9

What is a Gibbs state and what does it differ from a pure state? Say I have a two-level atom and it is described by a Gibbs state $\rho_G = \dfrac{e^{- \frac{H}{kT}}}{Z}$. I know $Z$ is a partition function. How can I express $\rho_G$ explicitly in the diagonalized form?

Let the eigenstates of $H$ be the ground state $\left|g\right>$ and the excited state $\left|e\right>$ of the atom. Is $\rho_G = \left|g\right>\left<g\right|$ in this case with near zero temperature?

Qmechanic
  • 201,751
diff
  • 681
  • @diff Check out this question, as well as the one it links to: http://physics.stackexchange.com/questions/112845/coherences-in-the-density-matrix – Rococo Jun 22 '16 at 03:24

1 Answers1

11

The expression you give, $$\rho_G=\frac 1Z e^{-H/kT},$$ is already explicit. If you want an explicitly diagonalized expression, then you can use the fact that a function $f:\mathbb C\to\mathbb C$ is defined as acting on Hilbert space operators $A$, using the eigenvector route, as giving $f(A)|a⟩=f(a)|a⟩$ whenever $A|a⟩=a|a⟩$. Thus if $$ H=\sum_n E_n|E_n⟩⟨E_n| $$ (taking a point spectrum for simplicity), with possible degeneracies and so on, then $$ \rho_G=\frac 1Z e^{-H/kT}=\sum_n \frac 1Z e^{-E_n/kT}|E_n⟩⟨E_n|. $$

As always, $Z$ is the partition function, chosen to make $\rho_G$ normalized to $\mathrm{Tr}(\rho_G)=1$, so $$ Z =\mathrm{Tr}\mathopen{}\left(e^{-H/kT}\right)\mathclose{} =\sum_n e^{-E_n/kT}. $$

Emilio Pisanty
  • 132,859
  • 33
  • 351
  • 666
  • How to define a Gibbs state if the Hamiltonian is non-Hermitian (e.g. $H=H_S +i\Gamma$ for effective system-bath coupling); where the density matrix become non-Hermitian? – donnydm Mar 22 '18 at 14:56