Opening the book…
Let’s look a bit further into the origin of mixed states and their intrinsically statistical nature. To this point we have discussed the time-dependent quantum mechanical properties of pure states, which are characterized by a single wavefunction derived from a Hamiltonian that specifies all particles and fields in the system of interest. In condensed phases, molecules interact with one another, and as a result no two molecules have the same state. We cannot write down an exact Hamiltonian for these problems because we cannot (and actually do not want to) specify the states of all degrees of freedom. It is more practical to concentrate our attention on one, or a few, degrees of freedom that are directly observed in a measurement and describe the influence of the surroundings on the observed degrees of freedom in a statistical manner. These observations lead to the concept of mixed states and statistical mixtures. A mixed state refers to any case in which we don’t have complete information on the observed system of interest and must instead describe its properties and behavior statistically. The most common example is a system at thermal equilibrium, but it applies to any large set of independently prepared states that have no phase relationship with each other.
How does a system get into a mixed state? Generally, if you have two systems and you put these in contact with each other, the interaction between the two will lead to a new joint system that is inseparable. Consider two systems described by \(H_{S}\) and \(H_{B}\). The eigenstates of \(H_{S}\) are \(\left| n \right\rangle \) and those of \(H_{B}\) are \(\left| \alpha \right\rangle \). Before these systems interact, the Hamiltonian is
\[H_{0} = H_{S} + H_{B} \label{eq12.4.01}\]
and the state of the joint system \(\left| \psi _{0} \right\rangle \) can be described as an expansion in product states of \(\left| n \right\rangle \) and \(\left| \alpha \right\rangle \).
\[\begin{gather}
\left| \psi _{0} \right\rangle = \left| \psi _S^{0} \right\rangle \left| \psi _B^{0} \right\rangle \vphantom{\sum\limits_{\alpha }} \label{eq12.4.02} \\
\left| \psi _S^{0} \right\rangle = \sum\limits_{n} s_{n}\left| n \right\rangle \qquad \left| \psi _B^{0} \right\rangle = \sum\limits_{\alpha } b_{\alpha }\left| \alpha \right\rangle \label{eq12.4.03}\\
\left| \psi _0 \right\rangle = \sum\limits_{n,\alpha } s_{n}b_{\alpha }\left| n\alpha \right\rangle \label{eq12.4.04}
\end{gather}\]
where \(s\) and \(b\) are expansion coefficients. After these states are allowed to interact, we have a new state \(\left| \psi (t) \right\rangle \). The new mixed state can still be expressed in the zero-order basis, although this does not represent the eigenstates of the new Hamiltonian with the interaction \(H_{SB}\):
\[H = H_{0} + H_{SB} \label{eq12.4.05}\]\[\left| \psi (t) \right\rangle = \sum\limits_{n,\alpha } c_{n,\alpha} \left| n \alpha \right\rangle \label{eq12.4.06}\]
For any point in time, \(c_{n,\alpha }\) is the complex amplitude for the basis state \(|n\alpha\rangle\). The coefficient \(c_{n,\alpha }\) encodes \(P_{n,\alpha } = \left| c_{n,\alpha } \right|^{2}\), the joint probability for finding particle of \(\left| \psi _S \right\rangle \) in state \(\left| n \right\rangle \) and simultaneously finding particle of \(\left| \psi _B \right\rangle \) in state \(\left| \alpha \right\rangle \). Generally speaking, any time after bringing the systems in contact \(c_{n,\alpha } \ne s_{n}b_{\alpha }\).
In the case of experimental observables, we are typically able to observe or manipulate one \(H_S\), the “system”, and are blind to the other \(H_B\), the “bath”. In condensed phases, the bath has vastly more degrees of freedom and is treated as a continuum. The key property of a mixed state or statistical mixture is that we don’t precisely know the state of the bath. For instance, in an equilibrium system, all we can do is specify the probability that the system occupies a state \(|n\rangle\), averaged over all possible states of the bath:
\[P_{n} = \sum\nolimits_{\alpha } \left| c_{n,\alpha } \right|^{2} = \overline {\left| c_{n} \right|^{2}} \label{eq12.4.07}\]
Here averaging over the bath removes the pureness of the state since we have lost information on the states of the bath. We refer to \(P_n\) as the population of state \(n\).
Now let’s look at how this emerges by describing ensembles. Imagine a dilute room temperature solution of molecules within a weakly interacting solvent, or equivalently. The key observation here is that the solute molecules can be treated as being completely independent entities that have no interactions with each other—only with the liquid. As a result, no quantum mechanical phase relationship exists between solute molecules. In isolation, each solute molecule has the same molecular Hamiltonian \(H_{\text{mol}}\) and eigenfunctions
\[H_{\text{mol}}\left| n \right\rangle = E_{n}\left| n \right\rangle \label{eq12.4.08}\]
As a result of their interaction with the solvent, the unique state of molecule \(i\) within the solution will be described by the wavefunction \(\left| \psi ^{(i)} \right\rangle \), which we will expand in this isolated molecule eigenbasis. This is reasonable for weak solute-solvent interactions, but generally one wants to ensure that the isolated system Hilbert space provides a complete basis for describing the system once you put it in the condensed phase. Once in the liquid, the details of the amplitudes of the eigenstates at any moment will differ from one molecule to the next depending on the time-dependent local environment.
To describe the solution, we introduce a Hamiltonian for the pure liquid \(H_{\text{liq}}\) into which we wrap all of the solvent degrees of freedom and which we have only minimal statistical information about. Finally, there is a solute-solvent interaction Hamiltonian \(H_{\text{mol–liq}}^{(i)}\) which describes how these interactions influence the state of a specific solute molecule indexed \(j\). Thus, the Hamiltonian for any individual molecule is
\[H^{(i)} = H_{\text{mol}} + H_{\text{liq}} + H_{\text{mol–liq}}^{(i)} \label{eq12.4.09}\]
To build a statistical description of the system, we don’t need to treat each molecule separately, only describe the variety of possible molecular interactions with the liquid. If we average eq. (\ref{eq12.4.09}) over all molecules, we can obtain the Hamiltonian:
\[\begin{aligned}
H & = \frac{1}{N}\sum\limits_{j = 1}^{N} H^{(i)} \\
& = H_{\text{mol}} + H_{\text{liq}} + \bar H_{\text{mol–liq}} \\
\end{aligned} \label{eq12.4.10}\]
where \(\bar H_{\text{mol–liq}}\) describes the average influence of the liquid on the solute. This Hamiltonian takes the form of the system-bath Hamiltonian in eq. (12.1.1), in which a molecular solute’s interactions with its surroundings are treated in a statistical manner.
We now turn our attention to the expectation values that we would measure in an experiment. First, we recognize that for the individual molecule \(i\), the expectation value for an operator would be expressed as
\[\left\langle A \right\rangle ^{(i)} = \left\langle \psi ^{(i)} \right|\hat A\left| \psi ^{(i)} \right\rangle \label{eq12.4.11}\]
In the case of the ensemble, we would also average the expectation value over the \(N\) molecules.
\[\left\langle A \right\rangle = \frac{1}{N}\sum\limits_{i = 1}^{N} \left\langle A \right\rangle ^{(i)} \label{eq12.4.12}\]
Equivalently, this could be expressed as an average over the probability that a molecule is described by wavefunction \(\left| \psi ^{(j)} \right\rangle \)
\[\left\langle A \right\rangle = \sum\limits_{j = 1}^{M} P_{j} \left\langle \psi ^{(j)} \right|\hat A\left| \psi ^{(j)} \right\rangle = \sum\limits_{j = 1}^{M} P_{j}\left\langle A \right\rangle ^{(j)} \label{eq12.4.13}\]
Since the isolated molecular eigenstates (\(n\)) and the liquid (\(\alpha\)) states form a complete basis set, the pure state of molecule \(i\) can be expressed as an expansion in this basis:
\[\left| \psi ^{(i)} \right\rangle = \sum\limits_{a,\alpha } c_{a,\alpha}^{(i)}\left| n \alpha \right\rangle \label{eq12.4.14}\]
Expanding eq. (\ref{eq12.4.12}) with the use of eqs. (\ref{eq12.4.11}) and (\ref{eq12.4.14}) gives us the expectation value
\[\left\langle A \right\rangle = \frac{1}{N}\sum\limits_{i = 1}^{N} \sum\limits_{n,m} \sum\limits_{\alpha ,\beta } c_{m,\beta}^{(i)}\left( c_{n,\alpha}^{(i)} \right)^{ * } \left\langle n\alpha \right|\hat A\left| m\beta \right\rangle \label{eq12.4.15}\]
This is an excruciating combination of sums over molecules, system eigenstates, and bath states, which reflects the daunting pure-state representation of the expectation value. However, for a mixed state we will average over the possible configurations of the bath and their influence on the different molecules. This results in an average of the product of complex system eigenfunction amplitudes for the system states:
\[\left\langle c_{m}c_n^{*} \right\rangle = \frac{1}{N}\sum\limits_{i = 1}^{N} \sum\limits_{\alpha ,\beta } c_{m,\beta} ^{(i)}\left( c_{n,\alpha}^{(i)} \right)^{ * } \label{eq12.4.16}\]
We also take the operator \(\hat{A}\) to only be a function of the system coordinates, so that \(\left\langle n\alpha \right|\hat A\left| m\beta \right\rangle = \left\langle n \right|\hat A\left| m \right\rangle \). Then the expectation value for the mixed state ensemble is
\[\left\langle A \right\rangle = \sum\limits_{n,m} \left\langle c_{m}c_n^{*} \right\rangle \left\langle n \right|\hat A\left| m \right\rangle \label{eq12.4.17}\]
The quantity \(\left\langle c_{n}c_m^{*} \right\rangle \) specifies the mixed state of the system. We retain information on the average state of solute molecules while losing precise information about the liquid environment. We will associate \(\left\langle c_{n}c_m^{*} \right\rangle \) with the matrix elements of the density matrix or density operator \(\rho\) that we describe in the following section:
\[\rho _{mn} = \left\langle c_{m}c_n^{*} \right\rangle \label{eq12.4.18}\]
Substituting this definition into eq. (\ref{eq12.4.17}), we find that the expectation value becomes
\[\begin{aligned}
\left\langle A \right\rangle &= \sum\limits_{n,m} \rho _{mn}A_{nm} \\
& = \sum\limits_{m} \left\langle m \right| \rho A\left| m \right\rangle \\
& = Tr\left[ \rho A \right] \\
\end{aligned} \label{eq12.4.19}\]
where the trace \(\text{Tr}\left( \cdots \right) = \sum\nolimits_{m} \left\langle m \right| \cdots \left| m \right\rangle \) refers to summing over the diagonal elements of a matrix.
It may not be obvious how eq. (\ref{eq12.4.17}) relates to the expected expression for an ensemble-averaged expectation value in eq. (12.3.07)
\[\left\langle A \right\rangle = \sum\limits_{n} P_{n}\left\langle n \right| {\hat A} \left| n \right\rangle \label{eq12.4.20}\]
The difference is that eq. (\ref{eq12.4.17}) applies to all systems, including non-equilibrium ones. Remember, in an equilibrium system we are dealing with a statistical mixture in which no coherences or phase relationships between members of the ensemble are present. The lack of coherence is the property that allows the equilibrium ensemble average of \(\left\langle c_{n}c_m^{*} \right\rangle \) to be equated with the thermal population \(p_n\). To evaluate this average, we recognize that these are complex numbers and that the equilibrium ensemble average of the expansion coefficients is equivalent to phase averaging over the expansion coefficients. Since at equilibrium all phases are equally probable,
\[\left\langle c_{n}c_m^{*} \right\rangle _{\text{eq}} = \frac{1}{2\pi }\int_{0}^{2\pi } c_{n}c_m^{*} \, d\phi = \frac{1}{2\pi } \left| c_{n} \right| \left| c_{m} \right|\int_{0}^{2\pi } e^{ -\mathrm{i}\phi _{nm}} \, d\phi _{nm} = \left| c_{n} \right| \left| c_{m} \right|\delta (\phi _{n} - \phi _{m}) \label{eq12.4.21}\]
where \(c_{n} = \left| c_{n} \right| e^{\mathrm{i}\phi _{n}}\) and \(\phi _{nm} = \phi _{n} - \phi _{m}\). The integral is quite clearly zero unless \(\phi _{n} = \phi _{m}\), giving an equilibrium average consistent with eq. (12.3.05):
\[\left\langle c_{n}c_m^{*} \right\rangle _{\text{eq}} = \left| c_{n} \right|^{2} = P_{n} = \frac{e^{ - \beta E_{n}}}{Z} \label{eq12.4.22}\]
This page titled 12.4: Mixed States was last modified on Thu, 04 Jun 2026 01:16:15 GMT and is shared under a CC BY-NC-SA 4.0 license and was authored, remixed, and/or curated by Andrei Tokmakoff via source content that was edited to the style and standards of the LibreTexts platform.
The book could not be loaded just now.