I noticed that the original work on Least-Square Tensor Hypercontraction ( LS-THC: http://scitation.aip.org/content/aip/journal/jcp/137/22/10.1063/1.4768233 ) was implemented in PSI4, and the core of the code appears to be in the source in libthce. However, it looks like none of the current public branch other than fisapt and some tangential usage in soscf actually uses this code, and there looks to be no direct interface to it through python either. Is there any known way of easily extracting the matrix Z and the weighted collocation factors X denoted in the original paper? I wouldn’t need anything as fancy as running a full calculation with it, but it would be extremely helpful for a small numerical experiment I wanted to run. Any help or comments someone can offer about the future of the THC feature set would be greatly appreciated!
THC is not exactly publicly released and may not work in its current form. For more information you would need to email Rob Parrish directly about this.