Mailman 3 April 2015 - Kwant-discuss

Re: [Kwant] Appcrash for very large systems
by Benoit GAURY April 27, 2015

April 27, 2015

Hi Sam, I just ran the code you gave, uncommenting the LDOS calculation. It went fine, no crash. My system is Debian GNU/Linux (Testing release) with Python 2.7.9 and Kwant version 1.0.1. My computer (laptop) has 8 Gb of RAM. FYI, I already performed several simulations of systems containing around 10^5 sites without any issue. Best, Benoit On Tue, Apr 28, 2015 at 1:08 AM, Sam LaGasse <lagassesam(a)gmail.com> wrote: > Greetings all, > > I have a question regarding using kwant for very large systems. In my > research I am trying to model graphene-based devices on the order of L = W > = 500 nm. I encounter a problem when trying to simulate electron density in > these devices with Kwant. > > The errors I get look like this: > > Faulting application name: python.exe, version: 0.0.0.0, time stamp: > 0x5488acd7 > Faulting module name: _mumps.pyd, version: 0.0.0.0, time stamp: 0x54d2a92a > Exception code: 0xc00000fd > Fault offset: 0x000000000076e357 > Faulting process id: 0x5018 > Faulting application start time: 0x01d0811ecc4f2ba7 > Faulting application path: > C:\WinPython-64bit-2.7.9.4\python-2.7.9.amd64\python.exe > Faulting module path: > C:\WinPython-64bit-2.7.9.4\python-2.7.9.amd64\lib\site-packages\kwant\linalg\_mumps.pyd > Report Id: d70446a6-ed26-11e4-9806-b8ac6f4eb471 > > A little info on the system: I am running 64 bit python installed from > the WinPython package. The IDE is Spyder. I've gotten all of the example > codes to work as well as several of my own (albeit for smaller versions of > the systems we would like to study). > > From the error and doing some testing with the code, I've devised that > it must be a problem occuring solver itself. I have enough memory to build > the system and I am pretty sure I am not running out of memory during the > actual solving, although I am not positive. I am wondering if the mumps > routine could be breaking for very large systems? When I run my code on a > laptop with much less ram (~60 gb for the desktop vs 4 gb on the laptop) I > end up with a memory error instead of a kernel crash. > > I've attached a modified version of one of the tutorial codes, where I > have added in a LDOS/electron density calculation and increased the > dimensions of the simulation to 500x1000. The code successfully finalizes > the device on my machine, but crashes the python kernel during the actual > calculation. > > I can get more ram for the machine if necessary, but I want to make sure > that it isn't a problem related to the solver before sending the purchase > order to my advisor. > > Kwant seems extremely nice so we're hoping to use it in our research in > the future! It is a lot better than the NEGF code I strung together in > matlab! > > Thank you! > > Sam LaGasse >

1 0

Armchair nanoribbon
by Lee Kwok-Long April 27, 2015

April 27, 2015

Dear authors, I want to consider the transport in armchair graphene nanoribbons, so I changed the definition of graphene lattice: graphene = kwant.lattice.general([(sqrt(3)/2, 1/2), (0, 1)], [(0, 0), (1 / (2*sqrt(3)),1/2)]) a, b = graphene.sublattices Now it gives an armchair grphene nanoribbon, but it gives wrong results when I attach the leads. Is there a method to solve my problem? Thanks in advance! Kwok-Long Lee My code is attached: from __future__ import division # so that 1/2 == 0.5, and not 0 from math import pi, sqrt, tanh import kwant # For computing eigenvalues import scipy.sparse.linalg as sla # For plotting from matplotlib import pyplot # Define the graphene lattice graphene = kwant.lattice.general([(sqrt(3)/2, 1/2), (0, 1)], [(0, 0), (1 / (2*sqrt(3)),1/2)]) a, b = graphene.sublattices def make_system(r=10, w=2.0, pot=0.1): #### Define the scattering region. #### # circular scattering region def circle(pos): x, y = pos return 0<=x<10 and 0<=y<10 #x ** 2 + y ** 2 < r ** 2 sys = kwant.Builder() def potential(site): (x, y) = site.pos d = y * cos_30 + x * sin_30 return pot * tanh(d / w) sys[graphene.shape(circle, (0, 0))] = potential sys[graphene.neighbors()] = -1 def lead0_shape(pos): x, y = pos return 0 <= y < 10 sym0 = kwant.TranslationalSymmetry(graphene.vec((-1, 0))) lead0 = kwant.Builder(sym0) lead0[graphene.shape(lead0_shape, (0, 0))] = -pot lead0[graphene.neighbors()] = -1 return sys, lead0 def main(): pot = 0.1 sys, lead0 = make_system(pot=pot) def family_colors(site): return 0 if site.family == a else 1 # Plot the closed system without leads. kwant.plot(sys, site_color=family_colors, site_lw=0.1, colorbar=False) sys.attach_lead(lead0) sys.attach_lead(lead0.reversed()) # Then, plot the system with leads. kwant.plot(sys, site_color=family_colors, site_lw=0.1, lead_site_lw=0, colorbar=False) # Finalize the system. sys = sys.finalized() # Call the main function if the script gets executed (as opposed to imported). # See <http://docs.python.org/library/__main__.html>. if __name__ == '__main__': main()

2 4

precision of calculations
by Sergey Slizovskiy April 26, 2015

April 26, 2015

Dear Authors, Is there any way to increase the internal numerical precision of calculations? In particular, I am interested in the wave-function. Currently the precision is around exp(-40), which is good, but, still, I would be happy to have more. Can kwant be combined with *mpmath *library for instance? Another question is about parallel computations: are there any examples of it's use with kwant? Thank you, Sergey

3 2

Formation of Hamiltonian with Rashba Interaction with up and down spin
by Sudin Ganguly April 17, 2015

April 17, 2015

Dear Sir, I want to study Spin Hall Effect with Rashba interaction and for this I've to separate the Hamiltonian for up and down spin and also the leads as mentioned in http://kwant-project.org/doc/1.0/tutorial/tutorial5#id1 So I did this. The attaching lead part is done. But I've some confusion with the Hamiltonian of the scattering region (square lattice). Specifically with the hopping terms between up and down spin. Can you please tell me how do I attach the effect of up spin in the Hamiltonian of down spin and the opposite part. I've attached a part of the Hamiltonian in which the spins have been separated. ================ lat_u = kwant.lattice.square(a, name='u') lat_d = kwant.lattice.square(a, name='d') sys = kwant.Builder() sys[(lat_u(x, y) for x in range(L) for y in range(W))] = 0.0 sys[(lat_d(x, y) for x in range(L) for y in range(W))] = 0.0 # hoppings for both up and down spins sys[lat_u.neighbors()] = -t sys[lat_d.neighbors()] = -t # Rashba term enters as hopping between # up and down spins for i in xrange(L): for j in xrange(W): # Hopping in y-direction if j > 0: sys[lat_u(i, j), lat_d(i, j - 1 )] = -1j*alpha sys[lat_d(i, j), lat_u(i, j - 1)] = -1j*alpha #Hopping in x-direction if i > 0: sys[lat_u(i, j), lat_d(i - 1, j)] = alpha sys[lat_d(i, j), lat_u(i - 1, j)] = -alpha sys = sys.finalized() ================================== With regards Sudin ================== Sudin Ganguly Research Scholar Dept. of Physics IIT Guwahati Assam,India-781039 ===================

2 1

Re: [Kwant] Kwant-discuss Digest, Vol 20, Issue 4
by Lee Kwok-Long April 13, 2015

April 13, 2015

Dear Xavier, Thanks a lot for the reply. It is very clear for me now. Regards, Kwok-Long Lee On Mon, Apr 13, 2015 at 8:01 PM, <kwant-discuss-request(a)kwant-project.org> wrote: > Send Kwant-discuss mailing list submissions to > kwant-discuss(a)kwant-project.org > > To subscribe or unsubscribe via the World Wide Web, visit > https://mailman-mail5.webfaction.com/listinfo/kwant-discuss > or, via email, send a message with subject or body 'help' to > kwant-discuss-request(a)kwant-project.org > > You can reach the person managing the list at > kwant-discuss-owner(a)kwant-project.org > > When replying, please edit your Subject line so it is more specific > than "Re: Contents of Kwant-discuss digest..." > > > Today's Topics: > > 1. Re: about ldos and wave function probability (Xavier Waintal) > > > ---------------------------------------------------------------------- > > Message: 1 > Date: Mon, 13 Apr 2015 10:06:21 +0200 > From: Xavier Waintal <xavier.waintal(a)cea.fr> > To: Lee Kwok-Long <kwoklonglee(a)gmail.com> > Cc: kwant-discuss(a)kwant-project.org > Subject: Re: [Kwant] about ldos and wave function probability > Message-ID: <15D07A0E-01BF-493E-A3AA-071571C396FD(a)cea.fr> > Content-Type: text/plain; charset=iso-8859-1 > > Dear Kwo-Long, > > Le 11 avr. 2015 à 07:41, Lee Kwok-Long <kwoklonglee(a)gmail.com> a écrit : > > > Dear all, > > I am learning Kwant through the tutorials and published paper, but I > feel confused about some Kwant scripts below: > > 1)Tutorial 2.7.1. > > ham = sys.hamiltonian_submatrix() > > evecs = la.eigh(ham)[1] > > wf = abs(evecs[:, n])**2 > > 2)New Jour nal of Physics 16(2014) 063065 figure14: > > wf=kwant.wave_function(sys, energy, args) > > return(abs(wf(lead_nr))**2).sum(axis=0) > > 3)New Jour nal of Physics 16(2014) 063065 figure4: > > local_dos = kwant.ldos(sys, energy=. 2) > > kwant.plotter.map(sys, local_dos, num_lead_cells=10) > > My questions are: > > Does "evecs = la.eigh(ham)[1]" give the same wave function as > "wf=kwant.wave_function(sys, energy, args)"? > > The first one simply calculates the eigenvectors of a FINITE system using > standard diagonalization routines > (as found in numpy - not in Kwant). The second calculates the scattering > eigenfunctions of an INFINITE system > with leads at energy= energy (The latter belong to a continuum). > > > What is the definition of local_dos? > > The usual one, i.e. ldos(E,r)= sum_a |wf_a(a)|^2 delta(E-E_a) > where wf_a and E_a are the eigen functions/energy of the system > > > Does"local_dos = kwant.ldos(sys, energy=. 2)" give the same results as > "wf = abs(evecs[:, n])**2"? > > Yes up to a factor 1/2pi. kwant.ldos is a simpler wrapper around > kwant.wave_function(). Please have > a look at the source code. > > Best, > Xavier > > > > Could you help me to know the differences and similarities? > > Kwok-Long Lee > > > > > > ------------------------------ > > Subject: Digest Footer > > _______________________________________________ > Kwant-discuss mailing list > Kwant-discuss(a)kwant-project.org > https://mailman-mail5.webfaction.com/listinfo/kwant-discuss > > > ------------------------------ > > End of Kwant-discuss Digest, Vol 20, Issue 4 > ******************************************** >

1 0

about ldos and wave function probability
by Lee Kwok-Long April 13, 2015

April 13, 2015

Dear all, I am learning Kwant through the tutorials and published paper, but I feel confused about some Kwant scripts below: 1)Tutorial 2.7.1. ham = sys.hamiltonian_submatrix() evecs = la.eigh(ham)[1] wf = abs(evecs[:, n])**2 2)New Jour nal of Physics 16(2014) 063065 figure14: wf=kwant.wave_function(sys, energy, args) return(abs(wf(lead_nr))**2).sum(axis=0) 3)New Jour nal of Physics 16(2014) 063065 figure4: local_dos = kwant.ldos(sys, energy=. 2) kwant.plotter.map(sys, local_dos, num_lead_cells=10) My questions are: Does "evecs = la.eigh(ham)[1]" give the same wave function as "wf=kwant.wave_function(sys, energy, args)"? What is the definition of local_dos? Does"local_dos = kwant.ldos(sys, energy=. 2)" give the same results as "wf = abs(evecs[:, n])**2"? Could you help me to know the differences and similarities? Kwok-Long Lee

2 1

kwant on OS X
by Bas Nijholt April 9, 2015

April 9, 2015

Small update on installing kwant on OS X: brew install gfortran python > brew install gcc python and why is there: brew tap samueljohn/python instead of brew tap homebrew/python? Best, Bas

1 0

reproducing graphene magnetoresistance curve (Novoselov paper)
by ANANT VIJAY VARMA April 1, 2015

April 1, 2015

I am trying to reproduce Novoselov paper the magnetoresistance curve.Plese help me correct this program. Thanks! #system Building def make(a=160,b=70,t=1,alpha= .0): def Square(pos): x , y = pos return 0<=x<=a and 0<y<=b def hopx(site1,site2,B=1): xi,yi = site1.pos xj,yj = site1.pos return -t * exp(-1j* B *(yj-yi)*(xj+xi)/2)*s_0 def hopy(site1,site2,B=1): xi,yi = site1.pos xj,yj = site1.pos return -t * exp(-1j* B *(yj-yi)*a) * s_0 def gate(site1,g=0): return g*s_0 sys = kwant.Builder() sys[lat.shape(Square,(2.46,0))] = gate sys[kwant.builder.HoppingKind((0, 0), A,B )]= -t * s_0 sys[kwant.builder.HoppingKind((0,1), A,B )] = hopx sys[kwant.builder.HoppingKind((-1,1), A,B )] = -t * s_0 #left lead sym0 = kwant.TranslationalSymmetry(lat.vec((-1,0))) def shape(pos): x, y = pos return (0< y <=b) lead0 = kwant.Builder(sym0) lead0[lat.shape(shape, (2.46, 0))] = gate lead0[kwant.builder.HoppingKind((0, 0), A,B )]= hopy lead0[kwant.builder.HoppingKind((0,1), A,B )] = hopy lead0[kwant.builder.HoppingKind((-1,1), A,B )] = hopy sys.attach_lead(lead0) #right lead sys.attach_lead(lead0.reversed()) #LEAD PAIR sym1 = kwant.TranslationalSymmetry(lat.vec((0,1))) def shape(pos): x, y = pos return (a/5< x <=2*a/5) lead1 = kwant.Builder(sym1) lead1[lat.shape(shape, (2.46*a/10, 0))] = gate lead1[lat.neighbors()] = -t * s_0 sys.attach_lead(lead1) #right lead sys.attach_lead(lead1.reversed()) #LEAD PAIR sym2 = kwant.TranslationalSymmetry(lat.vec((0,1))) def shape(pos): x, y = pos return (3*a/5< x <=4*a/5) lead2 = kwant.Builder(sym2) lead2[lat.shape(shape, (2.46*3*a/10, 0))] = gate lead2[lat.neighbors()] = -t * s_0 sys.attach_lead(lead2) #right lead sys.attach_lead(lead2.reversed()) return sys.finalized() def plot_conductance(sys, energy, gates): data = [] for g in gates: smatrix = kwant.smatrix(sys, energy,args=[g]) cond = np.array([[smatrix.transmission(i, j) for j in xrange(6)] for i in xrange(6)]) cond -= np.diag(cond.sum(axis=0)) cm = cond[:-1, :-1] nonlocal_resistance = np.linalg.solve(cm, [-1,1,0,0,0])[3] print nonlocal_resistance data.append(nonlocal_resistance) dos = kwant.ldos(sys, energy=1) x = np.empty(len(dos)/2, float) for i in xrange(len(dos)/2): x[i] = dos[2*i] pyplot.figure() pyplot.plot(gates, data) pyplot.xlabel("Vg") pyplot.ylabel("Resistance") pyplot.show() kwant.plotter.map(sys,dos, num_lead_cells=5) def main(): def family_site(site): return 'b' if site.family == A else 'y' sys = make() kwant.plot(sys) #plot_conductance(sys, energy=1, Bfields=[i* 0.006 for i in #xrange(100)]) plot_conductance(sys,energy= 0, gates= np.linspace(-1,1,100)) if __name__ == '__main__': main()

2 1