1

$$\underline{\text{**ATTENTION**}}$$

This question is not a duplicate of this one and this other question!

So let be $f:\Bbb R\rightarrow\Bbb R$ a periodic function, that is there exist $T\in\Bbb R$ such that $$ f(x+T)=f(x) $$ for any $x\in \Bbb R$. So I'd like to prove that $$ \int_x^{x+T}f(t)\,dt=\int_0^Tf(t)\,dt $$ for any $x\in\Bbb R$ and exactly I'd like to do this using the same arguments that André Nicolas gave in this answer but now I'd like do this with more details since I am sure that the mentioned user proves the result using the Fundamental Theorem of Calculus but unfortunately it is not clear how he applies it so that I thought to pay an apposite question where I ask to give this proof adding more details, that's all. Moreover it seems to me that if the period $T$ of $f$ is negative then the above integral could not be defined so that I ask also clarification about this.

So could anyone help me, please?

  • It is not generally true that $\int_T^{x+T} g(t),dt=\int_0^T g(s),ds$. – lulu Dec 30 '21 at 14:56
  • Sure, I know this... – Antonio Maria Di Mauro Dec 30 '21 at 14:57
  • 1
    And yet you claim that it is. All change of variables gets you is $\int_T^{x+T} g(t),dt=\int_0^T g(s+T),ds$ – lulu Dec 30 '21 at 14:58
  • No, I said that the result is very stranger! So I do not say what you claim. – Antonio Maria Di Mauro Dec 30 '21 at 14:59
  • Yes, you did. Right after " so that by the change variables theorem it follows that" – lulu Dec 30 '21 at 14:59
  • I wrote this before to say that the result is illogical so that I think it does not have sense: anyway if you like I can edit the question. – Antonio Maria Di Mauro Dec 30 '21 at 15:01
  • What you wrote is false, no matter what the reason you wrote it. I did not read beyond that error. – lulu Dec 30 '21 at 15:02
  • Okay, anyway I edited the question: I hope now it is more clear. Sorry, for the wrong. – Antonio Maria Di Mauro Dec 30 '21 at 15:04
  • 1
    Post edit, the false claim is still present. – lulu Dec 30 '21 at 15:04
  • Just take some non-periodic function, like $f(z)=z$, and go through your steps with that function to see that your argument fails. – lulu Dec 30 '21 at 15:05
  • The thing is that you actually do use the periodicity of $g$ in your proof, in the line that lulu points out. $\int_T^{x + T} g(t) dt = \int_0^x g(s+T) ds,$ and because $g(s + T) = g(s)$ for our periodic function we can rewrite as $\int_0^x g(s) ds,$ from which the rest of your argument can proceed. – Stephen Donovan Dec 30 '21 at 15:06
  • @lulu I repeate I know that the result is false: I put only it so that the people could see how I tried to prove the statement and so hoping that the proof can be adjusted and thus I cannot elimitate it, that's all. Sorry. – Antonio Maria Di Mauro Dec 30 '21 at 15:06
  • 1
    Once again: I have identified the line in your argument that is false for general functions. Of course, that line holds if you use periodicity. Not sure what else there is to say. – lulu Dec 30 '21 at 15:08
  • @lulu Okay, I understand what you want say: the problem is that I do not how apply the fundamental theorem of calculus to the function $\int_0^xg(s+T),ds$, that's all. So could you explain this rigorously? – Antonio Maria Di Mauro Dec 30 '21 at 15:09
  • You can still apply the fundamental theorem of calculus to that integral, but the result you would get is now $g(x + T),$ so you get $G'(x) = g(x+T) - g(x),$ which is not zero in general unless we have $g(x + T) = g(x).$ – Stephen Donovan Dec 30 '21 at 15:11
  • I don't understand. Your argument uses periodicity, of course. Periodicity tells you that $g(s+T)=g(s)$. Without periodicity, you can not advance your argument. So...what are you trying to do? – lulu Dec 30 '21 at 15:11
  • 2
    You ask where the wrong is, but it's been said: you claimed not to use the periodicity of $g$, but you actually did. – Randall Dec 30 '21 at 15:14
  • @StephenDonovan So if I understood I have to put $$\gamma(s):=g(s+T)$$ so that $$\frac d{dx}\int_0^xg(s+T),ds=\frac d{dx}\int_0^x\gamma(s),ds=\gamma(x)=g(x+T)$$ right? – Antonio Maria Di Mauro Dec 30 '21 at 15:15
  • I believe that would be the correct way to write it out more fully, yes. – Stephen Donovan Dec 30 '21 at 15:17
  • @Randall So first I pretended ** intentionally ** a uncorrect proof: I know that it is uncorrect because I claim explicitly that the result is false, okay? So I asked where is the wrong in the proof. Then I presented a good proof and I asked a clarification about it. Where is the problem? What did I do wrong? How I should have express my difficulties? – Antonio Maria Di Mauro Dec 30 '21 at 15:21
  • @StephenDonovan Oaky, so where's the problem in the first proof? that is, why without the periodicity of $g$ I can not apply the fundamental theorem of calculus? I'd like to know it because this can be usefull for the future: learning from your mistakes is good. – Antonio Maria Di Mauro Dec 30 '21 at 15:23
  • So again, you can use the fundamental theorem of calculus, but it doesn't yield a useful result. When you continue your proof with that result, you would arrive at $G'(x) = g(x + T) - g(x),$ which is not zero in general, but only if we assume $g(x + T) = g(x).$ So if we don't assume periodicity then we don't get that $G' = 0$ and we don't get $G(x) = G(0),$ which is the crux of the entire argument. – Stephen Donovan Dec 30 '21 at 15:27
  • @StephenDonovan Sorry, but I do not understand very well. So using the change of variable formula I proved that $$\int_T^{x+T}g(t),dt=\int_0^Tg(s),ds$$ (IS THIS ALWAYS TRUE?) then it would be $$G(x)=\int_0^xg(s),ds-\int_T^xg(t),dt$$ and so applying the fundamentel theorem (IS THIS POSSIBLE?) then we discover that $G$ is costant but this is false: so where I get a mistake? I get a mistake using the change variables or the fundamental theorem? Forgive my confusion. – Antonio Maria Di Mauro Dec 30 '21 at 15:36
  • Here your mistake is in the change of variables, as discussed before. In the first line the correct integrand for the right-hand side is $g(s + T),$ which we can only change to $g(s)$ if we assume periodicity. Continuing after this correction would lead to the problem I mentioned in that last comment. – Stephen Donovan Dec 30 '21 at 15:40
  • @StephenDonovan Oh yeah!!! indeed if $s=t-T$ then necessary $t=s+T$ so that the change variable formula implies that $$\int_{[T,x+T]}g(t),dt=\int_{\varphi^{-1}\big[[T,x+T]\big]}g(t(s)),ds=\int_{[0,T]}g(s+T)$$ right? – Antonio Maria Di Mauro Dec 30 '21 at 15:43
  • That looks correct to me, yes. – Stephen Donovan Dec 30 '21 at 15:48
  • @StephenDonovan Okay, now it is all clear: so thanks very much for your assistance!!! You was really kind with me, so thanks yet. – Antonio Maria Di Mauro Dec 30 '21 at 15:52

2 Answers2

1

First of all we observe that the period $T$ can be suppose not negative without loss of generality: indeed if $T$ was not positive then $-T$ would be not negative and it would be such that $$ f(x-T)=f\big((x-T)+T\big)=f(x) $$ for any $x\in\Bbb R$. So without loss of generality we can suppose that $$ x\le x+T $$ for any $x\in\Bbb R$ so that the function $F:\Bbb R\rightarrow\Bbb R$ defined as $$ F(x):=\int_x^{x+T}f(t)\,dt $$ for any $x\in\Bbb R$ is well defined.

So we let to prove now that $F$ is constant and precisely we are doing this proving that it is constant in $(-\infty,T]$ and in $[T,+\infty)$ respectively.

Previously we remember that the Change of Variables Theorem states that

Given a diffeomorphism $\phi:A\rightarrow B$ between open sets of $\Bbb R^n$ for any $C^r$ function $f:A\rightarrow\Bbb R$ with $r\ge 1$ the identity $$\int_Af=\int_B(f\circ \phi)|\det Dg|$$ holds.

whereas the Fundamental Theorem of Calculus states that

If $I$ is a (closed) interval of the real line $\Bbb R$ with extremities $x_1$ and $x_2$ then given a real valued function $f:I\rightarrow\Bbb R$ the function $F:I\rightarrow\Bbb R$ defined as $$ F(x):=\int_{[x_1,x]}f(\xi)\,d\xi $$ for any $x\in I$ is derivable in $I$ and it is such that $$ \operatorname{D}F(x)=f(x) $$ for any $x\in I$.

  1. So let be now $x\in(T+\infty]$ and thus through the additivity of the integral we observe that $$ F(x):=\int_{[x,x+T]}f(t)\,dt=\int_{[T,x+T]\setminus[T,x]}f(t)\,dt=\int_{[T,x+T]}f(t)\,dt-\int_{[T,x]}f(t)\,dt $$ for any $x\in[T,+\infty)$. Now let be $\varphi:\Bbb R\rightarrow\Bbb R$ the diffeomorphism defined through the equation $$ \varphi(x):=x+T $$ for any $x\in\Bbb R$. So through the Chain Rule we conclude that $$ \operatorname{D}F(x)=D\Biggl[\int_{\big[T,\varphi(x)\big]}f(t)\,dt-\int_{[T,x]}f(t)\,dt\Biggl]=\\ \operatorname{D}\Biggl[\int_{\big[T,\varphi(x)\big]}f(t)\,dt\Biggl]\cdot\operatorname{D}\varphi(x)-\operatorname{D}\Biggl[\int_{[T,x]}f(t)\,dt\Biggl]=\\ f\big(\varphi(x)\big)-f(x)=f(x+T)-f(x)=0 $$ for any $x\in[T,+\infty)$ and so the statement is proved for this such $x$.
  2. So let be now $x\in(-\infty,T]$ and provided that $x$ is not negative through the additivity of the integral we observe that $$ F(x):=\int_{[x,x+T]}f(t)\,dt=\int_{[x,T]\cup[T,x+t]}f(t)\,dt=\int_{[x,T]}f(t)\,dt+\int_{[T,x+T]}f(t)\,dt $$ Now let be $\phi:\Bbb R\rightarrow\Bbb R$ the diffeomorphism defined through the equation $$ \phi(x):=-x $$ for any $x\in\Bbb R$ so that through the change variables theorem we conclude that $$ F(x)=\int_{[x,T]}f(t)\,dt+\int_{[T,x+T]}f(t)\,dt=\int_{(x,T)}f(t)\,dt+\int_{(T,x+T)}f(t)\,dt=\\ \int_{\phi^{-1}\big[(x,T)\big]}\big(f\circ\phi\big)(t)\cdot\Big|\operatorname{det}\big(\operatorname{D}\phi(t)\big)\Big|\,dt+\int_{\big(x,\varphi(x)\big)}f(t)dt=\\ \int_{(-T,-x)}\big(f\circ\phi\big)(t)\,dt+\int_{\big(T,\varphi(x)\big)}f(t)\,dt=\int_{\big(-T,\phi(x)\big)}\big(f\circ\phi\big)(t)\,dt+\int_{\big(T,\varphi(x)\big)}f(t)\,dt $$ for any $x\in[0,T]$ and thus finally through the Fundamental Theorem of Calculus and through the Chain Rule we conclude that $$ \operatorname{D}F(x)= \operatorname{D}\Biggl[\int_{\big(-T,\phi(x)\big)}\big(f\circ\phi\big)(t)\,dt+\int_{\big(T,\varphi(x)\big)}f(t)\,dt\Biggl]=\\ D\Biggl[\int_{\big(-T,\phi(x)\big)}\big(f\circ\phi\big)(t)\,dt\Biggl]\cdot\operatorname{D}\phi(x)+\operatorname{D}\Biggl[\int_{\big((T,\varphi(x)\big)}f(t)\,dt\Biggl]\cdot\operatorname{D}\varphi(x)=\\ -\Big(f\circ\phi\Big)\big(\phi(x)\big)+f\big(\varphi(x)\big)=f\big(\varphi(x)\big)-f\Big(\phi\big(\phi(x)\big)\Big)=f(x+T)-f(x)=0 $$ for any $x\in[0,T]$ so that we proved that $F$ is effectively constant on the not negative real line. Whereas if $x$ is not positive through the additivity of the integral we observe that $$ F(x):=\int_{[x,x+T]}f(t)\,dt=\int_{[x,T]\setminus[x+T,x]}f(t)\,dt=\int_{[x+T]}f(t)\,dt-\int_{[x+T,x]}f(t)\,dt $$ for any $x\in(-\infty,0]$ so that through the change variables theorem we conclude that $$ F(x)=\int_{[x,T]}f(t)\,dt-\int_{[x+T,T]}f(t)\,dt=\\ \int_{\phi^{-1}\big[[x,T]\big]}(f\circ\phi)(t)\,dt-\int_{\phi^{-1}\big[[x+T,T]\big]}(f\circ\phi)(t)\,dt=\\ \int_{[-T,-x]}(f\circ\phi)(t)\,dt-\int_{[-T,-(x+T)]}(f\circ\phi)(t)\,dt $$ for any $x\in(-\infty,0]$ and thus finally through the Fundamental Theorem of Calculus and through the Chain Rule we conclude that $$ DF(x)=D\Biggl[\int_{[-T,-x]}(f\circ\phi)(t)\,dt-\int_{[-T,-(x+T)]}(f\circ\phi)(t)\,dt\Biggl]=\\ D\Biggl[\int_{[-T,\phi(x)]}(f\circ\phi)(t)\,dt-\int_{[-T,(\phi\circ\varphi)(x)]}(f\circ\phi)(t)\,dt\Biggl]=\\ D\Biggl[\int_{[-T,\phi(x)]}(f\circ\phi)(t)\,dt\Biggl]\cdot D\phi(x)-D\Biggl[\int_{[-T,(\phi\circ\varphi)(x)]}(f\circ\phi)(t)\,dt\Biggl]\cdot D(\phi\circ\varphi)(x)=\\ -(f\circ\phi)(\phi(x))-\Big(-(f\circ\phi)\Big(\big(\phi\circ\varphi\big)(x)\Big)\Big)=\\ f\big(\varphi(x)\big)-f(x)=f(x+T)-f(x)=0 $$ for any $x\in(-\infty,0]$ and this proves finally that $F$ is constant in $(-\infty,0]$ too.
0

You don't need advance technique. In fact, \begin{eqnarray} \int_{x}^{x+T}g(t)\,dt&=&\int_0^{x+T} g(t)\,dt-\int_0^{x} g(t)\,dt\\ &=&\int_{-T}^{x} g(u+T)\,dt-\int_0^{x} g(t)\,dt\\ &=&\int_{-T}^0g(t)\,dt=\int_{0}^Tg(T+t)\,dt\\ &=&\int_{0}^Tg(t)\,dt. \end{eqnarray}

xpaul
  • 47,821
  • Unfortunately there is a little problem: in the proof I implicitly assumed that $x\ge T$ since otherwise it was not possible to break the integral and analogous thing you do assuming that $x\ge0$. Anyway without loss of generality we can assume that $T\ge0$ so that my proof show that the result is true for $x\le 0$ and it seems that you do the *same. So what can you say about? Indeed I did not observe initially this. – Antonio Maria Di Mauro Dec 30 '21 at 18:33
  • Pheraps I found a rigorous answer as you can see below: so what do you think about it? is it correct? – Antonio Maria Di Mauro Dec 30 '21 at 23:31
  • @AntonioMariaDiMauro $x$ can be any value. – xpaul Jan 01 '22 at 16:46
  • Sorry, but inf you write $$\int_x^{x+T}g(t),dt=\int_0^{x+T}g(t),dt-\int_0^xg(t),dt$$ indeed necessarly it must by $x>0$ because otherwise it would be $$\int_x^{x+T}g(t),dt=\int_x^0g(t),dt+\int_0^{x+T}g(t),dt$$ and so by the change variable formula putting $$\varphi(x):=-x$$ you could write $$\int_x^{x+T}g(t),=\int_x^0g(t),dt+\int_0^{x+T}g(t),dt=\int_{\varphi^{-1}\big[(x,0)\big]}g\big(\varphi(t)\big),dt+\int_0^{x+T}g(t),dt$$ right? So where's my wrong? Excuse me for the confusion. – Antonio Maria Di Mauro Jan 02 '22 at 08:38
  • So substantially the problem is that for me the identity $$\int_a^b f(x)$$ has no meaning when $b<a$ because by definition $$\int_a^bf(x):=\int_{(a,b)}f(x)$$ that's all. – Antonio Maria Di Mauro Jan 02 '22 at 08:38
  • If $a>b$, the integral $\int_a^bf(x)dx$ is interpreted as $-\int_b^af(x)dx$. – xpaul Jan 04 '22 at 16:39
  • Okay, but it is only a convention, right? In other words, are there some FORMAL AND RIGOROUS results that can attribute a meaning to $\int_b^a(x) dx$? – Antonio Maria Di Mauro Jan 04 '22 at 16:44
  • Pheraps this can be done by inverting the usual order of $\Bbb R$? – Antonio Maria Di Mauro Jan 04 '22 at 16:44
  • Your answer is nothing wrong. It is just too long. – xpaul Jan 04 '22 at 17:06
  • I think it is long exactly because I did not define $\int_b^a f(x), dx$, that's all. – Antonio Maria Di Mauro Jan 04 '22 at 17:27