# Intuitive derivation of the Husimi Q function from the Wigner function

The Wigner function is given by

$$W(\alpha)=\frac{1}{\pi^2}\int \text{e}^{\alpha \beta^*-\alpha^*\beta}\text{Tr}\left(\hat \rho \hat D(\beta) \right) \text{d}^2\beta,$$

where $$\hat D(\beta)=\text e^{\beta \hat a^\dagger-\beta^*\hat a}$$ is the displacement operator and $$\hat \rho$$ is the density matrix of the state being analysed.

The Husimi Q function is the Weierstrass transform of the Wigner function. This means that it is integrated over a Gaussian filter

$$Q(\alpha)=\frac 2 \pi \int W(\beta) \text e^{-2|\alpha-\beta|^2}\text{d}^2 \beta.$$

The most common defintion of the Husimi Q function is given by

$$Q(\alpha)=\frac 1 \pi\langle \alpha |\hat \rho|\alpha\rangle$$

but how do you simplify the Gaussian filter such that the function can be written in this form?

• Hint: Try to evaluate trace in the definition of Wigner function in coherent state basis (BCH formula will be handy). – Sunyam Jun 2 '20 at 11:27

@Sunyam has mapped out your homework for you, but here are the two explicit steps that should allow you to unfold it, $$\frac{1}{\pi^2}\int {e}^{\alpha \beta^*-\alpha^*\beta}\text{Tr}\left(\hat \rho e^{ -\beta^*\hat a} e^{\beta \hat a^\dagger } e^{|\beta|^2/2}\right) \text{d}^2\beta= \\ \frac{1}{\pi^2}\int {e}^{\alpha \beta^*-\alpha^*\beta} e^{|\beta|^2/2} \text{Tr}\left( e^{\beta \hat a^\dagger }\hat \rho e^{ -\beta^*\hat a} \right) \text{d}^2\beta ,$$ while $$\text{Tr}\left( e^{\beta \hat a^\dagger }\hat \rho e^{ -\beta^*\hat a} \right)= \frac{1}{\pi}\int \!\! d^2\gamma ~ e^{\beta \gamma^*-\beta ^*\gamma}\langle \gamma|\hat \rho | \gamma \rangle .$$

I really don't know about "intuitive", though, unless that's what you call the Gaussian filtering.

Edit in response to comment : Maybe this and Ch 12 of W. Schleich's book could be helpful.

The double integrals are reduced by "completing the square" and performing the double Gaussian integrals. Here is an example/confirmation in a trivial special case, $$\hat \rho =|0 \rangle \langle|0$$. On the one hand, $$Q(\alpha)={1\over \pi} |\langle \alpha | 0\rangle |^2= {1\over \pi} e^{-|\alpha|^2}.$$ On the other hand, $$W(\beta)= {1\over \pi^2} \int \!\! d^2\alpha ~ e^{-|\alpha|^2/2 + \alpha^* \beta -\alpha \beta^*}= {2\over \pi} e^{-2|\beta|^2}.$$ Plugging this into the Gaussian filter expression, $${4\over \pi^2}\int \!\! d^2\beta ~e^{-2|\beta|^2 -2|\beta-\alpha|^2} = {4\over \pi^2}\int \!\! d^\beta ~e^{-4|\beta|^2 -2|\alpha|^2 +4(\alpha \beta^* + \alpha^* \beta)} \\ = {4\over \pi^2}\int \!\! d^\beta ~e^{-4|\beta-\alpha/2|^2 -|\alpha|^2 }={1\over \pi} e^{-|\alpha|^2},$$ the very previous expression, since the centering of the Gaussians is immaterial to its value.

• What is the trick to cancelling the three double integrals? – Cameron Jun 2 '20 at 14:29
• But how is the complex delta function defined when both variables are complex? – Cameron Jun 2 '20 at 15:26
• I was able to reduce $W(\alpha)$ to a form which included an integral over $\int d^2 \beta e^{(\alpha-\gamma)\beta^*+(\gamma^*-\alpha^*)\beta+|\beta|^2/2}$ but completing the square only reduced it to the form $\int d^2 \beta e^{1/2|\beta+2(\alpha-\gamma)|^2 -2|\alpha-\gamma|^2-2\beta(\alpha^*-\gamma^*)}$ which doesnt seem very helpful since we still have a $\beta$ – Cameron Jun 3 '20 at 14:20
• No! You did not "complete the square"! The exponent is a non-real complex expression, so you need a "square" of the form $(\beta -a) (\beta^*-b)$ which is definitely not real. Smells like a take-home exam. – Cosmas Zachos Jun 3 '20 at 14:45
• pp 59-61 of this booklet should tell one everything one needs to know about the tight equivalence between Husimi and Wigner, but it is in phase space language, not optical phase space, so coherent states are not relied on, exactly as when Husimi introduced it. – Cosmas Zachos Jun 3 '20 at 14:55

We take the Wigner function

$$W(\alpha)=\frac{1}{\pi^2}\int \text{e}^{\alpha \beta^*-\alpha^*\beta}\text{Tr}\left(\hat \rho \hat D(\beta) \right) \text{d}^2\beta,$$

and write the displacement operator as $$\hat D(\beta)=e^{\beta\hat a^\dagger-\beta^*\hat a}=e^{-\beta^*\hat a}e^{\beta\hat a^\dagger}e^{\frac 1 2|\beta|^2}$$ using the BCH formula such that

$$W(\alpha)=\frac{1}{\pi^2}\int \text{e}^{\alpha \beta^*-\alpha^*\beta}\text{Tr}\left(\hat \rho e^{-\beta^*\hat a} e^{\beta\hat a^\dagger}e^{\frac 1 2|\beta|^2} \right) \text{d}^2\beta.$$

Using the cyclic property of the trace, this can be rewritten as

$$W(\alpha)=\frac{1}{\pi^2}\int \text{e}^{\alpha \beta^*-\alpha^*\beta}e^{\frac 1 2|\beta|^2}\text{Tr}\left( \hat e^{\beta\hat a^\dagger} \hat \rho e^{-\beta^*\hat a} \right) \text{d}^2\beta.$$

The trace can be evaluated as

$$\text{Tr}\left( \hat e^{\beta\hat a^\dagger} \hat \rho e^{-\beta^*\hat a} \right)=\frac 1 \pi \int \text d^2\gamma \langle \gamma |e^{\beta \hat a^\dagger}\hat\rho e^{-\beta^*\hat a}|\gamma\rangle=\frac 1 \pi \int \text d^2\gamma \langle \gamma |e^{\beta \hat \gamma^*}\hat\rho e^{-\beta^*\gamma}|\gamma\rangle=\frac 1 \pi \int \text d^2\gamma e^{\beta \gamma^*-\beta^*\gamma} \langle \gamma |\hat\rho |\gamma\rangle.$$

Therefore, the Wigner function can be expressed as

$$W(\alpha)=\frac{1}{\pi^3}\int \int \text{e}^{(\alpha-\gamma)\beta^*- (\alpha^*-\gamma^*)\beta}e^{\frac 1 2|\beta|^2} \langle \gamma |\hat\rho |\gamma\rangle \text{d}^2\beta \text d^2\gamma .$$

By completing the square we find

$$W(\alpha)=\frac{1}{\pi^3}\int \int \text{e}^{\frac 1 2 (\beta+2(\alpha-\gamma))(\beta^*- 2(\alpha^*-\gamma^*))+2|\alpha-\gamma|^2} \langle \gamma |\hat\rho |\gamma\rangle \text{d}^2\beta \text d^2\gamma .$$

which can be simplified to

$$W(\alpha)=\frac{2}{\pi^2}\int \text{e}^{2|\alpha-\gamma|^2} \langle \gamma |\hat\rho |\gamma\rangle \text d^2\gamma .$$

This shows that

$$W(\alpha)=\frac{2}{\pi}\int \text{e}^{2|\alpha-\gamma|^2} Q(\gamma) \text d^2\gamma .$$