Mixed-norm Brascamp-Lieb inequalities

Ruixiang Zhang

REVIEW 2 major objections 6 minor 41 references

Mixed-norm Brascamp-Lieb inequalities

T0 review · 2 major / 6 minor · reviewed 2026-08-27 · deepseek-v4-flash

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

Pith's one-line read A new mixed-norm Brascamp-Lieb inequality is proved at the endpoint, via a transference from discrete counting to analytic norms.

desk verdict A solid, honest paper that sets up mixed-norm Brascamp-Lieb, proves a new endpoint quadrilinear inequality via a transference principle, and connects to Kakeya; main caveat is a black-boxed counting estimate. read the letter →

arxiv 2608.17952 v2 pith:YMQZTGRF submitted 2026-08-18 math.CA math.CO

Ruixiang Zhang This is my paper · ORCID

classification math.CAmath.CO MSC 42B2526D15

keywords mixed-norminequalitiesBrascamp-LiebrestrictedweaktypeestimatestransferenceprincipletensorproducttrickKakeyaconjecturearithmeticprojectionsendpoint

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

This paper introduces mixed-norm Brascamp-Lieb inequalities as a subject of their own: inequalities of the form $\| \prod_{j=1}^m f_j(B_j(x,y))^{p_j}\|_{L^q_x L^r_y}\lesssim \prod \|f_j\|_1^{p_j}$ in which the left-hand side is measured in two different $L^p$ norms, one in $x$ and one in $y$. It argues that the case $q<2$ is a genuinely new regime, invisible to classical Brascamp-Lieb theory, and it proves one sharp quadrilinear example at the endpoint: Theorem 1.1, with $q=7/4$ and $r=7/3$. The proof is a two-part mechanism: a restricted weak type estimate borrowed from arithmetic projection counting, upgraded by a tensor product trick that removes the interpolation loss. A reader should care because the same machinery may yield many new inequalities, and because the paper connects the $q\to 1$ limit of such bounds to the Hausdorff version of the Kakeya conjecture.

What carries the argument

The load-bearing object is the mixed-norm Brascamp-Lieb inequality (5): $\| \prod_{j=1}^m f_j(B_j(x,y))^{p_j}\|_{L^q_x L^r_y}\le C\prod_{j=1}^m \|f_j\|_1^{p_j}$, with $\frac1q+\frac1r=1$ and $\sum_j p_j=1$, where each $B_j$ is a projection to a one-dimensional rational subspace of $\mathbb{R}^2$ not equal to the $x$-axis. The new mechanism that carries the argument is Theorem 3.3, a transference principle: if the discrete restricted weak type counterpart holds for every torsion-free abelian group uniformly, then (5) holds over $\mathbb{R}$. The proof uses a tensor product trick -- applying the logarithm-lossy discrete estimate to $N$-fold tensor products and letting $N\to\infty$ removes the logarithm factor. The discrete input is Lemma 4.2, a refinement of Katz-Tao's arithmetic projection estimate that counts 4-tuples satisfying the four linear constraints $\pi_{1,1},\pi_{-1,1},\pi_{0,1},\pi_{1,3}$, and the paper includes a self-contained proof of that lemma.

What would settle it

To test the central claim, search over finite subsets $S_1,S_2,S_3,S_4\subset\mathbb{Z}$ and a set $Z\subset\mathbb{Z}^2$ satisfying the four linear constraints $\pi_{1,1}(z)\in S_1,\pi_{-1,1}(z)\in S_2,\pi_{0,1}(z)\in S_3,\pi_{1,3}(z)\in S_4$ and the fiber bound from condition (ii) of Lemma 4.2, and check whether $|Z|\le C m^{1/4}|S_1|^{1/2}|S_2|^{1/2}|S_3|^{1/2}|S_4|^{1/4}$ fails for a fixed constant $C$ as $m$ grows. Any such failure would falsify the discrete estimate that Theorem 1.1 depends on; alternatively, a direct check of the number of 4-tuples produced by the counting step in the proof of Lemma 4.2 would show whether the lower bound and the upper bound match.

Watch

Extended reading notes

Core claim

The central claim is Theorem 1.1: for nonnegative functions $f_1,f_2,f_3,f_4:\mathbb{R}\to\mathbb{R}_{\ge 0}$, $$\left\| f_1(x+y)^{2/7} f_2(-x+y)^{2/7} f_3(y)^{2/7} f_4(x+3y)^{1/7}\right\|_{$L^{{7/4}}$_x $L^{{7/3}}$_y}\lesssim \|f_1\|$_1^{{2/7}}$\|f_2\|$_1^{{2/7}}$\|f_3\|$_1^{{2/7}}$\|f_4\|$_1^{{1/7}}$.$$ The paper obtains this by first proving a restricted weak type estimate (Lemma 4.2) for the same four linear forms on a discrete torsion-free abelian group, with a uniform implied constant. It then invokes a new transference principle (Theorem 3.3): a uniform discrete restricted weak type bound implies the honest analytic mixed-norm inequality, via tensorizing the functions and letting the tensor power tend to infinity to eliminate logarithmic losses. The same transference shows that Christ's earlier positive results in the $q<2$ regime are the first examples of the general mixed-norm phenomenon, and that the endpoint that interpolation left open is now covered.

Load-bearing premise

The proof of Lemma 4.2 takes Katz-Tao's counting lemmas (their Lemma 2.1 and the deductions to (13) and (18)) as black boxes; if those combinatorial inputs fail for torsion-free abelian groups without order-2 elements, or if their constants are not uniform, then the restricted weak type estimate and hence Theorem 1.1 would not follow from the argument.

Editorial extensions

If this is right

Christ's trilinear results are re-read as the first positive cases of the $q<2$ mixed-norm regime, and the endpoint of that regime is now proved for the quadrilinear example.
To prove new analytic mixed-norm Brascamp-Lieb inequalities it suffices to prove discrete restricted weak type estimates with uniform implied constants, which is a combinatorial problem.
If, for every $\alpha>1$, some instance of (5) holds with $q<\alpha$, then every Kakeya set in $\mathbb{R}^d$ has Hausdorff dimension at least $d$ (Theorem 5.3).
The perturbed version of (5) fails: individual tube directions cannot be wiggled independently while preserving a mixed-norm bound, so the methods that prove stability of the classical Brascamp-Lieb inequality do not transfer.

Reading between the lines

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

The same tensor product transference should work for other configurations of projections, so any Katz-Tao type counting estimate with uniform constants is a candidate for a new endpoint mixed-norm inequality.
If the Kakeya connection is taken at face value, Theorem 1.1's $q=7/4$ is far from the conjectured threshold $q\to 1$; reaching that threshold would require combinatorial bounds far stronger than the current arithmetic projection method.
The negative perturbed result suggests that valid $q<2$ inequalities are sensitive to exact rational alignment of the projections, and that generalizing them to approximate directions will need new ideas.
A testable extension is to verify the transference principle for $m>4$ factors or unequal exponents $p_j$; the uniform-constant hypothesis is the only essential input, so the same proof should go through.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Request a human review

A listed scientist reviews the paper for a fee and the review publishes here regardless of verdict. See the reviewers or get listed.

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

Reading this paper, the main thing to know: there is a real new framework here, and the flagship inequality Theorem 1.1 is probably right, but the proof rests on a counting estimate that is cited rather than proved, so I would want that checked before relying on it. The paper is not a trivial repackaging of Christ; it adds the transference principle (Theorem 3.3) that upgrades discrete restricted weak type estimates to honest mixed-norm inequalities via a tensor product trick, and uses it to get the endpoint case of Theorem 1.1. The algebraic core of Lemma 4.2 (Claims I and II) is correct; I checked the manipulation. The counting lower bound, however, is outsourced to Lemma 2.1 of [KT99] "just like the deduction of (18)", and the paper does not reproduce the argument or verify the uniformity in torsion-free groups. This is the load-bearing step for Theorem 1.1, so it is not a footnote issue. On the other hand, it is a known tool from a published paper, and the style of the field routinely cites such counting lemmas; I would describe it as a repairable gap, not a fatal flaw, provided [KT99]'s lemma applies verbatim. The stress-test is right to flag it, but also right to note that the rest of the argument checks out.

Smaller things: Theorem 5.3 and Appendix A have the dimension exponent misprinted; as written "d - 1/q + 1" exceeds d. It should be (d-1)/q + 1. That is purely typographical but confusing. The appendix is explicitly a proof sketch, so the Hausdorff Kakeya claim is not fully verified in this paper; fine as a remark, but not a theorem I would quote without checking against [Bou99]. The negative perturbed result (Theorem 5.8) is clean and necessary, and the framing of Conjecture 5.4 is honest about what is open. Citation practice is fine: Christ, Katz-Tao, Bourgain, and recent Szemeredi work are all credited; self-citations are to the author's own survey, which is appropriate there. No circularity detected and no invented entities.

For peer review: I would send this to a serious referee. The framework is useful, the endpoint inequality is a genuine advance over Christ's non-endpoint result, and the flaws are repairable. I would not desk reject. My own verdict would be conditional accept with the request that the Katz-Tao input be either proved or stated with exact hypotheses.

Referee Report

2 major / 6 minor

Summary. The paper introduces a framework of mixed-norm Brascamp-Lieb inequalities in R^2 with one-dimensional targets, under the normalizations sum p_j=1 and 1/q+1/r=1. Its main concrete result, Theorem 1.1, is the quadrilinear estimate || f1(x+y)^{2/7} f2(-x+y)^{2/7} f3(y)^{2/7} f4(x+3y)^{1/7} ||_{L^{7/4}_x L^{7/3}_y} ≲ ||f1||_1^{2/7} ||f2||_1^{2/7} ||f3||_1^{2/7} ||f4||_1^{1/7} for nonnegative functions on R. The proof combines a restricted weak type estimate (Lemma 4.2) with a transference principle (Theorem 3.3) that upgrades discrete restricted weak type bounds to the full mixed-norm inequality by a tensor product trick. The paper also connects q<2 mixed-norm Brascamp-Lieb inequalities to Kakeya conjectures (Theorem 5.3) and proves that a natural perturbed version of the inequality cannot hold (Theorem 5.8).

Significance. If the proof is completed, the endpoint Theorem 1.1 would be a genuine new mixed-norm Brascamp-Lieb inequality in a regime where the classical Brascamp-Lieb theory gives no information, and the tensor product transference (Theorem 3.3) is a clean and potentially reusable idea. The negative result Theorem 5.8 is explicit and correctly separates the mixed-norm problem from the classical perturbed Brascamp-Lieb setting. The paper is refreshingly honest about its reliance on external counting arguments, but this reliance is currently the main obstruction to accepting the central claim.

major comments (2)

[Section 4, proof of Lemma 4.2] The proof of the restricted weak type estimate is incomplete at its most load-bearing step. The lower bound on the number of admissible 4-tuples, stated as 'there are ≥ |Z|^4/(|S_1|^2|S_2||S_3|) tuples...', is justified only by a reference to 'Lemma 2.1 in [KT99] again just like the deduction of (18) in [KT99]'. This lower bound is what produces the exponents in (21), and (21) is what feeds Theorem 3.3 to prove Theorem 1.1. Because the paper claims to include a proof of Lemma 4.2 'for self-containment', the author should state Katz-Tao's Lemma 2.1 explicitly, prove the deduction in the required generality (in particular for torsion-free abelian groups, with constants uniform in G), and either prove or clearly attribute the counting estimate. Without this, Theorem 1.1 is not proved by the argument in the manuscript.
[Appendix A / Theorem 5.3] Theorem 5.3 is stated as a theorem, but the proof of the key implication sends the quantitative argument to Appendix A, which is presented only as a sketch. The two critical moves — that the recent quantitative Szemerédi results [LSS24b] provide arithmetic progressions long enough for an arbitrary fixed N_0, and that SD(α) can replace Bourgain's Lemma 2.83 in the concatenation argument — are asserted without verification. If Theorem 5.3 is to remain a theorem of the paper, this appendix needs a complete proof; otherwise the claim should be downgraded to a conditional observation.

minor comments (6)

[Section 4, Proposition 4.1] The hypothesis 'with every two elements in S having different first coordinates' appears to be a typo; it should refer to elements of Z, since S is a list of four subsets of G, not a subset of G^2.
[Section 4, proof of Lemma 4.2] In Claim I, after deriving 2x_1 from the data, the text says 'divide it by 2'. This uses injectivity of the doubling map in a group without order 2 elements; the inference is correct but should be stated explicitly.
[Section 3, Lemma 3.2] In the dyadic step, the sentence 'there are ∼ b ∏ c_j^{-p_j r} different y...' is essential for matching condition (ii) of Definition 3.1; making the '∼' precise with constants depending only on m and the exponents would improve readability.
[Section 3, Definition 3.1] The phrase 'every element in π_{1,0}(Z) has ≤m preimages in Z with respect to the map π_{1,0}(Z)' should read 'with respect to the map π_{1,0}: G^2 → G'.
[Section 5.2, proof of Theorem 5.3] The reduction from (5) to SD(q) for Z should specify the dependence of the implicit constant on the data {H_j, p_j}; otherwise the finite-set embedding to R^{d-1} may introduce a dimension-dependent constant.
[References] The paper relies on [Tao], an unpublished note, for an existing version of Lemma 4.2; please provide a stable reference or include the precise entropy inequality statement in the paper.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: Theorem 1.1 is derived from an external restricted weak type estimate plus a transference argument; no target quantity is used as its own input.

full rationale

The derivation chain for Theorem 1.1 is: Lemma 4.2 supplies a restricted weak type counting estimate for finite subsets of an abelian group without order-2 elements; Theorem 3.3 converts such estimates, when they hold with uniform implied constants, into genuine mixed-norm inequalities via dyadic pigeonholing and the tensor product trick; Theorem 1.1 then follows by applying Theorem 3.3 to the discrete estimate. The restricted weak type assumption (12) is not the target inequality (5) in disguise: it is a separate counting bound on |Z|, and the tensor product step proves the full l^1 estimate (13) from (12) without assuming (13). Lemma 4.2 itself is justified by the elementary Claims I and II together with the external Katz-Tao counting input from [KT99]; the cited sources are prior work by other authors, not load-bearing self-citations. The paper's self-citations, such as [Zha22], [Zha17], and [ZK20], are contextual or concern related results that are not needed for the main endpoint inequality. The reliance on [KT99] as a black box and the apparent typo in the statement of Proposition 4.1 are correctness and verification concerns, but they are not instances of definitional circularity, renamed predictions, or self-citation chains. No fitted parameter is renamed as a prediction, and no target quantity is built into its own hypothesis. The derivation is therefore self-contained in the sense relevant to circularity.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

No fitting or invented objects. The paper's dependencies are published theorems and standard analytic tools. The only questionable input is the unproved invocation of Katz-Tao's Lemma 2.1 in the proof of the central estimate.

assumptions (4)

standard math Bennett-Carbery-Christ-Tao theorem: the classical Brascamp-Lieb inequality (2) holds iff the scaling and dimension conditions are satisfied (Theorem 1.2).
Used in Section 2 to identify the q<2 regime as genuinely new and to motivate the mixed-norm question. It is a published theorem, not derived here.
standard math Katz-Tao Lemma 2.1 and the derivations of (13) and (18) in [KT99], used as black boxes in the proof of Lemma 4.2.
The counting lower bound for tuples in Lemma 4.2 is obtained by 'Applying Lemma 2.1 in [KT99] again just like the deduction of (18) in [KT99]' without reproducing the argument or its constant.
standard math Quantitative Szemerédi theorem bounds from [LSS24b] (and predecessors [GT09, GT17, LSS24a]), used to convert arithmetic projection properties into Hausdorff dimension bounds for Kakeya sets (Appendix A).
The appendix uses the density bound from [LSS24b] as an input to Bourgain's argument; the precise form of the bound is not proved or restated.
standard math Bourgain's Proposition 1.5 in [Bou99] and the general arithmetic-projection-to-Kakeya framework, assumed as a template for Theorem A.1.
The proof of Theorem A.1 is a modification of Bourgain's argument; the underlying structural reduction is taken from that paper.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Mixed-norm Brascamp-Lieb inequalities." pith.science (2026). https://pith.science/paper/YMQZTGRF

@misc{pith2026260817952,
  author       = {Pith},
  title        = {Pith review of: Mixed-norm Brascamp-Lieb inequalities},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YMQZTGRF}},
  note         = {Machine review of arXiv:2608.17952}
}

read the original abstract

Christ [Chr01] looked at a certain boundedness problem for trilinear operators. In this paper we set up a framework of mixed-norm Brascamp-Lieb inequalities and identify a new interesting regime not covered by the study of classical Brascamp-Lieb. Positive results in [Chr01] can be viewed as the first nontrivial progress in this new regime. We will also prove another mixed-norm Brascamp-Lieb inequality, showcasing how one can use a tensor product trick to slightly sharpen Christ's argument and also obtain the endpoint case. We then discuss the connections between mixed-norm Brascamp-Lieb and the Kakeya Conjectures, as well as unique new difficulties for these mixed-norm Brascamp-Lieb inequalities compared to the classical setting.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 40 canonical work pages

[1]

Some special solutions of the

Barcel. Some special solutions of the. Indiana University mathematics journal , pages=. 2007 , publisher=

work page 2007
[2]

Guth, Larry and Iosevich, Alex and Ou, Yumeng and Wang, Hong , journal=. On. 2020 , publisher=

work page 2020
[3]

Annals of mathematics , pages=

The proof of the l^2 decoupling conjecture , author=. Annals of mathematics , pages=. 2015 , publisher=

work page 2015
[4]

Some sharp inequalities of

Carbery, Anthony and Iliopoulou, Marina and Wang, Hong , year=. Some sharp inequalities of

work page
[5]

2022 , journal=

An enemy scenario in restriction theory , author=. 2022 , journal=

work page 2022
[6]

An improved result for

Du, Xiumin and Iosevich, Alex and Ou, Yumeng and Wang, Hong and Zhang, Ruixiang , journal=. An improved result for. 2021 , publisher=

work page 2021
[7]

Bounds on arithmetic projections, and applications to the

Katz, Nets Hawk and Tao, Terence , journal=. Bounds on arithmetic projections, and applications to the. 1999 , publisher=

work page 1999
[8]

Ruixiang Zhang , journal=. The. 2022 , publisher=

work page 2022

Show all 41 references

[9]

On the multilinear restriction and

Bennett, Jonathan and Carbery, Anthony and Tao, Terence , journal=. On the multilinear restriction and. 2006 , publisher=

2006
[10]

Bennett, Jonathan and Carbery, Anthony and Christ, Michael and Tao, Terence , journal=. The. 2008 , publisher=

2008
[11]

Finite bounds for

Bennett, Jonathan and Carbery, Anthony and Christ, Michael and Tao, Terence , journal=. Finite bounds for. 2008 , publisher=

2008
[12]

2012 , publisher=

Interpolation spaces: an introduction , author=. 2012 , publisher=

2012
[13]

On the dimension of

Bourgain, Jean , journal=. On the dimension of. 1999 , publisher=

1999
[14]

New bounds for

Katz, Nets Hawk and Tao, Terence , journal=. New bounds for. 2002 , publisher=

2002
[15]

From rotating needles to stability of waves; emerging connections between combinatorics, analysis and

Tao, Terence , journal=. From rotating needles to stability of waves; emerging connections between combinatorics, analysis and
[16]

Wang, Hong and Zahl, Joshua , journal=. Sticky
[17]

Improved bounds for the

Hickman, Jonathan and Rogers, Keith M and Zhang, Ruixiang , journal=. Improved bounds for the. 2022 , publisher=

2022
[18]

Zahl, Joshua , journal=. New. 2021 , publisher=

2021
[19]

The p -adic

Arsovski, Bodan , journal=. The p -adic
[20]

On the arithmetic

Green, Ben and Ruzsa, Imre Z , journal=. On the arithmetic. 2019 , publisher=

2019
[21]

Recent progress on the

Katz, Nets and Tao, Terence , journal=. Recent progress on the
[22]

New bounds for

Green, Ben and Tao, Terence , journal=. New bounds for
[23]

New bounds for

Green, Ben and Tao, Terence , journal=. New bounds for. 2017 , publisher=

2017
[24]

Mathematical Proceedings of the Cambridge Philosophical Society , volume=

Improved bounds for five-term arithmetic progressions , author=. Mathematical Proceedings of the Cambridge Philosophical Society , volume=. 2024 , organization=

2024
[25]

Improved bounds for

Leng, James and Sah, Ashwin and Sawhney, Mehtaab , journal=. Improved bounds for
[26]

Stability of the

Bennett, Jonathan and Bez, Neal and Flock, Taryn C and Lee, Sanghyuk , journal=. Stability of the. 2018 , publisher=

2018
[27]

The endpoint case of the

Guth, Larry , journal=. The endpoint case of the. 2010 , publisher=

2010
[28]

A short proof of the multilinear

Guth, Larry , booktitle=. A short proof of the multilinear. 2015 , organization=

2015
[29]

The endpoint perturbed

Zhang, Ruixiang , journal=. The endpoint perturbed. 2017 , publisher=

2017
[30]

2020 , publisher=

Zorin-Kranich, Pavel , journal=. 2020 , publisher=

2020
[31]

The endpoint multilinear

Carbery, Anthony and Valdimarsson, Stef. The endpoint multilinear. Journal of Functional Analysis , volume=. 2013 , publisher=

2013
[32]

Mathematical Research Letters , volume=

On certain elementary trilinear operators , author=. Mathematical Research Letters , volume=. 2001 , publisher=

2001
[33]

Mathematical Research Letters , volume=

Multilinear estimates and fractional integration , author=. Mathematical Research Letters , volume=. 1999 , publisher=

1999
[34]

Mathematische Annalen , volume=

Some remarks on multilinear maps and interpolation , author=. Mathematische Annalen , volume=. 2001 , publisher=

2001
[35]

An improved bound on the

Katz, Nets and Zahl, Joshua , journal=. An improved bound on the
[36]

Can be found at "

Entropy sum set estimates , author=. Can be found at "
[37]

Collectanea Mathematica , pages=

Elementary proofs and the sums differences problem , author=. Collectanea Mathematica , pages=
[38]

arXiv preprint arXiv:2511.15135 , year=

Sum-difference exponents for boundedly many slopes, and rational complexity , author=. arXiv preprint arXiv:2511.15135 , year=
[39]

Wang, Hong and Zahl, Joshua , journal=. The. 2025 , publisher=

2025
[40]

Volume estimates for unions of convex sets, and the

Wang, Hong and Zahl, Joshua , journal=. Volume estimates for unions of convex sets, and the
[41]

Generalized Arithmetic

Pohoata, Cosmin and Zakharov, Dmitrii , journal=. Generalized Arithmetic

Pith tools

Reviewed August 27, 2026 · model on record in the stance chip above.

[1] [1]

Some special solutions of the

Barcel. Some special solutions of the. Indiana University mathematics journal , pages=. 2007 , publisher=

work page 2007

[2] [2]

Guth, Larry and Iosevich, Alex and Ou, Yumeng and Wang, Hong , journal=. On. 2020 , publisher=

work page 2020

[3] [3]

Annals of mathematics , pages=

The proof of the l^2 decoupling conjecture , author=. Annals of mathematics , pages=. 2015 , publisher=

work page 2015

[4] [4]

Some sharp inequalities of

Carbery, Anthony and Iliopoulou, Marina and Wang, Hong , year=. Some sharp inequalities of

work page

[5] [5]

2022 , journal=

An enemy scenario in restriction theory , author=. 2022 , journal=

work page 2022

[6] [6]

An improved result for

Du, Xiumin and Iosevich, Alex and Ou, Yumeng and Wang, Hong and Zhang, Ruixiang , journal=. An improved result for. 2021 , publisher=

work page 2021

[7] [7]

Bounds on arithmetic projections, and applications to the

Katz, Nets Hawk and Tao, Terence , journal=. Bounds on arithmetic projections, and applications to the. 1999 , publisher=

work page 1999

[8] [8]

Ruixiang Zhang , journal=. The. 2022 , publisher=

work page 2022