Talk:Dehn invariant/GA1

GA Review
The edit link for this section can be used to add comments to the review.''

Reviewer: Kusma (talk · contribs) 09:31, 8 March 2023 (UTC)

Will review this one. Expect comments over the next few days. —Kusma (talk) 09:31, 8 March 2023 (UTC)

Section by section review

 * Lead: Looks like a reasonable summary of the article.
 * Might be helpful to notice that Dehn=0 is not sufficient for being space-filling
 * Better.
 * Related results: Any reason not to mention the higher dimensional results?
 * One reason for not citing those specific publications is that I don't read German and can't get enough information from the MR reviews to understand exactly how those connect to Dehn invariants specifically (rather than to the more general theory of additive functionals developed e.g. in Klain and Rota's Introduction to Geometric Probability). —David Eppstein (talk) 07:44, 9 March 2023 (UTC)
 * I do read German (better than English...) I'll have a look and report back. But perhaps a single sentence like in Encyclopedia of Mathematics would do the trick.
 * As I understand it, Hadwiger proves that equality of Dehn invariants (he writes them using certain functionals; looks like the dual space approach to the Hamel basis approach to me) is necessary for equidecomposability of higher polytopes in any dimension. I couldn't access the Jessen paper that proves the sufficiency, but Dupont-Sah MR article put this into their modern context (see remark on p. 25 about the 4d case). I'm not sure whether this is open or wrong for dimension 5 and higher.
 * Oh, ok, that reference helped put this into context. This relates to a cryptic remark at the end of the "Realizability" section which I have now expanded and summarized in the lead based on Dupont & Sah 1990 (probably it's also somewhere in their 2000 book). Basically it involves the definition of Dehn invariant by an exact sequence at the start of the "Realizability" section. This lets you define a group of polytopes modulo dissections, which turns out to be the same group in both 3d and 4d. It follows that in 4d, too, Volume+Dehn is a complete system of invariants, telling you everything you needed to know about dissectability. In higher dimensions you still get a Dehn invariant, which still has to be equal for a dissection to exist, but it's open whether that and volume are enough or whether there might be some other invariant that also needs to be equal. —David Eppstein (talk) 07:22, 12 March 2023 (UTC)
 * Simplified calculation: nontrivial methods in number theory "of"? "from"?
 * Ok, better, done. —David Eppstein (talk) 07:23, 12 March 2023 (UTC)
 * Related polyhedra: Like the cube, the Dehn invariant of any parallelepiped is also zero. The cube is not zero.
 * Reworded. —David Eppstein (talk) 07:25, 12 March 2023 (UTC)
 * For the parallelepiped, I was kind of expecting to see a "dissect-into-rectangular-cuboid" approach, but this is of course fine.
 * That could probably be sourced (maybe in one of Greg Frederickson's books?) but I think it would add more complication than enlightenment. —David Eppstein (talk) 07:25, 12 March 2023 (UTC)
 * Applications: For the reader interested in Hilbert's third problem, the fact that the Dehn invariant is indeed invariant under dissection is kind of the central point of the article. Could you discuss this in the body instead of as a footnote? It would also benefit from a picture, but I know that is a big ask.
 * Un-footnoted the explanation of why the new edges don't affect the invariant, and added an illustration of a cube dissected into orthoschemes. —David Eppstein (talk) 01:45, 13 March 2023 (UTC)
 * I don't think the namedrop of Hilbert 18 is necessary; interestingly, Gyrobifastigium doesn't mention it.
 * Ok, removed. —David Eppstein (talk) 01:45, 13 March 2023 (UTC)
 * As a tensor product: I think the definition of which polyhedra the invariant is applicable to could be clarified (which of the definitions given in polyhedron are such that the Dehn invariant is always defined? Are there any where it isn't?). I stumbled on "manifold" as my default assumption for that is "smooth manifold", not "topological manifold".
 * Changed to piecewise linear manifold. But the issue that this passage skirts around is that the definition of "polyhedron" in the literature (or rather multiplicities of definitions and failed attempts at definitions) is a total mess. See for more than you probably wanted to know about this, especially the Grünbaum quote about original sin. Using embedded PL manifolds has the advantage of being specific and valid, although overly restrictive. The "embedded" part is important so that it has an inside and an outside and you know which side the dihedral is on; "manifold" is less important but describing exactly how it might be relaxed could easily veer into original research. —David Eppstein (talk) 01:53, 13 March 2023 (UTC)
 * Using a Hamel basis: do you need "carefully"?
 * Rephrased to avoid using that word. The intent was merely to point out that not all Hamel bases work. —David Eppstein (talk) 01:54, 13 March 2023 (UTC)
 * axiom of choice: make more explicit that the existence of a Hamel basis is what needs the axiom of choice? But anyone who knows what a Hamel basis is probably knows this... In any case, the "this alternative formulation shows it is a real vector space" thing should come before the axiom of choice bit.
 * Reordered, and rephrased to state more clearly (I hope) that it is the general construction of Hamel bases that involves AC. —David Eppstein (talk) 02:11, 13 March 2023 (UTC)
 * Hyperbolic polyhedra: is there a comprehensible reason why this doesn't depend on the choice of horospheres?
 * Yes, actually, but I don't know if it can be sourced. It's because the cusp of the polyhedron that is cut off by a horosphere has the same dihedral angles as its limiting 2d Euclidean polygon, as if it were an infinitely tall Euclidean prism, so per unit length its dihedral angles sum to zero in the Dehn invariant. —David Eppstein (talk) 08:15, 12 March 2023 (UTC)
 * Realizability: linear subspace with respect to the reals, I guess, which should be made explicit. I like the geometric explanation of the vector space operations.
 * Ok, added "real" a couple of times here. The parts elsewhere in the article that mention tensor rank involve linearity over $$\mathbb{Q}$$ rather than $$\mathbb{R}$$ but maybe that's more confusing to explain than to leave in the background. —David Eppstein (talk) 08:26, 12 March 2023 (UTC)
 * triangular prisms: do you use a fixed base? Otherwise you'd have many triangular prisms with the same volume. Do you really assign a volume to each group element or are you just talking about the prism?
 * Your exact sequence is only a "short exact sequence" because the second group is zero. Suggest to drop "short".
 * You caught that. I had been wondering whether anyone would. Ok, done. —David Eppstein (talk) 08:06, 12 March 2023 (UTC)
 * Why do you drop the 2pi in the tensor product in this section?
 * Probably because the main source for that section did. It makes no difference mathematically. But thinking about this again, I think it's more confusing to change notation and explain that it makes no difference than to just keep the same notation throughout, so I have put back the 2πs. —David Eppstein (talk) 08:01, 12 March 2023 (UTC)
 * Totally random aside: an article about the Dehn invariant appeared in a birthday volume for David Epstein :)
 * I am at coauthorship distance 2 to DBAE through Mike Paterson, who at one point threatened to make us write a joint paper, but that hasn't happened. —David Eppstein (talk) 07:29, 12 March 2023 (UTC)
 * Related results: The number of citations for some of the sentences seems a little over the top. And it would be nice to hear about the history of the rectangle decomposition problem (according to Benko, the rectangle-from-squares theorem was proved by Dehn himself).
 * The Dehn 1903 reference for this was in the piles of citations. I trimmed them a little and added a more explicit callout to Dehn in the article text. —David Eppstein (talk) 07:57, 12 March 2023 (UTC)
 * Total mean curvature: this is quite a different object (naturally generalized from the smooth case to a case where curavture is concentrated on a lower dimensional subset). But it is probably just ontopic enough.
 * Copyedited to clarify that this is a generalization to polyhedral surfaces of the usual definition for smooth surfaces. —David Eppstein (talk) 08:05, 12 March 2023 (UTC)
 * A source comment: I would suggest to cite the English version of Gaifullin-Ignashchenko.
 * Reference 26 looks funny ("&Dupont")
 * Double vertical bar fixed. —David Eppstein (talk) 07:50, 12 March 2023 (UTC)

General comments and GA criteria
Overall quite a nice article about a famous concept and some deep connections. I think it has a good mix of understandable to the general public and requiring deeper expertise. I'll do image and source checking later, but other than the somewhat questionable 24 (Rich Schwartz, lecture notes?) I expect to have no major concerns. None of my comments above points to major issues with other criteria. —Kusma (talk) 18:05, 9 March 2023 (UTC)
 * Richard Schwartz (mathematician) passes the "established subject-matter expert" test of WP:SPS. —David Eppstein (talk) 02:15, 13 March 2023 (UTC)
 * Did a few spotchecks, all fine. The Benko PDF link in 23 is broken. I find the comments about Bricard's failed solution (but correct theorem) in that paper interesting (these could go into the history as well).
 * Found an archive link and added Bricard to history. —David Eppstein (talk) 05:57, 13 March 2023 (UTC)
 * Images are fine and relevant. Not many, though -- you could add one of Max Dehn if you like.

Thanks for the thorough review! I'll try to get to this over the weekend. —David Eppstein (talk) 22:56, 9 March 2023 (UTC)
 * Excellent changes, very nice article (even more so than before). I think the citation for Schwartz could be slightly more detailed (give the website etc.) but I am going to pass this now. —Kusma (talk) 09:45, 13 March 2023 (UTC)