Maximum-Entropy Density Operator

The goal of this section is to obtain the density operator $ \hat{\rho}$, with $ Tr\{ \hat{\rho}\}=1$, that maximizes the entropy $ S= -k$   Tr$ \{ \hat{\rho}$   ln$ \hat{\rho} \}$ of a system characterized by an ensemble average internal energy
$\displaystyle E =$   Tr$\displaystyle \{\hat{\rho} \hat{H}\},$ (25)

and fix extensive properties $ X$ such as $ X=(V,N)$ (i.e., canonical and microcanonical ensembles). This is accomplished by implementing the Method of Lagrange Multipliers to maximize the function

$\displaystyle f(\hat{\rho}) \equiv -k$   Tr$\displaystyle \{\hat{\rho}$   ln$\displaystyle \hat{\rho} \} + \gamma (E -Tr\{\hat{\rho} \hat{H}\}) + \gamma'(1-Tr\{\hat{\rho} \}),$ (26)

where $ \gamma$ and $ \gamma'$ are Lagrange Multipliers. We, therefore, solve for $ \hat{\rho}$ from the following equation

$\displaystyle \frac{\partial f}{\partial \hat{\rho}} \Bigg )_X= 0,$ (27)

and we obtain that the density operator that satisfies Eq. (27) must satisfy the following equation:

$\displaystyle Tr \{ -k$   ln$\displaystyle \hat{\rho} - k - \gamma \hat{H} - \gamma' \} =0.$ (28)

Therefore,

$\displaystyle -$ln$\displaystyle \hat{\rho} = 1 + \frac{\gamma}{k} \hat{H} + \frac{\gamma'}{k}.$ (29)

Exponentiating both sides of Eq. (29) we obtain

$\displaystyle \hat{\rho} =$   exp$\displaystyle (-(1 + \frac{\gamma'}{k}))$   exp$\displaystyle (-\frac{\gamma}{k} \hat{H}),$ (30)

and, since Tr$ \{\hat{\rho}\}$=1,

exp$\displaystyle (-(1 + \frac{\gamma'}{k}))=\frac{1}{Z},$ (31)

where $ Z$ is the partition function

$\displaystyle Z \equiv Tr \{ exp(-\beta \hat{H}) \},$ (32)

with $ \beta \equiv \gamma/k$. Substituting Eqs. (32) and (31) into Eq. (30), we obtain that the density operator that maximizes the entropy of the ensemble, subject to the contraint of average ensemble energy $ \bar{E}$, is

$\displaystyle \hat{\rho} = Z^{-1}$   exp$\displaystyle (-\beta \hat{H}).$ (33)

Note that

$\displaystyle \frac{\partial \hat{\rho}}{\partial t} = 0,$ (34)

when $ \hat{\rho}$ is defined according to Eq. (33) and, therefore, the system is at equilibrium.

Exercise 6: Use Eqs. (17) and (33) to prove Eq. (34).