CIS GBs JC talk
•Download as PPTX, PDF•
0 likes•53 views
Research group journal club talk.
Recommended
Recommended
More Related Content
More from Suzanne Wallace
More from Suzanne Wallace (15)
Recently uploaded
Recently uploaded (20)
CIS GBs JC talk
- 1. WMD JOURNAL CLUB: 3RD FEB 2015 Hole barriers and reduced recombination at grain boundaries in polycrystalline chalcopyrites
- 2. • Supercell: 40 56 atoms • LDA LDA+U • Inhibit artificial charge transfer between top and bottom surface slab using pseudo-hydrogen atoms • Experimental confirmation
- 3. Polycrystalline CuInSe2 solar cells Single-crystalline CuInSe2 ~13% Polycrystalline CuInSe2 ~20% • Polycrystalline devices usually have lower efficiencies than corresponding single-crystalline devices • e.g Si or GaAs devices • Poor carrier transport in in polycrystalline materials • Interfaces between grains are sinks for defects • GB defects form recombination centres for optically generated e- and h+
- 4. Anomalous grain boundary physics • Cation-terminated CIS and CGS GB’s had a VBM offset w.r.t to grain boundary • Repels holes at GB • Deprives e- from recombining at GB defects GIs possibly closer to being perfect than in single- crystal
- 6. lower VBM (repels h+) cation-terminated GB anion-terminated GB grain interior CBM offset VBM offset larger VBM offset CIS CGS
- 7. Formation of VCu • Analogy between structure of GB ‘internal surfaces’ and surface structure of CIS films • Polar surface is more stable than the non-polar surface • Unlike in conventional semiconductors (e.g. GaAs) • Polar surface reconstructs to remove dipole • Reconstruction creates rows of Vcu in cation-terminated GB • Charge neutral defects
- 8. Lowering of the valence band maximum (VBM) • Interface between Cu-poor GB and closely Cu stoichiometric GI • Cu d-orbitals repel Se p-based VBM upwards Lack of Cu at GB results in less repulsion of VBM here
- 9. Lowering of the valence band maximum (VBM) CIS CGS grain interior grain boundary
- 10. Experimental confirmation of model • Micro-Auger electron spectroscopy Up to 50% Cu deficiency at CIS GB • Pump-power dependent cathodluminescence Strongly reduced recombination at GB Limited supply of one charge carrier at GB • Scanning tunneling microscopy Decrease in photon emission intensity at GB relative to GI Demonstrates reduced hole density at GB • 2D device simulations Strongly reduced recombination at GB Model charge-neutral hole reflector at GB
- 11. Key conclusions 1. Anion-terminated GBs in CIS and CGS have negligible VB offsets 2. Cation-terminated GBs have a VB offset larger in CGS than in CIS 3. Cation-terminated CGS GB has a CBM offset relative to the GI Attracts e- to the GB Also reduces recombination
- 12. Key conclusions (cont.) 4. Poor solar cell performance of Ga rich (>30% Ga) CIGS is not due to less hole repulsion at the GB CBM offset and smaller Δg in CGS Fermi level pinning more detrimental to voltage Tunnelling-assisted recombination possible 4. CIS and CGS have different electrostatic barriers Shown by kelvin-probe studies Explained by calculation of lower formation enthalpy of III-on-Cu antisite double donor defect in CIS than CGS
- 13. Final comments • Phenomena is material specific Unlikely to be applicable for processing other high efficiency polycrystalline devices • Nice example of modeling to answer questions posed from experimental observation • Generally interesting physical phenomena!
Editor's Notes
- The topic of the paper I’m presenting on is possibly a bit dangerously close to my research area but this GB-enhanced carrier collection phenomena in CIGS cells seemed pretty exciting and I thought journal club would be a good excuse to look into it! It’s a pretty short paper but they seem to have crammed a lot of interesting results into it.
- So the lower paper here I’m presenting on ‘….title…’, builds on the work done 2 years before by what more or less looks like repeating the calculations for CaGaSe2 GB as well as a CuInSe2 GB and performing calculations more accurately with a larger supercell, using LDA+U instead of just LDA and inhibiting the artificial charge transfer between the top and the bottom surface slab by using pseudo-hydrogen atoms with fractional charges and allowing dipole-compensating relaxation. It also includes some more recent complementary experimental work.
- So the motivation for the original study is that polycrystalline CuInSe2 solar cells break the usual rule of the low-cost polycrystalline cell outperforming the more expensive single-crystalline versions; as is the case for Si or GaAs devices. Polycrystalline devices usually have lower efficiencies than corresponding single-crystal devices because the materials contain a large number of grain boundaries which are sinks for defects which usually form recombination centres for optically-generated electrons and holes.
- But instead, this work showed that cation-terminated GBs of CuInSe2 and CaGaSe2 had a valence band offset w.r.t to the grain boundary interior which repels holes from the GB. This then deprives electrons from recombining at the GB defects. An additional benefit here of the polycrystalline device is that most defects and impurities migrate to the GBs during growth which could leave the grain interior even closer to being perfect than when attempting to grow perfect single-crystal materials.
- So this figure shows the main results from the paper with both the cation- and anion-terminated grain boundaries in CIS and CGS included in the figure. I thought I should include a plain version of this figure since I scrawled over it quite a lot in the next slide.
- Looking first at the CIS GBs, the offsets at the anion-terminated GB are negligible and so is the CBM offset at the cation-terminated GB but there’s a significant VBM offset to repel electrons from the GB. Now looking at the CGS GBs, the offsets at the anion-terminated GB are also negligible but the cation-terminated GB has both a CBM and VBM offset relative to the GI. But also, notice that the VBM offset in the CGS GB is larger than that in the CIS GB.
- The reason for the lowering on the VBM is related to Cu vacancies at the GB. In the paper they use the analogy between the structure of GB ‘internal surfaces’ and the surface structure of CIS films. Total-energy minimization of the surface structure of CIS showed that unlike in conventional semiconductors, such as GaAs, the polar surface is more stable than the non-polar surface. The polar surface must reconstruct to remove the dipole which in the case of the cation-terminated surface results in the creation of rows of Cu vacancies and this reconstruction costs little energy Cu’s in the sublattice are weakly bonded. These surface Cu vacancies however are charge neutral because they have been used to cancel the electrostatic dipole.
- Due to the Cu vacancies the interface between the GB and GI represents an interface between two materials of different chemical compositions – one strongly Cu poor and one closely Cu stoichiometric. This leads to a lowering of the VBM at the Cu poor GB. This is because the Cu d-orbitals repel the Se p-based VBM upwards, and so without them there is a lowering on the VBM.
- Due to the Cu vacancies the interface between the GB and GI represents an interface between two materials of different chemical compositions – one strongly Cu poor and one closely Cu stoichiometric. This leads to a lowering of the VBM at the Cu poor GB. This is because the Cu d-orbitals repel the Se p-based VBM upwards, and so without them there is a lowering on the VBM. Here I’ve just cropped out the anion-terminated DOS since they aren’t very interesting to show the difference between the GI and GB for the cation-terminated GBs more clearly.
- The model presented in the paper to explain the anomalous behaviour of GBs in polycrystalline CIGS devices was a new concept in the 2003 paper but has since been studied in various ways experimentally and by device simulations. I followed up some of the experimental papers to look into how the top two techniques work but I’m still a little unsure! Micro-Auger spectroscopy measurements found that there was a large deficiency of Cu at the CIS GB as predicted by the polar surface reconstruction model. (Uses characteristic emissions due to the Auger effect to gauge composition?) Pump-power dependent cathodluminescence showed a strongly recombination at the GB and indicated that there one a limited supply of one type of charge carrier there. (uses electron beam to excite electrons at GB but because holes are repelled recombination and subsequent photon emission cannot occur?) STM scans at low voltage showed a decrease in photon emission intensity at the GB compared to the GI which demonstrates a reduced hole density there. And lastly, 2D device simulations of the model of the neutral band offset at the GB/ GI interface indicated a strongly reduced recombination at the GB.
- So firstly, as I’ve already mentioned the anion-terminated GBs in this study are a bit boring and don’t have any VB offset so don’t repel holes and are therefore prone to recombination so preferably these types of GBs should be minimized during growth for solar cell materials. Cation-terminated GBs have a VB offset and this is larger in CGS than CIS. This was found to be due to a larger p-d repulsion because the Cu-Se bond length is shorter in CGS than CIS. There is no CBM offset for the CIS GBs but there is for the cation-terminated CGS GB so this will attract electrons to the GB which actually also reduced recombination because it implies that photo-generated electron-hole pairs will dissociate at the interface (like in organic solar cells) with the electron being attracted to the GB and the hole to the GI.
- 4) Alloying CIS with the wider gap CGS seems advantageous to increase VOC but it has been shown than Ga rich CIGS solar cells with more than 30% GA have a lesser performance. This isn’t due to less hole repulsion at GBs in CGS because we’ve already discussed that this effect is larger than in CIS. It instead relates to the fact that only CGS has a CBM offset and so has a smaller value for delta g: Firstly, the maximum attainable voltage can be reduced more in CGS than CIS by Fermi level pinning caused by the formation of Vcu’’s even though in both cases the pinning is Ev+0.8 eV. I think this is because of the change in the position of the CBM in CGS, so the pinning could result in a larger decrease in voltage across the whole device. Secondly the smaller delta g between the GI and GB in the CGS means that at abrupt GI/GB interfaces tunnelling-assisted electron-hole recombination can occur at these GBs. 5) This study found that although CGS has a larger neutral VB offset than CIS but CIS has a larger electrostatic offset. Kelvin-probe studies have shown that there is a higher surface potential on the CIS GB (meaning a larger downward movement of the VBM at the GB) indicating that there are charges there. Calculations explain this from the lower formation enthalpy of of the III-on-Cu antisite double donor in CIS than CGS.
- So this advantageous effect is clearly pretty unique to the material and not a general way to improve the efficiency of all PV devices but I thought it was pretty exciting when I first came across this phenomena making cheaper versions of the device actually more efficient! But also it’s a really nice example of computational modelling thoroughly answering questions from experimental observations and the hole repulsion at GBs reducing recombination just seemed like a generally pretty cool phenomena!