Tensor products of finitely presented functors

Martin Bies; Sebastian Posur

REVIEW 2 major objections 5 minor 13 references

Tensor products of finitely presented functors

T0 review · 2 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Every finitely presented promonoidal structure on a small additive category extends to a right exact monoidal structure on the category of finitely presented functors, and the lifted tensor coincides with Day convolution.

desk verdict The constructive framework is valuable and the Day-convolution comparison is illuminating, but the proof of the main universal property rests on a misstated right-exactness lemma that must be corrected. read the letter →

arxiv 1909.00172 v1 pith:XJZJQAQW submitted 2019-08-31 math.CT hep-th

Martin Bies , Sebastian Posur This is my paper · ORCID

classification math.CThep-th MSC 18E1018E0518A25

keywords FreydcategoryfinitelypresentedfunctorcomputableabelianpromonoidalstructuremonoidalDayconvolutionrightexacttensorproductconstructivetheory

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper establishes that a tensor product on the category of finitely presented functors over a small additive category can be specified by finite data on the base objects and then extended, constructively, to a genuine right exact tensor product on all finitely presented functors. The input data, called a finitely presented promonoidal structure, is a bilinear protensor $T\colon A \times A \to \mathcal{A}(A)$ together with associativity, unit, and braiding isomorphisms whose coherence identities are only required on objects of $A$. The main result is that such data lifts to a right exact monoidal, braided monoidal, or closed monoidal structure on the Freyd category $\mathcal{A}(A)$, which is equivalent to the category $\mathrm{fp}(A^{\mathrm{op}}, \mathrm{Ab})$ of finitely presented functors. Since the constructions use only cokernels of finite data, the lifted tensor products are computable, and Theorem 4.2.1 shows they coincide with the restriction of Day convolution. The payoff is a unified machine: one mechanism yields tensor products on finitely presented modules, finitely presented graded modules, iterated Freyd categories, and free abelian categories, a setting relevant to algebraic motives.

What carries the argument

The load-bearing machinery is the multilinear 2-categorical universal property of Freyd categories (Theorem 2.3.1). A Freyd category $\mathcal{A}(A)$ is the universal way to adjoin cokernels to an additive category $A$: objects are morphisms $\rho\colon a \leftarrow r$ of $A$, thought of as formal cokernels, and $\mathcal{A}(A) \simeq \mathrm{fp}(A^{\mathrm{op}}, \mathrm{Ab})$. The theorem says that an $n$-ary multilinear functor $F\colon A^n \to B$ into a category with cokernels extends to exactly one right exact functor $\widehat{F}\colon \mathcal{A}(A)^n \to B$, given on formal cokernels as $\widehat{F}(A_1,\dots,A_n) = \operatorname{coker}\big((F(\mathrm{id}_{A_i}; \rho_{A_i}))_i\big)$. Because this is an equivalence of functor categories, natural transformations — and hence associators, unitors, braidings, and the coherence identities they must satisfy — are uniquely determined by their restrictions to the embedded objects of $A$. The paper packages the input as a finitely presented promonoidal structure and lets this universal property carry every coherence proof: each identity on $\mathcal{A}(A)$ reduces to its restricted version on $A$, and the extensions are given by explicit cokernel constructions that a computer can execute.

What would settle it

A concrete test: take $R = \mathbb{Q}[x]$, $A = \operatorname{Rows}_R$, and the protensor $T(A,B) = A \otimes_R B$; build the associator by Construction 5.4.6 and check the pentagon identity on four finitely presented non-projective modules such as $\mathbb{Q}[x]/(x)$, $\mathbb{Q}[x]/(x^2)$, $\mathbb{Q}[x]/(x-1)$, and $\mathbb{Q}[x]/(x^2+x+1)$ — any failure of the pentagon diagram in $R\text{-}\mathrm{fpmod}$ would disprove Lemma 3.3.2. A second test for Theorem 4.2.1: pick a non-representable finitely presented functor $F$ and a functor $G$, compute the Day convolution $F \ast_P G$ from the defining coend, and compare it with the lifted tensor $F \,\widehat{\otimes}_T\, G$; the paper predicts a natural isomorphism for all such pairs.

Watch

Extended reading notes

Core claim

The paper's central claim is that finitely presented promonoidal structures on an additive category are exactly the restrictions to the embedded $A$ of right exact monoidal structures on the Freyd category $\mathcal{A}(A)$, and that the extension is canonical. Concretely, a finitely presented proassociator $\Pi$ extends to an associator $\widehat{\Pi}$, finitely presented prounitors extend to unitors, and a finitely presented probraiding extends to a braiding (Lemmas 3.3.1 through 3.3.6); the pentagon, triangle, and hexagon identities hold on all of $\mathcal{A}(A)$ exactly when their restrictions to $A$ hold. The engine is Theorem 2.3.1, a multilinear 2-categorical universal property: $n$-ary multilinear functors from $A^n$ to a category with cokernels correspond bijectively to right exact $n$-ary functors from $\mathcal{A}(A)^n$, so every natural transformation between right exact multifunctors is determined by its values on $A$. Theorem 4.2.1 then shows the lifted tensor product, viewed on $\mathrm{fp}(A^{\mathrm{op}}, \mathrm{Ab}) \simeq \mathcal{A}(A)$, is naturally isomorphic to the Day convolution built from the profunctor $P(a,b,c) = T(a,b)(c)$ and restricted to finitely presented functors; the paper's constructions are therefore the finitely presented shadow of Day convolution, obtained without coends.

Load-bearing premise

The load-bearing premise is the constructive equivalence between the Freyd category $\mathcal{A}(A)$ and the category $\mathrm{fp}(A^{\mathrm{op}}, \mathrm{Ab})$ of finitely presented functors (Theorem 2.1.6): every claim about tensor products is proved in $\mathcal{A}(A)$ and transferred to finitely presented functors through this equivalence, so if the equivalence were not constructive the claimed computable tensor products would not follow; a secondary standing premise is that $A$ is small, needed for the Day-convolution coends in Section 4.

Editorial extensions

If this is right

Monoidal structures propagate through the hierarchy of iterated Freyd categories: if $A$ is additive, closed monoidal, and has weak kernels and weak cokernels, then so does every iterated Freyd category of $A$ (Theorem 5.2.1), and every promonoidal structure on $A^{\mathrm{op}}$ yields a right exact monoidal structure on the free abelian category of $A$ (Theorem 5.3.1).
The paper's tensor product on finitely presented functors is computable: Section 5.4 gives explicit formulas for tensoring morphisms, unitors, associators, and braidings purely in terms of the protensor data, which is exactly the algorithmic content needed for computer implementation.
For finitely presented modules over a ring $R$, the construction recovers the usual tensor product, and Example 5.1.1 shows why the protensor $T$ genuinely takes values in $\mathcal{A}(A)$ rather than in $A$: tensoring two row modules with a fixed module $M$ can land outside the row modules when $M$ is not a row module.
The lifted monoidal structure can fail to be closed even when the original category is closed: Example 5.1.2 exhibits a closed monoidal category without weak kernels, the ring $\mathbb{Q}\langle x_i, z\rangle/\langle x_i z\rangle$, whose induced tensor product on the Freyd category has no right adjoint, so internal homs must be checked separately.
On finitely presented functors, the Day-convolution comparison (Theorem 4.2.1) means the f.p. promonoidal data determine the associator, unitors, and braiding of the restricted Day convolution up to natural isomorphism, so the coherence of the convolution on f.p. functors is governed entirely by representable values.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

My inference: the restricted-coherence strategy should extend to any algebraic structure whose defining identities are equalities of natural transformations between right exact multilinear functors, for example Hopf-like structures or module actions, since Theorem 2.3.1 reduces each such identity to its values on the embedded base category $A$.
My inference: Theorem 4.2.1 suggests a 'finitely presented Day convolution' computed directly on $\mathrm{fp}(A^{\mathrm{op}}, \mathrm{Ab})$: given finite presentations of $F$ and $G$, the defining coend collapses to a finite colimit, which would make Day convolution executable over bases where internal homs fail.
My inference: one testable consequence of the cokernel-based lifting is an invariance statement the paper leaves implicit: different small generating categories inside the same Freyd category that present the same finitely presented functors should induce naturally isomorphic right exact tensor products.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Colleague,

The one thing you should know: this paper has a promising framework for right exact tensor products on fp(A^op, Ab) via Freyd categories, but the proof of the central technical tool, Theorem 2.3.1, relies on Lemma 2.2.2, and that lemma as stated is not a right-exactness criterion. For n=1 it would say F(coker α) is a subobject of F(a), which is left exactness. The sequence should run ⊕_j F(a_{n−j}; b_j) → F(a_n) → F(coker α_n) → 0. The lemma is used in Lemma 2.3.3 and Lemma 2.3.5, and the lifting results in Section 3 inherit the flaw. So the main theorem—that every f.p. promonoidal structure extends to a right exact monoidal structure—is not established as written.

That said, the paper is not a waste. The notion of f.p. promonoidal structure is natural, the explicit lifting constructions are useful, and the Day convolution comparison in Theorem 4.2.1 is honest and illuminating. The CAP implementation and concrete examples in Section 5 give the work practical value. The citation practice is fine: the self-citations to Posur's constructive Freyd categories are appropriate and not inflated.

The soft spots are concentrated in Sections 2.2–2.3. Besides the lemma statement, the proof of Lemma 2.2.2 is sketched as a 'diagram chase' without enough detail. Theorem 3.3.8 relies on weak kernels and is not fully constructive, which contrasts with the paper's advertised constructivity. If Lemma 2.2.2 is fixed—and I think it can be, since the intended criterion is standard—the rest of the formal development may well be sound.

My bottom line: this deserves serious refereeing, but not acceptance in its current form. Have the authors correct the lemma and re-check Section 2.3, and this could be a solid contribution.

Referee Report

2 major / 5 minor

Summary. The paper develops a constructive theory of right exact tensor products on the category of finitely presented functors fp(A^op,Ab), working mainly in the equivalent language of Freyd categories A(A). The central technical tool is a multilinear 2-categorical universal property of Freyd categories (Theorem 2.3.1). Using this property, the authors define finitely presented promonoidal structures on an additive category A and show that they extend to right exact monoidal, braided monoidal, and closed structures on A(A) (Section 3.3). They then compare the resulting tensor product with Day convolution restricted to finitely presented functors (Theorem 4.2.1). The final sections give applications to finitely presented modules, iterated Freyd categories, free abelian categories, and explicit constructive implementations of the tensor product, associator, unitors, and braiding.

Significance. If the proofs are repaired, the paper would provide a useful unified and computational framework for constructing right exact tensor products on categories of finitely presented functors. Its emphasis on explicit constructions and on implementability in the CAP project is a genuine strength, as is the connection established with Day convolution. The main theorems are standard in spirit and likely correct, but the proof of the central universal property contains a load-bearing technical error, so the present version is not acceptable without revision.

major comments (2)

[Lemma 2.2.2; used in Lemmas 2.3.3 and 2.3.5] The exactness criterion stated in Lemma 2.2.2 is not a criterion for right exactness. For n=1 it asserts exactness of 0 -> F(coker alpha) -> F(a) -> F(b), which says that F(coker alpha) is the kernel of F(a) -> F(b); right exactness, as defined in Definition 2.2.1, instead requires the sequence F(b) -> F(a) -> F(coker alpha) -> 0 to be exact. The proof of Lemma 2.3.3 checks the stated left-exact sequence, and Lemma 2.3.5 explicitly invokes Lemma 2.2.2, so the proof of Theorem 2.3.1 and all subsequent lifting results (Lemmas 3.3.2, 3.3.4, 3.3.6, and Theorem 4.2.1) do not go through as written. The lemma should be replaced by the correct right-exactness criterion, e.g. exactness of the sequence bigoplus_j F(a^{n-j}; b_j) -> F(a^n) -> F(coker alpha_n) -> 0, or right exactness of the extended functor should be proved directly from Definition 2.2.1.
[Construction 2.3.2 and Lemma 2.3.3] There is an apparent direction mismatch in the presentation of the extended functor. If objects of A(A) are written as A = (a <- rho_a r_a) with rho_a: r_a -> a, then A is the cokernel of rho_a, and a right exact extension should be presented as coker( bigoplus_j F(a^{n-j}; r_{a_j}) -> F(a^n) ). The displayed formula in Construction 2.3.2 and the exact rows 0 -> widehat F(A^n) -> F(a^n) -> bigoplus_j F(a^{n-j}; r_{a_j}) instead present widehat F(A^n) as a kernel. This direction error is consistent with the incorrect statement of Lemma 2.2.2 and should be corrected in the same revision.

minor comments (5)

[Definition 2.1.1] In Definition 2.1.1 the category is written as ApPq in two places; this should be ApAq.
[Lemmas 3.3.1 and 3.3.5] The quantifiers in Lemmas 3.3.1 and 3.3.5 say 'for all a,b,c in ApAq' and 'for all a,b in ApAq', but the proassociator and probraiding are only defined for objects of A; these should read 'for all a,b,c in A' and 'for all a,b in A'.
[Example 5.1.2, Equation (9)] The displayed description of Hom_R(R/<z>, R) in Equation (9) is typeset in a garbled way and should be re-set so that the reader can see the intended R-module structure.
[Lemma 2.2.2 proof] Even after correcting the statement, the proof of the converse direction is only described as a diagram chase; since this lemma is load-bearing, a fuller proof should be supplied for the corrected criterion.
[Theorem 5.2.1] The proof is labeled 'Proof by induction' and is very terse; the induction step for the closed monoidal structure on ApXq should at least cite the relevant lemmas from Section 3.3 explicitly.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the lifting theorems are proved from explicit constructions, and the Day-convolution comparison is an external cross-check.

full rationale

The paper's central claim is that every finitely presented promonoidal structure on an additive category A extends to a right exact monoidal structure on the Freyd category A(A). This is not circular: the f.p. promonoidal data (protensor product, proassociator, prounitors, probraiding) are genuine inputs, and the paper proves the multilinear universal property of Freyd categories from explicit cokernel constructions in Construction 2.3.2, Lemma 2.3.3, and Lemma 2.3.5, rather than importing it as a black box. The lifting lemmas in Section 3.3 then verify coherence identities on all of A(A) by reducing them to their restrictions to A via that proved universal property; this is a theorem about the extension, not a restatement of the definition. The equivalence A(A) ≃ fp(A^op, Ab) is standard and is used only to translate the setting, not to force the tensor product. Self-citations such as [Pos17] and [BP19] concern constructiveness and software implementation, and they are not load-bearing for the mathematical derivation. Theorem 4.2.1 compares the constructed tensor product with Day convolution, an independent and externally established construction; the comparison is proved by checking representables, not assumed. The skeptic's objection that Lemma 2.2.2 misstates right exactness is a mathematical correctness concern about the proof as written, but it is not a circularity: a false or misstated lemma would invalidate the derivation, yet it would not make the conclusion identical to the assumptions or reduce a prediction to a fit. Therefore no circular step is exhibited, and the appropriate score is 0.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

The central claim rests on standard category theory and two cited theorems about Freyd categories. The paper introduces the definitions of f.p. promonoidal structures, but these are explicit definitions supported by constructions and examples, not unverified postulates. No free parameters are fitted to data.

assumptions (5)

domain assumption A is a small additive category.
Stated in the introduction and used throughout; smallness is needed for the coends in the Day convolution of Section 4 and to keep functor categories well-defined.
domain assumption The target category B in the universal property is additive and has cokernels.
Required for Theorem 2.3.1 and for the extension of tensor products; in the main application B is the Freyd category A(A), which has cokernels by Construction 2.1.4.
domain assumption Equivalence A(A) isomorphic to fp(A^op, Ab) of Theorem 2.1.6.
Cited from Freyd and Posur; the paper expresses all results on finitely presented functors in the language of Freyd categories, so this equivalence is load-bearing.
domain assumption Freyd's theorem: A(A) is abelian if and only if A has weak kernels.
Cited from Freyd as Theorem 2.1.5; used in Theorem 3.3.8 to construct the internal hom as a kernel.
standard math Standard category theory: Yoneda lemma, coend calculus, left exactness of Hom.
Used in the proof of Theorem 3.3.8 and throughout the Day convolution comparison in Section 4 without further proof.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Tensor products of finitely presented functors." pith.science (2026). https://pith.science/paper/XJZJQAQW

@misc{pith2026190900172,
  author       = {Pith},
  title        = {Pith review of: Tensor products of finitely presented functors},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XJZJQAQW}},
  note         = {Machine review of arXiv:1909.00172}
}

read the original abstract

We study right exact tensor products on the category of finitely presented functors. As our main technical tool, we use a multilinear version of the universal property of so-called Freyd categories. Furthermore, we compare our constructions with the Day convolution of arbitrary functors. Our results are stated in a constructive way and give a unified approach for the implementation of tensor products in various contexts.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

13 extracted references · 12 canonical work pages

[1]

Pure Appl

Murray Adelman, Abelian categories over additive ones, J. Pure Appl. Algebra 3 (1973), 103--117. 0318265

work page 1973
[2]

2 (2000), 147--185

Apostolos Beligiannis, On the F reyd categories of an additive category , Homology Homotopy Appl. 2 (2000), 147--185. 2027559

work page 2000
[3]

Luca Barbieri-Viale , Annette Huber , and Mike Prest , Tensor structure for Nori motives , arXiv e-prints (2018), arXiv:1803.00809

work page Pith review arXiv 2018
[4]

Martin Bies and Sebastian Posur, FreydCategories , https://github.com/homalg-project/CAP_project/tree/master/FreydCategoriesForCAP, 2019

work page 2019
[5]

137, Springer, Berlin, 1970, pp

Brian Day, On closed categories of functors, Reports of the M idwest C ategory S eminar, IV , Lecture Notes in Mathematics, Vol. 137, Springer, Berlin, 1970, pp. 1--38. 0272852

work page 1970
[6]

II , Category S eminar ( P roc

Brian Day, On closed categories of functors. II , Category S eminar ( P roc. S em., S ydney, 1972/1973), Springer, Berlin, 1974, pp. 20--54. Lecture Notes in Math., Vol. 420. 0360747 (50 \#13194)

work page 1972
[7]

Peter Freyd, Representations in abelian categories, Proc. C onf. C ategorical A lgebra ( L a J olla, C alif., 1965), Springer, New York, 1966, pp. 95--120. 0209333

work page 1965
[8]

Sebastian Gutsche and Sebastian Posur, Cap: categories, algorithms, programming, Computeralgebra-Rundbrief , 64, 14--17, March, 2019, pp. 14--17

work page 2019

Show all 13 references

[9]

Hasan, M

Sebastian Gutsche, Sebastian Posur, and Øystein Skartsæterhagen, On the syntax and semantics of CAP , In: O. Hasan, M. Pfeiffer, G. D. Reis (eds.): Proceedings of the Workshop Computer Algebra in the Age of Types,Hagenberg, Austria, 17-Aug-2018

2018
[10]

Sebastian Gutsche, Øystein Skartsæterhagen, and Sebastian Posur, The CAP project -- C ategories, A lgorithms, P rogramming , (http://homalg-project.github.io/CAP_project), 2013--2018

2013
[11]

5, Springer-Verlag, New York, 1998

Saunders Mac Lane, Categories for the working mathematician, second ed., Graduate Texts in Mathematics, vol. 5, Springer-Verlag, New York, 1998. 1712872

1998
[12]

Sebastian Posur, A constructive approach to Freyd categories , ArXiv e-prints (2017), ( https://arxiv.org/abs/1712.03492 arXiv:1712.03492 )

2017 arXiv
[13]

121, Cambridge University Press, Cambridge, 2009

Mike Prest, Purity, spectra and localisation, Encyclopedia of Mathematics and its Applications, vol. 121, Cambridge University Press, Cambridge, 2009. 2530988

2009

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [1]

Pure Appl

Murray Adelman, Abelian categories over additive ones, J. Pure Appl. Algebra 3 (1973), 103--117. 0318265

work page 1973

[2] [2]

2 (2000), 147--185

Apostolos Beligiannis, On the F reyd categories of an additive category , Homology Homotopy Appl. 2 (2000), 147--185. 2027559

work page 2000

[3] [3]

Luca Barbieri-Viale , Annette Huber , and Mike Prest , Tensor structure for Nori motives , arXiv e-prints (2018), arXiv:1803.00809

work page Pith review arXiv 2018

[4] [4]

Martin Bies and Sebastian Posur, FreydCategories , https://github.com/homalg-project/CAP_project/tree/master/FreydCategoriesForCAP, 2019

work page 2019

[5] [5]

137, Springer, Berlin, 1970, pp

Brian Day, On closed categories of functors, Reports of the M idwest C ategory S eminar, IV , Lecture Notes in Mathematics, Vol. 137, Springer, Berlin, 1970, pp. 1--38. 0272852

work page 1970

[6] [6]

II , Category S eminar ( P roc

Brian Day, On closed categories of functors. II , Category S eminar ( P roc. S em., S ydney, 1972/1973), Springer, Berlin, 1974, pp. 20--54. Lecture Notes in Math., Vol. 420. 0360747 (50 \#13194)

work page 1972

[7] [7]

Peter Freyd, Representations in abelian categories, Proc. C onf. C ategorical A lgebra ( L a J olla, C alif., 1965), Springer, New York, 1966, pp. 95--120. 0209333

work page 1965

[8] [8]

Sebastian Gutsche and Sebastian Posur, Cap: categories, algorithms, programming, Computeralgebra-Rundbrief , 64, 14--17, March, 2019, pp. 14--17

work page 2019