Wikipedia talk:WikiProject Mathematics/Archive/2021/Jan

Reliability of tertiary sources such as of Encyclopedia of Mathematics

At Function (mathematics) § Injective, surjective and bijective functions, while getting some citations into this previously {{unreferenced section}}, I replaced a statement that a function is injective if and only if the preimage of each element of its codomain contains at most one element with a statement that a function is injective if and only if the preimage of each element of its range contains exactly one element (Special:Diff/999090776/999284001), because the latter statement (and not the former) appears in what I thought seemed like a convenient source, the "Function" article in the Encyclopedia of Mathematics, which WP:WikiProject Mathematics/Reference resources says [s]hould be regarded as a highly reliable source.

D.Lazard, when you reversed this change, saying EOM os a WP:tertiary source and therefore is not considered as a WP:reliable source in such a case (Special:Diff/999286540), what did you mean? I looked through the policy and guideline you linked, but, as far as I see, neither mentions any case in which being a tertiary source makes a source unreliable (unless the source is Wikipedia).

(This seems like more of a general matter than one specific to Function (mathematics), so I hope I may ask the question here rather than at Talk:Function (mathematics).)

—2d37 (talk) 13:40, 9 January 2021 (UTC)

WP:PSTS: Wikipedia articles should be based on reliable, published secondary sources and, to a lesser extent, on tertiary sources and primary sources.

WP:TERTIARY: Policy: Reliable tertiary sources can be helpful in providing broad summaries of topics that involve many primary and secondary sources, and may be helpful in evaluating due weight, especially when primary or secondary sources contradict each other.

So, the use of tertiary sources for specific technical definitions (as it is the case here) is not recommended, and not forbidden, although secondary sources and well known textbooks are preferred. Here there are many available textbooks, and the previous formulation is clearer than the formulation that is alleged to be closer than that of EOM. As both formulations are mathematically equivalent, and one may be confusing, we must keep the clearer one. If a citation would be needed, we would have to find a textbook that uses a closer formulation. However, as a sourced definition has been given in the preceding sentence, and the equivalence of the two formulations must be very easy for every body who understand them, WP:CALC applies, and I agree with your last edit removing EOM reference. D.Lazard (talk) 16:26, 9 January 2021 (UTC)

Facet theory

Can people here contribute to dealing with the issues raised by the maintenance tags atop the article titled Facet theory? Michael Hardy (talk) 17:54, 12 January 2021 (UTC)

This does not look like mathematics to me. JRSpriggs (talk) 21:29, 12 January 2021 (UTC)

It looks unfixable to me. XOR'easter (talk) 23:04, 12 January 2021 (UTC)

This appears to also have the same issues (with the same creator and topic for the most part) as Guttman scale. Both need a lot of work. — MarkH₂₁^talk 04:47, 13 January 2021 (UTC)

Delete "intersection" page?

Hello! I believe we should delete/merge the current intersection page. It should turn into a disambiguation page between intersection (set theory) and intersection (Euclidean geometry).

As it stands, I don't think a "general" intersection page is useful or necessary.

To my understanding, something is either a geometric intersection (so can go to intersection (Euclidean geometry)) or a set theory intersection (and can go to intersection (set theory)). Or maybe someone is looking for intersection theory. I think Intersection should turn into a disambiguation page. Otherwise, what should the "broad" page for intersection be? IMO all it would be is a description of Euclidean or set theory intersections. I don't know what reliable independent sources we could cite that give a broad explanation of both and more.

What do you think? Should we keep intersection? If so, what should we put there? Turn it into a disambiguation page?

I also started a discussion on the intersection (set theory) page.

I think it should be noted that:

• intersection is a start-quality page
• intersection (set theory) is a C-class page
• intersection (Euclidean geometry) is a C-class page

And if it is agreed not to change anything, at the very least I think intersection (set theory) needs a hat possibly directing to intersection (Euclidean geometry) or Intersection (disambiguation) and visa versa IllQuill (talk) 06:28, 16 January 2021 (UTC)

I agree that things must change, but I disagree with the proposed changes. IMO, the three article must be merged into a single article called intersection or intersection (mathematics). In fact, in modern mathematics, intersections are almost alway set-theoretical intersections. The only case where intersections are not set-theoretic, are the versions of incidence geometry where a line is not the set of its points. This is very marginal and could be treated in a section "In incidence geometry". This section could be rather short and shoud mainly explain that the two concepts of intersection are essentially the same even if they differ formally. D.Lazard (talk) 12:05, 16 January 2021 (UTC)

While that may be true conceptually, I think there's a very big difference in the set theoretic and geometric article perspectives: the latter is presenting methods for determining the set of points of intersection, hence equations of lines and curves, vectors, etc. The readership of the two articles themselves will also likely have different backgrounds and concerns. Incidentally, Intersection (set theory) as-is nicely complements the article Union (set theory). NeilOnWiki (talk) 14:07, 16 January 2021 (UTC)

Having one article called Intersection (mathematics) makes the most sense to me. It's the simple approach, and I like simple. Too often, we have lots of little articles that each carry a piece of a topic, and the pieces might overlap, telling the same story with contradictory notations — the inevitable consequence of editors independently adding what they feel like when and where they feel like it without agreeing on a curriculum first. So, every now and then we have to come through and reorganize. Such is life on a wiki! There's nothing wrong with having multiple perspectives in one article, particularly when displaying those perspectives together allows them to illuminate each other. Intersection (set theory) can always redirect to the proper section of the merged article. XOR'easter (talk) 17:03, 16 January 2021 (UTC)

I very much agree the general observation that "Too often, we have lots of little articles that each carry a piece of a topic, and the pieces might overlap, telling the same story with contradictory notations ... So, every now and then we have to come through and reorganize." But, in this particular case, Intersection (set theory) and Intersection (Euclidean geometry) are about very different aspects with no obvious overlap in content beyond the word intersection. The content in the geometry one is mainly vector algebra and solving simultaneous equations. These are applied mathematical techniques where a knowledge of set theory is irrelevant: eg. there's no mention of the word set even when any equation has multiple solutions, nor is there a need to. NeilOnWiki (talk) 19:32, 16 January 2021 (UTC)

I agree with D.Lazard and XOR'easter that we should rename the page Intersection (set theory) as Intersection (mathematics) and include the page Intersection (Euclidean geometry) in it. The overlap is not just in the use of the word "intersection"; there is just one notion of intersection here and it is applied in different areas of mathematics. What is the meaning of "intersection" in Euclidean geometry, if not the notion of intersection that is used throughout mathematics? If you disagree with combining the pages, do you think we should also have separate pages for Intersection (group theory) and Intersection (linear algebra) and ..., with one disambiguation page to rule them all, given that the methods for computing intersections in the different areas of mathematics may be different? To me, having all those separate pages would seem ridiculous. Ebony Jackson (talk) 21:15, 16 January 2021 (UTC)

It is possible to treat geometric lines as first-class objects that are not merely sets of points, and to define intersections of lines as an operation that is not merely intersection of sets of points. But it is not necessary to do so, and I don't see the point in doing so in our main intersection article. I agree that geometric intersections should be covered there, as a special case of set-theoretic intersections. —David Eppstein (talk) 21:38, 16 January 2021 (UTC)

Just chiming in to say that intersections in Algebraic Geometry and very much not set-theoretic, and a great deal of effort in intersection theory is made to understand them in terms of the more intuitive Euclidean intersection picture. This may be relevant when writing/updating the intersection article. Tazerenix (talk) 22:17, 16 January 2021 (UTC)

@David Eppstein: Yes, that is possible. If that is to be mentioned, perhaps it could be done in a "Variants" section of the page (or omitted from this page entirely, as you suggest).

@Tazerenix: Well, most of the time in algebraic geometry when one speaks of the intersection of two varieties, one does mean the intersection of the sets. Even if one is thinking scheme-theoretically, the underlying set is the set-theoretic intersection, and also the functor of points of the scheme-theoretic intersection is the functor whose values are the (set-theoretic) intersections of the values of the functor of points of the two subschemes being intersected. I suppose that you are thinking of the more sophisticated but less common kind of intersection, when one is working in the Chow group or the like, when the intersections are not proper. Anyway, sorry for going off-topic! This shouldn't affect the outcome of the merging discussion. Ebony Jackson (talk) 01:50, 17 January 2021 (UTC)

There are other geometrical aspects of intersections that don't fall into Euclidean Geometry. When studying Immersion (mathematics) the set of self-intersections of the map is of interest, but there is no requirement for either the source or target to be Euclidean. This would point to a wider intersection (mathematics) article.--Salix alba (talk): 06:55, 17 January 2021 (UTC)

Yes, the title does raise a few questions. Looking at its sources, I vaguely wonder if it should have been named Intersection (Computational geometry). But more fundamentally I wonder if we're concentrating too much on rarefied mathematical questions here and need to think more broadly about the encyclopaedia and its users. There's already a more general Intersection (disambiguation) page so in answer to the original question I'm unsure if we'd need a specifically mathematical one. As may be obvious, I'm very much against merging Intersection (set theory) into Intersection (mathematics) — which currently feels a bit of a lonely position. We need (I think) to consider what role is played by each article in disseminating mathematical information, what kind of readership will benefit from it, their background and prior knowledge, their aims and motivation. What's the role of Intersection (set theory)? It looks to me like it's there to explain a concept in (mostly naive) set theory at a fairly introductory level that's accessible to a wide readership. It has issues (not least in nullary intersection), but seems a fairly coherent, reasonably well-defined page if seen in that light, and one which nicely complements the Union (set theory) article. My fear is that we're proposing to turn it into something less accessible and less valuable if we try to merge it into a catch-all page in the way we're arguing. NeilOnWiki (talk) 09:38, 18 January 2021 (UTC)

+1 re: accessibility. --JBL (talk) 14:31, 18 January 2021 (UTC)

Yes, accessibility is important. I think accessibility and merging are not incompatible. An Intersection (mathematics) article could start with an elementary discussion of intersection of sets, and then later sections could mention how intersections in elementary geometry are calculated. If more advanced topics such as "scheme-theoretic intersection" are mentioned at all (and it's not clear that they should be), then they should appear only as remarks towards the end of the article. Ebony Jackson (talk) 17:23, 18 January 2021 (UTC)

Wikipedia:WikiProject Physics/Contest

Over at WT:PHYS, we decided to try a bit of an article improvement drive. As there might be an overlap of interest, I figured I'd also post a notice here. XOR'easter (talk) 16:55, 18 January 2021 (UTC)

I think that complex analysis is applicable to this contest. This page points out that there is a lack of explanation from a physics perspective, and this competition may improve it. However, since this competition is focused on physics, the improvement evaluated will be from the perspective of physics, which may complicate the evaluation. Page views are 16,786. --SilverMatsu (talk) 05:12, 19 January 2021 (UTC)

I'm going to change the page name of Hartogs's Theorem

I'm thinking of moving to Hartogs's theorem on separate holomorphicity. Also, looking at Talk:Hartogs's theorem, it seems that the maths rating template is not used. --SilverMatsu (talk) 11:52, 20 January 2021 (UTC)

About the rating template, anyone (e.g., you) is free to add it at any time. If you do move the article (I don't personally have an opinion), the page Hartogs's theorem should presumably be converted into a disambiguation page, pointing at the various targets currently in the note at the top of Hartogs's theorem. --JBL (talk) 13:43, 20 January 2021 (UTC)

I have fixed the hatnote of the article for using a standard format and removing a duplicate link. So, there are only two other links. As the theorem on infinite ordinals is clearly not a primary topic, there are only two candidates for being a primary topic. As these two Hartogs's theorems belong to the same theory (of holomorphic functions of several variables), the primary topic is certainly clear for the specialists. So, per WP:ONEOTHER a dab page seems unneeded.

I have no opinion on the move, but if it is done, the hatnote {{about}} must be replaced by a template {{redirect}}. D.Lazard (talk) 14:48, 20 January 2021 (UTC)

Huh? How is it "clear" that the ordinal theorem is "not a primary topic"? I don't think that's clear at all. --Trovatore (talk) 17:38, 20 January 2021 (UTC)

It is clear because of the size of the interested audience: complex analysis and these Hartogs's theorems are used in many areas of mathematics and physics, while, as far as I know, the properties of transfinite ordinals that are considered here are rarely used outside advanced set theory. D.Lazard (talk) 18:14, 20 January 2021 (UTC)

It's a really pretty, simple, and fundamental construction, from the early days of set theory. Everyone should really know it. The complex-analysis result is more a technical thing from deeper in the bowels of the subject.

That said, it's probably true that hardly anyone refers to the existence of the Hartogs number as "Hartogs' theorem". --Trovatore (talk) 19:17, 20 January 2021 (UTC)

Thank you for improving Hartogs's theorem. I checked the page that links to Hartogs's theorem, but it seems a bit confusing. Looking at the Friedrich Hartogs page, it seems that it is linked to the Hartogs number by writing Hartogs's theorem. Even for several complex variables, the link that explains that the continuity of the condition that the function becomes holomorphic can be derived from the separate holomorphicity was previously the Hartogs extension theorem, and the name was confused. This was the reason for trying to clarify this name.--SilverMatsu (talk) 06:04, 21 January 2021 (UTC)

@Kpgjhpjm: Thank you for the message. Please check out the discussion here.--SilverMatsu (talk) 07:14, 21 January 2021 (UTC)

@SilverMatsu:, Thanks for the ping , You may convert the page into a dab page as mentioned above. Kpgjhpjm 07:34, 21 January 2021 (UTC)

About the name of X-Pseudoconvex

There are multiple definitions for the domain called pseudoconvex, and each name seems to be called differently depending on the person, but at Wikipedia, I wanted to discuss how to call it. Perhaps the early treatises were written in French (although one option is to choose a name that is commonly used in English), and on this page I found a user whose native language is French. I thought I would consult on this page. thanks!--SilverMatsu (talk) 23:20, 10 January 2021 (UTC)

Do you mean to say domain of holomorphy, (open) pseudoconvex subset and Stein manifolds are all the same thing so Wikipedia should pick one term to refer to them all? It is usually a bad idea to try to mess with terminology in Wikipedia; since, for one thing, there are many anonymous editors who edit math articles and we cannot expect them to be aware of some terminological insider convention. It is desirable and is quite achievable to use some consistency within a single article, though. -- Taku (talk) 00:08, 11 January 2021 (UTC)

Thank you for your reply. In a narrower story, I would like to ask if the usage of the names p-pseudoconvex, Levi pseudoconvex, Strongly pseudoconvex, and Cartan pseudoconvex is popular(commonly). I'm not trying to define these in one way. Each of these definitions has its own advantages. These names are ambiguous to myself. thanks!--SilverMatsu (talk) 00:35, 11 January 2021 (UTC)

Ah, I see. Usually in Wikipedia, the best way to approach the problem like this is to pick and follow a standard and *recent* textbook on the subject. For example, in this case, we can follow Demailly, Complex Analytic and Differential Geometry. In Theorem 7.2. it is shown that various notions like strongly psuedoconvex or weakly psuedoconvex are equivalent and that equivalence is used to define the common notion "pseudoconvex". In Wikipedia, we can do the same; i.e., an open subset is pseudoconvex if it satisfies the following equivalent conditions are met. ..... For Levi pseudoconvex, you need (as I understand) a C_2 boundary so we can say if the boundary is C_2, pseudoconvexity can be characterized in the Levi form. (I'm happy to leave the matters to specialists (I am certainly not) but I am also happy to edit the article myself if needed). -- Taku (talk) 04:03, 11 January 2021 (UTC)

Thank you very much! It was very helpful.--SilverMatsu (talk) 06:54, 11 January 2021 (UTC)

By the way, on the wiki, typing \mathscr seems to give an error.--SilverMatsu (talk) 08:10, 11 January 2021 (UTC)

You can see what TeX math fonts are available at Wikipedia:LaTeX symbols#Fonts. There is \mathcal, but no \mathscr. —2d37 (talk) 11:04, 11 January 2021 (UTC)

Thank you for teaching me. I'll try.--SilverMatsu (talk) 11:09, 11 January 2021 (UTC)

Although it is a 1954 paper, I found a paper that can be used as a reference for the name of the pseudoconvex domain. See https://doi.org/10.2969/jmsj/00620177. With reference to this material, Levi Pseudoconvex can be called as it is, and Levi convex (Equivalent conditions 4) may be called Levi strongly-Pseudoconvex. It seems that it can be called strongly pseudoconvex or locally analytical convex, but if we use strong pseudoconvex for the pseudoconvex region defined using the Strictly plurisubharmonic function, or considering the relationship with the pseudoconvex, I'm thinking of calling it Levi strongly-Pseudoconvex. thanks! --SilverMatsu (talk) 07:03, 24 January 2021 (UTC)

I'm sorry I seemed to have misunderstood. Levi convex seems to have been convex to the analytical surface.--SilverMatsu (talk) 08:38, 25 January 2021 (UTC)

Help:MATH

I came to this page after a discussion with a new user who changed systematically html to latex, even for isolated variables. He referred to this help page, which was blatantly biased toward latex. Moreover, the page contained very technical details that was of no help for the normal users of this help page. So I have rewritten the beginning of this page.

The new version gives advices that are based on an implicit consensus that I have deduced from many discussions in this talk page. Please, review my version for improving it, and fixing it, if I have misunderstood something. D.Lazard (talk) 14:40, 25 January 2021 (UTC)

"Thagomizer Matroids"

This is extremely tangentially related to mathematics for which I apologise. "Thagomizer" is an informal term for the tail spikes of stegosaurian dinosaurs, originally coined in comic strip. I have nominated this article for deletion. Several support votes for the article being kept have advanced the naming of a mathematical concept entitled "Thagomizer matroids" (and to a lesser extent "Thagomiser graphs") after the term, which are associated with Kazhdan–Lusztig polynomials as evidence of notability. The term appears to have been coined in a 2017 article in Electronic Journal of Combinatorics. As a non-mathematician, the term seems like one of hundreds, perhaps thousands of minor mathematical terms used in the literature, and not really evidence of notability of the article. Can I have a second opinion on the prominence of this term? Thanks. Hemiauchenia (talk) 16:30, 25 January 2021 (UTC)

Two published math papers have used it. There's no secondary commentary on etymology in either paper. It's a cute joke, but this is not a good indicator of notability. --JBL (talk) 16:49, 25 January 2021 (UTC)

Properties of integers

Please could a number theorist(?) review recent additions by 109.106.227.16? There are a number of plausible claims which may be worth keeping but the text generally seems too detailed for its articles. Certes (talk) 21:28, 14 January 2021 (UTC)

@Certes: They appear wholly unreferenced and unnoteworthy. I couldn't find any references for these claims from a quick search either. — MarkH₂₁^talk 21:35, 14 January 2021 (UTC)

Is 2 the only prime cake number? It's the only one up to 10,000 but that's hardly a rigorous proof. This seems a simple enough conjecture to have a proof or counter-example or prize on offer, and I can find none of those. Certes (talk) 22:14, 14 January 2021 (UTC)

I have no idea! There isn't much literature on the topic, and I haven't found a reference for that fact either but I wouldn't be surprised if it's out there somewhere. — MarkH₂₁^talk 22:31, 14 January 2021 (UTC)

Oddly enough I'd searched for the cake proof in Yaglom & Yaglom, which you just cited for a different claim, but couldn't find anything relevant. Certes (talk) 22:37, 14 January 2021 (UTC)

Here's a quick and easy proof that just came to mind: $(n^{3}+5n+6)=(n+1)(n^{2}-n+6)$ is always divisible by 6, since $n^{2}-n+6$ is always even and is divisible by 3 when $n\not \equiv 2{\pmod {3}}$ while $n+1$ is divisible by 3 when $n\equiv 2{\pmod {3}}$. The two factors are also each larger than 6 when n > 5, so $C_{n}={\frac {n^{3}+5n+6}{6}}$ has two nontrivial integer factors for n > 5.

Technically, the above is OR. I don't think it's worth me putting this anywhere to circumvent that though. — MarkH₂₁^talk 22:52, 14 January 2021 (UTC)

Thanks; at least it's verifiable in the mathematical rather than the WP:V sense. I was halfway there but my maths is rusty. If we leave the {{cn}} then someone may find that in a book somewhere. Certes (talk) 23:01, 14 January 2021 (UTC)

Yaglom (Vol I, solution 45a) states that the differences are the 2D Lazy caterer's sequence (and uses the fact to derive the formula for the nth 3D cake number), so that may be sufficient proof of that assertion. Certes (talk) 00:13, 15 January 2021 (UTC)

Ah, thanks. Adding the citation now! — MarkH₂₁^talk 00:17, 15 January 2021 (UTC)

I've removed the additions from 37 and 89 again (and I'm sure the $16,000 will go to a good cause). Certes (talk) 20:02, 25 January 2021 (UTC)

about Plurisubharmonic function

I think it is correct that Oka defined the Plurisubharmonic function for the research of the pseudoconvex domain, but it seems to call it the pseudoconvex function. Seems to be called as fonction pseudoconvexe. See Oka, Kiyoshi (1942), "Sur les fonctions analytiques de plusieurs variables. VI. Domaines pseudoconvexes", Tohoku Mathematical Journal, First Series, 49: 15–52, ISSN 0040-8735, Zbl 0060.24006.　Need to annotate this? --SilverMatsu (talk) 05:44, 27 January 2021 (UTC)

So, I am not exactly knowledgable with the history of this area (even though Oka is prominent enough in Japan that his profile sometimes appear in a newspaper). But, speaking generally, it is always a good idea to give a historical note on the original terminology (so to help the readers going through old texts). -- Taku (talk) 04:37, 28 January 2021 (UTC)

Thank you for your reply. Apparently, it was also pointed out in Talk:Pseudoconvex function. I left a note on the page. I'm wondering whether to link the pseudoconvex function of convex analysis to the pseudoconvex function. The position to put the annotation may not be appropriate(I would like to put it outside the title ...). thanks!--SilverMatsu (talk) 10:18, 28 January 2021 (UTC)

I am not too sure if that makes sense or that's stretching the connection too much. As the lead of subharmonic notes nicely, there is certainly a sort of analogy between convexity and subharmonicity. On other hand, as far as I can tell, convex analysis and several complex variables tend to be fields with little interaction between the two (I am not saying that's good or bad). -- Taku (talk) 01:11, 31 January 2021 (UTC)

Definition of the limit of a function

In (ε, δ)-definition of limit, limit of a function and several other articles, the limit of a function is defined as

$$\lim _{x\to c}f(x)=L\iff (\forall \varepsilon >0,\,\exists \ \delta >0,\,\forall x\in D,\,0<|x-c|<\delta \ \Rightarrow \ |f(x)-L|<\varepsilon ).$$

Because of the condition $0<|x-c|,$ I call this definition the "punctured definition".

The definition that I have learnt more than 40 years ago, is the "unpunctured definition"

$\lim _{x\to c}f(x)=L\iff (\forall \varepsilon >0,\,\exists \ \delta >0,\,\forall x\in D,\,|x-c|<\delta \ \Rightarrow \ |f(x)-L|<\varepsilon ).$

I never remarked that there were two different definitions that are commonly given, before the recent discussion at Talk:Function of several real variables#Minor edit: Section on continuity and limit.

In French Wikipedia, it is the unpunctured definition that is given. In German Wikipedia, the punctured definition is given first, and later in the article the unpunctured definition is given in a section called (in German) "Newer definition". (The terms "punctured" and "unpunctured" are the translation of the German words that are used for comparing the definitions.)

I guess that the punctured definition is commonly used in US educational mathematics, while the unpunctured one is more commonly used in advanced mathematics. This needs verification.

I have no clear opinion which definition must be chosen for English Wikipedia, but whichever definition is kept, Wikipedia readers must be warned that both definitions are commonly used. As this implies to edit several articles, a discussion is needed here for fixing how we must proceed. D.Lazard (talk) 18:37, 25 November 2020 (UTC)

In my experience, the punctured definition is used universally in English. (But I am not an analyst, I cannot speak to what happens beyond the level of undergraduate education.) Both definitions are discussed in one of the articles, see Limit of a function#Deleted versus non-deleted limits where it is asserted (with citation) that punctured limits are "most popular". (Though it's not the issue here, I like "punctured" better than "deleted".) --JBL (talk) 18:43, 25 November 2020 (UTC)

The fact that the limit of a sequence/function can attain a value that is outside the original set of definition is a tremendously important fact at every level of real analysis. For example, how does one define the limit as $t\to \infty$ of a function $f:\mathbb {R} \to \mathbb {R}$ with the unpunctured definition of a limit? Mathematicians try their best not to treat infinity as a number, so with the punctured definition, it is straightforward. Many questions like this (compactness, closedness, the relationship between limits/continuity for sequences/functions) are much better understood if you remember from the beginning that limits can land outside the set you originally started with. I vote we use the punctured definition, and give an example such as the one above to explain pedagogically the difference between the two approaches in the article. Tazerenix (talk) 19:30, 25 November 2020 (UTC)

@Tazerenix: I think that's a different issue from the one here: if the point at which the limit is being taken is outside of the domain, then [D.Lazard's version of] the punctured and unpunctured definitions agree (see the universal quantifier $\forall x\in D$). The question here is rather whether (e.g.) the Kronecker delta function $\delta _{0,x}$ has a limit as $x\to 0$ or not. --JBL (talk) 19:38, 25 November 2020 (UTC)

Right. The punctured definition allows us to compute limits at removable singularities. Does the unpunctured definition essentially require the function to be continuous, or am I jumping the gun? Mgnbar (talk) 20:07, 25 November 2020 (UTC)

@Mgnbar: that's right, in the unpunctured definition, the condition "the limit exists at a point of the domain" means the same as "the function is continuous at that point". (That is the genesis of this discussion, see Talk:Function of several real variables -- it refers to Function_of_several_real_variables#Continuity_and_limit and in particular the sentence If a is in the interior of the domain, the limit exists if and only if the function is continuous at a in that section of the article.) --JBL (talk) 20:42, 25 November 2020 (UTC)

How is one supposed to discuss semi-continuity with the unpunctured definition? Ozob (talk) 16:12, 26 November 2020 (UTC)

@Ozob: Could you expand on what the issue is here? I was hoping someone better qualified than me would answer your question, but here's my understanding of why this may be a non-problem. Looking at the Semi-continuity article, the formal topological definition of upper semi-continuity at x₀ involving neighbourhood U isn't affected by switching between the punctured or unpunctured limit definitions, except possibly when f(x₀) = -∞. There's a lim sup formulation for metric spaces (hence for Rⁿ) which involves a limit, but lim sup for a function requires a one sided limit for ε>0 (so not punctured), where ε is the half-width of an interval about x₀. (Also, though probably irrelevant, the result is identical whether or not the sup is taken over a punctured or an unpunctured interval around x₀.) NeilOnWiki (talk) 15:30, 22 December 2020 (UTC)

@NeilOnWiki: Suppose that $f(z)=0$ for $z\neq 0$ and that $f(0)=1$. This function is upper semi-continuous, and the limit at $z=0$ exists (under the punctured definition). This kind of situation arises naturally, for example, in algebraic geometry (where $f$ is the fiber dimension of the blowup of the origin), but it is not even fully describable using the unpunctured definition.

My opinion is that the unpunctured definition is erroneous, and sources that use it are mistaken. I do not even see a reason for articles to discuss the unpunctured definition unless we can find sources stating that it is a common error. Ozob (talk) 16:02, 22 December 2020 (UTC)

Not sure about semicontinuity, but it would be good to pin down why the difference in definitions arises: eg. whether it's down to European vs. U.S. expectations; or recent trends; or if there's a compelling reason for choosing one definition rather than the other. Encyclopædia Britannica online uses the unpunctured version; as does my British-published Collins dictionary of Mathematics. But the Concise Oxford Dictionary of Mathematics (5 ed.) seems to favour the punctured version. Subjectively, the punctured version seems (to me) to introduce a condition that's irrelevant to the intuition that a function has a limit L at c if f(x) becomes more nearly equal to L as x moves increasingly close to c. Why delete c in the formal definition? It seems unnecessary. Under the unpunctured definition, if we need to exclude c (eg. at a singularity), then we can restrict the function domain accordingly. This, in effect, is what we do if we write $\lim _{\stackrel {x\to x_{0}}{x\neq x_{0}}}f(x)\to y$ (which we see in the Filters in topology article). This notation is redundant under the punctured definition, except for emphasis or disambiguation.

Interestingly, JBL's Limit of a function#Deleted versus non-deleted limits link observes that the unpunctured definition interacts more nicely with function composition (my wording). The source for this is several decades more recent (2015) than the multiple sources cited to support that the punctured definition is more popular (latest 1974). Looking at articles more widely in topology, it seemed to me that English Wikipedia is surprisingly consistent in preferring the punctured definition (generalised to open sets, neighbourhoods, nets, etc). The non-standard analysis topics may be less consistent, depending on whether or not an article considers 0 to be an infinitesimal (I'm fairly unconfident here).

Incidentally, the punctured definition vacuously implies (I think!) that if c is an isolated point, then any L in the codomain of f (and not just in the image of f) is a limit as x approaches c. The unique unpunctured limit is f(c). The latter seems less perverse — though it might just be that the example itself is fairly perverse. NeilOnWiki (talk) 20:46, 2 December 2020 (UTC)

I believe that at least in modern English sources, the punctured definition is used almost universally, at all levels of mathematics. Ebony Jackson (talk) 18:29, 5 December 2020 (UTC)

Yes, this is my experience as well. I find the "unpunctured definition" sort of bizarre, frankly. What is the point of taking a limit, if it has to actually be the value of the function at that point? It seems to be entirely redundant with the notion of continuity; it's not clear why you would need both. And it means that you can't, for example, write the definition of derivative as

$f'(x)=\lim _{h\rightarrow 0}{\frac {f(x+h)-f(x)}{h}}$

which is how it is usually presented in calculus classes. (I suppose you could quibble that $h=0$ is not in the natural domain of the right-hand side, but this strikes me as confusing and error-prone.) --Trovatore (talk) 00:14, 13 December 2020 (UTC)

@Trovatore: I think these are persuasive points. Even so, I'd like to stick up for the unpuncturists, as I think that, even if they're a minority, it's a mathematically valid position and one taken in at least some sources. I guess it'd be good to have an idea of how significant a minority they are.

On your first point, continuity isn't completely equivalent to the unpunctured limit existing (only when the point in question is in the function domain), so the unpunctured limit isn't "entirely redundant". To my mind, the unpunctured definition copes less bizarrely for Real valued functions on the Integers f: Z→R, where a puncturist could assert that 2x→0 as x→1 (remembering that x ∈ Z — admittedly I can't think why anyone would do this). These arguments may both be down partly to aesthetic preference.

It's a long time since I learnt calculus and I'm not a teacher. My impression is that any tutor using this definition makes resolutely clear in the preamble that h is non-zero, so there's no ambiguity over the domain. NeilOnWiki (talk) 15:23, 13 December 2020 (UTC)

Perhaps this is a matter of taste, but I find the punctured definition more appealing for functions $f\colon \mathbf {Z} \to \mathbf {R}$. This definition means that every real number is a limit of $f$ at every point. While ambiguity is not usually a desirable property, I think it is natural in vacuous situations like this. It is still the case that such an $f$ is continuous. Ozob (talk) 16:14, 22 December 2020 (UTC)

(←) Thanks, Ozob. Apologies for being so late replying. I agree this ambiguity is logically consistent, though it does also produce what seems like some odd results geometrically. For example, we now have a continuous f where there's a limit but not a unique one (even though R is a Hausdorff space). Hence, we may need to pause before writing that in general a Real-valued function f is continuous at c iff the limit exists and equals f(c), because we might have to choose our phrasing more carefully to account for non-uniqueness if there's a possibility that c is an isolated point (as happens with c ∈ Z above). Interestingly, the Net article has a definition of limit with a punctured flavour for a function from a metric space to a topological space, which does ensure uniqueness when the codomain is Hausdorff. It agrees with the punctured ε-δ definition when c is a cluster point (limit point), but not when c is isolated. Instead, in effect it avoids the vacuous condition for an isolated point and implies the limit either doesn't exist or uniquely equals f(c). (As far as I can tell, although it's not developed there, the obvious unpunctured counterpart would be fully consistent with the unpunctured ε-δ definition for both kinds of point.)

I found this a surprisingly interesting question, not least to see how this kind of Wikiepdia discussion is concluded. MOS:MATHS has a section on Mathematical conventions. Would it make sense to add an entry for Limit of a function there? I also wonder whether it might help future editors by adding a summary of some of the less obvious implications of the punctured definition (notably for function composition and isolated points), if this were put forward as the more popular approach. NeilOnWiki (talk) 14:21, 30 December 2020 (UTC)

Hi Everyone: In the absence of cries of "that's a terrible idea", I'm planning to make the edits to MOS:MATHS that I proposed in the previous paragraph, echoing the consensus here regarding the punctured version (and adding a pointer to this conversation on the MOS:MATHS Talk page). It strikes me that it would be a shame if the current conversation disappeared into the ether, especially considering D.Lazard's initial concerns over the plurality of articles and editors. It may take me a week or so to get round to it, so please stall me if my doing so seems inappropriate or somehow premature. NeilOnWiki (talk) 16:33, 13 January 2021 (UTC)

I'm sorry to raise a further question so late in the day, but in preparing for editing MOS:MATHS regarding the punctured definition, I've realised there's an additional question over which points c are permitted. Must c be a limit point of the domain D? And is the limit undefined if not (i.e. if it's an isolated point)?

Up until now, I'd tacitly assumed that it applied to any c in the closure of D (hence No to both preceding questions). But (ε, δ)-definition of limit § Precise statement for real-valued functions restricts c to being a limit point (likewise for metric spaces); and similarly in Limit of a function § More general subsets. Neither article says whether the limit is undefined (or defined differently) for isolated points; or if it extends to them unchanged. My guess is that we can indeed apply the punctured definition both to the limit points of D and its isolated points: it's just that authors may prefer not to lump them together because (as we've noted) the latter case would imply a condition which is satisfied vacuously and hence (generally) not uniquely. Can anyone here help clarify whether the generally accepted pre-conditions for the punctured definition really does exclude taking the limit at an isolated point? (Incidentally, there are implications on the relationship between limits and continuity of a function if so.) I'm not an analyst, and I don't have easy access to any definitive texts. Thanks. NeilOnWiki (talk) 16:37, 27 January 2021 (UTC)

THe definition without the $0<|x-c|$ condition seems strange to me. For a discontinuous function $f(x)={\begin{cases}1&{\text{for }}x=0\\0&{\text{otherwise}}\end{cases}}$ it would imply $f$ has no limit at $x=0$. --CiaPan (talk) 18:06, 27 January 2021 (UTC)

@CiaPan: I think you're in the majority in this. The question I had was about when you have points in the function domain which are "discontinuous" (what I meant by isolated), such as a function g: Z→R from the Integers to the Reals, e.g. g(n) = πⁿ. I originally thought that taking the limit was still well-defined for this case (though multi-valued, because you get a vacuous "for all" condition); but the Wikipedia articles I looked at seem to restrict the definition so you couldn't apply it to cases like this. I'd like to find out if this restriction really does hold and if the limit is therefore classed as undefined at c ∈ Z for a function on the Integers like g(n). NeilOnWiki (talk) 14:15, 28 January 2021 (UTC)

@NeilOnWiki: I have no idea whether defining a limit of a function at a separated (isolated) point of a domain makes any sense. The limit is a formalized way to express 'approaching to'. When $k\in D$ is an isolated point, the argument $x$ of a function $f$ can not 'approach' $k$. Then a function value $f(x)$ can not 'approach' $f(k)$, hence there is no need to define a limit of $f$ as $x\to k$. Additionally, we consider limits at boundary points of an open domain (e.g., $\lim _{x\to 0}{\frac {x}{x}}$, which is obviously 1); that will not apply to isolated points, either, because we have no value of a function anywhere 'close to' the isolated point. --CiaPan (talk) 17:29, 28 January 2021 (UTC)

A punctured definition seems like $\lim _{x\to c}\sup f(x)=L\ and\ \lim _{x\to c}\inf f(x)=L$, a non-punctured definition seems like $\lim _{x\to c}\max f(x)=L\ and\ \lim _{x\to c}\min f(x)=L$. When defining a function f for a real number, we often take a sequence of rational numbers $\{c_{n}\}_{n=1}^{\infty }$ that converges to the real number c.--SilverMatsu (talk) 01:28, 29 January 2021 (UTC)

To be fair, I think that if you have a continuous function on the Rationals, then you can still plug-in the non-punctured definition to extend it to the Reals, i.e. it still works for c ∉ D when c's a limit point not in D (so it isn't quite like max vs. sup). In answer to CiaPan (for which thanks): although I agree it jars with geometric sense, once you have a formal punctured definition like the one given by D.Lazard you can just mechanically work through the logic to test if a given f(x)→L as x→c. The formal reasoning goes through even when c is isolated, in which case it implies f(x)→L for any L. (It's a bit like the formal definition is saying "if you can get close to c, then f must get close to L", so if you can't then the requirement goes away, ie. gets satisfied vacuously.) This may seem odd, but I think it allows analysts to come up with an equivalence between continuity and limits: viz. a general function f: D→R is continuous at c iff the limit exists there and f(x)→f(c) as x→c. This would work formally even when D = Z, in which case we'd have continuity at every c ∈ Z in line with what a topologist would say. But our (ε, δ)-definition of limit article seems very careful to exclude taking the limit at an isolated point. I don't really see why it does that (unless it's established practice in analysis), as it seems unnecessary and weakens the continuity/limits equivalence. NeilOnWiki (talk) 16:49, 29 January 2021 (UTC)

Thank you for teaching me. What do you think of function that is continuous everywhere but differentiable nowhere?--SilverMatsu (talk) 01:14, 30 January 2021 (UTC)

I'm really not trying to teach anyone; just trying to understand whether it's right for our articles to restrict the punctured definition to limit points (cluster points). Thanks for the link to the Weierstrass function, which I'd not seen before. NeilOnWiki (talk) 17:13, 31 January 2021 (UTC)

Draft:Third Vote could use a bit of a review

Nikogosyan has written a draft for the third vote election method. While I'm normally pretty mathematically inclined, and have an interest in electoral methods, I think this draft is pretty far over my head. If there's anyone who's able to give a general "OK" that it should be ready for mainspace, that would be very helpful—my main concern is that the draft might be crossing the line of being too close to an essay, but I'm not sure to what extent that policy applies in a mathematical article. Perryprog (talk) 18:52, 31 January 2021 (UTC)

I don't think the extent to which this is mathematical is relevant. (I also don't think the complication of this method has much to do with mathematics.) I agree with you that it is very essay-like. --JBL (talk) 19:05, 31 January 2021 (UTC)

Perhaps you're right; I guess I didn't read the prose too in-depth and just got scared by the intimidating wikitables :). WP:TECHNICAL#Rules of thumb could also be of some use here, I think. Perryprog (talk) 19:21, 31 January 2021 (UTC)

Yes, the tables are intimidating and also poorly placed: each of them comes before any text that would allow the reader to make sense of it. Particularly egregious are the last two lines in Table 1 (Popularity+ and Parliament faction size+), which are not explained until the subsection "Faction equalization effect" of the Criticism section, practically at the end of the article. --JBL (talk) 19:28, 31 January 2021 (UTC)