October 2014 - Comp-phys-alps-users

Inconsistent comment in generate_spline_matrix in the DMFT folder
by Kuangshing Chen 22 Jul '15

22 Jul '15

Hi, The function generate_spline_matrix in fouriertransform.C in the dmft folder shows the comments as follows: // A is the matrix whose inverse defines spline_matrix // // 6 6 // 1 4 1 // 1 4 1 // A = ... // // 1 4 1 // -2 0 0 2 However, the following codes, dense_matrix A = 4*dt/6.*boost::numeric::ublas::identity_matrix<double>(Np1); for (int i=1; i<Np1-1; i++) { A(i,i-1) = dt/6.; A(i,i+1) = dt/6.; } A(0,0) = 1.; A(0, Np1-1) = 1.; A(Np1-1, 0) = -2.*dt/6.; A(Np1-1, 1) = -1.*dt/6.; A(Np1-1, Np1-2) = 1*dt/6.; A(Np1-1, Np1-1) = 2*dt/6.; the red lines will make the matrix looks like // 6 6 // 1 4 1 // 1 4 1 // A = ... // // 1 4 1 // -2 *-1* *1* 2 Is the comment correct or the code correct? Any reference for this algorithm? Thank you, Kuang-Shing Chen

5 25

cluster algorithm for classical spin models
by Rachele Nerattini 22 Oct '14

22 Oct '14

Dear ALPS community, I am writing for a question on the algorithm for classical Monte Carlo simulation of spin models. From the simulations I performed on classical O(2), O(3) and O(4) models it turned out that the autocorrelation time is quite large, let's say of about 80 o 90, for the energy density observable. I just would like to understand better if there is a particular reason for such a big tau and which kind of algorithm is really used for the simulation: Swendsen-Wang , Wolff single cluster or even something else? This is also because the term ”sweep” is quite unusual when cluster algorithms are concerned. I really thank you all for your help. Best regards, Rachele Nerattini -- --------------------------------------------------------------------------------------------- Rachele Nerattini Dip. di Fisica e Astronomia, Università di Firenze Via G. Sansone, 1, 50019 Sesto Fiorentino (FI), Italy ---------------------------------------------------------------------------------------------

2 3

query regarding double counting in classical monte carlo simulation
by D´Souza, Sunil Wilfred 22 Oct '14

22 Oct '14

Dear Alps Users, I am trying to simulate a binary ferromagnetic system i.e FeCo. I am running the classical monte carlo simulations by using the classical Heisenberg model . I am using the spinmc binary for my task. My confusion is regarding the double counting aspect of the Heisenberg model. since i have FeCo as the binary system. I will have two set of edges , which are as follows: Set 1 : { Fe-Fe, Fe-Co, Co-Co} Set 2 : {Fe-Fe, Fe-Co, Co-Fe, Co-Co}. My question is, In order to simulate the FeCo system correctly should i consider the edge Fe-Co ( J= 25 meV) only once OR should i consider both F e-Co and Co-Fe (J= 50 mev). I have tried to locate the exact equation of the Heisenberg Hamiltonian which is being used for the classical monte carlo simulation, but unfortunately i could not find this information. is the double counting aspect already taken care in the ALPS simulations. if so, then i think i have to define both Fe-Co and Co-Fe edges for the simulation. I am kindly requesting you to provide me insight into this aspect. I am caught up with this confusion, which is stopping me from proceeding further. Therefore, I am kindly requesting you to provide me your suggestions and help regarding the above mentioned issue. I thank you in advance for your kind consideration and time. with best regards Sunil

2 1

dwa & Heisenberg chain
by Rene John Kerkdyk 22 Oct '14

22 Oct '14

Dear ALPS developers and users, while using the dwa application I stumbled upon a segmentation fault. This happens when I try to simulate a S=1/2 Heisenberg chain. Below is the input file "dwa.parm" used and the output generated by ALPS. Is it not possible to calculate this spin model with the dwa application? Best, René Reference: I am using the alps-2.2.b3-r7462-src to build ALPS and work on the following operating system: Linux version 2.6.32-431.20.5.el6.x86_64 (mockbuild(a)c6b8.bsys.dev.centos.org) (gcc version 4.4.7 20120313 (Red Hat 4.4.7-4) (GCC) ) #1 SMP Fri Jul 25 08:34:44 UTC 2014 CentOS release 6.5 (Final) Kernel \r on an \m Input & Output: -bash-4.1$ cat dwa.parm LATTICE="chain lattice"; MODEL="spin"; LATTICE_LIBRARY="./lattices.xml" MODEL_LIBRARY="./models.xml" local_S = 1/2 J = 1 h=0 L=4; T = 0.001; SWEEPS=5000000; THERMALIZATION=100000; SKIP=500; MEASURE[Winding Number]=1 { t=0.01; } -bash-4.1$ parameter2xml dwa.parm Converting parameter file dwa.parm to dwa.parm.in.xml -bash-4.1$ dwa --write-xml dwa.parm.in.xml /***************************************************************************** /SNIP/ *****************************************************************************/ using the ALPS parallelizing scheduler copyright (c) 1994-2006 by Matthias Troyer <troyer(a)comp-phys.org>. see Lecture Notes in Computer Science, Vol. 1505, p. 191 (1998). based on the ALPS libraries version 2.2.b3 available from http://alps.comp-phys.org/ copyright (c) 1994-2013 by the ALPS collaboration. Consult the web page for license details. For details see the publication: B. Bauer et al., J. Stat. Mech. (2011) P05001. parsing task files ... Initialization stage 1 (Site states) ... starting ... finished. Initialization stage 2 (Hamiltonian) ... starting i. State minimum/maximum ... starting ... done. i. Onsite matrix ... starting a. Step 1: partial evaluate everything up to x,y,z terms ... starting ... done. b. Step 2: Initialize onsite matrix ... starting: ... done. ... done. iii. Bond strength matrix ... starting ... done. ... finished. Initialization stage 3 (Lookups) ... starting ... finished. Initialization stage 4 (Measurements) ... starting ... finished. Parameters ========== LATTICE = "chain lattice"; MODEL = spin; LATTICE_LIBRARY = ./lattices.xml; MODEL_LIBRARY = ./models.xml; local_S = 1/2; J = 1; h = 0; L = 4; T = 0.001; SWEEPS = 5000000; THERMALIZATION = 100000; SKIP = 500; MEASURE[Winding Number] = 1; t = 0.01; SEED = 856848597; Simulation ========== Sweeps : 5000000 Thermalization sweeps : 100000 Skip (Sweeps per measurement) : 500 Sweep counter : 0 Sweep failure counter : 0 Propagation counter : 0 Propagation failure counter : 0 Interested in time-of-flight images : 0 Waist correction : 0 Lattice ======= Name : chain lattice Dimension : 1 Total Sites/Bonds : 4 / 4 Periodic : 1 Model ===== Name : spin Charge model : 0 Spin model : 1 Site Basis : spin states : -0.5 0.5 Measurements ============ MEASURE : 1 MEASURE[Simulation Speed] : 1 MEASURE[Total Particle Number^2] : 0 MEASURE[Energy^2] : 0 MEASURE[Density^2] : 0 MEASURE[Energy Density^2] : 0 MEASURE[Winding Number] : 1 MEASURE[Local Kink:Number] : 0 MEASURE[Local Density] : 0 MEASURE[Local Density^2] : 0 MEASURE[Green Function] : 0 Created run 1 locally Starting task 1. Segmentation fault --------------------------------- René John Kerkdyk Institute for Theoretical Physics University of Goettingen Friedrich-Hund-Platz 1 D-37077 Goettingen Germany Office: A4.126 Phone: +49 551 39 7687

2 1

mps_optim checkpointing
by Mateusz Łącki 20 Oct '14

20 Oct '14

Dear All, I would like to answer to my own question. I have found that chkp_each works also for mps_optim (not only for mps_evolve) It enables to checkpoint every few iterations (for example every 100). BTW setting chkp_each to 0 causes „division by zero“ error due to line 113 of a file: ./applications/dmrg/mps/mps_optim/dmrg_sim.hpp 113 if (stopped || (sweep+1) % chkp_each == 0 || (sweep+1) == parms["nsweeps“]) It also shows that turning off checkpointing altogether is not possible, but this is the line which has to be modified to switch it off. Kind Regards, Mateusz Łącki

2 2

mps_optim checkpoints
by Mateusz Łącki 19 Oct '14

19 Oct '14

Dear All I would like to ask about checkpointing in mps_optim. Is it possible to turn off checkpointing in mps_optim altogether? I am doing computations on a network drive with volatile file system. I would like to limit write operations to increase reliability. Is it possible without ingeration in source code? It would also help to save a little bit of disk space... Kind Regards, Mateusz Łącki

1 0

Problem working out a specific Hamiltonian
by Niraj Ghimire 11 Oct '14

11 Oct '14

Dear alps users/developers, I am trying to study the properties of a system using a Hamiltonian that has terms describing nearest-neighbor and next-nearest-neighbor interactions. I am thinking I can do so by simply using one of the Hamiltonians described in "models.xml" but I don't know how. I went through all the tutorials given on the alps homepage but found no help. I will truly appreciate any help or guidance. Thank you, Niraj

2 2

vt_skip="true" in the lattice definition
by Vladimir Iglovikov 09 Oct '14

09 Oct '14

What does vt_skip="true" mean in the lattice definition?

2 1

Re: [ALPS-users] complexity estimation for the 3D ferromagnetic Heisenberg model
by Menchyshyn Oleh 05 Oct '14

05 Oct '14

Dear Prof. Todo, Thank you very much for your answers and time. ありがとうございます Best regards, Oleh

1 0

Re: [ALPS-users] complexity estimation for the 3D ferromagnetic Heisenberg model
by Menchyshyn Oleh 04 Oct '14

04 Oct '14

Thank you for your answers, I think I understand much more now. I would like to clarify just 2 points: 1) I don't get why should I divide 48 by 3? What do you call a bond? - Is it a $ J_{i,j} ( S^x_i * S^x_j + S^y_i * S^y_j + S^z_i * S^z_j ) = J_{i,j} ((S_i, S_j)) $ ? Here I've denoted a scalar product with the double parentheses - Or is it just an edge in the lattice graph? - For example ...................... <EDGE><SOURCE vertex="1" offset="0 0 0"/><TARGET vertex="2" offset="0 0 0"/></EDGE> <EDGE><SOURCE vertex="1" offset="0 0 0"/><TARGET vertex="4" offset="-1 -1 0"/></EDGE> ...................... where the former is an intracell edge, and the later is an intercell edge. I meant I had the 48 such intra- and intercell edges together. 2) I know about --time-limit option, and the tutorial says: -- gives the time (in seconds) which the program should run before writing a final checkpoint and exiting Can I restart the simulation again and make the ALPS using the obtained output data to proceed further? I know it sounds stupid to interrupt the simulation to restart it later. Otherwise, I would need to talk to sysadmin to change privileges for my account, so my simulations could exceed 48 hours. Best regards, Oleh > Dear Oleh, > > On 2014/10/01, at 21:40, Menchyshyn Oleh <oleh.menchyshyn(a)gmail.com> > wrote: > > So in my case, having the lattice with 16 atoms unit cell and 48 intra > and intercell bonds (coupling constants), > > means I would need to multiply the estimated expected time for some > simple reference model by 48? > > Since there are 3 bonds in a unit in my case, 48/3 = 16 would be the > correct factor to be used. > > > The THERMALIZATION and SWEEPS parameters are chosen to meet the relative > errors set up for observables being measured. > > If we increase the size of the system, namely L=8,12,16,24,..., but > leave the SWEEPS the same, the relative error increases and the accuracy of > simulation drops, aren't they? > > The increase of error bar will be very slow (or sometimes negligible) for > the loop algorithm, > as the autocorrelation time stays O(1) irrespective of the system size. > > > I'm asking that to clarify the issue as the cluster I use limits the > jobs running time to 48 hours. > > > > And is it possible in principle in ALPS to stop the calculation and > restart later as a new process? > > Or in that case when I need to analyse the QMC statistics manually? > > You can use the command line option ??time-limit? to specify the time > after which the program > will write final checkpoints and terminate. The simulation will continue > from the checkpoints when > the program is executed again. > > Please read > http://alps.comp-phys.org/mediawiki/index.php/ALPS_using_the_command_line#C… > for more details. > > Best, > Synge >

2 1