{"id":"be55aa6b-67f5-47d2-b794-90a979ac5a00","arxiv_id":"1908.05237","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"DFT calculations predict that thiocarbamide (TCA) forms the strongest, most trap-passivating interface between CH3NH3PbI3 and graphene among three Lewis bases, consistent with a measured PCE increase from 10.20% to 13.03%.","lead":"Researchers simulated interfaces between perovskite solar-cell material, three Lewis base molecules, and graphene using density functional theory, and found that sulfur-donor molecules bind more strongly and suppress trap states better than an oxygen-donor molecule. They also made solar cells with added thiourea (TCA) and measured a modest efficiency gain.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Trap-passivation claim rests on an unestablished I-rich growth regime and on GGA-level deep/shallow classification; both legs need testing before 'deep trap elimination' is accepted.","rationale":"The paper is a coherent DFT screening study: the binding-energy ranking in Table 1 and the migration barriers in Figure 3d are internally consistent, and the PCE improvement from 10.20% to 13.03% gives independent, if limited, experimental support. The load-bearing step for the headline passivation claim is the defect analysis around Table 2 and Figure 3: TCA is 'best' only if PbI is deep and absent. The absence condition is tied to the I-rich chemical-potential endpoint, which the paper does not establish for the actual growth conditions, and the lack of spin-orbit coupling or hybrid functionals leaves the deep/shallow classification unvalidated. The reader's weakest assumption identifies exactly this combination, so I agree. A recomputation with HSE06+SOC over the full allowed iodine chemical-potential range, together with a stoichiometry check of the experimental films, would settle the question. If that recomputation confirms both legs, the CONDITIONAL verdict can be upgraded; otherwise the passivation component should be softened. Thus I recommend no change to the reader's verdict.","tokens_in":11508,"tokens_out":7699,"duration_ms":93844,"concrete_test":"Recompute the four neutral defects and their (+/0) and (0/−) transition levels in the MAPI-TCA-G slab using HSE06+SOC, scanning the iodine chemical potential over the full Pb-rich to I-rich range with the same reference phases as Table 2, and determine the chemical potential fixed by a PbI2:MAI 1:1 precursor with 15 mg/mL TCA. If the PbI transition level lies more than 0.1 eV inside the gap, or if the PbI formation energy at the experimentally accessible chemical potential is below about 1 eV, the 'deep trap elimination' claim is not supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central 'most desirable' ranking depends on the passivation claim: MAPI-TCA-G suppresses deep traps because only PbI is deep and PbI is argued to be absent. That argument is made in Table 2 and Figure 3 under the I-rich endpoint, where E_f(PbI) = 5.15 eV for TCA, but the value drops to 1.39 eV at the Pb-rich endpoint. The paper asserts that the I-rich condition is 'more desirable' but never shows that the 15 mg/mL TCA device actually operates in the I-rich chemical-potential window; no I/Pb stoichiometry, no precursor-chemistry analysis, and no kinetic justification are given for why a thermodynamic formation energy should rule out non-equilibrium PbI incorporation during solution processing. If the accessible iodine chemical potential is not near the I-rich bound, the claimed suppression is not established, and the trap-passivation leg of the ranking collapses. Independently, the deep/shallow classification rests on GGA-DFT density-of-states analysis without spin-orbit coupling or hybrid functionals and without charged-defect transition levels. For a lead-based perovskite, this can misplace defect levels by several tenths of an electron volt, which is exactly the margin that separates 'deep' from 'shallow' near mid-gap. Both legs must be tested before 'deep trap states elimination' is accepted.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper combines DFT calculations and device experiments to study MAPI(110) interfaces with three Lewis bases (thiocarbamide, thioacetamide, dimethyl sulfoxide) and a graphene layer. The DFT part quantifies interface binding, charge redistribution, interface dipoles and work function changes, defect formation energies for vacancies and antisites under I-rich and Pb-rich conditions, density-of-states analysis of defect levels, and iodine migration barriers. The authors conclude that the MAPI-TCA-G system is the most desirable for efficiency and stability because it binds most strongly, suppresses deep trap states (only the PbI antisite is deep and it is argued to be absent under I-rich conditions), and gives the highest iodine migration barrier. The experimental part reports a PCE increase from 10.20% to 13.03% when 15 mg/mL TCA is added to the perovskite precursor, which the authors interpret as support for their theoretical predictions.","tokens_in":11725,"tokens_out":4154,"duration_ms":48493,"significance":"If established, the results would provide a mechanistic rationale for Lewis-base/graphene interface engineering in halide perovskite solar cells and a concrete design rule favoring S-donor molecules. The paper has genuine strengths: it uses first-principles calculations without parameter fitting to the target PCE, includes van der Waals corrections, uses Monte Carlo sampling for adsorbate configurations, and directly connects the simulation to a new experiment. However, the central 'deep trap states elimination' claim and the resulting 'most desirable' ranking rest on two untested assumptions: that the experimental growth conditions are iodine-rich, and that GGA-level DOS without spin-orbit coupling reliably classifies defect levels as deep or shallow. These issues need to be resolved before the mechanistic conclusion can be accepted.","major_comments":[{"comment":"The claim of 'deep trap states elimination' is not supported by a direct comparison to the bare MAPI(110) surface. Figure 3a shows TDOS only for the MAPI-mol-G interface systems, with and without each defect. To demonstrate that the interface eliminates deep traps, the same defect DOS must be shown for the bare MAPI(110) slab computed with the identical methodology. Without this baseline, the reader cannot distinguish whether the interface itself passivates the defects or whether the defects are intrinsically shallow or deep on the clean surface. This baseline is load-bearing for the abstract's 'deep trap states elimination' statement and for the ranking of the three interface systems.","section":"Defect analysis (Table 2, Figure 3)"},{"comment":"The passivation conclusion relies on the I-rich chemical-potential endpoint, but the manuscript never demonstrates that the experimental TCA devices are processed under I-rich conditions. Table 2 gives E_f(PbI) = 5.15 eV under I-rich, but this drops to 1.39 eV under Pb-rich; under Pb-rich the deep PbI antisite would be abundant and the claimed suppression fails. The experiments report no I/Pb stoichiometry, precursor ratio, or chemical-potential analysis, and the text offers no kinetic argument for why a thermodynamic formation energy should prevent non-equilibrium PbI incorporation during solution processing. The statement 'the I-rich condition is favorable for surface trap passivation' is therefore an unverified auxiliary hypothesis rather than a demonstrated property of the TCA interface.","section":"Defect formation energies (Table 2) and experimental conditions (Figure 4)"},{"comment":"The deep/shallow classification is made from neutral-defect DOS at the GGA level (BLYP) without spin-orbit coupling (SOC) and without charged-defect transition levels. For lead iodide perovskites, SOC strongly modifies the band-edge character and can shift defect levels by several tenths of an eV, which is exactly the margin that separates a deep mid-gap state from a shallow state. In addition, the thermodynamic identity of a defect as a recombination center is properly established through charge-state transition levels, not through the neutral DOS alone. The authors should either repeat key defect calculations with SOC and/or a hybrid functional, or compute transition levels, before asserting that TCA and TAA interfaces eliminate deep traps while DMSO does not.","section":"Methodology (Figure 1 caption and Figure 3)"}],"minor_comments":[{"comment":"The caption has a labeling inconsistency: panel (c) is used twice ('Top view of vacancy-mediated I atom migration... and (c) their activation barriers'), and Figure 3d is referenced in the main text. The fourth panel should be labeled (d).","section":"Figure 3 caption"},{"comment":"The SEM images are referred to in the text as '(Figure 4e)' and '(Figure 4f)', but the caption reads 'SEM images for the perovskite films coated on FTO (d) without and (f) with 15 mg/mL TCA'. The '(d)' appears to be a typo for '(e)'.","section":"Figure 4 caption and main text"},{"comment":"The formation energy E_f is listed in kJ/mol without specifying whether the value is per molecule, per surface cell, or per adsorbed monolayer. The text discusses 'adsorption energies per molecule' for different coverages, so Table 1 should state the normalization explicitly to allow comparison with the SI values.","section":"Table 1"},{"comment":"The decomposition of the interface dipole into bond dipole and molecular dipole components is essential for the work-function analysis, but the main text gives only the final numbers (e.g., TCA molecular dipole −1.2 D and bond dipole 5.4 D) and refers the reader to the SI. A one-sentence definition of how these components are computed would improve readability.","section":"Interface dipole decomposition"}],"recommendation":"major_revision","confidential_remarks":"The paper's core idea is timely and the combination of defect chemistry with Lewis-base/graphene interfaces is not common, but the main mechanistic claim depends on an unverified chemical-potential regime and a DFT level that may not resolve the deep/shallow distinction. The authors should be asked to add the bare-surface defect baseline and to justify or verify the I-rich assumption before the claim of trap elimination is accepted. The experimental part is suggestive but limited; more device statistics would strengthen the validation."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Kim, here's the paper. It's a DFT comparison of three Lewis bases (TCA, TAA, DMSO) sandwiched between MAPI(110) and graphene, with a short experimental follow-up. The new content is the systematic comparison: binding energies, interface dipoles, work function shifts, iodine migration barriers, and defect formation energies for the same interface models. That's a useful dataset. The S-donor > O-donor trend in binding is not a surprise to coordination chemists, but the specific numbers for these sandwiched interfaces are new and the migration barrier increase with TCA is a nice additional result. The calculations look carefully done, with vdW-inclusive DFT and a slab model that matches earlier work on MAPI-graphene contacts.\n\nThe soft spot is the 'deep trap elimination' claim. The paper shows defect DOS only for the interface systems; there's no bare-surface baseline, so 'elimination' is not directly demonstrated. And the argument that PbI is suppressed under I-rich conditions depends on the formation energy at the I-rich endpoint, but the paper never shows that the actual solution-processed film sits in that window. The stress-test note is right: the I-rich assumption is load-bearing and unverified. Also, the defect levels are calculated with GGA without spin-orbit coupling, and for a lead-based perovskite that can shift levels by a few tenths of an eV—exactly the margin between deep and shallow. The experiments are thin: one J-V curve per concentration, no statistics, no error bars. The PCE jump from 10.20% to 13.03% is consistent with the prediction but hardly a proof.\n\nSo the central ranking 'TCA is the most desirable' holds on binding and migration grounds, but the passivation leg is conditional. The paper should be revised to either qualify the deep-trap claim or add the missing baselines and chemical-potential analysis. I'd still send it out; the binding and migration data are worth having, and the defect analysis is a reasonable starting point. It's not a desk reject, but the referee should push on the trap-passivation mechanism.","headline":"A useful comparative DFT study of Lewis-base/graphene interfaces on MAPI, with a conditional trap-passivation claim that needs more support before the central ranking is accepted.","tokens_in":12305,"tokens_out":1899,"would_cite":true,"duration_ms":20503,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper claims that adding thiocarbamide and graphene to the CH3NH3PbI3 surface gives the strongest binding, removes deep trap states under iodine-rich conditions, and raises the iodine migration barrier, lifting measured efficiency…","keywords":["halide perovskite","Lewis base","graphene","work function","defect passivation","iodine migration","solar cell","density functional theory"],"falsifier":"Grow thiocarbamide-treated $\\mathrm{CH_3NH_3PbI_3}$ under deliberately iodine-poor (Pb-rich) conditions and measure sub-gap absorption or deep-level photoluminescence; if a mid-gap state near the predicted lead-on-iodine antisite level appears, the deep-trap-elimination claim fails for those conditions.","tokens_in":11304,"feed_emoji":"☀️","tokens_out":10730,"duration_ms":92804,"temperature":0.7,"pith_summary":"The paper tries to establish why Lewis-base molecules plus a graphene layer improve hybrid perovskite solar cells, and which of three Lewis bases works best. Density functional theory on $\\mathrm{CH_3NH_3PbI_3}(110)$ surfaces shows sulfur-donor thiocarbamide (TCA) and thioacetamide (TAA) bind more strongly to the perovskite and to graphene than oxygen-donor dimethyl sulfoxide (DMSO), with TCA the strongest. The calculations also indicate that, when the perovskite is processed under iodine-rich conditions, the only defect creating deep mid-gap states in the TCA and TAA interfaces is the Pb-on-I antisite, whose high formation energy means it should be suppressed, whereas DMSO leaves deep traps from all studied defects. The same interfaces raise the activation barrier for vacancy-mediated iodine migration, implying better stability. Supporting devices show that adding 15 mg/mL TCA raises power-conversion efficiency from 10.20% to 13.03%.","feed_headline":"Thiocarbamide plus graphene lifts perovskite cell efficiency to 13.03%","feed_subtitle":"DFT: the S-donor molecule binds strongest, removes deep traps, and slows iodine drift.","key_machinery":"The central object is a slab model of the $\\mathrm{PbI_2}$-terminated $\\mathrm{CH_3NH_3PbI_3}(110)$ surface covered by a monolayer of one of three Lewis bases (TCA, TAA, or DMSO) and a graphene layer, studied with density functional theory using a van der Waals-corrected functional. The argument runs through four connected quantities: adsorption and interlayer binding energies per carbon atom for binding strength; the interface dipole and work-function change for electron extraction; defect formation energies together with total and local density of states to classify deep versus shallow trap levels; and activation barriers for vacancy-mediated iodine migration along the edge of a $\\mathrm{PbI_5}$ octahedron. The S-donor lone pair binds to Pb atoms while the $\\mathrm{NH_2}$ and $\\mathrm{CH_3}$ groups hydrogen-bond to iodine atoms, which is why TCA binds strongest and, under iodine-rich conditions, leaves no deep traps.","core_discovery":"The central discovery is a mechanistic ranking of three $\\mathrm{CH_3NH_3PbI_3}$-Lewis base-graphene interfaces. In the MAPI-TCA-G interface, the S-donor TCA forms the strongest adsorption and interlayer binding, the shortest graphene-to-surface distance, and a work-function reduction directed toward graphene that should promote electron extraction. Defect-formation analysis on the $\\mathrm{PbI_2}$-terminated (110) surface shows that under iodine-rich conditions the only defect creating deep trap states in the TCA and TAA interfaces is the Pb-on-I antisite ($\\mathrm{Pb_I}$), whose formation energy is high enough (5.15 eV for TCA) that it is unlikely to form, while the DMSO interface retains deep traps for all tested defects. Vacancy-mediated iodine migration barriers are highest with TCA, indicating slower degradation. A companion experiment with thiocarbamide at 15 mg/mL raises the efficiency from 10.20% to 13.03%, in line with these predictions.","pith_inferences":["The iodine-rich assumption could be tested by varying halide stoichiometry during TCA-treated film growth; the model predicts deep traps should reappear under Pb-rich conditions because the Pb-on-I antisite becomes cheap to form.","The same Lewis-base-graphene design may extend to other hybrid perovskites such as formamidinium lead iodide, since the acid-base binding mechanism does not depend on the methylammonium cation specifically.","Time-resolved photoluminescence on TCA-treated films should show longer carrier lifetimes if the predicted deep-trap removal is real, a measurement the paper does not report.","Devices made with and without the graphene layer but with the same TCA treatment could isolate how much of the efficiency gain comes from the molecule versus the graphene contact."],"forward_implications":["If the central claim is right, S-donor Lewis bases such as thiocarbamide should outperform the O-donor DMSO in perovskite-graphene devices, with TCA the best of the three studied.","Under iodine-rich processing, TCA and TAA interfaces should show no deep mid-gap defect states, so non-radiative recombination should be suppressed, while DMSO interfaces should retain traps.","The higher iodine migration barrier with the TCA interface implies slower iodine-vacancy-mediated degradation and improved operational stability.","The work-function reduction at the MAPI-TCA-G interface should favor electron transfer from perovskite to graphene, improving charge extraction.","The measured efficiency gain, from 10.20% to 13.03% with 15 mg/mL TCA, is the experimental counterpart of the predicted trap removal and stronger interface bonding."],"supporting_citations":[{"why":"Supplies the comparative Lewis-base concept and the choice of TCA and DMSO as O- and S-donor adducts.","marker":"[12]"},{"why":"Provides the experimental thiourea/TCA treatment and carbon-electrode device context that the paper's own experiments build on.","marker":"[30]"},{"why":"Identifies surface defects of the perovskite as charge-carrier trapping centers, motivating the trap-passivation analysis.","marker":"[19]"},{"why":"Supplies the MAPI(110)-graphene supercell matching and a reference binding energy for the graphene interface.","marker":"[40]"},{"why":"Establishes the PbI2-terminated (110) surface as the most stable nondefective termination used in the slab model.","marker":"[42]"},{"why":"Provides the defect-formation-energy framework and the meaning of dominant vacancy and antisite defects in CH3NH3PbI3.","marker":"[50]"},{"why":"Addresses how spin-orbit coupling and self-interaction affect DFT classification of defect levels, the basis for the deep-versus-shallow assignment.","marker":"[48]"},{"why":"Establishes vacancy-mediated iodine migration as the relevant stability/degradation pathway whose barriers the paper computes.","marker":"[54]"},{"why":"Supplies the first-principles method for computing iodine diffusion activation barriers in perovskite solar cell materials.","marker":"[56]"}],"fun_headline_variants":["S-donor interface: stronger binding, no deep traps, higher barrier -> 13.03%","Traps eliminated, iodine pinned: S-donor perovskite hits 13.03%","DFT: S-donor interface kills deep traps, slows iodine, yields 13.03%","S-donor-graphene interface: trap-free, iodine-stable, cell 13.03%","Thiocarbamide-graphene: traps gone, iodine stays, cell hits 13.03%"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The deep-trap-elimination claim depends on the perovskite being formed under iodine-rich conditions, where the only deep trap (a lead atom sitting on an iodine site) costs 5.15 eV to form, and the paper does not verify that the actual devices were iodine-rich.","fun_headline_variants_meta":{"raw":{"variants":["S-donor interface: stronger binding, no deep traps, higher barrier -> 13.03%","Traps eliminated, iodine pinned: S-donor perovskite hits 13.03%","DFT: S-donor interface kills deep traps, slows iodine, yields 13.03%","S-donor-graphene interface: trap-free, iodine-stable, cell 13.03%","Thiocarbamide-graphene: traps gone, iodine stays, cell hits 13.03%"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00163,"raw_usage":{"total_tokens":6482,"prompt_tokens":948,"completion_tokens":5534,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":564,"completion_tokens_details":{"reasoning_tokens":5404}},"tokens_in":564,"tokens_out":5534,"duration_ms":42496,"temperature":1.0,"reasoning_tokens":5404,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:15:59.849066+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Grow thiocarbamide-treated $\\mathrm{CH_3NH_3PbI_3}$ under deliberately iodine-poor (Pb-rich) conditions and measure sub-gap absorption or deep-level photoluminescence; if a mid-gap state near the predicted lead-on-iodine antisite level appears, the deep-trap-elimination claim fails for those conditions.","supporting_citations":[{"cited_title":"Lee, H.-S","cited_arxiv_id":null,"evidence_quote":"Supplies the comparative Lewis-base concept and the choice of TCA and DMSO as O- and S-donor adducts."},{"cited_title":"Ko, G.-I","cited_arxiv_id":null,"evidence_quote":"Provides the experimental thiourea/TCA treatment and carbon-electrode device context that the paper's own experiments build on."},{"cited_title":"Uratani, K","cited_arxiv_id":null,"evidence_quote":"Identifies surface defects of the perovskite as charge-carrier trapping centers, motivating the trap-passivation analysis."},{"cited_title":"Zibouche, G","cited_arxiv_id":null,"evidence_quote":"Supplies the MAPI(110)-graphene supercell matching and a reference binding energy for the graphene interface."},{"cited_title":"Haruyama, K","cited_arxiv_id":null,"evidence_quote":"Establishes the PbI2-terminated (110) surface as the most stable nondefective termination used in the slab model."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the defect-formation-energy framework and the meaning of dominant vacancy and antisite defects in CH3NH3PbI3."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Addresses how spin-orbit coupling and self-interaction affect DFT classification of defect levels, the basis for the deep-versus-shallow assignment."},{"cited_title":"Eames, J","cited_arxiv_id":null,"evidence_quote":"Establishes vacancy-mediated iodine migration as the relevant stability/degradation pathway whose barriers the paper computes."},{"cited_title":"Haruyama, K","cited_arxiv_id":null,"evidence_quote":"Supplies the first-principles method for computing iodine diffusion activation barriers in perovskite solar cell materials."}],"review_version":1}