r =0.94 ,~) relative to Sn 4+ (r=0.83 A) [24]. However, the evidence from XPS favours occupation of both Cd and Sn sites, as discussed below. 3.21 X-rayphotoemission spectroscopy A1 K s excited X-ray photoemission spectra in the region of Cd 3d, In 3d and Sn 3d core levels for In-doped Cd2SnO4 containing 0.5, 1.0 and 2.0 a t % In before annealing are shown in Fig. 5. i i i i i A Cd:3ds/2 Sn:3ds/2 i Cd:3d3/2 . Clearly the intensity of In 3d peaks increases as the bulk In content increases, but is much higher than that expected from the nominal In concentration, indicating that there is a pronounced segregation of In to the surface. It is also noticeable that the surface In concentration reaches a high value after initial doping, but then shows only a very slight increase as the dopant concentration further increases. Fig. 6 shows the corresponding spectra after U H V annealing. Clearly the In 3d intensity is much reduced after the annealing process. To quantify the cation distributions probed by XPS, the areas of Cd 3d5/2 In 3d~/2 and Sn 3d5/2 peaks were divided by atomic sensitivity factors. The resulting cation percentages are plotted in Fig. 7. For undoped material one expects n ( C d ) = 66.7% and n(Sn)= 33.3%: for air-annealed, as pres e n t e d material we found n ( C d ) = 6 6 . 0 % and Cd:3d6/2 Sn:3ds/2 A I Cd:3d:3/:z • Sn:3d3/2 ISn:3d~/2 B B I 390 I I 420 450 I I 480 510 binding energy / eV Fig. 5. A1 Kc~ X-ray photoemission spectra of (A) 0.5, (B) 1.0 and (C) 2.0 at% In-doped Cd2SnO 4 in the Cd 3d, In 3d and Sn 3d region before the vacuum annealing. Structure due to satellite radiation has been subtracted from the spectra. I 390 I | 420 450 I 480 I 510 binding energy / eV Fig. 6. AI K~ X-ray photoemission spectra of (A) 0.5, (B) 1.0 and (C) 2.0 at% In-doped Cd2SnO 4 in the Cd 3d, In 3d and Sn 3d region after vacuum annealing. Structure due to satellite radiation has been subtracted from the spectra. Y. Dou, R.G. Egdell/Surface Science 372 (1997)289-299 80 A nil- In" I~i Cd[ 70 . . . . . . . . . . . . . T____m____s_,_. . . . . . . . . . _] 60 0 40 . . . . . . . . . . . . 30 ~ 20 o ~ "~ "o "~" "- 80 70 60 50 B ~ 40 30 20 10 00.0 0.5 1.0 1.5 2.0 bulk in content / • atom 2.5 Fig. 7. Atomicpercentages of indium,tin and cadmiumon the surface, derived from XPS, for In-doped Cd2SnO4 (A) before and (B) after annealing in vacuum. The data for undoped Cd2SnO4 are representedby dashedlines in (A). n(Sn)=34.0% [21], in good agreement with the "ideal" value. For the 0.5 at% In-doped sample both n(Cd) and n(Sn) are lower than for undoped sample, suggesting that In occupies both Cd and Sn sites although substitution into Cd sites predominates. Additionally, it is clear that there is a dramatic segregation of In to the surface, with n(In)~10%, a factor of 30 higher than expected from the bulk doping level. The segregation of In in this system is highly reminiscent of segregation of Sb in doped SnO 2 [4] and of Sn in In20 a E5]. In these systems the high concentration of surface dopant has been attributed to occupation of cation sites in the topmost ionic layer by dopant ions in the (N--2) oxidation state, N here being the group oxidation state. The (N--2) ions carry an electron lone pair whose energy is lowered by the sp hybridisation allowed a t surface sites where there 295 are non-centro symmetric electric fields. In this model the high concentration of surface dopant atoms is not accompanied by a high concentration of conduction electrons because the ( N - 2 ) ions are not donor centres. This situation certainly pertains in the present system because as we have seen the surface carrier concentration probed by EELS is actually less eventhan the nominal doping level. Thus, the high surface concentration of In found in the present work is attributed to In + in surface Cd sites. Progressive In doping beyond 0.5 at% leads to only small subsequent increases in the surface In concentration suggesting that even at the lowest dopant level the surface is effectively saturated with a monolayer of In + . The effects of UHV annealing revealed in Fig. 6 and quantified in Fig. 7 are highly surprising: the In concentration drops dramatically as compared with the as-presented samples. This behaviour contrasts with that found in In-doped CdO [ 6 ] where the surface In concentration increases after UHV annealing. The decrease in the In concentration in the present study is accompanied by a small but significant decrease in the Cd: Sn ratio. This behaviour is also found in undoped CdzSnO4 and reflects the volatility of CdO. After UHV annealing the outmost ionic layer is therefore depleted of Cd, leaving more Sn cation sites on the surface. As with SnO2 itself, there will be surface reduction from Sn4+ to Snz + after UHV annealing, However, the 5s-5p hybrid state associated with Snz+ must be at lower energy than the corresponding In + state. Thus there is no thermodynamic driving force for accommodation of In + in surface Sn2+ sites and the In intensity in XPS drops dramatically. 3.3. Ultraviolet photoemission spectroscopy Ultraviolet photoelectron spectra excited with He(I)~ (hv =21.22 eV) and He(II)0~ (hv=40~81 eV) radiation were recorded. Fig. 8 shows He(I) spectra for 1 and 2 at% In doped Cd2SnO4 taken after annealing under UHV conditions compared with the spectrum for undoped material [20]. Fig. 9 gives the He(II)spectrum of 2 at% In-doped Cd2SnO 4 after UHV annealing. The He(I) spectra 296 Y. Dou, R. G. Egdell/ Surface Science 372 (1997) 289-299 i i , i , I i i i i i Cd:4d I -10 -5 I 0 5 I I 10 t5 20 binding energy/eV 02p c I( C s ! -I0 -5 Fig. 9. He(II) photoemission spectra of 2.0 at% In-doped Cd2SnO 4 after annealing under UHV conditions Structure due to O 2p valence electrons and shallow core level Cd 4d electrons are labelled. Binding energies are given relative to the Fermi energy of a silver sample stub. Structure due to satellite radiation has been subtracted from tlie spectrum. I I I I I 0 5 10 15 20 25 binding energy / eV Fig. 8. He(I) photoemission spectra of (A) undoped, (B) 1.0 and (C)2.0 at% In-doped CdzSnO4 taken after UHV annealing. Structure due to O 2p valence electrons, shallow core level Cd 4d electrons and secondary electron emission (S) is labelled in (C). Binding energies are given relative to the Fermi energy of a silver sample stub. Structure due to satellite radiation has been subtracted from the spectra. are in each case dominated by the oxygen O 2p valence band, the onset of which varies from about 2.80 to 3.22 eV below the Fermi energy as the In doping level increases. The Cd 4d shallow core level peak is at about 11.0 eV and is superimposed on the secondary electron background which is beyond 15,0 eV. N o major effects of indium doping on the valence band structure can be noted from the H e ( I ) spectra. The H e ( I I ) spectra show a better definition of the Cd 4d core level peak since the secondary electron tail onset is at higher energy than in the H e ( I ) spectrum. The binding energy of the Cd 4d level is 11.7 eV relative to EF or 8.5 eV relative to the valence band edge. The overall O 2p bandwidth is found to be about 8.5 eV. A very weak peak close to the Fermi energy in the H e ( I ) spectra can b e observed on an expanded scale. This peak arises from occupancy of the conduction band. The conduction band must be constructed [18] from both Sn 5s and Cd 5s atomic orbitals which have very low one-electron ionisation cross-sections in comparison with the O 2p states at h v = 2 1 . 2 e V (a(Cd 5 s ) = 3 . 2 5 x 1 0 -2 Mb; a(Sn 5s)=3.25 x 10 -2 Mb; o-(O 2p)=2.67 M b [26]). Coupled with the relatively low doping levels, this accounts for the low intensity of the c o n d u c t i o n band structure. Nevertheless, after removing the Satellites resulting from H e i r (hv = 23.09 eV) and HeI~ (hv =23.75 e V ) r a d i a t i o n from raw spectra the conduction band structure can be observed with adequate signal-to-noise to define the conduction band shape and width. It can be seen from Fig. 8 that the intensity of the conduction band is higher for doped samples than for the undoped sample. In t h e latter, the carriers are produced by oxygen vacancies. The ratios of the m a x i m u m intensities of the conduction bands to those of t h e valence band are given in Table 2 together with that for undoped sample. The values for the doped samples increase slowly with In doping, as expected from a free electron model 297 Y. Dou, R. G. Egdell/ Surface Science 372 (1997)289-299 Table 2 The intensity ratios between conduction and valence bands and the maxima of O 2p valence bands relative to the Fermi energy EF in He(I) UPS for In-doped Cd2SnO4; the data for undoped Cd2SnO4 [20] are also listed for comparison In content (at%) Intensity ratio between conduction and valence band maxima (x 104) Maximum O 2p valence band relative to EF (eV) 0.0 0.5 1.0 1.5 2.0 4.6 8.4 8.7 8.9 9.5 4.30 4.54 4.58 4.67 4.76 where the Fermi level intensity should vary with (cartier concentration) ~/3. This idealised behaviour is, however, modified by the effective mass variation described in Eq. (2). An interesting feature in H e ( I ) U P S is that the m a x i m u m of the O 2p valence band moves gradually towards the higher b i n d i n g energy as the bulk doping level increases. This parallels a shift of the low binding energy edge of the valence band to higher binding energy. The m a x i m u m of the O 2p valence band measured from the H e ( I ) spectra for each of the doped samples is also listed in Table 2, together with the data for u n d o p e d sample for comparison. The shift is analogous to the so called M o s s Burnstein shift in optical spectroscopy, where it is found that a shift of the fundamental optical absorption edge towards shorter wavelengths (higher energy) accompanies increasing carrier density. This p h e n o m e n o n has been observed with thin films of Cd2SnO 4 [8,12,27], C d O [28], In203 [29,30] and In203-doped SnO2 [31] and is discussed in detail in Refs. [2,13,32]. Fig. 10 shows the narrow scan of H e ( I ) spectra of In-doped and undoped Cd2SnO 4. The low binding energy edge of the valence band is at a binding energy of 2.8 eV for undoped CdESnO 4 and 3.2 eV for 2 a t % In-doped Cd2SnO4. The shift of the edge of the valence band is 0.4eV for the doped sample relative to the u n d o p e d one. The shift m a y be attributed to an increase ~ the Fermi energy within the conduction band resulting from the higher cartier concentration. Unfortunately, we can not / AZ B x250 I -2 -1 I I I I I 0 1 2 3 4 binding energy / eV 5 Fig. 10. He(I) photoemission spectra of the narrow scan of (A) undoped and (B) 2 at% In-doped CdzSnO4 after UHV annealing. The Fermi edge and the lower edge of the valence band are indicated by a dashed and black line, respectively,in each case. Binding energies are given relative to the Fermi energy of a silver sample stub. Structure due to satellite radiation has been subtracted from the spectra. compare the two conduction band widths directly because of the difficulty in locating the exact position of the high binding energy edge of the conduction band for the undoped sample, as seen in Fig. 10. This is owing to the relatively lower carrier concentration and a large number of bandgap states which obscure the high binding energy edge of the conduction band. However, we can make a reasonable estimate of the expected conduction band widths using a free-electron model [20,33,34]. In this model the Fermi energy, EF, defined relative to the b o t t o m of the conduction band, equals the conduction band width and can be derived from the effective mass, m*, and the carrier concentration, n, through the following equation: EF = (h2/8rczm *)(3zc2n)2/3, (3) 298 Y. Dou, ~ G. Egdell/Surface Science 372 (1997) 289-299 where h is Planck's constant. For undoped Cd2SnO4, n and m* can be obtained by assuming that the plasmon energy is not more than 0.3 eV [-20]. The conduction band width determined in this way is 0.39 eV for undoped CdzSnO 4 and 0.64 eV for 2 at% In-doped Cd2SnO4, giving an approximate shift of the valence band edge of 0.25 eV. The discrepancy between this estimate and the observed shift of 0.4 eV is probably due to the uncertainty of the plasmon energy and the over simplified model for the conduction band. Actually, the conduction band shape does not conform exactly to a simple expression such as that above owing to non-parabolicity and the variation of m* within the band. interstitial Cd. It thus seems probable that the change in plasmon energy for In-doped Cd2SnO4 again reflects the distribution of In between Cd and Sn sites. Substitution of In into Sn sites is basically producing p-type doping (which compensates n-type substitution for Cd), but p-type doping of oxides requires highly oxidising conditions and therefore annealing in UHV should favour a migration of In to Cd sites. To summarise, In-doped Cd2SnO4 is a complex system which shows qualitatively different behaviour to both Sb-doped Cd2SnO 4 and In-doped CdO. References 4. Concluding remarks The carrier concentration determined from EELS for In-doped CdzSnO 4 is always lower than bulk doping level. This situation is different from that in Sb-doped Cd2SnO 4 [-20], where it is found that the carrier concentration is always slightly greater than the bulk dopant concentration. The result from our own study indicates [6] that no significant compensation of electrons by Cd z+ cation vacancies occurs in indium-doped CdO. This leads us to consider that In 3+ ions are substituted for both Cd 2+ and Sn4+ ions. XPS demonstrates that the atomic percentages of both cadmium and tin are lower in In-doped Cd2SnO 4 than in undoped Cd2SnO 4. This gives the evidence for the above consideration, although the data relate to an as-presented sample. From Fig. 4 and the discussion in Section 3.1 it is known that high temperature vacuum annealing increases the plasmon energy of 2.0 at% In-doped Cd2SnO 4 by 0.12 eV, which is consistent with an increase in the carrier concentration of 5.0 x 1019 atoms crn -3. These carriers are unlikely to be introduced by oxygen vacancies produced at high temperature under vacuum because EELS data for Sb-doped CdzSnO4 show no difference between plasmon energies for samples before and after annealing. In fact in In-doped CdO the surface plasmon energies show a small decrease as a result of UHV annealing, apparently due to loss of donor [1] V.T. Agekyan, Phys. Status Solidi A 4 (1977) 11. [2] I. Hamberg, C.G. Granquist, K.F. Bergren, B.E. Sernelius and L. Engstrom, Phys. Rev. B 30 (1984) 3240. [3] H. K6stlin, Fest6rpeprobleme XXII (1982) 229. [4] R.G. Egdell, W.R. Flavell and P.J. Tavener, J. Solid State Chem. 51 (1984) 345. [5] P.A. Cox, W.R. Flavell and R.G. Egdell, J. Solid State Chem. 68 (1987) 340. [6] T. Fishlock, Part II Thesis, University of Oxford, UK, 1995. [7] This gap is indirect. See Semiconductors, Eds. O. Madelung, M. Schulz and H. Weiss, Landor B6rnstein Numerical Data and Functional Relationships in Science and Technology, New Series III/17b (Springer, Berlin, 1982). [8] A.J. Nozik, Phys. Rev. B 6 (1972) 453. [9] F.P. Koffyberg and F.A. Benko, Appl. Phys. Lett. 37 (1980) 320. [10] F. Golestani-Fard, C.A. Hogarth and D.N. Waters, J. Mater. Sci. Lett. 2 (1983) 505. [11] C.M. Cardile, Rev. Solid State Sci. 5 (1991) 31. [12] N. Miyata, K. Miyake, K. Koga and T. Fukushima, J. Electrochem. Soc. 127 (1980) 918. [13] M.S. Setty, J. Mater. Sci. Lett. 6 (1987) 909. [14] M. TrSmel, Z. Anorg. Allg. Chem. 371 (1969) 237. [15] G. Haacke, H. Ando and W.E. Mealmaker, J. Electrochem. Soc. 124 (1977) 1923. [16] G. Haacke, W.E. Mealmaker and L.A. Siegel, Thin Solid Films 55 (1978) 67. [17] M.S. Setty and A.P.B. Sinha, Thin Solid Films 144 (1986) 7. [18] K.J.D. MacKenzie, C.M. Cardile and R.H. Meinhold, J. Phys. Chem. Solids 52 (1991) 969. [19] R.D. Shannon, J.L. Gillson and R.G. Bouchard, J. Phys. Chem. Solids 38 (1977) 877. [20] Y. Dou and R.G. Egdell, J. Mater. Chem. 6 (1996) 1369. Y. Dou, 1LG. Egdell/Surface Science 372 (1997) 289-299 [21] N. Miyata, K. Miyake, H. Nakaoka and Y. Digaku, Kogakuin Diagaku Kenkyu Hokoku 28 (1978) 235. [22] If the loss intensity were dominated by a bulk loss process, the value of n/m* deduced from the EELS data would be decreased by a factor ~/e(oo)/[¢(oo)+l]= x/(4/5) = 0.89. [23] N. Miyata, K. Miyaka, K. Koga and T. Fukushima, J. Electrochem. Soc. 127 (1980) 918. [24] R.D. Shannon, Acta Crystallogr. Sect. A 32 (1976) 751. [25] D. Briggs and M.P. Seah, Eds., Practical Surface Analysis (Wiley, Chichester, 1983). [26] J.J. Yeh and I. Lindau, At. Data Nucl. Data Tables 32 (1985) 1. 299 [27] E. Leja, K. Budzynska, T. Pisarkiewics and T. Stapinski, Thin Solid Films 100 (1983) 203. [28] H. Finkenrath, Z. Phys. 159 (1960) 112. [29] K.L. Chopra, S. Major and D.K. Pandya, Thin Solid Films 102 (1983) 1. [30] Z.M. Jarzebski, Phys. Status Solidi A 71 (1982) 13. [31] G. Haacke, Ann. Rev. Mater. Sci. 7 (1977) 73. [32] G. Sanon, R. Rup and A. Mansingh, Phys. Rev. B 44 (1991) 5672. [33] R.G. Egdell, in: Science of Ceramics Interfaces II, Ed. J. Nowotny (Elsevier, Amsterdam, 1994) p. 527. [34] P.A. Cox, R.G. Egdell, C. Harding, W.R. Patterson and P. Tavener, Surf. Sci. 123 (1982) 179.