Pairing by disorder of dopants in a magnetic spin ladder

Abstract

I demonstrate a pairing mechanism of dopants in a magnetic lattice, emerging from the underlying high-temperature disorder of the spins. The effect is demonstrated in a mixed-dimensional model, where dopants travel along a two-leg ladder, while the spins feature Ising interactions along the rungs and the legs of the ladder. The dynamics following the sudden immersion of two nearest neighbor dopants in an infinite temperature spin lattice shows that thermal spin disorder frustrates the relative motion of the dopants and enforces them to co-propagate. The predictions are shown to be realistically testable in quantum simulation experiments.

Understanding pairing of dopants in magnetic lattices is an immense challenge in modern quantum many-body theory. The main goal is to gain a deeper understanding of strongly correlated systems, arising e.g. in the notoriously difficult high-temperature superconductors Schrieffer and Brooks (2007); Trugman (1988); Dagotto (1994). While solid state experiments continue to pursue such inquiries after decades of research O’Mahony et al. (2022), a highly interesting and promising pathway has appeared in quantum simulation experiments with ultracold atoms in optical lattices Gross and Bloch (2017); Ji et al. (2021); Wang et al. (2021); Prichard et al. (2023); Koepsell et al. (2019, 2021); Lebrat et al. (2024); Hirthe et al. (2023). Single-site resolution capabilities Bakr et al. (2009); Sherson et al. (2010) makes it possible to measure low-temperature spatial correlations Koepsell et al. (2019, 2021); Prichard et al. (2023); Lebrat et al. (2024), as well as track them over time by repeated realizations of the same initial state. In particular, single hole dynamics consistent with magnetic polaron formation Nielsen et al. (2022) in quantum antiferromagnets has been measured Ji et al. (2021). Moreover, in such experiments one can tinker with the simulated models, otherwise fixed in the solid state. This has enabled the observation of dopant pairing in a mixed-dimensional magnetic lattice Hirthe et al. (2023). This has also been seen to be a plausible explanation Schlömer et al. (2023) for high-temperature superconductivity in a specific nickelate compound Sun et al. (2023). More broadly speaking, the appearance of pairing is conventionally tied to the distortion of some underlying order. In the aforementioned example, it is the distortion of the underlying AFM order that gives rise to the induced attraction Bohrdt et al. (2022); Nielsen (2023), which is then further enhanced by restricting motion to one dimension. In conventional superconductors, electrons distort the underlying crystal lattice, inducing an attraction between them Fröhlich (1950); Bardeen et al. (1957); Bruus and Flensberg (2016). In materials supporting high-temperature superconductivity, there may be an important counterexample to this generic scenario, in which preformed Cooper pairs Geshkenbein et al. (1997); Randeria et al. (1992) appearing above the superconducting phase transition might explain Li et al. (2010); Seo et al. (2019) the pseudogap phase Stanescu and Phillips (2003).

Figure 1: Pairing dynamics. (a) Setup: holes move in a magnetic (red: $\!\!\uparrow$, blue: $\!\!\downarrow$) two-leg ladder with Ising interactions $J_{\perp},J_{\parallel}$ via nearest-neighbor hop** $t$ along the ladder. (b) Hole-hole correlator $C_{\rm hh}(x_{1},x_{2};\tau)=\braket{\hat{n}_{h}(x_{1})\hat{n}_{h}(x_{2})}_{\tau}$ for $J_{\parallel}=0$ at indicated times normalized to its maximum after the holes are initialized on the same rung, $(l,x_{1})=(1,0),(l,x_{2})=(2,0)$. (c) Dynamics of the root-mean-square dynamics of the center-of-mass $X=(x_{1}+x_{2})/2$ and relative $\Delta=(x_{1}-x_{2})/2$ coordinates, (d) correlation coefficient ${\rm corr}(x_{1},x_{2})=\braket{x_{1}x_{2}}_{\tau}/[\sigma(x_{1})\sigma(x_{2})]$, and (e) relative correlation function for $\Delta=0$. Steady state values are shown in short dashed lines. In all: $J_{\perp}=5t$.

In this Letter, we demonstrate a pairing mechanism that does not rely on being in nor in the vicinity of a phase transition to an ordered phase. In particular, the discovered mechanism remains at infinite temperatures and is a direct consequence of the thermally disordered spin lattice. I analyze the effect in a simplistic setup of a mixed-dimensional model [Fig. 1(a)], where two holes are placed in two legs of a spin ladder. The spin degree of freedom is assumed to be at infinite temperature and completely disordered. Conventional wisdom, therefore, suggests that the holes should not be able to pair. Nevertheless, when the holes are initialized on the same rung in the infinite temperature spin lattice, the strong thermal disorder of the system frustrates the relative motion of the dopants. Indeed, tracking the hole-hole correlator in time [Fig. 1(b)], shows a significant correlation along the diagonal $x_{1}=x_{2}$, where the holes are on the same rung of the ladder. One may equivalently observe that the inter-hole distance is always smaller than the traversed distance of their center-of-mass [Fig. 1(c)]. This results in a strictly positive correlation between the holes’ position, signaling an effective attraction between them [Fig. 1(d)]. Finally, we observe that the probability of finding the holes on the same rung is easily increased by a factor of $4$ as compared to what one expects from their marginal distributions, described by the relative hole-hole correlator in Fig. 1(e). This phenomenon is closely tied to the recent discovery of thermally induced localization in such systems Nielsen (2024a, b). In particular, the holes experience an effective induced interaction that has a disordered character, encouraging the holes to co-propagate. I emphasize that the discovered effect would appear in a wide range of systems, though the robustness of the pairing in more generic setups is presently unclear.

Model.-

I consider spins in a two-leg ladder with Ising interactions and hop** along the ladder,

$${\hat{H}}=\sum_{{\bf i},{\bf j}}\!\frac{J_{{\bf i}-{\bf j}}}{2}\hat{S}^{(z)}_{\bf i}\hat{S}^{(z)}_{\bf j}-t\!\sum_{\braket{{\bf i},{\bf j}}_{\parallel},\sigma}\!\!\!\left[\tilde{c}^{\dagger}_{\sigma{\bf i}}\tilde{c}_{\sigma{\bf j}}+\tilde{c}^{\dagger}_{\sigma{\bf j}}\tilde{c}_{\sigma{\bf i}}\right],$$ (1)

i.e. a mixed-dimensional $t$-$J_{z}$ model Grusdt and Pollet (2020). I focus on the case of nearest-neighbor couplings along, $J_{\parallel}=J_{\pm{\bf e}_{x}}$, and across, $J_{\perp}=J_{\pm{\bf e}_{y}}$, the ladder, with $\hat{S}^{(z)}_{\bf i}$ the $z$-projection of the spin operator $\hat{\bf S}_{\bf i}$. The hop**, ${\hat{H}}_{t}$, is only allowed along the ladder and only to vacant sites, enforced by the constrained operators $\tilde{c}^{\dagger}_{\sigma{\bf i}}=\hat{c}^{\dagger}_{\sigma{\bf i}}[1-\hat{n}_{\bf i}]$, where $\hat{n}_{\bf i}=\sum_{\sigma}\hat{n}_{\sigma{\bf i}}=\sum_{\sigma}\hat{c}^{\dagger}_{\sigma{\bf i}}\hat{c}_{\sigma{\bf i}}$ is the local density operator. The effective spin degree of freedom generally describes an internal degree of freedom with the Ising-type coupling in ${\hat{H}}_{J}$, the first term in Eq. (1), and since there are no exchange processes of the dopants, both fermions and bosons can realize the model. To have an efficient description of the hole and spin degrees of freedom, we perform a Holstein-Primakoff transformation on top of the ferromagnetic ground state $\ket{{\rm FM}}=\ket{\cdots\uparrow\uparrow\cdots}$, with all spins in the $\ket{\uparrow}$ state. The spin Hamiltonian

$$\!\!\!\hat{H}_{J}\!=\!\sum_{{\bf i},{\bf j}}\!\frac{J_{{\bf i}-{\bf j}}}{2}\Big{[}\frac{1}{2}\!-\!\hat{s}^{\dagger}_{{\bf i}}\hat{s}_{{\bf i}}\Big{]}\Big{[}\frac{1}{2}\!-\!\hat{s}^{\dagger}_{{\bf j}}\hat{s}_{{\bf j}}\Big{]}\Big{[}1\!-\!\hat{h}_{{\bf i}}^{\dagger}\hat{h}_{{\bf i}}\Big{]}\Big{[}1\!-\!\hat{h}_{{\bf j}}^{\dagger}\hat{h}_{{\bf j}}\Big{]},\!\!$$ (2)

hereby, describes the interaction of bosonic spin flip excitations $\hat{s}^{\dagger}_{\bf i}$, corresponding to the creation of a spin-$\downarrow$ on site ${\bf i}$, and holes created by the operator $\hat{h}^{\dagger}_{\bf i}$, which inherit the statistics of the underlying particles, be it fermionic or bosonic Nielsen (2023). Equivalently, the hop** Hamiltonian

$${\hat{H}}_{t}=t\sum_{\braket{{\bf i},{\bf j}}_{\parallel}}\! \Big{[}\hat{h}^{\dagger}_{{\bf j}}F(\hat{h}_{{\bf i}},\hat{s}_{{\bf i}})F(\hat{h}_{{\bf j}},\hat{s}_{{\bf j}})\hat{h}_{{\bf i}}$$

$$+\hat{h}^{\dagger}_{{\bf j}}\hat{s}^{\dagger}_{\bf i}F(\hat{h}_{{\bf i}},\hat{s}_{{\bf i}})F(\hat{h}_{{\bf j}},\hat{s}_{{\bf j}})\hat{s}_{\bf j}\hat{h}_{{\bf i}}\Big{]}+{\rm H.c.}$$ (3)

describes two distinct ways in which holes may hop: (1) a hole hop** from site ${\bf i}$ to ${\bf j}$ in the absence of a spin flip on site ${\bf j}$, and (2) hop** in the presence of a spin flip, whereby the hole and spin flip swap places. The constraint function $F(\hat{h},\hat{s})=\sqrt{1-\hat{s}^{\dagger}\hat{s}-\hat{h}^{\dagger}\hat{h}}$ ensures at most one spin per site. The initial state at time $\tau=0$ of two holes on the same rung of the ladder in a spin background ${\boldsymbol{\sigma}}$ is then

$$\ket{\Psi_{\boldsymbol{\sigma}}(\tau=0)}=\prod_{l=1}^{2}\hat{h}^{\dagger}_{(l,0)}\prod_{j\in S_{{\boldsymbol{\sigma}}}^{l}}\hat{s}^{\dagger}_{(l,j)}\ket{{\rm FM}},$$ (4)

in which $S_{{\boldsymbol{\sigma}}}^{l}$ is the subset of spins in leg $l$, which are initially in the $\ket{\downarrow}$ state. Here, we write ${\bf i}=(l,j)$, with leg index $l=1,2$ and position $j$ along the leg. Since the spins can only exchange places with the holes along the ladder, the state at any later stage is

$$\ket{\Psi_{\boldsymbol{\sigma}}(\tau)}= \sum_{x_{1},x_{2}}\!a_{\boldsymbol{\sigma}}(x_{1},x_{2};\tau)\prod_{l=1}^{2}\hat{h}^{\dagger}_{(l,x_{l})}$$

$$\times\!\!\!\!\!\!\!\!\prod_{\begin{subarray}{c}j\in S_{{\boldsymbol{\sigma}}}^{l}\\ j\in[{\rm sgn}(x_{l}),x_{l}]\end{subarray}}\!\!\!\!\!\!\!\hat{s}^{\dagger}_{(l,j-{\rm sgn}(x_{l}))}\!\!\!\!\!\prod_{\begin{subarray}{c}j\in S_{{\boldsymbol{\sigma}}}^{l}\\ j\notin[{\rm sgn}(x_{l}),x_{l}]\end{subarray}}\!\!\!\!\!\hat{s}^{\dagger}_{(l,j)}\!\ket{{\rm FM}}.$$ (5)

Here, ${\rm sgn}(x)$ is the sign function. Despite the rather formal description, the physics is simple and clear: if the hole in leg $l$ moves $|x_{l}|$ times to the right, it surpasses $|x_{l}|$ spins that all move one step to the left. If the hole moves to the left, the surpassed spins move one step to the right. The corresponding probability amplitudes $a_{{\boldsymbol{\sigma}}}(x_{1},x_{2};\tau)$ of finding the holes at positions $x_{1},x_{2}$ at time $\tau$, follow the equations of motion

$$i\partial_{\tau}a_{\boldsymbol{\sigma}}(x_{1},x_{2};\tau) =V_{\boldsymbol{\sigma}}(x_{1},x_{2})a_{{\boldsymbol{\sigma}}}(x_{1},x_{2};\tau)$$

$$+t\!\!\!\!\!\!\!\sum_{\begin{subarray}{c}\delta_{1},\delta_{2}\\ |\delta_{1}|+|\delta_{2}|=1\end{subarray}}\!\!\!\!\!\!\!a_{{\boldsymbol{\sigma}}}(x_{1}+\delta_{1},x_{2}+\delta_{2};\tau),$$ (6)

derived from the Schrödinger equation, $i\partial_{\tau}\ket{\Psi_{\boldsymbol{\sigma}}(\tau)}=\hat{H}\ket{\Psi_{\boldsymbol{\sigma}}(\tau)}$, with the initial condition that holes are on the same rung: $a_{\boldsymbol{\sigma}}(x_{1}=0,x_{2}=0;\tau=0)=1$. The sum extends over configurations reached in a single hop. The induced interaction $V_{\boldsymbol{\sigma}}(x_{1},x_{2})$ emerges, because the motion of the holes alter the spin bonds in the ladder [Fig. 2(a)]. For a random spin background, the induced interaction is also random [Fig. 2(b)], characterized in the relative and center-of-mass coordinates $\Delta=(x_{1}-x_{2})/2$ and $X=(x_{1}+x_{2})/2$.

Figure 2: Pairing mechanism. (a) As the holes move (green shaded region), spin bonds in the initial configuration are broken across (dark blue) and within (light blue) the ladder, and new bonds are established (dark and light red, respectively), resulting in a random induced interaction between the holes. (b) Induced interaction for vanishing center-of-mass, $X=0$, as a function of the relative coordinate $\Delta$ for $50$ realizations of the spin background (grey and colored lines), performing a random walk with standard deviation $\sim|J_{\perp}|\sqrt{|\Delta|}$ (black lines). (c) Induced interaction at $\Delta=0$, as a function of $X$. Here, $J_{\parallel}=J_{\perp}$ is used.

I indeed show SM that if all spin backgrounds are equally likely – as in an infinite temperature spin state – $V_{\boldsymbol{\sigma}}(x_{1},x_{2})$ has zero mean and a variance ($x_{1},x_{2}\neq 0$)

$${\rm Var}[V_{{\boldsymbol{\sigma}}}(x_{1},x_{2})]=\frac{J_{\perp}^{2}}{4}\left[|\Delta|-\frac{1}{4}\right]+\frac{J_{\parallel}^{2}}{4}$$ (7)

linear in the relative distance $2|\Delta|$ between the holes, describing a classical random walk in the values of the potential [Fig. 2(b)]. I can focus on the evolution in specific spin backgrounds by resolving the dynamics of the density matrix in terms of these states. Starting from the initial density matrix

$$\hat{\rho}(\tau=0)=\!\!\!\sum_{l=1,2,\sigma_{(l,0)}}\!\!\!\hat{c}_{(l,0)\sigma{(l,0)}}\hat{\rho}_{J}\hat{c}^{\dagger}_{(l,0)\sigma{(l,0)}},$$ (8)

where $\hat{\rho}_{J}=e^{-\beta{\hat{H}}_{J}}/Z_{J}$ is the Gibbs state of the spins at inverse temperature $\beta=1/(k_{B}T)$ and $\sigma_{(l,0)}=\uparrow,\downarrow$ designates the spin configuration at the origin of each leg, $l$, the ensuing dynamics $\hat{\rho}(\tau)=e^{-i\hat{H}\tau}\hat{\rho}(\tau=0)e^{+i\hat{H}\tau}$ – provided that the system is assumed to be closed – is the Boltzmann-weighted sum Nielsen (2024a, b)

$$\hat{\rho}(\tau)=\sum_{{\boldsymbol{\sigma}}}\frac{e^{-\beta E_{J}({\boldsymbol{\sigma}})}}{Z_{J}}\ket{\Psi_{{\boldsymbol{\sigma}}}(\tau)}\bra{\Psi_{{\boldsymbol{\sigma}}}(\tau)},$$ (9)

where $E_{J}({\boldsymbol{\sigma}})$ is the magnetic energy of the spin realization ${\boldsymbol{\sigma}}$ before the spins $\sigma_{(1,0)},\sigma_{(2,0)}$ are removed. In particular, at infinite temperatures all pure state evolutions $\ket{\Psi_{{\boldsymbol{\sigma}}}(\tau)}$ are sampled with equal probability, $e^{-\beta E_{J}({\boldsymbol{\sigma}})}/Z_{J}=2^{-N}$, for system size $N$. In this high-temperature limit, the dynamics is exactly the same for antiferromagnetic ($J_{\perp},J_{\parallel}>0$) and ferromagnetic couplings ($J_{\perp},J_{\parallel}<0$). The calculation of the hole dynamics then follows the recipe: (1) generate a sample $S$ of $N_{S}=400$ random spin configurations ${\boldsymbol{\sigma}}$ and calculate $V_{{\boldsymbol{\sigma}}}(x_{1},x_{2})$, (2) solve the equations of motion in Eq. (6) using exact diagonalization for $N=161\times 2$, and (3) compute the hole-hole correlator

$$C_{\rm hh}(x_{1},x_{2};\tau)=\braket{\hat{n}_{h}(x_{1})\hat{n}_{h}(x_{2})}_{\tau}=\sum_{{\boldsymbol{\sigma}}\in S}\frac{|a_{{\boldsymbol{\sigma}}}(x_{1},x_{2};\tau)|^{2}}{N_{S}},$$ (10)

whereby it may be tracked in time [Fig. 1(b)]. This describes the probability of simultaneously finding holes at positions $(1,x_{1}),(2,x_{2})$. Here, $\hat{n}_{h}(x_{l})=\hat{h}^{\dagger}_{(l,x_{l})}\hat{h}_{(l,x_{l})}$ is the hole density operator. Figure 1(b) manifestly breaks the individual inversion symmetries $x_{1}\to-x_{1}$ or $x_{2}\to-x_{2}$ of uncorrelated holes, and shows a clear tendency for them to be on the same rung ($x_{1}=x_{2}\Leftrightarrow\Delta=0$). Their motion is further analyzed by the root-mean-square of the center-of-mass and relative coordinates [Fig. 1(c)],

$$A_{\rm rms}(\tau) =\left[\sum_{X,\Delta}A^{2}C_{\rm hh}(X+\Delta,X-\Delta;\tau)\right]^{1/2},$$ (11)

for $A=X,\Delta$. Since $X_{\rm rms}>\Delta_{\rm rms}$, the holes are closer together than they have propagated into the lattice. This is further characterized by the correlation coefficient

$${\rm corr}(x_{1},x_{2})=\frac{\braket{x_{1}x_{2}}_{\tau}}{\sigma(x_{1})\sigma(x_{2})}=\frac{X_{\rm rms}^{2}-\Delta_{\rm rms}^{2}}{X_{\rm rms}^{2}+\Delta_{\rm rms}^{2}},$$ (12)

which is always restricted to the interval $[-1,+1]$. Positive values of ${\rm corr}(x_{1},x_{2})$, as seen in Fig. 1(d), correspond to $X_{\rm rms}>\Delta_{\rm rms}$, and signify attraction. Likewise, negative values correspond to an effective repulsion, while ${\rm corr}(x_{1},x_{2})=0$ corresponds to no linear correlations. To further substantiate the findings, I analyze the relative hole-hole correlator

$$C_{\rm hh}^{\rm rel}(\Delta,\tau)=\frac{\sum_{X}C_{\rm hh}(X+\Delta,X-\Delta;\tau)}{\sum_{X}C_{\rm hh}^{\rm disc}(X+\Delta,X-\Delta;\tau)},$$ (13)

giving the probability of finding the holes at distance $2|\Delta|$ relative to what is expected from their marginals, i.e. the disconnected correlator $C^{\rm disc}_{\rm hh}(\Delta,\tau)=\sum_{X}\braket{\hat{n}_{h}(X+\Delta)}\braket{\hat{n}_{h}(X-\Delta)}$. This is shown for $\Delta=0$ in Fig. 1(e), demonstrating a significant increase in the probability of finding the holes on the same rung. On long timescales, the system settles into a steady state, by which $X_{\rm rms}\to l_{X},\Delta_{\rm rms}\to l_{\Delta}$. For $J_{\perp}\geq 2t$, there are negligible finite-size corrections for the chosen length of $N/2=161$, and I, therefore, confine the analysis to this regime. Figure 3(a), hereby, shows that $l_{X}>l_{\Delta}$ remains true over this entire range of parameters. The associated asymptote of the correlation coefficient [Fig. 3(b)] shows very strong attractive correlations for $J_{\perp}\gg t$ and $J_{\parallel}=0$.

Figure 3: Steady state. (a) The rms distances $l_{X}$ and $l_{\Delta}$ of the center-of-mass, $X$, and relative, $\Delta$, coordinates of the holes as a function of $J_{\perp}/t$ for indicated values of $J_{\parallel}$ in the steady state. (b) Associated correlation coefficient ${\rm corr}(x_{1},x_{2})=\braket{x_{1}x_{2}}/[\sigma(x_{1})\sigma(x_{2})]$. (c) Relative hole-hole correlator [Eq. (13)] for holes on the same rung ($\Delta=0$) vs $J_{\perp}/t$. (d) Relative hole-hole correlator vs $|\Delta|$ for indicated values of $J_{\perp}/t$ and $J_{\parallel}$.

The dynamical binding – a finite $l_{\Delta}$ – has the exact same origin as the thermal localization phenomenon discovered recently for the motion of single holes Nielsen (2024a, b), by which the separation of the holes is avoided, because the induced interaction $V_{\boldsymbol{\sigma}}(x_{1},x_{2})$ inevitably fluctuates to large values in the relative distance $2|\Delta|=|x_{1}-x_{2}|$, as seen in Fig. 2(b). Therefore, at the very least, they back-reflect once they meet their classical turning points, corresponding to an emergent kind of Anderson localization Anderson (1958) in the presence of strong disorder Lagendijk et al. (2009). The retained localization of the holes – a finite $l_{X}$ – is more subtle, because $V_{{\boldsymbol{\sigma}}}(x,x)$ is restricted to take values between $\pm|J_{\parallel}|$ [Fig. 2(c)], such that it vanishes for holes on the same rung in the extreme limit of $J_{\parallel}=0$. Even in this case, however, the holes localize [Fig. 3(b)]. The reason is that there are many ways in which the holes can move to sites $(1,x)$ and $(2,x)$. Along each path, the wave builds up a phase directly depending on the randomly fluctuating induced interaction they meet along the way. The ensuing destructive interference of these pathways corresponds to Anderson localization for weak disorder Lagendijk et al. (2009). Finally, the relative hole-hole correlator in Figs. 3(c) and 3(d) show that the probability of finding holes in close proximity can be several times higher than expected from their marginals. Overall, it is clear that at large $J_{\perp}/t$ the strength of the pairing sensitively depends on $J_{\parallel}$, essentially because the holes localize over very few lattice sites if $J_{\parallel}\sim J_{\perp}$. However, already at intermediate values of $J_{\perp}/t$, the pairing for $J_{\parallel}=J_{\perp}$ is comparable to the case of $J_{\parallel}=0$. I expect this to persist at lower $J_{\perp}/t$.

Rydberg-dressed atoms.-

The predicted results may be feasibly tested in ultracold quantum simulators based on Rydberg-dressed atoms in optical lattices Zeiher et al. (2016). Indeed, such schemes natively realizes spin-specific density-density interactions

$${\hat{H}}_{J}=\frac{1}{2}\sum_{{\bf i}\neq{\bf j}}J_{{\bf i}-{\bf j}}\hat{n}_{\uparrow{\bf i}}\hat{n}_{\uparrow{\bf j}}$$ (14)

of the internal atomic state $\ket{\uparrow}$, by optically dressing it with a higher-lying Rydberg state. Here, $J_{{\bf i}-{\bf j}}$ takes on a soft core shape Henkel et al. (2010); SM . Moreover, such systems were recently demonstrated to enter the itinerant regime Weckesser et al. (2024), by which the desired $t$-$J$ model may be realized. In particular, the spin model in Eq. (14) is obtained by using a Feshbach resonance to enhance the onsite interaction deep into the Mott-insulating phase. This ensures at most one spin per site and negligible spin superexchange. Precise control of the temperature in such systems is difficult, but the hole dynamics may still be investigated in the following manner Nielsen (2024a). First, initialize holes at $x_{1}=x_{2}=0$ by applying a strong repulsive light field to those sites Ji et al. (2021). Second, rotate all spins from the fully polarized spin sample in the non-interacting $\ket{\downarrow}$ state into the equal superposition state $\ket{\Psi_{0}}=\prod_{{\bf i}\neq{\bf 0}}\frac{1}{\sqrt{2}}\left[\hat{c}^{\dagger}_{{\bf i}\uparrow}+\hat{c}^{\dagger}_{{\bf i}\downarrow}\right]\ket{0}$. Third, turn off the focused light fields, freeing the holes to move along the ladder Hirthe et al. (2023). Although this at face value seems distinct from the infinite temperature scenario above, the hole-hole correlator takes on the exact same form

$$C_{\rm hh}(x_{1},x_{2};\tau) =2^{-N}\sum_{{\boldsymbol{\sigma}}}|a_{\boldsymbol{\sigma}}(x_{1},x_{2};\tau)|^{2},$$ (15)

such that the holes move as though the system is at infinite temperature. The polarized Ising form of the interaction is crucial for this exact equivalence, and one should note that spin correlators would presumably differ from their infinite temperature counterpart. The computation now follows the same recipe as before, and the resulting correlation coefficient dynamics [Eq. (12)] and relative correlation function [Eq. (13)] are given in Figs. 4(a) and 4(b), respectively. This is averaged over $N_{S}=1000$ realizations and given in two instances [inset of Fig. 4(a)]: (1) isotropic couplings akin to the $J_{\parallel}=J_{\perp}$ case, and (2) anisotropic couplings where the $30P_{3/2}$ orbital of the Rydberg state is rotated to lie along the rungs using an external magnetic field Zeiher et al. (2016), suppressing coupling in other directions similar to low values of $J_{\parallel}$ SM . Crucially, we observe that the appearing disorder-induced attraction between the holes can be probed for experimentally feasible system sizes and timescales. Indeed, stroboscopic dressing Weckesser et al. (2024) makes $20$ tunneling events feasible, while longer timescales are inhibited by the intrinsic lifetime of the Rydberg state Weckesser (2024). I note that the rung case gives a much more favorable signal, just as was the case for $J_{\parallel}=0$ vs $J_{\parallel}=J_{\perp}$.

Figure 4: Rydberg dressing. (a) Dynamics of the correlation coefficient for isotropic (orange) and strong rung (purple) spin couplings. (b) Associated relative correlator for holes on the same rung, $\Delta=0$. Dashed lines: steady state values. Nearest neighbor rung interaction: $J_{{\bf e}_{y}}=8t$. Insets: orientation of $30P_{3/2}$ Rydberg orbitals enforced by shown magnetic fields ${\bf B}$. System size: $N=21\times 2$.

Conclusions and outlook.-

I have demonstrated that thermal disorder of spins can frustrate the relative motion of dopants and lead to a dynamical pairing mechanism. The effect is closely related to the thermally induced localization phenomenon recently discovered for single holes Nielsen (2024a, b). By carefully analyzing its origin and behavior at infinite temperature, it is, furthermore, clear that the effect persists for all values of the spin coupling versus the hop**. For diminishing temperatures, ferromagnetic interaction is expected to lead to a crossover to delocalized and uncorrelated holes. For antiferromagnetic interactions, on the other hand, a crossover from fluctuation induced pairing to confinement of the holes emerges at low temperatures Nielsen (2023). Moreover, it demonstrates that the localization phenomenon Nielsen (2024a, b) is not sensitive to further do**. The results finally show that it is feasible to experimentally test my predictions using currently available techniques with ultracold atoms in optical lattices. One could also engineer similar models using Rydberg arrays Schauss et al. (2015); Labuhn et al. (2016); Bernien et al. (2017); Guardado-Sanchez et al. (2018); Lienhard et al. (2018) or polar molecules Gorshkov et al. (2011). In the future, the generality of the discovered effect may be tested by, e.g., tuning in trans-leg hop** or spin superexchange. The robustness of the phenomenon to such alterations is presently unclear. The nature of the effect is, however, very general – it comes down to a randomly fluctuating field emerging as the dopants separate, frustrating their relative motion.