Top Page > Browsing

OpenMX crashes with Orbital Optimization Date: 2005/05/20 19:31 Name: Michele Gusso   <michele.gusso@brindisi.enea.it> Dear all, I have tried to run Openmx2.3 on Cadmium dimer using only 5s2 and 4d10 valence electrons (so no p electrons) but the program crashes when using the option Orbital Optimization Restricted(Error segmantation fault). Unfortunatly the debugger is not able to give the place of the error probably because I receive the following warning from GDB: shared library handler failed to enable breakpoint. Anyway, after some cheks I realized that the problem is in the routine Initial_CntCoes.c. In the lines 833-840 the array tmp_index2[L0][Mul0][M0] is initialized. But in this case, since there are not p electrons the array is not filled for L0=1 (in fact Mul0=0; Mul0<Spe_Num_Basis[wan][L0]). On the other hand in the successive loop on L0 which start at line 844 we have the instruction p=tmp_index2[L0][0][M0] (line 848) which for L0=1 assigns a value to p which is indefined. Since I don't know the way this code works, I haven't tried to solve this problem. By the way, there is a way to generate with ADAPAK pseudopotential components l for electrons having occupancy=0 (so that to include pseudopotentials for states that are nominally empty in the ground state)? Best regards, Michele Gusso.

Page: [1]

Re: OpenMX crashes with Orbital Optimization ( No.1 ) Date: 2005/05/22 01:31 Name: T.Ozaki Hi, Thank you for your detailed report. We did not assume such a skip of component L in the specification of basis orbitals as you did. In fact, the skip causes a segmentation fault by putting the undefined tmp_index2[L0][0][M0] (=p) as argument into C[p+1][mu] lying in the next line. I have modified this part so that a undefined component L can be skipped. Please replace the subroutine by http://staff.aist.go.jp/t-ozaki/bugfixed/05May22/Initial_CntCoes.c > By the way, there is a way to generate with ADAPAK pseudopotential components l for > electrons having occupancy=0 (so that to include pseudopotentials for states that > are nominally empty in the ground state)? A scheme to generate a pseudopotential of component L with occupancy 0 is not supported in ADPACK. Usually, we slightly promote a partial electron to this state in order to generate the pseudopotential, while often a charged atom is taken into account in this case. Regards, T.Ozaki

Page: [1]