Cookbook¶
The first part writes out the Hamiltonian of the common models, ready for
set_hamiltonian. The second part gives short recipes for calculations
that combine several pages of the guide. Every chain is built as in
Models, n is its number of sites, and the operators are
those of its class. On a fermionic chain, h + h.get_dagger() adds the
Hermitian conjugate of every hopping and pairing term, so we write each
such term once.
Models¶
XXZ chain¶
sc = spinchain.Spin_Chain(["S=1/2"]*n)
Delta = 0.5
h = sum(sc.Sx[i]*sc.Sx[i+1] + sc.Sy[i]*sc.Sy[i+1] + Delta*sc.Sz[i]*sc.Sz[i+1]
for i in range(n-1))
\(\Delta=1\) is the Heisenberg chain, sc.SS(i, i+1); \(\Delta=0\) is the XX
chain, which maps to free fermions; \(|\Delta|>1\) is gapped.
Transverse-field Ising chain¶
sc = spinchain.Spin_Chain(["S=1/2"]*n)
J, hx = 1.0, 0.5
h = -J*sum(sc.Sz[i]*sc.Sz[i+1] for i in range(n-1)) - hx*sum(sc.Sx)
With spin-1/2 operators the critical point is at \(h=J/2\); with Pauli matrices (\(4S^zS^z\), \(2S^x\)) it is at \(h=J\).
J1-J2 chain¶
sc = spinchain.Spin_Chain(["S=1/2"]*n)
J1, J2 = 1.0, 0.4
h = J1*sum(sc.SS(i, i+1) for i in range(n-1)) + J2*sum(sc.SS(i, i+2) for i in range(n-2))
For \(J_2/J_1>0.2411\) the chain is dimerized and gapped.
Dzyaloshinskii-Moriya interaction¶
sc = spinchain.Spin_Chain(["S=1/2"]*n)
D = 0.3
h = sum(sc.SS(i, i+1) + D*(sc.Sx[i]*sc.Sy[i+1] - sc.Sy[i]*sc.Sx[i+1]) for i in range(n-1))
Spin-1 chain with anisotropy and biquadratic exchange¶
s1 = spinchain.Spin_Chain(["S=1"]*n)
D, beta = 0.2, -0.3
h = sum(s1.SS(i, i+1) + beta*s1.SS(i, i+1)*s1.SS(i, i+1) for i in range(n-1))
h = h + D*sum(s1.Sz[i]*s1.Sz[i] for i in range(n))
\(\beta=D=0\) is the Haldane chain, \(\beta=-1/3\) the AKLT point, and a large \(D>0\) gives the trivial large-\(D\) phase.
Hubbard chain¶
fc = fermionchain.Spinful_Fermionic_Chain(n)
t, U = 1.0, 4.0
h = -t*sum(fc.Cdagup[i]*fc.Cup[i+1] + fc.Cdagdn[i]*fc.Cdn[i+1] for i in range(n-1))
h = h + h.get_dagger()
h = h + U*sum((fc.Nup[i]-0.5)*(fc.Ndn[i]-0.5) for i in range(n))
The shifted interaction puts the chemical potential at zero, so the ground state is at half filling. With \(U\sum_in_{i\uparrow}n_{i\downarrow}\) instead, fix the filling with a sector.
Interacting spinless fermions¶
fc = fermionchain.Fermionic_Chain(n)
t, V = 1.0, 2.0
h = -t*sum(fc.Cdag[i]*fc.C[i+1] for i in range(n-1))
h = h + h.get_dagger()
h = h + V*sum((fc.N[i]-0.5)*(fc.N[i+1]-0.5) for i in range(n-1))
This is the XXZ chain in fermionic language, with \(t=J/2\) and \(V=J\Delta\). For \(V>2t\) the chain is a charge density wave.
Kitaev chain¶
fc = fermionchain.Fermionic_Chain(n)
t, Delta, mu = 1.0, 0.5, 0.3
h = sum(-t*fc.Cdag[i]*fc.C[i+1] + Delta*fc.C[i]*fc.C[i+1] for i in range(n-1))
h = h + h.get_dagger()
h = h - mu*sum(fc.N)
For \(|\mu|<2t\) the open chain has Majorana end modes and its two lowest
states, of opposite fermion parity, are split only by \(e^{-n/\xi}\). DMRG
then returns a superposition of the two, and the
parity of the state is not \(\pm1\) (-0.06
on 8 sites in our run). Take the two lowest states from
get_excited_states(n=2) and measure the parity of each, or use
mode="ED" on a short chain.
SSH chain¶
fc = fermionchain.Fermionic_Chain(n)
t1, t2 = 1.0, 0.5
h = sum((t1 if i % 2 == 0 else t2)*fc.Cdag[i]*fc.C[i+1] for i in range(n-1))
h = h + h.get_dagger()
The chain is topological, with a zero mode at each end, when the weak bond is at the edge; see the Zak phase in Topological invariants.
Bose-Hubbard chain¶
bc = bosonchain.Bosonic_Chain(n, maxnb=[4]*n) # at most 3 bosons per site
t, U, mu = 1.0, 2.0, 1.0
h = -t*sum(bc.Adag[i]*bc.A[i+1] for i in range(n-1))
h = h + h.get_dagger()
h = h + 0.5*U*sum(bc.N[i]*(bc.N[i]-1) for i in range(n)) - mu*sum(bc.N)
maxnb cuts the local space; check that \(\langle n_i\rangle\) and the
energy do not change when it is increased.
\(\mathbb Z_3\) clock model¶
pc = parafermionchain.Parafermionic_Chain(n, Z=3)
J, f = 1.0, 0.7
h = -J*sum(pc.Sigd[i]*pc.Sig[i+1] for i in range(n-1)) - f*sum(pc.Tau)
h = h + h.get_dagger()
Sig, Sigd, Tau, Taud are the clock and shift operators and their
adjoints. The model is self-dual at \(f=J\), where it is critical.
Kondo lattice¶
The local moments are spin sites between the fermion sites of a mixed chain, each coupled to the two conduction sites next to it:
mc = mixedchain.Mixed_Spin_Fermion_Chain(["F", "1/2", "F", "1/2", "F"]) # c S c S c
t, JK = 1.0, 0.8
h = -t*(mc.Cdagup[0]*mc.Cup[2] + mc.Cdagdn[0]*mc.Cdn[2]
+ mc.Cdagup[2]*mc.Cup[4] + mc.Cdagdn[2]*mc.Cdn[4])
h = h + h.get_dagger()
h = h + JK*(mc.SS(0, 1) + mc.SS(1, 2) + mc.SS(2, 3) + mc.SS(3, 4))
mc.SS(i, j) is \(\mathbf S_i\cdot\mathbf S_j\) with the fermion spin on a
fermion site and the local moment on a spin site. See
examples/fermion_models/mixed_spin_fermion_chain.
Recipes¶
Central charge of the critical Ising chain¶
For \(H=\sum_iS_i^zS_{i+1}^z+\tfrac12\sum_iS_i^x\), at the critical field, the entanglement entropy of the open chain follows the Calabrese-Cardy formula with \(c=1/2\) (Entanglement):
sc.maxm = 200 # critical chains need a large bond dimension
wf = sc.get_gs()
c = wf.get_CFT_central_charge()
Haldane gap of the spin-1 chain¶
The open chain has four low states from the two spin-1/2 edge states. The bulk gap is the fifth state (Excited states):
s1 = spinchain.Spin_Chain(["S=1"]*n)
s1.set_hamiltonian(sum(s1.SS(i, i+1) for i in range(n-1)))
es = s1.get_excited(n=6)
gap = es[4] - es[0]
Gaps of a Hubbard chain¶
Here fc is the Hubbard chain above. get_gap() is the gap to the first
excited state, and get_charge_gap(d=...) is \(E_0(N+d)-E_0(N)\)
(Excited states):
fc.get_gap()
fc.get_charge_gap(d=1)
fc.get_charge_gap(d=2)
Fidelity susceptibility across the Ising transition¶
import numpy as np
from dmrgpy import fidelity
h0 = 4*sum(sc.Sz[i]*sc.Sz[i+1] for i in range(n-1)) # Ising coupling
h1 = 2*sum(sc.Sx[i] for i in range(n)) # transverse field
lambdas = np.linspace(0.5, 1.5, 11)
chis = [fidelity.get_fidelity(sc, h0, h1, lam) for lam in lambdas]
Dynamical structure factor \(S(q,\omega)\)¶
See Dynamical correlators. We compute it from the site-resolved correlators and a Fourier transform, \(S(q,\omega)=\frac1n\sum_{ij}e^{iq(i-j)}S_{ij}(\omega)\):
es = np.linspace(-0.5, 3, 200)
q = np.pi
Sqw = 0
for i in range(n):
for j in range(n):
x, y = sc.get_dynamical_correlator(submode="KPM", es=es, delta=0.1,
name=(sc.Sz[i], sc.Sz[j]))
Sqw = Sqw + np.exp(1j*q*(i-j))*y/n
Extrapolation to the thermodynamic limit¶
See Post-processing tools. Here we extrapolate the
energy per site computed on several lengths Ls:
from dmrgpy import extrapolate
e_inf = extrapolate.size_extrapolation(Ls, e_per_site, n=1)