---
title: Biased Spin-Boson Model
url: https://www.emergentmind.com/topics/biased-spin-boson-model
type: topic
---

# Biased Spin-Boson Model

The biased spin-boson model is a dissipative two-state system in which a two-level degree of freedom with tunneling amplitude \(\Delta\) and static bias \(\epsilon\) is longitudinally coupled to a bosonic environment. In its standard form,
\[
H=-\frac{\Delta}{2}\sigma_x+\frac{\epsilon}{2}\sigma_z+\sum_i \omega_i a_i^\dagger a_i+\frac{\sigma_z}{2}\sum_i \lambda_i(a_i+a_i^\dagger),
\]
with bath spectral density \(J(\omega)\), it provides a canonical setting for quantum dissipation, decoherence of two-level systems, open-system dynamics, and impurity criticality. Finite bias explicitly favors one \(\sigma_z\) state and breaks the parity symmetry present at \(\epsilon=0\), so several zero-bias phase transitions are replaced by crossover structures once \(\epsilon\neq 0\) [1701.05831][1201.5650][1608.04854].

## 1. Hamiltonian, conventions, and bath structure

The basic biased Hamiltonian is used directly in several works on equilibrium, dynamics, and transport,
\[
H=-\frac{\Delta}{2}\sigma_x+\frac{\epsilon}{2}\sigma_z+\sum_i \omega_i a_i^\dagger a_i+\frac{\sigma_z}{2}\sum_i \lambda_i(a_i+a_i^\dagger),
\]
where \(\Delta\) is the bare tunneling amplitude, \(\epsilon\) is the bias field, and the bath couples through the same \(\sigma_z\) operator that defines the localized-state population [1701.05831]. This longitudinal coincidence is central: the external bias and the dissipative dressing act in the same channel, so bias and bath renormalization do not simply add; they compete and reinforce each other in a strongly model-dependent way.

The bath is usually encoded by a power-law spectral density. A common sub-Ohmic/Ohmic choice is
\[
J(\omega)=2\pi\alpha\,\omega^s\omega_c^{1-s}, \qquad 0\le \omega\le \omega_c,
\]
or, with an exponential cutoff,
\[
J(\omega)=2\pi\alpha\,\omega^s\omega_c^{1-s}e^{-\omega/\omega_c}.
\]
Here \(s<1\) is sub-Ohmic, \(s=1\) is Ohmic, and \(s>1\) is super-Ohmic; \(\alpha\) is the dimensionless dissipation strength and \(\omega_c\) the high-frequency cutoff [1701.05831][1608.04854][1406.4983]. In sub-Ohmic baths, the enhanced low-frequency spectral weight makes the environment especially slow and history-dependent, which is why biased sub-Ohmic dynamics is typically much more non-Markovian than simple weak-coupling intuition suggests.

Several papers use rotated spin conventions. One field-theoretic formulation writes
\[
\mathcal H=-B S_z-b S_x+\lambda S_x X+\mathcal H_X,
\]
where the source field \(b\) is conjugate to the ordering component \(S_x\) and therefore plays the role analogous to a bias field in conventional spin-boson notation; this is explicitly described as a spin-axis relabeling of the standard model [2501.12457]. A different extension introduces hybrid diagonal and off-diagonal system-bath coupling and, after a rotation, generates an effective bias
\[
\widehat{\epsilon}=\Delta_0\sin\varphi,\qquad \widehat{\Delta}=\Delta_0\cos\varphi,
\]
showing that “bias” can also emerge from basis rotation and tunneling renormalization rather than from an explicit bare \(\epsilon\sigma_z/2\) term [2009.12296].

## 2. Equilibrium structure under finite bias

For the sub-Ohmic model at \(\epsilon=0\), the standard picture is a continuous quantum phase transition at \(\alpha=\alpha_c\) between delocalized and localized phases. Finite bias \(\epsilon\neq 0\) explicitly breaks the parity symmetry
\[
a_i\to -a_i,\qquad \sigma_z\to -\sigma_z,
\]
so the sharp delocalized-localized transition is replaced by a smooth crossover from weak-coupling delocalized-like behavior to strong-coupling localized-like behavior [1701.05831]. This is one of the most important structural facts about the biased problem: in most equilibrium observables the nearby zero-bias critical fixed point still organizes the physics, but it no longer appears as a true phase transition once the symmetry-breaking field is finite.

The equilibrium dynamical correlation function
\[
C(t)=\frac12\langle[\sigma_z(t),\sigma_z(0)]_+\rangle,\qquad
C(\omega)=\frac{1}{2\pi}\int_{-\infty}^{+\infty}C(t)\,dt
\]
has a particularly simple low-frequency law in the biased sub-Ohmic problem. Bosonic NRG finds
\[
C(\omega)\propto \omega^s,\qquad \omega\to 0,
\]
for essentially the entire \((\alpha,\epsilon)\) parameter space, including the biased strong-coupling regime, with the only exception at the exact unbiased critical point \(\alpha=\alpha_c,\epsilon=0\), where
\[
C(\omega)\propto \omega^{-s}.
\]
The same study reports \(C(\omega)\propto \chi^2\omega^s\) over a wide parameter range, supporting the generalized Shiba relation even for \(\epsilon\neq 0\) and \(\alpha>\alpha_c\) [1701.05831]. Finite bias therefore cuts off the zero-bias infrared singularity and restores the \(\omega^s\) continuum tail at sufficiently low frequency.

Near \(\alpha_c\), the biased crossover is controlled by the competition between the unbiased crossover scale
\[
\omega_0^*\equiv T_0^*\propto |\alpha-\alpha_c|^{z\nu}
\]
and a bias scale
\[
\epsilon^*\propto (\omega_0^*)^{1/\theta}.
\]
At the critical coupling,
\[
\omega^*(\alpha=\alpha_c,\epsilon)\propto \epsilon^\theta,
\]
with
\[
\theta=\frac{2}{3s}\quad (0<s\le 1/2),\qquad
\theta=\frac{2}{1+s}\quad (1/2<s<1).
\]
The regime \(\epsilon\ll \epsilon^*\) is controlled by the nearby unbiased fixed point, whereas for \(\epsilon\gg \epsilon^*\) the bias dominates, suppresses low-frequency continuum weight, and transfers spectral weight into the \(\delta(\omega)\) component associated with finite \(|\langle\sigma_z\rangle|\) [1701.05831].

A rotated-basis field theory expresses the same bias idea through the equation of state
\[
\langle S_x\rangle \propto b^{(1-s)/(1+s)}
\]
at criticality, so that
\[
\frac{1}{\delta}=\frac{1-s}{1+s}.
\]
In that formulation \(b\) is the field conjugate to the order parameter and is explicitly identified as the analog of the conventional bias field after spin-axis relabeling [2501.12457]. By contrast, a mean-field analysis of boson-state truncation shows that the finite-bias exponent can be strongly distorted for \(0<s<1/2\) at finite \(N_b\): the true mean-field \(\delta=3\) crosses over to an apparent truncation-dominated
\[
\delta=\frac{1+s}{1-s},
\]
so low-bias critical fits in truncated calculations are artifact-sensitive unless the \(N_b\to\infty\) limit is controlled explicitly [1005.4977].

## 3. Nonequilibrium dynamics and bath-induced bias renormalization

Real-time studies make the biased model qualitatively richer than a simple detuned damped two-level system. For the deep sub-Ohmic case \(s=0.1\), TDVMPS simulations of
\[
H=-\frac{\Delta}{2}\sigma_x+\frac{\epsilon}{2}\sigma_z+\sum_n\omega_n b_n^\dagger b_n+\frac{\sigma_z}{2}\sum_n g_n(b_n^\dagger+b_n)
\]
show three distinct nonequilibrium regimes. At weak coupling, the spin exhibits damped oscillations close to the bare precession frequency
\[
\Omega=\sqrt{\Delta^2+\epsilon^2}/2,
\]
but the late-time behavior departs from a Markovian master equation because slow bath modes acquire a small net displacement and act back on the spin as an additional \(z\)-field. At strong coupling, the environment rapidly forms a localized bosonic cloud, strongly suppresses tunneling, and the spin freezes near its initial state. Between them lies a nonperturbative intermediate regime with “sequential, time-retarded” dynamics: the spin first behaves approximately as in weak coupling and is only later dragged into bath-dominated evolution once the slow modes have sufficiently reorganized [1707.05524].

That study is unusual in explicitly following both spin and environment. The projected mode displacements
\[
f_{\uparrow,\downarrow}(n)=\left\langle \frac{\mathbb 1\pm \sigma_z}{2}\frac{b_n^\dagger+b_n}{2}\right\rangle
\]
distinguish two mechanisms. If \(f_\uparrow\) and \(f_\downarrow\) drift together, the bath develops a net displacement and hence an effective dynamical bias field. If they remain strongly anticorrelated, the bath response is predominantly polaronic and suppresses tunneling instead [1707.05524]. The interpretation is that the interaction term generates a time-dependent environmental contribution to the \(z\)-field proportional to the mode displacements; this is the paper’s microscopic explanation of environment-induced bias renormalization.

In the weak-coupling Ohmic problem at arbitrary bias, a consistent expansion beyond Bloch-Redfield gives a complementary picture. The renormalized tunneling and Rabi frequency are
\[
\tilde{\Delta}=\Delta\left(\frac{\Omega}{D}\right)^\alpha,\qquad
\Omega=\sqrt{\epsilon^2+\tilde{\Delta}^2},
\]
and the three nonstationary modes are
\[
z_0=-i\Gamma,\qquad z_\pm=\pm\Omega-i\Gamma/2,
\]
with
\[
\Gamma=\pi\alpha\frac{\tilde{\Delta}^2}{\Omega}.
\]
The time evolution has the generic structure
\[
\langle\sigma_\alpha\rangle(t)=\langle\sigma_\alpha\rangle_{\mathrm{st}}
+F_\alpha^0(t)e^{-\Gamma t}
+F_\alpha^c(t)e^{-\Gamma t/2}\cos(\Omega t)
+F_\alpha^s(t)e^{-\Gamma t/2}\sin(\Omega t),
\]
but the preexponential functions are not constants: they contain logarithmic terms in \(\Omega t\), branch-cut contributions, and power-law crossovers that are absent from simple Bloch-Redfield theory. After real-time RG resummation, the oscillating modes cross over from exponent \(2\alpha\) at exponentially small times to the bias-dependent exponent
\[
2\alpha\,\epsilon^2/\Omega^2
\]
at exponentially large times [1802.09846]. Bias therefore changes not only the oscillation frequency but also the long-time asymptotic structure of the decay.

An important reference point is the symmetric sub-Ohmic problem. Numerically exact TEMPO calculations at \(\epsilon=0\) find a three-region dynamical diagram with coherent, incoherent, and strong-coupling “pseudo-coherent” behavior, where the oscillation-like turnaround is controlled by the bath timescale \(t_{\min}\propto \omega_c^{-1}\) rather than by intrinsic tunneling [2208.10313]. Those results are direct only for the unbiased case. A plausible implication is that similar bath-driven turning points can survive under bias, but their interpretation must be reformulated because the relaxation target is no longer symmetry-related.

## 4. Variational and nonperturbative descriptions of the biased ground state

The biased Ohmic ground state is poorly captured by symmetric polaron pictures. A generalized biased ansatz introduces asymmetric bath displacements
\[
\alpha_{\mathbf k,0}=\frac{g_{\mathbf k}(\epsilon-\omega_{\mathbf k})}{\omega_{\mathbf k}(\omega_{\mathbf k}+\chi)},\qquad
\alpha_{\mathbf k,1}=\frac{g_{\mathbf k}(\epsilon+\omega_{\mathbf k})}{\omega_{\mathbf k}(\omega_{\mathbf k}+\chi)},
\]
with
\[
\chi=\sqrt{\Delta_r^2+\epsilon^2},
\]
leading to the self-consistent renormalized tunneling
\[
\Delta_r=\Delta\exp\!\left[-2\int_0^\infty d\omega\,\frac{J(\omega)}{(\omega+\chi)^2}\right].
\]
For the Ohmic spectrum this gives
\[
\Delta_r=\Delta\left(\frac{\chi}{\chi+\omega_c}\right)^\alpha e^{\alpha\omega_c/(\chi+\omega_c)}.
\]
Because \(\chi\) remains nonzero when \(\epsilon\neq 0\), the infrared divergence is cut off and \(\Delta_r\) stays finite and continuous. The same construction produces an environment-induced bias renormalization
\[
R=\frac{2\alpha\omega_c\epsilon}{\chi+\omega_c},
\]
and a smooth ground-state magnetization
\[
M=\langle\sigma_z\rangle
=-\frac{\epsilon\left(1+\frac{2\alpha\omega_c}{\chi+\omega_c}\right)}{\eta},
\]
with
\[
\eta=\sqrt{\Delta_r^2+\epsilon^2\left(1+\frac{2\alpha\omega_c}{\chi+\omega_c}\right)^2}.
\]
This removes the unphysical discontinuous collapse of \(\Delta_r\) and the spurious jump in \(M\) produced by Silbey-Harris theory for biased systems, and matches Bethe-Ansatz results much more closely in the scaling limit [1201.5650].

A more systematic ground-state description is the multi-polaron coherent-state expansion,
\[
|\Psi_{\rm GS}^{\rm bias}\rangle=
\frac{1}{\sqrt{\mathcal N}}
\sum_{n=1}^{N_{\rm pols}}
\left[p_n|f_\uparrow^{(n)}\rangle\otimes|\uparrow\rangle
-q_n|f_\downarrow^{(n)}\rangle\otimes|\downarrow\rangle\right].
\]
Its central message is that finite bias does not eliminate antipolarons. High-frequency modes remain polaronic, intermediate-energy modes still develop tunneling-stabilized antipolaronic displacements, and only the low-frequency asymptote changes qualitatively: in the biased case all displacements saturate to the same finite value rather than vanishing as \(\omega_k\to 0\). The resulting observables
\[
\langle\sigma_x\rangle=
-\frac{\sum_{n,m}p_nq_m\langle f_\uparrow^{(n)}|f_\downarrow^{(m)}\rangle}
{\sum_{n,m}[p_np_m\langle f_\uparrow^{(n)}|f_\uparrow^{(m)}\rangle+q_nq_m\langle f_\downarrow^{(n)}|f_\downarrow^{(m)}\rangle]},
\]
\[
\langle\sigma_z\rangle=
\frac{\sum_{n,m}[p_np_m\langle f_\uparrow^{(n)}|f_\uparrow^{(m)}\rangle-q_nq_m\langle f_\downarrow^{(n)}|f_\downarrow^{(m)}\rangle]}
{\sum_{n,m}[p_np_m\langle f_\uparrow^{(n)}|f_\uparrow^{(m)}\rangle+q_nq_m\langle f_\downarrow^{(n)}|f_\downarrow^{(m)}\rangle]}
\]
reproduce Bethe-Ansatz results very accurately; typically \(N_{\rm pols}=6\), and sometimes \(N_{\rm pols}=8\), is sufficient in the biased regimes studied [1406.4983].

The displaced-Fock-state improvement of Silbey-Harris pursues the same correction from another angle. Starting from the asymmetric biased ansatz
\[
D(\alpha_k)|0\rangle|\uparrow\rangle+D(\beta_k)r|0\rangle|\downarrow\rangle,
\]
it adds orthogonal displaced Fock sectors order by order. In the biased case, this removes the discontinuous magnetization jump reported in earlier SH-based work and yields smooth, rapidly convergent magnetization curves for both \(s=1\) and \(s=0.4\); for \(s=0.4\), convergence is essentially reached already by the third-order ansatz [1803.10142].

## 5. Transport, generated bias, and exact structures

In the nonequilibrium spin-boson model with two bosonic reservoirs at different temperatures,
\[
H_s=\frac{\varepsilon}{2}\sigma_z+\frac{\Delta}{2}\sigma_x,\qquad
H_I=\sigma_z\sum_{j,v=L,R}g_{j,v}(b_{j,v}^\dagger+b_{j,v}),
\]
the bias \(\varepsilon\) becomes a control parameter for energy transfer rather than only for local polarization. A polaron-transformed NEGF treatment gives a Landauer-like steady-state heat current
\[
J_L=\frac{\alpha\xi}{4\pi}\int_0^\infty d\omega\,\omega\,\tilde{\chi}_z''(\omega)\tilde I(\omega)\,[n_L(\omega)-n_R(\omega)],
\]
and reveals a bias-induced nonmonotonic energy conductance in the intermediate-coupling regime. The physical explanation given is resonance: increasing bias changes the effective spin splitting relative to the thermal window of the baths, so conductance can first increase and later decrease as \(\varepsilon\) is tuned [1609.05598].

Hybrid diagonal/off-diagonal coupling extends the same theme by showing that bias can be generated rather than imposed. Starting from
\[
H=\frac{\Delta_0}{2}\sigma_x+\sum_i\omega_i a_i^\dagger a_i+
\frac{bI+\cos\varphi\,\sigma_z+\sin\varphi\,\sigma_x}{2}\sum_i c_iX_i,
\]
a unitary rotation maps the model onto an ordinary spin-boson Hamiltonian with
\[
\widehat{\epsilon}=\Delta_0\sin\varphi,\qquad
\widehat{\Delta}=\Delta_0\cos\varphi.
\]
The paper distinguishes this rotated-frame effective bias from a separate “dynamically generated bias” produced by the renormalized tunneling amplitude \(\widehat{\Delta}_r\). In the overdamped regime the long-time population has the form
\[
P_z(t)\approx (A-B\tanh(-\beta E/2))e^{-\widehat{\Gamma}t}+B\tanh(-\beta E/2),
\]
so that \(P_z(\infty)\neq 0\) even though no explicit bare \(\epsilon\sigma_z/2\) term was present in the original Hamiltonian [2009.12296].

The biased Ohmic model also has an exact-overlap and integrability sector. With Hamiltonian
\[
{\boldsymbol H}=\int_0^\infty dk\,b_k^\dagger b_k\,\sigma_0-J\,\sigma_1-h\,\sigma_3
-\sqrt{\frac{g}{2}}\int_0^\infty dk\,(b_k^\dagger+b_k)\sigma_3,
\]
the bias is the local field \(h\), the renormalized impurity scale is \(E^\star\), and the magnetization is
\[
m=-\left(\frac{\partial E_0}{\partial h}\right)_{E^\star}.
\]
For ground-state overlaps between two parameter sets, the orthogonality exponent is
\[
d_{21}=\frac{g}{4}(m_2-m_1)^2,
\]
with generalized form
\[
d_{21}(a)=\frac{g}{4}(m_2-m_1-2a)^2.
\]
This connects bias variation directly to Anderson orthogonality, and at the Toulouse point \(g=1/2\) the exact magnetization is
\[
m=\frac{2}{\pi}\arctan\!\left(\frac{\pi h}{2E^\star}\right)
\]
[1506.01358].

## 6. Limits, controversies, and open problems

Several technical and conceptual cautions recur across the literature. First, some apparent critical exponents are method-sensitive. In the mean-field spin-boson model, boson-state truncation is a strong relevant operator with respect to the Gaussian fixed point for \(0<s<1/2\), and the finite-bias law
\[
m(\alpha=\alpha_c,\epsilon)\propto |\epsilon|^{1/3}
\]
crosses over at finite \(N_b\) to the truncation-dominated
\[
m(\alpha=\alpha_c,\epsilon)\sim |\epsilon|^{(1-s)/(1+s)}.
\]
This implies that the low-bias exponent \(\delta\) is especially vulnerable to numerical artifacts in truncated calculations [1005.4977].

Second, the status of Ohmic criticality remains nonuniform across formulations. A Schwinger-Keldysh/Majorana RG analysis claims a second-order quantum phase transition for both \(0<s<1\) and \(s=1\), with
\[
\frac{1}{\delta}=\frac{1-s}{1+s},
\]
and explicitly states that this is contrary to the conventional quantum-classical mapping expectation of a Kosterlitz-Thouless transition in the Ohmic case [2501.12457]. By contrast, other Ohmic treatments retain the standard BKT picture, for example the hybrid-coupling study with
\[
\alpha_c=1+O(\Delta_0/\omega_c)
\]
and the integrable Ohmic discussion that places the Kosterlitz-Thouless transition at \(g=1\) [2009.12296][1506.01358]. These statements are not framed in identical conventions, but they mark a genuine interpretive tension in the present literature.

Third, several influential dynamical results are direct only in the unbiased limit. The TEMPO study of the sub-Ohmic “pseudo-coherent” phase is explicit that \(\epsilon=0\) throughout and that any extension to \(\epsilon\neq 0\) is an inference rather than a reported result [2208.10313]. Likewise, exact SSE studies with added Markovian dephasing treat only the unbiased Ohmic model, so their frequency-renormalization formulas are baseline results rather than direct biased-model statements [2305.00110].

Finally, rigorous mathematical results are narrower than the standard biased \(\sigma_z\)-coupled Hamiltonian. The small-coupling ground-state existence theorem for the massless spin-boson model applies to an off-diagonal coupling structure in the energy basis and is not a direct treatment of the conventional biased model; its main structural lesson is that the absence of diagonal coupling terms is what makes the infrared problem tractable in that setting [1003.5923]. The massive one-boson scattering analysis, although it already describes an energy-asymmetric two-level system with
\[
K=\begin{pmatrix}e_1&0\\0&e_0\end{pmatrix},\qquad 0=e_0<e_1,
\]
is likewise not written in the conventional \(-\Delta\sigma_x/2+\epsilon\sigma_z/2+\sigma_z\otimes\Phi\) form, even though its resonance structure depends directly on the level asymmetry \(e_1-e_0\) [1810.09135].

Taken together, these results support a stable qualitative picture. Finite bias removes the exact \(\sigma_z\to-\sigma_z\) symmetry, rounds zero-bias critical singularities into crossovers, produces a biased stationary or equilibrium polarization, and changes both infrared scaling and long-time dynamics. At the same time, the detailed location and meaning of “delocalized,” “localized,” “coherent,” or “pseudo-coherent” behavior remain strongly dependent on bath exponent, cutoff, initial preparation, and approximation scheme. The biased spin-boson model is therefore less a single solved system than a family of closely related impurity problems whose shared core is the competition among tunneling, explicit asymmetry, and bath-induced renormalization.

Source: https://www.emergentmind.com/topics/biased-spin-boson-model