
3
ab
Photoassociation
0.2 0.22 0.24 0.26 0.28 0.3
0
0.01
0.02
0.03
0.04
pDoublon
0.2 0.21
0
0.01
0.02
0.03
0.04
pDoublon
0.27 0.28
Flux φ/2πFlux φ/2π
φcφc
Normal FQH Normal FQH
Separation
FIG. 3. Suppression of two-body interactions. We find a
reduction of the doublon fraction in the FQH state, revealing
the screening of interactions in the many-body wave function.
a, Doublon fraction measurement by photoassociation, con-
verting the doubly occupied lattice site into an empty one. b,
Doublon fraction measurement by separating the particles on
two different lattice sites prior to fluorescence imaging. Solid
lines show exact calculations for the ground state, dashed
lines take into account the finite ground-state overlap [34].
All errorbars denote the s.e.m.
tivity.
The system is governed by the interacting Harper-
Hofstadter model (Fig. 1c), which describes the motion
of particles on a square lattice in the presence of a mag-
netic field. In our setup the effective magnetic field is
realized by Floquet engineering complex tunneling ampli-
tudes with Raman-assisted tunneling processes [34, 36].
Within the Floquet-engineered Hamiltonian, we achieve
independent control of the flux φ/2πper unit cell, the
tunneling rates Kand Jalong xand y, respectively, as
well as the gradients ∆xand ∆y. The on-site interac-
tion Uremains constant and large compared to all other
energy scales.
The state preparation begins from an initial state of
two localized atoms in the absence of any tunneling.
First, we increase the tunneling Jalong yin the pres-
ence of a gradient ∆y, while tunneling along xremains
inhibited (Fig. 2a). Since Jremains approximately con-
stant for the remainder of the protocol [34], it sets the
tunneling time τ=~/J = 4.3(1) ms and the interac-
tion strength U= 6.7(1) J. Subsequently ∆yis adiabat-
ically removed and we obtain a one-dimensional system
in its ground state. A similar procedure is then per-
formed along x: tunneling Kis increased at constant
flux φ/2π= 0.27 and gradient ∆x, then the gradient is
adiabatically removed. Up to this point we keep the tun-
neling ratio at K/J = 1.19(3). In a final step we bring
the tunneling amplitudes to K=Jto reach the tar-
get state. At each step of the evolution we measure the
system’s density distribution and find agreement with an
exact numerical prediction at the respective Hamiltonian
parameters (Fig. 2b).
Our preparation scheme at asymmetric tunneling K >
J(Fig. 2c,d) can be understood in a coupled-wire picture
[34, 42], where the parabolic dispersions of weakly cou-
pled rows are offset in momentum by multiples of φ/a.
While the excited state energies in each parabola are af-
fected by the tilt ∆xdue to their directional momentum,
the ground state energies remain independent. As the
tunneling ratio approaches K=J, anti-correlated densi-
ties on neighbouring sites, reminiscent of a charge density
wave state, are converted into Laughlin orbitals.
The success probability of the state preparation is
given by the fidelity F=hψ0|ˆρFinal|ψ0i, which measures
the overlap between the density operator ˆρFinal describing
the state after the preparation protocol, and the ground
state |ψ0iat the final Hamiltonian parameters. Since |ψ0i
is a delocalized and entangled state, measuring Fdirectly
with local observables is not possible. Instead, we map
|ψ0iback to the initial state by following the prepara-
tion with an identical, but reversed protocol. Assuming
that the evolution during both ways is independent, the
final ground state population is given by F2, which can
be directly measured because it equals the probability to
measure the initial density distribution. We find a dom-
inant ground state population throughout the evolution,
and a fidelity of F= 43(6)% to prepare the final state
(Fig. 2e).
The flux φ/2πat which the lowest bosonic Laughlin
state is expected to stabilize corresponds to a filling fac-
tor of ν=N/Nφ= 1/2, such that the system contains
twice the number of magnetic flux quanta Nφthan the
number of charge carriers N. In systems with sharp edges
this condition is only approximate and a systematic un-
derstanding of the finite size effects on the filling factor
is still lacking [28]. In order to map out the transition
between the normal state and the FQH state we use the
adiabaticity of the preparation scheme as a spectroscopic
signature for the energy gap (Fig. 2g). The fidelity Fis
limited by the smallest energy gap during the preparation
protocol, and therefore decreases when the energy of the
ground and excited states approach each other. Repeat-
ing the protocol for different flux values shows a break-
down of the adiabaticity at φ/2π≈0.25, indicating the
location of the transition point. The observed transition
point is in agreement with exact numerical calculations of
the gap diagram (Fig. 2f,g). In contrast, when repeating
the measurement at K > J no breakdown of adiabatic-
ity is visible, indicating that the many-body gap remains
open until we reach homogeneous tunneling K=J.
A hallmark of Laughlin-type FQH states is the screen-
ing of on-site interactions due to the pairwise vortex
motion. In our two-particle system, the interaction en-
ergy simplifies to Eint =hψ0|PiUˆni(ˆni−1)/2|ψ0i=
U×pDoublon, where ˆniis the number operator on site i
and pDoublon is the global probability to observe the two
particles on the same lattice site. We measure the dou-
blon probability in two different ways. In a first set of
measurements we perform photo-association of the dou-
blons into molecular states, whose excess energy ejects