Brakke inequality and the existence of Brakke-flow for volume preserving mean curvature flow

Andrea Chiesa; Keisuke Takasao

REVIEW 3 major objections 4 minor 28 references

Brakke inequality and the existence of Brakke-flow for volume preserving mean curvature flow

T0 review · 3 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read This paper proves a volume-preserving Brakke inequality and global-in-time existence of integral varifolds satisfying it on the torus, via the phase-field method.

desk verdict A clever new Brakke-type inequality for volume-preserving MCF, but the proof omits a necessary estimate on the discrepancy measure. read the letter →

arxiv 2505.23222 v2 pith:DVY5KCFC submitted 2025-05-29 math.AP

Andrea Chiesa , Keisuke Takasao This is my paper

classification math.AP MSC 35K9353E10

keywords volumepreservingmeancurvatureflowBrakkeAllen-CahnequationphasefieldmethodvarifoldsL2-flowintegraltorus

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 proposes a volume-preserving analogue of Brakke's inequality for mean curvature flow and proves that, on the $d$-dimensional torus, the phase-field (Allen-Cahn) approximation produces a family of integral varifolds satisfying this inequality for all times. The new inequality is the classical Brakke inequality plus an extra error term $r^{d-1} C(1+t_2-t_1)\lVert\phi\rVert_{L^\infty}$ supported on balls of radius $r$, which absorbs the a priori uncontrollable contribution of the Lagrange multiplier $\lambda$ that enforces volume preservation. In the smooth setting the error term vanishes as $r\to 0$, so the inequality still determines the classical normal velocity. If correct, the result gives global-in-time weak solutions of volume-preserving mean curvature flow in a Brakke sense on the torus, strengthening the earlier $L^2$-flow existence result and aligning the constrained flow with the Brakke framework used for unconstrained mean curvature flow.

What carries the argument

The machinery is the Allen-Cahn phase-field equation (18) with the nonlocal volume-preserving term $-\lambda^\varepsilon\sqrt{2W(\varphi^\varepsilon)}/\varepsilon$, together with the surface-energy measures $\mu^\varepsilon_t$ and the discrepancy measure $\xi^\varepsilon=\frac{1}{\sigma}(\frac{\varepsilon\lvert\nabla\varphi^\varepsilon\rvert^2}{2}-\frac{W(\varphi^\varepsilon)}{\varepsilon})L^{d+1}$. The proof differentiates $\mu^\varepsilon_t(\phi)$ in time, obtaining (28), where the mean-curvature and velocity terms appear with signs that persist in the limit via lower semicontinuity, while the Lagrange-multiplier term is bounded by the density estimate. The discrepancy measure is asserted to converge to zero as Radon measures, which lets the approximate velocity measure $\tilde\mu^\varepsilon=\frac{\varepsilon}{\sigma}\lvert\nabla\varphi^\varepsilon\rvert^2 L^{d+1}$ be identified with the limiting varifold measure and produces the $-\lvert v\rvert^2/2$ dissipation.

What would settle it

Find an explicit sequence of solutions of the Allen-Cahn equation (18) on the torus, or a numerical simulation with initial data satisfying (16)-(17), for which $\xi^\varepsilon$ does not converge to zero as Radon measures; then the computation $\tilde\mu^\varepsilon\to\mu$ in the proof of Theorem 3.2 would fail, the inequality (11) would not follow, and the theorem as stated would be false.

Watch

Extended reading notes

Core claim

The central discovery is that a volume-preserving Brakke inequality is a viable weak formulation for the constrained flow: Definition 2.2 keeps the mean-curvature dissipation $-\lvert h\rvert^2/2$ and velocity dissipation $-\lvert v\rvert^2/2$ from the classical inequality, but replaces the unmanageable $\lvert\lambda\rvert^2$ term by the small error $r^{d-1}C(1+t_2-t_1)\lVert\phi\rVert_{L^\infty}$. The proof shows that the Allen-Cahn measures $\mu^\varepsilon_t$ from (20) converge to Radon measures $\mu_t$ that satisfy (11), using the approximate energy identity (28), weak lower semicontinuity for the $\lvert h^\varepsilon\rvert^2$ and $\lvert v^\varepsilon\rvert^2$ terms, and a density estimate $\mu^\varepsilon_t(B_r(x_0))\le c r^{d-1}$ plus the $L^2$-bound on $\lambda^\varepsilon$ to control the remainder.

Load-bearing premise

The proof depends on the unproved assertion that the discrepancy measure $\xi^\varepsilon$—the difference between the gradient-energy part and the potential part of the Allen-Cahn measure—tends to zero as a measure; if it does not, the approximate velocity need not be the velocity of the limiting varifold and the Brakke inequality's velocity term may disappear.

Editorial extensions

If this is right

If Theorem 3.2 is correct, the limiting family $\{\mu_t\}_{t\in[0,T)}$ is a volume-preserving Brakke-flow in the sense of Definition 2.2, giving global-in-time weak solutions on the torus.
Since every volume-preserving Brakke-flow is an $L^2$-flow (Proposition 2.1), the theorem upgrades the earlier $L^2$-flow existence result to a stronger dissipation-based notion.
In the smooth regime the extra error term vanishes in the local blow-up limit, so the new inequality characterizes the classical normal velocity just as the standard Brakke inequality does.
The $L^2$-bound (23) on the Lagrange multiplier $\lambda^\varepsilon$ and the density estimate $\mu^\varepsilon_t(B_r)\le c r^{d-1}$ are what make the problematic $\lvert\lambda\rvert^2$ term controllable.

Reading between the lines

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

A natural next step is to prove the missing convergence $\xi^\varepsilon\to 0$, for instance by showing that the gradient and potential parts of the Allen-Cahn energy become asymptotically balanced in $L^1$; if that convergence is established, the proof of Theorem 3.2 closes as written.
The same estimate-and-limsup strategy may transfer to other constrained or multiphase curvature flows, where a Lagrange multiplier produces an error term that can be absorbed by a density bound.
The $r^{d-1}$-error formulation suggests a quantitative 'almost-Brakke' inequality that might be useful for numerical schemes such as thresholding, where approximate solutions satisfy the inequality up to a controlled discretization error.
On a compact manifold like the torus, one might hope to localize or remove the spatial cut-off $B_r(x_0)$ in (11), yielding a cleaner inequality at the price of stronger compactness assumptions; this is not addressed in the paper.

Share X Bluesky LinkedIn Reddit HN

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

Dear colleague,

I've read the Chiesa–Takasao paper. The new thing is a volume-preserving version of Brakke's inequality that adds an r^{d-1} error term, which lets them bypass the nonlocal |lambda|^2 term that blocks the classical Brakke inequality. They then prove that the global L2-flow constructed in Takasao's earlier paper satisfies this inequality via the Allen–Cahn approximation. The r^{d-1} term is well chosen: it vanishes in the local-rescaling argument that characterizes smooth normal velocity, so the inequality is not vacuous.

The paper is mostly well executed. The estimate for the |lambda_epsilon|^2 term uses the density bound from [25] and is straightforward. The treatment of the h^2 and v^2 terms follows the standard lower-semicontinuity route. The writing is clear and the dependence on [25] is explicit, not circular.

The problem is in Section 3.2. They define e-mu_epsilon := (epsilon/sigma)|nabla phi_epsilon|^2 L^{d+1} and state e-mu_epsilon = mu_epsilon − xi_epsilon, where xi_epsilon := (1/sigma)(epsilon|nabla phi_epsilon|^2/2 − W(phi_epsilon)/epsilon)L^{d+1}. The correct relation is e-mu_epsilon = mu_epsilon + xi_epsilon. That's a harmless sign typo, because if xi_epsilon → 0 it makes no difference.

The real issue is the assertion, right after that, that xi_epsilon → 0 as Radon measures. This is not proved, not in the stated Theorem 3.1, and not cited. It is not a trivial fact: it is a sharp-interface estimate showing the difference of the kinetic and potential energy densities vanishes in the limit. Without it, the weak limit of e-mu_epsilon is unknown, and the −(1/2)∫phi|v|^2 d(mu) term in (11) does not follow. This is load-bearing. The rest of the proof is conditional on that statement.

I suspect the claim is true, and fixable with known tools, but it needs to be proved. The paper, as written, has a gap in its central theorem. That said, the definition is natural and the overall strategy is sound.

This paper deserves a serious referee, but it needs major revision. I'd engage with it.

Best

Referee Report

3 major / 4 minor

Summary. This paper introduces a weak notion of volume-preserving Brakke flow for varifolds, in which the problematic |λ|^2 term of the classical inequality is replaced by an error term r^{d-1}C(1+Δt)||ϕ||∞ for test functions supported in a ball of radius r. The authors prove, in the periodic setting, that the phase-field/Allen–Cahn approximation of volume-preserving mean curvature flow constructed in [25] satisfies this Brakke inequality, thereby upgrading the L^2-flow existence result of [25] to a Brakke-type existence theorem. The proof follows the standard Allen–Cahn strategy: derive an approximate dissipation identity (28), pass to the limit using known convergence results, and control the new |λ_ε|^2 term by the density estimate and the L^2-bound on λε from [25].

Significance. If established, Theorem 3.2 gives global-in-time integral varifold solutions to volume-preserving mean curvature flow in a Brakke sense on the torus, extending [25] and providing a varifold framework in which tools such as regularity and weak-strong uniqueness might be applied. The paper is careful to motivate the modified inequality and shows that it still characterizes the normal velocity in the smooth setting. The estimate for the |λε|^2 term is clean and correctly uses the density bound from [25]. The proof is not circular: it relies on prior results from [25] rather than on the desired conclusion. However, the proof contains a load-bearing gap concerning the discrepancy measure ξε, and an algebraic sign error in the definition of eµε, so the central claim is not presently fully justified.

major comments (3)

[Section 3.2, proof of Theorem 3.2] The displayed identity eµε = µε − ξε is false. With µε = (1/σ)(ε|∇φε|^2/2 + W(φε)/ε)L^{d+1} and ξε = (1/σ)(ε|∇φε|^2/2 − W(φε)/ε)L^{d+1}, the correct relation is eµε = µε + ξε, because ε|∇φε|^2/σ equals (1/σ)(ε|∇φε|^2/2 + W(φε)/ε) plus (1/σ)(ε|∇φε|^2/2 − W(φε)/ε). The stated minus sign gives µε − ξε = (2/σ)(W(φε)/ε)L^{d+1}. Please correct this identity.
[Section 3.2, proof of Theorem 3.2] The proof asserts, without proof or citation, that ξε → 0 as Radon measures. This convergence is not stated in Theorem 3.1 and is not established in the text. It is the key step that identifies the weak limit of eµε with µ, and it is then used, together with Hutchinson's lower-semicontinuity theorem [11, Thm. 4.4.2], to pass the liminf on the velocity term and produce the −ϕ|v|^2/2 term in (11). If ξε does not converge to zero, the weak limit of eµε is unidentified and the dissipation term for the velocity is lost, so inequality (11) would not follow. A proof of ξε → 0 from the available estimates, or a citation to a result that contains it, is required.
[Section 3.2, proof of Theorem 3.2] In passing to the limit in the boundary terms of (28), the proof invokes Theorem 3.1(a), but that convergence is stated only for times outside a countable exceptional set B. Since inequality (11) is asserted for all 0 ≤ t1 < t2 ≤ T, the proof should either choose t1 and t2 outside B and then approximate, or otherwise justify the passage to the limit at all times. As written, the inequality is only proven for times outside a countable set.

minor comments (4)

[Equations (28) and (29)] In equations (28) and (29), the symbol 'W(ϕε)' appears where 'W(φε)' is meant; this is a typo that should be fixed.
[Abstract] The abstract contains a typo: 'Morever' should be 'Moreover'.
[Proof of Theorem 3.2] In the treatment of the |hε|^2 term, the text writes 'letting η → 1_[t1,t2]' after taking a limsup in ε; this interchanges a limit in η with a limsup in ε and should be justified explicitly, for instance by a diagonal argument or by monotonicity in η.
[Definition 2.2(6)] The statement of the Brakke inequality in Definition 2.2(6) could be clarified by stating explicitly that the constant C is independent of r, t1, t2, and ϕ, as is implied by the proof.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular reduction: the new Brakke inequality is derived from independent results in [25]; the noted ξ_ε → 0 assertion is an unproved gap, not a circular step.

full rationale

The proof of Theorem 3.2 reduces the target inequality (11) to convergence and density facts already proved in [25], all of which are independent of (11): the Radon convergence με_t → μ_t (Theorem 3.1(a)), the convergence of the nonlocal force term (Theorem 3.1(d)), the L2-flow velocity convergence (Theorem 3.1(e)), the lower-semicontinuity of |h|^2 (Theorem 3.1(e)/(27)), the L2-bound for λ_ε (Theorem 3.1(c)), and the density estimate με_t(B_r(x_0)) ≤ c r^{d-1} from [25, Cor. 1]. None of these inputs asserts the Brakke inequality; the genuinely new |λ_ε|^2 estimate is bounded by an independent density estimate. Thus the central claim does not reduce by construction to its inputs. The one notable weakness is not circular: in Section 3.2, after defining ξ_ε, the proof states 'By Theorem 3.1 (a), and the fact that ξε → 0 as Radon measures, we have eμε → μ as Radon measures' without proving or citing that convergence. This is a load-bearing gap needed to identify the weak limit of the approximate velocity measure and produce the |v|^2 term; it is a correctness risk, not a circularity. There is also a sign typo in the displayed identity eμ_ε = μ_ε − ξ_ε, since the definitions give eμ_ε = μ_ε + ξ_ε; the sign is immaterial if ξ_ε → 0. Overall, the derivation is not equivalent to its inputs and no circular step is present.

Assumptions & free parameters 1 free parameters · 5 assumptions · 0 invented entities

The central claim inherits substantial machinery from [25]; the new proof adds one unproved technical fact (discrepancy convergence) and relies on standard GMT tools. No new physical or mathematical entities are postulated; the new volume-preserving Brakke-flow is a definition, not an entity.

free parameters (1)

α = α ∈ (0,1), arbitrary fixed
Exponent in the nonlocal penalization λε (equation (19)). The paper fixes α ∈ (0,1) and the proof uses only L2 bounds that depend on α via [25]; no α-independence of the limit is shown.

assumptions (5)

domain assumption Theorem 3.1 (a)-(e) from [25]
The proof of Theorem 3.2 imports convergence of measures, λε bounds, velocity and mean-curvature convergence, and L2-flow properties from the authors' previous paper [25]. These theorems are not re-proved.
domain assumption Upper density estimate µε_t(B_r(x0)) ≤ c r^{d-1} ([25, Cor. 1])
Used in Theorem 3.2 to control the |λε|^2 term by r^{d-1} C; not proved here.
ad hoc to paper Discrepancy measure convergence ξε → 0 as Radon measures
Asserted in the proof of Theorem 3.2 without proof or citation; needed to identify the weak limit of ε|∇φε|^2 dxdt.
standard math Standard geometric measure theory tools ([11, Thm 4.4.2], [28, Lemma 3.1], [23, Thm 3.5])
Compactness for approximate velocities, the bound |∇ζ|^2/ζ ≤ C||ζ||_{C^2}, and density characterizations are used as black boxes.
domain assumption Initial data assumptions (16)-(17)
The initial set U_0 must satisfy a uniform density bound and be approximable by smooth sets with converging perimeters; the phase-field construction depends on this.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Brakke inequality and the existence of Brakke-flow for volume preserving mean curvature flow." pith.science (2026). https://pith.science/paper/DVY5KCFC

@misc{pith2026250523222,
  author       = {Pith},
  title        = {Pith review of: Brakke inequality and the existence of Brakke-flow for volume preserving mean curvature flow},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DVY5KCFC}},
  note         = {Machine review of arXiv:2505.23222}
}

abstract

In this paper, we propose a new notion of Brakke inequality for volume preserving mean curvature flow. We show the existence of integral varifolds solving the flow globally-in-time in the corresponding Brakke sense using the phase field method. Moreover, such varifolds are solutions to volume preserving mean curvature flow in the $L^2$-flow sense as well. We thus extend a previous result by one of the authors [25].

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

28 extracted references · 27 canonical work pages

[25]

K. Takasao. The existence of a weak solution to volume preserving mean curvature flow in higher dimensions. Arch. Ration. Mech. Anal. 247 (2023), no.3, Paper No. 52, 53 pp

work page 2023
[1]

Almgren, J

F. Almgren, J. Taylor, L. Wang. Curvature-driven flows: a variational approach. SIAM J. Control Optim. 31 (1993), no. 2, 387–438

work page 1993
[2]

Ambrosio, N

L. Ambrosio, N. Fusco, D. Pallara. Functions of bounded variation and free discontinuity problems. The Claren- don Press, Oxford University Press, New York.(2000)

work page 2000
[3]

Bonforte, F

M. Bonforte, F. Maggi, D. Restrepo. Asymptotic behavior of a diffused interface volume-preserving mean cur- vature flow. preprint (2024), https://arxiv.org/abs/2407.18868

arXiv 2024
[4]

K. A. Brakke. The motion of a surface by its mean curvature . Math. Notes, 20 Princeton University Press, Princeton, NJ, 1978. i+252 pp

work page 1978
[5]

K. Ecker. Regularity theory for mean curvature flow. Progr. Nonlinear Differential Equations Appl., 57 Birkh¨auser Boston, Inc., Boston, MA, 2004

work page 2004
[6]

Escher, G

J. Escher, G. Simonett. The volume preserving mean curvature flow near spheres. Proc. Amer. Math. Soc. 126 (1998), no. 9, 2789–2796

work page 1998
[7]

The local structure of the energy landscape in multiphase mean curvature flow: Weak-strong uniqueness and stability of evolutions

J. Fischer, S. Hensel, T. Laux and T. M. Simon. The local structure of the energy landscape in multiphase mean curvature flow: Weak-strong uniqueness and stability of evolutions. preprint (2020), https://arxiv.org/abs/2003.05478

work page Pith review arXiv 2020

Show all 28 references

[8]

M. Gage. On an area-preserving evolution equation for plane curves. Nonlinear problems in geometry (Mobile, Ala., 1985), 51–62. Contemp. Math., 51 American Mathematical Society, Providence, RI, 1986

1985
[9]

Golovaty

D. Golovaty. The volume-preserving motion by mean curvature as an asymptotic limit of reaction-diffusion equations. Quart. Appl. Math. 55 (1997), no. 2, 243–298

1997
[10]

G. Huisken. The volume preserving mean curvature flow. J. Reine Angew. Math. 382 (1987), 35–48

1987
[11]

J. E. Hutchinson. Second fundamental form for varifolds and the existence of surfaces minimising curvature. Indiana Univ. Math. J. 35 (1986), 45–71

1986
[12]

Julin, J

V . Julin, J. Niinikoski. Quantitative Alexandrov theorem and asymptotic behavior of the volume preserving mean curvature flow. Anal. PDE 16 (2023), 679–710. THE EXISTENCE OF BRAKKE-FLOW FOR VOLUME PRESERVING MEAN CURV ATURE FLOW 13

2023
[13]

Kasai, Y

K. Kasai, Y . Tonegawa. A general regularity theory for weak mean curvature flow.Calc. Var. Partial Differential Equations 50 (2014), No. 1-2, 1–68

2014
[14]

I. Kim, D. Kwon. V olume preserving mean curvature flow for star-shaped sets. Calc. Var. Partial Differential Equations 59 (2020), no. 2, 40 pp

2020
[15]

Lambert, E

B. Lambert, E. M ¨ader-Baumdicker. Nonlocal estimates for the volume preserving mean curvature flow and ap- plications. Calc. Var. Partial Differential Equations62 (2023), Paper No. 202, 32

2023
[16]

T. Laux. Weak-strong uniqueness for volume-preserving mean curvature flow.Rev. Mat. Iberoam.40 (2024), no. 1, 93–110

2024
[17]

T. Laux, F. Otto. Convergence of the thresholding scheme for multi-phase mean-curvature flow.Calc. Var. Partial Differential Equations 55 (2016), no. 5, Art. 129, 74 pp

2016
[18]

T. Laux, D. Swartz. Convergence of thresholding schemes incorporating bulk effects. Interfaces Free Bound. 19 (2017), no.2, 273—304

2017
[19]

T. Laux, T. Simon. Convergence of the Allen-Cahn equation to multiphase mean curvature flow. Comm. Pure Appl. Math. 71 (2018), no. 8, 1597—1647

2018
[20]

Luckhaus, T

S. Luckhaus, T. Sturzenhecker. Implicit time discretization for the mean curvature flow equation. Calc. Var. Partial Differential Equations3 (1995), no. 2, 253–271

1995
[21]

Mugnai, M

L. Mugnai, M. R ¨oger. The Allen-Cahn action functional in higher dimensions.Interfaces Free Bound.10 (2008), no.1, 45–78

2008
[22]

Mugnai, C

L. Mugnai, C. Seis, E. Spadaro. Global solutions to the volume-preserving mean-curvature flow.Calc. Var. Partial Differential Equations 55 (2016), no. 1, Art. 18, 23 pp

2016
[23]

L. Simon. Lectures on geometric measure theory. Proceedings of the Centre for Mathematical Analysis, Aus- tralian National University. 3 (1983)

1983
[24]

Stuvard, Y

S. Stuvard, Y . Tonegawa. On the existence of canonical multi-phase Brakke flows.Adv. Calc. Var.17 (2024), no. 1, 33–78

2024
[26]

Takasao, Y

K. Takasao, Y . Tonegawa. Existence and regularity of mean curvature flow with transport term in higher dimen- sions. Math. Ann. 364 (2016), no.3-4, 857–935

2016
[27]

Tonegawa

Y . Tonegawa. A second derivative H¨older estimate for weak mean curvature flow.Adv. Calc. Var.7 (2014) No. 1, 91–138

2014
[28]

Tonegawa.Brakke’s mean curvature flow

Y . Tonegawa.Brakke’s mean curvature flow. An introduction. Springer Briefs Math. Springer, Singapore, 2019. xii+100 pp. (Andrea Chiesa) U NIVERSITY OF VIENNA , F ACULTY OF MATHEMATICS AND VIENNA SCHOOL OF MATHE- MATICS , O SKAR -M ORGENSTERN -PLATZ 1, A-1090 V IENNA , AUSTRIA...

2019

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [25]

K. Takasao. The existence of a weak solution to volume preserving mean curvature flow in higher dimensions. Arch. Ration. Mech. Anal. 247 (2023), no.3, Paper No. 52, 53 pp

work page 2023

[2] [1]

Almgren, J

F. Almgren, J. Taylor, L. Wang. Curvature-driven flows: a variational approach. SIAM J. Control Optim. 31 (1993), no. 2, 387–438

work page 1993

[3] [2]

Ambrosio, N

L. Ambrosio, N. Fusco, D. Pallara. Functions of bounded variation and free discontinuity problems. The Claren- don Press, Oxford University Press, New York.(2000)

work page 2000

[4] [3]

Bonforte, F

M. Bonforte, F. Maggi, D. Restrepo. Asymptotic behavior of a diffused interface volume-preserving mean cur- vature flow. preprint (2024), https://arxiv.org/abs/2407.18868

arXiv 2024

[5] [4]

K. A. Brakke. The motion of a surface by its mean curvature . Math. Notes, 20 Princeton University Press, Princeton, NJ, 1978. i+252 pp

work page 1978

[6] [5]

K. Ecker. Regularity theory for mean curvature flow. Progr. Nonlinear Differential Equations Appl., 57 Birkh¨auser Boston, Inc., Boston, MA, 2004

work page 2004

[7] [6]

Escher, G

J. Escher, G. Simonett. The volume preserving mean curvature flow near spheres. Proc. Amer. Math. Soc. 126 (1998), no. 9, 2789–2796

work page 1998

[8] [7]

The local structure of the energy landscape in multiphase mean curvature flow: Weak-strong uniqueness and stability of evolutions

J. Fischer, S. Hensel, T. Laux and T. M. Simon. The local structure of the energy landscape in multiphase mean curvature flow: Weak-strong uniqueness and stability of evolutions. preprint (2020), https://arxiv.org/abs/2003.05478

work page Pith review arXiv 2020