MightyLev: An acoustic levitator for high-temperature containerless processing of medium- to high-density materials

High-temperature levitation is a valuable technique for processing liquid metals and oxide materials. By removing the need for a container, levitation eliminates contamination, chemical reactions, and heterogeneous nucleation at the sample-container interface. This facilitates the deep supercooling of liquids several 100 s of Kelvin below their melting temperatures, enabling the fabrication of solid glasses and other metastable phases that cannot be formed using conventional furnace methods.¹ The absence of a container is advantageous for in situ spectroscopy or diffraction analysis of liquids, as it enables the accurate measurement of a stably confined sample with minimal contribution from the sample environment.^2–5 

Established techniques for high-temperature containerless processing and analysis of liquid metals and oxides include aerodynamic levitation (ADL),⁴ electromagnetic levitation (EML),⁶ and electrostatic levitation (ESL).⁷ In ADL, the sample is suspended by aerodynamic forces generated by a stream of gas flowing through a diverging conical nozzle. In EML, eddy currents are induced in a sample in response to a magnetic field produced by a radio frequency coil. These currents cause induction heating and generate a Lorentz force, which counteracts gravity. In ESL, levitation is achieved using the Coulomb force acting on charged samples within an electrostatic field between opposed electrodes, with the sample position regulated via high-speed visual feedback control. In both ESL and ADL techniques, the sample is normally heated and melted using infrared (IR) lasers.

EML and ESL have been employed extensively in both ground- and space-based measurements of liquid metals and alloys.^3,6–13 Aerodynamic levitation has been most widely used for the ground-based synthesis and characterization of oxide liquids and glasses.^{2,4,5,14–21} Although these levitation techniques are well established as effective methods for high-temperature synthesis and analysis of materials, they do have limitations; EML requires the sample to be an electrical conductor and the temperature is indirectly controlled using an external gas flow, and ESL requires an electrically charged sample within a vacuum chamber. This limits the materials that can be processed and the conditions that can be achieved using these methods. In ADL, the sample’s proximity (≲1 mm) to the levitation nozzle can cause instabilities and brief contact, resulting in crystallization and a risk of contamination. In addition, the levitation gas is at room temperature as it enters the nozzle, which cools the lower portion of the sample, giving rise to large temperature gradients across the sample.

Mid-air levitation at acoustic standing-wave nodes has been used for many years²³ and is a potential alternative approach for containerless processing of a wide-variety of materials. However, realizing the full potential of acoustic levitation requires the ability to levitate medium density oxide materials (⁠$∼3$–6 g cm⁻³) and high-density metals (up to $∼12$–16 g cm⁻³) at high-temperatures. Until recently, acoustic levitation of materials of medium density (a few g cm⁻³) has required the use of single or opposed pairs of high-powered Langevin horns to generate high acoustic pressures. These systems rely on highly tuned transducers and cavities. During operation, these transducers heat up, causing a change in their resonant frequency and effective height of the cavity, resulting in a loss of the stringent cavity resonance condition required for stable levitation. Practically, this means it is difficult to heat a levitated sample without continuous monitoring and tuning of the levitator.²⁴ Nonetheless, Langevin horn based acoustic levitation systems have been reported to heat samples within external furnace chambers up to $∼1573$ K on Earth^25–27 and up to 1800 K in Earth orbit during the Space Shuttle STS-61A mission for samples of density up to 4.2 g cm⁻³.²⁷ A system of three orthogonal single-axis acoustic levitators in combination with Xenon lamps has been reported to successfully trap and heat samples on Earth at temperatures exceeding 2000 K.²⁸ Aero-acoustic systems, using a gas jet for levitation and Langevin-horn type transducers for stabilization, have been shown to levitate materials with densities up to 6.5 g cm⁻³ at temperatures up to 3500 K using laser heating.^29,30 Despite these successful attempts, high-temperature containerless processing using acoustic levitation remains far from routine. Overcoming the inherent difficulty in maintaining the strict resonance conditions of Langevin-horn type transducers is crucial for the optimization and wider adoption of this method.

Recent advances in acoustic levitation have employed single-axis, non-resonant devices with phased arrays of low-voltage ultrasonic transducers. The “TinyLev” acoustic levitation device consists of 72 × 10 mm diameter 40 kHz ultrasonic piezoelectric “parking sensor” transducers arranged in a focusing condition to produce stable confinement of a variety of liquid- and solid-state materials of moderate density up to $∼3gcm−3$⁠.³¹ As TinyLev does not rely on a cavity resonance, it is insensitive to small changes in temperature and is capable of translating levitated samples by continuous phase adjustments of one bank of transducers. Using a precisely engineered and tuned TinyLev device, Contreras and Marzo³² demonstrated sustained levitation of steel and liquid mercury, although the high excitation voltages required to levitate such high density samples impair their levitation stability. The “BigLev” device is a 160% scaled up version of TinyLev employing larger 16 mm diameter transducers for routine levitation of medium density samples up to 6.5 g cm⁻³.³¹ 

In this paper, we describe the design and performance of “MightyLev,” a new non-resonant multi-transducer acoustic levitator with 16 mm transducers arranged in more closely spaced arrays compared to BigLev, with additional rows of transducers positioned toward the equatorial edges of the hemispherical arrays to enhance lateral stability. MightyLev is capable of highly stable levitation of medium- to high-density materials (up to at least 11.3 g cm⁻³). In combination with CO₂ laser-heating, we demonstrate sustained levitation of materials in MightyLev at temperatures above 1500 K and examine heating induced instabilities using a schlieren imaging system.

MightyLev consists of two opposed 3D printed hemispheres of radius 9.5 cm on which 236 × 16 mm piezoelectric transducers (Manorshi, MSO-A1640H10T) are mounted [Fig. 1(a)]. Each transducer array is subdivided into quadrants, driven by separate L297N Dual H-Bridge motor drivers to amplify a 40 kHz square wave signal output from an Arduino Nano microcontroller. The Ultraino code³³ was used to predict the acoustic pressure field in MightyLev using Huygens’ principle and the acoustic forces using the Gor’kov model.²² The simulated acoustic pressure field for in-phase and anti-phase configurations of the transducer arrays is shown in Figs. 1(b) and 1(c), respectively. The calculated force experienced by a 1.5 mm diameter sphere of mid-ocean ridge basalt (MORB) glass, with density ρ = 2.90 g cm⁻³ and speed of sound v = 6 × 10³ ms⁻¹,³⁴ placed at the central acoustic nodes for the in-phase and anti-phase configurations, is shown in Fig. 1(d) as a function of distance from the center of the particle. The ratio between the peak lateral force, F_x,y, and the peak longitudinal force, F_z, is ∼1:3, as compared to the ∼1:5 ratio calculated for the same sample in the TinyLev and BigLev designs.³¹ This is due to the additional rows of transducers in MightyLev, positioned toward the equatorial edge of the hemispherical arrays and the closer separation of the arrays, which results in a more uniformly distributed trapping force in all directions and improves the lateral stability.

The levitation capability of MightyLev was assessed by systematic levitation of millimeter-scale spheres of different densities in the central acoustic trap and determining the minimum peak-to-peak voltage (V_pp) required for levitation before the sample dropped. The minimum V_pp required for levitation of a range of solid samples is shown in Fig. 2 in comparison to the original TinyLev³¹ and optimized TinyLev variants.³² The response follows a linear trend within the experimental uncertainties. The slight deviations from linearity observed for samples of density 6–8 g cm⁻³ can be attributed to small variations in sample diameter. Since the damping constant in an acoustic trap is proportional to mass, smaller mass samples are less stable than larger mass samples of similar density at the same levitation voltage.

The minimum power consumption required to levitate a sample of the heavy metal Pb (density, ρ ∼ 11.3 g cm⁻³) in MightyLev is $∼20$ W at V_pp ∼ 20 V. For routine stable levitation of Pb, a typical power consumption of $∼30$–40 W is required. This is comparable to the minimum power consumption of $∼40$ W required to levitate Pb in the TinyLev device at V_pp ∼ 38 V.^31,32 However, while the three-dimensional stability of low-density samples (e.g., polystyrene and liquid water) levitated in TinyLev is good, the input voltages required to levitate high-density materials in TinyLev are close to or above the maximum recommended driving voltage of the piezoelectric transducers (typically V_RMS ≤ 30 V). Driving the transducers at high voltages causes heating of the piezoelectric ceramic and a shift in the resonance frequency. High-density samples levitated in TinyLev are visibly unstable, particularly with respect to lateral oscillations.³² In contrast, dense solids, including Pb, can be easily placed in the MightyLev trap for highly stable and sustained levitation at voltages well within the operational specifications of the transducers, with ample scope to increase the driving voltage if required to levitate higher-density materials or for heating experiments. As shown in video V1 of the supplementary material, there is no significant visible movement of a range of medium- to high-density samples levitated in MightyLev.

Levitation stability can be quantified by the mean amplitude of oscillation and its standard error.³⁵ Mean displacements of 19(8) μm (longitudinal) and 189(30) μm (lateral) are observed for a 1.4 mm diameter Pb sphere after equilibration in the central trap of MightyLev. Smaller mean displacements of 7(4) μm (longitudinal) and 41(18) μm (lateral) are observed for a lower density 1.29 mm diameter MORB sphere. The exceptional stability of medium- to high-density materials in MightyLev makes the device highly suitable for chemical and structural analysis using, e.g., micro-focused optical spectroscopy and x-ray scattering techniques.

To further quantify levitation stability and trap strength, high-frame rate (240 fps) digital videos were recorded for $∼1.4$ mm diameter MORB spheres levitated in (i) TinyLev, (ii) a variant of BigLev with reduced array separation, and (iii) MightyLev, during and after perturbation by a metal wire. Additional recordings were made for a larger (2.4 mm diameter) MORB sphere, a 1.4 mm diameter sphere of Pb, and a 3.2 mm diameter stainless steel ball bearing. The videos were recorded against a white background using a Panasonic Lumix GH5 II digital camera equipped with a Laowa 50 mm f/2.8 2X Ultra Macro lens. The image sequences were analyzed frame-by-frame to track the sample position using the connected components analysis function “regionprops” available in the MATLAB Image Processing Toolbox. The centroid of connected components, as computed within a region of interest of a binarized image of the sample, was used to determine the horizontal and vertical position of the sample at each time step.

The longitudinal and lateral damping and spring constants were derived for the different levitator designs using Eqs. (3) and (4) are shown in Fig. 4 as a function of applied electrical power. In general, the damping and spring constants increase linearly with increasing power delivered to the transducers. The values obtained for a small ∼1.4 mm diameter sphere of MORB glass levitated in MightyLev are comparable to those obtained from the BigLev variant. However, steeper gradients are observed in these values for the larger 2.4 mm diameter MORB sphere and the denser 1.4 mm diameter Pb sphere levitated in MightyLev. This is consistent with the direct proportionality of b to m and the dependence of k on both b and m. However, the values obtained for a 3.2 mm diameter stainless steel ball bearing levitated in MightyLev deviate from this trend, indicating the system may be operating at its performance limit or as a result of the Gor’kov criterion, which requires r ≪ λ, where r is the object radius and λ = 8.65 mm is the wavelength of 40 kHz sound waves in the air. For the 3.2 mm ball bearing, the Gor’kov criterion is not strictly satisfied. The ratio of the lateral and longitudinal spring constants k_x/k_z in TinyLev and MightyLev are consistent, within experimental uncertainties, with the theoretical values 0.076 and 0.157, respectively, computed from the acoustic field simulated in Ultraino³³ with no strong dependence on transducer power (Fig. 4).

A class-I laser installation was constructed consisting of two class-IV 60 W mid-IR CO₂ lasers (Synrad Firestar Ti60) mounted on a 2 × 1 m² solid vibration isolation optical table (GZT-MOT-F20-10, Photonic Solutions) on two sides of an interlocked enclosure (Fig. 5). The enclosure was constructed from 2 × 2 cm² aluminum profile and 2 mm thick blackened steel sheets, with three access hatches (two front, one rear) secured individually by tongue operated fail-safe interlock switches (Trojan 5, Allen Bradley). The laser beams, with a fixed beam height of 32 cm from the base of the optical table, are incident into the interlocked enclosure at an angle of 15° from the horizontal axis via aluminum beam tubes and brought to convergence at a common focal point at the center point of the enclosure using 100 mm focal length plano–convex ZnSe lenses (Edmund Optics). IR absorptive graphite beam blocks (Thorlabs LB2/M, 1–12 μm, 80 W Max Avg. Power) are installed inside the enclosure to ensure the direct beam cannot impinge on the walls. During operation, a “Laser On” annunciator illuminates if the interlock condition is satisfied and one or both lasers are in operation. The enclosure is fitted with a feedthrough panel equipped with mains power, DC, USB-3, BNC, and Cat5e Ethernet sockets. MightyLev is mounted on a Genmitsu 1810 CNC router stage, modified for full three-axis translation, with motion controlled externally using a GRBL stepper motor control module. The levitated sample is monitored during heating using two Raspberry Pi HQ camera modules, equipped with Sony IMX477 12.3 Megapixel (4056 × 3040 pixels) image sensors and 16 mm telephoto C-mount lenses, each connected to individual 4 GB Raspberry Pi 4 (RPi-4) single board computers within the enclosure. The internal RPi-4 machines are linked to an external RPi-4 server via a gigabit Ethernet switch (D-Link, DGS-105). Data are shared and stored on the server using a network file sharing (NFS) protocol, with real-time images displayed using VNC remote access software. High frame rate (up to 240 fps) digital videos of the laser-heated levitated materials are recorded using the Panasonic Lumix camera operated remotely via a USB-tether.

Images from the low power laser heating tests of a 1.5 mm polystyrene sphere and gallium sample are shown in Fig. 6 (see also videos V2 and V3 of the supplementary material, respectively). For the polystyrene sample [Fig. 6(a)], each laser was operated at 2.5 W, totaling 5 W. An indentation caused by the incident laser beam is visible on the sample after 4.2 ms. At ∼60 ms, the sample undergoes expansion prior to melting at ∼120 ms (melting point ∼513 K), resulting in the eventual formation of a solid transparent disk of 1 mm diameter and 100 μm in thickness. For the gallium sample [Fig. 6(b)], due to its near room temperature melting point, single-sided laser heating was employed only with a maximum power of$∼4$ W. As the sample underwent melting, it elongated laterally within the acoustic trap, yet remained stably levitated with a slow rotation. On further heating, the molten gallium underwent further lateral elongation before fragmenting into two distinct parts, which were expelled from the levitator.

The results of higher power laser heating tests, conducted using a MORB glass sample, are shown in Fig. 7. The lasers were operated at 6 W each, for a total of 12 W. Figure 7(a) shows a MORB sample during controlled heating, maintaining stable levitation throughout the process. After $∼0.5$ s of heating, the sample began to glow, reaching maximum luminosity ∼1 s later, after which the sample was quenched while remaining levitated (see video V4 of the supplementary material). The results of a subsequent heating experiment are shown in Fig. 7(b). Here, the sample was laser heated for a more extended period, resulting in more pronounced luminosity. A pyrometer reading indicated a temperature surpassing 1200 K, although the significant instability of the sample precluded a precise temperature measurement. After $∼5$ s of heating, the sample appeared elongated laterally, indicating a molten sample, suggesting a temperature about 1500 K from the known melting point of MORB. During heating, lateral and longitudinal heating instabilities intensified, and after $∼10$ s of continuous heating, the sample was ejected from the trap (see video V5 of the supplementary material).

The heating tests reveal significant lateral oscillations induced in levitated samples during heating, even at low laser power. These oscillations increase with both laser power and duration. When levitation fails during heating, samples are typically ejected laterally across the traps rather than vertically. While MightyLev’s enhanced lateral stability helps to mitigate this effect, understanding the origin of the induced lateral oscillations during heating is crucial for the future optimization of high-temperature mid-air acoustic levitation.

To investigate the cause of the instabilities that manifest during heating, a single-mirror on-axis schlieren³⁶ apparatus was constructed to image the acoustic field. The system, illustrated in Fig. 8, consisted of a spherical mirror (Luftvis Science, model D200F1280) of diameter ø = 0.2032(20) m and focal length f = 1.28 m (focal ratio f/6.4) held in a three-point adjustable mount. A bright (107 lm) white LED (Osram Duris E2835) light source was placed behind the two A2 tool steel blades of an adjustable mechanical slit (Thorlabs, VA100), with separation 2.8(1) mm, situated at a distance of 2f = 2.56(1) m along the optical axis of the mirror. The light from the LED source illuminates the mirror through the test region and is back reflected along the optical axis. A 25 mm visible wavelength non-polarizing beam splitter cube (Edmund Optics, A47-009) positioned in front of the light source diverts the reflected light to the imaging camera. The beam splitter was enclosed in a black fabric lined 3D printed case to prevent light external to the imaging axis from entering the optical system. To produce the schlieren effect, a 0.5 mm knife-edge was mounted on a high-precision manual 360° rotation stage and placed at the reflected image point. Light deflected by refractive index gradients in the test region, caused by, e.g., local variations in air density within an acoustic pressure field, can by-pass the knife-edge filter, resulting in a schlieren image. Variations in the brightness of the image can thus be used to infer pressure gradients and the structure of the acoustic field.

The apparatus was tested first using a TinyLev device and IR lamp (Osram, HLX64635) focused on a levitated 1 mm diameter blackened sphere of polystyrene. The TinyLev device was oriented horizontally to direct heat from the transducers out of the device, preventing it from obscuring the image of the acoustic field. The knife edge was oriented vertically to determine the pressure gradients along the axis of the levitator. During heating, the sample is observed to oscillate vertically in its trap. The schlieren image, shown in Fig. 9 (see also video V6 of the supplementary material), reveals a jet of lower density air propagating away from the heated sample and directed along the longitudinal anti-nodes of the acoustic field. This discovery provides a possible explanation for the instabilities observed during laser heating: the interaction between the sample’s reaction to the force of the jet, propagating along the anti-nodes of the acoustic field, and the restoring force of the field generates a parametric oscillation in the direction of acoustic pressure.

To verify this hypothesis, it was necessary to collect schlieren imagery during laser heating. However, due to MightyLev’s size, the position of the mirror, and the need to operate CO₂ lasers in an interlocked enclosure, integrating MightyLev with the schlieren imaging system was not feasible. Instead, a 500 mW red-orange (638 nm) adjustable focus laser (OdicForce, DM-R500) was employed to heat a sample levitated in a TinyLev device, oriented vertically to replicate the geometry employed for laser-heating. A safety box enclosure (Euro Stacker, 600 × 400 × 400 mm³) was placed over the mirror to control reflections and cover the open beam path of the class 3B laser. The laser was incident on the sample, and fast frame rate (240 fps) schlieren images were recorded through an aperture cut into the box enclosure using the Panasonic Lumix GH5 II digital camera. As for the IR lamp test, a jet of hot, lower-density air was observed propagating from the sample along the anti-nodes of the field, producing concomitant lateral oscillations of the sample.

Due to the vertical orientation of the levitator, heat from the transducers impaired the schlieren image quality. To compensate for this, the knife-edge was set horizontally and adjusted to enhance the contrast of the acoustic field. This adjustment resulted in cropping of the schlieren image, producing a circular shadow on the mirror and aberrations around the image of the levitation device. Despite this, the image quality and visibility of refractive index gradients were sufficiently good for quantitative measurements of the sample and jet displacements to be made.

We attempted to mitigate this effect by heating the sample with a laser incident vertically from above. However, lateral instability remained. Consequently, we suggest that dynamic control mechanisms will be necessary to stabilize acoustic levitation during laser heating sufficiently to enable the application of in situ analysis of levitated samples at high-temperatures. Potential strategies include dynamic modulation of the acoustic field or pulse modulation of the heating laser to counteract the parametric oscillation.

“MightyLev” is a new multi-emitter ultrasonic acoustic levitator capable of extremely stable levitation of medium- to high-density materials up to at least 11.3 g cm⁻³. The extremely small (micro-scale) displacements observed during levitation make the device ideally suited for integration with micro-focused optical spectroscopy or x-ray scattering techniques for chemical and structural analysis. MightyLev’s capability to process high-density materials presents opportunities for the safe, contact-free handling and analysis of hazardous heavy-metal materials, such as uranium-containing molten salt coolants for next-generation fission reactors or intermediate-level nuclear waste.

In combination with mid-IR laser heating, we demonstrate stable levitation and melting of a synthetic polymer above 513 K, melting of the liquid metal gallium, and high-temperature cycling and melting of an oxide glass above 1500 K. Instabilities in particle confinement were investigated using schlieren imaging, revealing that parametric oscillations of the sample during heating are driven by hot jets of air propagating from the sample, coupled with the restoring force of the acoustic trap. This provides valuable insight for future optimization of the technique to enable the wider adoption of mid-air acoustic levitation for the synthesis and characterization of novel oxide and metallic glasses using in situ analysis at high-temperatures.

Videos V1–V7 referenced in the main text are provided as the supplementary material. Video captions are provided below. V1: Video demonstrating sustained acoustic levitation of medium- to high-density solids in MightyLev with exceptional stability. V2: Slow motion (240 fps) video showing melting of levitated polystyrene using 5 W of mid-IR laser power. V3: Slow motion (240 fps) video showing melting of levitated gallium using 4 W of mid-IR laser power. V4: Slow motion (240 fps) video showing sustained levitation of a levitated MORB glass sample during laser heating and quenching. V5: Slow motion (240 fps) video showing melting of levitated MORB using 12 W of mid-IR laser power. V6: Schlieren imaging of a heat lamp focused on a levitated polystyrene sample in a TinyLev device, revealing a focused jet of hot air ejected from the trapping node. V7: Schlieren imaging (240 fps) analysis of the sample position and lateral extents of the jet during heating of a levitated polystyrene sphere using a 500 mW 638 nm laser.

We gratefully acknowledge Tom Kennedy, Patrick Alexander, Adrian Crimp, and Joe Nunn for technical assistance. We thank Asier Marzo for helpful discussions and construction of the BigLev variant device. This research was funded by EPSRC standard Grant No. EP/V001736/1.

The authors have no conflicts to disclose.

James W. E. Drewitt: Conceptualization (lead); Formal analysis (lead); Funding acquisition (equal); Investigation (lead); Methodology (lead); Software (lead); Writing – original draft (lead); Writing – review & editing (lead). Barnaby Emmens: Investigation (supporting); Methodology (supporting); Writing – review & editing (supporting). Zhe-Hui Kong: Investigation (supporting); Methodology (supporting); Writing – review & editing (supporting). Bruce W. Drinkwater: Conceptualization (supporting); Funding acquisition (supporting); Methodology (supporting); Writing – review & editing (supporting). Adrian C. Barnes: Conceptualization (supporting); Funding acquisition (equal); Methodology (supporting); Writing – review & editing (supporting).

Data are available at the University of Bristol data repository, data.bris, at https://doi.org/10.5523/bris.6hu7y9f09t12l1goe5ln0vhz.