Magnetic fields are ubiquitous in the Universe. The energy density of these fields is typically comparable to the energy density of the fluid motions of the plasma in which they are embedded, making magnetic fields essential players in the dynamics of the luminous matter. The standard theoretical model for the origin of these strong magnetic fields is through the amplification of tiny seed fields via turbulent dynamo to the level consistent with current observations. However, experimental demonstration of the turbulent dynamo mechanism has remained elusive, since it requires plasma conditions that are extremely hard to re-create in terrestrial laboratories. Here we demonstrate, using laser-produced colliding plasma flows, that turbulence is indeed capable of rapidly amplifying seed fields to near equipartition with the turbulent fluid motions. These results support the notion that turbulent dynamo is a viable mechanism responsible for the observed present-day magnetization.
Diffuse radio-synchrotron emission observations and Faraday rotation measurements1 have revealed magnetic field strengths ranging from a few nG and tens of μG in extragalactic disks, halos and clusters, up to hundreds of TG in magnetars, as inferred from their spin-down2. That turbulence is of central importance in the generation and evolution of magnetic fields in the Universe is widely accepted3. Plasma turbulence can be found in myriads of astrophysical objects, where it is excited by a range of processes: cluster mergers, supernovae explosions, stellar outflows, etc.4,5,6,7. If a turbulent plasma is threaded by a weak magnetic field, the stochastic motions of the fluid will stretch and fold this field, amplifying it until it becomes dynamically significant8,9. According to the current standard picture, the amplification can be summarized in two basic steps10,11. First, when the initial field is small the magnetic energy grows exponentially (kinematic phase). This phase terminates when the magnetic energy reaches approximate equipartition with the kinetic energy at the dissipation scale. Beyond this point, the magnetic energy continues to grow linearly in time (nonlinear phase) until, after roughly one outer-scale eddy-turnover time, it saturates at a fraction of the total kinetic energy of the fluid motions10,12. This is what is referred to as the turbulent dynamo mechanism for magnetic field amplification.
The seed fields that the dynamo amplifies can be produced by a variety of different physical processes. In many astrophysical environments where the plasma is initially unmagnetized, and most certainly at the time when proto-galaxies were forming, baroclinic generation of magnetic fields due to misaligned density and temperature gradients—the Biermann battery mechanism—can provide initial seeds13. The same starting fields also occur in laser-produced plasmas14,15.
Theoretical expectations that turbulent dynamo must operate go back more than half a century8,9,16 and the first direct numerical confirmation of this effect was achieved 35 years ago17. A significant body of theoretical work has developed over the years18,19,20 that has greatly expanded our understanding of the mechanism—for recent reviews of the current state of affairs see refs. 21 and 22. Despite these advancements, demonstrating turbulent dynamo amplification in the laboratory has remained elusive. This is primarily because of the difficulty of achieving experimentally magnetic Reynolds numbers (Rm = u L L/μ, where u L is the flow velocity at the outer scale L, and μ is the magnetic diffusivity) above the critical threshold of a few hundred required for dynamo23. Such a demonstration would not only establish experimentally the soundness of the existing theoretical and numerical expectations for one of the most fundamental physical processes in astrophysics24, but also provide a platform25 to investigate other fundamental processes that require a turbulent magnetized plasma, such as particle acceleration and reconnection.
To date, experimental investigation of magnetic field amplification has primarily been carried out in liquid-metal experiments, such as the von Kármán swirling flow of ref. 26 and the Riga dynamo experiment27,28 inspired by the Ponomarenko dynamo29. The driven dynamos achieved in these experiments depended on a particular fluid flow rather than a purely turbulent effect leading to a stochastic field. More recent work has focused on laser-driven plasmas25,30,31, but studying a regime that is a precursor to dynamo, because of the modest magnetic Reynolds numbers that could be achieved.
In the following, we describe experiments in which we reach magnetic Reynolds numbers above the expected dynamo threshold. We detail the experimental configuration and present diagnostic measurements that fully characterize the plasma state of the magnetized turbulence. By utilizing two independent magnetic field diagnostics we are able to demonstrate the turbulent dynamo amplification of seed magnetic fields to dynamical equipartition with the kinetic energy of the turbulent motions.
The experiments were performed at the Omega laser facility at the Laboratory for Laser Energetics of the University of Rochester32 using a combined platform that builds on our previous work on smaller laser facilities15,30,31. Laser ablation of a chlorine-doped plastic foil launches a plasma flow from its rear surface. The plasma then passes through a solid grid and collides with an opposite moving flow, produced in the same manner. In order to increase the destabilization of the motions as the flows collide, the two grids have hole patterns that are shifted with respect to each other. Further details on the experimental setup are given in Fig. 1. A set of diagnostics has been fielded to measure the properties of the flow, its turbulence and the magnetic field generated by it (see Figs. 2 and 3).
Extensive two-dimensional and three-dimensional simulations done prior to the experiments using the radiation-magnetohydrodynamics (MHD) code FLASH informed their design (see Fig. 2b and Supplementary Methods), including the details of the targets and the grids, and the timing of the diagnostics33,34.
Characterization of the turbulent plasma
X-ray emission can be used to characterize the interaction of the colliding flows and assess properties of the resulting plasma inhomogeneities. The presence of a small amount of chlorine in the plasma enhances the emission in the soft wavelength region (<2 keV). Soft X-ray images taken at t = 35 ns from the start of the laser drive, which is after the flows collide, indicate a broad non-uniform spatial distribution of the emission over a region more than 1 mm across.
In order to characterize the state of the turbulent plasma produced by the collision of two laser-produced jets (as shown in Fig. 1), power spectra of the X-ray intensity fluctuations were extracted from the experimental data using a two-dimensional fast Fourier transform (see Fig. 2a, c). Under the assumption of isotropic statistics, fluctuations in the detected X-ray intensity are directly related to density fluctuations (see discussion in Supplementary Methods). The power spectrum of the density fluctuations extracted from the X-ray data is consistent with a Kolmogorov power law (k−5/3 scaling). Experimental data from other diagnostics indicate that the plasma motions are mainly subsonic (Mach number ≲1 at the outer scale); as a result, density fluctuations injected at large scales behave as a passive scalar and the spectra of the density and velocity fluctuations should be the same35,36. We conclude that the X-ray emission supports the notion that turbulent motions are present in the interaction region. This is also confirmed by FLASH simulations34, which predict subsonic motions of the plasma following the flow collision. Furthermore, the power spectrum of density and velocity fluctuations can be calculated directly from FLASH, and the results are consistent the same power law scaling for both (Fig. 2c, d).
The Thomson scattering diagnostic (see Fig. 1 and Supplementary Methods) allows us to measure simultaneously three different velocities associated with the flow37. First, the bulk plasma flow velocity—composed of a mean flow velocity U and outer-scale turbulent velocity u L —is obtained from the measurements of blueshifts (in frequency) of the scattered light resulting from the bulk plasma moving towards grid B. Second, the separation of the ion-acoustic waves is an accurate measure of the sound speed and thus of the electron temperature, Te. Third, the FLASH prediction of equal ion and electron temperatures allows us to infer from the broadening of the ion-acoustic features the turbulent velocity on the scale μm (the Thomson scattering focal spot)31,38.
Based on these measurements, we find the following. Before the collision, the two plasma flows move towards each other with axial mean velocity in the laboratory rest frame, and have an electron temperature Te ≈ 220 eV (see Supplementary Figure 1). After the collision, the axial flow slows down to 20–40 km s−1, with motions being converted into transverse components. The electron temperature increases considerably, reaching Te ≈ 450 eV (Fig. 3). The measured time-averaged (RMS) turbulent velocity at scale is . If has Kolmogorov scaling, the turbulent velocity at the outer scale must therefore be . Electron density estimates can be obtained from the measured total intensity of the Thomson scattered radiation, to give a value ne ≈ 1020 cm−3, which is also consistent with values predicted by FLASH simulations34. As shown in Supplementary Methods, plasmas with these parameters can be well described as being collisional, and in the resistive MHD regime.
For an MHD-type plasma, we can estimate the characteristic fluid and magnetic Reynolds numbers attained in our experiment. We find Re = u L L/ν ~ 1200 (ν is the viscosity), and Rm~600, using L ~ 600 μm, the characteristic driving scale determined by the average separation between grid openings. We have thus achieved conditions where Rm is comfortably larger than the expected critical magnetic Reynolds number required for turbulent dynamo23. The experiment also lies in the regime where the magnetic Prandtl number is Pm ≡ Rm/Re < 1.
Magnetic field measurements
Magnetic fields were inferred using both Faraday rotation (Fig. 3) and proton radiography (Fig. 4). The rotation of the polarization angle of Thomson scattered light provides a measure of the variation of the longitudinal component of the magnetic field integrated along the beam path, weighted by the electron density. Assuming a random field with correlation length , we estimate B||,rms ≈ 120(Δθ/3°)(ne/1020 cm−3)−1, where B||,rms is the root mean square (RMS) value of the magnetic field component parallel to the probe beam, Δθ is the rotation angle, and is the scale length of the electron density along the line of sight. Estimating is more challenging, but as a reasonable estimate we can take the size of the grid aperture ( μm). The choice of was corroborated by the FLASH simulations via synthetic Faraday rotation measurements (see Supplementary Figure 11 and the discussion in Supplementary Methods).
The appearance of strong, sharp features in proton radiographs provides independent evidence for large magnetic fields39,40. Such structures are the result of initially divergent proton rays produced by an imploding D3He capsule being focused by magnetic forces as they traverse the plasma. This leads, at the detector plane, to localized regions where the proton counts greatly exceed their average value and other regions where they are strongly depleted. The detailed spatial structure of the path-integrated magnetic field can be reconstructed from the experimental images assuming that the protons undergo small deflections as they pass through the plasma, and that the paths of neighboring protons do not cross before reaching the screen. Assuming isotropic statistics, this is sufficient information to calculate the power spectrum of the magnetic energy E B (k), and then the RMS magnetic field strength Brms of fluctuating fields via (see refs. 41,42 and Supplementary Methods). Figure 4a shows that the magnetic field during the early phases of the collision is small, as no strong flux features appear in the radiographic image. The corresponding reconstructed RMS magnetic field strength Brms obtained from Fig. 4d gives .
Magnetic fields before the collision, and in the absence of any strong turbulence, are presumably Biermann battery fields produced at the laser spots and then advected by the flow, as indeed is confirmed by FLASH simulations. In contrast, Fig. 4b, c—corresponding to a later stage of the turbulent plasma’s lifetime for the 10 and 5 ns pulse shapes, respectively—do indeed show strong features, indicative of increased fields strengths and altered morphology. These strong features are absent in single-flow experiments (see Supplementary Figure 8 in Supplementary Methods), suggesting that the interaction of the counter-propagating flows and subsequent development of turbulence is essential for magnetic field amplification. The reconstruction algorithm can also be applied to the images in Fig. 4b, c (see Fig. 4e, f); in the latter case, we obtain Brms ≈ 100 kG (see Supplementary Methods). This is consistent with our previous estimates based on Faraday rotation. We claim that this increase of the magnetic field during the collision cannot be simply explained by the compression of the field lines due to the formation of shocks (this would only account for a factor of two increase at most), nor by further generation by Biermann battery as the temperature gradients are not strong enough. This view is supported by FLASH simulations (see Supplementary Methods). Collisionless (magnetic field-generating) plasma processes such as the Weibel or filamentation instabilities cannot be responsible for the amplification of the magnetic field either, on account of the plasma’s high collisionality (see Supplementary Methods).
In Fig. 4g we show the spectrum of the magnetic energy, E B (k), calculated from the reconstructed path-integrated magnetic field. This is the spectrum on which the estimate of Brms is based. The peak of this spectrum occurs at a wavenumber consistent with the claim that energetically dominant magnetic structures have a size μm. The steeper slope of the spectrum at small wavelengths ( μm) is not a property of the true spectrum, but is due to diffusion of the imaging beam caused by small-scale magnetic fields, and the underestimation of the magnetic energy by the reconstruction algorithm in the presence of small-scale caustics (see Supplementary Methods for a discussion of these effects). In the FLASH simulations the magnetic field spectrum appears to be consistent with a ~ k−1 power-law dependence, as shown in Fig. 2d, in agreement with the spectra of tangled fields near and above the dynamo threshold found in ref. 23 and by other investigators43,44,45, in the Pm < 1 regime.
Our experiment thus indicates that, as the two plasma flows collide, a strongly turbulent plasma, with magnetic Reynolds number above the threshold for dynamo action, is generated. The magnetic field grows from an initial value to ~100–120 kG. We assume this to be near the saturated value because the Faraday rotation measurement begins over 2 ns (comparable to dynamical times) before the proton imaging diagnostic, and we infer similar magnetic field strengths from both. Note that the expected timescale for saturation to be reached is of the order of an outer-scale eddy-turnover time, L/u L ~ 6 ns, a period that is comparable to the time that has elapsed between the initial flow collision and the magnetic field measurements. That the magnetized plasma is in a saturated state is corroborated by the FLASH simulation results (see Supplementary Figure 11 in Supplementary Methods).
If saturation is reached, the magnetic field energy should become comparable to the turbulent kinetic energy at the outer scale. We find (where ρ is the plasma mass density and we have taken Brms ≈ 120 kG). Because the field distribution is expected to be quite intermittent and because Rm in our experiment is unlikely to be asymptotically large compared to the dynamo threshold value, it is reasonable that the mean magnetic energy density is quantitatively smaller than the kinetic energy density10,11,17. However, a good indication that the magnetic field has reached a dynamically saturated state is that it is dynamically strong in the most intense structures, which are not necessarily volume filling. To find an upper experimental bound on the maximum field, Bmax, we assume that the deflections acquired by the imaging protons across the plasma come from an interaction with a single structure. The strongest individual structure in the reconstructed path-integrated image has scale μm with a path-integrated field of 6 kG cm. This gives , which leads to , consistent with dynamical strength.
Our results appear to provide a consistent picture of magnetic field amplification by turbulent motions, in agreement with the longstanding theoretical expectation that turbulent dynamo is the dominant process in achieving dynamical equipartition between kinetic and magnetic energies in high magnetic Reynolds number plasmas found in many astrophysical environments.
Experimental facility and diagnostics
The laser-driven experiments presented in this work were carried out at the Omega laser facility at the Laboratory for Laser Energetics of the University of Rochester32, under the auspices of the National Laser User Facilities (NLUF) program of the U.S. Department of Energy (DOE) National Nuclear Security Administration (NNSA). In order to fully characterize the plasma state and measure the magnetic field amplification we fielded a number of experimental diagnostics, including Thomson scattering, X-ray imaging, Faraday rotation, and proton radiography. Detailed discussions on each diagnostic, along with full descriptions on how the experimental measurements were analyzed, are given in Supplementary Methods of the Supplementary Information.
MHD and numerical simulations
The experimental platform was designed using radiation-MHD simulations with the publicly available, multi-physics code FLASH46,47. The numerical modeling of the platform employed the entire suite of High Energy Density Physics capabilities33,34 of the FLASH code. We performed an extensive series of moderate-fidelity 2D cylindrical FLASH radiation-MHD simulations on the Beagle 2 cluster at the University of Chicago followed by a smaller set of high-fidelity 3D FLASH radiation-MHD simulations on the Mira supercomputer at the Argonne National Laboratory. The simulation campaign is described in detail in a companion paper34 and presented in Supplementary Methods. The applicability of the MHD approximation and considerations regarding the collisionality of the turbulent plasma are discussed in Supplementary Methods of the Supplementary Information.
All data that support the findings of this study are available from the authors upon request.
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
The research leading to these results has received funding from the European Research Council under the European Community’s Seventh Framework Programme (FP7/2007-2013)/ERC grant agreements no. 256973 and 247039, the U.S. Department of Energy under Contract No. B591485 to Lawrence Livermore National Laboratory, Field Work Proposal No. 57789 to Argonne National Laboratory, grants no. DE-NA0002724 and DE-SC0016566 to the University of Chicago, DE-NA0003539 to the Massachusetts Institute of Technology, and Cooperative Agreement DE-NA0001944 to the Laboratory for Laser Energetics University of Rochester. We acknowledge support from the National Science Foundation under grant PHY-1619573. This work was also supported in part by NIH through resources provided by the Computation Institute and the Biological Sciences Division of the University of Chicago and Argonne National Laboratory, under grant S10 RR029030-01. Awards of computer time were provided by the U.S. Department of Energy Innovative and Novel Computational Impact on Theory and Experiment (INCITE) and ASCR Leadership Computing Challenge (ALCC) programs. This research used resources of the Argonne Leadership Computing Facility at Argonne National Laboratory, which is supported by the Office of Science of the U.S. Department of Energy under contract DE-AC02-06CH11357. We acknowledge funding from grants 2007-0093860 and 2016R1A5A1013277 of the National Research Foundation of Korea. Support from AWE plc., the Engineering and Physical Sciences Research Council (grant numbers EP/M022331/1 and EP/N014472/1) and the Science and Technology Facilities Council of the United Kingdom is acknowledged.
Electronic supplementary material
About this article
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.