June 2009 - Comp-phys-alps-users

Problems with worms??
by Jakub Zakrzewski 27 Oct '10

27 Oct '10

Hello, Sometimes QMC worm code (used on Bose-Hubbard) terminates with the following error: here come first earlier succesful simulations then ... Created run 85 remote on Host ID: 84 Created run 86 remote on Host ID: 85 Created run 87 remote on Host ID: 86 Created run 88 remote on Host ID: 87 All processes have been assigned Checking if Simulation 1 is finished: not yet, next check in 120 seconds ( 0% done). q = -0 state1 = 0 state2 = 0 bond_type = 0 zero matrix element in remove_jump application called MPI_Abort(MPI_COMM_WORLD, -2) - process 67 Could somebody advice what I can do? Best regards Kuba

2 2

complex hopping parameter
by Jakub Kolasinski 19 Jun '09

19 Jun '09

Dear Mr Troyer, I allow myself to use Your kindness one more time. My question is, how can I make hopping parameter complex (if that is possible at all)? I get an error that "can not convert complex number into a real one". I'm using worm. Thank You in advance for Your answer Jakub

2 1

evaluation of <b>
by Jakub Kolasinski 18 Jun '09

18 Jun '09

I'm using worm. Would it be, that there is no way (besides changing the source code) to evaluate it with worm? Many thanks Jakub

2 1

evaluation of <b>
by Jakub Kolasinski 18 Jun '09

18 Jun '09

Thank You for quick response, but the main part of my question is the problem with defining custom measurements. I would like to measure MEASURE_CORRELATION[name]="bdag:b" and I get an information that it can't be measured since it is offdiagonal. I gave an example of MEASURE_LOCAL[name]=b; because it doesn't work for me too. Jakub

2 1

evaluation of <b>
by Jakub Kolasinski 17 Jun '09

17 Jun '09

Hello, One question (maybe trivial - at least I hope so) - how can I evaluate <b>, where b is standard annihilation operator in Bose Hubbard model? Theoretically there is an information on the ALPS website (about defining custom measurements) how to do it, but... for example let's try simple thing: MEASURE_LOCAL[name]=b; What I get, is an information, that: "Will not measure "name" since it is off-diagonal or not a site operator". "b" is a site operator, so the problem is whether it is diagonal or not - but the only one constraint mentioned on the website is that operator must be a site operator - so is that information not full, or am I wrong somewhere here? Help will be very apreciated Jakub

2 1

Stiffness definition
by Masaki Oshikawa 11 Jun '09

11 Jun '09

Hello, Many thanks to the ALPS developers for providing such a great software. I have a question on the stiffness (I am dealing with spin system now.) To my knowledge, the stiffness is usually defined by the gradient expansion of the free energy density ¥[ f ¥sim ¥mbox{const.} + ¥frac{¥rho_s}{2} (¥nabla ¥theta)^2, ¥] where $¥theta$ is the phase angle (rotation angle in the xy plane, in the case of a planar magnet). For the groundstate ($T=0$) of a 1D spin chain, the above implies ¥[ ¥rho_s = L ¥frac{¥partial^2 E_0(¥varphi)}{¥partial ¥varphi^2}, ¥] where $E_0(¥varphi)$ is the groundstate energy of a ring of circumference $L$ for the twisted boundary condition with the twist angle $¥varphi$. ($E_0$ is $L$ times the groundstate energy density.) (I am following, e.g. N. Laflorencie et al., Eur. Phys. J. B 24, 77 (2001) and references therein. ) I was bit confused by the results from ALPS so I tried to check S=1/2 XXZ chain, for which exact values are known. Just for example I chose Jxy=1, Jz=0.5 (easy-plane antiferromagnetic chain). I haven't done a very systematic study, but for a given length (I did $L=30, 60, 120$) the stiffness seems to converge to a constant as the temperature is lowered. I took the "stiffness" output of ALPS at low enough temperature as the groundstate value. (I mostly used ALPS/looper code here, but I checked that dirloop_sse also produces similar results.) L=30 : 0.0104 L=60 : 0.00517 L=120: 0.00258 This decreases for larger size $L$. On the other hand, the exact value of the stiffness is known to be $9 ¥sqrt{3}/(16 ¥pi) ¥sim 0.310$ (in the infinite size limit, at T=0). This is in apparent contradition to the ALPS result if they refer to the same quantity. Actually if I multiply the output of ALPS by the length $L$, it agrees well with the exact value. So it seems that the "stiffness" output of ALPS is $1/L$ times the standard(??) definition. (Am I missing something trivial?) For the $S=1/2$ chain I think I can convert the output to what I want. But more generally what is the definition of the "stiffness" in ALPS? I looked at ALPS documentation but could not find it. I would appreciate if you could give me a pointer. I am now interested in S>1/2 case, and would like to make sure I will not make wrong observation due to mismatch in definition/convention. Thanks. Regards, Masaki -- Masaki Oshikawa <oshikawa(a)issp.u-tokyo.ac.jp> ---------------------------------------------------------------------------------------- Institute for Solid State Physics, University of Tokyo 5-1-5 Kashiwanoha, Kashiwa, 277-8581 Japan URL http://oshikawa.issp.u-tokyo.ac.jp/ ----------------------------------------------------------------------------------------

2 1

the total momentum quantum number
by khalid hassan 10 Jun '09

10 Jun '09

Dear alps team for the exact diaginalization program fulldiag and sparsediag, the output is divided into total momentum quantum numbers. also in many papers the spectrum is shown according to total momentum quantum number. for example for a square 2 by 2 lattice the program output with fulldiag will be divided as total momentum = 0 0 total momentum = Pi 0 total momentum = 0 Pi total momentum = Pi Pi Now i want to know that is these are written in k-space as these are not possible in real space how they correspond to the values in real space for example we have Nup and Ndown known as alps gives for a certain momentum eg Pi 0 can we know Nup and Ndown at each sites in real space for example for 3 up and 1 down in square lattice can we know at site 1 there is up, down or double occupied etc. is there some program statement in alps that tells this I also want to know the total spin S for each case in square lattice can alps helps. alps web site says something about total momentum but everyone knows it I cant find my answers there. actually I want the energies output by alps to correspong to situation in real space how can I associate these specific situations with these energies best regards khalid loane

2 1

lanczos algorithm and fulldiag_mpi
by Junping Shao 07 Jun '09

07 Jun '09

Hi, I'm interested in the internal workings of the sparse and full diagonalization programs, because I want to parallelize and adapt it to my studies. Could someone point out to me the code section in which the Lanczos is performed, especially the parts where a)preconditioning according to lattice symmetries, and b)reorthonormalization is done? (I'm a newbie at C++) Also, I can't find the source code which corresponds to the fulldiag_mpi binary in the fulldiag tutorial directory. Would someone show me where it is, or kindly link me to it, if possible? Much appreciation is due from me! (I'm doing an undergrad research project) Shao

2 1

Wave Functions from Exact Diagonalization
by Jesse Kinder 03 Jun '09

03 Jun '09

I am doing some variational Monte Carlo calculations. I would like to calculate the overlap of my optimized trial wave functions with the exact ground state for some small systems. Is there some way to have the sparsediag or fulldiag routines print energy eigenvectors to a file? If so, is there some documentation that describes what the output file will contain? Jesse

1 0

Bond Types in a Custom Lattice
by Miles Stoudenmire 03 Jun '09

03 Jun '09

Hi - I'm creating an application based on the ALPS classical Monte Carlo app that uses a custom lattice where a subset of the spins are to be designated as impurities. Here's an example lattice I'm using (just a 4 site chain): <LATTICES> <GRAPH name="test lattice" dimension="1" vertices="4" edges="4"> <VERTEX id="1" type="1"><COORDINATE>0</COORDINATE></VERTEX> <VERTEX id="2" type="0"><COORDINATE>1</COORDINATE></VERTEX> <VERTEX id="3" type="0"><COORDINATE>2</COORDINATE></VERTEX> <VERTEX id="4" type="0"><COORDINATE>2</COORDINATE></VERTEX> <EDGE source="1" target="2" id="1" type="0" vector="1"/> <EDGE source="2" target="3" id="2" type="0" vector="1"/> <EDGE source="3" target="4" id="3" type="0" vector="1"/> <EDGE source="4" target="1" id="4" type="0" vector="1"/> </GRAPH> </LATTICES> Note how site 1 has type 1, which means it's to be treated as an impurity. The problem I'm having is that when I'm looping over the bonds/edges using the following code, instead of detecting the bond type as 0 as written in the XML, it's detecting bond 1 as being of type 0, bonds 2 and 3 as being of type 1 and bond 4 as being of type 2. When I make all of the spins/vertices of type 0, however, all of the bonds come out as type 0. Here's the code (in a member function of the SpinSim object, which is a graph_helper object within spinsim.h): parent::bond_iterator bi; bi = (this->bonds()).first; for (; bi!=(this->bonds()).second; ++bi) { // get bond type int edge_type = this->edge_type(*bi); std::cout << " Bond of type " << edge_type; } How can I get ALPS to use the bond types that are written instead of making up its own? Thanks! -- -=Miles Stoudenmire=- miles(a)physics.ucsb.edu miles.stoudenmire(a)gmail.com http://www.crazytheory.com/

2 5