{"id":"21a6526d-c27a-44c7-83c6-c20a4870f23d","arxiv_id":"1910.02987","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A semi-analytic scheme evaluates Coulomb integrals for arbitrary real powers by expanding distance operators in Gaussians.","lead":"This paper presents formulas for computing electron interaction integrals using a sum of Gaussian functions to approximate distance powers like 1/r. It extends standard quantum chemistry methods to arbitrary real powers, which may help correlation and wavefunction corrections.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The advertised test against known n,m=0,1,2 integrals is never performed; accuracy of the single finite-interval fit for u=1 is the unverified load-bearing step.","rationale":"I agree with the reader's weakest assumption: the method's nonstandard ingredient is the Gaussian expansion of |r|^-u, and its accuracy over the full integration range is not demonstrated. The Gaussian product and integration algebra in the paper is standard and independently checkable, so the central risk is not the derivation but the uncontrolled approximation of the distance operator. The paper itself states that only one fit, for u=1, is exhibited, and it never reports the promised comparison with the known n,m=0,1,2 cases. Since the final integrals are linear in the fit coefficients, small pointwise or area deviations on the fitting interval do not automatically imply small integral errors; the Gaussian factors can sample the poorly fitted small-r region or the unquantified region outside (0.04,50), and cancellation among large alternating coefficients can worsen the result. A direct comparison against Eqs. (3) and (5) would settle this. I also note that the phrase 'any real (n,m)' needs a domain restriction: for n or m >= 3 with s-type Gaussian factors the original Coulomb integral diverges at coalescence, while the Gaussian expansion yields a finite number; this should be stated as a limitation rather than presented as a universal method. These concerns do not overturn the paper's conditional status, but they do mean the headline claim should not be accepted without the missing validation and a clear statement of the admissible range of n,m.","tokens_in":17320,"tokens_out":6136,"duration_ms":63502,"concrete_test":"Use the Appendix 2 (ai,ci) set to evaluate the u=1 expansion in Eq. (9), insert it into Eq. (3) for a grid of p and R_CP values (e.g., p=0.1,0.5,2 and R_CP=0,0.5,2,5,10), and compare with the exact closed form (2π/p)F0(p R_CP^2). Repeat for the two-electron case Eq. (5) with representative p,q,R_PQ. Report relative errors and also isolate contributions from r<0.04 and r>50. If any relative error exceeds 0.1%, the finite-interval fit does not support the central claim.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing assumption is that a finite fit of |r|^-u on a bounded interval (Eq. 8) can be substituted into all-space Gaussian integrals with controlled error. The paper exhibits exactly one fit, for u=1 on (0.04,50), with coefficients in Appendix 2, and reports only area and pointwise deviations within that interval. It never computes a single Coulomb integral with the expansion, even though it announces that the known (n,m)=(0,1),(1,0),(0,2),(2,0),(1,1),(2,2) cases 'serve as test.' Without such a comparison, the central claim that integrals for any real n,m reduce to controlled linear combinations of elementary Gaussian integrals is unverified: errors outside (0.04,50), especially r<0.04 where 1/r^u is largest, are not bounded, and cancellation among the large alternating coefficients (coefficients up to ~10^7 in Appendix 2) can amplify pointwise fit error. In addition, the 'any real (n,m)' phrasing is too broad: for n or m >= 3 with s-type Gaussian factors, e.g. ∫ exp(-p|r1-RC|^2)|r1-RC|^-3 dr1, the true integral diverges at the nucleus while the Gaussian expansion returns a finite number; the paper does not state this restriction.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a semi-analytic scheme for evaluating one-, two-, and three-electron Coulomb-type integrals in which the distance operator is one of W = R_{C1}^{-n} R_{D1}^{-m}, R_{C1}^{-n} r_{12}^{-m}, or r_{12}^{-n} r_{13}^{-m}. The central idea is to approximate the radial factor r^{-u} on a finite interval by a linear combination of Gaussians and r^2-weighted Gaussians (Eqs. 8-10), then substitute this expansion into the integral. Because Gaussian products and exponent shifts are elementary (Appendices 3-5), the resulting integrals reduce to sums of elementary one-dimensional Gaussian integrals. The paper provides one explicit least-squares fit for u=1 on (0.04,50) with coefficients listed in Appendix 2, and it argues that this extends the known analytic cases n,m=0,1,2 to arbitrary real powers. No actual Coulomb integral is computed numerically with the expansion, despite the introduction announcing that the n,m=0,1,2 cases \"serve as test.\" The algebraic steps are standard and mostly correct, but the central accuracy claim is not validated.","tokens_in":17779,"tokens_out":2989,"duration_ms":31381,"significance":"If the method were validated, it would provide a unified semi-analytic route to Coulomb-like integrals with non-integer and higher powers, which could be useful for correlation corrections, R12-type theories, and approximate Hamiltonians. The paper also collects useful Gaussian product and rotation formulas in the appendices, and it provides explicit fit coefficients for the u=1 case, which is a reproducible concrete artifact. The main mathematical construction is transparent: the approximation is openly fitted, so there is no hidden circularity in the derivation itself; the fit, however, is the load-bearing approximation. The significance is currently limited by the absence of numerical validation against exact integrals and by an unaddressed divergence for powers n>=3 with Gaussian factors centered on the nucleus. These issues are fixable, but they are essential to the paper's central claim.","major_comments":[{"comment":"The introduction states that the known n,m=0,1,2 cases \"serve as test,\" but the manuscript never performs this test. Equations (3)-(7) list the exact results for those cases, yet no numerical comparison is made between those formulas and the Gaussian-expansion results. This is load-bearing because the paper's claim of an \"alternative solution\" for the known cases and of controlled accuracy for general (n,m) depends on such a comparison. I request a table comparing, for representative geometries, the Gaussian-expansion evaluations of Eqs. (11) and (12) against the exact values from Eqs. (3)-(6), including error columns.","section":"Section 'A good fit for 1/r^u' and Appendix 2"},{"comment":"The fit r^{-u} ≈ Σ c_i r^{2k} exp(-a_i r^2) is determined on the finite interval (b1,b2) = (0.04,50), but the Coulomb integrals are over all space. The paper reports only the area deviation and pointwise deviation inside that interval. For r < 0.04, where r^{-u} is largest for u>0, no error bound is given; the added correction C1 exp(-A1 r^2) + C2 exp(-A2 r^2) improves the fit but does not quantify the error, and cancellation among the large alternating coefficients in Appendix 2 (some ~10^7) can amplify small local errors. The paper needs an error estimate for the all-space integrals, or at least a convergence test as the fit interval and term count are varied.","section":"Eq. (8) and Section 'Evaluation of generalized Coulomb integrals with Gaussian expansion'"},{"comment":"The claim that the method works for \"any real (n,m)\" is too broad. For u >= 3, the integral ∫ exp(-p|r1-RP|^2) |r1-RC|^{-u} d^3r1 diverges when RC = RP, because the integrand behaves as r1^{-u} near the nucleus. The Gaussian expansion, however, returns a finite value since each term r^{2k} exp(-a r^2) is integrable. The paper mentions singularities of the incomplete gamma function only in a passing parenthesis, but the abstract and the main claims do not state the restriction u < 3. The scope must be restricted accordingly (or a regularization must be introduced), and the abstract amended.","section":"Eqs. (11)-(12) and abstract"},{"comment":"Although the method is presented for arbitrary real u, the only fitted coefficients provided are for u=1. No numerical demonstration is given for non-integer u or for u = 2, despite those cases being central to the claim of a general procedure. The derivation shows how the integrals would be evaluated once a fit is available, but it does not show that the least-squares fit performs adequately for other u values. At minimum, the manuscript should provide fits or a fitting procedure with error measures for the u values actually used in the applications it cites.","section":"Section 'A good fit for 1/r^u' and Section 'Evaluation of generalized Coulomb integrals'"}],"minor_comments":[{"comment":"The text contains numerous typographical errors and formatting artifacts (e.g., garbled subscripts in displayed formulas, missing equation numbers, and repeated symbols) that make the derivation harder to follow than necessary. A careful copyedit is recommended.","section":"Throughout"},{"comment":"References [9] and [10] are generic web links to Wolfram Alpha and Wikipedia. The integrals and identities used should be cited to standard textbooks or primary sources with page numbers, so that the reader can verify the statements independently.","section":"References [9] and [10]"},{"comment":"The text says the effective M0 is 35 because two correction Gaussians are added, but the listed coefficients in Appendix 2 still show M0=33 and M2=17. The relationship between the tabulated coefficients and the claimed effective count should be clarified.","section":"Appendix 2 and Section 'A good fit for 1/r^u'"}],"recommendation":"major_revision","confidential_remarks":"The paper is a plausible algebraic construction but not yet a validated method. The absence of any numerical test of a single Coulomb integral, combined with the unqualified 'any real n,m' claim and the divergence for n>=3, makes the central claim unsupported as it stands. I believe these issues can be fixed by adding validation and restricting the scope, so I recommend major revision rather than rejection. I also note that the citation list is heavy on the author's own prior work and light on the current literature for Gaussian expansions and integral evaluation."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Kristyan's paper offers a genuinely unified semi-analytic route to a family of Coulomb integrals with arbitrary real powers n,m in distance operators, by expanding |r|^{-u} in even-power Gaussians and reducing the resulting integrals to elementary forms. The algebra is standard and, as far as I can see, correct; the use of r^{2k} exp(-a r^2) to avoid square roots is a sensible trick. The paper is also honest about what is being approximated: the Gaussian fit is openly fitted to 1/r^u, and the integral evaluation is an exact consequence of that fit. The one documented fit (u=1 on (0.04,50)) is reproducible from Appendix 2, which is good.\n\nThe central problem is that the paper never performs the test it announces. The introduction says the known n,m=0,1,2 cases 'serve as test,' but no Coulomb integral is ever computed numerically and compared. We only get the fit deviation (area difference 0.05%) within the fit interval, which says nothing about how the fit behaves when multiplied by a steep Gaussian and integrated over all space. The small-r region r<0.04, where 1/r^u is largest, is patched with ad hoc correction terms, but again there is no integral validation. This is the load-bearing missing piece.\n\nThe other real issue is the 'any real (n,m)' claim. For s-type Gaussian densities centered on the singular point, integrals with n or m >= 3 diverge at the nucleus; the Gaussian expansion returns a finite number. The paper needs to state this restriction explicitly.\n\nThere are also practical concerns: the coefficients in Appendix 2 are large and alternating (up to ~10^7), which invites cancellation and numerical instability in the final linear combination. The paper does not address conditioning.\n\nNone of these are fatal. The method might well work when the fit is validated and the validity range stated. But right now it is a plausible procedure without evidence that the procedure actually produces accurate integrals.\n\nThis paper is for a specialist who needs such integrals and is willing to do the validation themselves. It deserves a serious referee, but the referee should insist on the missing numerical comparison, fits for representative non-integer u, and a caveat on divergent powers. If those are added, it could be a usable methods note.","headline":"Plausible semi-analytic scheme for generalized Coulomb integrals, but the promised validation against known cases is absent and the 'any real n,m' claim needs a divergence caveat.","tokens_in":18123,"tokens_out":3128,"would_cite":false,"duration_ms":30648,"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":"A Gaussian expansion of $|r|^{-u}$ turns all the title's Coulomb integrals into sums of elementary Gaussian integrals, for any real powers $n$ and $m$.","keywords":["Coulomb integrals","Gaussian expansion","distance operators","arbitrary real powers","electron correlation","three-electron integrals","Gaussian type orbitals","Boys function"],"falsifier":"Compute the two-electron integral $\\int\\int \\exp(-p|r_1-R_P|^2-q|r_2-R_Q|^2)|r_1-r_2|^{-1/2}\\,dr_1dr_2$ numerically and by the Gaussian-expansion formula with a fitted $u=1/2$ expansion; if the summed expansion deviates by more than the claimed fit accuracy while the $u=1$ case passes, the claim that arbitrary real $n,m$ are covered is not established.","tokens_in":17142,"feed_emoji":"⚛️","tokens_out":12695,"duration_ms":115694,"temperature":0.7,"pith_summary":"The paper sets out to show that one expansion, $|r|^{-u}$ as a sum of Gaussians times even powers of $r$, can serve as a universal substitution in Coulomb integrals. With this substitution, integrals over one, two, or three electron densities with distance operators $R_{C1}^{-n}R_{D1}^{-m}$, $R_{C1}^{-n}r_{12}^{-m}$, and $r_{12}^{-n}r_{13}^{-m}$ become sums of elementary Gaussian integrals. The method is claimed to work for any real $n$ and $m$, not just the familiar integers 0, 1, 2, and it recovers the known cases as tests. If true, this gives a common algorithmic route to correlation corrections, higher moments of inter-electronic distances, and operators with negative powers such as $r_{12}/r_{13}$. The paper supplies explicit formulas (Eqs. 11, 12 and their generalizations) and one concrete fit for $u=1$.","feed_headline":"One Gaussian trick unlocks arbitrary-power Coulomb integrals","feed_subtitle":"Expanding the distance operator into Gaussians lets 1-, 2-, and 3-electron integrals handle any real powers.","key_machinery":"The load-bearing object is the Gaussian expansion of the Coulomb operator, $|r|^{-u} \\approx \\sum_{k=0}^{L}\\sum_{i=1}^{M} C_{ik} r^{2k} \\exp(-A_{ik} r^2)$, with $L=1$ recommended. Its role is to eliminate the square root $|r|=\\sqrt{x^2+y^2+z^2}$ by replacing it with even powers $r^{2k}$, so that products of Gaussians remain Gaussians. The paper constructs the coefficients in two ways: a least-squares fit of $1/r$ on the interval $(0.04,50)$ with $M=50$ terms, the only concrete fit tabulated, and modified Taylor expansions in $s=r^2$ around $r_0=1$ that match derivatives while allowing boundary pivoting. Once the expansion is inserted, the Boys-function and erf arithmetic of standard formulas is replaced by sums over Gaussian terms; the remaining steps are routine Gaussian product formulas, binomial re-centering of polynomial factors, and rotation of the quadratic form to remove cross terms.","core_discovery":"The central claim is that if one fixes a Gaussian expansion of $|r|^{-u}$ as a sum of terms $c_i r^{2k_i} e^{-a_i r^2}$, then every integral in the title reduces to sums of products of one-dimensional Gaussian integrals. Because $r^{2k}$ is an even power under the substitution $r = \\sqrt{x^2+y^2+z^2}$, the square root inside $|r|$ disappears, and no Laplace transform restricted to integer $n$ is needed. The paper shows this for one-electron density integrals (Eq. 11), two-electron integrals (Eq. 12), and three-electron integrals with $r_{12}^{-n}r_{13}^{-m}$, where mixed terms such as $x_1x_2$ are removed by rotating the quadratic form in the exponent. The same expansion serves both nucleus-electron and electron-electron distances, so the coefficient set can be reused. The known $n,m=0,1,2$ results are recovered as special cases, which the paper treats as validation.","pith_inferences":["The method's practical reach will be set by how well $u\\neq 1$ fits perform; the paper tabulates only $u=1$, so the fractional-power claim is a testable extension rather than a demonstrated fact.","For very diffuse Gaussians or very large inter-particle separations, the integration weight can extend beyond the fitted interval $(b_1,b_2)$, so accuracy may degrade even where the $u=1$ area check looks good.","The same Gaussian expansion plus orthogonal diagonalization should carry over to four-electron operators such as $r_{12}^{-n}r_{34}^{-m}$, with larger symmetric matrices replacing the $2\\times2$ rotation.","Precomputing coefficient sets once per $u$ would let existing Gaussian-basis codes adopt the scheme without changing their integral drivers."],"forward_implications":["For any real $n$ and $m$, the same Gaussian-product and diagonalization steps evaluate the one-, two-, and three-electron integrals named in the title; changing $u$ only changes the precomputed expansion coefficients.","The known cases $(n,m)=(1,0)$, $(0,1)$, $(1,2)$ and the three-electron $r_{12}^{-1}r_{13}^{-1}$ integral are recovered as special cases, giving an internal consistency check.","Non-integer and negative powers become usable, opening correlation corrections and operators such as $r_{12}/r_{13}$ to the same treatment.","The Boys-function/erf arithmetic in standard formulas is replaced by sums over Gaussian terms, trading special functions for a larger number of elementary integrals.","Because the expansion coefficients depend only on $u$ and the fitting interval, one fitted set can serve both nucleus-electron and electron-electron distance operators across many geometries and basis sets."],"supporting_citations":[{"why":"Original introduction of a Gaussian expansion of $1/|r|$, which the paper generalizes to $|r|^{-u}$ and extends to arbitrary real powers.","marker":"[8]"},{"why":"Standard textbook treatment of Gaussian-type-orbital Coulomb integrals that defines the known cases the new scheme must reproduce.","marker":"[1]"},{"why":"Gives analytic expressions for $n,m=0,1,2$ and a three-electron $r_{12}^{-1}r_{13}^{-1}$ integral used as validation targets.","marker":"[2]"},{"why":"Motivates arbitrary real powers by showing correlated-electron methods use operators such as $r_{12}/r_{13}$, i.e. negative powers.","marker":"[3]"},{"why":"Review of Gaussian functions, geminals, and Boys-function techniques whose Laplace-transform evaluation the expansion replaces.","marker":"[4]"},{"why":"Review of Boys-function and Gaussian-integral evaluation methods that the paper positions its expansion as an alternative to.","marker":"[6]"},{"why":"Provides the analytic two-electron Schr\\\"odinger solution with $r_{12}^{-1}$, used to motivate correlation-factor integrals.","marker":"[5]"}],"fun_headline_variants":["Gaussian expansion cracks Coulomb integrals for any power","Semi-analytic integrals: arbitrary power Coulomb via Gaussians","Coulomb integrals at any power from one Gaussian trick","1,2,3-electron Coulomb integrals via Gaussian expansion"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"Everything rests on the premise that a finite fit of $|r|^{-u}$ over a limited range of distances stays accurate wherever the Gaussian electron clouds give the integral significant weight; the paper shows this only for $u=1$ between 0.04 and 50 bohr, and gives no error bound for other powers or other distance ranges.","fun_headline_variants_meta":{"raw":{"variants":["Gaussian expansion cracks Coulomb integrals for any power","Semi-analytic integrals: arbitrary power Coulomb via Gaussians","Coulomb integrals at any power from one Gaussian trick","1,2,3-electron Coulomb integrals via Gaussian expansion"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000335,"raw_usage":{"total_tokens":1982,"prompt_tokens":1192,"completion_tokens":790,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":808,"completion_tokens_details":{"reasoning_tokens":725}},"tokens_in":808,"tokens_out":790,"duration_ms":7143,"temperature":1.0,"reasoning_tokens":725,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:59:32.250787+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the two-electron integral $\\int\\int \\exp(-p|r_1-R_P|^2-q|r_2-R_Q|^2)|r_1-r_2|^{-1/2}\\,dr_1dr_2$ numerically and by the Gaussian-expansion formula with a fitted $u=1/2$ expansion; if the summed expansion deviates by more than the claimed fit accuracy while the $u=1$ case passes, the claim that arbitrary real $n,m$ are covered is not established.","supporting_citations":[{"cited_title":"Chem.: Intro","cited_arxiv_id":null,"evidence_quote":"Standard textbook treatment of Gaussian-type-orbital Coulomb integrals that defines the known cases the new scheme must reproduce."}],"review_version":1}