{"id":"e8a7c791-48da-4bd7-a61c-4989a5e4c70f","arxiv_id":"1908.05208","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"A continuum dielectric model reproduces explicit DFT substrate effects on charged defect ionization energies in 2D materials to within 0.05 to 0.16 eV, enabling cheap substrate-aware defect screening.","lead":"This paper introduces a fast computational method to include the electric screening of a substrate when calculating charged defects in two-dimensional materials, avoiding the huge cost of simulating the substrate atom by atom. If accurate, it enables high-throughput screening of quantum defects in materials like MoS2 and hBN on realistic substrates.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Single-Gaussian calibration of the substrate cavity is the weakest link: it is fit to one monopole observable and may not transfer to defects with different charge geometry or band-edge states.","rationale":"The paper is a solid methods paper: the continuum model is calibrated without reference to target defect energies, benchmarked against explicit substrate calculations for donors and acceptors, and the authors disclose major limitations such as semi-local DFT, frozen substrate geometry, and use of the optical dielectric constant. The reported agreement (0.05 eV for MoS2/SiO2, 0.10-0.16 eV for hBN substrates) supports the method for the tested class of in-plane, localized deep defects on insulating substrates. The weakest point remains the transferability of the single-Gaussian calibration. I can identify a specific mechanism that makes this more than a need-more-data complaint: Eq. 6 measures the total interaction with the DFT substrate, which includes the bare substrate potential, while the continuum model represents only the dielectric response; fitting z0 to one scalar can therefore mix a term that should cancel in charge transition levels with the image response that should not. This is exactly the kind of hidden assumption that a targeted calculation can expose. The proposed two-sign test-charge calibration and one out-of-plane or anisotropic defect benchmark would settle whether the concern lands. Since the reader already conditioned acceptance on broader validation, this stress test does not move the verdict.","tokens_in":11720,"tokens_out":16953,"duration_ms":196467,"concrete_test":"Recompute the substrate continuum parameters using E_int^DFT for a negative Gaussian test charge as well as the positive one, and compare the resulting z0 values; if they differ significantly, the fitted profile is mixing dielectric response with the bare substrate potential and cannot be transferred to acceptors. Then run one deliberately anisotropic or out-of-plane defect, such as an alkali interstitial above the monolayer or a charged vacancy on one sublattice, in explicit substrate DFT and in the continuum model; if the ionization-energy error exceeds roughly 0.2 eV while in-plane localized defects stay within 0.1 eV, the single-Gaussian calibration is not universal.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that a continuum dielectric slab fitted to the substrate response of a single Gaussian test charge at the monolayer center (Section II.B, Eq. 6) accurately predicts charged-defect transition levels on substrates for arbitrary defects. The load-bearing premise is that one fitted boundary position z0, constrained by one scalar interaction energy E_int^DFT, fully determines the substrate's response to every defect charge distribution and to the band-edge reference states used in Eqs. 3 and 4. This premise is not directly established by the benchmarks: all test defects are in-plane and fairly localized, so they mainly probe the same monopole-like response used for calibration. The calibration observable itself is a total-energy difference that includes not only the dielectric (image-charge) response but also the interaction of the test charge with the bare substrate's electrostatic potential; a dielectric-only continuum model cannot separate these contributions, so z0 can absorb a substrate-potential term that should cancel only in the special case where the defect charge and the band edge respond identically. For an acceptor versus a donor, or for a defect whose charge centroid is displaced out of plane or whose charging involves an anisotropic orbital, the same z0 is not guaranteed to be correct. Equation 7 reduces the substrate effect to a Gaussian self-energy plus a band-edge shift, thereby asserting defect-independence; Table I supports this for the tested set, but it is an empirical observation, not a derived property. Since the paper generalizes to arbitrary combinations of defects, 2D materials, and substrates, the unexercised sensitivity of the fit to charge geometry is the most load-bearing uncertainty.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a continuum dielectric model for including substrate effects in DFT calculations of charged defects in 2D materials. The substrate is represented by a smooth dielectric profile with parameters sigma and z0; z0 is calibrated against the DFT interaction energy of a Gaussian test charge with the explicit substrate. Defect ionization energies are then computed in the free-standing 2D material with the continuum substrate, plus band-edge shifts from an explicit perfect-layer/substrate calculation. The method is benchmarked against explicit substrate DFT for ReMo/NbMo in MoS2/SiO2 and for selected hBN defects on SiO2 and diamond, with reported agreement of 0.05 eV and 0.10-0.16 eV, respectively. The paper also introduces Eq. 7, a defect-independent estimate of the ionization-energy shift, and applies the method to predict substrate effects on ten hBN defects.","tokens_in":11938,"tokens_out":4518,"duration_ms":44180,"significance":"The work addresses a real bottleneck: charged-defect supercells with explicit substrates are costly. If the method holds, it enables high-throughput screening of defects in 2D materials on substrates at 40-200x reduction in defect-supercell cost. Strengths: benchmarks against explicit DFT for two material/substrate combinations; the sigma independence is demonstrated; Eq. 7 is a simple, falsifiable prediction; calculations use open-source software. Main limitation: the calibration and defect-independence assumptions are tested only on in-plane, fairly localized defects, leaving transferability to general defects unproven.","major_comments":[{"comment":"The calibration condition E_int^s(z0) = E_int^DFT fits z0 to the total DFT interaction energy of one Gaussian test charge. That energy includes the electrostatic interaction of the test charge with the bare substrate's potential, not just the dielectric image-charge response. A dielectric-only continuum model cannot distinguish these contributions, so z0 can absorb a substrate-potential term that may not transfer to defects with different charge centroids or orbital character. The current benchmarks are all in-plane localized defects, so they mainly probe the same monopole-like response used for calibration. To support the general claim, the authors should test z0 transferability, for example by calibrating with test charges at different positions or with a dipole/quadrupole charge distribution, and by benchmarking one defect with significant out-of-plane charge density.","section":"II.B, Eq. 6"},{"comment":"The claim that Eq. 7 estimates ΔIE with ~0.02 eV accuracy is supported only by comparison with the continuum model's own self-consistent calculations. Both columns in Table I derive from the same Gaussian self-energy in the same dielectric profile, so agreement is partially by construction. The authors should compare Eq. 7 directly with the explicit DFT ΔIE values (e.g., for ReMo/NbMo in MoS2/SiO2 and the explicit hBN cases) or clearly state that Eq. 7 is a fast approximation within the continuum model whose accuracy relative to explicit DFT has not been separately benchmarked.","section":"III.C, Eq. 7 and Table I"},{"comment":"The claim of \"computational expediency of calculating defects in free-standing 2D materials\" should be qualified, because the method still requires one explicit DFT calculation of the perfect 2D material on the substrate to determine ΔεVBM and ΔεCBM (Fig. 1(d), Section III.B). The 40-200x speedup applies only to the defect-containing supercells; the required substrate-plus-perfect-layer calculation remains an explicit, nontrivial cost. This should be stated clearly so readers can assess the true computational savings.","section":"Abstract and Section III.B"}],"minor_comments":[{"comment":"The notation for the continuum interaction energy is inconsistent; \"Es(z)int\" should be defined (e.g., E_int^s(z)) and distinguished from E_int^DFT.","section":"II.B"},{"comment":"Headers \"Self-consistent\" and \"Continuum Model\" in the two sub-tables are inconsistent; use the same labels for the same quantity.","section":"Table I"},{"comment":"The comparison for MoS2/SiO2 is against previous explicit calculations (Ref. 40), not new explicit calculations; this should be stated explicitly in the text.","section":"III.A"},{"comment":"The sentence \"The defects are all deep in hBN with ionization energies in the range of 2.14-4.01 eV\" appears to refer to both donors and acceptors; specify which panel and whether this range includes both.","section":"III.B"},{"comment":"The text says \"we select the cavity parameters such that E_int^s(z)=E_int^DFT\" and then mentions two parameters but fixes sigma; please clarify that only z0 is fitted.","section":"II.B"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know two things up front. First, this paper does something useful: it adapts continuum solvation to charged defects in 2D materials on substrates, and benchmarks it against explicit DFT for MoS2/SiO2 and hBN on SiO2/diamond. The agreement is honestly good—0.05 eV for the MoS2 cases, 0.10–0.16 eV for hBN—for a 40–200x cost cut. Second, the weakest link is the calibration: they fit the substrate cavity boundary z0 to the response of a single Gaussian test charge at the monolayer center, then use that same z0 for every defect and for band-edge shifts. That transferability is asserted more than demonstrated.\n\nWhat's genuinely new: a calibrated dielectric shape function for substrates, the explicit band-edge shift procedure (electrostatic shift from continuum plus core-level-aligned DOS corrections), and the Eq. 7 one-number estimator for ionization-energy reduction. Table I shows that estimator works to ~0.02–0.04 eV for the tested defects, which is a nice practical tool. The paper is also honest about its limits—semi-local DFT, optical dielectric constant, no Pauli repulsion in the continuum—and the calibration target is an auxiliary DFT quantity, not the defect energies themselves, so the circularity burden is low.\n\nNow the soft spots. The single-Gaussian fit is the load-bearing assumption. E_int^DFT in Eq. 6 includes not just the image-charge response but also the test charge's interaction with the bare substrate potential. A dielectric-only continuum can't separate those, so z0 can absorb a substrate-potential term. The band-edge shifts are handled separately, but the cancellation between defect and band-edge states is not automatic, especially for acceptors versus donors or defects with out-of-plane charge centroids. The benchmarks don't exercise that: all tested defects are roughly in-plane and localized, so they probe the same monopole response used to fit. This doesn't invalidate the method for the tested cases—it clearly works there—but it does undercut the abstract's \"arbitrary combinations\" claim. The supplemental data (input files, Table S1–S5) is referenced but not included in the arXiv version, so the reproducibility claim can't be checked from the manuscript alone.\n\nThe citation pattern is appropriate; it builds on the authors' prior work without hiding that. This is a solid methods paper for defect engineering and quantum-defect screening in 2D materials. I'd send it to a serious referee. My recommendation: accept conditionally, asking for either more diverse validation (different substrate strongly varying potential, defect with out-of-plane charge) or a careful statement limiting the regime of validity to localized, in-plane defects.","headline":"A pragmatic continuum-solvation extension that passes its own benchmarks, but the single-Gaussian calibration makes the blanket generality claim thinner than the abstract suggests.","tokens_in":12525,"tokens_out":2482,"would_cite":true,"duration_ms":27570,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["71.55.-i","71.15.Mb"],"model":"deepseek-v4-flash","headline":"A calibrated continuum dielectric profile reproduces explicit-substrate charged-defect energies in 2D materials to within 0.16 eV.","keywords":["charged defects","two-dimensional materials","continuum dielectric model","substrate effects","density-functional theory","charge transition levels","hexagonal boron nitride","molybdenum disulfide"],"falsifier":"Compute, with explicit substrate DFT, a charged defect whose charge density is strongly delocalized or shaped very differently from a Gaussian (for example, a shallow donor with a diffuse orbital, or a defect whose wavefunction hybridizes with the substrate), and compare its ionization energy to the continuum-model prediction. A disagreement larger than the 0.05–0.16 eV band found for the localized defects studied here would show that a single Gaussian test charge does not calibrate the substrate for all defects.","tokens_in":11486,"feed_emoji":"⚛️","tokens_out":8256,"duration_ms":76473,"temperature":0.7,"pith_summary":"Charged defects in two-dimensional materials are routinely calculated as free-standing slabs, yet real devices place the monolayer on a substrate whose dielectric screening shifts defect levels by hundreds of meV. This paper argues that the substrate can be removed from the DFT supercell and replaced by a smooth continuum dielectric profile, calibrated with a single DFT calculation of the bare substrate responding to one Gaussian test charge. With that calibration, solving a modified Poisson equation inside the DFT self-consistency loop reproduces explicit-substrate ionization energies to within 0.05 eV for MoS2/SiO2 and 0.10–0.16 eV for hBN on SiO2 and diamond, at 40–200x lower cost. If the method holds, high-throughput screening of quantum defects with realistic substrate environments becomes practical, and a further simplification (Eq. 7) shifts all free-standing defect levels by a single defect-independent number.","feed_headline":"Continuum model reproduces substrate shifts of 2D defect levels","feed_subtitle":"One Gaussian-charge calibration matches explicit-substrate DFT within 0.16 eV at 40-200x lower cost, enabling fast screening.","key_machinery":"The load-bearing object is the dielectric shape function $s(z) = \\frac{1}{2}\\mathrm{erfc}\\left(\\frac{z_0 - z}{\\sigma \\sqrt{2}}\\right)$ and the associated modified Poisson equation solved in cylindrical coordinates within the DFT self-consistency loop; it converts the substrate's many atoms into a smooth dielectric slab whose bound surface charge screens the defect. Its boundary position $z_0$ is calibrated from one Gaussian test charge, and $\\sigma$ is fixed at 0.2 Å with results insensitive below 0.3 Å. A secondary mechanism, Eq. 7, uses the isolated self-energy of the same Gaussian model charge in the 2D-only versus 2D-plus-substrate dielectric environments, plus band-edge offsets, to give a defect-independent estimate of the ionization-energy shift.","core_discovery":"The central claim is that a continuum model of the substrate, given by a dielectric profile $\\epsilon(z) = 1 + (\\epsilon_b - 1)s(z)$ with $s(z) = \\frac{1}{2}\\mathrm{erfc}\\left(\\frac{z_0 - z}{\\sigma \\sqrt{2}}\\right)$, contains the decisive electrostatic physics of charged defects in 2D materials on substrates. The boundary position $z_0$ is not guessed; it is fixed by requiring the interaction energy of a Gaussian test charge at the monolayer center with the continuum substrate to equal the corresponding DFT interaction energy of the explicit substrate. Once calibrated, the modified Poisson equation replaces the Hartree term in the DFT calculation, yielding charged-defect total energies whose ionization energies agree with explicit substrate calculations. For band edges, an additional explicit calculation of the perfect 2D material on the substrate supplies the non-electrostatic VBM/CBM offsets, which are small relative to the roughly 0.94–1.00 eV rigid electrostatic shift. The paper also shows the ionization-energy reduction is nearly identical across defects, so a single Gaussian self-energy shift (Eq. 7) converts all free-standing defect levels to substrate values within about 0.02 eV.","pith_inferences":["One testable extension: because the shift is nearly defect-independent for localized defects, the method's range could be mapped by testing delocalized or substrate-hybridized charge states, where the single-Gaussian calibration is likely to be least accurate.","The paper leaves implicit that the same dielectric-profile construction applies to encapsulated multilayers and heterostructures by superposing profiles for each interface; the calibration cost would be paid once per material pair, not per defect.","A direct prediction the benchmarks do not test: replacing the optical dielectric constant with the low-frequency value should reproduce explicit calculations in which substrate atoms relax, a comparison that would validate the model's treatment of substrate degrees of freedom.","The explicit band-edge offset calculation remains a one-time DFT run per (2D material, substrate) pair, so the 40–200x speedup applies to the per-defect screening loop rather than the whole pipeline; the economics favor many defects sharing one substrate."],"forward_implications":["Charged-defect transition levels in 2D materials on realistic substrates can be computed from free-standing DFT plus a calibrated dielectric environment, at 40–200x lower cost than explicit substrate calculations.","Substrate screening lowers donor and acceptor ionization energies: by 0.10–0.15 eV for ReMo and NbMo in MoS2/SiO2, by 0.27–0.33 eV for hBN defects on SiO2, and by 0.47–0.64 eV on diamond.","All ten hBN defects studied remain deep on SiO2 and diamond even after substrate screening, so they are expected to retain the long excited-state lifetimes needed for quantum applications.","Because the ionization-energy reduction is nearly independent of the defect, a single Gaussian-charge shift computed once per material/substrate pair can update every free-standing defect level with about 0.02 eV accuracy.","The continuum construction applies to arbitrary combinations of defects, 2D materials, and substrates, enabling screening of both defect candidates and material–substrate pairs for targeted properties."],"supporting_citations":[{"why":"Supplies the model-charge correction scheme and the Bessel-function Poisson solver used for charged-defect total energies and for the continuum substrate interaction.","marker":"[17]"},{"why":"Provides the anisotropic dielectric response and free-standing 2D charged-defect methodology that the substrate model extends, including the hBN defect baseline.","marker":"[18]"},{"why":"Provides the continuum solvation framework of replacing the environment by a dielectric cavity with a smooth shape function and a modified Poisson equation, which the substrate model adapts.","marker":"[32]"},{"why":"Gives the earlier explicit-substrate DFT results for ReMo and NbMo in MoS2/SiO2 against which the continuum model is benchmarked.","marker":"[40]"},{"why":"Supplies the free-standing hBN defect transition-level results that define the baseline for the substrate shifts predicted here.","marker":"[23]"},{"why":"Supplies the band-structure unfolding technique used to locate hBN VBM and CBM offsets on the incommensurate hBN/SiO2 interface.","marker":"[47]"}],"fun_headline_variants":["Continuum model nails substrate shifts for 2D defects","Cheap substrate corrections for 2D defect levels","One calibration captures substrate shifts for 2D defects","Fast substrate-aware defect energies via continuum model","High-throughput screening of 2D quantum defects with substrates"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that one dielectric profile, fitted to the response of the bare substrate to a single Gaussian test charge at the monolayer center, stays accurate for every defect's charge distribution, position, and orbital character, and also for the band-edge shifts.","fun_headline_variants_meta":{"raw":{"variants":["Continuum model nails substrate shifts for 2D defects","Cheap substrate corrections for 2D defect levels","One calibration captures substrate shifts for 2D defects","Fast substrate-aware defect energies via continuum model","High-throughput screening of 2D quantum defects with substrates"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001338,"raw_usage":{"total_tokens":5451,"prompt_tokens":965,"completion_tokens":4486,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":581,"completion_tokens_details":{"reasoning_tokens":4409}},"tokens_in":581,"tokens_out":4486,"duration_ms":29538,"temperature":1.0,"reasoning_tokens":4409,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:20:09.859800+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute, with explicit substrate DFT, a charged defect whose charge density is strongly delocalized or shaped very differently from a Gaussian (for example, a shallow donor with a diffuse orbital, or a defect whose wavefunction hybridizes with the substrate), and compare its ionization energy to the continuum-model prediction. A disagreement larger than the 0.05–0.16 eV band found for the localized defects studied here would show that a single Gaussian test charge does not calibrate the substrate for all defects.","supporting_citations":[{"cited_title":"Sundararaman \\ and\\ author Y","cited_arxiv_id":null,"evidence_quote":"Supplies the model-charge correction scheme and the Bessel-function Poisson solver used for charged-defect total energies and for the continuum substrate interaction."},{"cited_title":"Wu , author A","cited_arxiv_id":null,"evidence_quote":"Provides the anisotropic dielectric response and free-standing 2D charged-defect methodology that the substrate model extends, including the hBN defect baseline."},{"cited_title":"Gunceler , author K","cited_arxiv_id":null,"evidence_quote":"Provides the continuum solvation framework of replacing the environment by a dielectric cavity with a smooth shape function and a modified Poisson equation, which the substrate model adapts."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the free-standing hBN defect transition-level results that define the baseline for the substrate shifts predicted here."}],"review_version":1}