Abstract
Phase transitions are ubiquitous in our threedimensional world. By contrast, most conventional transitions do not occur in infinite uniform lowdimensional systems because of the increased role of thermal fluctuations. The crossover between these situations constitutes an important issue, dramatically illustrated by BoseEinstein condensation: a gas strongly confined along one direction of space may condense along this direction without exhibiting true longrange order in the perpendicular plane. Here we explore transverse condensation for an atomic gas confined in a novel trapping geometry, with a flat inplane bottom, and we relate it to the onset of an extended (yet of finiterange) inplane coherence. By quench crossing the transition, we observe topological defects with a mean number satisfying the universal scaling law predicted by KibbleZurek mechanism. The approach described can be extended to investigate the topological phase transitions that take place in planar quantum fluids.
Introduction
Bose–Einstein condensation (BEC) is a remarkably simple phase transition that can in principle occur in a fluid even in the absence of interatomic interactions. As a mere result of singleparticle statistics, a phasecoherent fraction appears in the fluid, described by a uniform wavefunction spanning the whole system. During the last two decades, cold atom experiments have been used to probe many aspects of BEC^{1,2,3}. However, most of these cold atom studies are performed in the presence of a harmonic confinement. BEC becomes in this case a local transition: the condensate forms at the centre of the trap where the density is the largest, and interactions between particles play a dominant role in the equilibrium state of the fluid. In this geometry, the nonhomogeneous character of the gas makes it difficult to address some important features of BEC, such as the existence of longrange phase coherence. The recent achievement of a threedimensional (3D) Bose gas undergoing BEC in a boxlike potential^{4,5} constitutes an important step forward, realizing the textbook paradigm of an extended and uniform coherent matter wave.
When turning to lowdimensional (lowD) systems, subtle effects emerge because of the entangled roles of Bose statistics and thermal fluctuations. First, in an infinite lowD ideal gas, no BEC is expected at nonzero temperature, because of the modification of the singleparticle density of states with respect to the 3D case^{6}. In other words, the phase coherence between two points tends to zero when their distance increases, contrary to the 3D situation. Second, Bose statistics can play a key role in the dimensional crossover between the 3D situation and a lowD one, by facilitating the freezing of some directions of space. Consider the uniform twodimensional (2D) case obtained by imposing a tight harmonic trapping potential (frequency ν_{z}) along the third direction z. The transverse condensation phenomenon^{7,8,9} allows one to reach an effective 2D situation even in the ‘thermally unfrozen’ regime, where the quantum hν_{z} is smaller than the thermal energy k_{B}T (h and k_{B} stand for Planck’s and Boltzmann’s constants). Third, for the 2D case in the presence of interactions between the particles, the situation gets more involved with the possibility of a superfluid, Berezinskii—Kosterlitz—Thouless (BKT) transition^{10,11} for a large enough phasespacedensity, even though the absence of true longrange coherence remains valid^{12,13}. This superfluid transition has been identified and characterized over the recent years with nonhomogeneous, harmonically trapped Bose gases^{14,15,16,17,18,19,20}.
Another key feature of phase transitions for uniform systems is the time needed to establish the coherence/quasicoherence over the whole sample. As is well known for critical phenomena^{21}, the coherence length and the thermalization time diverge at the transition point, thus limiting the size of the phasecoherent domains that are formed at its crossing. The KibbleZurek (KZ) theory^{22,23} allows one to evaluate the scaling of the domain size with the speed of the crossing. Once the transition has occurred, these domains start merging together. During this coarsening dynamics singularities taking the form of topological defects can be nucleated at their boundaries, with a spatial density directly related to the characteristic domain size. The KZ mechanism has been studied in a variety of experimental systems (see for example, refs 24, 25, 26, 27, 28, 29), including cold atomic gases^{30,31,32,33,34,35,36}. In 2D quantum fluids, the singularities take the form of quantized vortices, that is, points of zero density around which the macroscopic wavefunction of the gas has a ±2π phase winding.
In this paper, we present an experimental realization of a uniform atomic Bose gas in a quasi2D geometry, addressing both the steady state of the fluid and its quench dynamics. First with the gas in thermal equilibrium, we characterize the threshold for the emergence of an extended phase coherence by two independent methods, respectively, based on the measurement of the atomic velocity distribution and on matter–wave interferences. We show in particular that for the thermally unfrozen case (with ζ=k_{B}T/hν_{z}), the transverse condensation phenomenon induces an extended inplane coherence. Second, we explore the quench dynamics of the gas prepared in an initial state such that and observe density holes associated to vortices. We study the relation between the cooling rate and the number of vortices that subsist after a given relaxation time, and we compare our results with the predictions of the KZ theory.
Results
Production of uniform gases in quasi2D geometries
We prepare a cold 3D gas of rubidium (^{87}Rb) atoms using standard laser and evaporative cooling techniques. Then we transfer the gas in a trap formed with two orthogonal laser beams at wavelength 532 nm, shorter than the atomic resonance wavelength (780 nm), so that the atoms are attracted towards the regions of low light intensity (Fig. 1a). The strong confinement along the z direction (vertical) is provided by a laser beam propagating along the x direction. It is prepared in a HermiteGauss mode, with a node in the plane z=0, and provides a harmonic confinement along the z direction with a frequency ν_{z} in the range 350–1,500 Hz. For the confinement in the horizontal xy plane, we realize a boxlike potential by placing an intensity mask on the second laser beam path, propagating along the z direction (Fig. 1b). Depending on the study to be performed, we can vary the shape (disk, square, double rectangle) and the area (from 200 to 900 μm^{2}) of the region accessible to the gas in the plane. The relevance of our system for the study of 2D physics is ensured by the fact that the size of the ground state along the z direction is very small compared with the inplane extension . The number of atoms N that can be stored and reliably detected in this trap ranges between 1,000 and 100,000. We adjust the temperature of the gas in the interval T~10−250 nK by varying the intensity of the beam creating the box potential, taking advantage of evaporative cooling on the edges of this box. The ranges spanned by ν_{z} and T allow us to explore the dimensional crossover between the thermally frozen regime () and the unfrozen one (). Examples of in situ images of 2D gases are shown in Fig. 1c–e.
Phase coherence in 2D geometries
For an ideal gas, an important consequence of BoseEinstein statistics is to increase the range of phase coherence with respect to the prediction of Boltzmann statistics. Here coherence is characterized by the onebody correlation function , where [resp. ] annihilates (resp. creates) a particle in r, and where the average is taken over the equilibrium state at temperature T. For a gas of particles of mass m described by Boltzmann statistics, G_{1}(r) is a Gaussian function , where is the thermal wavelength.
Consider the particular case of a 2D Bose gas (for example, ). When its phasespacedensity becomes significantly larger than 1 (ρ stands for the 2D spatial density), the structure of G_{1}(r) changes. In addition to the Gaussian function mentioned above, a broader feature develops, with the characteristic length that increases exponentially with (see ref. 37 and Supplementary Note 1)
Usually, two main effects amend this simple picture:

In a finite system, when the predicted value of becomes comparable to the size L of the gas, one recovers a standard BoseEinstein condensate, with a macroscopic occupation of the ground state of the box potential^{38}. The G_{1} function then takes nonzero values for any r≤L and the phase coherence extends over the whole area of the gas.

In the presence of weak repulsive interactions, the increase of the range of G_{1} for is accompanied with a reduction of density fluctuations, with the formation of a ‘quasicondensate’ or ‘presuperfluid’ state^{15,16,39}. This state is a medium that can support vortices, which will eventually pair at the superfluid BKT transition for a larger phasespacedensity, around for the present strength of interactions^{39}. At the transition point, the coherence length diverges and above this point, G_{1}(r) decays algebraically.
Role of the third dimension for inplane phase coherence
When the thermal energy k_{B}T is not negligibly small compared with the energy quantum hν_{z} for the tightly confined dimension, the dynamics associated to this direction brings interesting novel features to the inplane coherence. First, we note that the function G_{1}(r) can be written in this case as a sum of contributions of the various states j_{z} of the z motion (see Supplementary Note 1). The term with the longest range corresponds to the ground state j_{z}=0, with an expression similar to equation (1), where is replaced by the phasespacedensity associated to this state. Now, consider more specifically the unfrozen regime . In this case, one expects that for very dilute gases only a small fraction f_{0} of the atoms occupies the j_{z}=0 state; Boltzmann statistics indeed leads to . However, for large total phasespace densities , BoseEinstein statistics modifies this result through the transverse condensation phenomenon (BEC_{⊥})^{7}: the phasespacedensity that can be stored in the excited states j_{z}≠0 is bounded from above, and can thus become comparable to . This large value of leads to a fast increase of the corresponding range of G_{1}(r), thus linking the transverse condensation to an extended coherence in the xy plane. This effect plays a central role in our experimental investigation.
Phase coherence revealed by velocity distribution imaging
To characterize the coherence of the gas, we study the velocity distribution, that is, the Fourier transform of the G_{1}(r) function. We approach this velocity distribution in the xy plane by performing a 3D timeofflight (3D ToF): we suddenly switch off the trapping potentials along the three directions of space, let the gas expand for a duration τ, and finally image the gas along the z axis. In such a 3D ToF, the gas first expands very fast along the initially strongly confined direction z. Thanks to this fast density drop, the interparticle interactions play nearly no role during the ToF and the slower evolution in the xy plane is governed essentially by the initial velocity distribution of the atoms. The timeofflight (ToF) duration τ is chosen so that the size expected for a Boltzmann distribution is at least twice the initial extent of the cloud. Typical examples of ToF images are given in Fig. 2a–f. Whereas for the hottest and less dense configurations, the spatial distribution after ToF has a quasipure Gaussian shape, a clear nonGaussian structure appears for larger N or smaller T. A sharp peak emerges at the centre of the cloud of the ToF picture, signalling an increased occupation of the lowmomentum states with respect to Boltzmann statistics, or equivalently a coherence length significantly larger than λ_{T}.
In order to analyse this velocity distribution, we chose as a fit function the sum of two Gaussians of independent sizes and amplitudes, containing N_{1} and N_{2} atoms, respectively (see Fig. 2d–f). We consider the bimodality parameter Δ=N_{1}/N defined as the ratio of the number of atoms N_{1} in the sharpest Gaussian to the total atom number N=N_{1}+N_{2}. A typical example for the variations of Δ with N at a given temperature is shown in Fig. 2g for an initial gas with a square shape (side length L=24 μm). It shows a sharp crossover, with essentially no bimodality () below a critical atom number N_{c}(T) and a fast increase of Δ for N>N_{c}(T). We extract the value N_{c}(T) by fitting the function Δ ∝ (1−(N_{c}/N)^{0.6}) to the data. We chose this function as it provides a good representation of the predictions for an ideal Bose gas in similar conditions (see Methods).
Phase coherence revealed by matter–wave interference
Matter–wave interferences between independent atomic or molecular clouds are a powerful tool to monitor the emergence of extended coherence^{4,14,40,41,42}. To observe these interferences in our uniform setup, we first produced two independent gases of similar density and temperature confined in two coplanar parallel rectangles, separated by a distance of 4.5 μm along the x direction (see Fig. 1e). Then we suddenly released the box potential providing confinement in the xy plane, while keeping the confinement along the z direction (2D ToF). The latter point ensures that the atoms stay in focus with our imaging system, which allows us to observe interference fringes with a good resolution in the region where the two clouds overlap. A typical interference pattern is shown in Fig. 3a, where the fringes are (roughly) parallel to the y axis, and show some waviness that is linked to the initial phase fluctuations of the two interfering clouds.
We use these interference patterns to characterize quantitatively the level of coherence of the gases initially confined in the rectangles. For each line y of the pixelized image acquired on the CCD camera, we compute the xFourier transform of the spatial density ρ(x, y) (Fig. 3b). For a given y, this function is peaked at a momentum k_{p}(y)>0 that may depend (weakly) on the line index y. Then we consider the function that characterizes the correlation of the complex fringe contrast along two lines separated by a distance d
Here * denotes the complex conjugation and the average is taken over the lines y that overlap with the initial rectangles. If the initial clouds were two infinite, parallel lines with the same G_{1}(y), one would have γ(d)=G_{1}(d)^{2} (ref. 43). Here the nonzero extension of the rectangles along x and their finite initial size along y make it more difficult to provide an analytic relation between γ and the initial G_{1}(r) of the gases. However, γ(d) remains a useful and quantitative tool to characterize the fringe pattern. For a gas described by Boltzmann statistics, the width at 1/e of G_{1}(r) is and remains below 1 μm for the temperature range investigated in this work. Since we are interested in the emergence of coherence over a scale that significantly exceeds this value, we use the following average as a diagnosis tool
For the parameter Γ to take a value significantly different from 0, one needs a relatively large contrast on each line, and relatively straight fringes over the relevant distances d, so that the phases of the different complex contrasts do not average out.
For a given temperature T, the variation of Γ with N shows the same thresholdtype behaviour as the bimodality parameter Δ. One example is given in Fig. 3c, from which we infer the threshold value for the atom number N_{c}(T) needed to observe interference fringes with a significant contrast.
Scaling laws for the emergence of coherence
We have plotted in Fig. 4 the ensemble of our results for the threshold value of the total 2D phasespacedensity as a function of ζ=k_{B}T/hν_{z}, determined both from the onset of bimodality as in Fig. 2g (closed symbols) or from the onset of visible interference as in Fig. 3c (open symbols). Two trapping configurations have been used along the z direction, ν_{z}=1,460 Hz and ν_{z}=365 Hz. In the first case, the z direction is nearly frozen for the temperatures studied here (). In the second one, the z direction is thermally unfrozen (). All points approximately fall on a common curve, independent of the shape and the size of the gas: varies approximately linearly with ζ with the fitted slope 1.4 (3) for and approaches a finite value ~4 for .
In the frozen case, a majority of atoms occupy the vibrational ground state j_{z}=0 of the motion along the z direction, so that essentially represents the 2D phasespacedensity associated to this single transverse quantum state. Then for , we know from equation (1) and the associated discussion that a broad component arises in G_{1} with a characteristic length that increases exponentially with the phasespacedensity. The observed onset of extended coherence around can be understood as the place where starts to exceed significantly λ_{T}. The regime around is reminiscent of the presuperfluid state identified in refs 15, 16. It is different from the truly superfluid phase, which is expected at a higher phasespacedensity () for our parameters^{39}. Therefore, the threshold is not associated to a true phase transition, but to a crossover where the spatial coherence of the gas increases rapidly with the control parameter N.
For ν_{z}=365 Hz, the gas is in the ‘unfrozen regime’ (), which could be naively thought as irrelevant for 2D physics since according to Boltzmann statistics, many vibrational states along z should be significantly populated. However, thanks to the BEC_{⊥} phenomenon presented above, a macroscopic fraction of the atoms can accumulate in the j_{z}=0 state. This happens when the total phasespacedensity exceeds the threshold for BEC_{⊥} (cf. Supplementary Note 1 and Supplementary Fig. 1):
In the limit ζ→∞, BEC_{⊥} corresponds to a phase transition of the same nature as the ideal gas BEC in 3D (see Supplementary Note 1 and Supplementary Fig. 2). In the present context of our work, we emphasize that although BEC_{⊥} originates from the saturation of the occupation of the excited states along z, it also affects the coherence properties of the gas in the xy plane. In particular when rises from 0 to , the coherence length in xy increases from ~λ_{T} (the nondegenerate result) to ~a_{z}, the size of the ground state of the z motion (see Supplementary Note 1, Supplementary Fig. 3 and Supplementary Table 1). This increase can be interpreted by noting that when BEC_{⊥} occurs (equation 4), the 3D spatial density in the central plane (z=0) is equal to , where g_{s} is the polylogarithm of order s and g_{3/2}(1)≈2.612 (see Supplementary Note 1 and Supplementary Fig. 4). For an infinite uniform 3D Bose gas with this density, a true BEC occurs and the coherence length diverges. Because of the confinement along the z direction, such a divergence cannot occur in the present quasi2D case. Instead, the coherence length along z is by essence limited to the size a_{z} of the j_{z}=0 state. When the same limitation applies in the transverse plane, giving rise to coherence volumes that are grossly speaking isotropic. When is increased further, the coherence length in the xy plane increases, while remaining limited to a_{z} along the z direction. The results shown in Fig. 4 are in line with this reasoning. For , the emergence of coherence in the xy plane occurs for a total phasespacedensity , with a proportionality coefficient α=1.4(3) in good agreement with the prediction π^{2}/6≈1.6 of equation (4).
We have also plotted in Fig. 4 contour lines characterizing the coherence range in terms of ζ and . Using ideal Bose gas theory, we calculated the onebody coherence function G_{1}(r) and determined the distance r_{f} over which it decreases by a given factor f with respect to G_{1}(0). We chose the value f=20 to explore the long tail that develops in G_{1} when phase coherence emerges. The contour lines shown in Fig. 4 correspond to given values of r_{20}/λ_{T}; they should not be considered as fits to the data, but as an indication of a coherence significantly larger than the one obtained from Boltzmann statistics (for which r_{20}≈λ_{T}). The fact that the threshold phasespace densities follow quite accurately these contour lines validates the choice of tools (nonGaussian velocity distributions, matter–wave interferences) to characterize the onset of coherence.
Observation of topological defects
From now on we use the weak trap along z (ν_{z}=365 Hz) so that the onset of extended coherence is obtained thanks to the transverse condensation phenomenon. We are interested in the regime of strongly degenerate, interacting gases, which is obtained by pushing the evaporation down to a point where the residual thermal energy k_{B}T becomes lower than the chemical potential μ (see Methods for the calculation of μ in this regime). The final box potential is ~k_{B} × 40 nK, leading to an estimated temperature of ~10 nK, whereas the final density (~50 μm^{−2}) leads to μ≈k_{B} × 14 nK. In these conditions, for most realizations of the experiment, defects are present in the gas. They appear as randomly located density holes after a short 3D ToF (Fig. 5a,b), with a number fluctuating between 0 and 5. To identify the nature of these defects, we have performed a statistical analysis of their size and contrast, as a function of their location and of the ToF duration τ (Fig. 5c,d). For a given τ, all observed holes have similar sizes and contrasts. The core size increases approximately linearly with τ, with a nearly 100% contrast. This favours the interpretation of these density holes as single vortices, for which the 2π phase winding around the core provides a topological protection during the ToF. This would be the case neither for vortex–antivortex pairs nor phonons, for which one would expect large fluctuations in the defect sizes and lower contrasts.
Dynamical origin of the topological defects
In principle, the vortices observed in the gas could be due to steadystate thermal fluctuations. BKT theory indeed predicts that vortices should be present in an interacting 2D Bose gas around the superfluid transition point^{11}. Such ‘thermal’ vortices have been observed in nonhomogeneous atomic gases, either interferometrically^{14} or as density holes in the trap region corresponding to the critical region^{20}. However, for the large and uniform phasespace densities that we obtain at the end of the cooling process (), ref. 44 predicts a vanishingly small probability of occurrence for such thermal excitations. This supports a dynamical origin for the observed defects.
To investigate further this interpretation, we can vary the two times that characterize the evolution of the gas, the duration of evaporation t_{evap} and the hold duration after evaporation t_{hold} (see Fig. 1a). For the results presented in this section, we fixed t_{hold}=500 ms and studied the evolution of the average vortex number N_{v} as a function of t_{evap}. The corresponding data, given in Fig. 6a, show a decrease of N_{v} with t_{evap}, passing from N_{v}≈1 for t_{evap}=50 ms to N_{v}≈0.3 for t_{evap}=250 ms. For longer evaporation times, N_{v} remains approximately constant around 0.35(5).
The decrease of N_{v} with t_{evap} suggests that the observed vortices are nucleated via a KZ type mechanism^{22,23,45}, occurring when the transition to the phase coherent regime is crossed. However, applying the KZ formalism to our setup is not straightforward. In a weakly interacting, homogeneous 3D Bose gas, BEC occurs when the 3D phasespacedensity reaches the critical value g_{3/2}(1). For our quasi2D geometry, transverse condensation occurs when the 3D phasespacedensity in the central plane z=0 reaches this value. At the transition point, the KZ formalism relates the size of phasecoherent domains to the cooling speed . For fast cooling, KZ theory predicts domain sizes for a 3D fluid that are smaller than or comparable to the thickness a_{z} of the lowest vibrational state along z; it can thus provide a good description of our system. For a slower cooling, coherent domains much larger than a_{z} would be expected in 3D at the transition point. The 2D nature of our gas leads in this case to a reduction of the inplane correlation length. In the slow cooling regime, we thus expect to find an excess of topological defects with respect to the KZ prediction for standard 3D BEC.
More explicitly we expect for fast cooling, hence short t_{evap}, a powerlaw decay with an exponent d given by the KZ formalism for 3D BEC. The fit of this function to the measured variation of N_{v} for t_{evap}≤250 ms leads to d=0.69(17) (see Fig. 6a). This is in good agreement with the prediction d=2/3 obtained from the critical exponents of the socalled ‘F model’^{21}, which is believed to describe the universality class of the 3D BEC phenomenon. For comparison, the prediction for a pure meanfield transition, d=1/2, is notably lower than our result.
For longer t_{evap}, the above described excess of vortices due to the quasi2D geometry should translate in a weakening of the decrease of N_{v} with t_{evap}. The nonzero plateau observed in Fig. 6a for t_{evap}≥250 ms may be the signature of such a weakening. Other mechanisms could also play a role in the nucleation of vortices for slow cooling. For example, because of the box potential residual rugosity, the gas could condense into several independent patches of fixed geometry, which would merge later during the evaporation ramp and stochastically form vortices with a constant probability.
Lifetime of the topological defects
The variation of the number of vortices N_{v} with the hold time t_{hold} allows one to study the fate of vortices that have been nucleated during the evaporation. We show in Fig. 6b the results obtained when fixing the evaporation to a short value t_{evap}=50 ms. We observe a decay of N_{v} with the hold time, from N_{v}=2.3 initially to 0.3 at long t_{hold} (2 s). To interpret this decay, we modelled the dynamics of the vortices in the gas with two ingredients: first the conservative motion of a vortex in the velocity field created by the other vortices, including the vortex images from the boundaries of the box potential^{46}; second the dissipation induced by the scattering of thermal excitations by the vortices, which we describe phenomenologically by a friction force that is proportional to the nonsuperfluid fraction of atoms in the gas^{47}. During this motion, a vortex annihilates when it reaches the edge of the trap or encounters another vortex of opposite charge. The numerical solution of this model leads to a nonexponential decay of the average number of vortices, with details that depend on the initial number of vortices and their locations.
Assuming a uniform random distribution of vortices at the end of the evaporation, we have compared the predictions of this model to our data. It gives the following values of the two adjustable parameters of the model, the initial number of vortices N_{v,0}=2.5(2) and the superfluid fraction 0.94(2); the corresponding prediction is plotted as a continuous line in Fig. 6b. We note that at short t_{hold}, the images of the clouds are quite fuzzy, probably because of nonthermal phononic excitations produced (in addition to vortices) by the evaporation ramp. The difficulty to precisely count vortices in this case leads to fluctuations of N_{v} at short t_{hold} as visible in Fig. 6b. The choice t_{hold}=500 ms in Fig. 6a was made accordingly.
The finite lifetime of the vortices in our sample points to a general issue that one faces in the experimental studies on the KZ mechanism. In principle, the KZ formalism gives a prediction on the state of the system just after crossing the critical point. Experimentally we observe the system at a later stage, at a moment when the various domains have merged, and we detect the topological defects formed from this merging. In spite of their robustness, the number of vortices is not strictly conserved after the crossing of the transition and its decrease depends on their initial positions. A precise comparison between our results and KZ theory should take this evolution into account, for example, using stochastic meanfield methods^{48,49,50,51}.
Discussion
Using a boxlike potential created by light, we developed a setup that allowed us to investigate the quantum properties of atomic gases in a uniform quasi2D configuration. Thanks to the precise control of atom number and temperature, we characterized the regime for which phase coherence emerges in the fluid. The uniform character of the gas allowed us to disentangle the effects of ideal gas statistics for inplane motion, the notion of transverse condensation along the strongly confined direction and the role of interactions. This is to be contrasted with previous studies that were performed in the presence of a harmonic confinement in the plane, where these different phenomena could be simultaneously present in the nonhomogeneous atomic cloud.
For the case of a weakly interacting gas considered here, our observations highlight the importance of Bose statistics in the emergence of extended phase coherence. This coherence is already significant for phasespace densities , well below the value required either for the superfluid BKT transition or for the full BEC in the ground state of the box (see Supplementary Note 2 and Supplementary Fig. 5). For our parameters, the latter transitions are expected around the same phasespacedensity (~8–10) meaning that when the superfluid criterion is met, the coherence length set by Bose statistics is comparable to the box size.
By cooling the gas further, we entered the regime where interactions dominate over thermal fluctuations. This allowed us to visualize with a very good contrast the topological defects (vortices) that are created during the formation of the macroscopic matterwave, as a result of a KZ type mechanism. Here we focused on the relation between the vortex number and the cooling rate. Further investigations could include correlation studies on vortex positions, which can shed light on their nucleation process and their subsequent evolution^{52}.
Our work motivates future research in the direction of strongly interacting 2D gases^{19}, for which the order of the various transitions could be interchanged. In particular, the critical for the BKT transition should decrease, and reach ultimately the universal value of the ‘superfluid jump’, (ref. 53). In this case, the emergence of extended coherence in the 2D gas would be essentially driven by the interactions. Indeed, once the superfluid transition is crossed, the onebody correlation function is expected to decay very slowly, G_{1}(r) ∝ r^{−α}, with α<1/4. It would be interesting to revisit the statistics of formation of quenchinduced topological defects in this case, for which significant deviations to the KZ powerlaw scaling have been predicted^{54,55}.
Methods
Characterization of the boxlike potential
We create the boxlike potential in the xy plane using a laser beam that is bluedetuned with respect to the ^{87}Rb resonance. At the position of the atomic sample, we image a dark mask placed on the path of the laser beam. This mask is realized by a metallic deposit on a wedged, antireflectivecoated glass plate. We characterize the boxlike character of the resulting trap in two ways. First, the flatness of the domain where the atoms are confined is characterized by the root mean square intensity fluctuations of the inner dark region of the beam profile. The resulting variations of the dipolar potential are δU/U_{box}~3%, where U_{box} is the potential height on the edges of the box. The ratio δU/k_{B}T varies from ~40% at the loading temperature to ~10% at the end of the evaporative cooling (ramp of U_{box}, see below). In particular, it is of ~20% at the transverse condensation point for the configuration in which the vortex data have been acquired. Second, the sharp spatial variation of the potential at the edges of the boxlike trapping region is characterized by the exponent α of a powerlaw fit U(r) ∝ r^{α} along with a radial cut. We restrict the fitting domain to the central region where U(r)<U_{box}/4 and find α~10–15, depending on the size and the shape of the box.
Imaging of the atomic density distribution
We measure the atomic density distribution in the xy plane using resonant absorption imaging along z. We use two complementary values for the probe beam intensity I. First, we use a conventional lowintensity technique with I/I_{sat}≈0.7, where I_{sat} is the saturation intensity of the Rb resonance line, with a probe pulse duration of 20 μs. This procedure enables a reliable detection of lowdensity atomic clouds, but it is unfaithful for highdensity ones, especially in the 2D geometry because of multiple scattering effects between neighbouring atoms^{56}. We thus complement it by a highintensity technique inspired from ref. 57, in which we apply a short pulse of 4 μs of an intense probe beam with I/I_{sat}≈40. ToF bimodality measurements (where the cloud is essentially 3D at the moment of detection) were performed with the lowintensity procedure. This was also the case for the matter–wave interference measurements, for which we reached a better fringe visibility in this case. In situ images in Fig. 1 and all data related to vortices (for example, Fig. 5a,b) in the strongly degenerate gases were taken with highintensity imaging. We estimate the uncertainty on the atom number to be of 20%.
Ideal gas description of trapped atomic samples
We consider a gas of N noninteracting bosonic particles confined in a square box of size L in the xy plane, and in a harmonic potential well of frequency ν_{z} along z. The eigenstates of the singleparticle Hamiltonian are labelled by three integers j_{x}, j_{y}≥1, j_{z}≥0:
where a_{z}=(h/mν_{z})^{1/2}/(2π) and χ_{j} is the normalized jth Hermite function. Their energies and occupation factors are
where μ<0 is the chemical potential of the gas and N=∑_{j}n_{j}. The average value of any onebody observable can then be calculated:
Interaction energy estimate for weakly interacting gases
We estimate the local value of the interaction energy per particle , where a=5.1 nm is the 3D scattering length characterizing swave interactions for ^{87}Rb atoms and ρ^{(3D)}(r) the spatial 3D density estimated using the ideal gas description. It is maximal at trap centre r=0. For example, using a typical experimental condition with N=40,000 atoms in a square box of size L=24 μm at T=200 nK, we find a maximal 3D density of ρ^{(3D)}(0)=13.8 μm^{−3}. The meanfield interaction energy for an atom localized at the centre of cloud is then . We note that is negligible compared with k_{B}T and hν_{z} for all atomic configurations corresponding to the onset of an extended phase coherence. In this case, the interactions play a negligible role in the 2D ToF expansion that we use to reveal matter–wave interferences.
Temperature calibration
All temperatures indicated in the paper are deduced from the value of the box potential, assuming that the evaporation barrier provided by U_{box} sets the thermal equilibrium state of the gas. This hypothesis was tested, and the relation between T and U_{box} calibrated, using atomic assemblies with a negligible interaction energy. For these assemblies, we compared the variance of their velocity distribution Δv^{2} obtained from a ToF measurement to the prediction of equation (8). The calibration obtained from this set of measurements can be empirically written as
where the values of the dimensionless parameter η and of the reference temperature T_{0} slightly depend on the precise shape of the trap. For the square trap of side 24 μm, we obtain T_{0}=191(6) nK and η=3.5(3). The reason for which T saturates when the box potential increases to infinity is due to the residual evaporation along the vertical direction, above the barrier created by the horizontal Hermite–Gauss beam.
Power exponent for fitting N_{c}
We estimate the behaviour of Δ(N) at fixed T using Bose law for a ideal gas. We compute from equation (8) the equilibrium velocity distribution Then we estimate the spatial density after a ToF of duration τ (for a disk trap of radius R) via , where stands for the convolution operator and Θ for the Heaviside function. We fit ρ(r) to a double Gaussian and compute the atom fraction in the sharpest Gaussian Δ, similar to the processing of experimental data. To simulate our experimental results, we consider ν_{z}=350 Hz, R=12 μm, τ=14 ms and T varying from 100 to 250 nK. For a given T, we record Δ while varying the total atom number N from 0.06 to 4 times the theoretical critical number for (see Supplementary Note 1). We fit Δ(N) between N_{min}=0.06 N_{c,th} and a varying N_{max} in 1.1–4 N_{c,th}, to f(N)=(1−(N_{c}/N)^{α}) with N_{c} as a free parameter and a fixed α. For all considered T and N_{max}, choosing α=0.6 provides both a good estimate of N_{c} (between 0.93 and 0.99 N_{c,th}) and a satisfactory fit (average coefficient of determination 0.94).
Chemical potential in the degenerate interacting regime
To compute the chemical potential μ of highly degenerate interacting gases, we perform a T=0 meanfield analysis. We solve numerically the 3D Gross–Pitaevskii equation in imaginary time using a splitstep method, and we obtain the macroscopic groundstate wavefunction ψ(r). Then we calculate the different energy contributions at T=0—namely the potential energy E_{pot}, the kinetic energy E_{kin} and interaction energy E_{int}—by integrating over space:
with ω_{z}=2πν_{z}. We obtain the value of the chemical potential μ by taking the derivative of the total energy with respect to N and subtracting the singleparticle groundstate energy:
In the numerical calculation, we typically use time steps of 10^{−4} ms and compute the evolution for 10 ms. The 3D grid contains 152 × 152 × 32 voxels, with a voxel size 0.52 × 0.52 × 0.26 μm^{3}.
Analysis of the density holes created by the vortices
We first calculate the normalized density profile , where the average is taken over the set of images with the same ToF duration τ. Then we look for density minima with a significant contrast and size. Finally, for each significant density hole, we select a square region centred on it with a size that is approximately three times larger than the average hole size for this τ. In this region, we fit the function
to the normalized density profile, where A_{0} accounts for density fluctuations. We also correct for imaging imperfections (finite imaging resolution and finite depth of field) by performing a convolution of the function defined in equation (14) by a Gaussian of width 1 μm, which we determined from a preliminary analysis.
Additional information
How to cite this article: Chomaz, L. et al. Emergence of coherence via transverse condensation in a uniform quasitwodimensional Bose gas. Nat. Commun. 6:6162 doi: 10.1038/ncomms7162 (2015).
References
 1
Pethick, C. & Smith, H. BoseEinstein Condensation in Dilute Gases Cambridge Univ. (2002).
 2
Pitaevskii, L. & Stringari, S. BoseEinstein Condensation Oxford Univ. (2003).
 3
Leggett, A. J. Quantum Liquids Oxford Univ. (2006).
 4
Gaunt, A. L., Schmidutz, T. F., Gotlibovych, I., Smith, R. P. & Hadzibabic, Z. Bose–Einstein condensation of atoms in a uniform potential. Phys. Rev. Lett. 110, 200406 (2013).
 5
Gotlibovych, I. et al. Observing properties ofan interacting homogeneous BoseEinstein condensate: Heisenberglimited momentum spread, interaction energy, and freeexpansion dynamics. Phys. Rev. A 89, 061604 (2014).
 6
Huang, K. Statistical Mechanics Wiley (1987).
 7
van Druten, N. J. & Ketterle, W. Twostep condensation of the ideal Bose gas in highly anisotropic traps. Phys. Rev. Lett. 79, 549–552 (1997).
 8
Armijo, J., Jacqmin, T., Kheruntsyan, K. & Bouchoule, I. Mapping out the quasicondensate transition through the dimensional crossover from one to three dimensions. Phys. Rev. A 83, 021605 (2011).
 9
RuGway, W. et al. Observation of transverse BoseEinstein Condensation via Hanbury BrownTwiss Correlations. Phys. Rev. Lett. 111, 093601 (2013).
 10
Berezinskii, V. L. Destruction of longrange order in onedimensional and twodimensional system possessing a continuous symmetry group  II. quantum systems. Soviet Phys. JETP 34, 610 (1971).
 11
Kosterlitz, J. M. & Thouless, D. J. Ordering, metastability and phase transitions in twodimensional systems. J. Phys. C Solid State Phys. 6, 1181 (1973).
 12
Mermin, N. D. & Wagner, H. Absence of ferromagnetism or antiferromagnetism in one or twodimensional isotropic Heisenberg models. Phys. Rev. Lett. 17, 1133 (1966).
 13
Hohenberg, P. C. Existence of longrange order in one and two dimensions. Phys. Rev 158, 383 (1967).
 14
Hadzibabic, Z., Krüger, P., Cheneau, M., Battelier, B. & Dalibard, J. BerezinskiiKosterlitzThouless crossover in a trapped atomic gas. Nature 441, 1118–1121 (2006).
 15
Cladeé, P., Ryu, C., Ramanathan, A., Helmerson, K. & Phillips, W. D. Observation of a 2D Bose gas: from thermal to quasicondensate to superfluid. Phys. Rev. Lett. 102, 170401 (2009).
 16
Tung, S., Lamporesi, G., Lobser, D., Xia, L. & Cornell, E. A. Observation of the presuperfluid regime in a twodimensional Bose gas. Phys. Rev. Lett. 105, 230408 (2010).
 17
Hung, C.L., Zhang, X., Gemelke, N. & Chin, C. Observation of scale invariance and universality in twodimensional Bose gases. Nature 470, 236 (2011).
 18
Desbuquois, R. et al. Superfluid behaviour of a twodimensional Bose gas. Nat. Phys. 8, 645–648 (2012).
 19
Ha, L.C. et al. Strongly interacting twodimensional Bose gases. Phys. Rev. Lett. 110, 145302 (2013).
 20
Choi, J.y., Seo, S. W. & Shin, Y.i. Observation of thermally activated vortex pairs in a quasi2D Bose gas. Phys. Rev. Lett. 110, 175302 (2013).
 21
Hohenberg, P. C. & Halperin, B. I. Theory of dynamic critical phenomena. Rev. Modern Phys. 49, 435 (1977).
 22
Kibble, T. W. Topology of cosmic domains and strings. J. Phys. A Math. Gen. 9, 1387 (1976).
 23
Zurek, W. Cosmological experiments in superfluid helium? Nature 317, 505–508 (1985).
 24
Chuang, I., Durrer, R., Turok, N. & Yurke, B. Cosmology in the laboratory: defect dynamics in liquid crystals. Science 251, 1336–1342 (1991).
 25
Ruutu, V. et al. Vortex formation in neutronirradiated superfluid 3He as an analogue of cosmological defect formation. Nature 382, 334–336 (1996).
 26
Bauerle, C., Bunkov, Y. M., Fisher, S., Godfrin, H. & Pickett, G. Laboratory simulation of cosmic string formation in the early Universe using superfluid 3 He. Nature 382, 332–334 (1996).
 27
Ulm, S. et al. Observation of the KibbleZurek scaling law for defect formation in ion crystals. Nat. Commun. 4, 2290 (2013).
 28
Pyka, K. et al. Topological defect formation and spontaneous symmetry breaking in ion Coulomb crystals. Nat. Commun. 4, 2291 (2013).
 29
Monaco, R., Mygind, J., Rivers, R. & Koshelets, V. Spontaneous fluxoid formation in superconducting loops. Phys. Rev. B 80, 180501 (2009).
 30
Sadler, L. E., Higbie, J. M., Leslie, S. R., Vengalattore, M. & StamperKurn, D. M. Spontaneous symmetry breaking in a quenched ferromagnetic spinor BoseEinstein condensate. Nature 443, 312–315 (2006).
 31
Weiler, C. N. et al. Spontaneous vortices in the formation of BoseEinstein condensates. Nature 455, 948–951 (2008).
 32
Chen, D., White, M., Borries, C. & DeMarco, B. Quantum quench of an atomic mott insulator. Phys. Rev. Lett. 106, 235304 (2011).
 33
Lamporesi, G., Donadello, S., Serafini, S., Dalfovo, F. & Ferrari, G. Spontaneous creation of KibbleZurek solitons in a BoseEinstein condensate. Nat. Phys. 9, 656–660 (2013).
 34
Braun, S. et al. Emergence of coherence and the dynamics of quantum phase transitions. Preprint at http://arxiv.org/abs/1403.7199 (2014).
 35
Corman, L. et al. Quenchinduced supercurrents in an annular Bose gas. Phys. Rev. Lett. 113, 135302 (2014).
 36
Navon, N., Gaunt, A. L, Smith, R. P. & Hadzibabic, Z. Critical dynamics of spontaneous symmetry breaking in a homogeneous Bose gas. Preprint at http://arxiv.org/abs/1410.8487 (2014).
 37
Hadzibabic, Z. & Dalibard, J. Twodimensional Bose fluids: An atomic physics perspective. Rivista del Nuovo Cimento 34, 389 (2011).
 38
Petrov, D. S., Holzmann, M. & Shlyapnikov, G. V. BoseEinstein condensation in quasi2D trapped gases. Phys. Rev. Lett. 84, 2551 (2000).
 39
Prokof'ev, N. V. & Svistunov, B. V. Twodimensional weakly interacting Bose gas in the fluctuation region. Phys. Rev. A 66, 043608 (2002).
 40
Andrews, M. R. et al. Observation ofinterference between two Bose condensates. Science 275, 637 (1997).
 41
Hofferberth, S., Lesanovsky, I., Fischer, B., Schumm, T. & Schmiedmayer, J. Nonequilibrium coherence dynamics in onedimensional Bose gases. Nature 449, 324–327 (2007).
 42
Kohstall, C. et al. Observation of interference between two molecular BoseEinstein condensates. New J. Phys. 13, 065027 (2011).
 43
Polkovnikov, A., Altman, E. & Demler, E. Interference between independent fluctuating condensates. Proc. Natl Acad. Sci. USA 103, 6125 (2006).
 44
Giorgetti, L., Carusotto, I. & Castin, Y. Semiclassical field method for the equilibrium Bose gas and application to thermal vortices in two dimensions. Phys. Rev. A 76, 013613 (2007).
 45
Anglin, J. R. & Zurek, W. H. Vortices in the wake of rapid BoseEinstein condensation". Phys. Rev. Lett. 83, 1707 (1999).
 46
Donnelly, R. J. Quantized Vortices in Helium II Cambridge Univ. (1991).
 47
Fedichev, P. O. & Shlyapnikov, G. V. Dissipative dynamics of a vortex state in a trapped Bosecondensed gas. Phys. Rev. A 60, R1779–R1782 (1999).
 48
Bisset, R. N., Davis, M. J., Simula, T. P. & Blakie, P. B. Quasicondensation and coherence in the quasitwodimensional trapped Bose gas. Phys. Rev. A 79, 033626 (2009).
 49
Mathey, L. & Polkovnikov, A. Light cone dynamics and reverse KibbleZurek mechanism in twodimensional superfluids following a quantum quench. Phys. Rev. A 81, 033605 (2010).
 50
Das, A., Sabbatini, J. & Zurek, W. H. Winding up superfluid in a torus via Bose Einstein condensation. Sci. Rep. 2, 352 (2012).
 51
Cockburn, S. P. & Proukakis, N. P. Ab initio methods for finitetemperature twodimensional Bose gases. Phys. Rev. A 86, 033610 (2012).
 52
Freilich, D. V., Bianchi, D. M., Kaufman, A. M., Langin, T. K. & Hall, D. S. Realtime dynamics of single vortex lines and vortex dipoles in a Bose–Einstein condensate. Science 329, 1182–1185 (2010).
 53
Nelson, D. R. & Kosterlitz, J. M. Universal jump in the superfluid density of twodimensional superfluids. Phys. Rev. Lett. 39, 1201 (1977).
 54
Jelicé, A. & Cugliandolo, L. F. Quench dynamics of the 2d XY model. J. Stat. Mech. Theory Exp. 2011, P02032 (2011).
 55
Dziarmaga, J. & Zurek, W. H. Quench in 1D BoseHubbard model: topological defects and excitations from KosterlitzThouless phase transition dynamics. Sci. Rep 4, 5950 (2014).
 56
Chomaz, L., Corman, L., Yefsah, T., Desbuquois, R. & Dalibard, J. Absorption imaging of a quasitwodimensional gas: a multiple scattering analysis. New J. Phys. 14, 055001 (2012).
 57
Reinaudi, G., Lahaye, T., Wang, Z. & GueryOdelin, D. Strong saturation absorption imaging of dense clouds ofultracold atoms. Opt. Lett. 32, 3143 (2007).
 58
Ketterle, W. & Van Druten, N. Evaporative cooling of trapped atoms. Adv. Atomic Mol. Ppt. Phys. 37, 181–236 (1996).
Acknowledgements
We thank J. Palomo and D. Perconte for the realization of the intensity masks and Z. Hadzibabic for several useful discussions. This work is supported by the IFRAF, ANR (ANR12 247 BLANAGAFON), ERC (Synergy grant UQUAM) and the Excellence Cluster CUI. L.Ch. and L.Co. acknowledge the support from the DGA, and C.W. acknowledges the support from the EU (PIEFGA2011 299731).
Author information
Affiliations
Contributions
L.Ch., L.Co. and T.B. performed the experiment and carried out the preliminary data analysis, with contributions from J.B., S.N. and J.D.; R.D. and C.W. participated in the preparation of the experimental setup. L.Ch. performed the detailed data analysis and their modelling. S.N. developed the simulation on vortex dynamics. L.Ch. and J.D. wrote the manuscript with contributions from all authors.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing financial interests.
Supplementary information
Supplementary Information
Supplementary Figures 15, Supplementary Table 1, Supplementary Notes 12 and Supplementary References. (PDF 519 kb)
Rights and permissions
About this article
Cite this article
Chomaz, L., Corman, L., Bienaimé, T. et al. Emergence of coherence via transverse condensation in a uniform quasitwodimensional Bose gas. Nat Commun 6, 6162 (2015). https://doi.org/10.1038/ncomms7162
Received:
Accepted:
Published:
Further reading

Tan’s twobody contact across the superfluid transition of a planar Bose gas
Nature Communications (2021)

Phase coherence in outofequilibrium supersolid states of ultracold dipolar atoms
Nature Physics (2021)

Berezinskii–Kosterlitz–Thouless phase induced by dissipating quasisolitons
Scientific Reports (2021)

Observation of first and second sound in a BKT superfluid
Nature (2021)

Universal statistics of vortices in a newborn holographic superconductor: beyond the KibbleZurek mechanism
Journal of High Energy Physics (2021)
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.