The variance of the photon number distribution is : V_n=\langle \Delta n^2\rangle=\langle n^2\rangle-\langle n\rangle^2= \left\langle \left(a^{\dagger}a\right)^2\right\rangle-\langle a^{\dagger}a\rangle ^2. Using commutation relations, this can be written as : V_n=\langle {(a^{\dagger}})^2a^2 \rangle+\langle a^{\dagger}a\rangle-\langle a^{\dagger}a\rangle ^2. This can be written as : V_n-\langle n\rangle=\langle (a^\dagger)^2 a^2\rangle -\langle a^{\dagger}a\rangle^2. The second-order intensity
correlation function (for zero delay time) is defined as : g^{(2)}(0)={{\langle (a^\dagger)^2 a^2\rangle}\over{\langle a^{\dagger}a\rangle^2}}. This quantity is basically the probability of detecting two simultaneous photons, normalized by the probability of detecting two photons at once for a random photon source. Here and after we assume stationary counting statistics. Then we have : {{1}\over{(\langle n\rangle)^2}}(V_n-\langle n\rangle) =g^{(2)}(0)-1. Then we see that sub-Poisson photon statistics, one definition of photon antibunching, is given by g^{(2)}(0) . We can equivalently express antibunching by Q where the
Mandel Q parameter is defined as : Q\equiv \frac{V_n}{\langle n \rangle}-1. If the field had a classical stochastic process underlying it, say a positive definite probability distribution for photon number, the variance would have to be greater than or equal to the mean. This can be shown by an application of the Cauchy–Schwarz inequality to the definition of g^{(2)}(0). Sub-Poissonian fields violate this, and hence are nonclassical in the sense that there can be no underlying positive definite probability distribution for photon number (or intensity). Photon antibunching by this definition was first proposed by Carmichael and Walls and first observed by
Kimble,
Mandel, and Dagenais in
resonance fluorescence. A driven atom cannot emit two photons at once, and so in this case g^{(2)}(0)=0. An experiment with more precision that did not require subtraction of a background count rate was done for a single atom in an ion trap by Walther et al. A more general definition for photon antibunching concerns the slope of the correlation function away from zero time delay. It can also be shown by an application of the
Cauchy–Schwarz inequality to the time dependent intensity
correlation function : g^{(2)}(\tau)={{\langle a^{\dagger}(0)a^{\dagger}(\tau)a(\tau)a(0)\rangle}\over{\langle a^{\dagger}a\rangle^2}}. It can be shown that for a classical positive definite probability distribution to exist (i.e. for the field to be classical) g^{(2)}(\tau) \leq g^{(2)}(0). Hence a rise in the second order intensity correlation function at early times is also nonclassical. This initial rise is photon antibunching. Another way of looking at this time dependent correlation function, inspired by quantum trajectory theory is : g^{(2)}(\tau)={{\langle a^{\dagger}a\rangle_C}\over{\langle a^{\dagger}a\rangle}} where : \langle O \rangle_C \equiv \langle \Psi_C |O|\Psi_C\rangle. with |\Psi_C\rangle is the state conditioned on previous detection of a photon at time \tau=0. ==Experiments==