Crow instability of vortex lines in dipolar superfluids

Abstract

In classical inviscid fluids, parallel vortices perturbed by Kelvin waves may exhibit the Crow instability, where the mutual interaction of the Kelvin modes renders them dynamically unstable. This results in the mutual approach and reconnection of the vortices, leading to a cascaded decay into smaller and smaller vortex loops. We study the Crow instability of quantum vortex lines in a superfluid enjoying the anisotropic, long-ranged dipole-dipole interaction through mean-field simulations of the dynamics of the superfluid order parameter. We observe that both the strength and direction of the dipole-dipole interaction play a crucial role in determining the Kelvin modes that are dynamically favoured. This is shown to be linked to effective enhancements or suppressions of the vortices’ curvature and can be explained by the effective dipole-dipole interaction between the vortex lines themselves. This paves the way to a deeper understanding of vortex reconnection dynamics, vortex loop cascading and turbulence in dipolar superfluids.

I Introduction

Classical inviscid fluids play host to a variety of instabilities arising from the interaction of waves and vortices, and understanding their properties is a central tenet of fluid dynamics. One such example is the celebrated Crow instability, where a pair of vortices with mutually antiparallel circulations is unstable against transverse symmetric perturbations that induce helical Kelvin waves upon the vortex tubes. These Kelvin waves grow in amplitude until the vortices reconnect, thereby forming a series of vortex loops. The subsequent cascade of reconnections to ever-smaller vortex loops eventuates in dissipation of the loops, as seen in the mitigation of wingtip vortices of aircraft by the interaction with the corresponding contrails [1]. The existence and properties of analogues of such hydrodynamic instabilities has proved to be a fruitful source of inspiration for studying quantum fluids, such as Bose-Einstein condensates, that exhibit superfluidity. Unlike classical fluids described theoretically by the Navier-Stokes equations, superfluids can flow without viscous dissipation and are characterised by a locally coherent order parameter [2]. The phase coherence of this order parameter ensures that the vorticity of a superfluid can only be nonzero along discrete, infinitesimally thin lines, each a topological defect, such that each vortex line boasts a quantised circulation and a total depletion of the superfluid density at its core [3]. As such, dynamical processes involving quantum vortices, such as reconnections and annihilations, [4, 5] cannot change their total net circulation as it is a topologically conserved quantity [3]. Whereas the ground state of an ensemble of quantum vortices – typically, a triangular vortex lattice – is well-understood [6, 7, 8, 9], the interwoven phenomena of instabilities, turbulence and disorder of incompressible (vortices) and compressible (phonon) excitations in superfluids have been of great interest for some time [10, 11, 12, 13, 14]. While certain superfluid instabilities such as the snake instability [15] have no classical counterpart it has been shown that superfluids, or mixtures thereof, can become turbulent through quantum analogues of classical hydrodynamic instabilities such as the Kelvin-Helmholtz [16, 17, 18], Rayleigh-Taylor [19, 20] and Richtmyer-Meshkov [21] instabilities. The Crow instability, too, manifests itself in the dynamics of pairs of quantum vortices of opposite circulations when subjected to either an environmental perturbation or if there are pre-existing Kelvin waves on the vortex lines [15, 22, 23, 24, 25]. An example of the latter scenario is depicted in Fig. 1, showing how the magnitude of the transverse excitations of the vortex lines grows in time till the vortices’ cores overlap at a point in space to the extent that they are able to reconnect into a pair of loops .

Figure 1: The superfluid Crow instability of an antiparallel quantum vortex pair initially separated along the $y$-axis. With $t_{\mathrm{r}}$ the time till the first reconnection, the vortices are depicted at $t=t_{\mathrm{r}}/3$, $t=2t_{\mathrm{r}}/3$ and $t=t_{\mathrm{c}}\equiv t_{\mathrm{r}}-0.5$.

The vast majority of investigations of hydrodynamic instabilities in superfluids have focussed on systems where the interactions between its constituents can be approximated as being effectively isotropic and short-ranged. Dipolar Bose-Einstein condensates (dBECs), comprised of certain lanthanide atoms with large, permanent magnetic dipole moments, represent a counterpoint to this paradigm. When its atomic dipole moments are uniformly polarized by an applied magnetic field, a dBEC exhibits a wealth of exotic phenomena that arise from a delicate interplay between the long-ranged and anisotropic dipole-dipole interaction (DDI) and the residual short-ranged, effectively isotropic van der Waals repulsion [26, 27, 28]. The foremost consequence of this competition is the tendency of a dBEC to exhibit anisotropy, whether it be its density profile under external confinement [29, 30], the speed of sound and superfluid critical velocity [31, 32], or the core of an embedded quantum vortex [33, 34, 35, 36]. Another consequence is a tendency towards stratification or short-ranged order, often associated with the presence of intermediate-wavelength roton excitations, resulting in the manifestation of novel features such as non-triangular vortex lattices with density striping [37, 38, 27, 39, 35], density ripples about the cores of quasi-two-dimensional point vortices [34], stratification during turbulence [40], pattern formation [41, 42], and the hitherto elusive supersolid phase [43, 44, 45]. Theoretical studies of quantum vortex dynamics in dBECs, which are now of greater relevance than ever after the experimental realization of dipolar quantum vortices [46, 47], have also demonstrated evidence for considerable deviations away from the nondipolar limit of properties such as the velocities and trajectories of vortex pairs [34, 48, 49, 36, 50] and the Kelvin wave spectra of single vortices [51, 52]. Together, these results would suggest that the onset of a hydrodynamic instability such as the Crow instability is sensitive to the magnitude and direction of the dipole moment polarization.

Thus, in this article we present the results of a systematic analysis of the Crow instability manifesting itself in a dBEC with a uniform background density and elucidate the ways in which the various properties of the instability depend on the parameters of the DDI. In Section II we provide an overview of the dipolar Gross-Pitaevskii equation (dGPE) as employed to model perturbed quantum vortices in a uniform, three-dimensional dBEC. We also discuss the numerical methods used for propagating the dGPE and identifying vortex structures in the solutions. In Sec. III we introduce a scenario where each of a pair of antiparallel vortices is subjected to a symmetric, random perturbation and provide a qualitative discussion of the results of one such simulation, thereby illustrating the influence that the DDI has upon the vortex profiles that emerge when the Crow instability is triggered. Section IV presents a spectral study of these vortex profiles, where tracking the populations of each Kelvin wave mode allows us to identify the modes that are excited most strongly by the Crow instability, and their corresponding rates of amplitude growth, as well as the line-averaged curvature of the vortices. These findings are summarised in Sec. V along with some implications and possible generalisations of this work.

II Formalism

In this article, we investigate the dynamics of perturbed pairs of quantum vortices in a dBEC through propagating the dipolar Gross-Pitaevskii equation (dGPE) governing the superfluid order parameter, $\psi(\mathbf{r},t)$. For a system composed of a single atomic species of mass $m$ and magnetic dipole moment $\mu_{\mathrm{d}}$, polarised uniformly by a magnetic field parallel to $d$, the dGPE is given by [26, 27, 28],

$$i\hbar\frac{\partial\psi}{\partial t}=\left\{-\frac{\hbar^{2}}{2m}\nabla^{2}+\frac{4\pi\hbar^{2}a_{\mathrm{s}}}{m}\left[\delta(\mathbf{r})+3\varepsilon_{\mathrm{dd}}V_{\mathrm{dd}}\right]\circledast n-\mu\right\}\psi.$$ (1)

The dGPE encapsulates the effects of both the short-ranged two-body interaction, mediated by $a_{\mathrm{s}}$, the scattering length of the atom-atom scattering potential, and the dipole-dipole interaction (DDI), which is defined in real and reciprocal space as [26]

$$V_{\mathrm{dd}}(\mathbf{r}) =\frac{1}{4\pi}\left[\frac{1-3(\hat{\mathbf{B}}\cdot\hat{\mathbf{r}})^{2}}{r^{3}}\right],$$ (2)

$$\widetilde{V}_{\mathrm{dd}}(\mathbf{q}) =(\hat{\mathbf{B}}\cdot\hat{\mathbf{q}})^{2}-\frac{1}{3},$$ (3)

respectively. We represent the ratio between the interaction strengths of the DDI and the short-ranged interaction through the parameter $\varepsilon_{\mathrm{dd}}=m\mu_{0}\mu_{\mathrm{d}}^{2}/(12\pi\hbar^{2}a_{\mathrm{s}})$, where $\mu_{0}$ is the permeability of free space. Additionally, in Eq. (1) the superfluid’s density and chemical potential are represented as $n=|\psi|^{2}$ and $\mu$, respectively.

As encapsulated by Eq. (2), the DDI between any two atoms in the dBEC is repulsive (attractive) when the angle between them, $\arccos(\hat{\mathbf{B}}\cdot\hat{\mathbf{r}})$, is less (more) than the critical angle $\theta_{\mathrm{c}}=\arccos(1/\sqrt{3})\approx 54.7\,\mathrm{deg}$. The resulting tendency of the superfluid atoms to realign themselves to minimise their dipolar energy is responsible for magnetostrictive effects where the condensate density exhibits anisotropy where a nondipolar condensate would not. When $\varepsilon_{\mathrm{dd}}>1$, the solutions of Eq. (1) are unstable to collapse as the anisotropic mutual dipolar attraction of the atoms overwhelms the short-ranged repulsion, necessitating the inclusion of a beyond-mean-field energy correction [53, 54, 55] in such regimes to account for the existence and stability of exotic states such as quantum droplets [56, 57] and supersolids [43, 44, 45]. For the sake of simplicity, we focus on the regime $\varepsilon_{\mathrm{dd}}<1$, where no such correction is necessary [28]. This renders it convenient to work in natural units where energies are scaled by the ground state value of the chemical potential, $\mu_{\mathrm{g}}=4\pi\hbar^{2}a_{\mathrm{s}}n_{0}/m$, for a uniform background density $n_{0}$. Simultaneously, space and time are scaled in units of the healing length $\xi=\hbar/\sqrt{m\mu_{\mathrm{g}}}$, and $\hbar/\mu_{\mathrm{g}}$, respectively, while the order parameter is scaled by $\sqrt{n_{0}}$ [58].

Our results are based on numerical solutions of the dGPE obtained by propagation via the split-step pseudospectral method [59]. Throughout this article, we work in a domain with dimensions $\{L_{x},\,L_{y},\,L_{z}\}=\{100\xi,\,100\xi,\,200\xi\}$ along the $x$, $y$- and $z$-axes, respectively, corresponding to a spatial grid of $256\,\times\,256\,\times\,512$ points. The use of discrete Fourier transforms in the pseudospectral scheme to calculate spatial derivatives results in periodic boundary conditions being imposed upon the solutions of Eq. (1). The dGPE admits quantum vortex solutions, characterised by a quantized circulation [3], viz.

$$\Gamma=\oint\,\mathrm{d}\mathbf{s}\cdot\mathbf{v}=2\pi q\,:\,q\in\mathbb{Z},$$ (4)

of the superfluid velocity, $\mathbf{v}=\nabla\left[\mathrm{Im}(\log\psi)\right]$, around the vortex core. As $\psi$ is singly-valued, this implies that the superfluid density vanishes at the vortex core. To efficiently identify regions of the spatial domain where the superfluid is characterized by both a quantized circulation and a vanishing superfluid density and a quantized circulation, we compute the superfluid’s pseudovorticity [60],

$$\boldsymbol{\omega}_{\mathrm{ps}}=\frac{1}{2}\nabla\times\left(n\mathbf{v}\right)\equiv\nabla\mathrm{Re}[\psi]\times\nabla\mathrm{Im}[\psi].$$ (5)

This quantity is only nonzero in the vicinity of a quantum vortex core and has thus been used extensively in recent studies of superfluid vortex dynamics to locate and track vortices [25, 61, 62, 63]. Subsequently, we employ the Newton-Raphson method to obtain subgrid interpolations of the space curve defining each vortex line in the domain.

III Evolution of the Vortex Lines

In a realistic experimental scenario, the Crow instability of two antiparallel vortices would arise from an initial perturbation by a source of compressible energy, where vortex-sound interactions seed Kelvin waves oscillations along the vortices [1]. In a trapped Bose-Einstein condensate, one such scenario would occur when a pair of vortices is located in a region where the trapping potential landscape is not flat, thereby resulting in background density gradients that perturb the vortices [23, 24]. In this article we have no such external perturbation and, instead, the perturbation is implicit in the initial conditions of the vortices [15, 22, 25]. Thus, we initialize our simulations with the superfluid phase of an unperturbed pair of straight, parallel vortex lines with mutually opposed circulations and apply a random perturbation to the loci of the vortex lines along their respective lengths. Let us assume without loss of generality that the two respective vorticities of the unperturbed vortices are (anti-)parallel to the $z$-axis. Noting that the phase of $\psi$, $S$, is effectively a superfluid velocity potential, i.e. $\mathbf{v}=\nabla S$, let us recall the classical expression for the velocity potential generated by a pair of straight, antiparallel vortices in an incompressible fluid as defined in the periodic domain $x\in[0,L_{x}),\,y\in[0,L_{y})$ [64, 65, 66]:

$$S(x,y)=\sum_{p=-\infty}^{\infty}\biggl{\{} \mp\operatorname{atan}\left[\tanh\left(\frac{\pi Y_{\pm}}{L_{y}}+p\pi\right)\tan\left(\frac{\pi X_{\pm}}{L_{y}}-\frac{\pi}{2}\right)\right]$$

$$+\pi\left[\Theta(X_{+})-\Theta(X_{-})\right]\biggr{\}}-\frac{2\pi(x_{+}-x_{-})y}{L_{x}L_{y}},$$ (6)

Here, $(X_{\pm},Y_{\pm})=(x-x_{\pm},y-y_{\pm})$ where the vortex with circulation $\pm 2\pi$ is located at $(x_{\pm},y_{\pm})$ . Let us define the following random perturbation profile,

$$\delta w(z)=\sum_{q=-20}^{20}\exp\left[i\pi(2\eta_{q}-1)+\frac{2\pi iqz}{L_{z}}\right]-\exp\left[i\pi(2\eta_{0}-1)\right],$$ (7)

representing the lowest $40$ excited modes along the $z$-axis equally populated with a random phase $\eta_{q}$ drawn from the uniform distribution in the interval $[0,1)$. The initial phase of $\psi$ in our simulations of the Crow instability is then given by Eq. (6) with $(x_{\pm},y_{\pm})\mapsto[x_{\pm}-\mathrm{Re}(\delta w),y_{\pm}-\mathrm{Im}(\delta w)]$. Starting from a spatially uniform profile as an initial ansatz for the superfluid density, the dGPE is evolved in imaginary time, viz. under the substitution $t\mapsto it$, with our phase ansatz held fixed until convergence of $\psi$ is achieved.

In the dimensionless dGPE theory there exist two independent parametric degrees of freedom: $\mathbf{B}$ and $\varepsilon_{\mathrm{dd}}$. In our work we consider $\varepsilon_{\mathrm{dd}}\in\{0,0.1,0.2,\mathellipsis 0.8,0.9\}$. This allows us to investigate the emergence of the Crow instability across a range of regimes from the nondipolar limit to just below the upper limit of the range of validity of the mean-field theory. For each value of $\varepsilon_{\mathrm{dd}}$, the initial unperturbed separation of the vortices in the $x$-$y$ plane is taken to be $d=6.25\xi$ along the $y$-axis and $\mathbf{B}$ is allowed to be (anti)parallel to each vortex line ($\mathbf{B}\parallel\hat{z}$), parallel to the mean vortex separation ($\mathbf{B}\parallel\hat{y}$) or parallel to the net velocity of the vortex pair ($\mathbf{B}\parallel\hat{x}$). For each choice of $\mathbf{B}$ and $\varepsilon_{\mathrm{dd}}$, we have conducted simulations of the Crow instability over an ensemble consisting of $4$ randomly selected sets of the initial Kelvin wave phase profile, $\{\eta_{q}\}$.

Figure 2: The Crow instability depicted for $\varepsilon_{\mathrm{dd}}=0.9$ and $\mathbf{B}$ parallel to $\hat{x}$ (a), $\hat{y}$ (b) and $\hat{z}$ (c).

First, let us inspect the profiles of the vortex filaments during the evolution of the dGPE. In Fig. 1 we presented snapshots of two vortex lines exhibiting the Crow instability in the nondipolar limit. Below, in Fig. 2, we depict the corresponding snapshots of the vortex pair with the same initial conditions but with $\varepsilon_{\mathrm{dd}}=0.9$ and $\mathbf{B}$ being parallel to $\hat{x}$ (first row), $\hat{y}$ (second row), or $\hat{z}$ (third row). In each of these subplots, the vortices are plotted at the times $t=t_{\mathrm{r}}/3$, $t=2t_{\mathrm{r}}/3$ and $0.5$ units of dimensionless time before the first reconnection occurs at $t=t_{\mathrm{r}}$; for the sake of brevity we denote this value of $t$ as $t_{\mathrm{c}}=t_{\mathrm{r}}-1/2$. Note that $t_{\mathrm{r}}$, and thus $t_{\mathrm{c}}$, is a nontrivial function of $\mathbf{B}$ and $\varepsilon_{\mathrm{dd}}$ as well as initial conditions, and that the translation speed of the mean orientation of the vortex pair is also a function of $\mathbf{B}$ and $\varepsilon_{\mathrm{dd}}$ [36], such that in each row $t_{\mathrm{r}}$ attains a different value and the vortices are displaced by different distances along the $y$-axes from their initial positions.

Figure 2 highlights that the final vortex profiles shortly before the reconnection induced by the Crow instability are strongly dependent on the direction of the dipolar polarisation relative to the orientation of the vortices. Compared to the corresponding vortex lines in the nondipolar limit in Fig. 1, the vortex lines when $\mathbf{B}\parallel\hat{z}$ are noticeably straighter for each of the three times as depicted in Fig. 1 (c); whereas the nondipolar limit is characterised by a highly agitated pair of vortex lines when $t=t_{\mathrm{c}}$ with multiple points along each vortex line where the vortices are curved towards each other and appear to heading for reconnection, only one such point is evident in the $z$-polarized case. The opposite case occurs for a dipole polarization along the $y$-axis, which is parallel to the mean separation of the vortices and is depicted in Fig. 2 (b). In this case, while the first reconnection occurs at approximately the same location along the $z$-axis as in the nondipolar limit, the spatial separation of the vortices at the other two separation minima is appreciably smaller, suggesting that the vortices would undergo their second and third reconnections sooner than when $\varepsilon_{\mathrm{dd}}=0$. The third case, where $\mathbf{B}\parallel\hat{z}$, is distinguishable from both of the other polarizations as the points where the vortex separation attains local minima are characterised by larger displacements of each vortex from its line-averaged mean position along the $x$-axis than for any of the other polarizations; this is particularly evident when $t=t_{\mathrm{c}}$.

IV Properties of the Vortex Lines

In Sec. III we have shown that the profiles of antiparallel vortex lines subjected to a small initial perturbation depend considerably on the strength and orientation of the DDI experienced by the background dipolar superfluid. This would suggest that the populations of the Kelvin waves that are excited through the mechanism underlying the Crow instability are dependent in some manner on the specifics of the DDI. In this section we investigate the spectral properties of the vortex line profiles and characterise the Kelvin wave populations during the simulations of the dGPE presented in the previous section.

Let us proceed by representing the coordinates of the $n$th vortex line as $w_{n}(z)=x_{n}(z)+iy_{n}(z)$. Shifting the average of these coordinates over the $z$-axis to the origin of the $x$-$y$ plane yields the quantity $\tilde{w}_{n}(z,t)=w_{n}(z,t)-\langle w_{n}\rangle(t)$, where $\tilde{w}_{n}(z,t=0)=(-1)^{n}\delta w(z)$ is the initial Kelvin wave perturbation given by Eq. (7). The corresponding mode amplitudes, $W_{n}(k_{z},t)$, are the discrete Fourier transforms of $\tilde{w}_{n}(z,t)$ and the periodicity of the domain ensures a restriction of the mode wavenumber, $k_{z}$, such that $k_{z}=2\pi q/L_{z},\,q\in\mathbb{Z}$. We also note that, by construction, $W_{n}(k_{z}=0,t)=0$. Initially, we study the temporal evolution of these mode amplitudes, focussing on those initially occupied at $t=0$. Let us define the proportional amplitude of a mode, $W^{\prime}(k_{z},t)$, as

$$W^{\prime}(k_{z},t)=\frac{|W_{1}(k_{z},t)|+|W_{2}(-k_{z},t)|}{\sum_{k_{z}>0}\left[|W_{1}(k_{z},t)|+|W_{2}(-k_{z},t)|\right]}.$$ (8)

Figure 3 depicts $W^{\prime}(k_{z},t)$, with $k_{z}$ restricted to the $20$ lowest-lying modes with positive wavenumber, for the nondipolar system (a) as well as the maximally dipolar regime, $\varepsilon_{\mathrm{dd}}=0.9$, with $\mathbf{B}$ parallel to $\hat{x}$ (b), $\hat{y}$ (c), or $\hat{z}$ (d). For each distinct regime, we average the results over the $4$ sets of initial conditions over the duration of time in which all $4$ simulations have not yet undergone a vortex reconnection.

Figure 3: The relative occupation of the first 20 Kelvin wave modes with positive $k_{z}$ during time evolution, averaged over the ensemble of initial state Kelvin wave profiles. In (a), (top left), $\varepsilon_{\mathrm{dd}}=0$ and in the remaining subfigures, $\varepsilon_{\mathrm{dd}}=0.9$; $\mathbf{B}$ is parallel to $\hat{x}$ in (b), $\hat{y}$ (c) and $\hat{z}$ (d).

An inspection of Fig. 3 affords us the following insights into the evolution of the Kelvin modes on the vortices. Regardless of the dipolar regime, we can see that at early times the interaction of the modes results in fluctuations of the relative mode populations, accompanied by the suppression of strongly energetic high order modes with $p\gtrsim 10$. It is not until the vortices enter their final stage of evolution – that is, with reference to the snapshots of the vortex states in Figs. 1 and  2, when $t>2t_{\mathrm{r}}/3$ – that unambiguous signatures of a dependence of the Kelvin mode population on $\mathbf{B}$ and $\varepsilon_{\mathrm{dd}}$ become apparent. For large values of $\varepsilon_{\mathrm{dd}}$ with the dipole moments polarised (anti-)parallel to the vortex lines, as is the case in Fig. 3(d), the Kelvin mode populations become monotonically decreasing as a function of $k_{z}$ as $t\rightarrow t_{\mathrm{r}}$. A comparison with Fig. 2 (c) shows that these spectral features are consistent with the qualitative appearance of the vortex lines at late times since, for both vortices at $t=t_{\mathrm{c}}$, only one pair of antinodes of the vortex displacement from its mean position is apparent. By contrast, it is clear that polarising the dipole moments orthogonal to the vorticity of either vortex stimulates the population of higher Kelvin modes. The averaged spectrum for polarisations along the translational axis of the vortex pair in Fig. 3(b) illustrates clearly that the $q=3$ mode becomes overwhelmingly preferred as the Crow instability manifests itself and the vortices approach their first reconnection. In line with our observations for the $\mathbf{B}\parallel\hat{z}$ case, this aspect of the Kelvin wave spectrum for the $\mathbf{B}\parallel\hat{x}$ case is consistent with the plot of the vortices’ state at $t=t_{\mathrm{c}}$ in Fig. 2 (a) where $3$ distinct antinode pairs are distinguishable on each vortex. Intriguingly, while Fig. 3 (c) shows that odd values of the Kelvin mode index $q$ appear to be preferred as $t\rightarrow t_{\mathrm{c}}$ when the dipole polarisation is parallel to the vortices’ mean separation, it is evident that no single mode is overwhelmingly dominant in this regime and, similarly to the nondipolar limit in Fig. 3 (a), the first $6$ modes contribute to the majority of the mode population when $\mathbf{B}\parallel\hat{y}$. Indeed, while the two antinodes of the vortices at $z\approx-20$ and $z\approx-50$ when $t=t_{\mathrm{c}}$ appear to be closer to instigating reconnections in Fig. 2 (b) than the corresponding antinodes are in Fig. 1, the nondipolar vortices are otherwise much more similar to the dipolar vortices in Fig. 2 (b) than those in (a) and (c) when $t=t_{\mathrm{c}}$.

Figure 4: The wavenumber, $k_{z}$, of the Kelvin mode that is maximally occupied by the vortex pair immediately prior to their mutual reconnection as a function of $\varepsilon_{\mathrm{dd}}$ where the dipole polarisation is parallel to $\hat{x}$ (a), $\hat{y}$ (b), and $\hat{z}$ (c). In each of these subplots, distinct curves correspond to distinct Kelvin wave initial conditions indexed by $\eta\in\{1,\,2,\,3,\,4\}$.

With these insights on the evolution of $W^{\prime}(k_{z},t)$ over all of the low-lying modes for the maximally dipolar regime, $\varepsilon_{\mathrm{dd}}=0.9$, we proceed to study specific features of the mode spectrum for the full range of $\varepsilon_{\mathrm{dd}}$ from $0$ to $0.9$. Initially, let us consider the maximally unstable Kelvin mode for a given choice of $\{\mathbf{B},\varepsilon_{\mathrm{dd}}\}$, viz. the value of $k_{z}$, denoted hereafter as $k_{\mathrm{c}}$, that maximises $W^{\prime}(k_{z},t_{\mathrm{c}})$. Figure 4 presents plots of this wavenumber as a function of $\varepsilon_{\mathrm{dd}}\in[0.0.9]$, with $\mathbf{B}$ parallel to $\hat{x}$ (a), $\hat{y}$ (b), and $\hat{z}$ (c), where each choice of initial condition has been plotted as a distinct curve in each subplot. In the nondipolar limit we find that, depending on the initial conditions, either the 3rd or 5th Kelvin modes are most strongly excited through the Crow instability. Taking the higher of the two modes and comparing its wavenumber, $k_{\mathrm{c}}\xi=10\pi/200\approx 0.157$, to earlier studies of the nondipolar Crow instability shows that this is consistent with predictions that $dk_{c}\sim 1$ [22].

When the dipolar interaction strength is nonzero the behaviour of $k_{\mathrm{c}}$ is dependent to the dipole polarisation relative to the vortex configuration. When $\mathbf{B}\parallel\hat{y}$, $k_{\mathrm{c}}$ is particularly insensitive to $\varepsilon_{\mathrm{dd}}$ and the specific mode that is most strongly excited as a consequence of the Crow instability appears to be determined only by our choice of initial conditions. This, we note, is consistent with the ensemble-averaged relative mode populations in Fig. 3 being qualitatively similar for both the nondipolar limit (a) and for the regime $\varepsilon_{\mathrm{dd}}=0.9,\,\mathbf{B}\parallel\hat{y}$ (c) as $t\rightarrow t_{\mathrm{c}}$. However, such robustness against changes in the relative dipolar interactions strength is not evident when the dipole moments are polarised along either $\mathbf{x}$ or $\mathbf{z}$. In Fig. 2 (d), for the regime $\varepsilon_{\mathrm{dd}}=0.9,\,\mathbf{B}\parallel\hat{z}$ it was apparent that the $q=1$ Kelvin mode becomes overwhelmingly preferred as $t\rightarrow t_{\mathrm{c}}$. This is supported by the behaviour of $k_{\mathrm{c}}$ in Fig. 4 (c) for this regime where we see that only one of the $4$ initial conditions results in the $q=2$ mode being most strongly excited with the remainder preferentially occupying the $q=1$ mode. We also find that, regardless of the initial conditions applied, $k_{\mathrm{c}}$ is a decreasing function of $\varepsilon_{\mathrm{dd}}$ when $\mathbf{B}\parallel\hat{z}$, such that the vortex loops formed as a consequence of the Crow instability are unambiguously larger for more strongly dipolar superfluids when the dipole polarisation is (anti-)parallel to the initial vorticity of each vortex. The case when $\mathbf{B}\parallel\hat{x}$ is a little more ambiguous, though, as the evident drift to lower $k_{\mathrm{c}}$ for higher $\varepsilon_{\mathrm{dd}}$ is not obeyed for all of the choices of perturbation initial conditions. We note that each of the initial conditions imply the same mode, $q=3$ becoming preferential for the largest value of $\varepsilon_{\mathrm{dd}}$ considered in our work, $\varepsilon_{\mathrm{dd}}=0.9$, but an explanation for the volatility of this quantity with respect to the initial conditions for this particular dipole alignment is not readily apparent.

Having identified the maximally unstable Kelvin mode, $k_{\mathrm{c}}$, in a given dipolar regime, we have also studied the evolution of the corresponding mode amplitude, $W(k_{\mathrm{c}},t)$, from the start of each dGPE simulation till the reconnection induced by the Crow instability. After an early transient arising from nonlinear interactions with the other Kelvin modes, we find that $W(k_{\mathrm{c}},t)$ grows approximately exponentially in time, i.e. $W(k_{\mathrm{c}},t)\sim\exp(\sigma t)$ with $\sigma$ the growth rate. This is consistent with a linear dynamical instability of this mode arising from the Crow mechanism, which is evident in Fig. 5 (a) where a plot of $\log_{10}\,W(k_{\mathrm{c}},t)$ is approximately linear as a function of $t$ after sufficient time has passed. Here, $\varepsilon_{\mathrm{dd}}=0.9$ and $\mathbf{B}\parallel\{\hat{x},\hat{y},\hat{z}\}$. Figure 5 (a) also illustrates that the time taken till the first reconnection varies considerably depending on the dipole polarisation, though it is known from prior investigations that the velocities of vortex pairs are themselves dipole-dependent [36].

Figure 5: (a) $\log_{10}\,W(k_{\mathrm{c}},t)$ as a function of $t$, with $k_{\mathrm{c}}$ the maximally unstable mode wavenumber and $\varepsilon_{\mathrm{dd}}=0.9$. The initial conditions correspond to those in Figs. 1 and 2. (b) The ensemble-average of $\sigma$, the exponential growth rate of $W(k_{\mathrm{c}},t)$, as a function of $\varepsilon_{\mathrm{dd}}$. In both plots, $\mathbf{B}\parallel\{\hat{x},\hat{y},\hat{z}\}$.

To excise the transient, the exponential growth rate $\sigma$ has been computed through linear regression fitting of $\log\,W(k_{\mathrm{c}},t)$ to $t$ for times after the its growth is exponential, which for the sake of consistency we take to be the interval $t\in[t_{\mathrm{r}}/2,t_{\mathrm{c}}]$. The ensemble average of $\sigma$ over the set of four Kelvin wave initial conditions is plotted as a function of $\varepsilon_{\mathrm{dd}}$ as distinct curves for $\mathbf{B}\parallel\{\hat{x},\hat{y},\hat{z}\}$ in Fig. 5 (b). In the nondipolar limit, we can compare our computed value of the growth rate, $\sigma\approx 0.0118$, to predictions of $\sigma$ obtained in the literature through linear stability analysis of the corresponding Bogoliubov-de Gennes (BdG) equations [15, 22]:

$$\sigma(k_{\mathrm{c}})^{2}\sim\frac{k_{\mathrm{c}}^{2}}{d^{2}}\left[\ln(\sqrt{2}d)+0.38\right].$$ (9)

This yields the estimation $\sigma\approx 0.04$; while the two predictions are of the same order of magnitude, the discrepancy between the two may be due to the nonlinear growth of $W(k_{\mathrm{c}},t)$ evident in Fig. 5 (a). For finite values of $\varepsilon_{\mathrm{dd}}$, our results show that $\sigma$ is smaller when $\mathbf{B}\parallel\hat{x}$ than for the other two polarisations regardless of $\varepsilon_{\mathrm{dd}}$ and that as $\varepsilon_{\mathrm{dd}}\rightarrow 1$, a hierarchy such that $\sigma(\hat{x})<\sigma(\hat{z})<\sigma(\hat{y})$ manifests itself. Now, given that the initial mean separation of the vortices is identical for all of the simulations we have conducted, one would expect that longer reconnection times are consistent with smaller values of $\sigma$ and the hierarchy of $\sigma$ in Fig. 5 (b) is consistent with that of the ensemble-averaged reconnection times, $t_{\mathrm{r}}(\hat{x})>t_{\mathrm{r}}(\hat{z})>t_{\mathrm{r}}(\hat{y})$, seen in Fig. 2 in the regime $\varepsilon_{\mathrm{dd}}=0.9$.

From our spectral analysis of the vortex lines during their evolution, it is clear that there is a significant dependence on the polarisation of the dipole moments relative to the initial vortex alignment and the DDI interaction strength. The dependence of the maximally unstable mode $k_{\mathrm{c}}$ on the characteristics of the DDI, in particular, suggests that local geometric properties of the vortex line are sensitive to these parameters. One such quantity is the curvature of the vortices, which we now proceed to as a function of time for each of the simulations conducted. Parametrising a given vortex line as $\boldsymbol{\gamma}(s)=\gamma_{i}(s)\hat{e}_{i}$, with $\gamma_{z}(s)\equiv s$, its curvature is defined through this parametrisation as [67]

$$\kappa(s)=\frac{|\boldsymbol{\gamma}^{\prime}(s)\times\boldsymbol{\gamma}^{\prime\prime}(s)|}{|\boldsymbol{\gamma}^{\prime}(s)|^{3}}.$$ (10)

In Eq. (10), we compute the derivatives directly from the mode amplitudes $\{W_{n}(k_{z},t)\}$ by utilising the standard properties of the discrete Fourier transform and by applying a simple low-pass filter such that only the modes $L_{z}|k_{z}|/(2\pi)\leq 100$ contribute to the curvature .

Figure 6: The mean curvature over the length of the two vortices ensemble-averaged over the set of vortex initial conditions, $\langle\kappa\rangle$, at the time $t=t_{\mathrm{c}}$. The magnetic dipole orientation is parallel to $\hat{x}$ in (a), $\hat{y}$ in (b), and $\hat{z}$ in (c).

Let us focus on the vortex profiles when $t=t_{\mathrm{c}}$ and compare the curvature of the two vortices across the various regimes of $\varepsilon_{\mathrm{dd}}$ and $\mathbf{B}$. Finding the mean curvatures of both vortices over their respective line lengths and averaging the two resulting quantities yields the quantity $\langle\kappa(t_{\mathrm{c}})\rangle$, which is plotted in Fig. 6 as a function of $\varepsilon_{\mathrm{dd}}$. Here, $\mathbf{B}$ is assumed to be parallel to the $x$-, $y$- and $z$-axes in (a), (b) and (c), respectively, and for each choice of $\{\mathbf{B},\,\varepsilon_{\mathrm{dd}}\}$ we have performed an ensemble average over the $4$ initial vortex perturbation profiles. Figure 6 illustrates clearly the following salient characteristics of the vortex line profiles arising from the Crow instability. While fluctuations in the overall behaviour of the vortex lines occur due to the specific choice of initial conditions, it is evident in Fig. 6 (a) that $\langle\kappa(t_{\mathrm{c}})\rangle$ is largely independent of $\varepsilon_{\mathrm{dd}}$ when the dipole moments are polarised along the $x$-axis, the translation axis of the vortices, when compared to the $y$- and $z$-axes. As for dipole polarisations parallel to the $y$-axis, Fig. 6 (b) suggests that $\langle\kappa(t_{\mathrm{c}})\rangle$ generally increases with $\varepsilon_{\mathrm{dd}}$ and that the values it attains are comparable in magnitude to the corresponding values for polarisations along the $x$-axis.

These observations are in sharp contrast to the results presented in Fig. 6 (c), where the mean vortex curvature at $t=t_{\mathrm{c}}$ is seen to be strongly dependent on $\varepsilon_{\mathrm{dd}}$ when the dipole moments are polarised along the $z$-axis, which is (anti-)parallel to the mean vorticity of each axis. As $\varepsilon_{\mathrm{dd}}$ is increased, the curvature of the vortex is suppressed rather than enhanced. This has been attributed in earlier studies of Kelvin waves perturbing single, isolated vortices to the effects of magnetostriction, where there exists an effective interaction between the virtual dipole moments, induced in the vortex line, that are polarised antiparallel to the real dipole moments in the bulk superfluid [51, 68, 52]. When $\mathbf{B}\parallel\hat{z}$, viz. (anti-)parallel to the mean orientation of the vortex line, the effective dipolar interaction energy arising from the mutual interaction of the virtual dipole moments is minimised when the dipole moments are aligned along this mean orientation axis. Thus, a suppression of the curvature of the vortex arises since a larger curvature would effectively cause the configuration of the vortex’s virtual dipole moments to become misaligned with the magnetic field and thus increase the dipolar interaction energy. This also results in the wavenumber of the preferentially excited Kelvin mode, $k_{\mathrm{c}}$, being smaller when $\mathbf{B}\parallel\hat{z}$ than for polarisations along the other principal axes as seen in Fig. 4. These effects are naturally more pronounced for larger values of $\varepsilon_{\mathrm{dd}}$, which leads to smaller curvatures and thus also the lower values of $k_{\mathrm{c}}$ seen in Figs. 4 (c).

We also note that magnetostriction arising from the virtual dipole moments in the vortex cores serves to explain the salient features of the vortices when $\mathbf{B}\parallel\hat{x},\,\hat{y}$ as well. When the dipole polarisation are polarised along the $y$-axis, the energetic preference of the vortices’ virtual dipole moments is to align along this axis and such vortex curvature is stimulated along this axis. While this results in the monotonic behaviour of $\langle\kappa_{\mathrm{c}}\rangle$ in Fig. 6 (b), it is also responsible for the reconnections occurring on a faster timescale at fixed $\varepsilon_{\mathrm{dd}}$ than for the other polarisation orientations, as observed in Fig. 2 and, as we have argued, also explains why the exponential mode growth factor $\sigma$ is larger when $\mathbf{B}\parallel\hat{y}$ than the other polarisations. Similarly, when the dipole polarisation is parallel to the $x$-axis the contributions to the overall curvature arising from curvature orthogonal to this axis are suppressed, such that the majority of the curvature arises from curvature along the propagation axis of the vortex pair. While this results in the strongly transversely perturbed vortex profiles in the $x$-$z$ plane as is seen in Fig. 2 (a), curvature along the $y$-axis is greatly inhibited and the reconnection is suppressed. Thus, the exponential growth of the dominant Kelvin mode is suppressed by the DDI when $\mathbf{B}\parallel\hat{x}$, resulting in smaller values of the mode growth factor $\sigma$ and correspondingly longer reconnection times.

V Conclusion

Through studying the evolution of a pair of quantum vortices in a dipolar superfluid with an initial transverse, symmetric perturbation applied to each vortex, we have found that the quantum analogue of the Crow instability behaves in vastly different ways depending on the magnetic dipole polarisation and the relative strength of the DDI. When the dipole polarisation, $\mathbf{B}$, is (anti-)parallel to the initial mean vorticity of each vortex, the vortices’ curvature is strongly suppressed due to the influence of the interaction between virtual dipole moments inside the vortex core. This results in Kelvin waves of decreasing wavenumber contributing most strongly to the Crow instability as the relative dipolar interaction strength, $\varepsilon_{\mathrm{dd}}$, increases. This magnetostrictive effect has the converse effect when the dipole polarisation is orthogonal to the initial vorticities, with the curvature of each vortex as well as the growth of Kelvin modes of higher wavenumber being enhanced by the influence of the DDI. While the Kelvin mode occupation spectrum for a magnetic dipole polarisation parallel to the separation between the two vortices appears rather similar to its nondipolar analogue, these modes induce a greater curvature in the vortices as $\varepsilon_{\mathrm{dd}}$ increases and, correspondingly, a Crow instability that develops more rapidly. However, for dipole polarisations parallel to the binormal axis, the Kelvin modes become concentrated at one specific higher wavenumber mode, $p=3$, in the large-$\varepsilon_{\mathrm{dd}}$ limit and the magnetostrictive elongation of the vortex lines along the binormal axis suppresses the Crow instability when compared to the other two dipole polarisations considered here.

Our studies of the dependence of the Crow instability on the orientation of the dipole moments suggest that other hydrodynamical instabilities might exhibit directional dependence in the presence of a DDI. Notably, the snake instability of dark solitons in three-dimensional dipolar condensates has been shown to be suppressed when the dipole alignment is parallel to the solitonic phase slip [69]. The corresponding phase slip due to the vortex line pair in our system is parallel to the $x$-axis and our results demonstrate a suppression of the Crow instability for dipole polarisations parallel to this axis. Given the intrinsic correspondence of the snake instability of exhibited by dark solitons and the Crow instability of antiparallel vortex lines [15], this is likely not a coincidence. This suppression of the snake instability was demonstrated through linear stability analysis applied to the BdG spectra of a three-dimensional dark soliton [69]; a similar procedure was conducted in Ref. [22] to characterise the linear stability of antiparallel vortex lines in a nondipolar BEC against the transverse perturbations that induce the Crow instability. An application of linear stability analysis to the corresponding dipolar problem as studied in this article may thus prove to be illuminating. We also note that Ref. [69] studied only dipole orientations parallel to the phase slip for dark solitons in three-dimensional regimes, and that our analysis of the Crow instability for distinct dipole orientations suggest that it is prudent to also examine the snake instability for dipole orientations in the soliton plane.

We also note that our results bear implications for aspects of quantum vortex dynamics in dipolar superfluids that, to the best of our knowledge, have not yet been investigated thoroughly. For instance, in Sec. IV we have posited a relationship between the exponential growth factor of the dominant Kelvin mode, $\sigma$, and the time taken for the vortices to undergo their first reconnection, $t_{\mathrm{r}}$. While the values of $\sigma$, ensemble-averaged over the set of initial conditions for the vortex perturbation profiles, exhibited a clear dependence on $\mathbf{B}$ and $\varepsilon_{\mathrm{dd}}$, it would be desirable to investigate such scaling behaviour in a setting that is independent of the randomness of the initial conditions present in our system. In a dipolar superfluid, it would be of considerable interest to elucidate the relationship between the parameters of the DDI and the temporal scaling of quantities such as the inter-vortex separation, the angle between the normal axes of the vortex lines, and global properties such as the vortices’ curvature and torsion, and establish whether or not they exhibit deviations from the familiar nondipolar paradigm [67, 70]. Furthermore,in the nondipolar limit a considerable degree of insight may be gleaned into the dynamics of quantum vortex lines by assuming that the superfluid flow is purely compressible, which allows the use of the Biot-Savart law and the local induction approximation to model vortices before reconnective processes occur [71, 72]. In this article, a strong influence has been demonstrated of the magnetic dipole polarisation on the bending of vortex line pairs, while the dynamics of quasi-two-dimensional point-vortices and three-dimensional straight vortex lines have been shown to exhibit dipole-mediated deviations from the predictions of the point-vortex model, the two-dimensional limit of the Biot-Savart law. We also note that the Kelvin wave dispersion relation for a single vortex line is modified in the presence of the DDI [51, 52]. Given these pre-existing and new examples, an extension of the Biot-Savart law [71, 72] to account for the dipolar interaction between the virtual dipole moments in the vortex line cores would provide an intuitive understanding of how these disparate results can be generalised for a larger number of dipolar vortices. Furthermore, we also note that while this investigation assumed a value of $\varepsilon_{\mathrm{dd}}$ less than $1$, such that the dBEC does not exhibit the supersolid phase [43, 44, 45], the interaction of supersolid quantum droplets and vortices produces complicated dynamics that are inaccessible in the superfluid phase. It is feasible to suggest that this might result in Kelvin waves and subsequently the Crow instability in pairs of antiparallel vortices [73].