11

I'm given that

$$\left(x-c_1\frac{d}{dx}\right)^nf(x) = 0$$

I have to solve for $f(x)$ in terms of $n$.

  • For $n=0$: $$f(x)=0 \tag{0}$$

  • For $n= 1$:

$$\begin{align} xf(x) - c_1f'(x) &= 0 \\ \quad\implies\quad f(x) &= c_2\exp\left(\frac{x^2}{2c_1}\right) \tag{1} \end{align}$$

  • For $n=2$: $$\begin{align} \left(x-c_1\frac{d}{dx}\right)^1(xf(x) - c_1f'(x)) &= 0 \\[4pt] \quad\implies\quad x^2f(x) -xc_1f'(x) -c_1(f(x)+xf'(x)) +c_1^2f''(x) &=0 \\[4pt] \quad\implies\quad f(x) = k_1\exp\left(\frac{-x^2}{2c_1}\right) + k_2x\exp\left(\frac{-x^2}{2c_1}\right) & \tag{2} \end{align}$$

  • The case for $n= 3$ gets so complicated that I haven't put up the solution.

The solution is based on hermitian polynomials.

Blue
  • 83,939
  • 3
    Hmm, I mean there maybe a general answer in terms of $n$ but it's likely to be extremely , extremely long and complicated. Where did you come across this? – Riemann'sPointyNose Jul 18 '20 at 22:03
  • 3
    I told my professor that a problem could be solved like this instead of another way.So he suggested that I should try it myself :( – Tim Crosby Jul 19 '20 at 11:17
  • 3
    what method did your teacher say? – Riemann'sPointyNose Jul 19 '20 at 13:00
  • 1
    It is given that f(x) is square integrable and a solution to the schrodinger equation for the harmonic oscillator . It is also given that if we act the operator (x-d/dx) multiple times on any function that satisfies the equation, there will come a time after which further action on it will render 0. – Tim Crosby Jul 19 '20 at 18:32
  • 1
    So we assumed that $(x-\frac{d}{dx})f(x_0)= 0$ for and generated the solution for n=1,n=2, n=3, ..... – Tim Crosby Jul 19 '20 at 18:35
  • 1
    @Tim Crosby When you say it is a "solution", I suppose you mean it is the spacial part of the solution? – K.defaoite Jul 20 '20 at 23:22
  • 1
    I think your professor said you should try it so that you would find out that it is not solvable. – GEdgar Jul 21 '20 at 01:02
  • 1
    $-c_1^2 f''(x)$ should be $+c_1^2 f''(x)$. Then the equation has a simple solution. – Maxim Jul 21 '20 at 12:06
  • 1
    Ty @Maxim, fixed it – Tim Crosby Jul 21 '20 at 13:16
  • 1
    @Maxim, But still there is no solution for this equation in terms of n and n=3 will still be very complicated – Tim Crosby Jul 21 '20 at 13:24

3 Answers3

12

One can solve it recursively. For example, let $f_n(x)$ be such that $$\left(x-c_1\frac{d}{dx}\right)^nf_n(x)=0.$$ Then one needs to find $f_{n+1}(x)$ such that $$\left(x-c_1\frac{d}{dx}\right)^{n+1}f_{n+1}(x)=0$$ $$\Leftrightarrow \left(x-c_1\frac{d}{dx}\right)^n\left[\left(x-c_1\frac{d}{dx}\right)f_{n+1}(x)\right]=0$$ $$\Leftrightarrow \left(x-c_1\frac{d}{dx}\right)f_{n+1}(x)=f_n(x),$$ where the latter can be solved by the standard method (e.g. using integrating factor) to yield $$f_{n+1}(x)=e^{\frac {x^2}{2c_1}}\int\frac{f_n(x)}{-c_1}e^{-\frac{x^2}{2c_1}}~dx.\quad (1)$$

As Maxim observed in the comment, the answer turns out to be simple. No Hermitian polynomials are needed. Let $D=x-c_1\frac{d}{dx}$. There case $n=0$ being trivial, one needs to check that the solutions to $$D^nf(x)=0,n\geq 1$$ are given by $$f_n(x)=p(x)e^{\frac {x^2}{2c_1}},$$ where $p(x)$ is a polynomial with $\deg p\leq n-1.$

This can be proved by induction. You already obtained the case $n=1$. Assume the result is true for some $n\geq 1$, so the solutions to $D^nf(x)=0$ is of the form $$f_n(x)=p(x)e^{\frac{x^2}{2c_1}},\deg p\leq n-1.$$ Now by (1), one solves the equation $D^{n+1}f=0$ and obtains (up to constant multiple) $$f_{n+1}(x)=e^{\frac{x^2}{2c_1}}\int p(x)~dx=q(x)e^{\frac{x^2}{2c_1}},$$ for some polynomial $q(x)$ with $\deg q\leq n.$ QED

Pythagoras
  • 7,149
  • 1
    Since $f_0(x) = 0$, it follows that $f_n(x)$ is $e^{x^2/(2 c_1)}$ times a polynomial of degree (at most) $n - 1$. – Maxim Jul 21 '20 at 12:21
  • 1
    @Maxim Good observation! – Pythagoras Jul 21 '20 at 22:03
  • 1
    I dont know, But I'm getting $f_n(x) = p(x)e^{-x^2/2c_1}$ where $p(x)$ is a polynomial is of $\deg p=n$, – Tim Crosby Jul 22 '20 at 15:29
  • 1
    $f_{ 1} = e^{-x^2/2c_1} \int 0 = k_1e^{-x^2/2c_1}$ , $f_{2} = -\frac{k_1}{c_1}e^{-x^2/2c_1} $ – Tim Crosby Jul 22 '20 at 15:34
  • 1
    @TimCrosby You probably missed some signs. You got the case $n=1$ in your original post, but the signs in the exponent of the exponential functions should be the other way around in your comment above. You can check it for simplicity by taking $c_1=1$. – Pythagoras Jul 22 '20 at 17:39
  • 1
    :( Sorry ... careless mistake , $f_1=e^{x^2/2c_1}∫0=k_1e^{x^2/2c_1}$ , $f_2=−\frac{k_1}{c_1}e^{x^2/2c_1}$ – Tim Crosby Jul 22 '20 at 18:00
3

Alternatively, let $D$ denote the differential operator $$Df:=f'$$ for all differentiable function $f:\mathbb{R}\to\mathbb{C}$. Define the operator $M$ as $$(Mf)(x):=\exp\left(+\dfrac{x^2}{2c_1}\right)\,f(x)$$ for all $f:\mathbb{R}\to\mathbb{C}$ and $x\in\mathbb{R}$. Observe that $M$ is an invertible operator with the inverse $M^{-1}$ given by $$(M^{-1}f)(x)=\exp\left(-\dfrac{x^2}{2c_1}\right)\,f(x)$$ for all $f:\mathbb{R}\to\mathbb{C}$ and $x\in\mathbb{R}$. Now, we conjugate the differential operator $D$ by $M$ to obtain the operator $\Delta:=MDM^{-1}$ which satisfies $$(\Delta f)(x)=\left(\frac{\text{d}}{\text{d}x}-\frac{x}{c_1}\right)\,f(x)$$ for all differentiable function $f:\mathbb{R}\to\mathbb{C}$ and $c\in\mathbb{R}$. Therefore, the question asks for all $n$-time differentiable functions $f:\mathbb{R}\to\mathbb{C}$ in the kernel of $\Delta^n$, namely, $$\left(\Delta^n f\right)(x)=\left(\frac{\text{d}}{\text{d}x}-\frac{x}{c_1}\right)^n\,f(x)=0$$ for all $x\in\mathbb{R}$. Now, observe that $$\Delta^n=(MDM^{-1})^n=MD^nM^{-1}\,.$$ Thus, $f\in \ker(\Delta^n)$ if and only if $M^{-1}f\in\ker(D^n)$. Since $\ker(D^n)$ contains all polynomials of degree less than $n$, we conclude that there exists a polynomial function $p:\mathbb{R}\to\mathbb{C}$ of degree less than $n$ such that $$\exp\left(-\frac{x^2}{2c_1}\right)\,f(x)=\big(M^{-1}f\big)(x)=p(x)\,,$$ for each $x\in\mathbb{R}$. Thus, $$f(x)=(Mp)(x)=\exp\left(+\frac{x^2}{2c_1}\right)\,p(x)$$ for all $x\in\mathbb{R}$.

Furthermore, for any function $g:\mathbb{R}\to\mathbb{C}$ with an $n$-th antiderivative, all solutions $f:\mathbb{R}\to\mathbb{C}$ which are $n$-time differentiable and satisfy $$\Delta^n f=g\,,$$ or equivalently, $$\left(\frac{\text{d}}{\text{d}x}-\frac{x}{c_1}\right)^n\,f(x)=g(x)$$ for all $x\in\mathbb{R}$, are given by $$f(x)=\exp\left(+\frac{x^2}{2c_1}\right)\,\big(G(x)+p(x)\big)\,,$$ for all $x\in\mathbb{R}$, where $G$ is an $n$-th antiderivative of $M^{-1}g$, and $p:\mathbb{R}\to\mathbb{C}$ is a polynomial function of degree less than $n$. For example, one can take $$G(x):=\int_0^x\,\int_0^{x_1}\,\cdots\,\int_0^{x_{n-1}}\,\int_0^{x_n}\,\exp\left(-\frac{x_n^2}{2c_1}\right)\,g(x_{n})\,\text{d}x_{n}\,\text{d}x_{n-1}\,\cdots\, \text{d}x_2\,\text{d}x_1\,.$$

In general, if $h:\mathbb{R}\to\mathbb{C}$ has a first antiderivative $H$, then all $n$-time differentiable functions $f:\mathbb{R}\to\mathbb{C}$ such that $$\left(\frac{\text{d}}{\text{d}x}-h(x)\right)^n\,f(x)=0$$ for each $x\in\mathbb{R}$ take the form $$f(x)=\exp\big(+H(x)\big)\,p(x)$$ for all $x\in\mathbb{R}$, where $p:\mathbb{R}\to\mathbb{C}$ is a polynomial function of degree less than $n$. If $g:\mathbb{R}\to\mathbb{C}$ has an $n$-th antiderivative, then all all $n$-time differentiable functions $f:\mathbb{R}\to\mathbb{C}$ such that $$\left(\frac{\text{d}}{\text{d}x}-h(x)\right)^n\,f(x)=g(x)$$ for every $x\in\mathbb{R}$ take the form $$f(x)=\exp\big(+H(x)\big)\,\big(G(x)+p(x)\big)$$ for all $x\in\mathbb{R}$, where $p:\mathbb{R}\to\mathbb{C}$ is a polynomial function of degree less than $n$ and $G(x)$ is the $n$-th antiderivative of $\exp\big(-H(x)\big)\,g(x)$. We may take $$G(x):=\int_0^x\,\int_0^{x_1}\,\cdots\,\int_0^{x_{n-1}}\,\int_0^{x_n}\,\exp\big(-H(x_n)\big)\,g(x_{n})\,\text{d}x_{n}\,\text{d}x_{n-1}\,\cdots\, \text{d}x_2\,\text{d}x_1\,.$$

Batominovski
  • 50,341
3

Here is a fast track. Writing $D=\partial/\partial x$, $c=c_1$ we have the commutation relation: $$ ( x - c D) \; e^{\frac{x^2}{2c}} = e^{\frac{x^2}{2c}} (-cD) $$ Thus, $$ 0 = ( x - c D)^n f_n(x) = ( x - c D)^n e^{\frac{x^2}{2c}} e^{-\frac{x^2}{2c}} f_n(x) = e^{\frac{x^2}{2c}} (- c D)^n e^{-\frac{x^2}{2c}} f_n(x) \Leftrightarrow $$ $$ (- c D)^n e^{-\frac{x^2}{2c}} f_n(x) = 0\Leftrightarrow$$ $$ e^{-\frac{x^2}{2c}} f_n(x) = P_n(x), \ \ P_n\in {\Bbb R}_n[x]\Leftrightarrow$$ $$ f = P_n(x)e^{\frac{x^2}{2c}}, \ \ P_n\in {\Bbb R}_n[x].$$

A fairly general formula (less well-known but with the same proof) is obtained by considering $q\in C^\infty({\Bbb R})$. Then $$ (D - q'(x))^n f_n(x)=0 \ \ \Leftrightarrow \ \ f_n = P_n(x) e^{q(x)}, \ \ P_n\in {\Bbb R}_n[x].$$

H. H. Rugh
  • 35,992