def eval_rdm1BF(dmrg,flmps,fname='top',spinfo=None):
print '\n[mpo_dmrg_propsItrf.eval_rdm1BF]'
t0 = time.time()
nsite = dmrg.nsite
if not dmrg.ifs2proj:
rdm1 = numpy.zeros((nsite,nsite,2),dtype=dmrg_dtype)
for p in range(nsite):
for q in range(nsite):
if p > q: continue
fop = mpo_dmrg_propsMPO.genMPO_Epq(nsite,p,q,fname,dmrg.ifQt)
exphop = mpo_dmrg_props.evalProps(dmrg,flmps,flmps,fop,status='L')
fop.close()
rdm1[p,q] = exphop
rdm1[q,p] = exphop
rdm1 = rdm1.transpose(2,0,1)
rdm1t = rdm1[0]+rdm1[1]
rdm1s = 0.5*(rdm1[0]-rdm1[1])
else:
npts,sval,sz = spinfo
denorm,xts,wts = eval_P(dmrg,flmps,spinfo)
# RDM1 = <Epq>
rdm1t = numpy.zeros((nsite,nsite),dtype=dmrg_dtype)
for p in range(nsite):
for q in range(nsite):
if p > q: continue
fop = mpo_dmrg_propsMPORpt.genMPO_EpqRpt(nsite,p,q,fname,xts,dmrg.ifQt)
exphop = mpo_dmrg_props.evalProps(dmrg,flmps,flmps,fop,status='L')
fop.close()
rdm1t[p,q] = numpy.sum(wts*(exphop[:npts]+exphop[npts:]))/denorm
rdm1t[q,p] = rdm1t[p,q]
# Spin density matrix 0.5*<EpqA-EpqB>
# Generation of additional quadrature points
sz1 = sz-1.0
xts1,wts1 = s2quad(npts,sval,sz1,sz)
rdm1s = numpy.zeros((nsite,nsite),dtype=dmrg_dtype)
for p in range(nsite):
for q in range(nsite):
if p > q: continue
# We use the same quadrature point
fop = mpo_dmrg_propsMPORpt.genMPO_TpqRpt(nsite,p,q,fname,xts,dmrg.ifQt)
exphop = mpo_dmrg_props.evalProps(dmrg,flmps,flmps,fop,status='L')
fop.close()
# Summarize results with different weights - wts,wts1
term1 = numpy.sum(wts*(exphop[:npts]-exphop[npts:2*npts]))/math.sqrt(2)
term2 = numpy.sum(wts1*exphop[2*npts:])
term1 = term1*sval/(sval+1.0)
term2 = term2*math.sqrt(sval)/(sval+1.0)
tpq = (term1+term2)/math.sqrt(2)
rdm1s[p,q] = tpq/denorm
rdm1s[q,p] = rdm1s[p,q]
# Check skewness
skewt = numpy.linalg.norm(rdm1t-rdm1t.T.conj())
print ' skewness of rdm1t =',skewt
assert skewt < 1.e-12
skews = numpy.linalg.norm(rdm1s-rdm1s.T.conj())
print ' skewness of rdm1s =',skews
assert skews < 1.e-12
t1 = time.time()
print ' time for eval_rdm1BF = %.2f s'%(t1-t0)
return rdm1t,rdm1s
def eval_SiSj(dmrg,flmps,groups,fname='top',spinfo=None):
print '\n[mpo_dmrg_propsItrf.eval_SiSj]'
t0 = time.time()
nsite = dmrg.nsite
ngroup = len(groups)
expect = numpy.zeros((ngroup,ngroup),dtype=dmrg_dtype)
if not dmrg.ifs2proj:
for idx,ig in enumerate(groups):
for jdx,jg in enumerate(groups):
if idx > jdx: continue
fop = mpo_dmrg_propsMPO.genMPO_SiSj(nsite,ig,jg,fname,dmrg.ifQt)
exphop = mpo_dmrg_props.evalProps(dmrg,flmps,flmps,fop,status='L')
fop.close()
expect[idx,jdx] = numpy.sum(exphop)
expect[jdx,idx] = expect[idx,jdx]
else:
npts,sval,sz = spinfo
denorm,xts,wts = eval_P(dmrg,flmps,spinfo)
for idx,ig in enumerate(groups):
for jdx,jg in enumerate(groups):
if idx > jdx: continue
fop = mpo_dmrg_propsMPORpt.genMPO_SiSjRpt(nsite,ig,jg,fname,xts,dmrg.ifQt)
exphop = mpo_dmrg_props.evalProps(dmrg,flmps,flmps,fop,status='L')
fop.close()
expect[idx,jdx] = numpy.sum(wts*(exphop[:npts]+exphop[npts:2*npts]+exphop[2*npts:]))/denorm
expect[jdx,idx] = expect[idx,jdx]
t1 = time.time()
print ' time for eval_SiSj = %.2f s'%(t1-t0)
return expect
def eval_Local2(dmrg,flmps,groups,ikey,jkey,fname='top',spinfo=None):
print '\n[mpo_dmrg_propsItrf.eval_Local2] (ikey,jkey)=',(ikey,jkey)
t0 = time.time()
nsite = dmrg.nsite
ngroup = len(groups)
expect = numpy.zeros((ngroup,ngroup),dtype=dmrg_dtype)
if not dmrg.ifs2proj:
for idx,ig in enumerate(groups):
for jdx,jg in enumerate(groups):
if idx > jdx: continue
fop = mpo_dmrg_propsMPO.genMPO_Local2(nsite,ig,jg,ikey,jkey,1.0,fname,dmrg.ifQt)
exphop = mpo_dmrg_props.evalProps(dmrg,flmps,flmps,fop,status='L')
fop.close()
expect[idx,jdx] = exphop[0]
expect[jdx,idx] = expect[idx,jdx]
else:
denorm,xts,wts = eval_P(dmrg,flmps,spinfo)
for idx,ig in enumerate(groups):
for jdx,jg in enumerate(groups):
if idx > jdx: continue
fop = mpo_dmrg_propsMPORpt.genMPO_Local2Rpt(nsite,ig,jg,ikey,jkey,1.0,fname,xts,dmrg.ifQt)
exphop = mpo_dmrg_props.evalProps(dmrg,flmps,flmps,fop,status='L')
fop.close()
expect[idx,jdx] = numpy.sum(wts*exphop)/denorm
expect[jdx,idx] = expect[idx,jdx]
t1 = time.time()
print ' time for eval_Local2 = %.2f s'%(t1-t0)
return expect
def eval_Global(dmrg,flmps,key,fname='top',spinfo=None):
nsite = dmrg.nsite
if not dmrg.ifs2proj:
fop = mpo_dmrg_propsMPO.genMPO_Global(nsite,key,fname,dmrg.ifQt)
exphop = mpo_dmrg_props.evalProps(dmrg,flmps,flmps,fop,status='L')
fop.close()
expect = exphop[0]
else:
denorm,xts,wts = eval_P(dmrg,flmps,spinfo)
fop = mpo_dmrg_propsMPORpt.genMPO_GlobalRpt(nsite,key,fname,xts,dmrg.ifQt)
exphop = mpo_dmrg_props.evalProps(dmrg,flmps,flmps,fop,status='L')
fop.close()
expect = numpy.sum(wts*exphop)/denorm
return expect
def eval_Global(dmrg,flmps,key,fname='top',spinfo=None):
print '\n[mpo_dmrg_propsItrf.eval_Global] key=',key
t0 = time.time()
nsite = dmrg.nsite
if not dmrg.ifs2proj:
fop = mpo_dmrg_propsMPO.genMPO_Global(nsite,key,fname,dmrg.ifQt)
exphop = mpo_dmrg_props.evalProps(dmrg,flmps,flmps,fop,status='L')
fop.close()
expect = exphop[0]
else:
denorm,xts,wts = eval_P(dmrg,flmps,spinfo)
fop = mpo_dmrg_propsMPORpt.genMPO_GlobalRpt(nsite,key,fname,xts,dmrg.ifQt)
exphop = mpo_dmrg_props.evalProps(dmrg,flmps,flmps,fop,status='L')
fop.close()
expect = numpy.sum(wts*exphop)/denorm
t1 = time.time()
print ' time for eval_Global = %.2f s'%(t1-t0)
return expect
def eval_Local(dmrg,flmps,groups,key,fname='top',spinfo=None):
nsite = dmrg.nsite
ngroup = len(groups)
expect = numpy.zeros(ngroup,dtype=dmrg_dtype)
if not dmrg.ifs2proj:
for idx,ig in enumerate(groups):
fop = mpo_dmrg_propsMPO.genMPO_Local(nsite,ig,key,fname,dmrg.ifQt)
exphop = mpo_dmrg_props.evalProps(dmrg,flmps,flmps,fop,status='L')
fop.close()
expect[idx] = exphop[0]
else:
denorm,xts,wts = eval_P(dmrg,flmps,spinfo)
for idx,ig in enumerate(groups):
fop = mpo_dmrg_propsMPORpt.genMPO_LocalRpt(nsite,ig,key,fname,xts,dmrg.ifQt)
exphop = mpo_dmrg_props.evalProps(dmrg,flmps,flmps,fop,status='L')
fop.close()
expect[idx] = numpy.sum(wts*exphop)/denorm
return expect
def eval_P2(dmrg,flmps0,flmps1,spinfo,debug=False): npts,sval,sz = spinfo xts,wts = s2quad(npts,sval,sz,sz) nsite = dmrg.nsite fname = './top' fop = mpo_dmrg_propsMPORpt.genMPO_Rpt(nsite,fname,xts,dmrg.ifQt) exphop = mpo_dmrg_props.evalProps(dmrg,flmps0,flmps1,fop,status='L') fop.close() denorm = numpy.sum(wts*exphop) if debug: print '<P>=',denorm return denorm,xts,wts
def eval_SiSj(dmrg,flmps,groups,fname='top',spinfo=None):
nsite = dmrg.nsite
ngroup = len(groups)
expect = numpy.zeros((ngroup,ngroup),dtype=dmrg_dtype)
if not dmrg.ifs2proj:
for idx,ig in enumerate(groups):
for jdx,jg in enumerate(groups):
fop = mpo_dmrg_propsMPO.genMPO_SiSj(nsite,ig,jg,fname,dmrg.ifQt)
exphop = mpo_dmrg_props.evalProps(dmrg,flmps,flmps,fop,status='L')
fop.close()
expect[idx,jdx] = numpy.sum(exphop)
else:
npts,sval,sz = spinfo
denorm,xts,wts = eval_P(dmrg,flmps,spinfo)
for idx,ig in enumerate(groups):
for jdx,jg in enumerate(groups):
fop = mpo_dmrg_propsMPORpt.genMPO_SiSjRpt(nsite,ig,jg,fname,xts,dmrg.ifQt)
exphop = mpo_dmrg_props.evalProps(dmrg,flmps,flmps,fop,status='L')
fop.close()
expect[idx,jdx] = numpy.sum(wts*(exphop[:npts]+exphop[npts:2*npts]+exphop[2*npts:]))/denorm
return expect