5

I know how to solve heat equation where it's like $u_t=k\cdot u_{xx}$ (using Fourier Transform or using Separation of Variables) but this exercise is really difficult for me.

I have this:

$$u_t(x,t)=k \cdot u_{xx}(x,t)-a\cdot k \cdot u(x,t)$$ $$u_x(0,t)=0$$ $$u(x,0) = f(x)$$

with $x>0, t>0$ and $a, k$ are positive constants.

I have to find $u(x,t)$ and propose a possible $f(x)$

Any help? Thanks

I was told I cannot use Fourier Transform, I have to use Fourier Cosine Transform, and I don't know why

xzczd
  • 447
Unnamed
  • 439
  • 1
    You can bring the equation to the form you know how to solve with a simple change of variables. Define $v = u e^{ak t}$ then $v_t = k v_{xx}$ with $v_x(0,t) = 0$ and $v(x,0) = f(x)$. – Winther May 18 '17 at 20:15

3 Answers3

2

Indeed, this problem should be solved with Fourier cosine transform rather than Fourier transform. If you use Fourier transform, you won't be able to take the b.c. at $x=0$ into consideration, and the equation will then be solved in $-\infty<x<\infty$ with implicit b.c.s at $\pm\infty$. (For this part check this post for more information. )

The key point is, when $f(\infty)=0$, Fourier cosine transform

$$ \mathcal{F}_t^{(c)}[f(t)](\omega)=\sqrt{\frac{2}{\pi }}\int _0^{\infty } f(t) \cos (\omega t) d t $$

has the following property:

$$ \mathcal{F}_t^{(c)}\left[f''(t)\right](\omega)=-\omega^2 \mathcal{F}_t^{(c)}[f(t)](\omega)-\sqrt{\frac{2}{\pi }} f'(0) $$

You can easily verify this property using integration by parts.

So, by applying this property on your equation, we have

$$ \mathcal{F}_x^{(c)}\left[u^{(1,0)}(t,x)\right](\omega )=k \left(-\omega ^2 \mathcal{F}_x^{(c)}[u(t,x)](\omega )-\sqrt{\frac{2}{\pi }} u^{(0,1)}(t,0)\right)-a k \mathcal{F}_x^{(c)}[u(t,x)](\omega ) $$

Substitute the b.c. into the equation, we obtain a simple initial value problem (IVP) of linear ODE:

$$U'(t)=-a k U(t)-k \omega ^2 U(t)$$ $$U(0)=F$$

where $U(t)=\mathcal{F}_x^{(c)}[u(t,x)](\omega )$, $F=\mathcal{F}_x^{(c)}[f(x)](\omega)$.

If you have difficulty in solving the IVP, check the wikipedia page). Anyway, we can easily find its solution:

$$U(t)=F e^{-k t \left(a + \omega ^2\right)}$$

The last step is to transform back with inverse Fourier Cosine transform

$${\mathcal{F}_\omega^{(c)}}^{-1}[F(\omega)](t)=\sqrt{\frac{2}{\pi }} \int_0^{\infty } F(\omega ) \cos (t \omega) \, d\omega $$

and the solution is

$$ u(t,x)=\sqrt{\frac{2}{\pi }} \int_0^{\infty } e^{-k t \left(a + \omega ^2\right)} \mathcal{F}_x^{(c)}[f(x)](\omega ) \cos (\omega t) \, d\omega $$

Notice this solution is probably the same as the one given by doraemonpaul. I guess he has just chosen a different convention for Fourier parameters.

xzczd
  • 447
  • Can you explain what do the notations $u^{(1,0)},(t,x)$ and $u^{(0,1)},(t,0)$ mean? I am really confused with them. – Sam Wong May 04 '19 at 10:13
  • 1
    @sam $u^{(0,1)}(t,0)$ is equivalent to $\frac{\partial u(t,x)}{\partial x}\big|_{x=0}$. – xzczd May 04 '19 at 10:33
  • I see, thank you. But I still don't know why we can't use Fourier Transform and why we have to use Fourier Cosine Transform. Can you explain the reason? Thanks. – Sam Wong May 04 '19 at 10:46
  • 1
    @sam As already mentioned above, if we use Fourier transform, we won't be able to take the b.c. at $x=0$ into consideration, and the equation will then be solved in $-\infty<x<\infty$ with implicit b.c.s at $\pm\infty$. (Usually it's $f(\pm ∞)=0$. ) – xzczd May 04 '19 at 10:57
  • @sam BTW here is another example solving PDE with Fourier cosine transform: https://mathematica.stackexchange.com/a/158079/1871 – xzczd May 04 '19 at 11:13
  • Yea I saw that. But in order to use Fourier Cosine Transform, we will also need the implicit b.c. at $+\infty$ (i.e. $f(+\infty)=0$) right? In this problem, this necessary condition to use Fourier Cosine Transform would be $u(+\infty,t)=0$ for all $t\gt 0.$ But we don't have this extra condition. – Sam Wong May 04 '19 at 11:24
  • I mean, if you want to use integration by parts to calculate the Fourier Cosine Transform of $u_{xx}$ , then you will have to have the condition that $u(+\infty,t)=c$ for all $t\gt 0$ and some constant $c$. But here we don't have such condition. – Sam Wong May 04 '19 at 11:27
  • @sam Yes, the statement is not rigorous enough. Well, I would blame it on…Er… the looseness of the whole mathematical physics equation subject. Such kind of stuff is just everywhere. For example, if you check those materials about solving PDE using Fourier transform, you'll see the statement about b.c. at $\pm \infty$ is often missing, too. – xzczd May 04 '19 at 12:54
  • Yea, but I think @doraemonpaul 's post didn't use the b.c. at $\pm \infty$. I feel like it makes more sense. – Sam Wong May 04 '19 at 14:47
  • @SamWong Then why is separation of variables, which is rather counterintuitive, a valid method? – xzczd May 04 '19 at 15:09
1

Of course use separation of variables:

Let $u(x,t)=X(x)T(t)$ ,

Then $X(x)T'(t)=kX''(x)T(t)-akX(x)T(t)$

$X(x)(T'(t)+akT(t))=kX''(x)T(t)$

$\dfrac{T'(t)+akT(t)}{kT(t)}=\dfrac{X''(x)}{X(x)}=-s^2$

$\begin{cases}\dfrac{T'(t)}{T(t)}=-k(s^2+a)\\X''(x)+s^2X(x)=0\end{cases}$

$\begin{cases}T(t)=c_3(s)e^{-kt(s^2+a)}\\X(x)=\begin{cases}c_1(s)\sin xs+c_2(s)\cos xs&\text{when}~s\neq0\\c_1x+c_2&\text{when}~s=0\end{cases}\end{cases}$

$\therefore u(x,t)=\int_0^\infty C_1(s)e^{-kt(s^2+a)}\sin xs~ds+\int_0^\infty C_2(s)e^{-kt(s^2+a)}\cos xs~ds$

$u_x(x,t)=\int_0^\infty sC_1(s)e^{-kt(s^2+a)}\cos xs~ds-\int_0^\infty sC_2(s)e^{-kt(s^2+a)}\sin xs~ds$

$u_x(0,t)=0$ :

$\int_0^\infty sC_1(s)e^{-kt(s^2+a)}~ds=0$

$C_1(s)=0$

$\therefore u(x,t)=\int_0^\infty C_2(s)e^{-kt(s^2+a)}\cos xs~ds$

$u(x,0)=f(x)$ :

$\int_0^\infty C_2(s)\cos xs~ds=f(x)$

$\mathcal{F}_{c,s\to x}\{C_2(s)\}=f(x)$

$C_2(s)=\mathcal{F}^{-1}_{c,x\to s}\{f(x)\}$

$\therefore u(x,t)=\int_0^\infty\mathcal{F}^{-1}_{c,x\to s}\{f(x)\}e^{-kt(s^2+a)}\cos xs~ds$

doraemonpaul
  • 16,178
  • 3
  • 31
  • 75
  • 1
    Though the method is not Fourier transform, the answer is correct, I think. I suggest you to mention the convention of Fourier parameters in your answer. (It's different from the one chosen in my answer, right?) – xzczd Oct 20 '17 at 12:38
0

Let us apply the Fourier Transform with respect to the spatial coordinate $x$ :$$\partial_{t}\hat{u}\left(\xi,t\right)=-k\xi^{2}\hat{u}\left(\xi,t\right)-iak\xi\hat{u}\left(\xi,t\right)$$ i.e. $$\partial_{t}\hat{u}\left(\xi,t\right)=-k\left(\xi^{2}-ia\xi\right)\hat{u}\left(\xi,t\right)$$ hence$$\hat{u}\left(\xi,t\right)=Ce^{-k\left(\xi^{2}-ia\xi\right)t}$$ where $C$ is a constant to be determinated later. Here the convention used for the Fourier transform is$$\hat{u}\left(\xi,t\right)=\mathrm{TF}\left[u\right]\left(\xi,t\right)=\intop_{x\in\mathbb{R}}u\left(x,t\right)e^{-ix\xi}dx.$$ Now I let you come back to the initial space with the inversion formula$$u\left(x,t\right)=\mathrm{TF}^{-1}\left[\hat{u}\right]\left(x,t\right)=\frac{1}{2\pi}\intop_{x\in\mathbb{R}}Ce^{-k\left(\xi^{2}-ia\xi\right)t}e^{ix\xi}d\xi$$ (take care to the sign in the exponential). You must find a gaussian function.

Nicolas
  • 3,326