Contents

2 What’s new since sapt2002

2.1 New in revision sapt2012.2

• This is a bugfix release. No new features were added.
• Fixed gfortran compilation problem caused by the use of entry statements in integer*8 functions.
• Fixed several gfortran runtime errors.
• Restored compatibility with Fortran 77 compilers (such as g77).

2.2 New in revision sapt2012.1

• SAPT(CC): interaction energy contributions from relaxed monomer CCSD density matrices and their cumulants [24–31]. This functionality requires molpro and the patch supplied here.
• Relaxed third-order induction correction E_ind,resp⁽³⁰⁾ [35].
• Removed limitation of 1024 orbitals (current limitation is 65535). The option has to be compiled in (not enabled by default).
• Added code for calculating CKS exchange-dispersion (both non-DF and DF versions) with amplitudes obtained from CKS propagators in addition to version scaled from uncoupled versions. This correction, together with analogous exchange-induction corrections, is enabled by default when SAPT(DFT) calculations are requested. Old versions are retained for compatibility.
• orca interface for DF-SAPT(DFT).
• Gaussian09 interface.
• Gradient-regulated asymptotic correction (GRAC) [36] for the dalton program.
• Optional support for dimer.cnf in Angstroms.
• Lowered memory requirements in DF-SAPT(DFT) transformation. Added automatic automatic calculation of required memory in DF-tran.
• Added trailing zeroes to the name of output of Runlot. Helps with sorting.
• Fixed timing routines with Intel Fortran.
• Renamed sapt program to sapt.x enabling correct use of SAPT2012 on filesystems that are not case-sensitive (e.g., Mac OS X).
• Fix for DF-SAPT(DFT) going into infinite loop in some cases when compiled with gfortran.
• Fixed DF-SAPT(DFT) code in some calculations with large intermonomer distances.
• Fixed SAPT(DFT) geometry code with some versions of awk.
• Fixed integer overflows for large-scale SAPT(DFT) calculations.

2.3 New in revision sapt2008.2

• This is a bugfix release. No new features were added.
• Fixed crashing of very large (over 1000 shells in auxiliary basis sets) SAPT(DFT) jobs.
• Fixed crashing of DF-SAPT(DFT) for calculations with large intermonomer distances.
• Removed incorrect “integral unsupported for DF-TRAN” in density-fitting calculations for third-order induction calculations that prevented such calculations with DF-SAPT(DFT).
• Fixed a bug in the MOLPRO interface to prevent a crash when molpro is compiled with ifort.
• Fixed a compilation failure for TARGET=gfortran and GAMSI8=’YES’.
• Fixed a bug in the e1dcbs program that resulted in a crash of ccsddSAPT when GAMSI8=’YES’.
• Fixed SAPT script to correctly handle energies printed by molpro.
• Small fixes in the examples and scripts for running the examples.
• Attached new dalton patch with small compilation fixes and a fix to correctly handle ECP dalton runs.

2.4 New in revision sapt2008.1

• Density-fitting SAPT(DFT) [16], optionally with quadruple precision electrostatic component.
• Three-body SAPT(DFT) with or without density fitting [20].
• SAPT(DFT) for open-shell high-spin complexes [21].
• Faster transformation algorithm for non-four-virtual integrals, reusing partially transformed integrals whenever possible. Optimal performance is achieved when twice as much memory as for the regular in-core transformation is available. When additional memory is not available, the new implementation is still faster than the one from sapt2006.1, especially in dimer-centered basis sets.
• The four-virtual diagram in cc can now be evaluated in atomic orbitals, eliminating the need for a four-virtual transformation. The AO algorithm currently works only with atmol1024 and molpro interfaces.
• Faster (, not ) implementation of the correction, backported from parallel SAPT.
• Support for monomer-centered basis sets in the molpro interface.
• The C6H6_H2O_DCBS example replaced by a similar, C6H6_H2O_ADZM one. In the old example, the basis set was nearly linearly dependent which caused problems with the coupled-cluster convergence.
• A new example showing how to run SAPT with effective core potentials (ECPs) using the molpro interface.
• A new example showing how to calculate a relativistic SAPT interaction energy utilizing the second-order Douglas-Kroll-Hess Hamiltonian, with the relevant integrals generated by molpro.
• CKS program can now allocate more than 2 Gwords of memory.
• Fixed a bug resulting in a crash of the sapt.x program in an (extremely unlikely) situation when no transformed integrals of a given type were larger than the threshold so that no integrals of this type were written to disk.
• Fixed a bug causing the ccsdm program to allocate too little memory in some cases.
• Fixed a bug in misc/atmolstuff.F resulting in a compile error under HPUX.
• Fixed a bug in cc/tpdrvn.F resulting in a compile error when setting EXTRADEFS=’-DTWOGIGAMAX’.
• Fixed a bug in the molpro interface which caused SAPT to produce nonsensical results when used with molpro2006.1. Updated example input files for use with molpro2006.1 and molpro2008.1.
• Updated the gamess interface for use with the 2008 release of gamess.
• Corrected a POSIX-noncompliant use of the tail command in the Compall script that resulted in compile errors on some of the platforms.
• Replaced an advanced regular expression in the SAPT script by a simpler one which is compatible with mawk. The previous version was problematic for systems for which mawk was the default awk program.
• Added a preliminary version of the library with common utilities (saptlib/).

2.5 New in sapt2006

• SAPT(DFT) [14] calculations available, with DFT quantities extracted from dalton 2.0 [37].
• Third-order SAPT corrections: E_ind⁽³⁰⁾, E_exch-ind⁽³⁰⁾, E_ind-disp⁽³⁰⁾, E_{exch-ind-disp}⁽³⁰⁾, E_disp⁽³⁰⁾, and E_exch-disp⁽³⁰⁾ [38].
• Optional computation of E_elst⁽¹⁾ from relaxed-monomer CCSD densities. If a pure monomer basis set is used, the densities may be precomputed and reused, after translation and rotation, to compute the whole surface of E_elst⁽¹⁾(CCSD) using direct integrals.
• Optional calculation of the second-order dispersion energy using converged CCD amplitudes [39].
• The E_exch⁽¹⁰⁾(S²) correction reimplemented using a formula valid in both dimer-centered basis set (DCBS) and monomer-centered basis set (MCBS). The implementation of this correction in sapt2002 was valid in DCBS only, and for MCBS a wrong result was printed. This correction does not enter the final SAPT interaction energy, as the formula of infinite order in S is used.
• New, much faster algorithm for the calculation of E_exch-disp⁽²⁰⁾—the most demanding correction at the SAPT0 or SAPT(DFT) level. An out-of-core algorithm utilizing much less memory is also available for this correction.
• New, faster and more memory-efficient CHF routines.
• Frozen core implemented for all standard SAPT corrections [40].
• New interfaces: dalton 2.0, molpro [41], and gaussian03 [42].
• New architectures: AMD64 (Opteron, Athlon64) and Intel EMT64 with g77 (32 and 64 bit) and pgf compilers (32 and 64 bit), as well as HPUX (Itanium).
• DIIS [43] in coupled-cluster calculations. Turned on by default if compiled in.
• More efficient algorithms for the four-virtual diagram in CC calculations.
• Both energy-based and amplitude-based convergence criteria for CC available.
• Smarter scripts for running SAPT.
• Fixed a memory allocation bug on x86 Linux machines with kernel 2.6.x using g77.
• Several small bug fixes and enhancements.

Note that turning DIIS on/off, as well as changing the CC convergence criterion from energy-based to energy- and amplitude-based, affects the converged CC amplitudes as well as the E_exch⁽¹⁾(CCSD) SAPT correction if CC convergence thresholds are not very tight (as it is in the default case). This may lead to slight differences in the value of E_exch⁽¹⁾(CCSD) between sapt2006 and older versions of SAPT.