{"id":"07fa8df1-aa68-4086-99ca-38316f85af15","arxiv_id":"1908.02537","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":4,"one_line_summary":"PFC theory derived from DDFT through the standard approximations acquires spurious stripe phases, a spurious second spinodal, and a no-solution region; a one-mode approximation for log density is very accurate.","lead":"The authors show that the standard assumptions used to derive phase field crystal (PFC) theory from dynamical density functional theory (DDFT) introduce spurious stripe phases, a second freezing and melting transition, and a region with no solutions. They also show that approximating the logarithm of the density, rather than the density itself, is unexpectedly accurate and points to a better basis for PFC-type theories.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified: the artifact analysis is internally consistent, and the only soft spot (RPA baseline accuracy) is explicitly scoped and reference-supported rather than load-bearing.","rationale":"The reader's identified weakest assumption, RPA accuracy for GEM-4, is indeed the only place where the paper leans on external literature for its baseline. But the central negative claim about standard PFC derivations is supported by model-independent analytic arguments: the second spinodal arises purely from even-order truncation of the logarithmic ideal-gas term, and the DDFT-5 no-solution region arises from the gradient expansion of the nonlocal operator. These arguments are presented clearly in Sec. IV with supporting numerics and a singularity-balance analysis. The GEM-4 calculations serve as an illustration and for quantitative comparison, and the paper explicitly restricts to the regime where RPA is known to be accurate. The derivation sequence is internally consistent, and the naming and equation tracking in Table I and Sec. II are careful. I checked the algebraic route from Eq. (24) to Eqs. (34)-(35) and from the log expansion to PFC-beta/PFC-gamma; the coefficients and sign conventions are consistent. The continuation numerics are described in sufficient detail, including resolution checks and regridding. Overall, no load-bearing flaw was found, so the ACCEPT verdict should stand unchanged.","tokens_in":38038,"tokens_out":11628,"duration_ms":126481,"concrete_test":"Run equilibrium Monte Carlo or molecular dynamics simulations for 2D GEM-4 at state points inside the stripe and down-hexagon regions predicted by PFC-gamma and PFC-epsilon in Fig. 7(c,d), e.g., kBT/epsilon in 0.2-0.5 and average density rho0(1+nbar)R^2 in 8-12. Compare the observed stable phases against the DDFT-3 phase diagram in Fig. 7(a). If simulations show only the liquid and up-hexagon crystal with no reentrant liquid, stripe, or down-hexagon equilibrium phases, the artifact claim is confirmed; if stripes or down-hexagons appear as stable phases, the RPA baseline and the DDFT-3 comparison would need revision.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim is that the standard PFC approximations introduce qualitative artifacts. The most plausible weak point is the benchmark DDFT-3/RPA description of GEM-4: if RPA were inaccurate in the studied regime, the baseline phase diagram in Fig. 7(a) would be wrong and the PFC discrepancies might be corrections rather than artifacts. The paper itself flags this in Sec. III, citing Refs. [42,44,45] for RPA accuracy at high temperature/density. However, this concern does not land with enough force to change the verdict. The two headline artifacts are shown by arguments that do not depend on RPA accuracy: (i) the second spinodal follows from expanding log(1+n) to an even-order polynomial, because the spinodal condition P''(n_liq)=0 then has roots n_liq=0 and n_liq=1 for any even truncation (Sec. IV.A), independently of the pair potential; and (ii) the DDFT-5 no-solution region is an internal consequence of replacing the convolution with a finite-order gradient operator, balanced by a logarithmic singularity, as analyzed in Sec. IV.B and Fig. 8. The stripe and down-hexagon stability predictions are consequences of the same second spinodal. Thus even if RPA were imperfect, the qualitative conclusion that these PFC approximations create spurious phases and a no-solution regime would survive. The quantitative 'orders of magnitude' claim is specific to GEM-4, and there it rests on literature-supported RPA accuracy. No internal inconsistency or unsupported step was found in the derivation sequence from DDFT-0 through PFC-epsilon.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper systematically derives phase field crystal (PFC) theory from dynamical density functional theory (DDFT), identifying and naming each approximation in the chain: truncation of the functional Taylor expansion of the excess free energy, the Ramakrishnan–Yussouff/RPA treatment, replacement of the nonlocal convolution operator L by a gradient expansion, and finally the suite of changes that converts a DDFT into a PFC model (dropping the ∇·[n∇Ln] term, assuming constant mobility, and replacing the ideal-gas logarithm by a truncated polynomial). The authors construct a hierarchy of models DDFT-0 through DDFT-5 and PFC-β through PFC-ϵ, giving the free energy, chemical potential, and dynamics for each. They illustrate the consequences for the two-dimensional GEM-4 fluid, for which RPA-based DDFT is known to be accurate. Two central artifacts are documented: (i) expanding the logarithm and dropping ∇·[n∇Ln] introduces a second spinodal, producing erroneous stripe and down-hexagon phases and a re-entrant uniform liquid; (ii) making the gradient expansion while retaining the logarithm leads the density to approach zero at isolated points, with the logarithmic singularity balanced by a divergent fourth derivative, after which no real solutions exist.","tokens_in":38385,"tokens_out":11002,"duration_ms":108006,"significance":"This is a significant and carefully executed negative result for the PFC literature. The claim that standard polynomial-in-density PFC models cannot be derived as quantitatively faithful reductions of DDFT is supported by two arguments that are largely independent of the specific pair potential: the second spinodal follows from the algebraic structure of the truncated logarithm (Sec. IV.A), and the no-solution region follows from a local balance between a logarithmic singularity and a fourth derivative (Sec. IV.B and Fig. 8). The numerical work is reproducible: the continuation method is fully described in Appendix B, the singularity is verified at the grid level in Fig. 8(b), and a sample Matlab code is provided as supplementary material. The authors are appropriately careful in scoping their claims to the considered approximation sequence and to the GEM-4 model, and they explicitly flag the reliance on literature support for RPA accuracy in Sec. III. The constructive one-mode log-density result in Sec. V is an important positive step, as it indicates a principled route toward accurate PFC-type descriptions.","major_comments":[],"minor_comments":[{"comment":"The sentence 'we find it impossible to derive the PFC model as an accurate approximation to DDFT' is stronger than what is demonstrated, since the demonstration concerns the standard polynomial-in-density PFC models and the GEM-4 system; consider adding the qualifier 'for the density field n(x) within the approximation sequence considered here' in the abstract to prevent overgeneralization.","section":"Abstract, Sec. VI"},{"comment":"The phrase 'which involves the second derivative of Eq. (51) with respect to n' is imprecise: the spinodal condition (75) involves the derivative of the chemical potential, which after cancellation of the quadratic term is the second derivative of the log expansion minus a constant; rephrasing would improve clarity.","section":"Sec. IV.A"},{"comment":"The notation Lgrad-8 in Eq. (76) and in the surrounding text is slightly awkward because of the hyphen and the use of the same subscript 'grad'; consider a cleaner notation such as L_8 or L_grad^(8) for readability.","section":"Eq. (76)"},{"comment":"In the caption of Fig. 5, the circled region of quantitative agreement between DDFT-3 and PFC-γ is mentioned but not visually obvious in the small panels; adding a zoomed inset or a more prominent marker would help the reader locate this claim.","section":"Sec. III, Fig. 5"},{"comment":"The observation that equilibria with the same µ do not necessarily have the same mean density ¯n is important for interpreting the phase diagrams in Fig. 7; a brief additional remark on how the phase diagrams are constructed from the grand-potential comparison would be helpful for readers unfamiliar with this convention.","section":"Sec. III, Eq. (70)"}],"recommendation":"accept","confidential_remarks":"I found no load-bearing technical errors. The central negative claims are convincingly supported by analytic arguments and careful numerics, and the limitations are acknowledged explicitly in the manuscript. The minor comments are presentation-level suggestions that do not require another round of review."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The short version: this is the first derivation audit that actually follows the consequences instead of hand-waving terms away, and the two main results are real. Dropping the nabla-dot-[n nabla L n] term while expanding log(1+n) plants a second spinodal at n=1, and replacing the convolution L by a gradient expansion creates a genuine no-solution boundary. I went through the derivation sequence from DDFT-0 through PFC-epsilon and did not find a load-bearing error. The numerics are careful, the continuation is described well enough to reproduce, and Fig. 8(b) convinces me the DDFT-5 singularity is in the equation, not the grid.\n\nWhat is new: prior derivations drop the nabla-dot-[n nabla L n] term as \"higher order\" without checking. This paper shows it is the same order as the c(3) term and actually stabilizes the crystal. The second-spinodal explanation is clean and general: any even-order truncation of log(1+n) makes the spinodal condition have roots n=0 and n=1, so the liquid freezes and then melts again. The no-solution region in DDFT-5 is also convincingly analyzed, including a local asymptotic balance between the log and a fourth derivative. I also credit the closing observation that a one-mode ansatz for log(rho), rather than rho itself, is extremely accurate for the DDFT-3 profiles; that is a useful hint even though the theory is not fully developed.\n\nThe soft spot is the benchmark: DDFT-3 uses the RPA for GEM-4, and the paper explicitly scopes this and cites simulation/DFT literature for the regime. More importantly, the qualitative artifacts do not rest on RPA accuracy. The second spinodal follows from the algebraic structure of the expanded logarithm, and the no-solution region follows from replacing a convolution with derivatives. So the RPA dependence is a limitation of the quantitative orders-of-magnitude claim, not a flaw in the central argument. Minor point: the word \"impossible\" is stronger than the result; the authors do say \"impossible to derive PFC as an accurate approximation to DDFT\" within the considered approximation sequence and model, but a less careful reader may overread it. The one-mode log-density result is demonstrated only for GEM-4 stripes and hexagons; that is clearly stated, and the 3D discussion is an analogy from the 2D mechanisms, which is fine.\n\nThis paper deserves a serious referee. It will be useful to anyone who uses PFC, derives PFC from DDFT, or teaches either theory. I would bring it to reading group, cite it, and publish it after light revision.","headline":"This is the first honest derivation audit of PFC from DDFT that follows the approximations through to their phase-diagram consequences, and the two main artifacts it identifies are real.","tokens_in":38960,"tokens_out":2152,"would_cite":true,"duration_ms":25210,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["82C26","82D15","35Q82"],"pacs":[],"model":"deepseek-v4-flash","headline":"Standard phase field crystal theory cannot be derived from density functional dynamics without adding spurious states.","keywords":["phase field crystal","dynamical density functional theory","GEM-4 fluid","spinodal","stripe phase","gradient expansion","logarithmic ideal gas term","one-mode approximation"],"falsifier":"Simulate or experimentally measure two-dimensional GEM-4 at temperatures $k_BT/\\epsilon > 0.1$ across the density range where PFC predicts stripe, down-hexagon, and second-liquid states; if any of those appear as equilibrium phases, the claim that they are artefacts is wrong.","tokens_in":37873,"feed_emoji":"🧊","tokens_out":5332,"duration_ms":53901,"temperature":0.7,"pith_summary":"The paper follows every approximation in the standard derivation of phase field crystal (PFC) theory from dynamical density functional theory (DDFT), and shows where each step changes the physics. Its central conclusion is that the PFC model is not a faithful reduction of the accurate DDFT: dropping the $\\nabla\\cdot[n\\nabla L n]$ term and Taylor-expanding the ideal-gas logarithm introduces a second spinodal, so the liquid freezes and then melts again, and makes stripe and down-hexagon phases the equilibrium structures. Replacing the nonlocal operator $L$ by a gradient expansion instead creates a region of the phase diagram where no solution exists above a certain average density. These artefacts are demonstrated for the two-dimensional GEM-4 fluid, where the starting DDFT is known to be accurate.","feed_headline":"PFC theory cannot be derived from DDFT without artefacts","feed_subtitle":"Dropping one term makes liquids freeze, melt again, and form fake stripe phases.","key_machinery":"The load-bearing object is the linear operator $L$, defined by $L n(x)=-n(x)+\\rho_0\\int c^{(2)}(x,x_2)n(x_2)\\,dx_2$, whose Fourier eigenvalues $\\sigma(k)$ control the linear stability of the liquid; in DDFT it is a nonlocal convolution with the pair direct correlation function. The paper tracks the fate of the term $\\nabla\\cdot[n\\nabla L n]$ in the dynamics, which is linked to the density-dependent mobility and the logarithmic ideal-gas term. The PFC derivation replaces $L$ by the local gradient expansion $L_{\\mathrm{grad}}n = r n - \\gamma(1+\\nabla^2)^2 n$ and replaces $\\log(1+n)$ by its truncated Taylor polynomial; the number of real roots of the polynomial versus the one root of the logarithm is the mechanism behind the spurious second spinodal. The paper also constructs a one-mode ansatz for $\\varphi=\\log(\\rho/\\rho_0)$, using the decay of the convolution's Fourier coefficients at $|k|\\ge 2$ to show why a few modes in $\\varphi$ capture sharply peaked density profiles.","core_discovery":"On the paper's own terms, the discovery is that the standard PFC approximations are not harmless when applied to a microscopically accurate DDFT. The term $\\nabla\\cdot[n\\nabla L n]$, usually discarded as higher order, contributes at the same order as retained terms and helps stabilise the crystal relative to stripes. Dropping it forces a constant mobility and a polynomial replacement for $\\log(1+n)$; the polynomial has two roots where the logarithm has one, which is exactly the origin of the second spinodal and the erroneous stripe and down-hexagon equilibria. Replacing the convolution $L$ by the gradient operator $L_{\\mathrm{grad}}$ makes the density minimum tend to zero at finite chemical potential, with $\\log(1+n)\\to -\\infty$ balanced by $\\gamma n_{xxxx}\\to +\\infty$, so branches of solutions terminate. The paper therefore claims that PFC models are successful as phenomenological models but cannot be derived as accurate microscopic approximations for the density profile.","pith_inferences":["If the one-mode accuracy for $\\log\\rho$ carries to other potentials, a practical route to faithful PFC-type theories would be to write the free energy in terms of $\\varphi=\\log(\\rho/\\rho_0)$ and discard high-wavenumber convolution coefficients, rather than expanding in $n$.","The same mechanism, a polynomial replacing a logarithm with one root, implies that binary PFC models derived by the same route inherit analogous spurious re-entrant melting and lamellar phases; the paper notes that the $\\nabla\\cdot[n\\nabla L n]$ term is also dropped in binary derivations.","The second-spinodal artefact should appear generically for any even-order truncation of the logarithm, since the relevant equation keeps $n=0$ and $n=1$ as roots; this could be checked numerically with higher-order truncations of a model free energy.","For systems where the pair direct correlation function is not well described by the random-phase approximation, the quantitative size of the artefacts may differ, but the structural argument about roots and the gradient expansion should persist."],"forward_implications":["PFC phase diagrams for crystallising soft matter contain a second liquid spinodal, a stripe phase, and down-hexagons that are not present in the accurate DDFT, so quantitative predictions of crystal thermodynamics away from the coexistence region should not be trusted.","The gradient-expanded DDFT (DDFT-5) has a finite limit of validity: above a certain average density there is no smooth equilibrium profile, because the density touches zero and no solution exists.","Agreement between DDFT and PFC is quantitative only for small-amplitude states close to the spinodal; beyond that, agreement is at best qualitative.","A one-mode approximation for $\\log \\rho(x)$ reproduces the full DDFT-3 stripe and hexagon branches almost exactly, even when the density varies by orders of magnitude.","Extending the gradient expansion to higher order (EOF) improves the liquid compressibility but does not remove the no-solution singularity or the second spinodal; it only delays them."],"supporting_citations":[{"why":"Provides the standard PFC derivation from DDFT that drops the $\\nabla\\cdot[n\\nabla L n]$ term, which is the central approximation under scrutiny.","marker":"[24]"},{"why":"Introduces the gradient-expanded DDFT model (PFC1 in that paper), here called DDFT-5, and compares it with PFC2.","marker":"[23]"},{"why":"Review of PFC theory that lays out the accepted derivation route and the expected phase diagram features, serving as the baseline for what standard derivations claim.","marker":"[7]"},{"why":"Supplies the known phase diagram and simulation results for two-dimensional GEM-4 showing no stripe phase, establishing the benchmark that PFC fails to reproduce.","marker":"[42]"},{"why":"Establishes the GEM-4 model and its crystalline phases, providing the physical system on which the DDFT comparisons are based.","marker":"[41]"},{"why":"Gives the previously computed DDFT-3 phase diagram for GEM-4 that the paper reproduces and uses as the accurate reference.","marker":"[45]"},{"why":"Original gradient-expansion derivation of PFC that motivates replacing the convolution operator $L$ by $L_{\\mathrm{grad}}$.","marker":"[2]"},{"why":"Proposes the eighth-order fitting (EOF) gradient expansion, which the paper tests to show that higher-order terms only delay the spurious singular behaviour.","marker":"[9]"}],"fun_headline_variants":["PFC theory from DDFT fails without spurious artifacts","Dropping one term in PFC adds fake stripes and extra spinodal","PFC approximations from DDFT: re-melting and stripes are artifacts","Phase field crystal theory cannot be derived accurately from DDFT","PFC's gradient expansion can make solutions vanish entirely"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The random-phase approximation for the free energy is an accurate description of the GEM-4 fluid in the regime studied, so the extra phases predicted by PFC are artefacts rather than states the baseline theory already misses.","fun_headline_variants_meta":{"raw":{"variants":["PFC theory from DDFT fails without spurious artifacts","Dropping one term in PFC adds fake stripes and extra spinodal","PFC approximations from DDFT: re-melting and stripes are artifacts","Phase field crystal theory cannot be derived accurately from DDFT","PFC's gradient expansion can make solutions vanish entirely"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000264,"raw_usage":{"total_tokens":1669,"prompt_tokens":1076,"completion_tokens":593,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":692,"completion_tokens_details":{"reasoning_tokens":503}},"tokens_in":692,"tokens_out":593,"duration_ms":6419,"temperature":1.0,"reasoning_tokens":503,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:40:47.832004+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Simulate or experimentally measure two-dimensional GEM-4 at temperatures $k_BT/\\epsilon > 0.1$ across the density range where PFC predicts stripe, down-hexagon, and second-liquid states; if any of those appear as equilibrium phases, the claim that they are artefacts is wrong.","supporting_citations":[{"cited_title":"Pattern formation with a conservation law,","cited_arxiv_id":null,"evidence_quote":"Provides the standard PFC derivation from DDFT that drops the $\\nabla\\cdot[n\\nabla L n]$ term, which is the central approximation under scrutiny."},{"cited_title":"Spatial localization in dissipative sys- tems,","cited_arxiv_id":null,"evidence_quote":"Introduces the gradient-expanded DDFT model (PFC1 in that paper), here called DDFT-5, and compares it with PFC2."},{"cited_title":"Polymorphism, crystal nu- cleation and growth in the phase-field crystal model in 2D and 3D,","cited_arxiv_id":null,"evidence_quote":"Review of PFC theory that lays out the accepted derivation route and the expected phase diagram features, serving as the baseline for what standard derivations claim."},{"cited_title":"Microphase separation in two-dimensional systems with competing interactions,","cited_arxiv_id":null,"evidence_quote":"Supplies the known phase diagram and simulation results for two-dimensional GEM-4 showing no stripe phase, establishing the benchmark that PFC fails to reproduce."},{"cited_title":"A bidimensional fluid system with competing interactions: spontaneous and induced pattern formation,","cited_arxiv_id":null,"evidence_quote":"Establishes the GEM-4 model and its crystalline phases, providing the physical system on which the DDFT comparisons are based."},{"cited_title":"Hexatic phase and cluster crystals of two-dimensional GEM4 spheres,","cited_arxiv_id":null,"evidence_quote":"Gives the previously computed DDFT-3 phase diagram for GEM-4 that the paper reproduces and uses as the accurate reference."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Original gradient-expansion derivation of PFC that motivates replacing the convolution operator $L$ by $L_{\\mathrm{grad}}$."},{"cited_title":"Simulation of an atomistic dynamic field theory for monatomic liquids: Freezing and glass formation,","cited_arxiv_id":null,"evidence_quote":"Proposes the eighth-order fitting (EOF) gradient expansion, which the paper tests to show that higher-order terms only delay the spurious singular behaviour."}],"review_version":1}