Andriy Zakutayev November 2009 Photoemission spectroscopy study of SnZrCh3 [Ch=S, Se] and related materials Andriy Zakutayev November 2009
Experimental details SnZrSe3, SnSe, ZrSe2 - prepared by Annette Richard (add preparation details) SnZrS3 - prepared by Daniel Harada (add details) XPS measured by Andriy Zakutayev (August 2009, TU Darmstadt) Stored in food-grade vacuum and polished with sand paper before XPS (Escalab 250, monochromated Al anode x-rays, calibrated with clean Ag sample) ZrSe2 pellets degassed strongly in 10-9 mbar UHV
SnZrSe3 doping (2% and 5%): Sb on Sn, Bi on Zr site All the peaks in the survey spectra are due to either photoemission or Auger electrons from Sn, Zr, Se, O, Bi, and Sb . There is no Si or C after polishing with SiC sand paper
Core level spectra of the elements Sn 3d Sn-O Sn-Se Zr 3p Zr-Se Degree of the oxidation of Sn is random – the difference most likely comes from not-reproducibility of surface polishing. Zr is oxidized similarly for all the samples – it is likely that this indicates oxidation of Zr in the grain boundaries Se peaks ratios is not consistent with d-nature of the states. This means there is a large Se-O component in the signal O1s signal also has 2 peaks – most likely due to difference in (1) C-O and (2) Zr-O, Se-O, Sn-O bonds Position of the peaks does not change as a function of doping %. This means that the EF-EVBM is fixed Zr-O Sb 5% Sb 5% Sb 2% Sb 2% Bi 5% Bi 5% Bi 2% Bi 2% SnZrSe3 SnZrSe3 Se 3d Sn-Se and Zr-Se O 1s O-C and O-Zr/Se/Sn Se-O Sb 5% Sb 5% Sb 2% Sb 2% Bi 5% Bi 5% Bi 2% Bi 2% SnZrSe3 SnZrSe3
Other core level spectra that interfere Zr 4p and Sn 4d Zr 3d and Se LMM The most intense Zr 3d (179 and 181 eV) peak coincides with Se LMM Auger line (181 eV + weaker satellites at lower BE) Zr 4p (28eV) coincides with Sn 4d (25 eV)– apparent ratios of Sn 4d intensities is reverse because of it SnZrSe3 SnZrSe3 C 1s and Se LMM There is not much C around – broad peak is from Se LMM Auger line (C 1s should be sharper) Polishing with sand paper removes most of C-O atmospheric contamination, but not all. It is also obvious form useless UPS spectra (few slides down) SnZrSe3
Core level spectra of dopands Both Bi and Sb most intense peaks (arrows) co-inside with peaks of the other elements (upper panel), but I subtracted the backgrounds (lower panel). Signals of Bi and Sb are obvious for the 5% doping less clear for 2% doping Sb 3d and O 1s Sb doping Bi 4f and Se 3p Bi doping Sb 5% Bi 5% Sb 2% Bi 2% SnZrSe3 SnZrSe3 Sb 3d Bi 4f Sb doping Bi doping Sb 5% Bi 5% Sb 2% Bi 2%
Composition The atomic % of the elements is obtained from the integrated area under the photoelectron peaks from XPS. The accuracy of this method is 2-3 at%. Table 2 is just different interpretations of the data (Table1) In all samples Se/(Zr+Bi)=3.0…3.5, all the samples are Se-rich Sn/Zr fluctuates. Is it related to Sn low melting T? Need to be controlled better if we want reliable transport properties Sb-doped samples and 2%Bi sample have Sn/Zr=1. The two other samples are Sn-rich (Sn/Zr=1.5) Bi/Zr and Sb/Sn are consistent with 2% and 5% doping of the samples. O/Zr=2 in most samples. It is likely that ZrO2 is present at the surfaces (compare to Ba/O=1 in BaCuSeF where BaO might be present) Anion/Cation=2-2.5 (should be 1.5 in SnZrSe3). The surfaces are oxidized. atomic % Sn+Sb Zr+Bi Se O Bi Sb SnZrSe3 0.1963 0.1237 0.4405 0.2396 0.0000 2%Bi 0.1316 0.1621 0.4714 0.2349 0.0047 5%Bi 0.1775 0.1363 0.4532 0.2330 0.0064 2%Sb 0.1389 0.1402 0.4332 0.2878 0.0037 5%Sb 0.1523 0.1300 0.4268 0.2909 0.0078 Ratios (Sn+Sb)/(Zr+Bi) Anion/Cation Se/ (Zr+Bi) O/ Bi/ Sb/ (Sn+Sb) 1.5877 2.1253 3.5622 1.9373 0.8119 2.4051 2.9084 1.4494 0.0291 1.3028 2.1866 3.3257 1.7096 0.0469 0.9908 2.5840 3.0910 2.0531 0.0264 1.1710 2.5427 3.2828 2.2375 0.0511
Valence band spectra and band bending Sb 5% EF-EVBM is the same for all doping levels – the Fermi level at the surface is pinned at 0.65 eV above VB – almost in the middle of the gap. This agrees with no shift in core levels. VBM shape compares qualitatively well with the DFT predictions for DOS in SnZrS3. Calculate DOS for SnZrSe3 Assume that the Bi and Sb substitute for Sn or Zr and cause EF in the bulk to shift towards CBM (this might not be the case if the Fermi level is pinned in the bulk by the defects) Under these assumptions, the bands of SnZrSe3 must bend up towards the surface (see figures below) Sb 2% Bi 5% Bi 2% SnZrSe3 Surface Surface Surface 2% Doped Bulk 5% Doped Bulk Not doped Bulk 0.65 eV 0.65 eV 0.65 eV 0.7 eV 0.7 eV 0.7 eV
UPS VB spectra and secondary electron edge Totally useless UPS is much more surface sensitive compared to XPS, because the kinetic energy of the escaping electrons is lower The 1/x background slope in UPS spectra at high BE – secondary electrons The distance between the secondary electron edge (16.7) and HeI energy (21.2 eV) is a work function WF is 4.5 eV for all samples – characteristic for organic atmospheric contaminants The EF-EVBM is 1.0-1.2 eV – this is also quite typical for organic molecules Two broad peaks in the VB UPS spectra correspond to HOMO and some deeper state in organic atmospheric contaminants Full UPS spectra WF=4.5 eV Sb 5% Bi 5% Bi 2% SnZrSe3 VB UPS spectra 1.0-1.2 eV Sb 5% Bi 5% Bi 2% SnZrSe3
Comparison of SnZrSe3 with SnSe and ZrSe2 at % Sn Zr Se O SnSe 0.4706 0.0000 0.3750 0.1544 ZrSe2 0.2174 0.3925 0.3902 ratios Se/M (Se+O)/M Se/O O/M 0.7970 1.1251 2.4288 0.3281 1.8058 3.6008 1.0060 1.7951 ZrSe2: Surface are slightly Se deficient and strongly oxidized Se/Zr=1.8; O/Zr=1.8; Se/O=1 SnSe: Surfaces are slightly Se deficient and slightly oxidized Overall surfaces are anion-rich – oxidation Very many Se Auger lines in the spectra… bad element
Delete bad datapoints LINE O AT% Zr AT% Se AT% Sn AT% TOTAL Average 19.602 -0.01895 37.22513 43.117 94.6173 STDEV 14.94924 0.031107 11.07834 4.595691 7.861138 LINE O AT% Zr AT% Se AT% Sn AT% TOTAL Average 10.57689 29.41728 59.93863 0.000214 91.24688 STDEV 2.819077 2.041569 3.543106 0.012913 2.572845 LINE O AT% Zr AT% Se AT% Sn AT% Average 18.12335 25.935 55.87345 0.000992 STDEV 8.61367 5.546111 13.59622 0.008969 LINE O AT% Zr AT% Se AT% Sn AT% Average 16.58395 17.71895 55.08005 10.51583 STDEV 6.522282 4.192083 21.72282 11.05346
Comparison of core peaks of SnZrSe3, SnSe and ZrSe2 Sn 3d Zr 3p Degree of oxidation of Sn is surface-related and not conclusive based on SnZrSe3 results Zr oxidized the same in ZrSe2 and SnZrSe3 Se oxidizes more in ZrSe2 Amount of oxygen is ZrSe2>SnZrSe3>Se. It correlates with the reactivity of the elements with O. Also it correlates with the nominal anion/cation ratio: 2>1.5>1 Sn-O Zr-O ZrSe2 SnSe SnZrSe3 SnZrSe3 Se 3d SnSe O 1s Se-O SnSe ZrSe2 ZrSe2 SnZrSe3 SnZrSe3
Valence band spectra of SnZrSe3, SnSe and ZrSe2 CB 1 eV SnSe EF 0.36 eV VB (c) CB ZrSe2 EF 1.2 eV 0.9 eV SnZrSe3 VB (d) CB 1 eV..? EF 0.64 eV VB EF-VBM is different in 3 materials: Based on the known band gaps, the surfaces are: for SnSe is p-type; for ZrSe2 is n-type (less n-type than expected though); for SnZrSe3 EF is in the middle of the gap. Might be nice for i-absorber in pin cell if the bulk Fermi level is the same d-states in ZrSe2 VB are obvious. Need DFT DOS to assign the features of the VB – calculate it.
SnZrS3 grinded in air and in a glove-bag All the peaks in the survey spectra are due to either photoemission or Auger electrons from Sn, Zr, S, O and C. There is no Si after polishing with SiC sand paper Related thoughts: Sn, Zr and Te have similar atomic weight. Might be easy to make SnZrTe3 films by sputtering
SnZrS3 surface composition At.% O 1s Zr 3p3 Sn 3d5 S 2p Ar 0.19 0.17 0.18 0.46 air 0.21 0.16 0.43 Ratios An./Cat. O/Sn Sn/Zr S/Zr 1.86 1.08 1.07 2.70 1.79 1.19 2.63 The atomic % of the elements is obtained from the integrated area under the photoelectron peaks from XPS. The accuracy of this method is 2-3 at%. Table 2 is just different interpretations of the data (Table1) Sn/Zr=1 – nice! What is the difference in synthesis procedure compared to SnZrSe3? S/O=2, just like Se/O in SnZrSe3; S/Zr<3 and ZrS2 impurity is present – if add extra S in synthesis, will get phase-pure SnZrS3. Anion/cation>1.5 – oxidized surface. O/Sn=1 – SnO on the surface? That would be nice to have…
SnZrS3 spectra of core levels Sn 3d Zr 3p Sn is less oxidized than in SnZrSe3 Zr is oxidized the same way as in SnZrSe3 Oxidation shoulders of Sn and Se peaks are the same for the glove bag and air samples – because of the same surface preparation with sand-paper There is less oxygen signal for the glove bag sample, hence higher oxygen 1s signal should come from the grinding in the air Grind in glovebags!!!! Zr-O Sn-O air air bag bag O 1s O-C and O-Zr/S/Sn S-O..? air air bag bag
SnZrS3 valence band spectra air bag EF-EVBM is the same for both samples – EF is 1.2 eV above VB – closer to the CB. Why is it p-type than according to Seebeck? EF-EVBM=1.2 eV is much larger compared to 0.65 eV in SnZrSe3 Fixed EF agrees with no shift in core levels. VBM shape agrees well with the DFT predictions for DOS in SnZrS3. Calculate DOS deeper. Assume that SnZrS3 is p-type in bulk (if Seebeck measurements are bulk-sensitive). Under these assumptions, the bands of SnZrS3 must bend down towards the surface. (b) Bulk Surface 1.4 eV 1.2 eV
SnZrS3 EPMA - AR Delete bad datapoints LINE O AT% Zr AT% Sn AT% S AT% TOTAL Average 1.658943 20.06218 21.61608 56.63195 83.17313 STDEV 1.980805 0.646271 0.308424 2.10158 1.736416