Opening the book…
Separability and Product States
While single particles are rarely of interest in chemical applications, molecular wavefunctions are another extreme involving thousands of electrons and nuclei. In such complex systems, how do we identify which particles matter and which have little bearing on our problem? We must be able to identify ways of concentrating on the subset of interest, whether this is an isolated polyatomic molecule or a small molecule interacting with a complex solvent environment. Separation of the Hamiltonian into these subsets is the underpinning of all strategies.
To start, consider the example of a Hamiltonian that can be written as a sum of two contributions:
\[\hat{H}=\hat{H}_{1}+\hat{H}_{2} \label{eq21.2.2.01}\]
These subsystems refer to completely separable degrees of freedom, with no interactions between them. This example is rarely encountered, but it could refer to cases where interactions are negligible, such as the case of two isolated molecules. If so we can find solutions to the timeindependent Schrödinger equation (TISE) for each subsystem: \(H_{1}\left|\psi_{1, n}\right\rangle=E_{1, n}\left|\psi_{1, n}\right\rangle\) and \(H_{2}\left|\psi_{2, m}\right\rangle=E_{2, m}\left|\psi_{2, m}\right\rangle\) where \(n\) and \(m\) are the indices labeling the eigenstates and energy eigenvalues of the \(1\) and \(2\) subsystems, respectively. Since they are independent, with the operators only acting in their space, the wavefunction of the joint system \(\Psi\) can be represented in terms of product states of the eigenstates of the two sub-systems: \(\left|\Psi_{n, m}\right\rangle=\left|\psi_{1, n}\right\rangle\left|\psi_{2, m}\right\rangle\), and the energy of the full system is simply additive in the two subsystems
\[H|\Psi\rangle=\left(H_{1}+H_{2}\right)\left|\psi_{1} \psi_{2}\right\rangle=\left(E_{1}+E_{2}\right)\left|\psi_{1} \psi_{2}\right\rangle \label{eq21.2.2.02}\]
Furthermore, the full wavefunction can be expressed as an expansion in the product basis:
\[|\Psi\rangle=\sum_{n, m} c_{n} c_{m}\left|\psi_{1, n} \psi_{2, m}\right\rangle \label{eq21.2.2.03}\]
This reasoning extends to as many separable degrees of freedom in which the Hamiltonians can be expressed as a sum of independent terms. We will commonly specify the state of the multiparticle or multistate system using the quantum numbers for the subsystems: \(|\psi\rangle=\left|n_{0}, n_{1}, n_{2} \cdots\right\rangle\). In its matrix form, we write the Hamiltonian in eq. (\ref{eq21.2.2.03}) in a block diagonal form in which the individual contributions \(H_{i}\) form one of the blocks, and the ket wavefunction \(|\psi\rangle\) is a tensor product of the column vectors \(\left|\psi_{i}\right\rangle\)
As one might expect, this form of the Hamiltonian serves to dramatically simplify solutions to the Schrödinger equation. This form occurred above in the description of the \(3 \mathrm{D}\) particle in a box and the spherically symmetric potential in which we separated the radial Hamiltonian from the full Hamiltonian. Even when not strictly possible, approximations that allow us to cast the problem in this form are cornerstones of much of molecular quantum mechanics, for instance in the adiabatic approximation, which uses a time-scale separation to partition the problem by fast and slow degrees of freedom. Such strategies are often used for describing molecular quantum states in terms of product states in the electronic, nuclear, rotational, and spin degrees of freedom, where electron-nuclear, ro-vibrational, and spin-orbit couplings are treated perturbatively.
Interacting systems
Naturally, a non-interacting universe is uninteresting if not unphysical, and Hamiltonians cannot typically be fully decoupled. Let’s reexamine our two-subsystem example, but now allowing them to interact through a potential
\[\hat{H}^{\prime}=\hat{H}_{1}+\hat{H}_{2}+\hat{V}_{1,2} \label{eq21.2.2.04}\]
This form indicated that the two systems can influence the energy of each other but not exchange particles, as one might expect of intermolecular interactions. The product states that we determined in the absence of interactions (\(\hat{V}_{1,2} \rightarrow 0\)) are not eigenstates of the new Hamiltonian \(H^{\prime}\), as the eigenstates of \(H_{1}\) become entangled with those of \(H_{2}\). Nonetheless, the original product states still form a complete basis by which to express the new state of the system:
\[\left|\Psi^{\prime}\right\rangle=\sum_{n, m} C_{n . m}\left|\psi_{1, n} \psi_{2, m}\right\rangle \label{eq21.2.2.05}\]
However, \(C_{n, m} \neq c_{n} c_{m}\), generally speaking. To the extent that the interaction is weak, the degree of mixing is small and the corrections to the energy eigenvalues and eigenstates can be made with perturbation theory. We will be more quantitative about what "weak" means soon.
To look at this further, consider the full TISE for the Hamiltonian in eq. (\ref{eq21.2.2.04}) \(H^{\prime}\left|\Psi^{\prime}\right\rangle=E_\text{tot}\left|\Psi^{\prime}\right\rangle\) evaluated in the original orthonormal product state basis
\[\hat{H}^{\prime}\left|\psi_{1}^{\prime} \psi_{2}^{\prime}\right\rangle=\left[E_{1}+E_{2}+E_{1,2}\right]\left|\psi_{1}^{\prime} \psi_{2}^{\prime}\right\rangle \label{eq21.2.2.06}\]
Here \(\psi_{i}^{\prime}\) are the new subsystem wavefunctions which we added a prime to in order to indicate a change of projection onto the original basis set \(\left(C_{n, m}\right)\), and \(E_{1,2}=\left\langle\psi_{1}^{\prime} \psi_{2}^{\prime}\left|\hat{V}_{1,2}\right| \psi_{1}^{\prime} \psi_{2}^{\prime}\right\rangle\) is the energy change of the full system that results from turning on the interaction \(\hat{V}_{1,2}\). If we operate from the left with the wavefunction for space \(2,\left\langle\psi_{2}^{\prime}\right|\), we obtain
\[\left[H_{1}+\left\langle\psi_{2}^{\prime}\left|\hat{V}_{1,2}\right| \psi_{2}^{\prime}\right\rangle\right]\left|\psi_{1}^{\prime}\right\rangle=\left[E_{1}+\delta E_{1}\right]\left|\psi_{1}^{\prime}\right\rangle \label{eq21.2.2.07}\]
\(\delta E_{1}\) is the energy shift in system \(1\) due to the interaction with system 2. Notice in this process we have ended up with a TISE for system \(1\) with corrections due to its interaction with system 2. This correction is obtained by integrating the interaction potential over the full \(\psi_{2}\) space. We have thus retained our explicit description of system \(1\) but have only retained information on the average influence of the particles and fields in system \(2\) in the interaction expectation value, and thus this is considered a mean-field approximation.
To summarize, we have made multiple approximations in this strategy but gained an important advantage. We do not allow exchange of particles between subsystems, which assures us that the product states remain a complete basis before and after the interaction, and the interactions must be weak \(\left(E_{1} \ll \delta E_{1}\right)\) so that the two subsystems to not become significantly entangled. The key advantage is that we have found a strategy to study the structure, interactions, and dynamics of a molecule or other quantum system interacting with its environment, which in principle assume arbitrary complexity. Approximations that allow us to cast the problem in this form are important in many aspects of molecular quantum mechanics, and they form the basis for perturbation theory.
This page titled 21.2.2: Separation of Subsystems was last modified on Thu, 04 Jun 2026 01:16:37 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.