(289a) Development and Comparative Performance of Density Functionals That Correctly and Efficiently Treat Long-Range Correlations

Authors: 
Head-Gordon, M., University of California - Berkeley



Conventional density functionals (DF) of the generalized gradient approximation (GGA) and hybrid type have proven to be enormously useful for the computational modeling of a wide variety of catalytic processes.  Examples include, but are not limited to, catalysis on bulk surfaces, such as transition metals, on molecular sites as in homogeneous catalysis, and in mesoporous solids such as zeolites.  Nevertheless, it has become recognized over the past five years or so that the performance of standard GGA and hybrid functionals are quite inadequate to describe binding that is controlled at least in part by long-range dispersion forces. 

Therefore a key frontier in functional development has become the design of functionals that can correctly describe such interactions.  Three main approaches have emerged over recent years.  The first one is the addition of a damped dispersion potential (DFT-D methods), which is computationally least expensive, but is not strictly an approach that lies within density functional theory.  The second one is the development of non-local density-density correlation functionals that describe van der Waals interactions (VdW-DF methods), such as the approaches of Langreth et al, and Vydrov and Van Voorhis.  The third class are so-called double hybrid (DH) functionals that incorporate a fraction of non-local second order correlation (PT2), or related approaches that use the random phase approximation (RPA).

In this talk, I will describe some recent developments from my group regarding both the development and the testing of new density functionals that can correctly and reasonably efficiently describe long-range correlations.  I will focus on two examples.  The first is the extension of our widely used range-separated density functionals of the omega-B97 family to include VdW-DF contributions.  This design, training and testing of this new functional will be described in detail.  The testing will include comparison against other new generation functionals including DFT-D examples, hybrid meta GGA examples, and double hybrids.  The second example is work on a new class of double hybrid functionals that include orbital optimization.  I will argue why orbital optimization is essential, and present examples of the performance of at least one new functional that is designed from the outset to be fully orbital optimized.  If research progress permits, I will include examples of the application of these functionals to catalytically relevant examples.

Topics: