Suppose, I have some operator $\hat{A}$, such that in the $x$-basis, it is written as $f(x).$ I'm trying to calculate the expectation value of this operator in integral form. That is given by the following expression :

$$\langle\hat{A}\rangle = \int\psi^*(x)f(x)\psi(x)dx$$

We have assumed that the wave function is normalized here.

In this case, we have done the entire integral in the $x$-basis. I suppose, we can actually write the above integral as : $$\langle\hat{A}\rangle = \int\langle \psi|x\rangle\langle x|\hat{A}|x\rangle\langle x|\psi\rangle dx$$

However, suppose my wave function evolves and becomes $\psi(u)$, where $u=g(x)$. We can try to find the expectation value of this new wave function. We know : $$\langle\hat{A}\rangle = \frac{\int\psi^*(u)f(x)\psi(u)dx}{\int\psi^*(u)\psi(u)dx}$$

However, we need to change the variable of integration to $u$, and so we do the following:

$$\langle\hat{A}\rangle = \frac{\int\psi^*(u)f\space o\space g^{-1}(u)\psi(u)\frac{du}{g' \space o \space g^{-1}(u)}}{\int\psi^*(u)\psi(u)\frac{du}{g' \space o \space g^{-1}(u)}}$$

Now we can evaluate the integral, as everything is in the $u$ basis.

However, what we did here is first write the integral in the $x$ basis, and then use suitable transformations to make $u$ the variable of integration. However, how can I write the integral directly in the $u$ basis.

For example, we know: $$\langle\hat{A}\rangle=\frac{\langle\psi|\hat{A}|\psi\rangle}{\langle\psi|\psi\rangle}$$

I suppose we can insert any basis here in the following way:

$$\langle\hat{A}\rangle=\frac{\langle\psi|x\rangle\langle x|\hat{A}|x\rangle\langle x|\psi\rangle}{\langle\psi|x\rangle\langle x|\psi\rangle}=\frac{\langle\psi|u\rangle\langle u|\hat{A}|u\rangle\langle u|\psi\rangle}{\langle\psi|u\rangle\langle u|\psi\rangle}$$

Moreover, $$\frac{\langle\psi|x\rangle\langle x|\hat{A}|x\rangle\langle x|\psi\rangle}{\langle\psi|x\rangle\langle x|\psi\rangle}=\int\psi^*(x)f(x)\psi(x)dx$$

Hence, we must have :

$$\frac{\langle\psi|u\rangle\langle u|\hat{A}|u\rangle\langle u|\psi\rangle}{\langle\psi|u\rangle\langle u|\psi\rangle}=\frac{\int\psi^*(u)f\space o\space g^{-1}(u)\psi(u)\frac{du}{g' \space o \space g^{-1}(u)}}{\int\psi^*(u)\psi(u)\frac{du}{g' \space o \space g^{-1}(u)}}$$

Now I can see that $f(x)$ is replaced by $f\space o\space g^{-1}(u)$, as this is the representation of the operator in the $u$ basis. However there is also a $g'\space o \space g^{-1}(u)$ term below the differential $du$.

My question is, where is this factor coming from ? How do we formulate the entire thing using dirac notation ? I know the initial equation in terms of the $x$ basis, and so I use the transformations to convert the integral into the $u$ basis. This can be shown as :

$$\langle \hat{A}\rangle=\frac{\int\psi^*(g(x))f(x)\psi(g(x))dx}{\int\psi^*(g(x))\psi(g(x))dx}=\frac{\int\psi^*(u)f\space o\space g^{-1}(u)\psi(u)\frac{du}{g' \space o \space g^{-1}(u)}}{\int\psi^*(u)\psi(u)\frac{du}{g' \space o \space g^{-1}(u)}}$$

Using Bra-Ket notation, I can write the integral in $x$ basis, and then transform it into the $u$ basis. However, I don't see how to write the integral directly in the $u$ basis, using Bra-ket notation. It is clear to me, how the expression of the operator changes in the Bra-ket notation, in the new basis. However, it is unclear to me where the factor under the differential comes from, if we try to write the integral directly using Bra-Ket notation.

I hope I've been able to explain my confusion. Any help would be highly appreciated.


My initial intuition is that the term under the differential $du$ is the weight factor of the integral. However, I have no idea how it is represented in Dirac notation. For example, in the $x$ basis, the weight factor is clearly $1$ as there is nothing under the differential $dx$. Where does this come from in Dirac notation to integral notation conversion?

  • 1
    $\begingroup$ Linked. $\endgroup$ Oct 24, 2021 at 22:49
  • $\begingroup$ @CosmasZachos yeah, but this is a different question somewhat. Here, I'm asking about how to represent the entire thing in Dirac notation, as in, where do the individual integral terms originate from in Bra-ket notation. For example, $\langle u|\hat{A}|u \rangle = f\space o\space g^{-1}(u)$ and $\langle u|\psi\rangle = \psi(u)$. However, where does the term under the differential $du$ come from in Dirac notation ? $\endgroup$ Oct 24, 2021 at 22:53
  • $\begingroup$ You might need some integrals in the "Moreover" expression... Furthermore, you might denote the states the expectation values are for, since they are distinctly different. $\endgroup$ Oct 24, 2021 at 23:15
  • 3
    $\begingroup$ It looks like you read too quickly. 1. Inserting a complete set of states means inserting $\int dx \left | x \right > \left < x \right |$, not just $\left | x \right > \left < x \right |$. 2. Eigenstates of the position operator in Cartesian co-ordinates are also eigenstates of the position operator in any other system of co-ordinates. 3. You never have to change what basis or co-ordinate system you use because the wavefunction has evolved. $\endgroup$ Oct 25, 2021 at 5:18
  • 3
    $\begingroup$ Hmm, this sounds like an XY problem. Writing a time evolved wavefunction in terms of the old one in the form $\psi(g(x))$ is just a very strange thing to do -- why are you considering that in the first place? $\endgroup$
    – knzhou
    Oct 25, 2021 at 20:35

1 Answer 1


As you suggest, $u=g(x)$, taken invertible, and $$ [\hat u, \hat x]=[\hat A, \hat x]=[\hat u, \hat A]=0,\\ \hat u \equiv g(\hat x), ~~~\leadsto |u\rangle \propto |x\rangle ~~~\leadsto \hat u |x\rangle = g(\hat x)|x\rangle = g(x)|x\rangle . $$ Likewise, $$ \hat A |x\rangle= f(x) |x\rangle= f(g^{-1}(\hat u)) |x\rangle \leadsto \\ \hat A |u\rangle= f(g^{-1}( u)) |u\rangle . $$

In your sibling question you seem to appreciate that $\psi(x)$ does not "evolve" unitarily to $\psi (u)=\psi(g(x))\equiv \tilde \psi(x)$, since $|\psi\rangle$ and $|\tilde \psi\rangle$ have different normalizations--see example below.

Let's illustrate some of this with the trivial scaling case g(x)=ax suggested in the sibling question, so g'=a, so du=a dx. $$ {\mathbb I}=\int \!\!dx~~ |x\rangle \langle x|= \int \!\!du~~ \frac{1}{a}|x\rangle \langle x| \\ =\int \!\!du~~ |u\rangle \langle u| , $$ That is, $|u\rangle=\frac{1}{\sqrt{a}}|x\rangle$.

You then have, $$\psi(u)=\langle u|\psi\rangle =\frac{1}{\sqrt{a}}\langle x|\psi\rangle= \psi (x)/\sqrt{a} =\langle x|\tilde \psi\rangle=\tilde \psi(x)=\psi(ax),\\ |\tilde \psi\rangle= |\psi\rangle / \sqrt{a} ~~~\implies ~~~ \langle \tilde \psi | \tilde \psi \rangle = \langle \psi | \psi \rangle/a, $$ as remarked at the beginning. For this trivial case (only), the two expectation values you are computing turn out to be equal, but this is not a general feature, of course!

You may take it from here, $$|\tilde \psi\rangle= \frac{1}{\sqrt{g'(\hat x)}}|\psi\rangle, ~~ |u\rangle= {1\over \sqrt{g'}}|x\rangle\leadsto \\ \frac{\langle\psi|\hat{A}|\psi\rangle}{\langle\psi|\psi\rangle} = \int\!\! du |\psi(u)|^2 f(g^{-1}(u)) , ~~\mathbf{ but}\\ \frac{\langle\tilde\psi|\hat{A}|\tilde\psi\rangle}{\langle\tilde \psi|\tilde \psi\rangle} = {\int\!\! {du\over g'(g^{-1}(u))} |\psi(u)|^2 f(g^{-1}(u)) \over \int\!\! {du\over g'(g^{-1}(u))} |\psi(u)|^2 }~~, $$ as stressed! Check their identity for simple scaling, as per above remark.

  • $\begingroup$ Thanks for this, but I was trying to do something slightly different. The new basis $|u\rangle$ is not normalized to start with, so the resolution of identity in this new basis should be of the form $\int du\space h(u)|u\rangle\langle u|=\hat{1}$. Then we can set $|u\rangle \rightarrow \sqrt{h(u)}|u\rangle$ to normalize $|u\rangle$. $\endgroup$ Oct 26, 2021 at 14:26
  • $\begingroup$ I'm sorry, but I don't get your method of normalizing the basis vector. My idea was to keep it in that non-normalized form $\int du\space h(u)|u\rangle\langle u| = \hat{1}$, while introducing the identity operator. The equation $|u\rangle = \frac{|x\rangle}{\sqrt{g'(x)}}$ is something I'm not being able to understand intuitively. Should I not rather say that $\int du\space h(u)|u\rangle\langle u| = \hat{1}$, and then normalize $|u\rangle$, by taking the root of the function in front of it, and then absorbing it into the ket ? $\endgroup$ Oct 26, 2021 at 14:34
  • $\begingroup$ No, how did you get your original expression in the first place? Unless you orthonormalize your basis, all expressions such as $\langle u|\psi \rangle= \psi (u)$ etc are invalid. I illustrated the resolution of the identity for the simple scaling, so you get it... $\endgroup$ Oct 26, 2021 at 14:38
  • $\begingroup$ If you would kindly move this to chat, perhaps I could share my working and the specific part that I am not being able to understand. Thank you so much for taking the time. $\endgroup$ Oct 26, 2021 at 14:40
  • $\begingroup$ I'm not sure I know how to do this. $|u\rangle \propto |x\rangle$ does not fix the former's normalization. One adjusts it versus a normalized basis, the latter, so now the former is also normalized. $\endgroup$ Oct 26, 2021 at 14:45

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.