Talk:Chebyshev polynomials

Notes on transcription and correct reading

Name of this guy in original ia Пафнутий Львович Чебышёв Львович means that his father's name was Lev (There is a big cat from genus Panthera called lion) Чебышёв - should be transcripted as Chebyshov Ч should be read softly ё - here we have two dots over e so there cannot be e but o (Russians usually don't write these two dots but it does not mean that they are not there) Accent should be placed on the last syllable If you dont believe me consult it with Russian speaking guy

[Untitled]

Michael, it's not actually your browser. You can set that in your wikipedia preferences - by default it will render simple equations in normal text. And you apparently just complicated the equation by adding some white space ;-) snoyes 23:37 Mar 1, 2003 (UTC)

You're right; it works. Thanks. Why is the default the option that renders things as normal text, given that that is never what is intended when things are set with TeX? Michael Hardy 23:45 Mar 1, 2003 (UTC)

I don't know why that is the default option - best to ask one of the developers, maybe on the village pump. --snoyes 23:53 Mar 1, 2003 (UTC)

And BTW, there don't seem to be any other articles about special polynomial sequences on Wikipedia. Michael Hardy 23:45 Mar 1, 2003 (UTC)

---

//Here's the polynomial sequence on the page. Κσυπ Cyp   19:47, 25 Jan 2004 (UTC)

#include <stdio.h>

int asdf(int, int);
int choo(int, int);

int main() {
  int n, k;
  for(n=0;n<=16;++n) {
    printf("T<sub>%d</sub>(x)=", n);
    for(k=0;k<=n/2;++k) {
      printf("%s%d", k&&asdf(n, k)>=0?"+":"", asdf(n, k));
      if(n-2*k) printf("x");
      if(n-2*k>1) printf("<sup>%d</sup>", n-2*k);
      if(k==n/2) printf("<br>\n");
    }
  }
  return(0);
}

int tmp[1000];

int asdf(int n, int k) {
  int r, a;
  if(!n) return(1);
  if(!k) return(1<<(n-1));
  for(a=r=0;a<k;++a) r-=(asdf(n, a)*choo(n-2*a, k-a))>>(2*(k-a));
  return(r);
}

int choo(int n, int k) {
  int x, y;
  for(tmp[0]=y=1;y<=n;++y) for(tmp[x=y]=0;x;--x) tmp[x]+=tmp[x-1];
  return(tmp[k]);
}

---

If we recursively define a function (the letter ξ an arbitrary choice)

$\xi (n,0)=\lceil 2^{n-1}\rceil$

$\xi (n,k)=-\sum _{a=0}^{k-1}{\frac {\xi (n,a)}{4^{k-a}}}{n-2a \choose k-a}$

then

$T_{n}(x)=\sum _{k=0}^{\lfloor n/2\rfloor }\xi (n,k)x^{n-2k}$

Maxima and minima

It looks to me like the local maxima of the Chebyshev polynomials of the first kind all have value 1, and the local minima all have value -1. If so (I don't know enough to say whether it is actually true), then this is an interesting property that should be mentioned somewhere on the page. I'm guessing it probably pops right out of one of the definitions, but that's not transparent to a muggle like me. -dmmaus 04:01, 15 August 2005 (UTC)

I think that follows from the definition:

$T_{n}(x)=\left\{{\begin{matrix}&\cos(n\arccos(x))&{\mbox{ , }}\ x\in [-1,1]\\&\cosh(n\,\mathrm {arcosh} (x))&{\mbox{ , }}\ x\geq 1\\&(-1)^{n}\cosh(n\,\mathrm {arcosh} (-x))&{\mbox{ , }}\ x\leq -1\\\end{matrix}}\right.$

Obviously, for x in [-1, 1] since the behaviour is given by the cosine, the local maxima and minima are 1 and -1. If one can prove that there are no maxima/minima outside this interval, then you assertion would follow. Oleg Alexandrov 04:07, 15 August 2005 (UTC)

Ah, good point. I'm afraid my eyes glaze over when I see hyperbolic trig functions in the vicinity. :-) Still, I think it'd be nice to mention this property explicitly in the article, since it's the standout physical feature of the graphs when you look at them. -dmmaus 08:54, 15 August 2005 (UTC)

I don't know that much about these polynomials to attempt to write in the article. Let us hope somebody else will mention this fact. Oleg Alexandrov 17:04, 16 August 2005 (UTC)

I've mentioned it.

There's also an interesting paralel here, which is a bit OT. Notice that for a degree n polynomial (with fixed leading coeff) to have a minimal maximum absolute value on an interval, it has to reach this maximal absolute value the highest number of times possible: n+1 times. (Clearly a degree n polynomial can have at most n-1 places where its derivative is 0, so it's impossible that is has more than n+1 extremal point in an interval.) This paralels up nicely with a general theorem in numerical analysis that the polynomial p of degree n (with no restriction on the leading coeff) that approximates a continuous function f in the sense that max|p-f| on an interval is minimal, |p-f| has to take this maximum n+1 times. I'm not sure how this general theorem is stated exactly, and I also don't know if it has any connection to this. – b_jonas 21:45, 29 December 2005 (UTC)

You have cited famous approximation theorem by Chebyshev (not later than 1857). And yes, extremal properties of polynomials of first kind (minimal C-norm) follow from this theorem immediately. Mir76 (talk) 16:18, 16 August 2009 (UTC)

Let me say thanks to User:172.128.197.224 for his correction of the statement in the second paragraph of the article. – b_jonas 18:30, 18 February 2006 (UTC)

Optimal approximations to arbitrary functions

The theorem says that, if an n^th degree polynomial P approximates a function F on a closed interval in such a way that the error function P(x)-F(x) has n+2 maxima, and those maxima have alternating signs and the same absolute value (that is, P-F has maxima +ε, -ε, +ε, ...), then that polynomial is locally optimal. That is, there is no other polynomial close to P that has maximum error less than ε. This locality situation is the same as the situation in differential calculus: elementary calculus problems involve finding the maximum value of a function. One sets the derivative to zero. That gives a local maximum (or minimum), but there might be other solutions that are far away.

I will outline the proof below. I think this theorem, and its proof, are interesting, but don't seem to be part of the standard maths curriculum. (Orthogonal polynomials in general seem to be a mathematical orphan child.) I would suggest that this proof ought to be on a separate "proof" page. WP is trying to minimize the number of such pages, because WP isn't a textbook, but this would probably be a good candidate.

The proof really requires a graph to motivate it. I'm new to WP; it should be made by someone who is comfortable making and uploading graphs, such as the graphs on the Chebyshev page.

Here's the proof: Suppose P(x)-F(x) has n+2 maxima, alternating signs, all with absolute value ε. Suppose Q is another n^th degree polynomial very close to P, that is strictly better. Superimpose tha graph of Q-F on that of P-F. Since Q is close to P, they will be about the same, but at the n+2 points where P-F is maximal, Q-F must lie strictly inside P-F. This is because the maximal excursions of Q-F are δ, which is strictly less than ε. So (Q-F)-(P-F) is strictly nonzero at those n+2 points, with alternating signs. But (Q-F)-(P-F) = Q-P, an n^th degree polynomial. Since it has n+2 alternating nonzero values, it must have n+1 roots. But n^th degree polynomials can't do that.

Now here's an outline of why Chebyshev interpolation gets very close to this optimum polynomial: If the Taylor series for a function converges very rapidly (the way, for example, sin, cos, and exponential do), the error from chopping off the series after some finite number of terms will be close to the first term after the cutoff. That term will dominate all the later terms, and that term plus all the later ones are of course the error function. The same is true if we use a Chebyshev series instead of a Taylor series.

Now if we expand F(x) in its Chebyshev series, and chop it off after the T_n term, we have an n^th degree polynomial. The error is dominated by the T_n+1 term, so it looks like T_n+1. T_n+1 is level, with n+2 maxima, with alternating signs. This means the error function is approximately level, and hence, by the theorem above, this polynomial is approximately optimal.

If one wants a truly optimal polynomial, one "tweaks" the polynomial P so that its error excursions are truly level. Remes' algorithm does this.

Should there be a separate WP page for this stuff, linked from Chebyshev? What should it be called? William Ackerman 18:46, 21 February 2006 (UTC)

Thanks for stating the theorem. – b_jonas 21:13, 21 February 2006 (UTC)

The relationship

I removed the following:

Starting with T0( cos u ) = 1 and T1( cos u ) = cos u, one quickly calculates that cos 2u = 2 cos^2 u - 1, cos 3u = 4 cos^2 u - 3 cos u, cos 4u = 8 cos^4 u - 8 cos^2 u + 1, etc. By setting u = 0, whence cos u = 1, one can immediately detect that both sides of each equation evaluate to unity, which is a good clue that the relationship may be correct—which, of course, it is.

What does "the relationship may be correct" mean? which relationship? Not $T_{n}(\cos \theta )=\cos(n\theta )$ because thats a definition. It may show that the trignometric identities are true, but thats not the point of the article. PAR 13:27, 11 August 2006 (UTC)

reasoning not correct

I removed the following part because it is nonsense:

From reasoning similar to that above, one can develop a closed-form generating formula for Chebyshev polynomials of the first kind:

$\cos(n\theta )={\frac {e^{in\theta }+e^{-in\theta }}{2}}={\frac {(e^{i\theta })^{n}+(e^{i\theta })^{-n}}{2}}\,\!$

which, combined with DeMoivre's formula: THIS IS FALSE:

$\!e^{i\theta }=\cos \theta +i\sin \theta =\cos \theta +i{\sqrt {1-\cos ^{2}\theta }}=\cos \theta +{\sqrt {\cos ^{2}\theta -1}}\,\!$

yields:

$\cos(n\theta )={\frac {\left(\cos \theta +{\sqrt {\cos ^{2}\theta -1}}\right)^{n}+\left(\cos \theta +{\sqrt {\cos ^{2}\theta -1}}\,\right)^{-n}}{2}}\,\!$

which, of course, is far more expedient for determining the cosine of N times a given angle than is cranking through almost N rounds of the recursive generator calculation. Finally, if we replace $\cos(\theta )$ with x, we can alternatively write:

$T_{n}(x)={\frac {\left(x+{\sqrt {x^{2}-1}}\right)^{n}+\left(x+{\sqrt {x^{2}-1}}\right)^{-n}}{2}}.$

Linear operators

Should someone write a section about these as the eigenfunctions of a second order linear differential operator (which is symmetric wrt the inner product $(f,g)=\int _{-1}^{1}f(x)g(x){\frac {1}{\sqrt {1-x^{2}}}}\,dx$)? I would, but I'm not really comfortable with the theory yet... Marbini (talk) 17:54, 2 April 2008 (UTC)

Factoring

Can anything of interest be said about factorization of Chebyshev polynomials? Michael Hardy (talk) 00:04, 9 December 2008 (UTC)

Apparently. Googling for "factoring chebyshev" turned up this and this which seem to at least contain some pointers. Fredrik Johansson 00:15, 9 December 2008 (UTC)

So there's really not very much.....maybe. There is a sense in which the spread polynomials S_n, satisfying

$S_{n}(\sin ^{2}\theta )=\sin ^{2}(n\theta )\,$

are trivially equivalent to the Chebyshev polynomials T_n, satisfying

$T_{n}(\cos \theta )=\cos(n\theta ).\,$

But the article titled spread polynomials has some material on factorization, and maybe there the story is different. Or maybe not? Michael Hardy (talk) 21:36, 10 March 2009 (UTC)

Some interest in factoring the Chebyshev T-polynomials stems from reduction of cosines to rational numbers as PDF here. - R. J. Mathar (talk) 19:35, 12 March 2014 (UTC)

A few remarks about divisibility of Chebyshev polynomials by other Chebyshev polynomials have now been added in the section Composition and divisibility properties. Perhaps more could be said, but I wanted to keep it simple, and also didn't feel qualified to say more. Will Orrick (talk) 23:24, 8 January 2022 (UTC)

Explicit formulas

Can someone provide sources for the given explicit formulas, or at least some hints on how to obtain them?

I found some other formulas in Mason, J. C. "Chebyshev polynomials", 2003. CRC Press, 2003., pp.35-36 and in Snyder, M. A. "Chebyshev methods in numerical approximation", Englewood Cliffs, NJ, USA, 1966., p. 14.

In the first book, there are some other references to Rivlin (1974) and Clenshaw (1962). Mstempin (talk) 21:23, 3 August 2009 (UTC)

After expanding formula

$T_{n}(x)=\sum _{k=0}^{\lfloor {\frac {n}{2}}\rfloor }{n \choose 2k}x^{n-2k}(x^{2}-1)^{k}$

for n=8 i made conjecture that

$T_{n}(x)=\sum _{k=0}^{\lfloor {\frac {n}{2}}\rfloor }\sum _{m=k}^{\lfloor {\frac {n}{2}}\rfloor }(-1)^{k}{n \choose 2m}\cdot {m \choose k}x^{n-2k}$

This conjecture should be proven and in my opinion induction should be enough

For the sum

$\sum _{m=k}^{\lfloor {\frac {n}{2}}\rfloor }{n \choose 2m}\cdot {m \choose k}$

i found with help of Wolfram Alpha only piecewise definition

$\sum _{m=k}^{\lfloor {\frac {n}{2}}\rfloor }{n \choose 2m}\cdot {m \choose k}={\begin{cases}1\qquad n=0\\{\frac {n}{2}}\cdot {\frac {1}{n-k}}\cdot {n-k \choose k}\cdot 2^{n-2k}\qquad n>0\end{cases}}$ — Preceding unsigned comment added by 188.47.53.242 (talk) 11:56, 4 July 2023 (UTC)

History

I believe Chebyshev developed these polynomials while attempting to solve the problem of the inversor, a mechanical linkage that converts circular motion into linear motion. DonPMitchell (talk) 03:42, 13 March 2010 (UTC)

Dickson polynomial

The following discussion is closed. Please do not modify it. Subsequent comments should be made in a new section. A summary of the conclusions reached follows.

The result of this discussion was no consensus. NukeofEarl (talk) 16:24, 30 October 2014 (UTC)

---

I would like to propose merging Dickson polynomial into Chebyshev polynomials. All the properties, except the analysis over finite fields, is already here. — Arthur Rubin (talk) 00:27, 21 September 2011 (UTC)

Dickson polynomials are mainly studied over finite fields, when they are not equivalent to Chebyshev polynomials.

It appears in their application, these two polynomials groups do really differ, and I wouldn't support a merge... Gryllida (talk) 22:16, 2 January 2014 (UTC)

The discussion above is closed. Please do not modify it. Subsequent comments should be made on the appropriate discussion page. No further edits should be made to this discussion.

Proposed section to merge: Spread polynomials and some other Chebyshev-linked Polynomials

spread polynomials are in a sense equivalent to the Chebyshev polynomials of the first kind, but enable one to avoid square roots and conventional trigonometric functions in certain contexts, notably in rational trigonometry.

Dickson polynomials ,D_n(x), are equivalent to Chebyshev polynomials T_n(x), with a slight and trivial change of variable:

$D_{n}(2x)=2T_{n}(x)\,$

Dickson polynomials are given by:

$D_{n}(x)=\sum _{p=0}^{\lfloor n/2\rfloor }{\frac {n}{n-p}}{\binom {n-p}{p}}(-x^{n-2p}.$

The first few Dickson polynomials are

$D_{0}(x)=2\,$

$D_{1}(x)=x\,$

$D_{2}(x)=x^{2}-2\,$

$D_{3}(x)=x^{3}-3x\,$

$D_{4}(x)=x^{4}-4x^{2}+2\,$

Boubaker polynomials are also linked to Chebyshev polynomials. they can be defined by the recurrence relation:

$B_{n}(x)={\begin{cases}1,&{\mbox{if }}n=0,\\x,&{\mbox{if }}n=1,\\xB_{n-1}(x)-B_{n-2}(x),&{\mbox{if }}n\geq 3.\end{cases}},$

Lucas polynomials L_n(x) are a polynomial sequence defined by the following Chebyshev-like recurrence relation:

$L_{m}(x)={\begin{cases}2,&{\mbox{if }}m=0\\x,&{\mbox{if }}m=1\\xL_{m-1}(x)+L_{m-2}(x),&{\mbox{if }}m\geq 2.\end{cases}}$

References

Possible corrections

Section on Orthogonality and Example 1 has errors

The session on Orthogonality has a somewhat major issue, namely that the discrete orthogonality condition as written is wrong. The discrete orthogonality condition neglects to mention that the formula given is only valid for $0\leq i,j\leq N-1$. For other values of $i,j$ the aliasing condition gives alternating sign "tiled" versions of what's currently written.

Also, in Basis Set, Example 1, the closed-form Chebyshev coefficients $a_{n}$ for Log(1+x) is given correctly. However, immediately afterward it is written that these same coefficients $a_{n}$ can alternately be computed via the discrete Gauss-Lobatto sum. This is false, as is readily verified numerically. The $a_{n}$ in the Gauss-Lobatto discretization are the aliasing-folded versions of the $a_{n}$ computed in the continuous integral case, and are not the same. To a newcomer who does not know this already, this is completely non-obvious, especially considering that the same symbol $a_{n}$ is used for both. I suggest that different symbols should be used for the two cases to emphasize the fact that the two quantities are different. — Preceding unsigned comment added by MathDoobler (talk • contribs) 04:13, 3 January 2014 (UTC)

Wrong discrete orthogonality

There is a very wrong formula in the Orthogonality section. The discrete orthogonality shouldn't have the weight that appears in the continuous orthogonality condition. See for example Eq. (3.30) of https://www.siam.org/books/ot99/OT99SampleChapter.pdf Revision of 14 March 2016 should be undone because wrong! Pliskin14 (talk) 09:21, 12 May 2016 (UTC)

Third and Fourth Kind

Is there any reason why the Chebyshev polynomials V_n(x) and W_n(x) of the third and fourth kind are not mentioned (at least: defined) in the article? R. J. Mathar (talk) 19:29, 12 March 2014 (UTC)

I hadn't heard of them. Got a reference which is uses them? — Arthur Rubin (talk) 16:27, 18 March 2014 (UTC)

For example

1. J. C. Mason (2010). "1.2.3". Chebyshev Polynomials. CRC Press.
2. J. C. Mason. "Chebyshev polynomials of the second, third and fourth kinds in approximation...". doi:10.1016/0377-0427(93)90148-5.
3. J. C. Mason, G. H. Elliott (1995). "Constrained near-minimax approximation...". Num. Alg. No. 9. pp. 39–54. doi:10.1007/BF02143926.
4. S. E. Notaris. "Interpolatory quadrature.". doi:10.1016/S0377-0427(97)00018-6.
5. "A survey on third and fourthkind...". Appl. Math Comput. 2008. doi:10.1016/j.amc.2007.09.018.
6. G. Heinig (2001). "Chebyshev-hankel matrices...". Lin. Alg. Applic. pp. 181–196. doi:10.1016/S0024-3795(00)00300-1.

and many, many others R. J. Mathar (talk) 17:25, 1 May 2014 (UTC)

In response to my earlier edit being reverted, p_1 as listed are not orthogonal to the respective weight functions despite being consistent with the definitions. While definitions for V and W are sometimes interchanged, the ones used here are consistent with Mason and Handscomb's 2003 book (definition 1.3), but the weight functions are not (section 4.2.2). In regards to this page, I believe that the weight functions for V and W need to be switched. — Preceding unsigned comment added by Youboo4 (talk • contribs) 21:12, 21 July 2022 (UTC)

I now agree that the weights were incorrect and have switched them. Do you agree that everything else is consistent now? Will Orrick (talk) 22:15, 21 July 2022 (UTC)

Yes, I think that was the only issue. Youboo4 (talk) 23:45, 21 July 2022 (UTC)

Regarding the 24 May 2024 edits and reversions: it seems to have escaped the participants' notice that Chebyshev polynomials of the third and fourth kind were already in the article. My suggestion would be to roll the article back to its state prior to 24 May 2024 and to have a discussion before changing the article in such a drastic way. (As it is now, the article is only partially rolled back.) Will Orrick (talk) 06:50, 26 May 2024 (UTC)

I rolled back all the recent edits that introduced polynomials of the third and fourth kind early in the article. I'm willing to entertain contrary opinions, but my feeling is that 99% of readers coming to this article are here to learn about polynomials of the first or second kind, which have a history going back to Viète and are ubiquitous in mathematics and science. Polynomials of the third and fourth kinds, while implicit in the work of Jacobi, and having been studied in the aerospace community for more than half a century, have only started to be called by that name in the past few decades and have a much smaller literature. I feel that introducing them early in the article gives them undue weight.

Furthermore, it is untenable to have an article that opens by stating that there are two sets of polynomials, then immediately introduces four of them, but then goes back mostly to talking as if there are only two sets, only to reintroduce the other two sets at the very end of the article as if they hadn't been mentioned before. This is what we had prior to the rollback. If treating Chebyshev polynomials as four sets of polynomials is the right thing to do from the outset, then the article needs to maintain that point of view throughout.

I'm not saying that such a change would need to take place in a single edit—that would be impractical—but anyone proposing to make that change needs to provide a rationale that achieves broad support from other editors and needs to outline a plan for carrying it out. Will Orrick (talk) 16:09, 26 May 2024 (UTC)

I think it's possibly worth mentioning in the lead section, but perhaps not worth elaborating too much up front. My impression is that these are mostly used in the context of fast transforms, where a polynomial of type X is broken into sub-transforms of types X, Y, Z (precisely which ones depends on X and on the degree). –jacobolus (t) 07:24, 27 May 2024 (UTC)

The questions of whether and where to mention the third and fourth kinds comes down to how widespread usage of those terms is. I was the person who added mention of them to the article back in January 2022. I did that because R. J. Mathar's comment on this Talk page persuaded me that usage of the terms was common enough that we couldn't ignore it. On the other hand, my judgement was that this usage was restricted to certain specialties and was relatively new, so I restricted the mention to a late section of the article, "Families of polynomials related to Chebyshev polynomials".

Whenever terminology with longstanding use is adopted for a new purpose, the new thing is piggybacking on the fame of the older one. Sometimes the new usage catches on; sometimes it doesn't; sometimes it catches on in some quarters but not in others. I find it hard to judge whether the applications you mention are important enough to mention in the lead of this article. Will Orrick (talk) 11:46, 28 May 2024 (UTC)

Sure, I agree. I might even focus on the "first-kind" T polynomials, which are what people mean when they say "Chebyshev polynomials" without qualification, and seem by far the most important to discuss, and then lump the second, third, and fourth-kind spin-offs into a later section. –jacobolus (t) 15:59, 28 May 2024 (UTC)

This is valuable input, and shows that there is a wide variety of different views. It's hard to get many editors to weigh in on issues like this, but I hope we will hear from others.

I think I learned this topic from volume II of Bateman (1953), or maybe from Abramowitz and Stegun (1964), both of which use the language "first kind", "second kind" and deal with two sets of polynomials from the start. It's also interesting that polynomials essentially equivalent to T(x) and U(x) already appear together in Viète's book (1646). But it seems you may be right that this "first kind", "second kind" vocabulary is of relatively recent vintage. I find very little evidence of this language prior to Bateman. Regarding terminology, Bateman (p. 183) has this to say,

"Sometimes (especially in the French literature) orthogonal polynomials in general are called Tchebichef polynomials. There are also several special systems of orthogonal polynomials called Tchebichef polynomials. In this chapter we shall reserve the name Tchebichef polynomials of the first and second kind for the suitably standardized orthogonal polynomials associated with

$a=-1,\quad b=1,\quad w(x)=(1-x^{2})^{\mp 1/2},\quad X=1-x^{2}$."

A quick search yielded articles on Chebyshev polynomials of the fifth, sixth, seventh, and eighth kinds, so it looks like this is never going to end. Will Orrick (talk) 13:19, 29 May 2024 (UTC)

It's probably worth going with whatever is used in widely cited dedicated sources, e.g. Rivlin (1974) or Mason & Handscomb (2002). I don't feel particularly motivated to do a survey right now though. –jacobolus (t) 14:57, 29 May 2024 (UTC)

Just my perspective, but I've read through most of this page multiple times in the capacity of my research (computer science / applied math), and not yet encountered a reason to use Chebyshev polynomials of the third & fourth kind. My impression is that the first kind is by far the most important, and expressions involving the second kind come up often when analyzing the first kind, so it is important by proxy. I haven't seen any use for these other polynomials in the areas I work in, and to my knowledge it's not even common to find it in approximation theory textbooks. So, keeping weight off of the third and fourth kind makes sense to me. Fawly (talk) 20:26, 29 May 2024 (UTC)

So I guess the Mason of Mason and Handscomb is the Mason that wrote all those articles on 3rd and 4th kind polynomials mentioned by R. J. Mathar in the opening comment of this thread. It's therefore perhaps not surprising that the blurb for the book says that it treats "all four kinds" of Chebyshev polynomials, and, indeed, the table of contents bears that out.

Rivlin had a second edition in 1990, which was just about the time that the 3rd kind/4th kind terminology was being introduced, so one wouldn't expect to find it in that book. From what I can see of the first chapter, "the" Chebyshev polynomials are the T(x). But Rivlin does introduce the second kind polynomials as early as page 7.

The NASA document by Desmarais and Bland, cited in the bibliography discusses the original motivation for looking at the 3rd and 4th kind polynomials.

I don't work in approximation theory, so I don't have much of an opinion. Chebyshev polynomials, of both of the first two kinds, come up from time to time in my work in mathematical physics, combinatorics, and number theory, but that's about it. Will Orrick (talk) 13:18, 30 May 2024 (UTC)

Fox & Parker (1968) also seems to focus on the T polynomials. –jacobolus (t) 02:30, 31 May 2024 (UTC)

Boyd (1989) Chebyshev and Fourier Spectral Methods focuses on the T polynomials, but also has two appendices covering the "Chebyshev polynomials: ⁠$T_{n}(x)$⁠" and the "Chebyshev polynomials of the second kind: ⁠$U_{n}(x)$⁠", respectively. Davis (1963) Interpolation and Approximation also focuses on the T for Tschebyscheff polynomials, and when he later introduces the U polynomials, describes the two types as "Tschebyscheff polynomials (of the first kind)" and "Tschebyscheff polynomials of the second kind". –jacobolus (t) 02:47, 31 May 2024 (UTC)

Trefethen (2013) Approximation Theory and Approximation Practice focuses on the T polynomials. –jacobolus (t) 02:50, 31 May 2024 (UTC)

Cheney (1966) Introduction to Approximation Theory focuses on the "Tchebycheff polynomials" T, but introduces the "Tchebycheff polynomials of the second kind" U in an exercise, and then uses them a couple more times toward the end of the book. –jacobolus (t) 03:22, 31 May 2024 (UTC)

Series coefficients found using Gauss–Lobatto zeros are approximate

I edited the section "As a basis set," Example 1, to indicate that the coefficients found by exploiting the discrete orthogonality property and the Gauss–Lobatto zeros give approximate results. This is discussed in Numerical Recipes (which is about my level of expertise here).

Also, I believe that the "exact" results in the same example for n > 0 are wrong. I don't have a correct form but I have a short Mathematica notebook which compares these exact coefficients with the approximate ones using a large N and they aren't even close. — Preceding unsigned comment added by 97.117.195.123 (talk) 10:33, 22 September 2015 (UTC)

Oh--sorry. The comment not far above this, "Section on Orthogonality and Example 1 has errors," deals with the same subject. — Preceding unsigned comment added by 97.117.195.123 (talk) 10:35, 22 September 2015 (UTC)

External links modified

Hello fellow Wikipedians,

I have just modified one external link on Chebyshev polynomials. Please take a moment to review my edit. If you have any questions, or need the bot to ignore the links, or the page altogether, please visit this simple FaQ for additional information. I made the following changes:

• Added archive https://web.archive.org/web/20070529221407/http://math.fullerton.edu:80/mathews/n2003/ChebyshevPolyMod.html to http://math.fullerton.edu/mathews/n2003/ChebyshevPolyMod.html

When you have finished reviewing my changes, please set the checked parameter below to true or failed to let others know (documentation at {{Sourcecheck}}).

This message was posted before February 2018. After February 2018, "External links modified" talk page sections are no longer generated or monitored by InternetArchiveBot. No special action is required regarding these talk page notices, other than regular verification using the archive tool instructions below. Editors have permission to delete these "External links modified" talk page sections if they want to de-clutter talk pages, but see the RfC before doing mass systematic removals. This message is updated dynamically through the template {{source check}} (last update: 5 June 2024).

• If you have discovered URLs which were erroneously considered dead by the bot, you can report them with this tool.
• If you found an error with any archives or the URLs themselves, you can fix them with this tool.

Cheers.—InternetArchiveBot (Report bug) 23:37, 20 November 2016 (UTC)

Proposed remove mention of spread polynomials

They seem to be part of the research program of one author and to have not caught on more widely. Googling "spread polynomials" and "Chebyshev polynomials" gives me 1.5k vs 364k hits. Spread polynomials are introduced in a book which is not widely cited apart from by its author. The work may be good but that shouldn't be wikipedia's job to evaluate. This mathoverflow thread suggests that they are basically equivalent, and that it's enough to just describe shifted Chebyshev polynomials. I propose removing all mention of spread polynomials from the article. The only reason to keep them, I think, would be if we find the approach to geometry in the book "Divine Proportions..." to be an important enough application of Chebyshev polynomials to mention in the main article. But Chebyshev polynomials have many many important applications I'd mention first, like their role in filter design, linear algebra algorithms, etc. Aram.harrow (talk) 19:17, 23 March 2021 (UTC)

update: I've edited the article to remove them. Aram.harrow (talk) 20:02, 28 March 2021 (UTC)

Generalized Chebyshev polynomials

The term "generalized Chebyshev polynomials" appears to have a variety of uses: among others, it is used to refer to Shabat polynomials, to certain two-variable or multi-variable generalizations, and to generalizations to unions of disjoint intervals. As used in the section "Generalized Chebyshev polynomials", however, it appears to refer to objects that are not actually polynomials most of the time. I have not been able locate any sources attesting to this use of the term. Would anyone be able to help? I have placed a "dubious" tag on the section. Will Orrick (talk) 05:41, 9 January 2022 (UTC)

I have removed the section. It seems unlikely that this is a common use of the term "generalized Chebyshev polynomials" if extensive searching doesn't turn up any examples of the term being used in this way. Will Orrick (talk) 03:40, 15 January 2022 (UTC)

Hyperbolic trigonometric functions; romanization of "Chebyshev"

Your argumet of cosh is purely imaginary assuming both x,t in R in your exponential generating function of T_{n}
Would be better to use trig cos instead
By the way this guy was named Chebyshov because over last e there is diaeresis which Russians usually dont write
Russians usually dont write it but diareresis is still here
— Preceding unsigned comment added by 188.47.27.31 (talk) 17:57, 29 May 2023 (UTC)

Wikipedia's biographical article on Chebyshev mentions the romanization issue. Unfortunately, the current spelling is well established in the mathematics literature, and Wikipedia needs to reflect what's in the majority of authoritative sources. Will Orrick (talk) 01:31, 30 May 2023 (UTC)

Anyone interested in Chebyshev and romanization might enjoy Philip Davis's The Thread. –jacobolus (t) 16:09, 28 May 2024 (UTC)

Yes but current transcription leads to misreading this name Many people misread this name — Preceding unsigned comment added by 188.47.53.242 (talk) 23:38, 7 July 2023 (UTC)

Incorrect integration formula for n=0

In the integration "The last formula can be further manipulated to express the integral of Tn as a function of Chebyshev polynomials of the first kind only:" the derivation is incorrect for n=0. It uses T_1 T_n = 1/2 (T_{n+1} + T_{n-1}). But according to the below section "Products of Chebyshev polynomials" the T_{n-1} should be T_{|n-1|}. So the formula is incorrect for n=0. 100.7.38.140 (talk) 19:41, 26 August 2023 (UTC)