Quantum phase transition between topologically distinct quantum critical points

The system under study is a quantum Ising chain interpolated with a three-body cluster interaction. The model is defined by the following Hamiltonian:

Here, $\sigma_{i}^{x/y/z}$ represents the spin-$\frac{1}{2}$ Pauli matrices on each site $i$. The Hamiltonians $H_{TFI}$ and $H_{CI}$ correspond to the transverse field and the cluster Ising model, respectively. Notably, these models possess a $\mathbb{Z}_{2}$ spin-flip (generated by $P=\prod_{i}\sigma^{z}_{i}$) and a time-reversal symmetry denoted as $\mathbb{Z}^{T}_{2}$ (acting as the complex conjugation $T=K$). The $\lambda$ serves as a tuning parameter that governs the competition between two different quantum spin chains, ultimately leading to the emergence of an unconventional universality class.

Although both the spontaneous symmetry-breaking phase (e.g., ferromagnetic (FM) phase) to trivial PM or SPT phase transitions are described by Ising conformal field theory (CFT), the distinct behavior of the time-reversal symmetry towards the symmetry flux operator (also known as the disorder operator) gives rise to topologically distinct (symmetry-enriched ) QCPs or gSPT [27, 29]. To provide a brief overview of this distinction, it’s important to consider that an Ising CFT has a unique local primary field denoted as $\sigma$ with a scaling dimension $\Delta=1/8$, as well as a unique nonlocal primary field denoted as $\mu$ with the same scaling dimension. These primary fields correspond to the order parameters of the nearby phases. For instance, $\sigma(n)\sim\sigma^{x}_{n}$ is the Ising order parameter, whereas the nonlocal operator $\mu(n)$ is the Kramers-Wannier dual disorder order parameter of the disorder symmetric phases. More precisely, $\mu(n)\sim\prod_{j=-\infty}^{n}\sigma^{z}_{j}$ in the trivial PM phase, whereas $\mu(n)\sim\prod_{j=-\infty}^{n}\sigma^{x}_{j-1}\sigma^{z}_{j}\sigma^{x}_{j+1}$ in the SPT phase. Notably, the two Ising critical lines are distinguished by the discrete invariant $T\mu T=\pm\mu$, indicating that they must be separated by a phase transition, essentially representing the "transition" of a phase transition. Indeed, as depicted in Fig. 1, they converge at a multicritical Lifshitz point with the dynamical exponent $z=2$ [85, 3]. More broadly, one of the authors of this paper proposes that the conformal boundary condition can serve as a more general "topological invariant" for classifying topologically distinct quantum critical points, even in the absence of degenerate edge modes [28].

In this work, we denote the topological non-trivial case, characterized by the property where the nonlocal disorder operator is charged as $T\mu T=-\mu$, as the "symmetry-enriched ${\rm{Ising^{*}}}$" critical point, which exhibits degenerate zero-energy edge modes even the bulk is gapless. At a fundamental level, the boundary of the symmetry-enriched ${\rm{Ising^{*}}}$ critical point spontaneously breaks the $\mathbb{Z}_{2}$ symmetry, creating an intriguing degenerate boundary fixed point that remains stable and corresponds to a fixed boundary condition. Remarkably, the finite-size splitting of this edge mode $\sim 1/N^{14}$, is parametrically faster than the finite-size bulk gap $\sim 1/N$. Therefore, the degenerate edge mode can maintain stability even if the bulk remains gapless. In the subsequent few sections, our focus remains on the case where $h=h_{c}=1.0$ to ensure that the system resides within the QCPs.

Before delving into the analytical results, let’s summarize our main findings and outline the global quantum phase diagram of the model in Eq. (1). The schematic phase diagram is provided in Fig.1. The tuning parameters ($h$,$\lambda$) drive the system toward different phases, including FM, trivial PM, and $\mathbb{Z}_{2}\times\mathbb{Z}^{T}_{2}$ SPT phases [86, 87, 88]. The latter is sometimes referred to as the cluster or Haldane SPT phase. Furthermore, there exist rich QPTs between these quantum phases, including the 1+1D conformal (topologically trivial) Ising universality class (green solid line), symmetry-enriched (topologically nontrivial) ${\rm{Ising^{*}}}$ universality class (orange solid line), Gaussian universality class (blue solid line), and nonconformal Lifshitz criticality (red star).

To elaborate, using the integrability of the model, Eq. (1) can be reformulated as a free fermion model [89, 90] through the Jordan-Wigner transformation:

After applying the Fourier transformation $c_{k}=\frac{1}{\sqrt{N}}\sum_{j=1}^{N}e^{ikj}c_{j}$, where $k=2\pi m/N$ and $m$ ranges from $-(N-1)/2$ to $(N-1)/2$, we obtain the following free Hamiltonian:

where $y_{k}=-{\rm{sin}}(k)+(1-\lambda)h{\rm{sin}}(2k)$ and $z_{k}=\lambda h-{\rm{cos}}(k)+h(1-\lambda){\rm{cos}}(2k)$. Subsequently, the Hamiltonian takes on a bilinear form and can be diagonalized using the Bogoliubov transformation:

where $b_{k}$ ($b_{k}^{\dagger}$) is the Bogoliubov quasiparticle annihilation (creation) operator, $\epsilon_{k}=4\sqrt{y_{k}^{2}+z_{k}^{2}}$, ${\rm{tan}}(\theta_{k})=-\frac{y_{k}}{z_{k}}$, and the ground state is given by: $|G\rangle=\prod_{k>0}[{\rm{cos}}(\frac{\theta_{k}}{2})+\mathrm{i}{\rm{sin}}(\frac{\theta_{k}}{2})c^{\dagger}_{k}c^{\dagger}_{-k}]|{\rm{Vac}}\rangle$ ($|{\rm{Vac}}\rangle$ is the vacuum state of $c$ fermion).

Before delving into phase transitions, let’s explore the possible phases that appear in a phase diagram. As a preliminary step, we examine some limiting cases: When $\lambda=0.0$, the model is simplified to a cluster Ising model. By adjusting the parameter $h$, the model can achieve a phase transition from the FM (small $h$) to the cluster SPT phase (large $h$). Conversely, when $\lambda=1.0$, the model is reduced to the usual transverse field Ising model. At this time, adjusting the parameter $h$ can achieve the phase transition from the FM (small $h$) to the trivial PM phase (large $h$). In general cases ($\lambda,h$), to identify the possible quantum phase of matter, we calculated the FM correlation function (order parameter) and string order parameter:

As depicted in Fig. 2(a1) and (a2), we observe that when $\lambda<0.5$ ($\lambda=0.2$) and $h<1.0$ ($h=0.2,0.8$), the FM or string order parameter remains constant or tends to zero in the long-distance limit, suggesting the existence of FM long-range order in this region [1]. Conversely, when $\lambda<0.5$ ($\lambda=0.2$) and $h>1.0(h=2.0,3.0)$, the string order parameter or FM spin correlation function becomes constant or zero in the long-distance limit, indicating that the system resides in the cluster SPT phase in such a parameter region [88].

Similarly, as illustrated in Fig. 2(b1) and (b2), when $\lambda>0.5$ ($\lambda=0.8$) and $h<1.0$ ($h=0.2,0.8$), the FM or string order parameter remains constant or zero in the long-distance limit, implying FM long-range order dominate in this region. However, when $\lambda>0.5$ ($\lambda=0.8$) and $h>1.0$ ($h=2.0,3.0$), both the string order parameter and FM correlation tends to zero under long-distance limits, indicating that the system is in the trivial PM phase in such a region.

After delineating all the quantum phases in the phase diagram, we shift our focus to the more intriguing QPTs between these phases. While traditional discussions primarily concentrate on the phase transitions between gapped phases, it’s entirely plausible that there exists an unconventional QPTs between different gapless topological phases [91]. For simplicity, our attention is drawn to the intriguing "transition" between topologically distinct critical points. For our purposes, we set $h=1.0$ in the model Eq (1), and by manipulating the parameter $\lambda$, we first consider two tractable cases:

1) When $\lambda=0.0$: the model corresponds to a critical cluster Ising chain, thereby realizing the symmetry-enriched ${\rm{Ising^{*}}}$ universality class [27, 29].

2) When $\lambda=1.0$: the model transforms into a usual critical Ising chain, belonging to the 1+1D (topological trivial) Ising universality class.

These two (topologically) distinct Ising universality classes correspond to different conformal boundary conditions [28, 36], and it is infeasible to smoothly connect them without either breaking the symmetry or encountering a multicritical point. Therefore, akin to the unconventional phase transition between gapped topological phases [88], there may exist an unconventional QPT between topologically distinct quantum critical points or critical phases, which constitutes a largely unexplored area in statistical and condensed matter physics. In the following subsections, we provide evidence of unconventional phase transitions from different perspectives.

For large values of $h$, the Ising interaction term becomes negligible, resulting in the simplified Hamiltonian:

When $\lambda=0.0$, the ground state corresponds to the cluster SPT phase protected by $\mathbb{Z}_{2}\times\mathbb{Z}^{T}_{2}$ symmetry. Conversely, when $\lambda=1.0$, the transverse field term dominates, and the ground state resides in the trivial PM phase. Consequently, a topological phase transition is expected between these two distinct ground states [88]. To systematically investigate this phase transition, similar to the previous section, we utilized the Jordan-Wigner transformation to solve the model and calculated various physical quantities such as the second derivative of the ground state energy density, winding number, entanglement entropy, and fidelity susceptibility. Additionally, we determined the central charge and critical exponent through finite-size scaling (see Appendix E for details).

The numerical results reveal that the topological phase transition from the cluster SPT to the trivial PM phase is described by the free boson CFT with critical exponents $\nu=1.0$ and $z=1.0$[103, 104, 88]. More generally, when $h$ is finite and greater than $1.0$, this universality class of phase transitions remains stable (see Appendix E for details). In other words, a critical line (blue solid line in Fig. 1) characterized by the free boson CFT with $c=1$ exists in the global phase diagram.

For the sake of completeness and comparison, we briefly discuss the properties of the model around $h<1.0$. This line differs from the $h>1.0$, we find that regardless of how $\lambda$ is tuned, our results (see Appendix D for details) show that the system always exhibits FM long-range order.