REVIEW 2 major objections 3 minor 35 references
The Pythagoras number of fields of transcendence degree $1$ over $\mathbb{Q}$
T0 review · 2 major / 3 minor · reviewed 2026-08-06 · deepseek-v4-flash
Pith's one-line read The paper proves that every field of transcendence degree one over Q has Pythagoras number at most 5: any sum of squares is a sum of five squares, via a representation theorem for rank-5 quadratic forms over function fields of curves over…
desk verdict Resolves the 30-year-old Pop–Pfister question with a genuinely new representation theorem; the proof is dense but coherent, and the flagged gap in Lemma 4.2 is not a real gap. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The load-bearing mechanism is condition (ii) of Theorem 1.2, which compares a line bundle $M\otimes P^{\otimes 2}$ with $\mathcal{O}_{C_v}(\Delta-D)$ and thereby controls, divisorially, the zeros of the prospective solution at every place where $q_v$ is one hyperbolic plane away from a smaller form. The key new arithmetic input is Proposition 4.8: over a p-adic field $k$, for any line bundle $L$ and any nonzero $\sigma\in H^0(C,L^{\otimes 2})$ with reduced zero locus, and any nondegenerate quadratic form $q$ of rank $\ge 3$ over $k$, there exists a divisor $\Delta$ on $C$ with $L\simeq \mathcal{O}_C(\Delta)$ such that $\sigma$ is nonzero and represented by $q$ at every point of the support of $\Delta$. Its proof runs through global class field theory for function fields via a description of line bundles on curves over finite fields (Proposition 4.3) and through the Lang–Weil estimates, and it lets the proof take $M=\mathcal{O}_C$ at p-adic places. This step has no counterpart in the earlier polynomial-based proof for $k(t)$, and it is exactly why the rank-5 case, where p-adic places impose constraints, can now be handled.
What would settle it
Exhibit a field $F$ of transcendence degree one over $\mathbb{Q}$ and an element of $F$ that is a sum of squares but provably not a sum of five squares; Corollary 5.11 says no such element can exist. Alternatively, find a single p-adic curve $C$, a line bundle $L$, a section $\sigma\in H^0(C,L^{\otimes 2})$ with reduced zero locus, and a nondegenerate rank-3 quadratic form $q$ for which no divisor $\Delta$ with $L\simeq \mathcal{O}_C(\Delta)$ makes $\sigma$ represented by $q$ at every point of the support of $\Delta$; this would invalidate Proposition 4.8 and with it the proof's bridge from the representation theorem to the five-squares bound.
Extended reading notes
Core claim
The central claim is Theorem 1.1: if $F$ is a field of transcendence degree one over $\mathbb{Q}$, then $p(F)\le 5$. The paper reaches it by proving Theorem 1.2: for a geometrically connected smooth projective curve $C$ over a number field $k$ and a nondegenerate quadratic form $q$ of rank $r\ge 5$ over $k$, a nonzero rational function $f\in k(C)^*$, written $\operatorname{div}(f)=E-2D$ with $E$ reduced, is represented by $q$ in $k(C)$ if and only if there exists $M\in \operatorname{Pic}(C)$ satisfying: (i) at real places where $q_v$ is definite, $f$ has the corresponding sign and the Borel–Haefliger class $\operatorname{cl}_v(M)$ vanishes; and (ii) at places where $q_v = \tilde q_v \perp \langle 1,-1\rangle$, there are $P\in \operatorname{Pic}(C_v)$ and a divisor $\Delta$ with $M\otimes P^{\otimes 2}\simeq \mathcal{O}_{C_v}(\Delta-D)$ such that $f$ is invertible and $f(x)$ is represented by $\tilde q_v$ at every closed point $x$ in the support of $\Delta$. When $q$ is the five-square form, the sign condition at real places is the only obstruction: Corollary 5.4(c) shows that condition (ii) can always be met, so nonnegative rational functions on curves over number fields are sums of five squares.
Load-bearing premise
The proof of Theorem 1.1 relies on Proposition 4.8, which asserts that over a p-adic field, any section of the square of a line bundle with simple zeros can be made to take only values represented by a given nondegenerate rank-at-least-three form at all points of a divisor realizing that line bundle; this proposition is built on global class field theory for function fields and on Lang–Weil point counts, and if any step there fails, the deduction of the five-squares bound from the representation theorem collapses.
Editorial extensions
If this is right
- For every smooth curve $C$ over a number field $k$, every $f\in k(C)$ that is nonnegative at all real points is a sum of five squares in $k(C)$; in particular, Hilbert's 17th problem over such fields has the quantitative answer 'five'.
- The representation theorem supplies an if-and-only-if criterion for representability by any rank-$\ge 5$ form over $k(C)$, so it applies to forms other than the five-square form by checking the sign and line-bundle conditions.
- Pourchet's theorem for $k(t)$ is recovered as the special case $C=\mathbb{P}^1_k$, giving a uniform proof of the known bound $p(k(t))\le 5$.
- Because $p(\mathbb{Q}(t))=5$ is already known, the general bound $p(F)\le 5$ cannot be lowered for the whole class of transcendence-degree-one fields over $\mathbb{Q}$.
- The technique of solving $\sigma\alpha^2=\sum_i \beta_i^2$ with sections of line bundles, instead of bounding degrees of polynomials, is what extends the classical polynomial method to arbitrary curves.
Reading between the lines
- The p-adic statement of Proposition 4.8 may hold more generally, for higher-rank forms or for sections without the reduced-zero hypothesis; if so, the same mechanism could give representation theorems for quadratic forms over function fields of higher-dimensional varieties over number fields.
- Because the criterion is formulated through a line bundle $M$ and the classes $\operatorname{cl}_v(M)$, the exact value of the Pythagoras number for a specific curve could depend on the arithmetic of its Jacobian, suggesting that curves with different Mordell–Weil groups may exhibit finer structure between 4 and 5.
- The same device of forcing a divisor to realize a line bundle while controlling the values of a section could be adapted to measure other additive invariants of function fields of curves over number fields, such as levels or u-invariants.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper proves that every field F of transcendence degree 1 over Q has Pythagoras number at most 5 (Theorem 1.1), improving Pop's bound 6 and answering a question of Pop and Pfister. The proof is built around a representation theorem (Theorem 5.1) for quadratic forms of rank at least 5 over the function field k(C) of a curve over a number field: a form q represents f in k(C) if and only if a line bundle M exists satisfying explicit local conditions at real and p-adic places. The author develops a line-bundle formulation of the representation problem, a geometric analysis of quadric bundles over curves (Proposition 3.2), a real-variable criterion (Proposition 3.3), and arithmetic results over finite and p-adic fields (Section 4), culminating in Proposition 4.8. Theorem 1.1 is deduced by applying the criterion to the five-square form via Corollary 5.4(c).
Significance. If the proof is completed, this settles a long-standing open problem and extends Pourchet's bound p(k(t)) ≤ 5 from rational function fields to all transcendence-degree-one fields over Q. The line-bundle formulation of the local-global representation criterion is a genuine novelty, and the combination of Kato's local criterion with Lang-Weil estimates and global class field theory over finite fields goes substantially beyond earlier work. The paper is generally careful, identifies its external inputs explicitly, and shows no circularity: the main theorem is derived from published theorems of Kato, Merkurjev-Suslin, and Pourchet, together with standard facts on quadratic forms. The main technical claims are proven in the text, but there is a load-bearing gap in the verification of the p-adic part of condition (ii) of Theorem 5.1, detailed below, which affects Corollary 5.4(c) and hence the proof of Theorem 1.1 as written.
major comments (2)
- [§5.2, Remark 5.2(iii) and §4.3, Proposition 4.8] The assertion that for a p-adic place v condition (ii) of Theorem 5.1 is always satisfied for M = O_C does not follow from Proposition 4.8. Proposition 4.8 yields a divisor Delta with O_C(Delta) isomorphic to O_C(D) on which sigma is nonzero and represented by the relevant form, but condition (ii) also requires f to be invertible at every point of the support of Delta. Since f = sigma/tau^2, this requires the support of Delta to be disjoint from the support of D. The proof of Proposition 4.8 gives no control of Delta relative to D, and such control is not automatic: if C is an elliptic curve and D = p is a rational point, then every divisor Delta with O(Delta) isomorphic to O(p) has p in its support, since a representative avoiding p would differ from p by a principal divisor whose order at p is -1, i.e. a rational function with exactly a simple pole at p, which would give a degree-one map from an elliptic curve to P^1. Thus for f with a double pole at p, condition (ii) with M = O_C is impossible, and the claimed automatic verification fails. Corollary 5.4(c) and Theorem 1.1 rely on this step; the proof needs an additional argument, for instance a more careful choice of the horizontal divisors in Proposition 4.5 so that the resulting divisor avoids D, or a different choice of M.
- [§3.2, Proposition 3.2, proof of (i) ⇒ (iii)] In the proof of (i) implies (iii), after solving equation (3.2) the author chooses a general orthogonal transformation so that gamma_r does not vanish at any zero of sigma or alpha, and then sets Delta = {gamma_r = 0}. For the application to condition (ii) of Theorem 5.1 in Step 1 of the proof of that theorem, Delta must also be disjoint from the zero divisor D of tau, because f = sigma/tau^2 must be invertible on the support of Delta. The current argument does not impose this condition, so the divisor Delta produced may have support in D. This is repairable by additionally requiring gamma_r not to vanish on the support of D, which is possible because the vector of sections (beta_i) is not identically zero at such points; but as written the necessity direction of Theorem 5.1 is not complete.
minor comments (3)
- [§4.1, Lemma 4.2] The final inference that the morphism \hat{X}_d -> X_{F_{q^d}} is an isomorphism is terse. Since \hat{X}_d -> X factors through the degree-d cover X_{F_{q^d}} -> X, the degree delta_d of \hat{X}_d over X is a multiple of d; together with the bound delta_d ≤ d this forces delta_d = d. Please state this divisibility explicitly.
- [§2.2, Proposition 2.9, final sentence of proof] The last sentence of the proof says that by Proposition 2.3 the form represents sigma, but the argument shows that every tau in the neighborhood W is represented; the final sentence should refer to tau, not to sigma.
- [§4.3, Proposition 4.8, reduction to the anisotropic case] The reduction to the case where q is anisotropic says only that isotropic forms are universal. The proposition also requires sigma to be nonzero at every point of the support of Delta, so the isotropic case needs a sentence explaining that one can choose a divisor in the class of L whose support avoids the zeros of sigma; this is standard but should be stated explicitly.
Circularity Check
No significant circularity: the proof derives p(F) <= 5 from external representation theorems and does not assume its conclusion.
full rationale
The paper's central claim (Theorem 1.1) is obtained as Corollary 5.9 from Theorem 5.8, which in turn follows from Corollary 5.4(c) applied to q = <1,1,1,1,1>. Corollary 5.4(c) is deduced from the representation theorem Theorem 5.1, whose proof uses Kato's local-global principle and local criterion [Kat86], the Merkurjev-Suslin theorem, Witt's theorems, Pourchet's Proposition 8, and local geometric Propositions 3.2 and 3.3. The p-adic condition (ii) is supplied by Proposition 4.8, proved via Lang-Weil estimates and global class field theory. At no point is the desired bound p(F) <= 5 assumed or used as an input, and no parameter is fitted to the target quantity. The paper contains no self-citations by the author and invokes no uniqueness theorem or ansatz from the author's own prior work. The only flagged concern, the gap between delta_d <= d and delta_d = d in Lemma 4.2, is a potential completeness/correctness gap in a supporting counting argument; it does not make any theorem's conclusion its own hypothesis, so it is not circularity. The derivation is therefore self-contained modulo acknowledged external results.
Assumptions & free parameters
assumptions (6)
- domain assumption Kato's local-global principle [Kat86, Theorem 0.8(2)]: the restriction map H^3(k(C), Z/2) to product over places is injective.
- domain assumption Kato's local criterion [Kat86, Proposition 5.2]: residue maps induce a quasi-isomorphism controlling H^3 in terms of henselizations at codimension-1 points.
- standard math Merkurjev-Suslin theorem (norm residue homomorphism): for a field k with 2 invertible, a Pfister form represents f iff the cup product {a}.{b}.{f} vanishes in H^3(k, Z/2).
- standard math Hasse-Minkowski theorem over number fields: a quadratic form represents an element iff it does so at all completions.
- standard math Global class field theory for function fields (Hasse-Witt) and Lang-Weil estimates.
- domain assumption Pourchet's Proposition 8: after scaling, any 5-dimensional quadratic form over a number field contains a 2-fold Pfister subform.
Cite this review
Pith. "Pith review of The Pythagoras number of fields of transcendence degree $1$ over $\mathbb{Q}$." pith.science (2026). https://pith.science/paper/IMSXK64D
@misc{pith2026250621380,
author = {Pith},
title = {Pith review of: The Pythagoras number of fields of transcendence degree $1$ over $\mathbbQ$},
year = {2026},
howpublished = {\url{https://pith.science/paper/IMSXK64D}},
note = {Machine review of arXiv:2506.21380}
}
abstract
We show that any sum of squares in a field of transcendence degree $1$ over $\mathbb{Q}$ is a sum of $5$ squares, answering a question of Pop and Pfister. We deduce this result from a representation theorem, in $k(C)$, for quadratic forms of rank $\geq 5$ with coefficients in $k$, where $C$ is a curve over a number field $k$.
Reference graph
Works this paper leans on
-
[1]
write newline
" write newline "" before.all 'output.state := FUNCTION output.nonempty.mrnumber duplicate missing pop "" 'skip if duplicate empty 'pop " " swap * " " * write if FUNCTION fin.entry add.period write mrnumber output.nonempty.mrnumber newline INTEGERS nameptr namesleft numnames FUNCTION format.language language empty "" " (" language * ")" * if FUNCTION form...
-
[2]
T. M. Apostol, Introduction to analytic number theory, Undergraduate Texts in Math., Springer-Verlag, New York-Heidelberg, 1976
work page 1976
-
[3]
Artin, \" U ber die Z erlegung definiter F unktionen in Q uadrate , Abh
E. Artin, \" U ber die Z erlegung definiter F unktionen in Q uadrate , Abh. Math. Sem. Univ. Hamburg 5 (1927), no. 1, 100--115
work page 1927
-
[4]
Artin, Algebraic approximation of structures over complete local rings, Publ
M. Artin, Algebraic approximation of structures over complete local rings, Publ. Math. IH \'E S 36 (1969), 23--58
work page 1969
-
[5]
E. Artin and O. Schreier , Algebraische Konstruktion reeller K\"orper , Abh. Math. Sem. Univ. Hamburg 5 (1926), 85--99
work page 1926
-
[6]
E. Artin and G. Whaples, Axiomatic characterization of fields by the product formula for valuations, Bull. AMS 51 (1945), 469--492
work page 1945
-
[7]
J. Bochnak, M. Coste and M.-F. Roy, Real algebraic geometry, Ergeb. Math. Grenzgeb. (3), vol. 36, Springer-Verlag, Berlin, 1998
work page 1998
- [8]
Show all 35 references
-
[9]
Binda, A
F. Binda, A. Krishna and S. Saito, Bloch's formula for 0-cycles with modulus and higher-dimensional class field theory, J. Algebr. Geom. 32 (2023), no. 2, 323--384
2023
-
[10]
M. F. Becker and S. MacLane, The minimum number of generators for inseparable algebraic extensions, Bull. AMS 46 (1940), 182--186
1940
-
[11]
Borel, Linear algebraic groups, 2nd ed., Grad
A. Borel, Linear algebraic groups, 2nd ed., Grad. Texts Math., vol. 126, 1991
1991
-
[12]
Colliot-Th\'el\`ene, Appendix to a Hasse principle for two dimensional global fields by K
J.-L. Colliot-Th\'el\`ene, Appendix to a Hasse principle for two dimensional global fields by K . K ato , J. reine angew. Math. 366 (1986), 181--183
1986
-
[13]
Colliot-Th \'e l \`e ne and U
J.-L. Colliot-Th \'e l \`e ne and U. Jannsen, Sommes de carr \'e s dans les corps de fonctions , CRAS, Paris, S \'e r. I 312 (1991), no. 11, 759--762
1991
-
[14]
Elman, N
R. Elman, N. Karpenko and A. Merkurjev, The algebraic and geometric theory of quadratic forms, Colloq. Publ., AMS, vol. 56, 2008
2008
-
[15]
Euler, Demonstratio theorematis F ermatiani omnem numerum sive integrum sive fractum esse summam quatuor pauciorumve quadratorum , Novi Comment
L. Euler, Demonstratio theorematis F ermatiani omnem numerum sive integrum sive fractum esse summam quatuor pauciorumve quadratorum , Novi Comment. Acad. Sci. Imp. Petropol. 5 (1760), 13--58
-
[16]
Jannsen, Hasse principles for higher-dimensional fields, Ann
U. Jannsen, Hasse principles for higher-dimensional fields, Ann. Math. (2) 183 (2016), no. 1, 1--71
2016
-
[17]
Jannsen and R
U. Jannsen and R. Sujatha, Levels of function fields of surfaces over number fields, J. Algebra 251 (2002), no. 1, 350--357
2002
-
[18]
Kato, A Hasse principle for two dimensional global fields , J
K. Kato, A Hasse principle for two dimensional global fields , J. reine angew. Math. 366 (1986), 142--181
1986
-
[19]
T. Y. Lam, Introduction to quadratic forms over fields, Graduate Studies in Math., vol. 67, Amer. Math. Soc., Providence, RI, 2005
2005
-
[20]
Landau, \"U ber die D arstellung definiter F unktionen durch Q uadrate , Math
E. Landau, \"U ber die D arstellung definiter F unktionen durch Q uadrate , Math. Ann. 62 (1906), 272--285
1906
-
[21]
Lang, The theory of real places, Ann
S. Lang, The theory of real places, Ann. of Math. (2) 57 (1953), 378--391
1953
-
[22]
Lang and A
S. Lang and A. Weil, Number of points of varieties in finite fields, Amer. J. Math. 76 (1954), 819--827
1954
-
[23]
A. S. Merkurjev and A. A. Suslin, \(K\) -cohomology of Severi -- Brauer varieties and the norm residue homomorphism , Math. USSR Izv. 21 (1983), 307--340
1983
-
[24]
Pfister, Quadratic forms with applications to algebraic geometry and topology, LMS Lect
A. Pfister, Quadratic forms with applications to algebraic geometry and topology, LMS Lect. Note Ser., vol. 217, 1995
1995
-
[25]
Pieper, On E uler's contributions to the four-squares theorem , Historia Math
H. Pieper, On E uler's contributions to the four-squares theorem , Historia Math. 20 (1993), no. 1, 12--18
1993
-
[26]
Pollak, Orthogonal groups over global fields of characteristic 2, J
B. Pollak, Orthogonal groups over global fields of characteristic 2, J. Algebra 15 (1970), 589--595
1970
-
[27]
F. Pop, Summen von Q uadraten in arithmetischen F unktionenk \"o rpern , https://www2.math.upenn.edu/ pop/Research/files-Res/dimen1.pdf https://www2.math.upenn.edu/ pop/Research/files-Res/dimen1.pdf , 1991
1991
-
[28]
, On the Pythagoras number of function fields of curves over number fields , Isr. J. Math. 257 (2023), no. 2, 561--574
2023
-
[29]
Pourchet, Sur la repr \'e sentation en somme de carr \'e s des polyn \^o mes \`a une ind \'e termin \'e e sur un corps de nombres alg \'e briques , Acta Arith
Y. Pourchet, Sur la repr \'e sentation en somme de carr \'e s des polyn \^o mes \`a une ind \'e termin \'e e sur un corps de nombres alg \'e briques , Acta Arith. 19 (1971), 89--104
1971
-
[30]
Serre, Corps locaux, Hermann, Paris, 1968
J.-P. Serre, Corps locaux, Hermann, Paris, 1968
1968
-
[31]
C. L. Siegel, Darstellung total positiver Zahlen durch Quadrate , Math. Z. 11 (1921), 246--275
1921
-
[32]
A. J. de Jong et al., The Stacks Project , https://stacks.math.columbia.edu
-
[33]
B. L. van der Waerden, Algebra V ol. 1 , Frederick Ungar Publishing Co., New York, 1970
1970
-
[34]
Voevodsky, Motivic cohomology with \( Z/2\) -coefficients , Publ
V. Voevodsky, Motivic cohomology with \( Z/2\) -coefficients , Publ. Math. IH \'E S 98 (2003), 59--104
2003
-
[35]
o rper \
E. Witt, Zerlegung reeller algebraischer Funktionen in Quadrate . Schiefk \"o rper \"u ber reellem Funktionenk \"o rper , J. reine angew. Math. 171 (1934), 4--11
1934
Reviewed August 6, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.