Chlorine activated stacking fault removal mechanism in thin film CdTe solar cells: the missing piece

The conversion efficiency of as-deposited, CdTe solar cells is poor and typically less than 5%. A CdCl2 activation treatment increases this to up to 22%. Studies have shown that stacking faults (SFs) are removed and the grain boundaries (GBs) are decorated with chlorine. Thus, SF removal and device efficiency are strongly correlated but whether this is direct or indirect has not been established. Here we explain the passivation responsible for the increase in efficiency but also crucially elucidate the associated SF removal mechanism. The effect of chlorine on a model system containing a SF and two GBs is investigated using density functional theory. The proposed SF removal mechanisms are feasible at the 400 ∘C treatment temperature. It is concluded that the efficiency increase is due to electronic effects in the GBs while SF removal is a by-product of the saturation of the GB with chlorine but is a key signal that sufficient chlorine is present for passivation to occur.

Previous studies have already shown that Cl accumulation at the Sigma3 (112) GB can help remove deep defect levels and the common stacking fault is not harmful to the system. This is somewhat confirmed in this paper and not much is new. So, the Cl-induced removal of unharmful stacking fault is not important in this field, and they are also other mechanism that Cl can help remove stacking fault with or without GBs.
There are also some misleading and unclear statements in the paper: What is the position of the Cl, how many are on interstitial sites or substitutional sites? How are they determined. It's not clear by just looking at Figure 8.
For computational easiness, both Te-Core and Te-Core are included in the small unit cell. What happens if only Te core GBs exist? What's the effect of interaction between Cd-core and Te-core GBs?
What are the calculated electronics states, where is the defect levels introduced by the GBs AND Cl?
Please plot a DOS of the clean Sigma3 (112) grain boundary structure without a stacking fault layer. Is the existence of stacking fault important?
In Figure 9b, there are still many mid-gap defect states and they claim it is the result of Te `noise' throughout the structure related to the single atomic layer thickness necessary for the computationally expensive NEB calculations. What is their evidence for this statement and what do they mean 'noise'? Please also see [Phys. Rev. Lett. 101, 155501 (2008)], which show Cl cannot fully passivate the GBs.
They used the PBEsol functional for their calculations which underestimate the band gap. They said "A study of chlorine's effects on grain boundaries at the hybrid DFT level of theory is required to confirm and improve understanding of passivation mechanisms". But it is not clear whether they did it or not.
Based on the above observation, I think the paper should be rejected by Nature Communication.

Response to Reviewers
The authors would like to thank all three reviewers for their positive comments. We have addressed their individual points of concern below. Reviewer #1: This is a very satisfying wrap-up of an ongoing issue concerning the stacking faults that are so common in these solar cell materials. The authors have done a very thorough job of observing and calculating the energetics off the chlorine uptake and the stacking fault removal, nicely disentangling the mechanism. They conclude that the stacking fault removal is essentially a by-product of the chlorination and the improved properties result from the chlorine at the grain boundaries. This adds significantly to our understanding, resolving the issues, and does essentially tie up the whole story, and so I think it deserves publication.
• The only comment I would like to give is that the authors might consider citing another paper by Chen Li where the movie and image is better at showing the annihilation process and they also calculate that the energy cost is low. It is C. Li, Y. Y. Zhang, T. J. Pennycook, Y. Wu, A. R. Lupini, N. Paudel, S. T. Pantelides, Y. Yan, and S. J. Pennycook, "Column-by-column observation of dislocation motion in CdTe: Dynamic scanning transmission electron microscopy," Applied Physics Letters, 109, 143107-6 (2016).

Response:
The authors agree that this reference is suited to the discussion of stacking fault annihilation in the grain interior and have added this reference as [19].

Reviewer #2:
The paper is important for CdTe PV technology.
Recommend publication after major revision.
1. The phrase "whole story" in the title is misleading. The authors have looked at sigma3 grain boundary one of the simplest ones and no comments about what may happen with more complex boundaries.

Response:
We have now changed the title to "the missing piece" instead. We focussed on the sigma 3 boundary due to the complex nature of the calculations. Since the mechanism for stacking fault removal is very structural, we would expect a similar mechanism to remove the stacking faults as long as the structural conditions were met. A note to emphasise this has been added to the discussion, highlighted in blue.
Response: This is a good question and we thank the reviewer for bringing it to our attention. In CdTe deposited via close space sublimation, stacking faults are also removed during CdCl2 treatments which do not undergo significant recrystallization. This implies that there is some recrystallisationindependent mechanism at play. A discussion to clarify this has been added to the paper which reads "The chlorine activation process causes a number of changes to the microstructure and improves the electrical performance. However, recrystallisation is minimal and little grain growth is observed following CdCl2  Reviewer #3: The authors studied Cl passivation at the CdTe GBs and its relation to the stacking fault. They find that Cl doping at the GB helps the removal of staking fault.
Previous studies have already shown that Cl accumulation at the Sigma3 (112) GB can help remove deep defect levels and the common stacking fault is not harmful to the system. This is somewhat confirmed in this paper and not much is new. So, the Cl-induced removal of unharmful stacking fault is not important in this field, and they are also other mechanism that Cl can help remove stacking fault with or without GBs.

Response: We are fully aware that the most common stacking faults are not electrically harmful as reported previously by ourselves and others ref {14} in the paper. Whereas the stacking faults themselves may not be electrically harmful, the correlation between stacking faults removal and improved cell efficiency has never previously been explained and is important to understand since it is a by product of the chlorine treatment.
There are also some misleading and unclear statements in the paper: • What is the position of the Cl, how many are on interstitial sites or substitutional sites? How are they determined. It's not clear by just looking at Figure 8.

Response: The authors agree that the nature of these Cl atoms is an important concept so have added a few passages to indicate how the Cl was initially inserted and to point out the nature of some of the key Cl atoms in the stacking fault removal process. Unfortunately, due to the high level of disordering during relaxation of these boundaries it is impossible to classify which Cl are interstitials and which are substitutions as everything is completely mixed.
Text has been added to explain how the Cl were initially inserted "The structure in figure 8a) is identical to figure 5a) but Cl has been added in to interstitial and substitutional sites which have previously been found to be important sites for defect passivation [8]. This resulted in two interstitial Cl atoms in the Cd-core and one substitutional Cl in the Te core. A further 11 Cl atoms were placed initially interstitially in either the Cd-or Te-core. Figure 8 shows that after relaxation, some Te atoms in the grain boundary have been pushed to interstitial sites" • For computational easiness, both Te-Core and Te-Core are included in the small unit cell.
What happens if only Te core GBs exist? What's the effect of interaction between Cd-core and Te-core GBs?
Response: Due to the required periodicity of the cell, single core models would not be possible in the study of electronic defects since two oppositely polarised Sigma 3 cores are required to generate periodicity.
The length of the grain boundary simulation cell was chosen to simulate a bulk-like environment in the centre of the cell which was confirmed using the defect formation energy of a Cl defect in this region. This comment has been added to the paper and reads "The system length of 62.08 Å was methodically increased until the defect formation energy of a chlorine defect plateaued in the centre to give to give a bulk-like environment." In terms of the effect of the interaction between the two cores we believe this will be negligible in this study compared with the action of Cl • They used the PBEsol functional for their calculations which underestimate the band gap. They said "A study of chlorine's effects on grain boundaries at the hybrid DFT level of theory is required to confirm and improve understanding of passivation mechanisms". But it is not clear whether they did it or not.

Response:
To clarify further, this text has been changed to read "To study chlorine's effect on grain boundary defect passivation, hybrid DFT level theory was used with static calculations in a 4 X 1 X 2 k-point grid using the HSE06 hybrid functional."