Topological Modes in Monitored Quantum Dynamics

Haining Pan,1, 2 Hassan Shapourian,3 and Chao-Ming Jian1

1
Department of Physics, Cornell University, Ithaca, New York 14853, USA
2
Department of Physics and Astronomy, Center for Materials Theory,
Rutgers University, Piscataway, New Jersey 08854 USA
3
Cisco Quantum Lab, Los Angeles, California 90404, USA
Dynamical quantum systems both driven by unitary evolutions and monitored through measure-
ments have proved to be fertile ground for exploring new dynamical quantum matters. While the
entanglement structure and symmetry properties of monitored systems have been intensively stud-
ied, the role of topology in monitored dynamics is much less explored. In this work, we investigate
novel topological phenomena in the monitored dynamics through the lens of free-fermion systems.
Free-fermion monitored dynamics were previously shown to be unified with the Anderson local-
arXiv:2411.04191v1 [quant-ph] 6 Nov 2024

ization problem under the Altland-Zirnbauer symmetry classification. Guided by this unification,
we identify the topological area-law-entangled phases in the former setting through the topological
classification of disordered insulators and superconductors in the latter. As examples, we focus on
1+1D free-fermion monitored dynamics in two symmetry classes, DIII and A. We construct quan-
tum circuit models to study different topological area-law phases and their domain walls in the
respective symmetry classes. We find that the domain wall between topologically distinct area-law
phases hosts dynamical topological modes whose entanglement is protected from being quenched by
the measurements in the monitored dynamics. We demonstrate how to manipulate these topologi-
cal modes by programming the domain-wall dynamics. In particular, for topological modes in class
DIII, which behave as unmeasured Majorana modes, we devise a protocol to braid them and study
the entanglement generated in the braiding process.

Introduction.— Monitored dynamical quantum sys- This unification implies a correspondence between the
tems evolve through both unitary evolution and mea- phases in free-fermion monitored dynamics with an area-
surements that “monitor” the systems’ wave function. law entanglement entropy (EE) scaling and disordered
The ground-breaking discovery of the measurement- localized phases in the Anderson localization problem
induced phase transitions [1–6] in monitored dynamics of the same symmetry class. The latter is well-known
has sparked extensive research recently to explore the to further subdivide into topological insulators and su-
rich landscape of novel dynamical quantum matter in perconductors within each symmetry class (see review
such systems. Significant progress has been made in un- Refs. [47, 48]). The domain walls (DWs) between distinct
derstanding the changes of entanglement structure across topological localized phases host topologically protected
the measurement-induced phase transitions (see, for ex- modes. Through this correspondence, area-law phases in
amples, Refs. [1–4, 7–20] and reviews Refs. [5, 6]) and free-fermion dynamics should follow the same topological
the effect of global symmetries in enriching the phase classification, and the DWs between them should exhibit
diagram of monitored dynamical systems [21–32]. In dynamical topological domain-wall modes (DTDMs).
contrast, topological properties of monitored dynamics In this paper, we demonstrate the topology and the
have only been previously investigated in limited exam- DTDMs of 1D area-law phases of free-fermion monitored
ples [23, 33, 34]. This work investigates novel topological dynamics in AZ symmetry classes DIII and A as ex-
phenomena in the monitored dynamics through the lens amples, corresponding to 2D disordered localized phases
of free-fermion systems. classified by Z2 and Z, respectively [47, 48]. We first
In free-fermion monitored dynamics, both unitary evo- construct quantum circuit models to realize the topolog-
lutions and the measurements keep the fermionic states ically distinct area-law phases, expected from the corre-
non-interacting. That is, a Slater-determinant state (or spondence, within each symmetry class. Then, we design
its charge-non-conserved counterpart) always evolves into circuit models with DWs to investigate the entanglement
another Slater-determinant state in every quantum tra- structure and the topological classification of the DTDMs
jectory (labeled by the different measurement outcomes). via numerical simulations. We show that despite fre-
Despite being “free”, the intrinsic randomness in the quent measurements in the surroundings of the DWs, the
measurement outcomes can still lead to rich universal entanglement carried by the DTDMs remains protected
behavior in the entanglement structure, as shown in pre- from being quenched by the (symmetry-allowed) mea-
vious works [26–30, 34–45]. This work is motivated surements. We demonstrate how to manipulate the DT-
by a unification of the free-fermion monitored dynam- DMs by programming DWs in the circuits. Specifically,
ics in d spatial dimensions with the Anderson local- for class-DIII monitored dynamics, where each DTDM is
ization problem in d + 1 spatial dimensions under the effectively an unmeasured Majorana mode, we design a
Altland-Zirnbauer (AZ) symmetry classification [26, 46]. protocol to braid these DTDMs and study the entangle-
 2

ment generated in the braiding process. to simplify the parametrization. When podd (peven ) is
large, the actions on odd (even) links are predominantly
(a) (b)
t L = 32 measurements, while those on the even (odd) links are
0.4
L = 64 predominantly two-site random unitary gates. We will
L = 128
B A L = 256 determine the phase diagram of this monitored Majorana

IA,B
1
peven
chain parametrized by podd (abbreviated as p hereafter).
Area law Critical Area law
To distinguish different dynamical phases, we numer-
podd
ically calculate the average steady-state mutual infor-
0 x 0.0 p
0 pc,1 0.5 pc,2 1 mation (MI) IA,B between two antipodal regions with
LA = LB = L4 as shown in Fig. 1(b) (see Secs. I and III
FIG. 1. (a) Spacetime geometry of a symmetry-class-DIII in SM for details). The phase diagram in Fig. 1(b) illus-
monitored circuit on a Majorana chain of L sites (black dots). trates two phase transitions at pc,1 ≈ 0.4 and pc,2 ≈ 0.6
This circuit has a staggered pattern with different measure- indicated by crossings of IA,B for different system sizes
ment probabilities podd/even in different two-site gates. (b) L. For p ∈ [0, pc,1 ) and p ∈ (pc,2 , 1], IA,B vanishes as L
Average steady-state mutual information IA,B as a function
grows, signifying the dynamical phases with the area-law
of p ≡ podd = 1 − peven in the class-DIII monitored circuit
shows two distinct area-law phases, separated by a critical scaling of the steady-state EE. For p ∈ [pc,1 , pc,2 ], the
phase. Inset shows the configuration of two antipodal regions, system is in a critical phase (where the EE scales loga-
A and B, on a ring geometry to compute IA,B . rithmically). These area-law and critical phases of the
monitored Majorana chain have been previously identi-
Monitored dynamics in a free Majorana chain— We fied by Refs. [26–28, 35] in similar models. We use the
start with the monitored dynamics of a free Majorana phase diagram in Fig. 1(b) to guide our subsequent in-
chain without any symmetry constraint (except fermion vestigation of the DW between the topologically distinct
parity). This system belongs to the AZ symmetry class area-law phases at p < pc,1 and p > pc,2 .
DIII, which can be understood from both the transfer The physical intuition for the two topologically distinct
matrix for single-particle evolution and density matrix area-law phases is as follows: In the long-time limit, the
evolution in a doubled Hilbert space [26] [see Sec. II two types of dynamical phases are dominated by the mea-
in Supplemental Material (SM) for a review]. The dy- surements that dimerize the Majorana chain on the even
namical phases where the system evolves into states with (odd) links. Their spatial DW must host a protected
area-law EE scaling correspond to 2D disordered class- DTDM, equivalent to a Majorana mode that cannot be
DIII superconductors, whose topologies are Z2 -classified. measured by itself. In a complementary picture, the dy-
Below, we construct monitored quantum circuits to real- namics of the two area-law phases in the 1+1D space-
ize the two classes of area-law dynamical phases and show time correspond to the quantum states of the two classes
that each DW between them hosts a localized DTDM, of disordered topological superconductors in 2-spatial di-
equivalent to an effective unmeasured Majorana mode. mensions in class DIII [26]. Therefore, the spacetime DW
We design protocols to braid these DTDMs and study between the two area-law phases must exhibit nontrivial
the entanglement generation during the braiding process. dynamics of the DTDMs. Below, we confirm and reify
We consider a class-DIII free-fermion monitored dy- these intuitions by investigating concrete circuit models
namics driven by the monitored circuit shown in Fig. 1(a) with a DW between the two area-law phases.
acting on a chain with L Majorana modes γ̂i=1,2,...,L and Domain-wall dynamics—In exploring the phase dia-
the periodic boundary condition. For each pair of adja- gram Fig. 1, we keep the p homogeneous in spacetime. To
cent Majorana modes (γ̂i , γ̂i+1 ), we apply either a two- study DWs and DTDMs, we generalize p to a smoothly-
site random unitary gate, exp{θγ̂i γ̂i+1 } with a random varying spacetime-dependent probability p(i, t) for mea-
θ ∈ [0, 2π), with a probability 1 − p, or a two-site pro- suring odd links, while locally maintaining peven + podd =
jective measurement of the local fermion parity iγ̂i γ̂i+1 1. By programming p(i, t), we can create spacetime do-
with a probability of p. For each measurement, the sys- mains (with p < pc,1 and p > pc,2 ) dominated by one
tem evolves under the Kraus operators 1±iγ̂2i γ̂i+1 (which of the two area-law dynamical phases with DWs in be-
collapse the wave function at the measured sites) de- tween. As a concrete example, we consider a 128-site Ma-
pending on the measurement outcome. Both the uni- jorana chain with a spacetime configuration p(i, t) shown
tary gates and measurements evolve free-fermion states in Fig. 2(a). This configuration starts with a uniform
to free-fermion states. p < pc,1 at t = 0. A pair of DWs are introduced at i = L4
A staggered pattern of p in the quantum circuit al- and i = 3L4 at time t = 1, with two DTDMs expected to
lows access to different dynamical phases: In the first emerge. For later times, we program p(i, t) so that the
(second) half of each time step, the probability of mea- two DWs cross and continue propagating until t = 80.
suring the parity of Majorana pairs on odd (even) links, To examine the entanglement evolution in the physi-
(γ̂2i−1 , γ̂2i ) [(γ̂2i , γ̂2i+1 )], is denoted as podd (peven ) [dark cal Majorana chain, we introduce a reference Majorana
(light) blue gates in Fig. 1(a)]. We set podd + peven = 1 chain of the same length [gray dots in Fig. 2(d)] that
 3

(a) (b) (c)

p(i, t) sp (i) class-DIII disordered superconductors. From the circuit
t 0 1 t 0 0.5 t
80 perspective, when DWs are close, their DTDMs can be
64 measured in pairs, causing their disentanglement from
48 the reference chain [as illustrated in Fig. 2(d)]. As DWs
32 separate again for t > 49, their DTDMs reemerge and be-
16 come maximally entangled with each other. Hence, the
0 i i Sp
EC describing the physical system’s entanglement with
1 32 64 96 128 1 32 64 96 128 0 1 2 64 the reference system does not detect these DTDMs for
(d) t > 49. The circuit with p(i, t) shown in Fig. 2 demon-
Phys. system
strates the ability to control the DTDMs by program-
Ref. system ming the DW motions via p(i, t).
t=0 t=1

B A B A B A B A
β2β α2α1 β2β α1 β2 α1 β2α β1α1
α2 β1
t = 47 t = 48 t = 80 1 1 2
α2

FIG. 2. (a) Spacetime configuration p(i, t) in a class-DIII C C C C

2
monitored circuit acting on a 128-site Majorana chain. (b-c) IA,B

I [log 2]
Entanglement contour and total entanglement entropy (in bi- 1 IB,C
nary logarithm) of the physical chain in a typical quantum tra- IA,C
jectory. The peaks in the entanglement contour indicate the 0
t0 t1 t2 t3
spacetime position of the dynamical topological domain-wall
modes. (d) Schematics for the “ideal” entanglement structure
of physical and reference systems at different times t. Each FIG. 3. Top: Schematics for the four stages of the DTDM
dot presents a Majorana mode, with the DTDMs colored red. braiding protocol, where α1 , α2 and β1 , β2 represent the DT-
The colored strip behind physical sites represents the config- DMs. We numerically simulate the protocol for a T-junction
uration of p(i, t) shown in (a). Each pair of modes connected of three 64-site chains. Bottom: Mutual information between
by a wavy line is maximally entangled. the pair of chains (A, B), (B, C), and (C, A) in a typical quan-
tum trajectory as a function of time t.

is mode-by-mode maximally entangled with the physical Braiding.— Using the ability to control the DTDMs,
chain at t = 0. The EE Sp between the physical and we study a protocol to braid them in a T-junction ge-
reference chains evolves [Fig. 2(c)] as the physical chain ometry [see Fig. 3(a)] formed by Majorana chain A, B,
undergoes the monitored circuit evolution. Since we are and C. With a short-range entangled state on each chain
considering a free-fermion system, we can use the entan- at the beginning, the pairwise MI IA,B , IB,C , and IA,C
glement contour (EC) sp (i) to spatially resolve the phys- are all initially zero. At time t0 , we create a maximally-
ical chain’s entanglement with the reference chain [49]. entangled pair of DTDMs α1,2 (β1,2 ) on chain A (B) via
Note that sp (i) is not the EE at the site i and satisfies the dynamics shown in Fig. 2. In the following, we devise
P
i sp (i) = Sp (see Sec. I 2 for its definition). Figure 2(b)
a protocol that braids α2 and β1 and study the evolution
shows the time evolution of sp (i) for a typical quantum of MI IA,B , IB,C , and IA,C among the three chains in-
trajectory in the monitored dynamics governed by p(i, t) duced by the braiding in typical quantum trajectories.
shown in Fig. 2(a). Figure 3 outlines this braiding protocol and the evolu-
As the physical system evolves beyond t = 1, the total tion of IA,B , IB,C , and IA,C in the process (See Sec. III 4
EE quickly drops from the maximal value Sp = L2 log 2 = in SM for microscopic details). From time t0 to t1 , we
64 log 2 to Sp = log 2. Most of the physical chain disen- combine chain A and B into one single chain and program
tangles from the reference by the measurements. The EC the DW configurations to move the DTDM α2 from chain
sp (i) develops two peaks, each contributing 12 log 2 to the A to C, after which the two chains become entangled as
total EE, localized around the two DWs. Each peak cor- indicated by IA,C = 2 log 2 in Fig. 3. The DTDMs α1 and
responds to a DTDM [red dot in Fig. 2 (d)], equivalent β1,2 stand still during this time. From time t1 to t2 , we
to an effective Majorana mode whose entanglement with use the same method to move the DTDM β1 from chain
the reference chain is protected from being quenched by B to A keeping α1,2 and β2 fixed. From time t2 to t3 , we
the local measurements, consistent with our expectation. complete the braiding by moving β1 from chain C to A.
During t ∈ [17, 47], DWs are programmed via p(i, t) to This braiding process entangles chains A and B, leading
approach each other, and the DTDM follows the DWs to IA,B = 2 log 2 as expected. Our DTDM braiding pro-
with their EC peaks still integrated to 12 log 2 each, as tocol resembles the braiding protocol for Majorana zero
shown in Fig. 2(b). modes using nanowire T-junctions. [50, 51] However, be-
As the two DWs cross at t = 48, the EC peaks car- ing independent of any Hamiltonian, our protocol is not
ried by the DTDMs annihilate each other, manifesting a limited by any energy gap in the spectrum but only the
Z2 classification consistent with that of the DWs in 2D correlation length in the area-law phases.
 4

(a) (b)
t 0.5 evolution induced by (generalized) measurements [52]
L = 16 L = 64 n o
†
L = 32 L = 128 to take the form exp ±α(ci,A cj,B + h.c.) . α ∈ R
1
α1 Area law Area law parametrizes the measurement strength. Such Kraus

IA,B
operators can be implemented as the weak measure-
α0 ments of the respective occupancy on the fermion modes
θ1 θ2 √1 (ci,A ± cj,B ) followed by post-selection of the quantum
0 x 0.0 tanh2 α0 2
A B 0 tanh2 α0,c 1 trajectories (see Sec. IV 2 in SM for this post-selected
(c) (d) sp (i) (e) weak measurement). The α → ∞ limit corresponds to
t 0 r(i, t) 2 t 0 1 t
160 the projective measurements of these particle numbers
128 followed by post-selecting the two trajectories with a to-
+1 −1
96
tal occupancy 1 on the two associated modes.
64 +1 +1 −2
32 To understand the phases of the class-A moni-
+1 +1 −2 Sp tored dynamics, we study the circuit as shown in
0 i i
1 32 64 96 128 1 32 64 96 128 0 2 4 128 Fig. 4(a), where each two-site gate implements the post-
selected nweak measurementothrough the Kraus opera-
FIG. 4. (a) Spacetime geometry of a class-A monitored
tors exp ±α(c†i,A cj,B + h.c.) with the probability of the
circuit on a 1D complex fermion chain with each unit cell
(dashed box) containing two sublattices A and B. Two-site sign ± following the (post-selected) Born rule. To access
(post-selected) measurements (purple and orange gates) and different area-law phases, we consider different measure-
onsite unitary gates (gray gates) are defined in the main text. ment strengths α0,1 for the two types of measurements
(b) Average steady-state mutual information IA,B as a func- (purple and orange gates) in Fig. 4(a) corresponding to
tion of α0 with tanh2 α0 + tanh2 α1 = 1. (c) Spacetime con-
i = j and i = j + 1, with tanh2 α0 + tanh2 α1 = 1. Hence,
figuration r(i, t) for a class-A monitored circuit on a 128-site
chain, encoding DWs of different integer classes as labeled. when the pairs of fermion modes (ci,A , ci,B ) are strongly
(d-e) Entanglement contour and total entanglement entropy (weakly) measured, the pairs (ci,A , ci+1,B ) are weakly
(in binary logarithm) of the physical chain in a typical quan- (strongly) measured. Each gray gate is an onsite unitary
tum trajectory. gate defined above with independent random phases θ1,2 .
In Fig. 4(b), we numerically calculate the average
steady-state MI IA,B for two antipodal intervals A and B
Monitored dynamics with charge conservation and chi- [in the same geometry as in Fig. 1(b)] as a function of α0
ral symmetry— Now, we study the monitored dynamics to map out the phase diagram. This phase diagram con-
in a different AZ symmetry class, class A. The class-A tains two area-law phases (separated by α0 = α0,c ) where
area-law phases in one spatial dimension correspond to the chain develops a steady-state dimerization pattern
disordered insulators in the same class in two spatial di- between the two sublattices within each unit cell or span-
mensions [26], implying the Z classification for both area- ning neighboring unit cells. In general, there should be
law phases and their DWs. Below, we study the behavior an area-law dynamical phase associated with the range-
DTDMs in class-A monitored dynamics. R dimers between (ci,A , ci+R,B ) for every R ∈ Z, match-
A 1D monitored system in class A is constructed on ing the Z classification of the corresponding 2D class-
complex fermions on a bipartite lattice, with sublattices A disordered insulators [26, 47, 48]. The DW between
A and B [dark and light brown dots in Fig. 4(a)]. We two area-law dynamical phases with different dimeriza-
denote the fermion operators on the two sublattices in tion ranges R (on the left) and R′ (on the right) is ex-
the ith unit cell as c†i,A/B and ci,A/B . Monitored dynam- pected to carry |R − R′ | DTDMs, each being an effective
ics in class A are defined by the dynamics that respect complex-fermion mode of the same sublattice type (A or
the U(1) charge conservation and an anti-unitary chiral B depending on the sign of R−R′ ). The entanglement on
symmetry C, defined by ci,A → −c†i,A , ci,B → c†i,B , i → −i, the DTDMs is protected from being quenched by local
in each quantum trajectory (see Sec. IV 2 in SM for the measurements because the chiral symmetry C required by
identification of symmetry class A). In other words, the symmetry class A only allows modes from the sublattices
U(1) and C actions must commute with all unitary gates A and B to be measured in pairs. Therefore, each DW
and Kraus operators capturing the measurements. should be classified by ∆R = R′ − R ∈ Z.
To preserve U(1) and C, the unitary gates must act To numerically study the entanglement carried by the
only within each sublattice. Here, wenchoose the simplesto DTDMs, we generalize the two-site measurements in
case with onsite unitary gates, exp iθ1 c†i,A ci,A − 21 Fig. 4(a) to measurements with their ranges controlled by
n  o a smoothly-varying function r(i, t) in spacetime. For the
and exp iθ2 c†j,B cj,B − 12 , with θ1,2 ∈ R (see Sec. IV 3 unit cell i at time t, we measure (with post-selection) the
in SM for longer-range unitary gates). These symmetries fermion pairs (ci,A , ci+⌊r⌋,B ) and (ci,B , ci+⌊r⌋+1,A ) with
require the Kraus operators capturing the non-unitary respective measurement strengths tanh2 α = 0.04, 0.96,
 5

with a probability ⌊r⌋ + 1 − r. Otherwise, we measure Acknowledgements.— C.M.J thanks Andreas W. W.

the pairs (ci,A , ci+⌊r⌋+1,B ) and (ci,B , ci+⌊r⌋+2,A ) with the Ludwig for collaboration on related topics. H.P. is sup-
same strengths. Intuitively, the dominating measurement ported by the National Science Foundation (Platform
at spacetime coordinates (i, t) favors the dimers on the for the Accelerated Realization, Analysis, and Discov-
pair (ci,B , ci+R,A ) with the range R the closest integer to ery of Interface Materials (PARADIM)) under Coopera-
r(i, t)+1. With this generalization (and the unitary gates tive Agreement No. DMR-2039380, and US-ONR grant
unchanged), we can study the class-A monitored circuit No. N00014-23-1-2357. C.-M.J. is supported by the Al-
with various DWs incorporated through the r(i, t) con- fred P. Sloan Foundation through a Sloan Research Fel-
figuration shown in Fig. 4(a) on a 128-site chain. Similar lowship.
to class DIII, we introduce a reference chain, maximally
entangled with the physical chain at t = 0. We calculate
the EC sp (i) of the physical chain’s entanglement with
the reference, and the total EE Sp [Figs. 4(d-e)].
[1] B. Skinner, J. Ruhman, and A. Nahum, Measurement-
In Fig. 4(c), we mark the topological classes ∆R ∈ Z Induced Phase Transitions in the Dynamics of Entangle-
for each DW based on the dimerization favored by the ment, Physical Review X 9, 031009 (2019).
configuration r(i, t). Fig. 4(d) shows that the peaks of EC [2] Y. Li, X. Chen, and M. P. A. Fisher, Quantum Zeno ef-
always follow the DWs with each peak’s EC integrated to fect and the many-body entanglement transition, Physi-
exactly |∆R| log 2, as expected for |∆R| complex-fermion cal Review B 98, 205136 (2018).
DTDMs. DWs with the same sign of ∆R can pass [3] Y. Li, X. Chen, and M. P. A. Fisher, Measurement-
through each other, e.g., at t = 32, with their EC peaks driven entanglement transition in hybrid quantum cir-
cuits, Physical Review B 100, 134306 (2019).
unaffected. However, the EC peaks of the DWs with op- [4] A. Chan, R. M. Nandkishore, M. Pretko, and G. Smith,
posite signs of ∆R (partially) annihilate each other when Unitary-projective entanglement dynamics, Physical Re-
the DWs meet (e.g., at t = 96, 128), reflecting that the view B 99, 224307 (2019).
sign of ∆R decides the sublattice character of the DT- [5] A. C. Potter and R. Vasseur, Entanglement Dynamics
DMs. The EC analysis also indicate that when two DWs in Hybrid Quantum Circuits, in Entanglement in Spin
merge, their classes ∆R’s simply add. The EC peaks no Chains: From Theory to Quantum Technology Applica-
longer follow the DWs beyond t > 128 because the phys- tions, edited by A. Bayat, S. Bose, and H. Johannes-
son (Springer International Publishing, Cham, 2022) pp.
ical system has disentangled from the reference. The cir-
211–249.
cuits with longer-range unitary gates studied in Sec. IV 3 [6] M. P. A. Fisher, V. Khemani, A. Nahum, and S. Vi-
of SM confirm the robustness of the DTDMs as long as jay, Random Quantum Circuits, Annual Review of Con-
the symmetries U(1) and C are preserved. This robust- densed Matter Physics 14, 335 (2023).
ness is comprised when C is broken (see Sec. IV 4 in SM). [7] R. Vasseur, A. C. Potter, Y.-Z. You, and A. W. W. Lud-
Summary and discussion— In this work, we study the wig, Entanglement Transitions from Holographic Ran-
dom Tensor Networks, Physical Review B 100, 134203
topological properties of area-law entangled phases and
(2019).
their DWs in free-fermion monitored dynamics, specifi- [8] M. J. Gullans and D. A. Huse, Dynamical Purification
cally showcasing examples in AZ symmetry classes DIII Phase Transition Induced by Quantum Measurements,
and A in one spatial dimension. Via a correspondence to Physical Review X 10, 041020 (2020).
2D disordered topological insulators and superconduc- [9] M. J. Gullans and D. A. Huse, Scalable Probes of
tors in the respective symmetry classes, we show that Measurement-Induced Criticality, Physical Review Let-
the area-law phases in class-DIII (class-A) free-fermion ters 125, 070606 (2020).
[10] A. Zabalo, M. J. Gullans, J. H. Wilson, S. Gopalakrish-
monitored circuits on 1D chains follow a Z2 (Z) classifi-
nan, D. A. Huse, and J. H. Pixley, Critical properties of
cation. The DW between different area-law phases, also the measurement-induced transition in random quantum
classified by Z2 and Z respectively, carries DTDMs with circuits, Physical Review B 101, 060301 (2020).
robust entanglement protected from being quenched by [11] S. Choi, Y. Bao, X.-L. Qi, and E. Altman, Quan-
the measurement in the surrounding area-law dynamics. tum Error Correction in Scrambling Dynamics and
We verify the expected topological types of DTDMs and Measurement-Induced Phase Transition, Physical Re-
their robustness by designing and numerically simulating view Letters 125, 030505 (2020).
[12] Y. Li, X. Chen, A. W. W. Ludwig, and M. P. A. Fisher,
the class-DIII and class-A monitored circuits with DWs.
Conformal invariance and quantum nonlocality in critical
For symmetry class DIII where the DTDMs are effective hybrid circuits, Physical Review B 104, 104305 (2021).
unmeasured Majorana modes, we devise a braiding pro- [13] C.-M. Jian, Y.-Z. You, R. Vasseur, and A. W. W. Ludwig,
tocol for these DTDMs by controlling the DW dynam- Measurement-induced criticality in random quantum cir-
ics. Our study can be readily generalized to free-fermion cuits, Physical Review B 101, 104302 (2020).
monitored dynamics in higher dimensions in all AZ sym- [14] Y. Bao, S. Choi, and E. Altman, Theory of the phase
metries classes. It highlights an interesting direction to transition in random unitary circuits with measurements,
Physical Review B 101, 104301 (2020).
explore the topology within area-law entangled dynamics
[15] N. Lang and H. P. Büchler, Entanglement transition in
phases in monitored quantum systems.
 6

the projective transverse field Ising model, Physical Re- 10.48550/arXiv.2307.13038 (2023).
view B 102, 094204 (2020). [33] A. Lavasani, Y. Alavirad, and M. Barkeshli, Topological
[16] X. Turkeshi, R. Fazio, and M. Dalmonte, Measurement- Order and Criticality in (2 + 1)D Monitored Random
induced criticality in (2+1)-dimensional hybrid quantum Quantum Circuits, Physical Review Letters 127, 235701
circuits, Physical Review B 102, 014315 (2020). (2021).
[17] M. Ippoliti, M. J. Gullans, S. Gopalakrishnan, D. A. [34] G. Kells, D. Meidan, and A. Romito, Topological transi-
Huse, and V. Khemani, Entanglement Phase Transitions tions in weakly monitored free fermions, SciPost Physics
in Measurement-Only Dynamics, Physical Review X 11, 14, 031 (2023).
011030 (2021). [35] A. Nahum and B. Skinner, Entanglement and dynamics
[18] A. Nahum, S. Roy, B. Skinner, and J. Ruhman, Measure- of diffusion-annihilation processes with Majorana defects,
ment and Entanglement Phase Transitions in All-To-All Physical Review Research 2, 023288 (2020).
Quantum Circuits, on Quantum Trees, and in Landau- [36] X. Cao, A. Tilloy, and A. De Luca, Entanglement in
Ginsburg Theory, PRX Quantum 2, 010352 (2021). a fermion chain under continuous monitoring, SciPost
[19] A. Zabalo, M. J. Gullans, J. H. Wilson, R. Vasseur, Physics 7, 024 (2019).
A. W. W. Ludwig, S. Gopalakrishnan, D. A. Huse, and [37] O. Alberton, M. Buchhold, and S. Diehl, Entanglement
J. H. Pixley, Operator Scaling Dimensions and Multi- Transition in a Monitored Free-Fermion Chain: From Ex-
fractality at Measurement-Induced Transitions, Physical tended Criticality to Area Law, Physical Review Letters
Review Letters 128, 050602 (2022). 126, 170602 (2021).
[20] Y. Li, R. Vasseur, M. P. A. Fisher, and A. W. W. Ludwig, [38] M. Buchhold, Y. Minoguchi, A. Altland, and S. Diehl, Ef-
Statistical Mechanics Model for Clifford Random Tensor fective Theory for the Measurement-Induced Phase Tran-
Networks and Monitored Quantum Circuits, Physical Re- sition of Dirac Fermions, Physical Review X 11, 041004
view B 109, 174307 (2024). (2021).
[21] S. Sang and T. H. Hsieh, Measurement-protected quan- [39] I. Poboiko, I. V. Gornyi, and A. D. Mirlin, Measurement-
tum phases, Physical Review Research 3, 023200 (2021). Induced Phase Transition for Free Fermions above One
[22] Y. Han and X. Chen, Measurement-induced criticality Dimension, Physical Review Letters 132, 110403 (2024).
in Z2 -symmetric quantum automaton circuits, Physical [40] K. Chahine and M. Buchhold, Entanglement phases, lo-
Review B 105, 064306 (2022). calization, and multifractality of monitored free fermions
[23] A. Lavasani, Y. Alavirad, and M. Barkeshli, in two dimensions, Physical Review B 110, 054313
Measurement-induced topological entanglement transi- (2024).
tions in symmetric random quantum circuits, Nature [41] L. Fidkowski, J. Haah, and M. B. Hastings, How Dynam-
Physics 17, 342 (2021). ical Quantum Memories Forget, Quantum 5, 382 (2021).
[24] U. Agrawal, A. Zabalo, K. Chen, J. H. Wilson, A. C. [42] J. Merritt and L. Fidkowski, Entanglement transitions
Potter, J. H. Pixley, S. Gopalakrishnan, and R. Vasseur, with free fermions, Physical Review B 107, 064303
Entanglement and charge-sharpening transitions in U(1) (2023).
symmetric monitored quantum circuits, Physical Review [43] X. Chen, Y. Li, M. P. A. Fisher, and A. Lucas, Emer-
X 12, 041002 (2022). gent conformal symmetry in nonunitary random dynam-
[25] F. Barratt, U. Agrawal, S. Gopalakrishnan, D. A. Huse, ics of free fermions, Physical Review Research 2, 033017
R. Vasseur, and A. C. Potter, Field theory of charge (2020).
sharpening in symmetric monitored quantum circuits, [44] Q. Tang, X. Chen, and W. Zhu, Quantum criticality
Physical Review Letters 129, 120604 (2022). in the nonunitary dynamics of (2 + 1)-dimensional free
[26] C.-M. Jian, B. Bauer, A. Keselman, and A. W. W. Lud- fermions, Physical Review B 103, 174303 (2021).
wig, Criticality and entanglement in nonunitary quantum [45] H. Guo, M. S. Foster, C.-M. Jian, and A. W. W. Ludwig,
circuits and tensor networks of noninteracting fermions, Field theory of monitored, interacting fermion dynamics
Physical Review B 106, 134206 (2022). with charge conservation, arXiv:2410.07317 (2024).
[27] C.-M. Jian, H. Shapourian, B. Bauer, and A. W. W. Lud- [46] A. Altland and M. R. Zirnbauer, Nonstandard symme-
wig, Measurement-induced entanglement transitions in try classes in mesoscopic normal-superconducting hybrid
quantum circuits of non-interacting fermions: Born-rule structures, Physical Review B 55, 1142 (1997).
versus forced measurements, arXiv:2302.09094 (2023). [47] A. W. W. Ludwig, Topological phases: Classification
[28] M. Fava, L. Piroli, T. Swann, D. Bernard, and A. Nahum, of topological insulators and superconductors of non-
Nonlinear Sigma Models for Monitored Dynamics of Free interacting Fermions, and beyond, Physica Scripta T168,
Fermions, Physical Review X 13, 041045 (2023). 014001 (2016).
[29] I. Poboiko, P. Pöpperl, I. V. Gornyi, and A. D. Mirlin, [48] X.-L. Qi and S.-C. Zhang, Topological insulators and
Theory of free fermions under random projective mea- superconductors, Reviews of Modern Physics 83, 1057
surements, Physical Review X 13, 041046 (2023). (2011).
[30] M. Fava, L. Piroli, D. Bernard, and A. Nahum, A [49] Y. Chen and G. Vidal, Entanglement contour, Journal
tractable model of monitored fermions with conserved of Statistical Mechanics: Theory and Experiment 2014,
U(1) charge, arXiv:2407.08045 (2024). P10011 (2014).
[31] S. Majidy, U. Agrawal, S. Gopalakrishnan, A. C. Pot- [50] J. Alicea, Y. Oreg, G. Refael, F. von Oppen, and
ter, R. Vasseur, and N. Y. Halpern, Critical phase and M. P. A. Fisher, Non-Abelian statistics and topologi-
spin sharpening in SU(2)-symmetric monitored quantum cal quantum information processing in 1D wire networks,
circuits, Physical Review B 108, 054307 (2023). Nature Physics 7, 412 (2011).
[32] A. Chakraborty, K. Chen, A. Zabalo, J. H. Wilson, [51] T. Karzig, Y. Oreg, G. Refael, and M. H. Freedman,
and J. H. Pixley, Charge and Entanglement Critical- Universal Geometric Path to a Robust Majorana Magic
ity in a U(1)-Symmetric Hybrid Circuit of Qubits, Gate, Physical Review X 6, 031019 (2016).
 7

[52] M. A. Nielsen and I. L. Chuang, Quantum Computation

and Quantum Information: 10th Anniversary Edition,
1st ed. (Cambridge University Press, 2012).
[53] S. Bravyi, Lagrangian representation for fermionic lin-
ear optics, Quantum Information & Computation 5, 216
(2005).
[54] S. Bravyi and D. Gosset, Complexity of Quantum Impu-
rity Problems, Communications in Mathematical Physics
356, 451 (2017).
[55] M.-D. Choi, Completely positive linear maps on complex
matrices, Linear Algebra and its Applications 10, 285
(1975).
[56] A. Jamiolkowski, Linear transformations which preserve
trace and positive semidefiniteness of operators, Reports
on Mathematical Physics 3, 275 (1972).
[57] M. Ericsson and Centro di Ricerca Matematica En-
nio De Giorgi, eds., Quantum Information and Many
Body Quantum Systems: Proceedings, Centro Di Ricerca
Matematica Ennio De Giorgi / CRM Series No. 8 (Ed.
della Normale, Pisa, 2008).
[58] C. V. Kraus, N. Schuch, F. Verstraete, and J. I. Cirac,
Fermionic projected entangled pair states, Physical Re-
view A 81, 052338 (2010).
[59] J. T. Chalker and P. D. Coddington, Percolation, quan-
tum tunnelling and the integer Hall effect, Journal of
Physics C: Solid State Physics 21, 2665 (1988).
[60] P. W. Brouwer, A. Furusaki, I. A. Gruzberg, and
C. Mudry, Localization and Delocalization in Dirty Su-
perconducting Wires, Physical Review Letters 85, 1064
(2000).
[61] P. W. Brouwer, A. Furusaki, C. Mudry, and
S. Ryu, Disorder-induced critical phenomena–new uni-
versality classes in Anderson localization, arXiv:cond-
mat/0511622 (2006).
[62] A. W. W. Ludwig, H. Schulz-Baldes, and M. Stolz, Lya-
punov Spectra for All Ten Symmetry Classes of Quasi-
one-dimensional Disordered Systems of Non-interacting
Fermions, Journal of Statistical Physics 152, 275 (2013).
 1

Supplemental Materials for “Topological Modes in Monitored Quantum Dynamics”

CONTENTS where H and O are real and complex antisymmetric ma-

trix (with Hij , Im(Oij ) → ±∞ allowed), respectively.
I. Free-fermion monitored circuit 1 Here, the Gaussian operator O includes the unitary gates
1. Efficient simulation of Gaussian circuits 1 and the Kraus operators associated with measurements.
2. Entanglement measures in the covariance Our goal is to show how the state evolution of ρ un-
matrix formalism 2 der O, i.e., OρO† , can be efficiently simulated using co-
variance matrix formalism, which is also the numerical
II. Correspondence between a free-fermion monitored method in the main text.
circuit and Anderson localization problem 2 To this end, we first define the covariance matrix Γρ
for the Gaussian state ρ as
III. Class-DIII monitored circuit 3
1. Kraus operator description 3 i
[Γρ ]ij = tr([γ̂i , γ̂j ]ρ). (S-I.2)
2. Evolution time and ensemble size 3 2
3. Domain-wall dynamics 3 The covariance matrix for the Gaussian operator O needs
4. Braiding protocol 4 a special treatment following the Choi-Jamiolkowski iso-
morphism [45, 55, 56] which generalizes the original
IV. Class-A monitored circuit 4 Hilbert space spanned by γ̂i to a doubled Hilbert space
1. Class-AIII monitored circuit 4 spanned by η̂i and ξˆi (two sets of Majorana fermions in
2. Class-A monitored circuit with onsite unitary a different space from γ̂i ) in the following form
gates 5  
3. Class-A monitored circuit with longer-range ΓO,ηη ΓO,ηξ
ΓO = , (S-I.3)
unitary gates 5 ΓO,ξη ΓO,ξξ
4. Compromised domain-wall modes under broken
symmetries 6 where each block is defined as
a. Measurements on the same sublattice 6 [ΓO,αβ ]ij =
b. Unitary gates on different sublattices 6 !
1 P Y 1 + iη̂i ξˆi P ∗
η̂ O η̂ η̂ O η̂
tr e i,j i ij j e i,j i ji j i[α̂i , β̂j ] ,
2 i
2
Appendix I: Free-fermion monitored circuit
(S-I.4)

In this section, we provide a brief review of the free- with α, β taking η or ξ for the corresponding blocks in
fermion monitored circuit. In the following, we will refer Eq. (S-I.3).
to free-fermion states as Gaussian states and the opera- With these two covariance matrices representation Γρ
tors that keep the many-body fermionic states “free” as and ΓO for the state ρ and operator O, we can readily
Gaussian operators. Here, the Gaussianity refers to the obtain the resulting state ρ′ = OρO† in the covariance
fact that, in the free-fermion setting, the system’s den- matrices representation Γρ′ using the following contrac-
sity operator and the operations on them all take a Gaus- tion formula [53]
sian form, i.e., exponentiated fermion bilinears. We will
−1
demonstrate, through a concrete example, that a free- Γρ′ = ΓO,ξξ + ΓO,ξη ΓO,ηη + Γ−1
ρ Γ⊺O,ξη . (S-I.5)
fermion circuit consisting of Gaussian operators can be
efficiently simulated in a covariance matrix formalism. In numerical simulation, we initialize a random pure
For a more rigorous treatment, we refer the interested Gaussian state Γρ0 in a L-site Majorana chain as
readers to Refs. [53, 54]. L/2  
M 0 Πi
Γρ0 = V V ⊺, (S-I.6)
−Πi 0
i=1
1. Efficient simulation of Gaussian circuits
with a random matrix V ∈ SO(2L) and Πi = ±1, and
We start with a generic Gaussian state in its den- repeat the contraction formula Eq. (S-I.5) for various O
sity matrix form ρ and Gaussian operators O, both con- to simulate the circuit in O(L3 ) complexity.
structed in the Majorana basis γ̂i , in the following form For the class-DIII monitored circuit, we choose all Πi =
P P 1. For the class-A monitored circuit, due to the U(1)
iγ̂i Hij γ̂j γ̂i Oij γ̂j
ρ∝e i,j ,O ∝ e i,j , (S-I.1) charge conservation, we park ourselves at the half-filling
 2

P by choosing a random profile of Πi = ±1 subject

sector the Anderson localization problem in 2 spatial dimen-
to i Πi = 0. sions. We will focus on the version of this correspondence
After the contraction in Eq. (S-I.5), the resulting state between the 1+1D monitored circuit of free Majorana
could be slightly deviated away from a pure state as fermions without any constraints and the 2D Anderson
|Πi | < 1 due to the numerical error. Therefore, we purify localization problem of AZ symmetry class DIII. We will
the state once in a while by block-diagonalizing the co- extend the discussion to class AIII and class A. For more
variance matrix into Eq. (S-I.6), and resetting all Πi to details, we refer the interested readers to Ref. [13].
its nearest ±1. We start with a generic free-fermion monitored circuit
on a Majorana fermion chain without any additional sym-
metry, as shown in Fig. 1(a). We can treat the time and
2. Entanglement measures in the covariance matrix spatial axis on an equal footing and view the circuit as a
formalism Gaussian tensor network [57, 58] on a 2D square lattice.
Therefore, the evolution of the Gaussian state under the
Besides the efficient simulation of the circuit in the co- Gaussian operator O can be interpreted as the contrac-
variance matrix formalism, all entanglement measures for tion of Gaussian tensors. For example, in the covariance
a pure Gaussian state ρ can be also efficiently computed matrix formalism, this evolution performs the contrac-
in this formalism. Below, we will demonstrate how to tion between a Gaussian state ρ and a Gaussian operator
efficiently compute the entanglement entropy (EE), en- O following Eq. (S-I.5). In a given quantum trajectory,
tanglement contour (EC), and the mutual information the quantum gates in the circuit can be viewed as the
(MI) for a Gaussian state encoded in the covariance ma- many-body transfer operator of the corresponding Gaus-
trix Γρ . We refer the interested readers to Ref. [49] for a sian tensor network. These quantum gates (or trans-
more rigorous treatment. fer operators), due to their Gaussianity, admit a “first-
The von Neumann EE between a region A and its com- quantized” transfer matrix representation describing the
plement Ā of state ρ is obtained from its reduced density transformation of Majorana fermion operators γ̂i under
matrix ρA = trĀ ρ, whose covariance
P matrix representa- O, i.e.,
tion can be obtained as ΓA = i,j∈A [Γρ ]i,j |i⟩⟨j|. X
In the covariance matrix formalism, the von Neumann Oγ̂i O−1 = tij γ̂j . (S-II.1)
EE of ΓA is computed as SA = tr f (ΓA ) with j

  If O is a unitary gate, we know that t is a real orthogonal

1 + iΓA 1 + iΓA matrix in SO(N ) where N is the number of the Majorana
f (ΓA ) = − log . (S-I.7)
2 2 modes. However, when we include measurements, we can
show that the transfer matrix t become a complexified or-
The EC is a spatially-resolved EE [49], indicating the
thogonal matrix in SO(N, C), i.e. t is a complex N × N
contribution of each site to the total EE, which can be
matrix satisfying tT t = 1 [26]. We caution that, when O
obtained by taking the ith diagonal element of the matrix
is associated with a projective measurement, O−1 and,
of f (ΓA ), namely
consequently, t becomes singular. In this case, we should
view the projective measurements as a limit of weak mea-
sA (i) = [f (ΓA )]ii , (S-I.8)
surements with infinite measurement strength. For weak
with f (ΓA ) defined in Eq. (S-I.7). This definition ensures measurements of finite strength, the transfer matrix t is
that the entanglement contour non-singular. In a monitored circuit, different measure-
P on each site i sums up ment outcomes (and different random realizations of the
to the total EE of A, i.e., i∈A sA (i) = SA . Here, we
generalize the original version of the EC in Ref. [49] from unitary gates) correspond to different realizations of the
complex fermions to Majorana fermions. gates O in the circuit. Therefore, the 1+1D monitored
Finally, to identify the volume-law and area-law circuit of free Majorana fermions shown in Fig. 1(a) can
phases, we use MI IA,B between two antipodal regions be interpreted as a 2D random Gaussian tensor network
A and B, each of one-quarter of the system, defined as whose single-particle transfer matrices are sampled from
the complexified orthogonal group SO(N, C).
IA,B = SA + SB − SA∪B . (S-I.9) A 2D Gaussian tensor network with SO(N, C) single-
particle transfer can also be interpreted as a Chalker-
Coddington network model [59] that describes a network
Appendix II: Correspondence between a of disordered elastic scatters in the AZ symmetry class
free-fermion monitored circuit and Anderson DIII [26]. This Chalker-Coddington network model cap-
localization problem tures the Anderson localization problem in static disor-
dered Hamiltonian of non-interacting fermions in 2 spa-
In this section, we briefly review the correspondence tial dimensions in symmetry class DIII. It was shown
between a free-fermion monitored circuit in 1+1D and that, in a non-interacting-fermion Hamiltonian system
 3

as MDIII (p) = K+
 DIII DIII
in symmetry class DIII, namely a system equipped with (p), K− (p), KUDIII (p, θ) , where
a time-reversal symmetry (that squares to −1) and a
particle-hole (that squares to 1), the single-particle im- DIII √ 1 ± iγ̂i γ̂i+1
K± (p) = p ,
purity scattering can be captured by SO(N, C) transfer 2 (S-III.1)
KUDIII (p, θ) = 1 − p exp(θγ̂i γ̂i+1 ).
p
matrices [60–62]. This fact enables the aforementioned
identification between the Gaussian tensor network and
DIII
the Anderson localization problem. Here, K± captures the wave function collapse associ-
From the complementary perspective of the monitored ated with the two measurement outcomes iγ̂i γ̂i+1 = ±1
circuit, even if we do not impose any symmetry require- while KUDIII describes the unitary evolution.
√ As will be
√
ment on the Majorana chain (other than fermion parity), clear below, their prefactors p and 1 − p encode the
when we consider the evolution of the system’s density probabilities of applying a measurement and a unitary
matrix in a doubled Hilbert space, extra symmetries that operation at the given quantum gate.
relate the “bra” and “ket” copy of the Hilbert space can Given a (normalized) state |ψ⟩ undergoing this quan-
emerge. In the case of the monitored Majorana chain, the tum gate action, each K ∈ MDIII corresponds to a quan-
total emergent symmetry belongs to AZ symmetry class tum trajectory in which the state evolves as
DIII [26]. The correspondence between the monitored K |ψ⟩
circuit of free Majorana fermions in 1+1D and the 2D |ψ⟩ → . (S-III.2)
∥|Kψ⟩∥
disordered Anderson localization problem in symmetry
class DIII implies that the area-law entangled phases in The Born-rule probability for this quantum trajectory is
√
the former setting and the disordered topological super- given√by ⟨ψ|K † K|ψ⟩. One can see that prefactors p
conductors in the latter problem should share the same and 1 − p in Eq. (S-III.1) enter ⟨ψ|K † K|ψ⟩ as the
topological classification, which is Z2 [47, 48]. probabilities for applying a measurement or a unitary
When we add an extra symmetry constraint to the operation. This Kraus operator ensemble MDIII satisfies
monitored circuit, the group of transfer matrices changes, the positive-operator valued measure (POVM) condition,
i.e., K∈MDIII K † K = 1, which guarantees the normal-
P
and so does the symmetry class. The identification with
the Anderson localization problem remains correct in the ization of probability.
respective symmetry class. When we consider a moni-
tored circuit with U(1) charge conservation, the transfer
2. Evolution time and ensemble size
matrix belongs to GL(N, C), and the symmetry class be-
comes class AIII. Furthermore, adding an anti-unitary
chiral symmetry, as we do in the main text, limits the In Fig. 1(b), we show the average steady-state MI IA,B .
transfer matrix to the group U(N, N ), resulting in sym- Here, the steady state is reached by simulating the circuit
metry class A. Concrete examples of the quantum gates following the Born rule up to L time steps. The average
and their transfer matrices in class AIII and class A are MI IA,B is obtained by first averaging over the location
provided in Sec. IV 1 and Sec. IV 2. of the antipodal region A and B while keeping the size of
the two regions and their relative distance fixed, and then
sampling across 1000 realizations of different circuits and
measurement outcomes.
Appendix III: Class-DIII monitored circuit Both evolution time and ensemble size also apply to
the class-A and class-AIII monitored circuit, as shown in
Fig. 4 and Fig. S1, respectively.
In this section, we first provide a more formal treat-
ment of the two types of operators in the class-DIII mon-
itored circuit using the language of Kraus operators and 3. Domain-wall dynamics
then describe the details of the simulation, including the
evolution time, ensemble size, the domain-wall dynamics, In Fig. 2, we construct DWs in spacetime to explore
and the braiding protocol. area-law phases with different topologies and the DT-
DMs in between. This DW is obtained by different
parametrizations of the ensemble of Kraus operators. For
example, in Fig. 2(a) in the main text, the two class-
1. Kraus operator description
DIII area-law phases are obtained by p ≡ podd = 0.1
and 0.9, respectively. Therefore, we can parametrize a
In the class-DIII monitored circuit as shown in spacetime-dependent p(i, t) to interpolate these two p val-
Fig. 1(a), each gate, including all the effects of the proba- ues. In addition, to avoid DWs from changing abruptly,
bilistic application of the unitary gates and the measure- we smoothen the change across the two regions with dif-
ment, can be captured by an ensemble of Kraus operators ferent parametrizations using a tanh(ki) function. We
 4

choose k = 0.5 in numerics. This same logic also applies this assumption is not essential— In Fig. 3, we use a more
to the class-A monitored dynamics, as shown in Fig. 4(c) generic configuration by programming p(i, t) to vary be-
tween 0.15 and 0.85 with the smoothness of the DW k
being 0.5, to demonstrate that the braiding protocol is
4. Braiding protocol robust in the presence of symmetry-allowed unitary op-
erators.
In Fig. 3, we utilize a T-junction model consisting of
three L-site Majorana chains in open boundary condi-
tions (assuming L to be even) to swap DTDMs α2 and Appendix IV: Class-A monitored circuit
β1 . Below, we show the details of the braiding proto-
col, which consists of the following four stages, and then In this section, we extend the scope beyond the class-
explain the pairwise MI in each stage. DIII monitored circuit to include a U(1) symmetry. By
embedding each complex fermion into a pair of Majorana
1. We initialize the T-junction into a product state fermions, the circuit can be efficiently simulated within a
on all three chains, where each chain has a
fixed unified covariance matrix formalism. As discussed in the
a a
parity on odd Majorana pairs γ̂2i−1 , γ̂2i with main text, the class-A monitored circuit requires both a
a = {A, B, C}. At time t = t0 , we create two U(1) symmetry and an anti-unitary chiral symmetry C.
DWs by configuring p(i, t) to
dimerize even pairs We take a two-step process in the following. We will first
a a
of Majorana modes, γ̂2i , γ̂2i+1 , within the region discuss the monitored dynamics with only the U(1) sym-
enclosed by the two DWs, in chain a = {A, B}. metry required, which belongs to symmetry class AIII.
Without loss of generality, we place the DWs at The ensemble of Kraus operators in this class will be
L 3L
4 and 4 , creating two DTDMs, corresponding to provided. To obtain the monitored dynamics in symme-
unmeasured Majorana zero modes, α1 , α2 (β2 , β1 ) try class A, we post-select the quantum trajectories in
in chain A (B). Each pair of DTDMs is maximally class AIII that respect an extra anti-unitary chiral sym-
entangled within the chain, leading to all three pair- metry C. As it turns out, post selections are inevitable
wise MIs being 0. (The site index can be chosen for class-A free-fermion monitored circuits. For class A,
to run from the outermost to the center of the T- we present the Kraus operator description for the on-
junction.) site unitary gate and the measurement gate as shown in
2. At time t = t1 , we move the DTDM α2 from chain Fig. 4(a), and then generalize the onsite unitary gate to
A to chain C by moving the DW from chain A to a more generic long-range unitary gate. Finally, we show
chain C, treating the two chains as a single chain. that the robustness of the DTDMs in the class-A moni-
Therefore, we need to measure the two Majorana tored circuit is comprised when the required symmetries
sites at the intersection between chains A and C, are broken.
γ̂LA and γ̂LC , leading to the MI IA,C being 2 log 2
at t1 , as the long-range entangled Majorana pair
1. Class-AIII monitored circuit
contributes log 2 and the local entangled pair at the
intersection between A and C contributes another
log 2. We start with a class-AIII monitored circuit, which can
be obtained by imposing a U(1) symmetry to the class-
3. At time t = t2 , we move the DTDM β1 from chain DIII monitored circuit. The transfer matrix tAIII for a
B to chain A using the same procedure as in the 2D class AIII ensemble is in the complexified general lin-
previous step. To make β1 leave chain B and land ear group GL(N, C) [62]. Therefore, we can parametrize
on chain A, we need to measure the innermost Ma- it as tAIII = exp(µAIII ), with µAIII ∈ CN ×N . In order to
jorana modes between chains B and C, leading to later connect to class-A monitored circuit, we can choose
the MI IB,C being log 2 coming from the local en- N = 2 and construct the ensemble of Kraus operators
tangled pair at the intersection, and both IA,B and as MAIII (α, θ1 , θ2 , i, j) = {KsAIII
+ ,s−
(α, θ1 , θ2 , i, j)|s± ∈
IA,C being log 2 from their respective long-range en- {−1, +1}}, where
tangled DTDMs.
1 † 1 † 1
KsAIII
+ ,s−
(α, θ1 , θ2 , i, j) = es+ α(c+ c+ − 2 )+s− α(c− c− − 2 )
4. At time t = t3 , we move the DTDM α2 from chain 2 cosh α
† †
C to chain B using the same procedure by treating 1 1
eiθ1 (ci,A ci,A − 2 ) eiθ2 (cj,B cj,B − 2 ) .
the two chains as a single chain. Since chain C re-
(S-IV.1)
turns to its initial state, the only entangling chains
are A and B with the MI of 2 log 2. Here, the first line in Eq. (S-IV.1) corresponds to a weak
For the convenience of the description above, we assume
measurement with  strength α ∈ [0, ∞) on the fermion
an “ideal” case in a measurement-only limit. However, modes c†± = √1
2
c†i,A ± c†j,B for s+ = −s− = ±1, or on
 5

the fully-occupied (empty) fermion modes for s+ = s− = MA (α, θ1 , θ2 , i, j) = {K±

A
(α, θ1 , θ2 , i, j)} where
+1 (s+ = s− = −1). The second line in Eq. (S-IV.1)
1 †
corresponds to unitary gates where θ1 , θ2 are uniformly
−1
A
K± (α, θ1 , θ2 , i, j) = e±α(ci,A cj,B +h.c.)
distributed. The prefactor (2 cosh α) is to ensure the 2 cosh α (S-IV.2)
† 1 † 1
POVM condition. Due to the absence of topological eiθ1 (ci,A ci,A − 2 ) eiθ2 (cj,B cj,B − 2 ) .
phases in 2D class AIII, the class-AIII monitored cir-
cuit always shows an area-law phase corresponding to a This is the ensemble of Kraus operators used in Fig. 4.
trivial disordered insulator as shown in Fig. S1. Here, From Eq. (S-IV.1) for class AIII to Eq. (S-IV.2) for
we follow the same staggered pattern of α0 and α1 as in class A, we exclude the two possible outcomes with s+ =
Fig. 4 but with a different set of Kraus operators defined s− = ±1, corresponding to the fully-empty and fully-
in Eq. (S-IV.1). occupied state, respectively. In fact, even considering a
generic U(N ,N ) for the transfer matrix, which involves
measurement of N pairs of complex fermions, it becomes
L = 16 inevitable to sacrifice POVM for the chiral symmetry C in
0.06 L = 32
L = 64
class-A monitored circuit, because the fully-empty state
L = 128 and fully-occupied state are always decoupled and do not
0.04
respect the chiral symmetry C by themselves.
IA,B

Due to the absence of POVM, we adopt a post-

selection Born rule by renormalizing
 the Bornprobabili-
0.02 ties for the two fermion modes √2 c†i,A ± c†j,B as
1

A


p K± (α, θ1 , θ2 , i, j)
0.00 D
A

† A E
0.0 0.2 0.4 0.6 0.8 1.0 K± (α, θ1 , θ2 , i, j) K± (α, θ1 , θ2 , i, j)
tanh2 α0 = P D E.
†
(KsA (α, θ1 , θ2 , i, j)) KsA (α, θ1 , θ2 , i, j)
s=±1
Supplementary Figure S1. Average steady-state mutual in-
formation IA,B as a function of the measurement strength
(S-IV.3)
α0 in class-AIII monitored circuit, showing only one area-law
phase.
(a) (b) sp (i) (c)
t 0 1 t
0.4 160
L = 16 L = 64
L = 32 L = 128 128
96
IA,B

Area law Area law

2. Class-A monitored circuit with onsite unitary 64
gates
32
2 Sp
0.0 tanh α0 0 i
In Fig. 4, we study the class-A monitored circuit by fur- 0 tan2 α0,c 1 1 32 64 96 128 0 2 4 128
ther imposing an additional anti-unitary chiral symmetry
to the class-AIII monitored circuit. Below, we show that Supplementary Figure S2. (a) Phase diagram of steady-
this is equivalent to post-select the outcomes that respect state class-A monitored circuit with longer-range unitaries
the chiral symmetry in the class-AIII monitored circuit. in Eq. (S-IV.4). Average steady-state mutual information
IA,B as a function of the measurement strength α0 , with
We start with the formal description of transfer matrix tanh2 α0 + tanh2 α1 = 1, showing two topologically different
tA for a class A ensemble, which is an indefinite unitary area-law phases separated by a critical point. (b-c) Entan-
group U(N, N ) [62], The minimal example is U(1, 1) with glement contour and total entanglement entropy (in binary
each basis for the two sublattice A and B, leading to the logarithm) of the physical chain in a typical quantum tra-
constraint as tA † σz tA = σz , with σz being the Pauli Z jectory. The peaks in the entanglement contour indicate the
spacetime position of the dynamical topological domain-wall
matrix. One can show that this constraint on the single-
modes.
particle transfer matrix tA exactly matches the require-
ment of an anti-unitary chiral symmetry C, defined by
ci,A → −c†i,A , ci,B → c†i,B , i → −i, as introduced in the
main text. 3. Class-A monitored circuit with longer-range
Here, it can be shown that by setting s+ = −s− unitary gates
in Eq. (S-IV.1), the aforementioned constraint for the
transfer matrix tA can be satisfied, leading to the en- In Fig. 4, we choose the onsite unitary gate in the
semble of Kraus operators for the class-A monitored as Kraus operators defined in Eq. (S-IV.2). However, this
 6

choice is merely a convenient choice for illustrative pur- 4. Compromised domain-wall modes under broken
poses. In principle, any generic (nonlocal) unitary gate symmetries
acting on a range-R dimer from the same sublattice re-
spects the chiral symmetry and thus should demonstrate In previous sections for class-A monitored circuits, we
the same classification of area-law phases as in Fig. 4(b). ensure that unitary gates (measurements) always act on
Therefore, the Kraus operators can be generalized with the pair of complex fermion sites within the same sublat-
the longer-range unitary gate as per tice (across the opposite sublattices) to respect the chiral
symmetry C. In this section, we will show that if we in-
1 †
clude the measurements within the same sublattice, or
A
K̃± (α, θ1 , θ2 , i, j) = e±α(ci,A cj,B +h.c.) the unitary gates across the sublattices, the DTDMs will
2 cosh α (S-IV.4)
θ θ be comprised. This directly manifests that the DTDMs
i 21 (c†i,A cj,A −ci,A c†j,A ) i 22 (c†i,B cj,B −ci,B c†j,B )
e e . in Fig. 4 are indeed protected by the symmetry from be-
ing quenched by the local symmetry-allowed disorder.

The phase diagram of the class-A monitored circuit

using longer-range unitary gates is shown in Fig. S2(a). a. Measurements on the same sublattice
The difference between Fig. S2(a) and Fig. 4(b) is merely
quantitative, which does not change the existence of two We start with the class-A monitored circuit, and de-
area-law phases with different topologies separated by a sign the spacetime profile of r(i, t) to create two DWs at
critical point. However, due to the long-range nature of i = L4 and i = 3L 4 initially for time t < 32 as shown in
the unitary gate in Eq. (S-IV.4), it is harder to reach the Fig. S3(a). After t = 32, we introduce the Kraus opera-
IR limit, and thus we reduce the variance of the longer- tors, which allow measurement on complex fermion sites
range unitary gate from θ1,2 ∈ [−π, π] to [− π4 , π4 ], which within the same sublattice. Therefore, we can choose
effectively decreases the correlation length [13]. Eq. (S-IV.2) as the ensemble of Kraus operators, with
Following the same procedure as the DW dynamics in r× ≡ i − j (the subscript × is to emphasize that the
the class-A monitored circuit in Figs. 4(c-e), we can per- Kraus operators break the chiral symmetry C). Here, we
form the same DW dynamics with the longer-range uni- still parametrize r× (i, t) to have two “DWs” at i = L4 and
tary gate in Eq. (S-IV.4) as shown in Figs. S2(b-c). We i = 3L 4 for a direct comparison with the r(i, t), though
use the same spacetime profile of r(i, t) as in Fig. 4(c), strictly speaking, both regions are now in the same type
and verify the same DTDMs with Z classification, con- of area-law phases.
sistent with the onsite unitary gate in the main text. We present the EC to track the DTDMs in Fig. S3(b).
We find that DTDMs are immediately comprised after
introducing the measurements on the same sublattice.
(a) r(i, t) r× (i, t) (b) (c)
sp sp
t0 1 0 1 t 0 1 t 0 1
64
b. Unitary gates on different sublattices

32 Apart from measurements on the same sublattice, we

can also introduce the unitary gates that can couple the
0 i i i different sublattices, breaking the chiral symmetry C as
1 32 64 96 128 1 32 64 96 128 1 32 64 96 128 well. These chiral-symmetry-broken Kraus operators can
be chosen as
Supplementary Figure S3. (a) Parametrization of chiral-
1 †
symmetry-preserving r(i, t) before time t = 32, and chiral- K±,× (α, θ1 , θ2 , i, j) = e±α(ci,A cj,B +h.c.)
symmetry-breaking r× (i, t) after time t = 32 (b-c) Entangle- 2 cosh α (S-IV.5)
θ1 † † θ2 † †
ment contours (in binary logarithm) in the class-A monitored ei 2 (ci,A ci,B −ci,B ci,A ) ei 2 (cj,A cj,B −cj,B cj,A ) ,
circuit for t < 32. For t > 32, the chiral symmetry is broken
by (b) measurements on the same sublattice or (c) unitary
with r× ≡ i − j defined the same as before.
gates on different sublattices, both leading to the compromis-
ing of the DTDMs. Ensembles of Kraus operators are (b) In Fig. S3(c), we present the EC for such chiral-
Eq. (S-IV.2) and (c) Eq. (S-IV.5). symmetry-breaking Kraus operators, and find that the
DTDMs are also compromised once unitary gates on dif-
ferent sublattices are applied after t > 32.