8.3.2 Customization

Q-Chem offers a number of standard and special customization features. One of the most important is that of supplying additional diffuse functions. Diffuse functions are often important for studying anions and excited states of molecules, and for the latter several sets of additional diffuse functions may be required. These extra diffuse functions can be generated from the standard diffuse functions by applying a scaling factor to the exponent of the original diffuse function. This yields a geometric series of exponents for the diffuse functions which includes the original standard functions along with more diffuse functions.

When using very large basis sets, especially those that include many diffuse functions, or if the system being studied is very large, linear dependence in the basis set may arise. This results in an over-complete description of the space spanned by the basis functions, and can cause a loss of uniqueness in the molecular orbital coefficients. Consequently, the SCF may be slow to converge or behave erratically. Q-Chem will automatically check for linear dependence in the basis set, and will project out the near-degeneracies, if they exist. This will result in there being slightly fewer molecular orbitals than there are basis functions. Q-Chem checks for linear dependence by considering the eigenvalues of the overlap matrix. Very small eigenvalues are an indication that the basis set is close to being linearly dependent. The size at which the eigenvalues are considered to be too small is governed by the $rem variable BASIS_LIN_DEP_THRESH. By default this is set to 6, corresponding to a threshold of ${10}^{-6}$. This has been found to give reliable results, however SCF convergence failure (especially for large molecules or those with highly diffuse basis sets) may be a symptom of linear dependencies. The smallest overlap matrix eigenvalue is printed in the Q-Chem output file, and usually when this number goes below ${10}^{-5}$, numerical issues caused by basis function linear dependence may occur and the SCF calculation may not give reasonable solutions. If the smallest overlap matrix eigenvalue is less than the square root of the integral threshold, a warning message urging to tighten the integral threshold (e.g., setting THRESH = 14) will be printed out. In any case, when a linear dependence issue is suspected, tightening the integral threshold should be tried first. Especially for larger molecules in basis sets that contain diffuse functions, tightening the integral threshold sometimes has the nonintuitive effect of decreasing the time-to-solution, by significantly reducing the number of SCF cycles at only a modest per-cycle increase in cost.