1: \begin{abstract}

2: We present a new method for the optimization of large configuration

3: interaction (CI) expansions in the quantum Monte Carlo (QMC)

4: framework.  The central idea here is to replace the non-orthogonal

5: variational optimization of CI coefficients performed in usual QMC

6: calculations by an orthogonal non-Hermitian optimization thanks to the

7: so-called transcorrelated (TC) framework, the two methods yielding the

8: same results in the limit of a complete basis set.

9: By rewriting the TC equations as an effective self-consistent Hermitian

10: problem, our approach requires the sampling of a single quantity per Slater

11: determinant, leading to minimal memory requirements in the QMC code.

12: Using analytical quantities obtained from both the TC framework and

13: the usual CI-type calculations, we also propose improved estimators which

14: reduce the statistical fluctuations of the sampled quantities by more

15: than an order of magnitude.

16: We demonstrate the efficiency of this method on wave functions containing

17: $10^5-10^6$ Slater determinants, using effective core potentials or

18: all-electron calculations. In all the cases, a sub-milliHartree convergence is

19: reached within only two or three iterations of optimization.

20: \end{abstract}

21: