Opening the book…
Linear Combinations of Atomic Orbitals and H₂⁺
Let us remove one of the electrons to make \(\ce{H2^+}\). The Hamiltonian for \(\ce{H2^+}\) is
\[\hat{H} = -\frac{1}{2}\nabla^2 - \frac{1}{r_{A}} - \frac{1}{r_{B}} + E_N(R) \nonumber \]
(recall that \(E_N(R)\) is just \(1/R\)). As we said above, we want to find wavefunctions \(\psi(\vec{r};R)\) (\(\vec{r}\) is the electron coordinate vector– a variable– and \(R\) is the internuclear separation held constant) that are eigenfunctions of the operator. The wavefunctions are also called electron orbitals (just like the orbitals for an electron in an atom); the eigenvalues associated with each wavefunction are the energies that an electron would have if it occupied each orbital.
In the particular case of \(\ce{H2^+}\), it is possible solve the Schrödinger equation exactly (under the BO approximation, at least), but the solutions obtained would only be useful for \(\ce{H2^+}\), which does nothing to help us with the general problem of polyatomic molecules. Instead, we can learn something more generally useful by solving the equation approximately using the variational method. As you recall, the variational method involves using a trial function \(\ket{\psi^\prime}\) that includes coefficients that can be optimized as a guess for the true wavefunction. The variational principle says that for the lowest-energy state, the expectation value of the energy of the trial function (\(\bra{\psi^\prime}\hat{H}\ket{\psi^\prime}\)) will overestimate the true ground state energy unless the trial function is exactly the ground state wavefunction. This means that by tweaking the trial function to minimize the energy, we can obtain better and better approximations to the true solution. It also means that the better we do at guessing a trial function (i.e., the closer the form of the trial function is to the true answer), the better our approximation.
Since the solutions to the hydrogen atom problem are known, it makes sense to try using a linear combination of hydrogen atom \(1s\) orbitals for our trial function.
\[\ket{\psi} = c_1\ket{1s_A} + c_2\ket{1s_B}. \nonumber \]
In this equation, \(\ket{1s_A}\) is a hydrogen \(1s\) orbital centered on hydrogen nucleus A, \(\ket{1s_B}\) is another \(1s\) orbital centered on atom B, and \(c_1\) and \(c_2\) are weighting coefficients that we can adjust to change the shape of the molecular orbital \(\ket{\psi}\) using the variational method. Because is symmetric– i.e., each atom and orbital are the same on each side of the molecule– we can guess that the best guess of the molecular wavefunction will have \(|c_1| = |c_2|\). If we let \(c_1 = c_2 = 1\) and accept that the resultant function is not normalized (it will be easy to do this later), we can make the trial function
\[\ket{\psi_+} = \ket{1s_A} + \ket{1s_B}. \nonumber \]
Because we don’t necessarily know that \(c_1\) and \(c_2\) are both positive, we’ll also try a second trial function
\[\ket{\psi_-} = \ket{1s_A} - \ket{1s_B}. \nonumber \]
Now we need to calculate the variational energy of our trial functions \(\ket{\psi_+}\) and \(\ket{\psi_-}\). We’ll start with \(\ket{\psi_+}\), using the definition for variational energy:
\[E_+ = \frac{\bra{\psi_+}\hat{H}\ket{\psi_+}}{\langle \psi_+ | \psi_+ \rangle}. \label{chap:mot:eq:var} \]
Step 1: The Denominator
We can expand the denominator:
\[\braket{\psi_+}{\psi_+} = \braket{1s_A}{1s_A} + \braket{1s_B}{1s_B} + \braket{1s_A}{1s_B} + \braket{1s_B}{1s_A}. \nonumber \]
The individual H-atom wavefunctions are normalized, so the first and second terms (\(\braket{1s_A}{1s_A}\) and \(\braket{1s_B}{1s_B}\)) are equal to 1. The other two terms represent the overlap integrals of hydrogen \(1s\) wavefunctions located at different positions separated by \(R\) (Figure 1.2); we can denote them \(S(R)\) and \(S^*(R)\) because they are complex conjugates of one another.
Because the hydrogen \(1s\) functions are real, \(\braket{1s_A}{1s_B} = \braket{1s_B}{1s_A}\), so \(S(R) = S^*(R)\). If we multiply the values of the two \(1s\) functions and integrate them, we find that the result is \(0 \leq S(R) \leq 1\). This makes some physical sense: if \(R=0\) then the orbitals overlap completely, giving an overlap of 1 because the \(1s\) orbital function is normalized. As \(R\) increases, the centers of the \(1s\) orbitals move farther and farther apart, so that their overlap decreases with increasing \(R\): \(S(R) \to 0\) as \(R \to \infty\). In the end, the denominator evaluates to:
\[\braket{\psi_+}{\psi_+} = 2(1+S(R)). \label{chap:mot:eq:overlap} \]
Step 2: The Numerator
Now let’s expand the numerator of Equation \ref{chap:mot:eq:var}:
\[\begin{align}
\left\langle\psi_{+}\right| \hat{\mathbf{H}}\left|\psi_{+}\right\rangle= & \left\langle\psi_{+}\right| \hat{\mathbf{H}}\left|1 s_A\right\rangle+\left\langle\psi_{+}\right| \hat{\mathbf{H}}\left|1 s_B\right\rangle \label{2.10}\\[4pt]
= & \left\langle\psi_{+}\left|\left[-\frac{1}{2} \nabla^2-\frac{1}{r_A}\right]-\frac{1}{r_B}+\frac{1}{R}\right|1 s_A\right\rangle + \left\langle\psi_{+}\left|\left[-\frac{1}{2} \nabla^2-\frac{1}{r_B}\right]-\frac{1}{r_A}+\frac{1}{R}\right|1 s_B\right\rangle \label{2.11} \\[4pt]
= & \left\langle\psi_{+}\left|\left[-\frac{1}{2} \nabla^2-\frac{1}{r_A}\right]-\frac{1}{r_B}\right|1 s_A\right\rangle+\left\langle\psi_{+}\left|\left[-\frac{1}{2} \nabla^2-\frac{1}{r_B}\right]-\frac{1}{r_A}\right|1 s_B\right\rangle + \left\langle\psi_{+}\left| \frac{1}{R}\right|\psi_{+}\right\rangle \label{2.12} \\[4pt]
= & \left\langle\psi_{+}\right| \hat{\mathbf{H}}_0^A+\hat{\mathbf{V}}^A\left|1 s_A\right\rangle+\left\langle\psi_{+}\right| \hat{\mathbf{H}}_0^B+\hat{\mathbf{V}}^B\left|1 s_B\right\rangle+E_N(R)\left\langle\psi_{+} \mid \psi_{+}\right\rangle . \label{2.13}
\end{align} \nonumber \]
where the \(\hat{H}_0\) terms are exactly equivalent to hydrogen atom Hamiltonians for atoms A and B as isolated by the square brackets in Equation \ref{2.11}, and the \(V\) terms contain the potential energy from the electron interacting with the other nucleus. The \(E_N(R)\) term contains the internuclear repulsion energy, which does not depend on the electronic coordinates and can be factored out of the integrals. We know that \(\hat{H}_0\ket{1s} = E_{1s}\ket{1s}\) already, so the terms involving \(\hat{H}_0\) become \(E_{1s}\braket{\psi_+}{1s_A} + E_{1s}\braket{\psi_+}{1s_B}\). Because
\[E_{1s}\braket{\psi_+}{1s_A} + E_{1s}\braket{\psi_+}{1s_B} = E_{1s}\bra{\psi_+}\left(\ket{1s_A} + \ket{1s_B}\right) = E_{1s}\braket{\psi_+}{\psi_+}, \nonumber \]
from Equation \ref{chap:mot:eq:overlap}, we can simplify further:
\[E_{1s}\braket{\psi_+}{\psi_+} = 2E_{1s}(1 + S(R)). \nonumber \]
Likewise,
\[E_N(R)\braket{\psi_+}{\psi_+} = 2E_N(R)(1 + S(R)). \nonumber \]
Now let’s revisit Equation \ref{chap:mot:eq:var}, substituting the terms we just solved into the numerator:
\[\begin{align*} \bra{\psi_+}\hat{H}\ket{\psi_+} = & 2E_{1s}(1 + S(R)) \\[4pt] & + \bra{1s_A}\hat{V}^A\ket{1s_A} + \bra{1s_B}\hat{V}^B\ket{1s_B} \\[4pt] & + \bra{1s_B}\hat{V}^A\ket{1s_A} + \bra{1s_A}\hat{V}^B\ket{1s_B} \\[4pt] & + 2E_N(R)(1+S(R)).\end{align*} \]
The terms on the second line represent the average Coulombic interaction between a hydrogen atom and a proton; these are called and are denoted \(J_{AA}\) and \(J_{BB}\). On the third line, the terms are called (denoted \(K_{AB}\) and \(K_{BA}\)); they have no classical analog. They represent a quantum interaction of an electron in the the \(1s\) wavefunction on one atom with the other nucleus as if the electron were somehow also swapped into the other orbital (hence, “exchange integral”). Because the H atom \(is\) wavefunctions are real, in this case \(J_{AA} = J_{BB}\) and \(K_{AB} = K_{BA}\) (in general \(J_{mn} = J_{nm}^*\), and likewise for \(K_{mn}\)), so we will drop the subscripts for simplicity. Later, we will have to keep track of which orbitals are involved in each Coulomb and exchange integral, so we will use subscripts in the future.
In the end the numerator of Equation \ref{chap:mot:eq:var} is
\[\bra{\psi_+}\hat{H}\ket{\psi_+} = 2E_{1s}(1+S(R)) + 2J + 2K + E_N(R), \nonumber \]
and the variational energy is:
\[E_+ = \frac{\bra{\psi_+}\hat{H}\ket{\psi_+}}{\braket{\psi_+}{\psi_+}} = E_{1s} + \frac{J + K}{1 + S(R)} + E_N(R). \nonumber \]
If we repeat the same procedure for our other trial function, we find that \(\ket{\psi_-}\) has energy
\[E_- = \frac{\bra{\psi_-}\hat{H}\ket{\psi_-}}{\braket{\psi_-}{\psi_-}} = E_{1s} + \frac{J - K}{1 - S(R)} + E_N(R). \nonumber \]
Linear Combination of Atomic Orbitals-Molecular Orbitals (LCAO-MO)
The two wavefunctions we have used are molecular orbitals: they describe the states that are accessible to an electron in a molecule much in the same way that atomic orbitals do in atoms. Specifically, these are so-called LCAO-MOs: molecular orbitals built from linear combinations of atomic orbitals, as we could have chosen any mathematical function as our trial function rather than a weighted sum of atomic orbitals. The energies we’ve calculated are the energies of the molecular orbitals and because they were calculated under the Born-Oppenheimer approximation, they depend on the nuclear coordinates parametrically. In other words, if we move the nuclei to new positions and repeat the calculation, we will obtain a different value for the orbital energies. We can plot the energy of each orbital as a function of the nuclear coordinate; if there is a local minimum, then the orbital is a bonding orbital that “holds” the nuclei together. Otherwise, if there is no well, then the nuclei are not bound together, and the orbital is an antibonding orbital.
Defining \(\Delta E_\pm = E_\pm - E_{1s}\), we can compare the energy of the molecular orbitals with the energy of a proton and hydrogen atom separated infinitely (Figure Figure 2.3, left). We see that \(\ket{\psi_+}\) actually has lower energy than \(E_{1s}\) over a certain region and is a bound potential, while \(\ket{\psi_-}\) has greater energy and is not bound. Breaking down the contributions from \(J\) and \(K\) to \(E_+\) (Figure 2.3, right), we see that it is the exchange term \(K\) that gives rise to the stability of a bond: it’s purely quantum-mechanical! The curves \(\Delta E_+\) and \(\Delta E_-\) are examples of potential energy curves that result from making the Born-Oppenheimer approximation. Each curve represents the potential energy that the nuclei would experience from an electron placed into the corresponding orbital, depending on how far apart the nuclei are. We can therefore see that \(\ket{\psi_+}\) a bonding orbital, as it has a local minimum where the nuclei are held together, and \(\ket{\psi_-}\) an antibonding orbital, where no such stable minimum exists. When there are multiple electrons, MO theory says that the total electronic potential energy is the sum of the potential energies of all of the electrons in their respective orbitals.


When a molecule contains more than 2 nuclei, the dimensionality of their potential energy curves is greater than 1, the hypersurface of electronic energy at all nuclear positions is called a potential energy surface. These are generated by repeatedly calculating the energy of the electronic wavefunction at many nuclear positions. The potential energy surface is typically used as the potential energy term in the nuclear Hamiltonian for calculating the nuclear geometry (the equilibrium geometry is the minimum of the PES) and the vibrational energy (related to the derivative of the PES near a minimum).
In a PES, there may be multiple minima that correspond to different arrangements of the nuclei that lead to stable bonding patterns: these are the different conformers and structural isomers. For instance, the PES for \(\ce{C3,H8,O}\) will have two deep minima that correspond to 1-propanol and isopropanol, and near each minimum there are several other local minima that correspond to the different conformations of the methyl groups and directions of the \(\ce{OH}\). We will discuss the impact of potential energy surfaces on spectroscopy later.
Returning to the problem at hand, all that remains is to normalize the wavefunctions \(\ket{\psi_+}\) and \(\ket{\psi_-}\). We already calculated \(\braket{\psi_+}{\psi_+} = 2(1+S(R))\) and \(\braket{\psi_-}{\psi_-} = 2(1-S(R))\), so the normalized wavefunctions, which we’ll call \(\ket{\psi_b}\) and \(\ket{\psi_a}\) for bonding and antibonding, respectively, are:
\[\begin{aligned} \ket{\psi_b} & = \sqrt{\frac{1}{\braket{\psi_+}{\psi_+}}}\ket{\psi_+} \\[4pt] &= \sqrt{\frac{1}{2(1+S(R))}}\left(\ket{1s_A} + \ket{1s_B}\right) \\[4pt] \ket{\psi_a} & = \sqrt{\frac{1}{\braket{\psi_-}{\psi_-}}}\ket{\psi_-} \\[4pt] &= \sqrt{\frac{1}{2(1-S(R))}}\left(\ket{1s_A} - \ket{1s_B}\right)\end{aligned} \nonumber \]
They are drawn in Figure 2.3.
We’ve obtained two possible orbitals for and calculated their energies. Are they the “true” orbitals and “true” energies for ? Most likely not. We chose a very simple trial function, just a linear combination of two atomic orbitals. We could have, for instance, chosen to include both the \(1s\) and \(2s\) orbitals from each hydrogen atom. And if this were a polyatomic molecule, then we would need even more orbitals from the other atoms as well, leading to a real mess if we are just making random guesses for trial functions and calculating their energies! A more systematic approach is needed.
LCAOs and the Hamiltonian Matrix
Provided our trial functions remain linear combinations of functions, the Schrödinger equation can be converted into matrix form, and standard, scalable techniques in linear algebra can be used to find the solutions. We’ll illustrate this matrix formalism by first repeating the calculation of , then we will look at larger systems. Again we choose \(\ket{\psi_n} = c_{1n}\ket{1s_A} + c_{2n}\ket{1s_B}\) as our trial function, letting \(n\) be an index to keep track of the two solutions (e.g., \(\ket{\psi_1} = c_{11}\ket{1s_A} + c_{21}\ket{1s_B}\)). In matrix formalism, the trial function forms a vector space whose unit vectors correspond to our individual atomic basis functions. In this case, the vector space is two-dimensional:
\[\ket{1s_A} = \begin{pmatrix} 1 \\[4pt] 0 \end{pmatrix} \quad \text{and}\quad \ket{1s_B} = \begin{pmatrix} 0 \\[4pt] 1 \end{pmatrix} \nonumber \]
The trial wavefunction \(\ket{\psi}\) can be expressed as in this vector space like this:
\[\ket{\psi_n} = \begin{pmatrix} c_{1n} \\[4pt] c_{2n} \end{pmatrix}\quad \text{and}\quad \bra{\psi_n} = \begin{pmatrix} c_{1n}^* & c_{2n}^* \end{pmatrix} \nonumber \]
The Hamiltonian operator \(\hat{H}\) becomes a \(2 \times 2\) matrix in this vector space:
\[\begin{aligned} \hat{H} & = \left( \begin{array}{cc} \bra{1s_A}\hat{H}\ket{1s_A} & \bra{1s_A}\hat{H}\ket{1s_B} \\[4pt] \bra{1s_B}\hat{H}\ket{1s_A} & \bra{1s_B}\hat{H}\ket{1s_B} \end{array} \right) \nonumber \\[4pt] &= \left( \begin{array}{cc} \bra{1s_A}\hat{H}_0^A + \hat{V}^A\ket{1s_A} & \bra{1s_A}\hat{H}_0^B + \hat{V}^B\ket{1s_B} \nonumber \\[4pt] \bra{1s_B}\hat{H}_0^A + \hat{V}^A\ket{1s_A} & \bra{1s_B}\hat{H}_0^B + \hat{V}^B\ket{1s_B} \end{array} \right) \\[4pt] &= \left( \begin{array}{cc} E_{1s} + J & E_{1s}S(R) + K \\[4pt] E_{1s}S(R) + K & E_{1s} + J \end{array} \right) \label{chap:mot:eq:hmatrix} \end{aligned} \]
Here, we have removed the constant nuclear repulsion term \(E_N\) from the matrix (it is always possible to add or subtract a constant amount of energy from a Hamiltonian, as we can choose to define the zero of our energy scale). After solving this matrix equation and calculating the electronic energies, we will add \(E_N\) to the final electronic energy. You can verify that this is an equivalent formulation by calculating \(\bra{\psi_1}\hat{H}\ket{\psi_1}\):
\[\begin{aligned} \bra{\psi_1}\hat{H}\ket{\psi_1} & = \begin{pmatrix} c_{11}^* & c_{21}^* \end{pmatrix} \begin{pmatrix} \bra{1s_A}\hat{H}\ket{1s_A} & \bra{1s_A}\hat{H}\ket{1s_B} \\[4pt] \bra{1s_B}\hat{H}\ket{1s_A} & \bra{1s_B}\hat{H}\ket{1s_B} \end{pmatrix}\begin{pmatrix} c_{11} \\[4pt] c_{21} \end{pmatrix} \nonumber \\[4pt] & = c_{11}^2\bra{1s_A}\hat{H}\ket{1s_A} + c_{11}^*c_{21}\bra{1s_A}\hat{H}\ket{1s_B} + c_{21}^*c_{11}\bra{1s_B}\hat{H}\ket{1s_A} + c_{21}^2\bra{1s_B}\hat{H}\ket{1s_B}.\end{aligned} \nonumber \]
The goal is to solve the Schrödinger equation \(\hat{H}\ket{\psi_n} = E_n\ket{\psi_n}\) for all \(n\), which is expressed in matrix formalism as a linear system of equations:
\[\begin{pmatrix} H_{11} & H_{12} \\[4pt] H_{21} & H_{22} \end{pmatrix} \begin{pmatrix} c_{11} & c_{12} \\[4pt] c_{21} & c_{22} \end{pmatrix} = \begin{pmatrix} \braket{1s_A}{1s_A} & \braket{1s_A}{1s_B} \\[4pt] \braket{1s_A}{1s_B} & \braket{1s_B}{1s_B} \end{pmatrix}\begin{pmatrix} c_{11} & c_{12} \\[4pt] c_{21} & c_{22} \end{pmatrix}\begin{pmatrix} E_1 & 0 \\[4pt] 0 & E_2 \end{pmatrix}. \label{chap:mot:eq:coefeq} \]
We can rewrite this equation as:
\[\begin{pmatrix} c_{11} & c_{12} \\[4pt] c_{21} & c_{22} \end{pmatrix}^{-1} \begin{pmatrix} \braket{1s_A}{1s_A} & \braket{1s_A}{1s_B} \\[4pt] \braket{1s_A}{1s_B} & \braket{1s_B}{1s_B} \end{pmatrix}^{-1} \begin{pmatrix} H_{11} & H_{12} \\[4pt] H_{21} & H_{22} \end{pmatrix} \begin{pmatrix} c_{11} & c_{12} \\[4pt] c_{21} & c_{22} \end{pmatrix} = \begin{pmatrix} E_1 & 0 \\[4pt] 0 & E_2 \end{pmatrix}. \nonumber \]
In other words, there should exist some set of coefficients that can transform the Hamiltonian matrix into a diagonal form that contains only the energies of each wavefunction. Each column vector in the coefficient matrix corresponds to the weighting coefficients for one of the LCAO-MOs. Mathematically speaking, we wish to find the eigenvalues and eigenvectors of the Hamiltonian matrix (an eigenvector is a vector containing the coefficients \(c_{ij}\) that make up one of the columns of the coefficient matrix that solves Equation \ref{chap:mot:eq:coefeq}). Finding the eigenvalues and eigenvectors of a matrix is a well-known problem in linear algebra; it is also called “diagonalizing” a matrix. This particular case is a “generalized eigenvalue problem” because the basis functions we are using are non-orthogonal (i.e., the two \(1s\) orbitals have nonzero overlap integrals with each other): the overlap matrix \(\hat{S}\) is not equal to the identity matrix.
Writing out Equation \ref{chap:mot:eq:coefeq} gives a system of linear equations, and linear algebra tells us that a nontrivial solution exists only if the eigenvalues satisfy the determinantal equation:
\[\begin{aligned} \left| \begin{array}{cc} E_{1s} + J - \braket{1s_A}{1s_A}\lambda & E_{1s}S(R) + K - \braket{1s_A}{1s_B}\lambda \\[4pt] E_{1s}S(R) + K - \braket{1s_B}{1s_A}\lambda & E_{1s} + J - \braket{1s_B}{1s_B}\lambda \end{array} \right| && \nonumber \\[4pt] = \left| \begin{array}{cc} E_{1s} + J - \lambda & E_{1s}S(R) + K - S(R)\lambda \\[4pt] E_{1s}S(R) + K - S(R)\lambda & E_{1s} + J - \lambda \end{array} \right| & = 0 \label{chap:mot:eq:deteq}\end{aligned} \]
where \(\lambda\) is a dummy variable that represents all eigenvalues. Note that if the basis functions we were using were orthogonal, the off-diagonal elements of the determinant would not contain \(\lambda\) because the off-diagonal terms in the overlap matrix would be 0.
The determinantal equation is a polynomial equation in \(\lambda\); in the \(2\times 2\) case it is a quadratic equation that gives the two eigenvalues
\[\lambda_\pm = E_{1s} + \frac{J \pm K}{1 \pm S(R)}. \nonumber \]
Plugging in these values for \(\lambda\) back into the two equations in Equation \ref{chap:mot:eq:coefeq} as \(E_1\) and \(E_2\) and solving, we find that \(c_{11} = c_{21}\) and \(c_{12} = -c_{22}\). To obtain the actual values of the coefficients, we normalize them:
\[\begin{aligned} 1 & = \braket{\psi_1}{\psi_1} \\[4pt] & = c_{11}^2(\braket{1s_A}{1s_A} + \braket{1s_B}{1s_B} + \braket{1s_A}{1s_B} + \braket{1s_B}{1s_A}) \nonumber \\[4pt] 1 & = c_{11}^2(2+2S(R)) \nonumber \\[4pt] c_{11} = c_{21} & = \sqrt{\frac{1}{2(1+S(R))}}\end{aligned} \nonumber \]
and the same for the other wavefunction:
\[\begin{aligned} 1 & = \braket{\psi_2}{\psi_2} \\[4pt] &= c_{12}^2(\braket{1s_A}{1s_A} + \braket{1s_B}{1s_B} - \braket{1s_A}{1s_B} - \braket{1s_B}{1s_A}) \nonumber \\[4pt] 1 & = c_{12}^2(2-2S(R)) \nonumber \\[4pt] c_{12} = -c_{22} & = \sqrt{\frac{1}{2(1-S(R))}}\end{aligned} \nonumber \]
We find therefore that the matrix form gives exactly the same answer as before (after adding back in the nuclear repulsion \(E_N\)):
\[\begin{aligned} \ket{\psi_1} = \sqrt{\frac{1}{2(1+S(R))}}(\ket{1s_A} + \ket{1s_B}), & \quad & E_1 = E_{1s} + \frac{J+K}{1+S(R)} + E_N \\[4pt] \ket{\psi_2} = \sqrt{\frac{1}{2(1-S(R))}}(\ket{1s_A} - \ket{1s_B}), & \quad & E_2 = E_{1s} + \frac{J-K}{1-S(R)} + E_N\end{aligned} \nonumber \]
Most often, when doing electronic structure calculations, the constant nuclear repulsion energy (\(E_N\)) is excluded from the Hamiltonian when creating the electronic Hamiltonian matrix. After calculating the total electronic energy (more on this later), the nuclear repulsion energy is added back in.
It is easy to extend this matrix formalism to larger systems. In the previous section, we used a simple linear combination of H-atom \(1s\) orbitals as a trial finction, but instead it is possible to choose, for example:
\[\ket{\psi_n} = c_{1n}\ket{1s_A} + c_{2n}\ket{2s_A} + c_{3n}\ket{1s_B} + c_{4n}\ket{2s_B} \nonumber \]
With this choice of a trial function, \(\hat{H}\) has the general form seen in Table 2.1, and solving the eigenvalue equation would give four eigenvalues and four eigenfunctions. This is one of the properties of MO theory: the number of orbitals that come out of a MO calculation is equal to the number of orbitals put in. While the \(4\times4\) matrix would be tedious to solve by hand, numerous efficient algorithms are available for matrix diagonalization on computers. We will discuss computational chemistry in more detail separately.
| \(\hat{H}\) | \(\ket{1s_A}\) | \(\ket{2s_A}\) | \(\ket{1s_B}\) | \(\ket{2s_B}\) |
|---|---|---|---|---|
| \(\ket{1s_A}\) |
Atom A 1s Energy |
Atom A J |
Atom B ES(R) |
Atom B K |
| \(\ket{2s_A}\) |
Atom A J |
Atom A 2s Energy |
Atom B | Atom B |
| \(\ket{1s_B}\) |
Atom A ES(R) |
Atom A K |
Atom B 1s Energy |
Atom B J |
| \(\ket{2s_B}\) |
Atom A K |
Atom A ES(R) |
Atom B J |
Atom B 2s Energy |
Still working with \(\ce{H2^+}\), we can look qualitatively at what happens when we include the \(\ket{2p}\) H-atom wavefunctions for each atom as well (10 orbitals total). We find that the linear combinations that come out look approximately like (in order of increasing energy, and taking only the most dominant terms):
\[\begin{aligned} \ket{\psi_1} \propto & \ket{1s_A} + \ket{1s_B} \\[4pt] \ket{\psi_2} \propto & \ket{1s_A} - \ket{1s_B} \\[4pt] \ket{\psi_3} \propto & \ket{2s_A} + \ket{2s_B} \\[4pt] \ket{\psi_4} \propto & \ket{2s_A} - \ket{2s_B} \\[4pt] \ket{\psi_5} \propto & \ket{2p_{z,A}} - \ket{2p_{z,B}} \\[4pt] \ket{\psi_6} \propto & \ket{2p_{x,A}} + \ket{2p_{x,B}} \\[4pt] \ket{\psi_7} \propto & \ket{2p_{y,A}} + \ket{2p_{y,B}} \\[4pt] \ket{\psi_8} \propto & \ket{2p_{x,A}} - \ket{2p_{x,B}} \\[4pt] \ket{\psi_9} \propto & \ket{2p_{y,A}} - \ket{2p_{y,B}} \\[4pt] \ket{\psi_{10}} \propto & \ket{2p_{z,A}} + \ket{2p_{z,B}}\end{aligned} \nonumber \]
These orbitals are sketched in Figure 2.5, showing the familiar \(\pi\) and \(\pi^*\) orbitals. The specific orbitals that we’ve shown here are general for homonuclear diatomic species, though some of the energy orderings change depending on the specific nuclei involved.
This page titled 10.2: Origins of MO Theory - Linear Combinations of Atomic Orbitals was last modified on Mon, 28 Sep 2026 15:44:31 GMT and is shared under a CC BY 4.0 license and was authored, remixed, and/or curated by Kyle Crabtree.
The book could not be loaded just now.