The linear entropy is a measure of mixedness in quantum states. Its main feature is that it is easy to compute. It is a scalar defined as
$$S_L \, \dot= \, 1 - \mbox{Tr}(\rho^2) \,$$
where ρ is the density matrix of the state.
The linear entropy can range between zero, corresponding to a completely pure state, and (1 − 1/d) , corresponding to a completely mixed state. (Here, d is the dimension of the density matrix.)
Linear entropy is trivially related to the purity γ of a state by
SL = 1 − γ .
Motivation
The linear entropy is an approximation to the Von Neumann entropy S , which is defined as
$$S \, \dot= \, -\mbox{Tr}(\rho \ln \rho) \, .$$
The linear entropy is obtained by approximating lnρ with the first order term (ρ − 1) in the Mercator series
− Tr(ρlnρ) → − Tr(ρ(ρ − 1)) = Tr(ρ − ρ2) = 1 − Tr(ρ2) = SL
where the unit trace property of the density matrix has been used to get the second to last equality.
The linear entropy and Von Neumann entropy are similar measures of the mixedness of a state, although the linear entropy is easier to calculate because it does not require the diagonalization of the density matrix.
Alternate definition
Some authors define linear entropy with a different normalization
$$S_L \, \dot= \, \frac{d}{d-1} (1 - \mbox{Tr}(\rho^2) ) \, .$$
This ensures that the quantity ranges from zero to unity.