三角不等式

Let n be a natural number and let 0<x<π. Then, here are my questions.

Question 1: Is the following true?
k=1ncos(kx)k>1

Question 2: Is the following true?
k=1nsin(kx)k>0


 

This is a possible hint for solution; perhaps someone can finish it along these lines (it won't fit as a comment). We have
sinx+sin2x2+sin3x3++sinnxn=k=1n0xcosktdt,
2k=1ncoskt=sin((n+1/2)t)/sin(t/2)1 (by taking the real part of k=1neikt)
so we want to show
0x(sin(n+1/2)tsin(t/2)1)dt>0.
It's easy without that 1 (as 1/sin(t/2) decreases); to do it really (with 1) a better estimate is needed.


 

TEOREMA (L.Fejer-1910, D. Jackson-1912, T.J.Gronwall- 1912). Pentru x(0,π) au loc inegalitatile
(1)k=1nsinkxk>0,nN.

 

DEMONSTRATIA I .
Prin efectuarea unor calcule elementare se constata
ddφ{sin2kϕ(sinϕ)2k}=2ksin(2k1)ϕ(sinϕ)2k+1.
Efectuand calculele precum si substitutia ϕx2, prin integrare avem
sinkxk=2(sinx2)2kx2π2sin(2k1)ϕ(sinϕ)2k+1dϕ,kN.
Insumand pentru k{1,2,...,n} se obtine
k=1nsinkxk=2x2π2k=1n[r(x,ϕ)]ksin(2k1)ϕsinϕdϕ
unde r(x,ϕ):=(sinx2sinϕ)2[0,1]
pentru 0x2ϕ<π2. Egalitatea de mai sus implica
k=1nsinkxk=xπk=1n[r(x,ψ2)]ksin(k12)ψsinψ2dψ.
Aplicand identitatea lui Abel ("'insumarea prin parti"), adica
k=1nAkBk=k=1n1(AkAk+1)j=1kBj+Anj=1nBj,
si notand
{r:=r(x,ψ2),r[0,1]vezi()Fn(ψ):=k=1nsin(k12)ψ,Fn(ψ)0dac'ax[0,π]vezi L. Fej\'er, din identitatea lui Abel g'asim
k=1nsinkxk=xπ{(1r)k=1n1rk1Fk(ψ)+rnFn(ψ)}dψsinψ2.
Deoarece Fn(ψ)=1cos(nψ)2sinψ2, concludem cu
k=1nsinkxk=12xπ{(1r)k=1n1rk1(1cos(kψ))+rn(1cos(nψ))}dψ(sinψ2)2.
Aceasta reprezentare completeaza demonstra'tia I.

DEMONSTRATIA II.
Fie Pn(x) polinomul lui Legendre de gradul n, adic'a
Pn(x)=1n!2n[(x21)n)(n)=2F1(n,n+1;1;1x2)==12nk=0n2(n2k)(1)k(n+1)n2k(2k)!k!(nk)!xn2k
unde
2F1(n;b;c;z):=k=0n(1)k(nk)(b)k(c)kzk(d)k:=d(d+1)(d+k1),kN,(d)0:=1.
Se cunosc urmatoarele:
-- radacinile lui Pn(x) sunt reale,distincte, situate in (1,1);
-- |Pn(t)|1,t[1,1].
Demonstratia a II-a (vezi [16] precum si comentariile lui R.Askey )
se bazeaza pe identitatea:
k=1nsin(karccosx)k=1x21x1Pn(y)1ydyxy,x(1,1)..

Observatii.

1) Inegalitatea (1) a fost conjecturata de catre Leopold (Lipot) Fejer. Ulterior a fost demonstrata de catre D.Jackson-[14] si T.H.Gronwall-[12].
In prezent se cunosc peste 100 de demonstratii. Am ales pentru a Va prezenta pe cele pe care subsemnatul le considera "mai simple".

Se spune ca pana la moasrte L.Fejer a cautat sa gaseasca noi demonstratii a lui (1). L.Fejer a predat si la Universitatea din Cluj (sub numele de L.Weiss).
2)Desi simpla la prima vedere, inegalitatea (1) a dat de furca matematicienilor.
3) Inegalitatea (1) intervine in urmatoarele domenii: Serii Fourier (fenomenul lui Gibbs - vezi [13]), Polinoame ortogonale ([1]-[5],[18] ), Functii Complexe (Demonstratia conjecturii lui Bieberbach,functii univalente, [4],[6]), Teoria Aproximarii ([5]).

 

BIBLIOGRAFIE.
[1]R. Askey , Orthogonal Polynomials and Special Functions,
Regional Conf.Lect.Appl.Math., vol.21,SIAM,Philadelphia,Pa., 1975.
[2]R.Askey Positive quadrature methods and positive polynomial sums, 'in Approximation Theory V, Academic Press, 1986.
[3] R. Askey and J. Fitch , Integral reprezentations for Jacobi
polynomial amd some applications ,J.Math.Anal.Appl., 26 (1969)
411-437.
[4]R. Askey and G. Gasper , Positive Jacobi polynomial sums,(II),
Amer.J.Math., 98 (1976) 709-737.
[4]R. Askey and G. Gasper , Inequalities for polynomials}, 'in "The Bieberbach Conjecture", Proc.of the Symposium on the Occasion
of the Proof, Mathematical Surveys and Monographs, 21, Amer.Mathematical Society, 1986, 7-32.
[5]H. Bavinck , Jacobi Series and Approximation,
Mathematical Centre Tracts 39, Mathematisch Centrum Amsterdam 1972.
[6] L. de Branges , The Story of the Verification of the Bieberbach Conjecture, 'in " The Bieberbach Conjecture", Proc.of the Symposium on the Occasion of the Proof, Mathematical Surveys and Monographs, 21, Amer.Mathematical Society, 1986, 199-203.
[7] L. Fejer , Sur les fonctions bornees et integrables,
C.R.Acad.Sci.Paris 131 (1900) 984-987.
[8] L. Fejer , Sur le develpopment d'une fonction arbitraire suivant les fonctions de Laplace, C.R.Acad.Sci.Paris , 146 (1908) 224-227.
[9]L. Fejer , Ueber die Laplacesche Reihe, Math. Ann. (1909)76-109.
[10] L. Fejer , Einige Saetze , die sich auf das Vorzeichen einer ganzen rationalen Funktion beziehen u.s.w., Monta.fuer Math. und Phys., 35 (1928) 305-344.
[11]L. Fejer , Gesammelte Arbeiten (I)-(II), Birkhauser Verlag, Basel, 1970 .
[12]T.H. Gronwall , Ueber die Gibbsche Erscheinung und die trigonometrischen Summen sinx+(1/2)sin2x+...+(1/n)sinnx , Math.Ann., 72 (1912) 228-243.
[13] E. Hewitt and R.E. Hewitt , The Gibbs-Wilbraham phenomenon: an epsiode in Fourier analysis, Arch.Hist.Exact Sci., 21 (1979) 129-160.
[14]D. Jackson , Ueber eine trigonometrische Summe, Rend.Circ.Mat.Palermo 32(1911) 257-262.
[15]A. Lupas , Advanced Problem 6517, Amer.Math.Monthly (1986) p. 305 ; (1988) p.264.
[16]A. Lupas , Advanced Problem 6585, Amer.Math.Monthly
(1988) p. 880 ; (1990) p.859-860.
[17]L. Lupas , An identity for ultraspherical polynomials, Revue d,Analyse numerique et de Theorie de l'approximation, tome 24 , no.1-2 (1995) 181-185.
[18]G. Szego , Orthogonal Polynomials , Amer.Math.Soc.Colloq. Publications vol.23, fourth ed.,Amer.Math.Soc., Providence, R.I., 1975.

 


Călcat-up inegalitate k=1nsinkxk>x(1xπ)3 deţine pentru x(0,π).

参考这里Sawtooth wave


**In short:** let fn(x) denote the function on the lhs of the inequality. Of course, f1(x)=sinx0 on [0,π]. We will prove that fn(x)0 on [0,π] by induction on n. It is not too hard to determine the local minima of fn on [0,π] by investigating its derivative. Then Ma Ming observed that fn coincides with fn1 on these local minima. And the induction step follows easily. Of course, fn(0)=fn(π)=0. We will actually prove that
>fn(x)=k=1nsinkxk>0x(0,π).

**Remark:** it is worth noting that the fn's are the partial sums of the Fourier series of the same sawtooth function. Just [look at the case n=6][1], for instance, to see how they tend to approximate it nicely. [See here][2] to get an idea how to estimate the error in such approximations. As pointed out by math110, there are many proofs of this so-called Fejer-Jackson inequality. It can even be shown that the [fn's are bounded below][3] by a certain nonnegative polynomial on [0,π]. The proof below is at the calculus I level. I'm not sure it can be made more elementary.

**Proof:** first, f1(x)=sinx is positive on (0,π). Assume this holds for fn1 for some n2. Then observe that fn is differenbtiable on R with
fn(x)=k=1ncoskx=Rek=1n(eix)k.
For x2πZ, we have fn(x)=n. So the zeros of fn are the zeros of
Reeixeinx1eix1=Reei(n+1)x/2sin(nx/2)sin(x/2)=cos((n+1)x/2)sin(nx/2)sin(x/2).
This yields
nx2πZor(n+1)x2π2+πZ
i.e.
x2πnZorxπn+1+2πn+1Z.
Between 0 and π, these are ordered as follows:
0<πn+1<2πn<3πn+1<4πn<<2n/2πnπ.
The sign of fn changes at each of these zeros, starting from a positive sign on (0,π/(n+1)). It follows that fn is positive on the latter, positive on the last interval (if nontrivial, i.e. in the odd case), with local minima at
2jπnforj=1,,n/2.

But now here is Ma Ming's key observation: for these values, we have
fn(2jπn)=fn1(2jπn)+sin(n2jπn)=fn1(2jπn)>0
by induction step. It follows that fn(x)>0 on (0,π). QED.


[1]: http://www.wolframalpha.com/input/?i=sin%20%28x%29%2bsin%20%282x%29/2%2bsin%20%283x%29/3%2bsin%20%284x%29/4%2bsin%285x%29/5%2bsin%286x%29/6
[2]: https://math.stackexchange.com/questions/57054/asymptotic-error-of-fourier-series-partial-sum-of-sawtooth-function
[3]: https://math.stackexchange.com/questions/177995/a-pseudo-fejer-jackson-inequality-problem

 

参考:论坛1论坛1.1这里

论坛2

论坛3


 

How to prove that xR, nN, we have

(1)k=1n|sinkx|k|sinnx|?

Given n1 and xR, denote
fn(x)=k=1n|sinkx|kandgn(x)=fn(x)|sinnx|.


----------


> **Lemma:** For every n1,
>
> (i) fn is increasing on [0,πn+1];
> (ii) gn is increasing on [0,πn];
> (iii) fn1 on [π2n,π2].

**Proof of Lemma:** Note that when x[0,πn],
fn(x)=k=1nsinkxk,andgn(x)=fn(x)sinnx,
so
fn(x)=k=1ncoskxandgn(x)=fn(x)ncosnx.
(i) Given x[0,πn+1], noting that coskx20 for k=0,±1,,±(n+1), we have
fn(x)=k=1ncoskx+cos(n+1k)x2=cos(n+1)x2k=1ncos(n+12k)x20.


(ii) Since the cosine function is decreasing on [0,π], when x[0,πn], coskxcosnx for k=1,,n, so g(x)0.

(iii) When n=1, the statement is clearly true; when n=2, since f2 is concave on [π4,π2], f2(π4)>1 and f2(π2)=1, the statement is also true. By induction, we may assume that fn11 on [π2(n1),π2] for some n3, and the conclusion fn1 on [π2n,π2] follows from the facts below. Firstly, fnfn1; secondly, fn is increasing on [0,πn+1][π2n,π2(n1)]; thirdly,
sinπt2t, t[0,1]fn(π2n)=k=1nsinkπ2nk1.


----------
Now we can prove that gn0 by using the lemma. Since gn(π±x)=gn(x), we may focus on x[0,π2]. Since gn(0)=0, by (ii), we know that gn(x)0 on [0,πn]. Since gnfn1, by (iii) we know that gn0 on [π2n,π2].

参考:这里


 

In Iwaniec's book, *Topics in Classical Automorphic Forms*, pg. 4, he gives the statement:

 

{x}=12n=1Nsin2πnxπn+O((1+||x||N)1)
where {x} denotes the fractional part of xR, and ||x|| denotes the distance from x to the nearest integer, and the implied constant is "absolute" (which I take to mean that it is independent of x and N).


It suffices to prove this for 0x1/2. Let gN(x)=n=1Nsin2πnxπn+x12
To show gN(x)=O(11+xN), it suffices to show gN(x) is O(1xN) and also O(1). But it's hard to show O(1) directly so I'll actually show it is 1/2+O(log(1+xN)), which together with O(1xN) implies O(1).

Let DN(t)=1+2n=1Ncos2πnt=sin2π(N+1/2)tsinπt be the Dirichlet kernel. Observe that gN(x)=DN(x). Thus: gN(x)=0xDN(t)dt12

The Dirichlet kernel satisfies the inequalities |DN(t)|2N+1 and |DN(t)|12|t| for 0<|t|<1/2. We combine these to get |DN(t)|72N1+N|t|. Then we have: 0x|Dn(t)|dt720xN1+Ntdt=72log(1+xN)
Thus we get gN(x)=1/2+O(log(1+xN)).

It remains to show that gN(x)=O(1xN). To get that kind of bound, we need to use the trick of integrating by parts, but we can't do that on 0xDN(t)dt because the boundary terms will blow up. So we use the fact that DN(1t)=DN(t) and 01DN(t)dt=1 to rewrite gN(x) as: gN(x)=12x1xDN(t)dt for 0x1/2. Now we can integrate by parts to get:x1xDn(t)dt=1π(N+12)cos2π(N+12)xsinπx12(N+12)x1xcos2π(N+12)tdtsin2πt

Using the fact that sinπt2t for 0t1/2, the first term above is bounded by 12π1(N+1/2)x. A similar calculation for the second term gives a bound of 181(N+1/2)x. Putting this together, we get: |gN(x)|(14π+116)1(N+12)x

Thus, gN(x)=O(1xN) and so also O(1), thus gN(x)=O(11+xN), and we are done.

ADDED LATER:
In response to the question about the bound |DN(t)|72N1+N|t|, we consider two cases:

Case 1: When 12|t|2N+1, we have 1N|t|4+2/N6, since N is a positive integer. Therefore,
|DN(t)|12|t|12|t|71+1N|t|=12|t|7N|t|1+N|t|=72N1+N|t|.

Case 2: When 12|t|2N+1, since N1 we have
|DN(t)|2N+17N5N+2(2N+1)=72N1+N/(4N+2)72N1+N|t|.

 

 参考:这里

 

posted on   Eufisky  阅读(828)  评论(0编辑  收藏  举报

编辑推荐:
· 开发者必知的日志记录最佳实践
· SQL Server 2025 AI相关能力初探
· Linux系列:如何用 C#调用 C方法造成内存泄露
· AI与.NET技术实操系列(二):开始使用ML.NET
· 记一次.NET内存居高不下排查解决与启示
阅读排行:
· Manus重磅发布:全球首款通用AI代理技术深度解析与实战指南
· 被坑几百块钱后,我竟然真的恢复了删除的微信聊天记录!
· 没有Manus邀请码?试试免邀请码的MGX或者开源的OpenManus吧
· 园子的第一款AI主题卫衣上架——"HELLO! HOW CAN I ASSIST YOU TODAY
· 【自荐】一款简洁、开源的在线白板工具 Drawnix
< 2025年3月 >
23 24 25 26 27 28 1
2 3 4 5 6 7 8
9 10 11 12 13 14 15
16 17 18 19 20 21 22
23 24 25 26 27 28 29
30 31 1 2 3 4 5

导航

统计

点击右上角即可分享
微信分享提示