Weak Quantum Chaos

Ivan Kukuljan,1 Sašo Grozdanov,2 and Tomaž Prosen1

1
University of Ljubljana, Faculty of Mathematics and Physics, Jadranska ulica 19, SI-1000 Ljubljana, Slovenia
2
Instituut-Lorentz for Theoretical Physics, Leiden University,
Niels Bohrweg 2, Leiden 2333 CA, The Netherlands
(Dated: February 1, 2017)
Out-of-time-ordered correlation functions (OTOC’s) are presently being extensively debated as
quantifiers of dynamical chaos in interacting quantum many-body systems. We argue that in quan-
tum spin and fermionic systems, where all local operators are bounded, an OTOC of local observables
arXiv:1701.09147v1 [cond-mat.stat-mech] 31 Jan 2017

is bounded as well and thus its exponential growth is merely transient. As a better measure of quan-
tum chaos in such systems, we propose, and study, the density of the OTOC of extensive sums of
local observables, which can exhibit indefinite growth in the thermodynamic limit. We demonstrate
this for the kicked quantum Ising model by using large-scale numerical results and an analytic so-
lution in the integrable regime. In a generic case, we observe the growth of the OTOC density to
be linear in time. We prove that this density in general, locally interacting, non-integrable quan-
tum spin and fermionic dynamical systems exhibits growth that is at most polynomial in time—a
phenomenon, which we term weak quantum chaos. In the special case of the model being integrable
and the observables under consideration quadratic, the OTOC density saturates to a plateau.

Introduction.—Quantum chaos was an active area bility of semi-classical trajectories of electrons scattered
of research in the 80’s and 90’s [1–3]. The main suc- by impurities in a superconductor. Consequently, ex-
cess of the field was a random matrix theory (RMT) tended quantum systems were defined as chaotic if there
classification of universal properties of quantum systems exists a pair of local observables, w and v, such that the
whose classical counterparts are chaotic. The classical OTOC (1) grows exponentially at early times [11, 12]:
limits of such systems have positive Lyapunov expo-
nents, which characterise exponential sensitivity to ini- C (x, t) ∝ eλL (t−|x|/vB ) . (2)
tial conditions—the so-called butterfly effect. However,
since the (classical) definition of the Lyapunov exponent Motivated by the semi-classical picture, λL is referred to
is based on the concept of phase-space trajectories, one as the Lyapunov exponent and vB the butterfly velocity.
cannot unambiguously translate it to the quantum realm. A multitude of works examining the properties of
Nevertheless, it has been argued that a weaker prop- quantum chaos have recently been written both from the
erty of dynamical mixing—a decay of almost all con- high-energy perspective (typically in models with long-
nected temporal correlators—is sufficient to establish range interactions and in theories with holographic grav-
universal quantum chaotic behaviour, such as random ity duals) and from the condensed matter perspective
matrix statistics of energy spectra [4] or the universal (typically in experimentally more feasible models with
exponential decay of Loschmidt echoes [5]. In the theory local interaction) [12–43].
of dynamical systems, complex (mixing) dynamics that In this work, we investigate systems with local inter-
displays no exponential butterly effect is referred to as actions with extensive number N → ∞ of degrees of
weak chaos (see Ref. [6] and references therein). Exam- freedom, but with a finite local Hilbert space dimension
ples of such dynamical systems include generic polygonal D. In any model with a finite D (including all fermionic
billiards in which nearby trajectories deviate only lin- and spin lattice models), in which local operators u, v are
early with time, while correlation functions nevertheless bounded, the exponential growth in (2) can be bounded
exhibit mixing [7, 8]. by operator norm inequalities (the triangular inequality,
The study of dynamical mixing (now called scrambling) kabk ≤ kakkbk and haiβ ≤ kak):
and Lyapunov chaos in quantum mechanics was recently 2 2
revived by the high-energy physics community, initially C (x, t) ≤ 4 kvk kwk . (3)
in the context of the propagation of information in black Thus, the OTOC can only grow exponentially up to a fi-
hole backgrounds [9]. In 2014, Kitaev proposed to quan- nite (scrambling) time t∗ , after which it remains bounded
tify chaos in interacting quantum many-body systems by a constant. This is consistent with the observations
[10] in terms of the following out-of-time-ordered (four- made in other works on OTOC’s (of local observables) in
point) correlation function (OTOC): fermionic systems where OTOC’s were always observed
C (x, t) = −h[wx (t), v0 (0)]2 iβ , (1) to reach a plateau [27, 29–33]. As already noted in [28],
the only way for the exponential time evolution to persist
where wx , vx are local observables and h•iβ denotes the to late times is if there is a small prefactor multiplying
thermal expectation value at inverse temperature β. The the exponential function in (2). Even in the Sachdev-Ye-
concept is based on a work by Larkin and Ovchinnikov Kitaev (SYK) model with long-range interactions, this
[11] from 1969, where OTOC was connected to the insta- prefactor is 1/N , which becomes small as N → ∞ [15].
 2

Exponential growth (2) of the OTOC is therefore at best Moreover, we report below the results of extensive nu-
a transient effect in systems of interest to this work. merical and analytical calculations, which demonstrate
If interactions are local, C(x, t) can be further bounded that possibly the simplest non-trivial locally interacting
by the Lieb-Robinson theorem (LRT) [44] (see also [18]): quantum chaotic spin system: the kicked Ising (KI) quan-
tum spin chain [47, 48], exhibits linear growth of the
2 2
C (x, t) ≤ 4 kvk kwk e−µ max{0,|x|−vLR t} . (4) dOTOC of extensive magnetisation observables, c(t) ∝ t.
An exception is the integrable KI model (equivalent to
In this case, for t  t∗ = |x| /vLR , the OTOC is even a free fermion model), for which we show analytically
more suppressed. The interpretation of this effect is clear: that its dOTOC of extensive quadratic observables (in
namely, t∗ is the time in which C(x, t) enters the causal fermionic variables) saturates, c(t → ∞) = const. Since
cone. Before t∗ , C(x, t) is almost zero, while after t∗ , it is the KI model seems to be generic, we further conjec-
bounded by (3) and saturates at a plateau. The dynamics ture that the bound (6) is not optimal and that typical
can only be non-trivial near the edge of the causal-cone one-dimensional, non-integrable and locally interacting
(or for t ∼ t∗ ), where C(x, t) can vary greatly. This is models exhibit linear growth of dOTOC’s.
consistent with [27, 29]. As a consequence, theories under consideration in this
Another important fact is that momentum operators— work are not expected to exhibit any late-time butterfly
the observables that Ref. [11] originally used to compute effect, but as we know from results in the RMT, can
the Lyapunov exponent of the semiclassical trajectories— still be chaotic. In reference to classical mixing systems
are unbounded. Therefore, if we wanted to preserve the without the butterfly effect, we term the phenomenon of
semiclassical justification of the OTOC, which is nec- infinite polynomial growth of dOTOC’s weak quantum
essary to be able to speak about quantum chaos, the chaos.
quantum observables under consideration must have un- Kicked quantum Ising model.—The Hamiltonian
bounded spectra. of the one-dimensional KIPmodel consists of the Ising-
These observations can be summarised in the intuitive interaction term HIsing = j Jσjx σj+1x
and the kick term
statement that if chaos is to fully develop over long time, P z x


the observables have to provide enough “space” for this to Hkick = j h σj cos ϕ + σj sin ϕ :
happen; they need to be unbounded. Indeed, this is the X
case with general observables in bosonic systems (usu- H(t) = HIsing + Hkick δ (t − n) , (7)
ally studied in holography). However, this condition is n∈Z

not fulfilled by local observables in fermionic or spin sys- where σjα are local Pauli spin operators. The model has
tems, or more generally, in systems with a finite D. On three parameters: the Ising coupling J, the magnitude
the other hand, extensive observables in such theories do of the external magnetic field h and the inclination of
satisfy the unbounded spectrum criterium and therefore the external magnetic field ϕ. KI is a periodic (in time)
have the capacity to fully unveil the system’s dynamical system with the Floquet propagator:
properties and quantum chaos. Motivated by this fact, n R1 o
we propose a new measure of quantum chaos: the density U = T e−i 0 dtH(t)
of the OTOC P (dOTOC) of P (non-local) extensive opera- σjx σj+1
x
(σjz cos ϕ+σjx sin ϕ) .
P P
−ih
tors V ≡ x∈Λ vx , W ≡ x∈Λ wx , with wx , vx local. It = e−iJ j e j (8)
is defined on a d−dimensional lattice Λ with N sites as
Because of the temporal periodicity, KI dynamics can be
the centralised second moment of the commutator
viewed as discrete in time, or as a quantum cellular au-
1  2

tomaton. The effect of a perturbation on a single lattice
c(N ) (t) := − h[W (t), V (0)]2 iβ − h[W (t), V (0)]iβ .
N site propagates in a causal-cone with speed 1. Namely,
(5) information can spread only by one site, left or right,
The disconnected part, which is just the square of the within one period (kick of the magnetic field). Random
standard dynamical susceptibility (i.e. the response func- matrix analysis [49, 50] revealed that KI is chaotic.
tion), has been subtracted to make the dOTOC well de- The system has a further nice property of being inte-
fined in the thermodynamic limit (TL) for any temper- grable (quasi-free) for transverse magnetic field, ϕ = 0,
ature. Because of the cyclicity of the trace, this term and non-integrable (and interacting) for ϕ > 0. Thus,
vanishes at β = 0 (this will occur in the model that we ϕ serves as a handy parameter which allows us to study
study below). Using the LRT and the clustering prop- integrability breaking. See e.g. [48, 51] for a survey of
erty of thermal states, which holds for any temperature elementary dynamical properties of the KI model.
in d = 1 [45] and for sufficiently high temperature in Here, we study the KI chain with N spins and eval-
d > 1 [46], in Appendix A, we rigorously prove that the (N )
uate the dOTOC (5) cα (t) for a (non-local) extensive
dOTOC satisfies a uniform (in N ) polynomial bound PN α
magnetisation, W = V = Mα = j=1 σj , which can
c(N ) (t) ≤ At3d , (6) either be transverse (α = z) or parallel (α = x) to the
direction of the Ising interaction. We take β = 0 as an
where A is an (N, t)−independent constant. The same infinite-temperature Gibbs ensemble is the only mean-
bound equally holds in the TL, c(t) := limN →∞ c(N ) (t). ingful equilibrium state for periodically driven systems,
 3


which generically heat up to infinite temperature. We w(θ)
with shorthand notation w(θ) = . One can show
use three different approaches, two numerical methods w0 (θ)
for the general inclination (0 ≤ ϕ ≤ π2 ) and an analyt- (Appendix C) that the Floquet
 † propagator in the Heisen-
ical solution for the transverse field case ϕ = 0. In the U w(θ)U
berg picture, U w(θ) := , takes the follow-
first, appropriate for small system sizes (up to N ∼ 12), U † w0 (θ)U
we used the exact numerical Floquet operator (8). The ing form in Fourier transformed Majorana basis:
second method, used for intermediate system sizes (up
to N ∼ 22), was a Monte-Carlo wave-function sampling U (J, h, θ) = Ukick (J, h, θ)UIsing (J, h, θ) (14)
based on typicality arguments (explained in Appendix   iθ

B). The analytical solution in the TL for the integrable cos(2h) − sin(2h) cos(2J) e sin(2J)
= .
(transverse) case and transverse magnetisation Mz , was sin(2h) cos(2h) −e−iθ sin(2J) cos(2J)
found using fermionisation. We outline the main steps This 2 × 2 unitary matrix valued symbol can be diago-
for obtaining the analytical solution in what is to follow. nalised as:
Analytical solution.—For the transverse field (ϕ =
0), KI is a quasi-free model. If, furthermore, the (exten-
 iκ(J,h,θ) 
e
sive) observable of interest is simple enough, the dOTOC U (J, h, θ) = V † (J, h, θ) V (J, h, θ),
e−iκ(J,h,θ)
allows for an analytic solution in terms of Jordan-Wigner (15)
transformation of Pauli spins into staggered Majorana where
fermion operators
κ(J, h, θ) = arccos [cos(2J) cos(2h) +
   
Y Y
w2j =  σkz  σjx , w2j+1 =  σkz  σjy , (9) + cos(θ) sin(2J) sin(2h)] , (16)
k<j k<j

and V (J, h, θ) is given explicitly in Appendix D.

obeying the anti-commutation relations {wi , wj } = 2δij .
Knowing that the KI Majorana fermions in the Fourier
The Floquet operator (8) then takes the following form:
basis time evolve as w(θ, t) = U (θ)t w(θ, 0) allows us to
P P
define the real space propagator as:
U = e−J j w2j−1 w2j −h
e j w2j w2j+1
. (10)
kj
It is clear from (10) that the KI model is free for ϕ = 0 Kab (t) := hw2k+a−1 w2j+b−1 (t)i , (17)
[52]. Now, the transverse magnetisation can be expressed
as a sum of quadratic Majorana operators:
for a, b ∈ {1, 2}. K kj can then be computed from U (θ)
X (Appendix E):
Mz = −i w2j w2j+1 , (11)
j∈Z Z π
1
K kj (t) := K j−k (t) = dθe−iθ(j−k) U t (θ). (18)
which enables the analytic computation of the dOTOC 2π −π
of Mz [53]. Power-expanding the Floquet operator (10)
2
and using (wi wj ) = −1 for i 6= j, U further simplifies to Using the propagator (18), we can compute the infi-
Y nite temperature OTOC of the transverse magnetisation,
U= (cos (J) − w2j−1 w2j sin (J)) · (N )
cz (t). First, we express the terms in (5) using (11),
j
Y e.g. hσiz (t)σjz σkz (t)σlz i as an eight-fermion expectation
· (cos (h) − w2k w2k+1 sin (h)) = UIsing Ukick . (12) value hw2i (t)w2i+1 (t)w2j w2j+1 w2k (t)w2k+1 (t)w2l w2l+1 i.
k Then, using (17), these are expressed as the prod-
uct of four propagators (one for each time-dependent
Since the transverse field model is free, it is convenient fermion) times an equal-time eight-fermion expectation
to work in the Fourier transformed Majorana basis: value, with terms summed over four spatial and spin
X X indices (see Appendix F for details). Simple algebraic
w(θ) = w2j eiθj , w0 (θ) = w2j+1 eiθj , (13) manipulations then lead to the final expression for the
j j dOTOC in the TL:

j6=0 2
R (p ) R (p˜ )
X X p1 +p3 1 1
cz (t) = −4 (−1) KS(p 1 ),1
(t)KS(1p˜1 ),2
1
(t) ·
j,l1 ,l3 ∈Z s0 ,sj ,p1 ,p3 =1
h i
s +s R3 (p3 ) R (p˜3 ) R3 (p3 ) R (p˜3 )
· (−1) j 0 KS̃(p ),1
(t)KS̃(3p˜ ),2 (t) − KS(p 3 ),1
(t)KS(3p˜3 ),2 (t) , (19)
3 3
 4

where we used the following notation: R1 := (l1 − j, l1 ), any way. What we find is that the plateau height ceases
R3 := (l3 −j, l3 ), S := (sj , s0 ), together with the notation to be smooth for J = h. Beyond that, we also checked
v = (v(1), v(2)) for vector components and 1̃ := 2, 2̃ := 1. the slope of the OTOC for longitudinal magnetisation
We can use the formula (19) in two different ways. For Mx in the vicinity of this line. It turns out that the slope
intermediate times t ∼ 50, we can perform the integral exhibits a peak, but not exactly on the line J = h. This
in (18) exactly and evaluate the sums in (19), which, could be the effect of a small system size, which was nec-
because of the causal-cone spreading of information, now essary for numerics. It is plausible that the peak may
become finite sums (see Appendix G 1 for details). align with J = h in the TL.
Secondly, we can use the stationary phase approxi- This work should be considered as a starting point
mation in combined (18), (15) and (19) to compute the for future investigations of quantum, weakly chaotic sys-
large-t asymptotic behaviour of the dOTOC. In this way, tems, which exhibit dynamical late-time mixing but do
we prove that for large times, cz (t) is a constant (de- not display any exponential butterfly effect due locality
pendent only on J and h). In other words, the dOTOC of interactions and finiteness of the local Hilbert space.
of quadratic extensive observables in the integrable KI In such systems, the standard OTOC rapidly plateaus
model saturates to a plateau. Details are explained in and is therefore not a good measure of chaos. This ob-
Appendix G 2. servation led us to propose of a new measure of chaos:
density of the OTOC of non-local extensive operators.
Results and discussion.—In summary, we observe We have proven (Appendix A) that such correlators al-
two distinct behaviours of the OTOC density for exten- ways exhibit a polynomial bound and can thus be widely
sive observables in a one-dimensional KI model. For a used to diagnose and classify quantum chaos. In the case
generic situation, unless the model is integrable and the of the non-integrable KI model studied here, the growth
observable quadratic, the extensive dOTOC grows lin- is linear. Intuitively, it seems apparent that in locally in-
early with time. In fact, numerical results for finite sys- teracting systems, information propagates slower than in
tem sizes saturate to a plateau at t ∼ N/2, but this is an all-to-all interacting theory like the SYK model. The
simply due to a finite size effect—a consequence of the speed is limited by the Lieb-Robinson velocity. However,
causal cone coming around the periodic boundary. This what is less apparent is that such systems can still be
plateau grows with an increasing system size N and we chaotic; a result established by an RMT analysis [49].
expect that it disappears in the TL N → ∞. In the Lastly, we note that in order to study chaos in strongly
regime where the model is integrable (free) and the ob- coupled, large-N theories (even in those that do exhibit
servable is simple (quadratic in fermion operators), the the buttery effect), it would be interesting to extend holo-
dOTOC saturates to a genuine plateau despite the fact graphic calculations to computations of OTOC’s of non-
that the spectrum of the observable is unbounded. The local, smeared operators. For detailed future analyses,
latter statement was proven in this work by finding an ex- we will likely need to utilise the full machinery of holo-
plicit analytic solution for cz (t) from which the expression graphic n-point function calculations [54–56] that will ex-
for the height of the plateau for a given set of parameters tend beyond studying gravitational shock waves [13, 14].
J and h could be found. The results of the time depen- Acknowledgements.— We thank M. Medenjak for
dence of the extensive dOTOC for different scenarios are fruitful discussions and E. Ilievski for useful comments.
presented in Figure 1 and explained in the caption. The work has been supported by the ERC grant OMNES
For the integrable case with ϕ = 0, the quasiparticle and grants P1-0025, P1-0044 of Slovenian Research
spectral gap closes on the line of J = h in the param- Agency (ARRS). S. G. is supported in part by a VICI
eter space and the system exhibits a Floquet analogue grant of the Netherlands Organization for Scientific Re-
of a quantum phase transition, i.e. κ (J = h, θ = 0) = search (NWO) and by the Netherlands Organization for
−κ (J = h, θ = 0) = 0 (cf. (15), (16)). It is interesting Scientific Research/Ministry of Science and Education
to ask whether the OTOC also reflects this transition in (NWO/OCW).

[1] M. C. Gutzwiller, Chaos in Classical and Quantum Me- chaos, infinite ergodic theory, and anomalous dynamics
chanics (Springer-Verlag New York, 1990). in: X. Leoncini and M. Leonetti (Eds.), From Hamilto-
[2] F. Haake, Quantum Signatures of Chaos, 3rd ed. nian Chaos to Complex Systems (Springer, Berlin, 2013),
(Springer-Verlag, 2010). p.3-42.
[3] H.-J. Stöckmann, Quantum Chaos: An Introduction [7] G. Casati and T. Prosen, Phys. Rev. Lett. 83, 4729
(Cambridge University Press, 2007). (1999).
[4] A. V. Andreev, O. Agam, B. D. Simons, and B. L. Alt- [8] G. Casati and T. Prosen, Phys. Rev. Lett. 85, 4261
shuler, Phys. Rev. Lett. 76, 3947 (1996). (2000).
[5] T. Gorin, T. Prosen, T. H. Seligman, and M. Žnidarič, [9] Y. Sekino and L. Susskind, JHEP 10, 065 (2008),
Physics Reports 435, 33 (2006). arXiv:0808.2096 [hep-th].
[6] R. Klages, arXiv:1507.04255 [nlin], book chapter: Weak [10] A. Kitaev, “Hidden correlations in the hawking radiation
 5

FIG. 1. Density of the OTOC of extensive observables for one-dimensional KI model (7) with periodic boundary conditions
is presented for four possible regimes. In the upper panels (A, B), the magnetic field is transversal (ϕ = 0) so the system is
integrable (free), while in the lower panels (C, D) the field is tilted (ϕ = π4 ) so the model is non-integrable. In the left panels
(A, C) the observable is a sum of quadratic Majorana terms (11), while in the right panels (B, D), the observable is a sum of
terms composed of infinite Majorana strings (composite). Here J = 0.7 and h = 1.1 but the behaviour was found qualitatively
similar for other values of J, h. The numerically exact results for small system sizes are plotted with crosses. Results obtained
with numerical method based on typicality arguments (with a sample of 50 × 50 random vectors) are plotted with error bars.
The analytical solution for the integrable case and quadratic observable is plotted with a bold black line. The asymptotic
behaviour in the limits N → ∞ and t → ∞ is plotted with a dashed line. In the integrable + quadratic case the dashed line
is the result of our analytic solution. In other cases it is an extrapolation based on numerics. The numerical results start to
deviate around t ∼ N/2 due to finite size effects. The inset (i) shows the dependence of the plateau height on the parameters
J and h.

and thermal noise,” (2014), talk given at Fundamental [14] D. A. Roberts, D. Stanford, and L. Susskind, JHEP 03,
Physics Prize Symposium. 051 (2015), arXiv:1409.8180 [hep-th].
[11] A. I. Larkin and Y. N. Ovchinnikov, Soviet Journal of [15] J. Polchinski and V. Rosenhaus, JHEP 04, 001 (2016),
Experimental and Theoretical Physics 28, 1200 (1969). arXiv:1601.06768 [hep-th].
[12] J. Maldacena, S. H. Shenker, and D. Stanford, JHEP [16] J. Maldacena and D. Stanford, Phys. Rev. D94, 106002
08, 106 (2016), arXiv:1503.01409 [hep-th]. (2016), arXiv:1604.07818 [hep-th].
[13] S. H. Shenker and D. Stanford, JHEP 03, 067 (2014), [17] K. Jensen, Phys. Rev. Lett. 117, 111601 (2016),
arXiv:1306.0622 [hep-th]. arXiv:1605.06098 [hep-th].
 6

[18] D. A. Roberts and B. Swingle, Phys. Rev. Lett. 117, arXiv:1609.01707 [cond-mat].
091602 (2016), arXiv:1603.09298 [hep-th]. [43] W. Buijsman, V. Gritsev, and R. Sprik, (2016),
[19] M. Blake, Phys. Rev. Lett. 117, 091601 (2016), arXiv:1610.06751 [quant-ph].
arXiv:1603.08510 [hep-th]. [44] E. H. Lieb and D. W. Robinson, Comm. Math. Phys. 28,
[20] M. Blake, Phys. Rev. D94, 086014 (2016), 251 (1972).
arXiv:1604.01754 [hep-th]. [45] H. Araki, Comm. Math. Phys. 14, 120 (1969).
[21] M. Blake and A. Donos, (2016), arXiv:1611.09380 [hep- [46] M. Kliesch, C. Gogolin, M. J. Kastoryano, A. Riera, and
th]. J. Eisert, Phys. Rev. X 4, 031019 (2014).
[22] A. Lucas and J. Steinberg, JHEP 10, 143 (2016), [47] T. Prosen, Progress of Theoretical Physics Supplement
arXiv:1608.03286 [hep-th]. 139, 191 (2000).
[23] M. Miyaji, JHEP 09, 002 (2016), arXiv:1607.01467 [hep- [48] T. Prosen, Phys. Rev. E 65, 036208 (2002).
th]. [49] C. Pineda and T. Prosen, Phys. Rev. E 76, 061127
[24] D. Stanford, JHEP 10, 009 (2016), arXiv:1512.07687 (2007).
[hep-th]. [50] I. Kukuljan and T. Prosen, Journal of Statistical Me-
[25] B. Swingle and D. Chowdhury, (2016), arXiv:1608.03280 chanics: Theory and Experiment 2016, 043305 (2016).
[cond-mat.str-el]. [51] T. Prosen, Journal of Physics A: Mathematical and The-
[26] G. Turiaci and H. Verlinde, JHEP 12, 110 (2016), oretical 40, 7881 (2007).
arXiv:1603.03020 [hep-th]. [52] We note that even free theories can exhibit complicated
[27] I. L. Aleiner, L. Faoro, and L. B. Ioffe, Annals Phys. 375, entangled collective behaviour when one considers the
378 (2016), arXiv:1609.01251 [cond-mat.stat-mech]. dynamics of composite operators. See e.g. [57, 58].
[28] A. A. Patel and S. Sachdev, (2016), arXiv:1611.00003 [53] Longitudinal magnetisation P is instead
Q a sum of infinite
[cond-mat.str-el]. strings of fermions, Mx = j k<j (−iw 2k w2k+1 ) w2j .
[29] Y. Huang, Y.-L. Zhang, and X. Chen, Annalen der Computation of the OTOC for Mx is therefore signifi-
Physik (2016), 10.1002/andp.201600318. cantly more involved.
[30] R.-Q. He and Z.-Y. Lu, arXiv:1608.03586 [cond-mat.dis- [54] E. Barnes, D. Vaman, C. Wu, and P. Arnold, Phys. Rev.
nn] (2016). D82, 025019 (2010), arXiv:1004.1179 [hep-th].
[31] R. Fan, P. Zhang, H. Shen, and H. Zhai, [55] P. Arnold and D. Vaman, JHEP 11, 033 (2011),
arXiv:1608.01914 [cond-mat.quant-gas] (2016). arXiv:1109.0040 [hep-th].
[32] X. Chen, T. Zhou, D. A. Huse, and E. Fradkin, Annalen [56] S. Grozdanov and A. O. Starinets, JHEP 03, 007 (2015),
der Physik , 1600332 (2016). arXiv:1412.5685 [hep-th].
[33] P. Hosur, X.-L. Qi, D. A. Roberts, and B. Yoshida, JHEP [57] S. Grozdanov and J. Polonyi, Phys. Rev. D92, 065009
02, 004 (2016), arXiv:1511.04021 [hep-th]. (2015), arXiv:1501.06620 [hep-th].
[34] N. Y. Halpern, (2016), arXiv:1609.00015 [quant-ph]. [58] J. Polonyi, Proceedings, 7th International Workshop :
[35] W. Fu and S. Sachdev, Phys. Rev. B94, 035135 (2016), Spacetime - Matter - Quantum Mechanics. (DICE2014):
arXiv:1603.05246 [cond-mat.str-el]. Castiglioncello, Tuscany, Italy, September 15-19, 2014,
[36] B. Swingle, G. Bentsen, M. Schleier-Smith, and P. Hay- J. Phys. Conf. Ser. 626, 012021 (2015), arXiv:1502.03947
den, Phys. Rev. A94, 040302 (2016), arXiv:1602.06271 [hep-th].
[quant-ph]. [59] S. Bravyi, M. B. Hastings, and F. Verstraete, Phys. Rev.
[37] M. Gaerttner, J. G. Bohnet, A. Safavi-Naini, M. L. Wall, Lett. 97, 050401 (2006).
J. J. Bollinger, and A. M. Rey, (2016), arXiv:1608.08938 [60] E. Ilievski and T. Prosen, Communications in Mathemat-
[quant-ph]. ical Physics 318, 809 (2013).
[38] H. Shen, P. Zhang, and H. Z. Ruihua Fan, [61] T. Matsui, Annales Henri Poincaré 4, 63 (2003).
arXiv:1608.02438 [cond-mat.quant-gas] (2016). [62] M. Gerken, “Measure concentration: Levy’s lemma,”
[39] Y. Chen, (2016), arXiv:1608.02765 [cond-mat.stat- (2013), lecture notes for Talk 6, on Selected topics in
mech]. Mathematical Physics: Quantum Information, Heidel-
[40] R. A. Davison, W. Fu, A. Georges, Y. Gu, K. Jensen, and berg University.
S. Sachdev, (2016), arXiv:1612.00849 [cond-mat.str-el]. [63] M. P. Müller, D. Gross, and J. Eisert, Communications
[41] N. Tsuji, P. Werner, and M. Ueda, Phys. Rev. A 95, in Mathematical Physics 303, 785 (2011).
011601 (2017). [64] S. Širca and M. Horvat, Computational Methods for
[42] E. Rozenbaum, S. Ganeshan, and V. Galitski, (2016), Physicists, Compendium for Students (Springer-Verlag
Berlin Heidelberg, 2012) Chap. 1.3.5.
 7

Appendix A: Proof of the polynomial bound on the density of extensive OTOC

In this section, we first prove that the density of the extensive OTOC (dOTOC) for 1D locally interacting trans-
lationally invariant lattice systems with finite local Hilbert space dimension cannot grow faster than with the third
power of time. Then, we directly extend our theorem, Eq. (6) of the paper, to d−dimensional regular lattices.
To derive the bound, we will take advantage of two important theorems that hold for locally interacting lattice
systems. The first, the Lieb-Robinson theorem (LRT) [44] states that for any locally interacting lattice system there
exist positive constants ξ, µ and vLR , such that for any two operators a and b:

k[a(t), b]k ≤ ξ min {|supp(a)| , |supp(b)|} kak kbk e−µ max{0,d(supp(a),supp(b))−vLR t} . (A1)

Here, supp(a) ⊂ Z denotes the support of a local operator a and d (•, •) is the distance between two sets. Roughly
speaking, the theorem says that the commutator of two local observables grows in a causal-cone, spreading with
velocity vLR .
The authors of Ref. [59] have found an elegant and useful reformulation of the LRT. Let Γ be a subset of the lattice
of N sites and define
trΓC a
a|Γ := ⊗ 1ΓC , (A2)
tr1ΓC

where ΓC denotes the set complement, to be a projection of the operator a on the sublattice Γ. Note that supp(a|Γ ) =
Γ. Then, for a given locally interacting system, the LRT is equivalent to [59, 60]

ka(t) − a(t)|Γ k ≤ ξ |supp(a)| kak e−µ max{0,d(supp(a),Γ )−vLR t} .

C
(A3)

The second theorem that we will need is the exponential clustering property of thermal states [45, 61]. For a thermal
state of a one-dimensional locally interacting system, there exist positive constants χ and ρ, such that the following
inequality is satisfied by any two operators a and b:
c
ha, biβ ≤ χ kak kbk e−ρ d(supp(a),supp(b)) , (A4)

where, in order to make the expressions in this section more compact, we have introduced the notation for the
connected (bipartite) correlation function:
c
ha, biβ := habiβ − haiβ hbiβ . (A5)

An analogous result is true for locally interacting Hamiltonians on arbitrary d−dimensional lattices for sufficiently
high temperatures [46]. As we will show, the three bounds, (A1), (A3) and (A4), imply a polynomial bound for the
dOTOC.
Our goal is to compute an upper bound on the dOTOC:
c(t) := lim c(N ) (t)
N →∞
1 X c
:= − lim h[wi (t), vj ] , [wk (t), vl ]iβ
N →∞ N
i,j,k,l∈Z
X c
= − h[wi (t), vj ] , [wk (t), v0 ]iβ
i,j,k∈Z
X
=: cijk (t). (A6)
i,j,k∈Z

Note that in the third line, we used the translational invariance of the system.
The bound can be established by first using the triangular inequality
X
c(t) ≤ |cijk (t)| , (A7)
i,j,k∈Z

and then by finding the appropriate bounds for individual terms. To take advantage of the exponential clustering
property, at every time t, we will separate the i–k plane into two domains. The first, |k − i| ≤ 2vLR t, is the region
where the causal-cones of the two commutators in (A6) are overlapping. There, the exponential clustering cannot be
 8

FIG. 2. The illustration of the main concepts needed in proving the polynomial upper bound on the dOTOC. At a given
time, we can divide the i–k plane into two regions and use different techniques to bound the contribution to the total bound
on OTOC coming from each region. We will use the intuition implied by the LRT (A1) that a commutator spreads essentially
in causal-cone and is exponentially damped outside. The first region is the one where the causal-cones corresponding to the
two commutators in (A6) overlap (for the particular choice of i and k in the drawing, this is the case for example at time t0 ).
The leading order term in the bound on OTOC (∝ t3 ) will come from this region. The second is the region where the light
cones are well separated and can be embedded into semi-infinite intervals (Γi , Γk ) with growing distance between them. This
region will contribute subleading terms (∝ t2 ).

used, but the region is bounded in |k − i| which will yield a finite contribution to the upper bound. In the second
region, |k − i| > 2vLR t, the causal-cones are well separated so we will be able to use the exponential clustering to
produce a finite upper bound. The contributions from both regions will be summed up in the end to get the overall
upper bound on the OTOC. We will treat cijk (t) slightly differently in the two regions:
c
cijk (t) := − h[wi (t), vj ] , [wk (t), v0 ]iβ
c
(
− h[wi (t), vj ] , [wk (t), v0 ]iβ ; |k − i| ≤ 2vLR t
=     c . (A8)
− wi (t)|Γi + wi (t) − wi (t)|Γi , vj , wk (t)|Γk + wk (t) − wk (t)|Γk , v0 β
; |k − i| > 2vLR t
In the proof, the following obvious bound on the connected correlation functions will be useful,
c
|ha, biβ | ≤ |habiβ | + |haiβ hbiβ | ≤ 2 kak kbk . (A9)
We will also need a rigorous estimate for the norm of a projected operator a|Γ = a − (a − a|Γ ):
k ax (t)|Γ k ≤ kax (t)k + kax (t) − ax (t)|Γ k
≤ ka k + ξ ka k e−µ max{0,d(x,Γ )−vLR t}
C
x x
≤ (1 + ξ) kak . (A10)
In the first line, we used the triangular inequality. In the second line, we utilised the fact that unitary time evolution
preserves the norm, followed by an application of the second version of the LRT, Eq. (A3). In the last line, we used
the fact that the exponential of a non-positive function can be bound by 1.
For |k − i| ≤ 2vLR t :

|cijk (t)| ≤ 2 k[wi (t), vj ]k k[wk (t), v0 ]k

2 2
≤ 2 ξ 2 kwk kvk e−µ max(0,|i−j|−vLR t) e−µ max(0,|k|−vLR t) , (A11)
where we used Eq. (A9) to bound the connected correlator and the LRT (A1) to bound the norms of the commutators.
We can now sum over the corresponding domain to get the contribution to the overall upper bound:
X k+2v
XLR t X X k+2v
XLR t X
2 2
|cijk (t)| ≤ 2 ξ 2 kwk kvk e−µ max(0,|i−j|−vLR t) e−µ max(0,|k|−vLR t)
k i=k−2vLR t j k i=k−2vLR t j
  ! !
X X k+2v
XLR t
2 2 2 −µ max(0,|i−j|−vLR t)  −µ max(0,|k|−vLR t)
= 2 ξ kwk kvk e e 1
j k i=k−2vLR t
  µ    µ 
2 2
= 2 ξ 2 kwk kvk 2vLR t + coth
2vLR t + coth (1 + 4vLR t)
2 2
2 2 3
= 32 ξ 2 kwk kvk (vLR t) + O t2 .


(A12)
 9

In the second line, we used the fact that each individual sum is independent of all other coefficients.

For |k − i| > 2vLR t :

We can now use the linearity of commutators and thermal expectation values to write cijk (t) as a sum of
four terms and then bound it from above, again using the triangular inequality:
    c
|cijk (t)| ≤ wi (t)|Γi , vj , wk (t)|Γk , v0 β (I)
    c
+ wi (t) − wi (t)|Γi , vj , wk (t)|Γk , v0 β (II)
    c
+ wi (t)|Γi , vj , wk (t) − wk (t)|Γk , v0 β (III)
    c
+ wi (t) − wi (t)|Γi , vj , wk (t) − wk (t)|Γk , v0 β . (IV ) (A13)

Let us now find the upper bounds for each of the terms individually.
Term (I): Since the supports of the two commutators are well separated (the first commutator is different from
identity only on Γi , the second one only on Γk ), we can use the exponential clustering property of thermal states to
bound the term. This is the only place in the proof where this property of thermal states is used. However, here, it
is indeed crucial:
    c
wi (t)|Γi , vj , wk (t)|Γk , v0 β
wk (t)|Γk , v0 e−ρ d(Γk ,Γi )
   
≤ χ wi (t)|Γi , vj
2 2 2
≤ χξ 2 (1 + ξ) kwk kvk e−µ max(0,|i−j|−vLR t) Θ (j ∈ Γi ) e−µ max(0,|k|−vLR t) Θ (0 ∈ Γk ) e−ρ d(Γi ,Γk )
2 2 2
≤ χξ 2 (1 + ξ) kwk kvk e−µ max(0,|i−j|−vLR t) e−µ max(0,|k|−vLR t) e−ρ d(Γi ,Γk )
=: bound[I]ijk . (A14)
In the first line, we used the exponential clustering property. In the second line, we used the LRT together with
(A10) and the fact that the commutator is non-zero only if v0 and vj are located inside the supports of wi (t)|Γi and
wk (t)|Γk , respectively. In the third line, we used Θ (•) ≤ 1, where Θ is defined as Θ(true) = 1, Θ(false) = 0.
Note that if one wanted to compute the density of the disconnected OTOC, this bound would still be valid, but
only in the infinite temperature regime β = 0, where the expectation values of the commutators vanish because of
the cyclicity of the trace. At finite temperature,
P an estimate obtained using exponential clustering gives a divergent
P
contribution upon summation over |k−i|>2vLR t j , indicating that the disconnected dOTOC is generically not a
well defined quantity in the thermodynamic limit.
Term (II): Here, we will use (A9) to bound the term. To obtain the bound, which will give a non-divergent
P P
contribution when summed over |k−i|>2vLR t j , we will take advantage of a convenient fact that the first commutator
can be bound in two different ways—using two different versions of the LRT. One version will give us exponential
damping when |k − i| grows to infinity, the other version when |i − j| grows to infinity. For a given combination of
i, j, k, we then take the minimum of the two bounds, which results in a convergent bound upon summation over the
domain:
    c
wi (t) − wi (t)|Γi , vj , wk (t)|Γk , v0 β
 
wi (t) − wi (t)|
 Γi kvj k 
 
≤ 2 min wk (t)|Γk , v0
k[wi (t), vj ]k + wi (t)|Γi , vj
( )
−µ(d(i,ΓC i )−vLR t)
2 2
≤ 2 ξ (1 + ξ) kwk kvk min
2 e e−µ max(0,|k|−vLR t) Θ (0 ∈ Γk )
(1 + (1 + ξ) Θ (j ∈ Γi )) e−µ max(0,|i−j|−vLR t)
( )
−µ(d(i,ΓC
i )−vLR t)
2 2
≤ 2 ξ (2 + ξ) (1 + ξ) kwk kvk min
2 e e−µ max(0,|k|−vLR t)
e−µ max(0,|i−j|−vLR t)
  
2 2 d(i, ΓCi ) − vLR t
= 2 ξ 2 (2 + ξ) (1 + ξ) kwk kvk exp −µ max e−µ max(0,|k|−vLR t)
max (0, |i − j| − vLR t)
=: bound[II]ijk . (A15)
Since we only want to prove that the term will contribute to the upper bound no more (no faster) than polynomially
in time, we were allowed to make some of the terms in the third line larger by a constant factor. In the fourth line,
 10

we used the fact that the functions appearing in the exponent next to −µ are non-negative.

Term (III): In analogy with the previous term:

    c
wi (t)|Γi , vj , wk (t) − wk (t)|Γk , v0 β
 
  wk (t) − wk (t)|Γk kv0 k 
≤ 2 wi (t)|Γi , vj min
k[wk (t), v0 ]k + wk (t)|Γk , v0
( C
)
2 2
≤ 2 ξ (1 + ξ) kwk kvk e
2 −µ max(0,|i−j|−vLR t)
Θ (j ∈ Γi ) min e−µ(d(k,Γk )−vLR t)
(1 + (1 + ξ) Θ (0 ∈ Γk )) e−µ max(0,|k|−vLR t)
  
2 2 d(k, ΓC
k ) − vLR t
≤ 2 ξ 2 (1 + ξ) (2 + ξ) kwk kvk e−µ max(0,|i−j|−vLR t) exp −µ max
max (0, |k| − vLR t)
=: bound[III]ijk . (A16)

Term (IV ): Writing again all possible combinations of different versions of the LRT and taking the minimum:

    c
wi (t) − wi (t)|Γi , vj , wk (t) − wk (t)|Γk , v0 β
 

 wi (t) − wi (t)|Γi kvj k wk (t) − wk (t)|Γk kv0 k 


 wi (t) − wi (t)|Γi  kvj k k[wk (t), v0 ]k + w k (t)|Γ , v0

≤ 2 min 
k

 k[wi (t), vj ]k + wi (t)|Γi ,
vj wk (t) − wk (t)|Γk kv0 k 
 
k[wi (t), vj ]k + wi (t)|Γi , vj k[wk (t), v0 ]k + wk (t)|Γk , v0
 
C C

e−µ(d(i,Γi )−vLR t) e−µ(d(k,Γk )−vLR t)


 

C
e−µ(d(i,Γi )−vLR t) (1 + (1 + ξ) Θ (0 ∈ Γk )) e−µ max(0,|k|−vLR t)

 

2 2
≤ 2 ξ 2 kwk kvk min C


 (1 + (1 + ξ) Θ (j ∈ Γi )) e−µ max(0,|i−j|−vLR t) e−µ(d(k,Γk )−vLR t) 


(1 + (1 + ξ) Θ (j ∈ Γi )) e−µ max(0,|i−j|−vLR t) (1 + (1 + ξ) Θ (0 ∈ Γk )) e−µ max(0,|k|−vLR t)
 
C C
e−µ(d(i,Γi )−vLR t) e−µ(d(k,Γk )−vLR t)
 

 

−µ(d(i,ΓC )−vLR t) −µ max(0,|k|−vLR t)
 
(2 + ξ) e e
 i

2 2 2
≤ 2 ξ kwk kvk min C
 (2 + ξ) e−µ max(0,|i−j|−vLR t) e−µ(d(k,Γk )−vLR t) 

 

2
(2 + ξ) e−µ max(0,|i−j|−vLR t) e−µ max(0,|k|−vLR t)
 
( )
−µ(d(k,ΓC k )−vLR t)
2 2 2
≤ 2 ξ (2 + ξ) kwk kvk e
2 −µ max(0,|i−j|−vLR t)
min e
e−µ max(0,|k|−vLR t)
  
2 2 2 2 −µ max(0,|i−j|−vLR t) d(k, ΓCk ) − vLR t
= 2 ξ (2 + ξ) kwk kvk e exp −µ max
max (0, |k| − vLR t)
(2 + ξ)
= bound[III]ijk . (A17)
(1 + ξ)

In the fourth line, we used the fact that min cannot decrease if we simply omit a couple of (non-negative) functions
and if we multiply some of the remaining functions by a constant factor. We have taken the common term of the two
remaining functions out of the minimum.
We now have the estimates for all of the four terms so we are ready to sum them over the domain to get the
contribution to the bound for the dOTOC. Since the setting is reflection symmetric (upon exchanging i and k), it is
enough to compute (all the terms) for k > i + 2vLR t and double the result. In the case of k > i + 2vLR t, the subsets
are:
   
k − i − 2vLR t 2 1 1
Γi = −∞, i + vLR t + = −∞, i + k + vLR t ,
3 3 3 3
   
k − i − 2vLR t 2 1 1
Γk = k − vLR t − ,∞ = k + i − vLR t, ∞ , (A18)
3 3 3 3
 11

and the distances:

1
d(i, ΓC C
i ) = d(k, Γk ) = (k − i + vLR t) ,
3
1
d(Γi , Γk ) = (k − i − 2vLR t) . (A19)
3
We can plug these into the expressions (A14-A17) and evaluate the sums (all in the form of geometric series). We
find that the contribution to the upper bound on the dOTOC coming from the region |k − i| > 2vLR t is of order of t2 :
X X X 
(2 + ξ)
 
bound[III]ijk = O t2 ,


2 bound[I]ijk + bound[II]ijk + 1 + (A20)
j
(1 + ξ)
k i<k−2vLR t

since each of the bounds gives an O t2 contribution upon the summation.

By adding this that to the result for |k − i| ≤ 2vLR t, we arrive to the end of the proof. Hence, we have established
that the dOTOC cannot grow faster than with the third power of time:
X X X X
c(t) ≤ |cijk (t)| + |cijk (t)|
|k−i|≤2vLR t j |k−i|>2vLR t j
2 2
2 3 3
+O t 2


≤ 32 ξ kwk kvk vLR t . (A21)
This result can be straightforwardly extended in two ways:
• Even without taking the thermodynamic limit in (A6), we can still find a bound on c(N ) (t) by using exactly the
same formal steps. We only have to assume that the finite N lattice is periodic, so translational invariance can
be used.
• We may consider any regular d-dimensional lattice in the regime where the temperature is sufficiently high
for the generalisation of the exponential clustering property to hold [46]. In this case, each summation over a
positional index (with an appropriate constraint) yields a factor that scales as O(td ), rather than O(t). With
these results in hand, we finally arrive at the general polynomial bound stated in Eq. (6) of the main text.

Appendix B: Numerical evaluation of OTOC based on typicality

The approximative numerical method for evaluating the OTOC for intermediate system sizes, N ∼ 22, that we used
is based on Levy’s lemma (also referred to as the measure concentration, or typicality). The lemma states, roughly,
that in a large enough Hilbert space, the expectation value of a well-behaved observable on a single randomly chosen
quantum state will be exponentially close in probability to the ensemble average of the observable. That is, in a large
enough Hilbert space, almost any state is typical. For a precise formulation, see for example Refs. [62, 63]. Here, we
will approximate ensemble averages by averaging over a set {|Ψrand i} of random states in the Hilbert space. In this
case, for an observable A, typicality arguments lead to
1 X
haiβ=0 ≈ hΨrand | a |Ψrand i , (B1)
|{|Ψrand i}|
{|Ψrand i}

where |S | denotes the cardinality of the set S (i.e. the number of random states used in the calculation). Rather than
estimating the error of such an approximation by analytical arguments, we will estimate it numerically by computing
variances.
The numerical method for computing the OTOC is then constructed as follows. We generate two sets {|Ψ1 i} and
{|Ψ2 i} of random (normalised) vectors in the 2N dimensional Hilbert space. Then, we can compute
1 X
hW (t)V W (t)V iβ=0 ≈ hΨ1 | W (t)V W (t)V |Ψ1 i
|{|Ψ1 i}|
{|Ψ1 i}

1 2N X X
≈ hΨ1 | W (t)V |Ψ2 i hΨ2 | W (t)V |Ψ1 i
|{|Ψ1 i}| |{|Ψ2 i}|
{|Ψ1 i} {|Ψ2 i}

1 2N X X
= hΨ1 (t)| W |Ψ
f2 (t)i hΨ2 (t)| W |Ψ
f1 (t)i. (B2)
|{|Ψ1 i}| |{|Ψ2 i}|
{|Ψ1 i} {|Ψ2 i}
 12
P
In the second line, we inserted a partition of unity and approximated it by 1 = |Ψi∈H |Ψi hΨ| ≈
2N
P
{|Ψi} |Ψi hΨ|. In the third line, we defined |Ψi := V |Ψi and switched from the Heisenberg to the
e
|{|Ψi}|
Schrödinger picture. A similar expression can be derived for the other term appearing in the OTOC, namely
hV W (t)W (t)V iβ=0 = V 2 W 2 (t) β=0 .
In evaluating the dynamics of |Ψ1,2 (t)i, the computation of the action of one-spin (or two-spin) unitary operators
on a vector is numerically very efficient. This is because the operators act only on two (or four) among the composite
spin indices of the vector at the time. Thus, the operation only requires 2 · 2N (or 4 · 2N ) computational steps. The
evaluation of the
expression (B2) is composed only of one-spin and two-spin operations and can therefore be computed
within O N 2N computational steps. For a detailed discussion of such an algorithm, see for example [51].
Finally, the statistical error of such an approximation is estimated by
σc(N ) (t)
∆c(N ) (t) = p , (B3)
|{|Ψ1 i}| |{|Ψ2 i}|
n o
(N ) (N )
where σc(N ) is the variance of c|Ψ1 i,|Ψ2 i if c|Ψ1 i,|Ψ2 i is an expression like RHS of Eq. (B2) for a fixed state pair
|Ψ1 i , |Ψ2 i (omitting the averaging).

Appendix C: Kicked quantum Ising propagator in the Fourier basis

Propagators of translationally invariant free models can be simplified by writing them in the Fourier transformed
basis of Majorana fermions. This is useful for computing the action of powers of the propagator and therefore obtaining
real time dynamics.
The Fourier transformed Majorana fermions can be written as
X X
w(θ) = w2j eiθj , w0 (θ) = w2j+1 eiθj . (C1)
j j

In most of the calculations here, we can safely assume that the chain is infinite, N = ∞, and hence, θ ∈ [−π, π) is a
continuous quasi-momentum parameter. It is convenient to introduce the shorthand spinor notation
 
w(θ)
w(θ) = . (C2)
w0 (θ)
Written in the Heisenberg picture, the propagator in this basis acts as a 2 × 2 unitary matrix
 † 
U w(θ)U
U w(θ) := . (C3)
U † w0 (θ)U
As an example, let us explicitly compute the expression for the propagator of the transverse KI field (ϕ = 0) in the
Fourier basis:
P P
U = e−J j w2j−1 w2j e−h j w2j w2j+1 (C4)
Y Y
= (cos (J) − w2j−1 w2j sin (J)) (cos (h) − w2k w2k+1 sin (h)) (C5)
j k
= UIsing Ukick . (C6)
The kick term acts as
 † 
Ukick w(θ)Ukick
Ukick w(θ) = †
Ukick w0 (θ)Ukick
X  U† w U 
= kick 2j kick eiθj
†
j
U kick w2j+1 U kick

X  (cos (h) + w w 
2j 2j+1 sin (h)) w2j (cos (h) − w2j w2j+1 sin (h))
= eiθj
(cos (h) + w2j w2j+1 sin (h)) w2j+1 (cos (h) − w2j w2j+1 sin (h))
j
X  cos2 (h) − sin2 (h)
w − 2 sin (h) cos (h) w 
2j 2j+1
= eiθj
2 sin (h) cos (h) w2j + cos2 (h) − sin2 (h) w2j+1


j
 
cos(2h) − sin(2h)
= w(θ), (C7)
sin(2h) cos(2h)
 13

where in the first line, we used the definition of the propagator in the Heisenberg picture (C3). In the second line, we
employed the definition of the Fourier transformed Majorana fermions (C1). Then, in the third line, we used the fact
that wi has a non-trivial product only with terms containing wi (this follows from the Majorana anti-commutation
relations, {wi , wj } = 2δij ) and finally, in the last line, we again utilised the definition of the Fourier transform.
Similarly, for the Ising propagator,
!
X U † w2j UIsing
UIsing w(θ) = †
Ising
eiθj
j
U Ising w 2j+1 U Ising
!
†
X UIsing w2j UIsing eiθj
= †
j
UIsing w2j−1 UIsing eiθ(j−1)
X  (cos (J) + w2j−1 w2j sin (J)) w2j (cos (J) − w2j−1 w2j sin (J)) eiθj 
=
(cos (J) + w2j−1 w2j sin (J)) w2j−1 (cos (J) − w2j−1 w2j sin (J)) eiθ(j−1)
j
 
cos(2J) eiθ sin(2J)
= w(θ). (C8)
−e−iθ sin(2J) cos(2J)

We note that in the second line, we shifted the summation index.

Then, for the action of the entire Floquet propagator (note the correct order), we recover Eq. (14) from the main
text:
  
cos(2h) − sin(2h) cos(2J) eiθ sin(2J)
U = Ukick UIsing = . (C9)
sin(2h) cos(2h) −e−iθ sin(2J) cos(2J)

Appendix D: Spectrum of the kicked quantum Ising model

The Floquet propagator U of a general quadratic model in the Fourier basis is a 2 × 2 unitary matrix. We can
diagonalise it to the following form:
 
eiκ(θ)
U (θ) = V (θ)
†
V (θ), (D1)
eiλ(θ)

where V is a unitary eigenvector matrix with elements

 
v (θ) v12 (θ)
V (θ) = 11 . (D2)
v21 (θ) v22 (θ)
√

Diagonalising (C9) and using arccos (z) = −i ln z + z 2 − 1 , we find the eigenvalues to be

κ(J, h, θ) = arccos [cos(2J) cos(2h) + cos(θ) sin(2J) sin(2h)] , (D3)

λ(J, h, θ) = −κ(J, h, θ). (D4)

We see that the Floquet quasiparticle dispersion relation κ(J, h, θ) has three extrema: two maxima at θ = ±π and a
minimum at θ = 0. For J 6= h, we have κ (J 6= h, θ) > 0 so the system has a spectral gap (see Fig. 3). For J = h,
κ (J = h, θ = 0) = −κ (J = h, θ = 0) = 0, and so the gap closes. This is a Floquet-type analogue of a quantum critical
line in the J–h plane. On the J = h line, the eigenphase κ (J = h, θ) has only two extrema at θ = ±π.
For the eigenvectors, we find

e−iθ sin(2J) sin(2h) + cos(2J) cos(2h) − e−iκ(J,h,θ)

v11 (J, h, θ) = /norm1 (J, h, θ), (D5)
eiθ sin(2J) cos(2h) − cos(2J) sin(2h)
v12 (J, h, θ) = 1/norm1 (J, h, θ), (D6)
−iθ iκ(J,h,θ)
e sin(2J) sin(2h) + cos(2J) cos(2h) − e
v21 (J, h, θ) = /norm2 (J, h, θ), (D7)
eiθ sin(2J) cos(2h) − cos(2J) sin(2h)
v22 (J, h, θ) = 1/norm2 (J, h, θ), (D8)
 14

FIG. 3. Floquet quasiparticle spectrum (eigenphases) of the kicked quantum Ising model (D3), (D4). The full lines represents
generic curves for the case of J 6= h, for which the spectrum has a gap. The dashed lines represent generic curves for the case
of J = h, for which the gap closes and the system exhibits a Floquet analogue of a quantum phase transition.

where
s
2
2 [sin (κ[J, h, θ]) − sin(θ) sin(2J) sin(2h)]
norm1 (J, h, θ) = 1+ , (D9)
1 − cos (κ [2J, 2h, θ])
s
2
2 [sin (κ[J, h, θ]) + sin(θ) sin(2J) sin(2h)]
norm2 (J, h, θ) = 1+ . (D10)
1 − cos (κ [2J, 2h, θ])

At the extremal points, θ ∈ {−π, 0, π}, the eigenvectors simplify to

iΣ0/±π !
− √ √1
V0/±π = 2 2 , (D11)
iΣ0/±π
√ √1
2 2

where

Σ0 = sign [sin(2h − 2J)] , Σ±π = sign [sin(2h + 2J)] . (D12)

Note that Σ20/±π = 1.

Appendix E: Real space propagator

The propagator of a general translationally invariant Floquet system, quadratic in fermionic operators (that is,
a free model), can be expressed in the Fourier transformed basis as a unitary 2 × 2 matrix that depends on quasi-
momentum. The same formalism applies for general time-dependent situation through the application of the Trotter-
Suzuki formula. For example, we could obtain the results for the time-independent transverse field Ising model by
setting J → dt J, h → dt h and then taking the limit of dt → 0.
By using the expression for U (θ) and

w(θ, t) = U (θ)t w(θ, 0), (E1)

we compute the time-evolution of the Majorana fermions in the spatial basis; that is, the real-space propagator K,
defined by
 
w2j (t) X T
w2k w2k+1 K kj (t) ,


=: (E2)
w2j+1 (t)
k

or equivalently,
kj
Kab (t) := hw2k+a−1 w2j+b−1 (t)i , (E3)
 15

for a, b ∈ {1, 2}.

Let us start from Eq. (E1). Then,

w(θ, t) = U (θ)t w(θ). (E4)

At the same time, by definition (and by using Eq. (C1)), the above expression is equal to
 
X w2j (t)
w(θ, t) = eiθj . (E5)
w2j+1 (t)
j

Using the inverse Fourier transform

  Z π  
w2k 1 w(ϕ)
= dϕ e−iϕk , (E6)
w2k+1 2π −π
w0 (ϕ)

along with the definition (E2), the equation (E5) can then be further expressed as
Z π  
X X 1 w(ϕ)
= eiθj K kj (t) dϕ e−iϕk =. (E7)
2π −π w0 (ϕ)
j k

Now, taking into account the translational invariance of the system, i.e. K kj =: K j−k , and thereby introducing a
new index l := j − k, the expression becomes
Z π
1 X
= dϕ ei(θ−ϕ)k K l (t)eiθl w(ϕ) = . (E8)
2π −π
k,l

einx = 2π
P P
Summing over k and using n k δ (x − 2πk), we have
Z π X
= dϕδ(θ − ϕ) K l (t)eiθl w(ϕ) = . (E9)
−π l

Finally, by integrating over ϕ, we get

X
= K l (t)eiθl w(θ). (E10)
l

Comparing Eqs. (E4) and (E10), we see that

X
U (θ)t = K l (t)eiθl , (E11)
l

or equivalently, by performing the inverse Fourier transform

Z π
1
l
K (t) = dθe−iθl U t (θ). (E12)
2π −π

Appendix F: The dOTOC of transverse magnetisation

We now want to use the general form of the propagator to compute the high temperature limit of the dOTOC for
transverse magnetisation in free fermionic systems:
cz (t) := lim c(N )
z (t)
N →∞
1 D 2
E
:= − [Mz (t), Mz (0)]
N β=0
2 n o
= − hMz (t)Mz (0)Mz (t)Mz (0)iβ=0 − hMz (0)Mz (t)Mz (t)Mz (0)iβ=0
N
=: −2 {(I) − (II)} , (F1)
 16

with
X X
Mz := σjz = −i w2j w2j+1 . (F2)
j∈Z j∈Z

Plugging (F2) into (F1) and using the definition (E3), we obtain
1 X
(I) = hw2l1 (t)w2l1 +1 (t)w2j w2j+1 w2l3 (t)w2l3 +1 (t)w2k w2k+1 iβ=0
N
l1 ,j,l3 ,k
2 2
1 X X X l1 −¯
X X
l3 −¯
= Ks̄l11−
,1
l̄1
(t)K ¯
s̄1 ,2
l̄1
(t) Ks̄l33− l̄3 l̄3
,1 (t)Ks̄¯3 ,2 (t) ·
N
l1 ,j,l3 ,k l̄1 ,¯
l̄1 s̄1 ,s̄¯1 =1 l̄3 ,¯
l̄3 s̄3 ,s̄¯3 =1
D E
· w2l̄1 +s̄1 −1 w2¯l̄1 +s̄¯1 −1 w2j w2j+1 w2l̄3 +s̄3 −1 w2¯l̄3 +s̄¯3 −1 w2k w2k+1 . (F3)
β=0

Note that we changed the summation indices compared to those used in the main text (i, k to l1 , l3 and l to k).
Taking into account the translational invariance of the KI, we can fix one of the indices in the first sum and replace the
summation over that index with an overall multiplication by the number of particles in the system N , with N → ∞.
It is convenient to fix the last index to k = 0. The expression then simplifies to
2 2
XXXX X X ¯ ¯
(I) = Ks̄l11− l̄1 l1 −l̄1 l3 −l̄3 l3 −l̄3
,1 Ks̄¯1 ,2 Ks̄3 ,1 Ks̄¯3 ,2 ·
j l̄3 s̄1 ,s̄¯1 =1 s̄3 ,s̄¯3 =1
l̄1 l̄3 ,¯
l1 ,l3 l̄1 ,¯
D E
· w2l̄1 +s̄1 −1 w2¯l̄1 +s̄¯1 −1 w2j w2j+1 w2l̄3 +s̄3 −1 w2¯l̄3 +s̄¯3 −1 w0 w1 , (F4)
β=0

where the temporal dependence of K is omitted for compactness of notation. The expression is formal and will be
simplified in what is to follow. Furthermore, we also have
2 2
XXXX X X ¯ ¯
(II) = Ks̄l11− l̄1 l1 −l̄1 l3 −l̄3 l3 −l̄3
,1 Ks̄¯1 ,2 Ks̄3 ,1 Ks̄¯3 ,2 ·
j l1 ,l3 l̄1 ,¯ l̄3 s̄1 ,s̄¯1 =1 s̄3 ,s̄¯3 =1
l̄1 l̄3 ,¯
D E
· w2j w2j+1 w2l̄1 +s̄1 −1 w2¯l̄1 +s̄¯1 −1 w2l̄3 +s̄3 −1 w2¯l̄3 +s̄¯3 −1 w0 w1 . (F5)
β=0

The key to simplifying the expressions (I) and (II) are the anti-commutation relations

{wi , wj } = 2δij , (F6)

or the equation for the pair correlation function that follows from them:

hwi wj iβ=0 = δij . (F7)

We can use it together with the Wick’s theorem to compute the eight-fermion correlation functions appearing in
(F4) and (F5). We see that the infinite temperature expectation values in (F4) and (F5) are only non-zero if all of
the Majorana fermions in them appear in pairs. In particular, it is helpful to consider the cases of j 6= 0 and j = 0
separately.

a) j 6= 0

First, consider the terms in the correlator for (I) that have the form

h w2j w2j+1 w0 w1 iβ=0 , (F8)

where the empty slots ( ) have to be filled by w2j , w2j+1 , w0 , w1 , each appearing exactly once. This gives 24 possible
permutations. Then, for the corresponding correlator in (II) (with the first two pairs of terms interchanged), we want

hw2j w2j+1 w0 w1 iβ=0 = − h w2j w2j+1 w0 w1 iβ=0 , (F9)

so that the combination results in a non-zero term in (F1). Note that the only allowed combinations are those where
we have only one among w2j and w2j+1 in the first two slots. The remaining slot, among the first two slots, has to
 17

be filled by either w0 or w1 . This is also a direct consequence of the anti-commutation relations. We are left with 16
possible permutations, which we write out explicitly:
l1 −j l1 l3 −j l3
K1,1 K1,2 K2,1 K2,2 hw2j w0 w2j w2j+1 w2j+1 w1 w0 w1 iβ=0 ; sj = 1, s0 = 1, p1 = 1, p3 = 1,
| {z }
1
l1 −j l1 l3 l3 −j
K1,1 K1,2 K2,1 K2,2 hw2j w0 w2j w2j+1 w1 w2j+1 w0 w1 iβ=0 ; sj = 1, s0 = 1, p1 = 1, p3 = 2,
| {z }
−1
l1 l1 −j l3 −j l3
K1,1 K1,2 K2,1 K2,2 hw0 w2j w2j w2j+1 w2j+1 w1 w0 w1 iβ=0 ; sj = 1, s0 = 1, p1 = 2, p3 = 1,
| {z }
−1
l1 l1 −j l3 l3 −j
K1,1 K1,2 K2,1 K2,2 hw0 w2j w2j w2j+1 w1 w2j+1 w0 w1 iβ=0 ; sj = 1, s0 = 1, p1 = 2, p3 = 2,
| {z }
1

l1 −j l1 l3 −j l3
K1,1 K2,2 K2,1 K1,2 hw2j w1 w2j w2j+1 w2j+1 w0 w0 w1 iβ=0 ; sj = 1, s0 = 2, p1 = 1, p3 = 1,
| {z }
−1
l1 −j l1 l3 l3 −j
K1,1 K2,2 K1,1 K2,2 hw2j w1 w2j w2j+1 w0 w2j+1 w0 w1 iβ=0 ; sj = 1, s0 = 2, p1 = 1, p3 = 2,
| {z }
1
l1 l1 −j l3 −j l3
K2,1 K1,2 K2,1 K1,2 hw1 w2j w2j w2j+1 w2j+1 w0 w0 w1 iβ=0 ; sj = 1, s0 = 2, p1 = 2, p3 = 1,
| {z }
1
l1 l1 −j l3 l3 −j
K2,1 K1,2 K1,1 K2,2 hw1 w2j w2j w2j+1 w0 w2j+1 w0 w1 iβ=0 ; sj = 1, s0 = 2, p1 = 2, p3 = 2,
| {z }
−1

l1 −j l1 l3 −j l3
K2,1 K1,2 K1,1 K2,2 hw2j+1 w0 w2j w2j+1 w2j w1 w0 w1 iβ=0 ; sj = 2, s0 = 1, p1 = 1, p3 = 1,
| {z }
−1
l1 −j l1 l3 l3 −j
K2,1 K1,2 K2,1 K1,2 hw2j+1 w0 w2j w2j+1 w1 w2j w0 w1 iβ=0 ; sj = 2, s0 = 1, p1 = 1, p3 = 2,
| {z }
1
l1 l1 −j l3 −j l3
K1,1 K2,2 K1,1 K2,2 hw0 w2j+1 w2j w2j+1 w2j w1 w0 w1 iβ=0 ; sj = 2, s0 = 1, p1 = 2, p3 = 1,
| {z }
1
l1 l1 −j l3 l3 −j
K1,1 K2,2 K2,1 K1,2 hw0 w2j+1 w2j w2j+1 w1 w2j w0 w1 iβ=0 ; sj = 2, s0 = 1, p1 = 2, p3 = 2,
| {z }
−1

l1 −j l1 l3 −j l3
K2,1 K2,2 K1,1 K1,2 hw2j+1 w1 w2j w2j+1 w2j w0 w0 w1 iβ=0 ; sj = 2, s0 = 2, p1 = 1, p3 = 1,
| {z }
1
l1 −j l1 l3 l3 −j
K2,1 K2,2 K1,1 K1,2 hw2j+1 w1 w2j w2j+1 w0 w2j w0 w1 iβ=0 ; sj = 2, s0 = 2, p1 = 1, p3 = 2,
| {z }
−1
l1 l1 −j l3 −j l3
K2,1 K2,2 K1,1 K1,2 hw1 w2j+1 w2j w2j+1 w2j w0 w0 w1 iβ=0 ; sj = 2, s0 = 2, p1 = 2, p3 = 1,
| {z }
−1
l1 l1 −j l3 l3 −j
K2,1 K2,2 K1,1 K1,2 hw1 w2j+1 w2j w2j+1 w0 w2j w0 w1 iβ=0 ; sj = 2, s0 = 2, p1 = 2, p3 = 2. (F10)
| {z }
1

Now, we introduce the following notation:

     
l1 − j l3 − j s
R1 := , R3 := , S := j , (F11)
l1 l3 s0
 18

the spinor “inversion”:

1̃ := 2, 2̃ := 1, (F12)

and the following notation for the vector components

 
v(1)
v≡ . (F13)
v(2)

This allows us to summarise the above 16 lines in a compact formula:

2 2
R (p ) R (p˜ ) R (p3 ) R (p˜3 )
XX X X sj +s0 +p1 +p3
1 1
(I)j6=0 = KS(p 1 ),1
KS(1p˜1 ),2
1 3
KS̃(p KS̃(3p˜ ),2 (−1) , (F14)
3 ),1 3
j6=0 l1 ,l3 sj ,s0 =1 p1 ,p3 =1

where the summation runs over the 3 spatial indices: j, l1 , l3 and 4 permutations: sj , s0 , p1 , p3 . Here, sj denotes the
”spin” of the j-type fermion in the first pair of slots (w2j+sj −1 ) and s0 the ”spin” of the 0-type fermion in the first two
slots (ws0 −1 ). Furthermore, p1 denotes the permutation of fermions in the first pair of slots and p3 the permutation
of fermions in the third pair of slots. Note also that

(II)j6=0 = − (I)j6=0 . (F15)

b) j = 0

In this case, we have to fill in the following correlator:

h w0 w1 w0 w1 iβ=0 . (F16)

Since the present fermions are already contracted, we can fill the empty slots with arbitrary two pairs of fermions
w2j̄+sj −1 , w2j̄+sj −1 and w2l+s0 −1 ,w2l+s0 −1 . Again, we want to have

hw0 w1 w0 w1 iβ=0 = − h w0 w1 w0 w1 iβ=0 , (F17)

so that the terms in (F1) do not end up cancelling out. It is easy to check that the only way to achieve this is to set
either j̄ or l to zero, with the remaining index being non-zero. We choose l = 0 and j̄ 6= 0. As before, we again have
16 possible permutations:
l1 −j̄ l1 l3 −j̄ l3
K1,1 K1,2 K1,1 K1,2 w2j w0 w0 w1 w2j̄ w0 w0 w1 β=0
; sj̄ = 1, s0 = 1, p1 = 1, p3 = 1,
| {z }
−1
l1 −j̄ l1 l3 l3 −j̄
K1,1 K1,2 K1,1 K1,2 w2j̄ w0 w0 w1 w0 w2j̄ w0 w1 β=0
; sj̄ = 1, s0 = 1, p1 = 1, p3 = 2,
| {z }
1
l1 l1 −j̄ l3 −j̄ l3
K1,1 K1,2 K1,1 K1,2 w0 w2j̄ w0 w1 w2j̄ w0 w0 w1 β=0
; sj̄ = 1, s0 = 1, p1 = 2, p3 = 1,
| {z }
1
l1 l1 −j̄ l3 l3 −j̄
K1,1 K1,2 K1,1 K1,2 w0 w2j̄ w0 w1 w0 w2j̄ w0 w1 β=0
; sj̄ = 1, s0 = 1, p1 = 2, p3 = 2,
| {z }
−1

l1 −j̄ l1 l3 −j̄ l3
K1,1 K2,2 K1,1 K2,2 w2j̄ w1 w0 w1 w2j̄ w1 w0 w1 β=0
; sj̄ = 1, s0 = 2, p1 = 1, p3 = 1,
| {z }
−1
l1 −j̄ l1 l3 l3 −j̄
K1,1 K2,2 K2,1 K1,2 w2j̄ w1 w0 w1 w1 w2j̄ w0 w1 β=0
; sj̄ = 1, s0 = 2, p1 = 1, p3 = 2,
| {z }
1
l1 l1 −j̄ l3 −j̄ l3
K2,1 K1,2 K1,1 K2,2 w1 w2j̄ w0 w1 w2j̄ w1 w0 w1 β=0
; sj̄ = 1, s0 = 2, p1 = 2, p3 = 1,
| {z }
1
l1 l1 −j̄ l3 l3 −j̄
K2,1 K1,2 K2,1 K1,2 w1 w2j̄ w0 w1 w1 w2j̄ w0 w1 β=0
; sj̄ = 1, s0 = 2, p1 = 2, p3 = 2,
| {z }
−1
 19

l1 −j̄ l1 l3 −j̄ l3
K2,1 K1,2 K2,1 K1,2 w2j̄+1 w0 w0 w1 w2j̄+1 w0 w0 w1 β=0
; sj̄ = 2, s0 = 1, p1 = 1, p3 = 1,
| {z }
−1
l1 −j̄ l1 l3 l3 −j̄
K2,1 K1,2 K1,1 K2,2 w2j̄+1 w0 w0 w1 w0 w2j̄+1 w0 w1 β=0
; sj̄ = 2, s0 = 1, p1 = 1, p3 = 2,
| {z }
1
l1 l1 −j̄ l3 −j̄ l3
K1,1 K2,2 K2,1 K1,2 w0 w2j̄+1 w0 w1 w2j̄+1 w0 w0 w1 β=0
; sj̄ = 2, s0 = 1, p1 = 2, p3 = 1,
| {z }
1
l1 l1 −j̄ l3 l3 −j̄
K1,1 K2,2 K1,1 K2,2 w0 w2j̄+1 w0 w1 w0 w2j̄+1 w0 w1 β=0
; sj̄ = 2, s0 = 1, p1 = 2, p3 = 2,
| {z }
−1

l1 −j̄ l1 l3 −j̄ l3
K2,1 K2,2 K2,1 K2,2 w2j̄+1 w1 w0 w1 w2j̄+1 w1 w0 w1 β=0
; sj̄ = 2, s0 = 2, p1 = 1, p3 = 1,
| {z }
−1
l1 −j̄ l1 l3 l3 −j̄
K2,1 K2,2 K2,1 K2,2 w2j̄+1 w1 w0 w1 w1 w2j̄+1 w0 w1 β=0
; sj̄ = 2, s0 = 2, p1 = 1, p3 = 2,
| {z }
1
l1 l1 −j̄ l3 −j̄ l3
K2,1 K2,2 K2,1 K2,2 w1 w2j̄+1 w0 w1 w2j̄+1 w1 w0 w1 β=0
; sj̄ = 2, s0 = 2, p1 = 2, p3 = 1,
| {z }
1
l1 l1 −j̄ l3 l3 −j̄
K2,1 K2,2 K2,1 K2,2 w1 w2j̄+1 w0 w1 w1 w2j̄+1 w0 w1 β=0
; sj̄ = 2, s0 = 2, p1 = 2, p3 = 2, (F18)
| {z }
−1

which we can express as

2 2
R (p ) R (p˜ ) R (p ) R (p˜ )
XX X X p1 +p3
1 1
(I)j=0 = KS(p 1 ),1
KS(1p˜1 ),2
1 3
KS(p 3
3 ),1
KS(3p˜3 ),2
3
(−1) (−1) . (F19)
j̄6=0 l1 ,l3 sj̄ ,s0 =1 p1 ,p3 =1
   
li − j̄ s
where we used Ri := and S := j̄ .
li s0
Renaming the dummy summation index in (F19) from j̄ to j, summing (F19) and (F14) and plugging it together
with (I) = − (II) into the expression (F1) for cz (t), we finally recover Eq. (19) from the main text:
2 2
R (p ) R (p˜ )
XX X X p1 +p3 1 1
cz (t) = −4 (−1) KS(p 1 ),1
(t)KS(1p˜1 ),2
1
(t) ·
j6=0 l1 ,l3 sj ,s0 =1 p1 ,p3 =1
h i
s +s R3 (p3 ) R3 (p˜3 ) R3 (p3 ) R3 (p˜3 )
· (−1) j 0 KS̃(p ),1
(t)K S̃(p˜ ),2
(t) − K S(p 3 ),1 (t)KS( p˜3 ),2 (t) . (F20)
3 3

Eq. (F20) is a general expression for the dOTOC of transverse magnetisation and holds for any free fermion model.

Appendix G: Application of the formula for the dOTOC of Mz

The equation (F20) can be used in two ways: for exact numerical computation at intermediate times and for the
analytical computation of the long-time asymptotics.

1. Intermediate times

For intermediate times, t < 50, we proceed by first computing the power U t of KI Floquet propagator (C9) for
given numerical values of the parameters J and h:
t
X
U (J, h, θ) =
t
Un (J, h)einθ , (G1)
n=−t
 20

where Un are 2 × 2 matrices whose elements depend only on J and h. It is then easy to see from (E12) that

Z π
1
K l (t) = dθe−iθl U t (θ) = Ul , (G2)
2π −π

thereby enabling a direct computation of the real space propagator for given numerical values of J and h.
From the above calculation, we also learn that K l (t) 6= 0 only for |l| ≤ t. This is a direct observation of the fact
that the information in KI spreads in a sharp causal-cone with the speed of propagation equal to 1. Hereon, it follows
that the sums over j, l1 and l3 in (F20) do not need to be taken over the entire Z but only over the finite intervals
j ∈ [−2t, 2t] − {0}, l1 , l3 ∈ [max (−t, j − t) , min (t, j + t)]. The number of terms in this sum is proportional to t3 so
the summation can be efficiently carried out for intermediate t. The result is numerically exact.

2. Long-time asymptotics

To find the long-time asymptotic behavior of cz (t), we express (F20) using (E12):

2 2 4 Z π
 Z π Z π Z π
XX X X 1
p1 +p3
cz (t) = −4 (−1) dθ dθ1 dθ2 dθ3
2π −π −π −π −π
j6=0 l1 ,l3 sj ,s0 =1 p1 ,p3 =1

·e−iθR1 (p1 ) U t (θ) S(p ),1 e−iθ1 R1 (p˜1 ) U t (θ1 ) S(p˜ ),2 ·
   
1 1
n
sj +s0 −iθ2 R3 (p3 ) 
U (θ2 ) S̃(p3 ),1 e
t −iθ3 R3 (p˜3 )
U (θ3 ) S̃(p˜3 ),2 −
  t 
· (−1) e
o
− e−iθ2 R3 (p3 ) U t (θ2 ) S(p3 ),1 e−iθ3 R3 (p˜3 ) U t (θ3 ) S(p˜3 ),2 .
   
(G3)

Introducing the following notation

   
θ θ2
Θ1 = , Θ3 = , (G4)
θ1 θ3

we then have

2 2  4 Z π Z π Z π Z π
XX X X p1 +p3 1
cz (t) = −4 (−1) dθ dθ1 dθ2 dθ3
2π −π −π −π −π
j l1 ,l3 sj ,s0 =1 p1 ,p3 =1

·e−i(θ+θ1 )l1 e−i(θ2 +θ3 )l3 ei(Θ1 (p1 )+Θ3 (p3 ))j U t (θ) S(p1 ),1 U t (θ1 ) S(p˜1 ),2 ·
   
n o
s +s 
· (−1) j 0 U t (θ2 ) S̃(p3 ),1 U t (θ3 ) S̃(p˜3 ),2 − U t (θ2 ) S(p3 ),1 U t (θ3 ) S(p˜3 ),2 +
      

2 2  4 Z π Z π Z π Z π
X X X p1 +p3 1
+4 (−1) · dθ dθ1 dθ2 dθ3
2π −π −π −π −π
l1 ,l3 sj ,s0 =1 p1 ,p3 =1

·e−i(θ+θ1 )l1 e−i(θ2 +θ3 )l3 U t (θ) S(p1 ),1 U t (θ1 ) S(p˜1 ),2 ·
   
n o
s +s 
· (−1) j 0 U t (θ2 ) S̃(p ),1 U t (θ3 ) S̃(p˜ ),2 − U t (θ2 ) S(p ),1 U t (θ3 ) S(p˜ ),2 .
      
3 3 3 3
(G5)
 21

Here, we took the j sum over the entire Z in the first term and then subtracted the j = P
0 case in the second term.
Performing (formally) the j, l1 and l3 summations and taking into account n einx = 2π k δ (x − k2π), we get
P

2
X 2
X p1 +p3
cz (t) = −4 (−1) ·
sj ,s0 =1 p1 ,p3 =1
" Z π Z π Z π Z π
1
· dθ dθ1 dθ2 dθ3
2π −π −π −π −π

·δ(θ + θ1 )δ(θ2 + θ3 )δ (Θ1 (p1 ) + Θ3 (p3 )) U t (θ) S(p ),1 U t (θ1 ) S(p˜ ),2 ·
   
1 1
n o
sj +s0 
U (θ2 ) S̃(p ),1 U (θ3 ) S̃(p˜ ),2 − U (θ2 ) S(p ),1 U (θ3 ) S(p˜ ),2 −
t
  t   t   t 
· (−1)
3 3 3 3
 2 Z π Z π Z π Z π
1
− dθ dθ1 dθ2 dθ3
2π −π −π −π −π
·δ(θ + θ1 )δ(θ2 + θ3 ) U (θ) S(p1 ),1 U (θ1 ) S(p˜1 ),2 ·
 t   t 
#
n o
sj +s0 
U (θ2 ) S̃(p ),1 U (θ3 ) S̃(p˜ ),2 − U (θ2 ) S(p ),1 U (θ3 ) S(p˜ ),2 .
t
  t   t   t 
· (−1) (G6)
3 3 3 3

Finally, integrating over θ1 , θ2 and θ3 in the first term and θ1 and θ3 in the second, we get:
2
X 2
X p1 +p3
cz (t) = −4 (−1)
sj ,s0 =1 p1 ,p3 =1
"Z π
1
dθ U t (θ) S(p ),1 U t (−θ) S(p˜ ),2 ·
   
·
2π −π 1 1

 h  i h  i
s +s p1+p3 p1+p3
· (−1) j 0 U t − (−1) θ U t (−1) θ −
S̃(p3 ),1 S̃(p˜3 ),2
h  i h  i 
p1+p3 p1+p3
− U t − (−1) θ U t (−1) θ −
S(p3 ),1 S(p˜3 ),2
 2 Z π
1
dθ U t (θ) S(p U t (−θ) S(p˜ ),2 ·
   
− 1 ),1
2π −π
1
Z π
n
s +s 
dθ2 (−1) j 0 U t (θ2 ) S̃(p3 ),1 U t (−θ2 ) S̃(p˜3 ),2 −
  
·
−π
#
o
− U (θ2 ) S(p3 ),1 U (−θ2 ) S(p˜3 ),2
 t   t 
(G7)
Z π Z π Z π
=: dθ I1 (t, θ) + dθ dθ2 I2 (t, θ, θ2 ) . (G8)
−π −π −π
It can be shown that the integrand I2 vanishes:
I2 (t, θ, θ2 ) = 0. (G9)
The remaining integration over θ in I1 is in general difficult to perform but we can find the asymptotic behavior for
large t.
To complete this task we can take advantage of the fact that U is a unitary matrix. Using the form (D1), the
powers of U are simply:
 itκ(θ) 
e
U (θ) = V (θ)
t †
V (θ). (G10)
eitλ(θ)
We can then compute the integral (G8) in the large t regime by using the stationary phase approximation [64]:
Z b s
X 2π n h π 00
io
dθφ(θ)eitψ(θ) ∼ φ(ξj ) exp i tψ(ξ j ) + sign (ψ (ξ j )) , (G11)
a j
t |ψ 00 (ξj )| 4

where ξj denotes (all of) the local extrema of ψ(θ), i.e. ψ 0 (ξj ) = 0, on the interval [a, b].
 22

a. The case of the kicked quantum Ising model

The considerations so far have been general and can be applied to any quadratic fermion model. For the KI model,
we can use the expressions for eigenvalues and eigenvectors from Section D.
Let us introduce the following notation:
{V } (θ) := {v11 (θ), v12 (θ), v21 (θ), v22 (θ),
v11 (θ), v12 (θ), v21 (θ), v22 (θ),
v11 (−θ), v12 (−θ), v21 (−θ), v22 (−θ),
v11 (−θ), v12 (−θ), v21 (−θ), v22 (−θ)} . (G12)
Plugging the results from Section (D) into (G8), we see that integrand I1 can be written the following form:
I1 (t, θ) = P0 [{V } (θ)] +
+P−2 [{V } (θ)] e−2itκ(θ) + P2 [{V } (θ)] e2itκ(θ) +
+P−4 [{V } (θ)] e−4itκ(θ) + P4 [{V } (θ)] e4itκ(θ) . (G13)
Here, all P ’s are polynomials in their arguments. The indices denote the power of the term eitκ(θ) multiplying a
particular polynomial. All eigenvactor components
Rπ {V } and the eigenvalue κ also depend on parameters J and h.
The large t behavior of the integrals −π dθ of all the terms except for P0 can be obtained using stationary phase
approximation (G11). Plugging in the elements of the eigenvectors at stationary points of κ(θ), that is (D11), we
see that P−2 , P2 , P−4 , P4 vanish at these points. This means that in the large t regime, cz (t) is constant for the
transverse field KI.
The only remaining integral
Z π
lim cz (t) = dθ P0 [{V } (J, h, θ)] (G14)
t→∞ −π

can be evaluated numerically to get the asymptotic (constant) value of the cz (t) for any given J and h (See the inset
of Figure 1 of the main text).