Luttinger Liquid Preview

A Luttinger liquid is the universal low-energy phase of a broad class of gapless, compressible, one-dimensional quantum fluids. Its elementary long-wavelength excitations are collective waves of density and phase. In a generic interacting liquid, a microscopic particle insertion creates a correlated superposition of those waves. The free spinless-fermion fixed point is a special exception with a single-particle pole; the collective waves themselves are sharp excitations of the linear fixed-point theory.

For one gapless mode, the fixed-point Hamiltonian requires only two parameters:

$$H_{\mathrm{LL}} = \frac{\hbar v}{2\pi} \int dx\, \left[ K(\partial_x\theta)^2 + \frac1K(\partial_x\phi)^2 \right].$$

The velocity $v$ sets the light-cone slope of the collective mode. The dimensionless Luttinger parameter $K$ sets the relative cost of phase and density distortions and controls an entire family of correlation exponents.

This compact theory is powerful precisely because it forgets most microscopic details. Weakly interacting fermions, strongly interacting bosons, spin chains, and lattice particles can flow to the same operator structure while retaining different values of $v$ and $K$.

The price of that universality is a strict domain:

• one spatial dimension;
• at least one gapless sector;
• interactions short-ranged enough for constant infrared parameters;
• energies and momenta below microscopic crossover scales.

Outside that domain, a gap, additional modes, disorder, long-range forces, curvature, boundaries, or dimensional crossover can change the answer.

Helpful background. Low-Dimensional Quantum Gases explains confinement and infrared fluctuations; Field Operators supplies the second-quantized particle fields. Fermi Liquid Theory introduces the finite-residue pole used for comparison.

Why One Dimension Is Special

A Fermi surface becomes two points

For spinless fermions in one dimension, the zero-temperature Fermi sea occupies an interval,

$$-k_{\mathrm F} < k < k_{\mathrm F}.$$

Its low-energy boundary consists of only two Fermi points:

$$k \simeq +k_{\mathrm F} \qquad \text{or} \qquad k \simeq -k_{\mathrm F}.$$

A small momentum transfer can therefore connect many low-energy states near one branch, while momentum near $2k_{\mathrm F}$ connects the two branches. The same restricted kinematics repeatedly exposes an excitation to forward and backscattering processes.

No angular escape

In three dimensions, two particles near a Fermi surface can often scatter into many angular directions. In one dimension, the ordering of particles along the line and the two-branch kinematics leave far fewer alternatives. Density fluctuations generated by one particle cannot simply spread around it in transverse directions.

This does not mean that every one-dimensional interaction is strong. It means that even a weak interaction can reorganize the asymptotic low-energy expansion.

Many low-energy modes share the particle weight

A particle insertion excites arbitrarily many soft density modes. In the generic interacting spinless case this spreads its spectral weight into a continuum, and the finite-residue pole disappears. The correlation exponents below make this statement quantitative.

This resembles orthogonality after a local perturbation, but the relevant overlap must be chosen correctly. Ground states before and after a static scattering potential can have a vanishing overlap within the same particle-number sector. In contrast, an unmodified $N$-particle state and an $(N+1)$-particle state are already exactly orthogonal by number conservation; that elementary fact does not diagnose a quasiparticle residue. The residue instead involves a matrix element of the creation operator between those sectors.

Collective variables are economical

Long-wavelength density fluctuations can be described by their displacement and phase fields. Once written in those variables, the interacting low-energy theory is quadratic even though microscopic particle operators have nontrivial power-law correlations.

Density and Phase Fields

Bosonization formulas are convention-sensitive. Every exponent below is tied to the normalization stated here.

Fields and equal-time commutator

Use two compact, real fields $\phi(x)$ and $\theta(x)$ with

$$\left[ \phi(x), \partial_y\theta(y) \right] = i\pi\delta(x-y).$$

The smooth density fluctuation and current are

$$\delta\rho(x) = -\frac1\pi \partial_x\phi(x),$$ $$j(x) = \frac{vK}{\pi} \partial_x\theta(x).$$

The sign of $\theta$ is not universal across the literature. With the convention above, the annihilation part of a particle field contains $e^{+i\theta}$.

Hamiltonian

The universal one-mode Hamiltonian is

$$H_{\mathrm{LL}} = \frac{\hbar v}{2\pi} \int dx\, \left[ K(\partial_x\theta)^2 + \frac1K(\partial_x\phi)^2 \right].$$

Stability requires

$$v>0, \qquad K>0.$$

The cutoff $a$ used below is a short-distance regulator of order a lattice spacing, interaction range, healing length, or inverse bandwidth scale. It is not universal.

Momentum and frequency

The symbol $q$ denotes wave number and $\omega$ denotes angular frequency:

$$p = \hbar q, \qquad E = \hbar\omega.$$

The linear theory applies when

$$|q|a \ll 1, \qquad \hbar|\omega| \ll E_{\mathrm{UV}},$$

where $E_{\mathrm{UV}}$ is a microscopic crossover energy.

What is universal

The existence of the Gaussian fixed-point form, scaling dimensions, exponent relations, and ratios that eliminate nonuniversal amplitudes are universal low-energy statements.

The values of

$$v, \qquad K, \qquad a, \qquad A_m$$

are not all universal. In particular, the amplitudes $A_m$ multiplying oscillatory density harmonics depend on microscopic physics.

From Microscopic Fermions to Two Branches

Linearization

Resolve a slowly varying field into right- and left-moving pieces:

$$\psi(x) \simeq e^{ik_{\mathrm F}x}\psi_R(x) + e^{-ik_{\mathrm F}x}\psi_L(x).$$

For a smooth microscopic dispersion,

$$\epsilon(k) -\mu \simeq \begin{cases} +\hbar v_{\mathrm F}(k-k_{\mathrm F}), & k\simeq+k_{\mathrm F}, \\[4pt] -\hbar v_{\mathrm F}(k+k_{\mathrm F}), & k\simeq-k_{\mathrm F}. \end{cases}$$

The free low-energy Hamiltonian is

$$H_0 = -i\hbar v_{\mathrm F} \int dx\, \left( \psi_R^\dagger\partial_x\psi_R - \psi_L^\dagger\partial_x\psi_L \right).$$

Linearization introduces a branch cutoff $\Lambda$:

$$|k-rk_{\mathrm F}| < \Lambda, \qquad r=\pm1.$$

Results at distances $x\gg\Lambda^{-1}$ cannot depend on the detailed shape of that cutoff, although nonuniversal amplitudes do.

Forward-scattering couplings

In a minimal spinless forward-scattering model, $g_4$ couples density fluctuations on the same branch and $g_2$ couples opposite branches. A common convention gives

$$v_N = v_{\mathrm F} + \frac{g_4+g_2}{2\pi\hbar},$$ $$v_J = v_{\mathrm F} + \frac{g_4-g_2}{2\pi\hbar}.$$

The collective parameters are then

$$v = \sqrt{v_Nv_J},$$ $$K = \sqrt{ \frac{v_J}{v_N} }.$$

For repulsive forward scattering with $g_2>0$, this convention normally gives

$$K<1.$$

For attractive forward scattering, it normally gives

$$K>1.$$

These formulas are a weak-coupling matching relation, not the definition of a Luttinger liquid. Exchange conventions and the precise definition of $g_2$ and $g_4$ vary. Backscattering and Umklapp processes must be analyzed separately because they can renormalize the parameters or open a gap.

Top: tangent lines to a quadratic band at its two zero-energy Fermi points define slowly varying left and right fields. Their fast phases occur only in the reconstruction shown. Bottom: in a gapless spinful liquid, two right-moving responses launched at the same point separate by $(v_c-v_s)t$ when $v_c>v_s>0$. The illustrative slopes use $v_c=2v_s$; neither their ordering nor that ratio is universal. Left-moving contributions and packet broadening are omitted. Oppositely moving packets would instead separate at the sum of their speed magnitudes.

The Harmonic-Fluid Theory

Equations of motion

The commutator and Hamiltonian give

$$\partial_t\phi = vK\,\partial_x\theta,$$ $$\partial_t\theta = \frac vK \partial_x\phi.$$

Applying another time derivative yields

$$\left( \partial_t^2 - v^2\partial_x^2 \right) \phi = 0,$$ $$\left( \partial_t^2 - v^2\partial_x^2 \right) \theta = 0.$$

Every oscillator therefore has the linear dispersion

$$\omega(q) = v|q|.$$

Continuity equation

Using

$$\delta\rho = -\frac1\pi\partial_x\phi,$$

one finds

$$\partial_t\delta\rho = -\frac{vK}{\pi} \partial_x^2\theta.$$

With

$$j = \frac{vK}{\pi} \partial_x\theta,$$

this becomes

$$\partial_t\delta\rho + \partial_xj = 0.$$

The density mode is therefore not an analogy imposed from outside. It is the collective carrier of the conserved particle number.

Density and current stiffnesses

It is often useful to write

$$v_N = \frac vK, \qquad v_J = vK.$$

Then

$$H_{\mathrm{LL}} = \frac{\hbar}{2\pi} \int dx\, \left[ v_J(\partial_x\theta)^2 + v_N(\partial_x\phi)^2 \right].$$

The inverse relations are

$$v = \sqrt{v_Nv_J}, \qquad K = \sqrt{ \frac{v_J}{v_N} }.$$

The quantity $v_N$ measures the energy cost of changing the density, while $v_J$ measures the energy cost of a phase twist or persistent current. Their geometric mean fixes propagation; their ratio fixes the interaction-dependent scaling.

Thermodynamic and Response Meaning of the Parameters

Compressibility

For a uniform density change $\delta n$, the field has

$$\partial_x\phi = -\pi\delta n.$$

After subtracting the linear chemical-potential term $\mu L\delta n$, the quadratic energy density is

$$\frac{\Delta E}{L} = \frac{\pi\hbar v}{2K} (\delta n)^2.$$

Comparing with

$$\frac{\Delta E}{L} = \frac12 \frac{\partial\mu}{\partial n} (\delta n)^2$$

gives

$$\frac{\partial n}{\partial\mu} = \frac{K}{\pi\hbar v}.$$

The conventional isothermal compressibility per equilibrium density $n$ is

$$\kappa = \frac1{n^2} \frac{\partial n}{\partial\mu} = \frac{K}{\pi\hbar v n^2}.$$

These formulas apply to the single density mode in the normalization used here. Spin degeneracy or a different charge-field normalization changes prefactors.

Stiffness

A uniform phase twist has

$$\partial_x\theta = \frac{\Phi}{L}.$$

Within a fixed winding branch and for a sufficiently small dimensionless twist $\Phi$, its energy cost is

$$\Delta E = \frac{\hbar vK}{2\pi L} \Phi^2.$$

Thus a twist or Drude-weight measurement probes the product

$$vK = v_J,$$

whereas compressibility probes

$$\frac K v = \frac1{v_N}.$$

Measuring both determines $v$ and $K$ without fitting a correlation exponent.

Galilean-invariant continuum

For one spinless density mode in a Galilean-invariant continuum,

$$vK = \frac{\pi\hbar n}{m}.$$

This relation follows because a uniform phase gradient is a center-of-mass boost and the total current is tied to conserved momentum. A lattice breaks continuous Galilean invariance, so the same formula need not hold there.

Density response

For an external scalar potential coupled as

$$\delta H = \int dx\, U(x,t)\rho(x,t),$$

the retarded smooth-density polarization for this positive-potential source is

$$\Pi_{\rho\rho}^{R}(q,\omega) = \frac{Kvq^2} {\pi\hbar \left[ (\omega+i0^+)^2-v^2q^2 \right] }.$$

The poles lie at

$$\omega = \pm v|q|.$$

The static limit is

$$\Pi_{\rho\rho}^{R}(q,0) = -\frac K{\pi\hbar v}.$$

The minus sign appears because a positive external potential raises the local particle energy and reduces the density. For the source $-\int dx\,f\rho$ used in Susceptibilities, the physical susceptibility is $\chi^R=-\Pi^R$. The displayed static limit takes $\omega\to0$ before $q\to0$; at exactly $q=0$, conserved total number has no nonzero-frequency retarded response.

Bosonization Dictionary

Bosonization is an operator correspondence between the low-energy sectors of one-dimensional particle theories and a compact bosonic field theory. It does not claim that microscopic fermions have changed their exchange statistics.

Chiral fermion operators

For spinless fermions,

$$\begin{aligned} \psi_r(x) ={}& \frac{\eta_r}{\sqrt{2\pi a}} e^{-i[r\phi(x)-\theta(x)]}, \\ r ={}& \pm1. \end{aligned}$$

These are the slowly varying fields: the factors $e^{\pm ik_{\mathrm F}x}$ appear once, when reconstructing $\psi(x)=e^{ik_{\mathrm F}x}\psi_R+e^{-ik_{\mathrm F}x}\psi_L$.

Here:

• $r=+1$ labels the right branch and $r=-1$ the left branch;
• $a$ is a short-distance cutoff;
• $\eta_r$ are anticommuting Hermitian Klein factors in the local oscillator convention;
• the vertex operator is defined with a regulator and an ordering prescription.

The Klein factors enforce anticommutation between distinct branches:

$$\{\eta_r,\eta_{r'}\} = 2\delta_{rr'}.$$

They enforce relative branch signs. A finite-ring operator that changes particle number must also shift the appropriate number and winding zero modes; a fully finite-size construction often combines these shifts with non-Hermitian Klein ladder operators. Hermitian branch factors alone do not implement all global-sector changes (von Delft and Schoeller, 1998).

Density harmonics

The full long-distance density expansion is

$$\rho(x) ={} \rho_0 - \frac1\pi\partial_x\phi(x) + \sum_{m=1}^{\infty} A_m \cos \left( 2m[\pi\rho_0x-\phi(x)] \right).$$

For spinless fermions,

$$k_{\mathrm F} = \pi\rho_0,$$

so the first oscillatory harmonic has wave number

$$2k_{\mathrm F}.$$

The smooth term is fixed by number conservation. The amplitudes $A_m$ are nonuniversal.

Bosonic particle field

For a one-dimensional bosonic fluid, the annihilation field has the asymptotic representation

$$\Psi_B(x) \sim{} \left[ \rho_0 - \frac1\pi\partial_x\phi(x) \right]^{1/2} \times e^{+i\theta(x)} \times \sum_{m=-\infty}^{\infty} B_m e^{2im[\pi\rho_0x-\phi(x)]}.$$

The creation field is the Hermitian conjugate and contains $e^{-i\theta}$. The coefficients $B_m$ depend on the cutoff and microscopic model.

Fermionic and bosonic fields differ in their allowed vertex combinations, exchange phases, and harmonic content. The quadratic Hamiltonian can nevertheless be the same.

Compactness

The density harmonics are unchanged under

$$\phi \longrightarrow \phi+\pi.$$

Particle-number quantization and boundary conditions also constrain windings of $\phi$ and $\theta$. Treating the fields as unconstrained real variables can reproduce local oscillator correlators while losing zero-mode sectors, parity rules, and the allowed operator lattice.

Vertex Operators and Scaling Dimensions

Consider

$$\mathcal O_{m,n}(x) = e^{i[m\phi(x)+n\theta(x)]}.$$

In the present normalization, its bulk scaling dimension is

$$\Delta_{m,n} = \frac14 \left( m^2K + \frac{n^2}{K} \right).$$

At equal time,

$$\left\langle \mathcal O_{m,n}(x) \mathcal O_{m,n}^{\dagger}(0) \right\rangle \propto \frac1{|x|^{2\Delta_{m,n}}}.$$

Important cases are

$$\Delta[e^{2i\phi}] = K,$$ $$\Delta[e^{2i\theta}] = \frac1K.$$

The reciprocal appearance of $K$ expresses density–phase duality: making density fluctuations costly makes phase fluctuations softer, and conversely.

The displayed dimensions assume the single-mode Hamiltonian above. Rescaling the fields changes the numerical labels $m$, $n$, and $K$ together while leaving physical exponents invariant.

Correlations in a Spinless Fermion Liquid

Density correlations

At zero temperature and distances $|x|\gg a$,

$$\left\langle \delta\rho(x) \delta\rho(0) \right\rangle \sim{} -\frac{K}{2\pi^2x^2} + C_1 \frac{\cos(2k_{\mathrm F}x)} {|x|^{2K}} + \cdots.$$

The first term is the universal smooth-density contribution. The coefficient $C_1$ is nonuniversal, but its exponent is fixed by $K$.

Higher harmonics behave as

$$\frac{\cos(2mk_{\mathrm F}x)} {|x|^{2m^2K}}, \qquad m=2,3,\ldots.$$

One-particle correlation

For a parity-symmetric state, the leading equal-time correlator has a power-law envelope but also a chiral phase:

$$\langle\psi^\dagger(x)\psi(0)\rangle \sim C\,\frac{\sin(k_{\mathrm F}x)} {x\,|x|^{\eta-1}}, \qquad \eta=\frac{K+K^{-1}}2.$$

Here $C$ is nonuniversal; the formula is asymptotic for $|x|\gg a$, with subleading harmonics omitted. The factor $1/x$ carries the relative phase of the two branches. It cannot be replaced by a cosine over $|x|^\eta$: at the free point direct integration over occupied momenta gives

$$\int_{-k_{\mathrm F}}^{k_{\mathrm F}}\frac{dk}{2\pi}e^{-ikx} =\frac{\sin(k_{\mathrm F}x)}{\pi x}.$$

The arithmetic–geometric mean inequality gives

$$\frac12 \left( K+\frac1K \right) \ge 1,$$

with equality only at

$$K=1.$$

For $K\ne1$, the envelope therefore decays faster than the free spinless-fermion result. The statement concerns the envelope, not the zeros of its oscillating factor.

Density wave versus pairing

The leading $2k_{\mathrm F}$ density-wave correlation decays as

$$C_{\mathrm{CDW}}(x) \propto |x|^{-2K}.$$

A point-split odd-parity pair, whose low-energy component is $\psi_R\psi_L$, has

$$C_{\mathrm{pair}}(x) \propto |x|^{-2/K}.$$

The coincident microscopic product $\psi(x)\psi(x)$ vanishes by Pauli exclusion; the pair here is not an on-site spinless pair.

Consequently:

• For $K<1$, density-wave correlations decay more slowly.
• For $K>1$, pair correlations decay more slowly.

These are dominant correlations, not automatically broken-symmetry order. In an ordinary short-ranged one-dimensional system, both vanish at asymptotically large separation within the gapless phase.

Correlations in a One-Dimensional Bose Liquid

The leading one-body correlation is

$$\left\langle \Psi_B^\dagger(x) \Psi_B(0) \right\rangle \propto |x|^{-1/(2K)}.$$

The leading oscillatory density term is

$$\left\langle \delta\rho(x) \delta\rho(0) \right\rangle_{2\pi\rho_0} \propto \frac{ \cos(2\pi\rho_0x) } {|x|^{2K}}.$$

For the repulsive Lieb–Liniger gas, common conventions give

$$K \ge 1.$$

The strongly repulsive Tonks–Girardeau limit approaches

$$K \longrightarrow 1,$$

while weak repulsion gives larger $K$. This does not make the weakly interacting gas a true condensate in the infinite one-dimensional thermodynamic limit; the one-body correlation remains algebraic at zero temperature.

Momentum Distribution and Vanishing Residue

Define

$$\alpha = \frac12 \left( K+\frac1K-2 \right) \ge 0.$$

Near a Fermi point, the nonanalytic part of the momentum distribution behaves schematically as

$$n(k)-n(k_{\mathrm F}) \propto -\operatorname{sgn}(k-k_{\mathrm F}) |k-k_{\mathrm F}|^{\alpha}.$$

For

$$K\ne1,$$

one has

$$\alpha>0,$$

so there is no finite jump:

$$Z = n(k_{\mathrm F}^{-}) - n(k_{\mathrm F}^{+}) = 0.$$

This is the sharpest contrast with a Landau Fermi liquid, where the jump equals the quasiparticle residue in the idealized zero-temperature translationally invariant setting.

For $\alpha>1$, analytic background terms can dominate the first derivative of $n(k)$; the power law still identifies a subleading nonanalyticity. At special integer exponents, logarithmic factors can appear. A numerical fit should therefore include regular background terms rather than forcing the entire curve into one cusp.

At the exactly free point,

$$K=1, \qquad \alpha=0,$$

and the step is restored. For a fixed interaction that changes $K$ away from one, the thermodynamic and low-energy limits give $Z=0$ however small that change is. An interaction tuned to retain $K=1$, such as pure same-branch forward scattering in the linear model, is an exception; $K=1$ alone does not establish a microscopically noninteracting Hamiltonian.

Real-Time Green Function and Spectral Thresholds

For a right mover at zero temperature, define the full right-branch greater function $G_R(x,t)=-i e^{ik_{\mathrm F}x}\langle\psi_R(x,t)\psi_R^\dagger(0,0)\rangle$, including its fast phase. Its regulated asymptotic form is

$$G_R(x,t) \propto{} e^{ik_{\mathrm F}x} \left[ \frac{a} {a+i(vt-x)} \right]^{\gamma_+} \times \left[ \frac{a} {a+i(vt+x)} \right]^{\gamma_-}.$$

The time-ordered function requires different ordering prescriptions for positive and negative time; it is not obtained by treating this one greater function as valid for both orderings. Here

$$\gamma_{\pm} = \frac14 \left( K+\frac1K\pm2 \right).$$

The exponents satisfy

$$\gamma_+ + \gamma_- = \frac12 \left( K+\frac1K \right).$$

At the free point,

$$\gamma_+ = 1, \qquad \gamma_- = 0,$$

so only the right-moving light-cone factor remains.

For $K\ne1$, both factors carry noninteger powers in general. Writing $q=k-k_{\mathrm F}$, Fourier transformation produces threshold singularities and continuum weight rather than a finite-residue Landau pole of the form

$$G^R(q,\omega) \not\sim \frac{Z} {\omega-v_{\mathrm F}q+i0^+}.$$

The local tunneling density of states of a spinless bulk liquid obeys

$$\nu_{\mathrm{bulk}}(\omega) \propto |\omega|^{\alpha}.$$

At an open end,

$$\nu_{\mathrm{end}}(\omega) \propto |\omega|^{\alpha_{\mathrm{end}}},$$

with

$$\alpha_{\mathrm{end}} = \frac1K-1$$

for the simplest spinless boundary fixed point.

Bulk and boundary exponents differ because a boundary relates right and left movers. Contacts, finite temperature, finite length, and an actual tunneling circuit can further modify the measured conductance.

The exactly linear model concentrates the smooth density response on

$$\omega = v|q|.$$

Band curvature and irrelevant interactions broaden or reshape finite-energy thresholds. The resulting nonlinear-Luttinger-liquid problem is beyond the fixed-point preview; fitting an exact delta line at nonzero experimental resolution is not a test of the asymptotic theory.

Spin–Charge Separation

Charge and spin fields

For spin-$1/2$ fermions, define

$$\phi_c = \frac{ \phi_\uparrow+\phi_\downarrow }{\sqrt2}, \qquad \phi_s = \frac{ \phi_\uparrow-\phi_\downarrow }{\sqrt2},$$

and similarly

$$\theta_c = \frac{ \theta_\uparrow+\theta_\downarrow }{\sqrt2}, \qquad \theta_s = \frac{ \theta_\uparrow-\theta_\downarrow }{\sqrt2}.$$

When symmetry allows the sectors to decouple, the fixed-point Hamiltonian is

$$H = \sum_{\nu=c,s} H_\nu,$$ $$H_\nu = \frac{\hbar v_\nu}{2\pi} \int dx\, \mathcal K_\nu(x),$$

where

$$\mathcal K_\nu = K_\nu(\partial_x\theta_\nu)^2 + \frac1{K_\nu} (\partial_x\phi_\nu)^2.$$

The charge and spin waves generally have different velocities:

$$v_c \ne v_s.$$

What an injected electron creates

An electron carries both charge and spin. Its creation operator contains vertex operators from both sectors. A localized injection therefore launches a superposition of charge and spin excitations.

After time $t$, characteristic packet centers can be separated by

$$\Delta x(t) \simeq |v_c-v_s|t.$$

This is spin–charge separation. It does not mean that the electron breaks into two microscopic fragments that can be assigned independent permanent identities. It means that charge and spin correlations propagate through distinct collective sectors.

Spin-rotation symmetry

For a gapless spin sector with exact spin-rotation invariance,

$$K_s = 1$$

at the infrared fixed point in the standard spin normalization. A marginally irrelevant spin backscattering operator can generate logarithmic corrections, so finite-size data need not look like a clean power law immediately.

Spinful exponents

With a gapless spin-rotation-invariant spin sector, the bulk tunneling exponent is

$$\alpha_{\mathrm{spinful}} = \frac14 \left( K_c+\frac1{K_c}-2 \right).$$

Representative equal-time exponents are

$$C_{2k_{\mathrm F}\mathrm{\ density}}(x) \propto |x|^{-(K_c+1)},$$ $$C_{\mathrm{singlet\ pair}}(x) \propto |x|^{-(K_c^{-1}+1)},$$

up to oscillatory factors, amplitudes, and possible logarithmic corrections.

If one sector is gapped, these formulas change. A Luther–Emery liquid, for example, has one gapless charge mode but a spin gap. Calling every spinful one-dimensional metal a two-component gapless Luttinger liquid is therefore too broad.

Central Charge and Low-Temperature Thermodynamics

Each independent nonchiral compact-boson mode has

$$c=1.$$

Thus one spinless or bosonic mode has $c=1$, gapless spin and charge modes together have $c=2$, and a spinful system with one gapped sector and one gapless sector has $c=1$ in its infrared theory.

For one mode on an infinite line, the low-temperature free-energy density is

$$f(T) = e_0 - \frac{\pi c}{6} \frac{(k_{\mathrm B}T)^2} {\hbar v} + o(T^2).$$

The specific heat per length is

$$\frac{C_V}{L} = \frac{\pi c}{3} \frac{k_{\mathrm B}^2T} {\hbar v} + o(T).$$

For several decoupled modes,

$$\frac{C_V}{L} = \frac{\pi k_{\mathrm B}^2T}{3\hbar} \sum_\nu \frac{c_\nu}{v_\nu} + o(T).$$

A linear heat capacity alone does not uniquely identify a Luttinger liquid, but its coefficient provides a stringent consistency test when $v$ is known independently.

Finite Temperature

Define the thermal length scale

$$\ell_T = \frac{\hbar v} {\pi k_{\mathrm B}T}.$$

For an operator of scaling dimension $\Delta$, the zero-temperature power law maps to

$$\left( \frac a{|x|} \right)^{2\Delta} \longrightarrow \left[ \frac{ \pi a k_{\mathrm B}T/(\hbar v) } { \sinh( \pi k_{\mathrm B}T|x|/(\hbar v) ) } \right]^{2\Delta}.$$

At short separation,

$$|x| \ll \ell_T,$$

the zero-temperature power law is recovered. At long separation,

$$|x| \gg \ell_T,$$

the same correlator decays exponentially:

$$C_{\mathcal O}(x,T) \propto \exp \left[ -\frac{ 2\pi\Delta k_{\mathrm B}T }{\hbar v} |x| \right].$$

Different operators have different $\Delta$, so there is no unique correlation length shared by every observable. The scale $\ell_T$ marks the common quantum-to-thermal crossover.

For spin and charge sectors with different velocities, each factor carries its own thermal length:

$$\ell_{T,\nu} = \frac{\hbar v_\nu} {\pi k_{\mathrm B}T}.$$

Finite Size and Zero Modes

Ground-state Casimir term

For one periodic gapless mode on a ring of circumference $L$, in the conformal vacuum sector,

$$E_0(L) = Le_\infty - \frac{\pi\hbar vc}{6L} + o(L^{-1}).$$

The coefficient contains the central charge, but extraction requires an independently known velocity and the correct vacuum sector. A forced nonzero winding or particle-number offset adds a positive term of order $1/L$; fitting it as part of the Casimir term would give the wrong central charge.

For an open chain, the boundary condition changes the coefficient. One must not fit periodic data with the open-chain formula or conversely.

Number and current sectors

Let $\Delta N$ be the particle-number change relative to a reference sector and let $J$ be the winding or current quantum number. After subtracting $\mu\Delta N$, a common periodic-ring excitation spectrum is

$$\Delta E ={} \frac{\pi\hbar v}{2L} \left[ \frac{(\Delta N)^2}{K} + KJ^2 \right] + \frac{2\pi\hbar v}{L} \left( N_R+N_L \right).$$

The nonnegative integers $N_R$ and $N_L$ count total right- and left-moving oscillator levels. This formula displays three distinct probes:

$$\begin{array}{c|c} \text{number change} & v/K \\[3pt] \text{current or twist} & vK \\[3pt] \text{oscillator spacing} & v \end{array}$$

The allowed combinations of $\Delta N$ and $J$ depend on statistics, microscopic filling, boundary conditions, and parity sector. The quadratic energy is universal; the selection rules are part of the compactification data and must be matched to the microscopic model.

Entanglement check

A periodic $c=1$ liquid has interval entropy

$$S(\ell,L) = \frac13 \ln \left[ \frac{L}{\pi a} \sin \left( \frac{\pi\ell}{L} \right) \right] + s_1$$

at leading order. This is a valuable central-charge diagnostic, but oscillatory and marginal corrections can bias short-chain fits. Entanglement Entropy gives boundary variants, Rényi formulas, and finite-size cautions.

Perturbations Around the Fixed Line

The Luttinger Hamiltonian is a fixed-point theory. Microscopic symmetries determine which additional operators are allowed. Their scaling dimensions determine whether they fade, remain marginal, or grow under coarse-graining.

Bulk cosine

Consider

$$\delta H_p = g_p \int dx\, \cos(2p\phi).$$

Its bulk scaling dimension is

$$\Delta_p = p^2K.$$

Writing $y_p$ for the dimensionless version of the dimensional Hamiltonian coefficient $g_p$, its leading flow is

$$\frac{dy_p}{d\ell} = \left( 2-p^2K \right)y_p + \cdots.$$

Therefore:

$$\begin{array}{c|c} p^2K<2 & \text{relevant} \\[3pt] p^2K>2 & \text{irrelevant} \\[3pt] p^2K=2 & \text{marginal at tree level} \end{array}$$

At marginality, higher-order flow decides the outcome. A relevant cosine can pin $\phi$ near one of its minima and gap the corresponding density sector.

Spinless half-filled Umklapp

For a translation-invariant spinless lattice at half filling, the leading allowed Umklapp term has

$$\delta H_{\mathrm U} \propto \int dx\, \cos(4\phi).$$

Its dimension is

$$\Delta_{\mathrm U} = 4K.$$

It becomes relevant when

$$K < \frac12.$$

At a continuous interaction-driven transition of this type, the long-distance parameter approaches $K=1/2$ from the liquid side. This universal value should not be imposed on an arbitrary bare Gaussian coupling: the cosine and $K$ flow together. Spinless Fermion Chains applies this criterion to the $t$–$V$ model.

Commensurate bosons

For bosons at the simplest integer commensurability in a periodic lattice, the leading pinning term is commonly

$$\delta H_{\mathrm{lat}} \propto \int dx\, \cos(2\phi).$$

Its dimension is

$$\Delta_{\mathrm{lat}} = K,$$

so the corresponding BKT threshold occurs at

$$K=2$$

in this normalization for the interaction-driven transition at fixed integer filling. This is distinct from doping out of a Mott lobe: the commensurate–incommensurate transition has limiting $K=1$ for the simplest integer case. Different commensurabilities produce higher harmonics and different thresholds.

Pairing perturbation

If particle-number conservation is explicitly broken, for example by proximity coupling, a spinless pairing perturbation can contain

$$\delta H_{\mathrm{pair}} \propto \int dx\, \cos(2\theta).$$

Its dimension is

$$\Delta_{\mathrm{pair}} = \frac1K.$$

It is relevant at tree level when

$$K > \frac12.$$

The density-pinning and phase-pinning cosines are dual competitors. They cannot both be treated as simultaneously sharp classical fields because $\phi$ and $\theta$ are conjugate.

Single impurity

A local backscattering impurity at $x=0$ produces

$$\delta H_{\mathrm{imp}} = \lambda \cos[2\phi(0)].$$

As a boundary operator, it has dimension

$$\Delta_{\mathrm{imp}} = K.$$

Its leading flow is

$$\frac{d\lambda}{d\ell} = (1-K)\lambda + \cdots.$$

Thus a weak impurity grows for repulsive spinless liquids with $K<1$. The dual weak-link amplitude $t_{\mathrm{link}}$ has boundary dimension

$$\Delta_{\mathrm{link}} = \frac1K$$

and flow

$$\frac{dt_{\mathrm{link}}}{d\ell} = \left( 1-\frac1K \right)t_{\mathrm{link}} + \cdots.$$

This weak-barrier/weak-link duality describes the asymptotic isolated Luttinger liquid. Leads, finite length, multiple channels, resonant levels, and additional symmetries can change the measured transport crossover.

Spin and charge gaps

In a spinful system, backscattering can generate a cosine in the spin sector, while commensurate Umklapp can generate one in the charge sector. Depending on its sign and scaling flow, a cosine may:

• remain marginally irrelevant and leave the sector gapless;
• become relevant and pin the field;
• produce a spin gap with gapless charge;
• produce a charge gap with gapless or gapped spin;
• leave two gapless sectors.

The phrase “Luttinger liquid” should therefore be accompanied by the number and identity of the gapless modes.

Representative Microscopic Realizations

Repulsive contact bosons

The Lieb–Liniger Model Preview treats continuum bosons with repulsive delta-function interactions. Its low-energy phase is a one-mode Luttinger liquid throughout the repulsive regime. Exact Bethe-ansatz thermodynamics determines $v$ and $K$, and the relations above then determine long-distance exponents.

Spinless lattice fermions

The half-filled $t$–$V$ chain is a Luttinger liquid between phase separation and charge-density-wave order. Its free point has $K=1$; repulsion lowers $K$, and the commensurate charge-order threshold is reached at $K=1/2$.

XXZ spin chain

The easy-plane spin-$1/2$ XXZ chain has one gapless compact-boson mode. Spin operators map to uniform and staggered vertex operators. The exactly known dependence of $K$ and $v$ on anisotropy belongs to XXZ Spin Chain.

One-dimensional Bose–Hubbard superfluid

The compressible side of the one-dimensional Bose–Hubbard model is described by a Luttinger liquid. At commensurate filling, a lattice cosine can become relevant and drive a Mott transition. The universal threshold is stated above; Bose–Hubbard Chain develops the one-dimensional Hamiltonian, exact limits, and finite-size calculations, while the general lattice treatment remains at Bose–Hubbard Model.

Repulsive Hubbard chain away from commensurability

Away from a charge-gapping commensurability, the repulsive one-dimensional Hubbard model has separate gapless charge and spin sectors. Spin-rotation symmetry fixes the infrared spin normalization, while interactions renormalize the charge parameter and separate the velocities.

Quantum wires and cold-atom tubes

Semiconductor wires, carbon nanotubes, atomic gases confined to tubes, and chains in crystalline solids can display Luttinger-liquid scaling over a finite window. Real samples also contain contacts, transverse modes, disorder, long-range Coulomb interactions, and finite temperature. Demonstrating the fixed point requires showing a common parameter set across several observables, not merely fitting one power law.

How to Extract the Luttinger Data

Small-momentum structure factor

Define the equal-time continuum structure factor per length by

$$s(q) = \frac1L \left\langle \delta\rho_q \delta\rho_{-q} \right\rangle,$$

with

$$\delta\rho_q = \int_0^L dx\, e^{+iqx}\delta\rho(x).$$

The fixed point predicts

$$s(q) = \frac{K|q|}{2\pi} + o(|q|).$$

Hence

$$K = 2\pi \lim_{q\to0} \frac{s(q)}{|q|}.$$

On a lattice, the definition of density per site and the Fourier normalization must be translated before applying this prefactor.

Compressibility and stiffness

The two independent combinations are

$$\frac K v = \pi\hbar \frac{\partial n}{\partial\mu},$$ $$vK = v_J.$$

Their product and ratio give

$$K = \sqrt{ \left( \pi\hbar \frac{\partial n}{\partial\mu} \right) v_J },$$ $$v = \sqrt{ \frac{ v_J }{ \pi\hbar \partial n/\partial\mu } }.$$

The dimensions of $v_J$ depend on the stiffness convention. The equations above use the velocity units defined by the Hamiltonian.

Finite-size levels

The oscillator spacing gives

$$\Delta E_{\mathrm{osc}} \simeq \frac{2\pi\hbar v}{L}.$$

Number and winding gaps then separate $v/K$ and $vK$. This method is powerful in exact diagonalization and density-matrix renormalization group calculations, provided the correct symmetry and parity sectors are identified.

Correlation exponents

One can fit several predictions:

$$\eta_{2k_{\mathrm F}} = 2K,$$ $$\eta_{\mathrm{pair}} = \frac2K,$$ $$\eta_{\mathrm{fermion}} = \frac12 \left( K+\frac1K \right).$$

A reliable identification asks whether one value of $K$ explains all accessible exponents after finite-size, boundary, temperature, and logarithmic corrections are included.

Velocity-resolved response

The slope of a low-$q$ density mode gives $v$. In a spinful system, charge- and spin-sensitive probes can reveal separate slopes:

$$\omega_c(q) \simeq v_c|q|,$$ $$\omega_s(q) \simeq v_s|q|.$$

Spectral continua and matrix elements can obscure a simple line, so one should track thresholds and integrated response rather than only the brightest pixel.

Central charge

Finite-size ground-state energies and entanglement scaling can test

$$c=1$$

per gapless nonchiral mode. Central charge counts modes; it does not determine $K$. Two $c=1$ liquids can have different exponents.

Where the Linear Theory Stops

Band curvature

The exact microscopic dispersion is not linear:

$$\epsilon(k) -\mu = \hbar v_{\mathrm F}\delta k + \frac{\hbar^2}{2m_*} (\delta k)^2 + \cdots.$$

Curvature is irrelevant in the strict fixed-point sense but reorganizes finite-energy threshold singularities. It can determine decay channels, line shapes, and asymmetry near spectral edges.

Gapped phases

If a relevant cosine pins a field, the asymptotic theory is no longer the gapless Gaussian Hamiltonian. Correlations of some operators become exponential, solitons or massive particles can replace the sound mode, and the $c=1$ finite-size formulas cease to apply below the gap.

Long-range interactions

Unscreened or weakly screened Coulomb interactions can make the effective parameters momentum dependent:

$$K \longrightarrow K(q), \qquad v \longrightarrow v(q).$$

The resulting correlations need not be simple powers with one constant exponent. Calling such a system an ordinary short-ranged Luttinger liquid can hide physically important logarithms.

Disorder

Random backscattering introduces a spatially varying operator rather than one isolated boundary impurity. Its relevance criterion and localized phase are not obtained by applying the single-impurity result point by point. Renormalization Group Preview explains the flow; disorder-specific flows require their own treatment.

Multiple channels

Two bands, valleys, legs, or internal components produce a matrix of density and phase couplings. Diagonalization may yield several collective modes with different velocities and compactification data. A single pair $(v,K)$ is then insufficient.

Boundaries and leads

An open boundary folds right movers into left movers and changes local exponents. A clean interacting wire connected adiabatically to noninteracting reservoirs can have a two-terminal conductance governed by the leads even though its local tunneling density of states shows interaction-dependent powers.

Dimensional crossover

Interchain hopping or transverse excitation can become relevant at sufficiently low energy. A material may show Luttinger scaling over an intermediate window and cross over to higher-dimensional coherent motion, ordering, or Fermi-liquid behavior at longer scales.

Chiral edge liquids

A fractional quantum Hall edge can be described by a chiral Luttinger theory. Its anomaly, compactification, charge assignments, and one-way propagation differ from the nonchiral right-plus-left liquid developed here. The shared word “Luttinger” does not make the two theories interchangeable.

Integrability

Integrability is not required for Luttinger-liquid universality. It is one way to calculate $v$, $K$, and exact crossover functions in special microscopic models. Generic nonintegrable short-ranged systems can flow to the same fixed point.

Worked Checks

Recover the free spinless point

Set

$$K=1, \qquad v=v_{\mathrm F}.$$

Then

$$\gamma_+ = 1, \qquad \gamma_- = 0,$$

and

$$\alpha = 0.$$

The right-moving Green function reduces to one chiral denominator, the momentum distribution regains a jump, and the density and pairing exponents are both $2$. The fixed-point Hamiltonian therefore contains the free gas as one special point, not as a separate formalism.

A repulsive example

Take a spinless liquid with

$$K = \frac34.$$

The leading density-wave exponent is

$$\eta_{\mathrm{CDW}} = 2K = \frac32,$$

while the pairing exponent is

$$\eta_{\mathrm{pair}} = \frac2K = \frac83.$$

Density correlations decay more slowly. The momentum cusp exponent is

$$\alpha = \frac12 \left( \frac34 + \frac43 - 2 \right) = \frac1{24}.$$

The exponent is small, so a finite system may look almost discontinuous even though the true thermodynamic jump is zero.

Reconstruct the velocity and Luttinger parameter

Suppose finite-size energies yield

$$v_N = 1.8u, \qquad v_J = 0.8u,$$

where $u$ is a velocity unit. Then

$$v = \sqrt{1.44}\,u = 1.2u,$$

and

$$K = \sqrt{ \frac{0.8}{1.8} } = \frac23.$$

The inferred $K<1$ is consistent with repulsive spinless behavior. The conclusion is stronger if the $2k_{\mathrm F}$ density exponent also approaches

$$2K = \frac43.$$

Exercises

Exercise 1: Wave equation and continuity

Starting from

$$\partial_t\phi = vK\partial_x\theta, \qquad \partial_t\theta = \frac vK\partial_x\phi,$$

derive the wave equation for $\phi$. Then use the density definition to identify the conserved current.

Solution

Differentiate the first equation with respect to time:

$$\partial_t^2\phi = vK \partial_x\partial_t\theta.$$

Insert the second equation:

$$\partial_t^2\phi = vK \partial_x \left( \frac vK \partial_x\phi \right) = v^2\partial_x^2\phi.$$

Therefore

$$\left( \partial_t^2-v^2\partial_x^2 \right)\phi = 0.$$

Since

$$\delta\rho = -\frac1\pi\partial_x\phi,$$

one has

$$\partial_t\delta\rho = -\frac{vK}{\pi} \partial_x^2\theta.$$

Writing this as

$$\partial_t\delta\rho + \partial_xj = 0$$

identifies

$$j = \frac{vK}{\pi} \partial_x\theta.$$

Exercise 2: Density and current velocities

A finite-size calculation gives

$$v_N = 3u, \qquad v_J = u.$$

Find $v$ and $K$. Which qualitative spinless interaction regime does the value of $K$ suggest?

Solution

The collective velocity is

$$v = \sqrt{v_Nv_J} = \sqrt3\,u.$$

The Luttinger parameter is

$$K = \sqrt{ \frac{v_J}{v_N} } = \frac1{\sqrt3}.$$

Because

$$K<1,$$

the result is characteristic of repulsive spinless interactions in this normalization. That inference can fail if the field or species convention differs, so the Hamiltonian normalization must accompany the quoted number.

Exercise 3: Free limit of the chiral correlator

Evaluate $\gamma_+$, $\gamma_-$, and $\alpha$ at $K=1$. Explain what happens to the counterpropagating factor in $G_R(x,t)$.

Solution

At $K=1$,

$$\gamma_{\pm} = \frac14 (1+1\pm2).$$

Therefore

$$\gamma_+ = 1, \qquad \gamma_- = 0.$$

The factor involving $vt+x$ is raised to the zeroth power and disappears. The right-moving correlator becomes

$$G_R(x,t) \propto e^{ik_{\mathrm F}x} \frac a{a+i(v_{\mathrm F}t-x)}.$$

Also,

$$\alpha = \frac12(1+1-2) = 0,$$

consistent with the free momentum-distribution step.

Exercise 4: A cusp without a jump

For a spinless liquid with $K=1/2$, compute the momentum-distribution exponent. Does $n(k)$ have a jump at $k_{\mathrm F}$?

Solution

The exponent is

$$\alpha = \frac12 \left( \frac12 + 2 - 2 \right) = \frac14.$$

Near the Fermi point,

$$n(k)-n(k_{\mathrm F}) \propto{} -\operatorname{sgn}(k-k_{\mathrm F}) \times |k-k_{\mathrm F}|^{1/4}.$$

The function is continuous, although its slope is singular. Hence

$$Z=0.$$

Exercise 5: Competing correlations

For $K=3/4$, compare the leading spinless density-wave and pairing exponents. Which correlation decays more slowly?

Solution

The density-wave exponent is

$$\eta_{\mathrm{CDW}} = 2K = \frac32.$$

The pairing exponent is

$$\eta_{\mathrm{pair}} = \frac2K = \frac83.$$

Because

$$\frac32 < \frac83,$$

the $2k_{\mathrm F}$ density correlation decays more slowly. This identifies the dominant algebraic tendency, not true charge-density-wave long-range order.

Exercise 6: Arrival-time separation

A probe injects an electron a distance $d$ from detectors sensitive to both charge and spin. The charge and spin velocities are $v_c$ and $v_s$, with $v_c>v_s>0$. Find the separation between the arrival times. What does the detector observe if its time resolution is worse than this separation?

Solution

The characteristic arrival times are

$$t_c = \frac d{v_c}, \qquad t_s = \frac d{v_s}.$$

Their separation is

$$\Delta t = t_s-t_c = d \left( \frac1{v_s} - \frac1{v_c} \right).$$

If the instrumental time resolution exceeds $\Delta t$, the two responses overlap and may appear as one broadened pulse. Failure to resolve two peaks therefore does not establish $v_c=v_s$. A charge-only or spin-only detector can also miss the other sector even with adequate time resolution.

Exercise 7: Classify four perturbations

At a spinless fixed point with

$$K=0.4,$$

classify the following at tree level:

1. half-filled Umklapp $\cos(4\phi)$;
2. a local impurity $\cos[2\phi(0)]$;
3. a weak link;
4. a bulk pairing term $\cos(2\theta)$.

Solution

The half-filled Umklapp dimension is

$$\Delta_{\mathrm U} = 4K = 1.6.$$

Because this is a bulk operator and

$$1.6<2,$$

it is relevant.

The local impurity has boundary dimension

$$\Delta_{\mathrm{imp}} = K = 0.4<1,$$

so it is relevant.

The weak link has boundary dimension

$$\Delta_{\mathrm{link}} = \frac1K = 2.5>1,$$

so it is irrelevant near the cut-chain fixed point.

The bulk pairing cosine also has dimension

$$\Delta_{\mathrm{pair}} = \frac1K = 2.5>2,$$

so it is irrelevant at tree level.

The comparison threshold is $2$ for bulk operators and $1$ for boundary operators.

Exercise 8: Finite-size spectroscopy

For a periodic one-mode liquid, compare the lowest energy costs of the sectors

$$(\Delta N,J)=(1,0)$$

and

$$(\Delta N,J)=(0,1),$$

ignoring oscillator excitations and subtracting $\mu\Delta N$. First treat both sectors as allowed; then adapt the result to spinless fermions with $\Delta N$ and $J$ of the same parity.

Solution

The number-changing gap is

$$\Delta E_N = \frac{\pi\hbar v}{2LK}.$$

The winding gap is

$$\Delta E_J = \frac{\pi\hbar vK}{2L}.$$

Their ratio is

$$\frac{\Delta E_J}{\Delta E_N} = K^2.$$

Therefore

$$K = \sqrt{ \frac{\Delta E_J}{\Delta E_N} }.$$

For spinless fermions with $\Delta N=\Delta N_R+\Delta N_L$ and $J=\Delta N_R-\Delta N_L$, the two integers have the same parity. Neither displayed unit sector is then allowed. Using $(2,0)$ and $(0,2)$ multiplies both quadratic gaps by four, leaving their ratio $K^2$ unchanged. Other boundary sectors must be matched separately.

Cross-Links

• Quantum Wires for the material realization, transverse subbands, threshold density of states, and the distinction between one-particle subbands and Luttinger-liquid behavior.
• Low-Dimensional Quantum Gases for confinement criteria, one-dimensional kinematics, and infrared fluctuation constraints.
• Fermi Liquid Theory Preview for the finite-residue quasiparticle framework that Luttinger liquids replace.
• Fractionalization for the cross-dimensional physical meaning beyond one-dimensional field theory.
• Collective Modes for response poles, damping, and general mode language.
• Perturbation Theory in Many-Body Systems for why fixed-order expansions can fail in the one-dimensional infrared.
• Spinless Fermion Chains for the $t$–$V$ model and exact Luttinger data.
• XXZ Chain for a concise microscopic model and exact zero-field Luttinger parameters.
• XXZ Spin Chain for a spin realization with a $c=1$ critical interval.
• Renormalization Group Preview for fixed lines, scaling dimensions, relevance, and crossover.
• Spectral Functions for spectral normalization, thresholds, broadening, and experimental intensity.
• Entanglement Entropy for central-charge extraction and finite-size fitting.
• Entanglement and Criticality for the $c=1$ fixed-line signature, chord-length fits, and finite-entanglement scaling.
• Why Many-Body QM Leads to QFT for the broader operator-field-theory bridge.
• Condensed Matter Roadmap for a study sequence through quantum matter.
• Bridge to QFT Roadmap for the continuation from many-body bosonization to field theory.

References

• M. A. Cazalilla, “Bosonizing One-Dimensional Cold Atomic Gases”, Journal of Physics B 37, S1–S47 (2004).
• M. A. Cazalilla, R. Citro, T. Giamarchi, E. Orignac, and M. Rigol, “One Dimensional Bosons: From Condensed Matter Systems to Ultracold Gases”, Reviews of Modern Physics 83, 1405–1466 (2011).
• T. Giamarchi, Quantum Physics in One Dimension, Oxford University Press (2004).
• T. Giamarchi and H. J. Schulz, “Anderson Localization and Interactions in One-Dimensional Metals”, Physical Review B 37, 325–340 (1988).
• F. D. M. Haldane, “‘Luttinger Liquid Theory’ of One-Dimensional Quantum Fluids. I. Properties of the Luttinger Model and Their Extension to the General 1D Interacting Spinless Fermi Gas”, Journal of Physics C 14, 2585–2609 (1981).
• A. Imambekov, T. L. Schmidt, and L. I. Glazman, “One-Dimensional Quantum Liquids: Beyond the Luttinger Liquid Paradigm”, Reviews of Modern Physics 84, 1253–1306 (2012).
• Y. Jompol et al., “Probing Spin-Charge Separation in a Tomonaga–Luttinger Liquid”, Science 325, 597–601 (2009).
• C. L. Kane and M. P. A. Fisher, “Transport in a One-Channel Luttinger Liquid”, Physical Review Letters 68, 1220–1223 (1992).
• E. H. Lieb and W. Liniger, “Exact Analysis of an Interacting Bose Gas. I. The General Solution and the Ground State”, Physical Review 130, 1605–1616 (1963).
• E. H. Lieb and F. Y. Wu, “Absence of Mott Transition in an Exact Solution of the Short-Range, One-Band Model in One Dimension”, Physical Review Letters 20, 1445–1448 (1968).
• A. Luther and I. Peschel, “Calculation of Critical Exponents in Two Dimensions from Quantum Field Theory in One Dimension”, Physical Review B 12, 3908–3917 (1975).
• A. Luther and V. J. Emery, “Backward Scattering in the One-Dimensional Electron Gas”, Physical Review Letters 33, 589–592 (1974).
• J. M. Luttinger, “An Exactly Soluble Model of a Many-Fermion System”, Journal of Mathematical Physics 4, 1154–1162 (1963).
• D. C. Mattis and E. H. Lieb, “Exact Solution of a Many-Fermion System and Its Associated Boson Field”, Journal of Mathematical Physics 6, 304–312 (1965).
• S. Tomonaga, “Remarks on Bloch’s Method of Sound Waves Applied to Many-Fermion Problems”, Progress of Theoretical Physics 5, 544–569 (1950).
• J. Voit, “One-Dimensional Fermi Liquids”, Reports on Progress in Physics 58, 977–1116 (1995).
• J. von Delft and H. Schoeller, “Bosonization for Beginners — Refermionization for Experts”, Annalen der Physik 7, 225–305 (1998).