Photophoretic Levitation of Macroscopic Nanocardboard Plates

Communication

www.advmat.de Photophoretic Levitation of Macroscopic Nanocardboard Plates

John Cortes, Christopher Stanczak, Mohsen Azadi, Maanav Narula, Samuel M. Nicaise, Howard Hu, and Igor Bargatin*

in size, has demonstrated the ability to Scaling down miniature rotorcraft and flapping-wing flyers to sub-centimeter lift off, hover, and move in the horizontal dimensions is challenging due to complex electronics requirements, plane using power tethers or concentrated manufacturing limitations, and the increase in viscous damping at low sunlight.[5,6] Reynolds numbers. Photophoresis, or light-driven fluid flow, was previously Many emerging applications for aerial vehicles, such as search and rescue mis- used to levitate solid particles without any moving parts, but only with sizes sions, jet engine inspection, and even of 1–20 µm. Here, architected metamaterial plates with 50 nm thickness are satellites could benefit from even smaller, leveraged to realize photophoretic levitation at the millimeter to centimeter “smart dust-”scale flyers;[7] however, fur- scales. Instead of creating lift through conventional rotors or wings, the ther decreasing the size of conventional nanocardboard plates levitate due to light-induced thermal transpiration microflyers introduces many challenges. One key challenge is the increasing dif- through microchannels within the plates, enabled by their extremely low ficulty of miniaturizing both mechanical mass and thermal conductivity. At atmospheric pressure, the plates hover structures and electronic components. above a solid substrate at heights of ≈0.5 mm by creating an air cushion Another, more fundamental challenge is beneath the plate. Moreover, at reduced pressures (10–200 Pa), the increased the fact that the propulsion efficiency of speed of thermal transpiration through the plate’s channels creates an rotorcraft and winged flyers decreases as air jet that enables mid-air levitation and allows the plates to carry small the vehicle scale is reduced and Reynolds number decreases, making viscous forces payloads heavier than the plates themselves. The macroscopic metamaterial more dominant.[8] These limitations can structures demonstrate the potential of this new mechanism of flight to be overcome by using alternative methods realize nanotechnology-enabled flying vehicles without any moving parts in of propulsion, such as the photophoretic the Earth’s upper atmosphere and at the surface of other planets. forces described below. Solids illuminated by light while immersed in a gas experience a force variously known in different litera- Miniaturized unmanned aerial vehicles or drones typically rely ture domains as the photophoretic, radiometric, or Knudsen on rotorcraft or flapping wing propulsion, with recent advances force.[9] Different from the light pressure caused by a photon’s made possible by new actuators, processors, batteries, sen- momentum, this force appears when incident light is absorbed sors, and communication electronics.[1-3] The current record by a solid, heating the solid relative to the ambient gas. The holder for the smallest self-powered maneuverable drone is temperature difference can then drive gas flows that result in the Piccolissimo, which is about 28 mm in its largest dimen- a corresponding reaction force acting on the solid. The photo- sion, weighs ≈2.5 g, and uses rotorcraft propulsion.[4] Another phoretic/radiometric/Knudsen force has been studied for over approach is provided by winged robots inspired by insects. The a century, including by luminaries such as Reynolds, Maxwell, RoboBee, which weighs only ≈80 mg and is a few centimeters Einstein, and Knudsen, who primarily focused on explaining a common device called a Crookes radiometer or light mill, which [10,11] Dr. J. Cortes, M. Azadi, M. Narula, Dr. S. M. Nicaise, Prof. H. Hu, rotates when illuminated by sufficiently intense light. The Prof. I. Bargatin force pushing on the rotating centimeter-scale paper vanes of Department of Mechanical Engineering and Applied Mechanics a light mill is typically a few orders of magnitude smaller than University of Pennsylvania the weight of the vanes, which is enough to produce rotation on Philadelphia, PA 19104, USA E-mail: [email protected] a low-friction bearing, but not levitation. However, for microm- C. Stanczak eter-sized particles, the same force can exceed their weight. [12,13] [14] Vagelos Integrated Program in Energy Research Both experimental and theoretical studies have shown University of Pennsylvania the levitation of 1–20 µm diameter spherical particles in mid- Philadelphia, PA 19104, USA air by using light to create a temperature difference and thus The ORCID identification number(s) for the author(s) of this article gas flow across the particle surface. The corresponding relative can be found under https://doi.org/10.1002/adma.201906878. movement of particles and gas is called photophoresis, from DOI: 10.1002/adma.201906878 Greek words for light and motion.

A i ii iii

x = 0 mm x = 5 mm x = 13 mm 4 mm

B C Gas Flow

Absorbing Th CNT Film

D E

400 µm 10 µm

Figure 1. A) Sequential screenshots of a low-height levitation test at atmospheric conditions inside a glass enclosure at an incident optical flux of 10 000 W m−2 (≈10 Sun). B) Schematic of the photophoretic gas flow through one nanocardboard channel. C) Cross-sectional schematic of the optimized nanocardboard structure, showing a cutaway view of the internal channels. D) Scanning electron microscopy (SEM) image of the five-channel-array optimized nanocardboard plate structure. E) SEM image of the 200 nm thick carbon nanotube film on the edge of one of the nanocardboard channels.

Much of the recent research on photophoretic forces focused images.[21] Yet, despite multiple theoretical proposals,[22–24] sus- on the settling of aerosols in the atmosphere[15,16] and led to a tained levitation of larger, macroscale bodies has not yet been better understanding of how pollutants such as volcanic ash, demonstrated because it requires materials with both very low carbon soot, and other greenhouse gases affect atmospheric density and thermal conductivity. This creates an opportunity conditions.[17] Researchers have also modeled the photophoretic for metamaterials specifically engineered for photophoresis to lift forces generated on aerosol nanoparticles as a function realize new mechanisms of flight. of altitude, which could be useful for climate engineering While the previous research on photophoresis focused on efforts.[18,19] Finally, some of the most detailed theoretical the gas flows around solid bodies, such as spheres or plates, models of photophoretic forces have been developed to study photophoretic flows can also occur inside tubes and channels if the effect of photophoresis on planetary disk accretion.[20] On a temperature gradient is present along the inner channel wall, the more applied side, Smalley et al. recently created a 3D dis- as shown in Figure 1B. This phenomenon, known as thermal play by photophoretically trapping and moving 10 µm cellulose transpiration or thermal creep, has been widely studied both particles in free space while continuously illuminating them fundamentally[25–27] and with the goal of creating new types of with red, green, and blue light, providing highly detailed 3D gas pumps without any moving parts.[28,29] As we show below,

thermal transpiration can also be used to pump air through chemically in xenon difluoride (XeF2) gas. The resulting hollow an architected plate to maximize the photophoretic forces and plates of alumina were manually cut with a razor blade to the realize photophoretic hovering and propulsion of macroscopic desired size (typically, ≈1 cm). For more details of the fabrica- structures at both atmospheric and reduced pressures. tion process, see the main text in the Supporting Information. The physical mechanism behind the photophoretic force Figure 1A and Movie S1 in the Supporting Information depends on the Knudsen number, Kn = λ/L, where λ is show our plates hovering above a transparent substrate at equi- the molecular mean free path and L the characteristic length librium heights ranging from 400 to 600 µm for arbitrarily long scale of the flow. For reference, the mean free path for air at periods of time. The levitation was most reliably observed above atmospheric pressure and room temperature is ≈70 nm and a micropatterned low-stiction glass substrate (see the main text scales proportionally to the absolute temperature and inversely in the Supporting Information). We observed the hovering in proportionally to the pressure.[30] In the free-molecular regime over 50 flights with a wide variety of samples. After around (Kn ≫ 1), the gas molecules that collide with the hotter sur- 50 tests, the corners and edges of the sample typically experi- face depart at a larger thermal velocity in comparison to the ence damage that prohibits further trials. The damage typically molecules colliding with the cooler surface. This momentum occurs when the samples land and their corners lock into the exchange results in a reaction force that is proportional to the mesh grid below. number of molecules colliding with the solid per unit time and, The levitation heights were estimated from microscope therefore, proportional to both the active area and the pressure video footage, such as the one shown in Figure 1A, by com- of the ambient gas (Figure S3A, Supporting Information). In paring the distance between the plate and its reflection to the contrast, in the continuum regime (Kn ≪ 1), the ambient gas 500 µm thickness of the substrate. The incident optical flux of will flow along the side walls of the object due to the thermal up to ≈1 W cm−2 was produced by a 100 W LED diode array, creep phenomenon,[18] producing a reaction force that usually as shown in Figure 2B. The light was absorbed by the CNT scales with the perimeter length of the plate and inversely pro- film on the underside of the plate, causing the temperature portionally to the pressure (see the Supporting Information for gradient along the channels and therefore the pumping of air details). As a result, the force is typically maximum at pressures through the channels into the gap below the plate, creating an that correspond to the transition regime (Kn ≈ 1). In all these air cushion just like in a conventional hovercraft. The levitation regimes, the resulting force is directed from the hot side of the mechanism is also similar to the one used in air bearings and device toward the cold side, meaning that the incident light air hockey, where a toy puck is levitated on an air cushion cre- should be primarily absorbed by the bottom side of the solid to ated by gas pumped through a perforated substrate. create photophoretic lift. To predict the lift force as a function of the levitation height, Figure 1 shows a version of our previously reported we developed a theoretical model based on the well-known nanocardboard[31]—a microarchitected metamaterial—that is lubrication theory for thin-film flows under a disk (for example, designed to maximize photophoretic lift force while minimizing in air bearings). Figure 2D shows the lift force, the weight, and the weight. The design features arrays of five channels, with the theoretical equilibrium height, at which the two are equal, in-plane dimensions of 25 × 600 µm and 75 µm gaps for both the single-channel design that we used previously for (Figure 1D), and results in air flowing through the plate at an mechanical tests[31] and the optimized five-channel nanocard- effective “flow-though” velocity on the order of 1–10 mm s−1. board plates. Plates of both designs were successfully levitated, Though these centimeter-scale plates are ultralight, with areal although the observed gaps varied from sample to sample density of ≈1 g m−2, they are robust enough to handle with twee- and over time, with the full ranges shown by colored bands in zers and even by hand without causing permanent damage. Figure 2D. One reason for the larger range of gaps in experi- The design in Figure 1 resulted from a numerical optimization ments is that our model assumes perfectly flat plates whereas, of the ratio of the flow-through velocity to the mass of the plate in reality, the plates curve slightly upward near the edges due (Figure S13, Supporting Information), which maximizes the to differential thermal expansion. Another potential source of ability of the plate to levitate. error is the inconsistency in thickness of the absorbing carbon These ultralight plates were fabricated on a 4 in. silicon nanotube layer, which can vary by as much as 50 nm for a on insulator (SOI) wafer, with silicon device layers varying nominal thickness of 200 nm. This inconsistency results in between 10 and 200 µm. Photolithography was used to pattern both weight and absorption variations that ultimately affect the the device layer, followed by deep reactive-ion etching (DRIE) plate’s ability to generate the gas flow through the plate. to create the channel openings. After the device layer was sep- For highly insulating materials in the continuum regime, arated from the carrier wafer, it was conformally coated with the thermal creep velocity and the flow-through velocities are alumina via atomic layer deposition (ALD). This alumina layer, on the order of I/(10P), where I is the incident light flux and P typically 50 nm thick, coated all exposed surfaces, including the is the ambient pressure.[25,32] Therefore, the volumetric flow rate channel walls. A mat of single-wall carbon nanotubes (CNTs) through our plates can be increased dramatically by reducing was then deposited by drop-casting on one side of the plate and the ambient pressure, resulting not only in hovercraft-like selectively etched in oxygen plasma to form the absorbing layer levitation on an air cushion but also in mid-air flight. Figure 3A (as seen in Figure 1D,E). The ≈200 nm thickness of the CNT and Movie S2 in the Supporting Information show sequential film was chosen to achieve> 80% absorption of incident light screenshots of a plate in mid-air flight at a reduced pressure of (measured using optical power meter on finished plates) while ≈100 Pa inside a glass vacuum chamber. Unlike the hovering adding a relatively small extra mass. The samples were then behavior previously discussed where the gaps underneath the cleaved to expose the encapsulated silicon, and then etched plate were less than 1 mm, these flights featured heights as high

A B Levitang Sample 500 µm thick substrate

Glass Substrate Microscope LED Camera

C D

Gas Flow

Glass Levitaon Substrate Height

Figure 2. A) Microscope camera picture showing a 60 µm thick sample levitating at a small height above a 500 µm thick substrate, which is used as a measuring reference. B) Graphical schematic of the image acquisition setup (not to scale). C) Schematic of the levitation mechanism at atmospheric pressure based on thermal creep flows (not to scale). D) Prediction of the lift force based on the lubrication theory as a function of the gap height. Also shown are equilibrium gap heights (circles) for unoptimized single-channel and optimized five-channel designs levitating above a substrate for the incident optical flux of 1 W cm−2 (≈10 Sun). The colored bands show the range of experimentally observed gap heights.

A t = 0 s t = 3 s Plate Plate

10 mm

B C

Figure 3. A) Sequential screenshots from Movie S2 in the Supporting Information, showing levitation at reduced pressures inside a glass vacuum chamber at an illumination of 0.8 W cm−2 (≈8 Sun). The tested plate has lateral dimensions of ≈0.5 cm × 1 cm. B) The streamlines around a small plate levitating in mid-air. The inset zooms in on the plate to illustrate the flow through the channels. C) Theoretical model prediction of the pressure regimes under which mid-air levitation is possible, based on a comparison of the lift force and weight for a plate measuring 0.5 cm × 1 cm at an illumination of 1 W cm−2 (≈10 Sun). The black arrows show the experimental results, with up arrows corresponding to successful levitation at that pressure and down arrows to no levitation.

Figure 4. A-i–iii) Sequential screenshots of a levitation test consisting of a nanocardboard rectangular plate (6 × 13 mm) carrying a microfabricated silicon payload (shown sitting on top of the plate) at an illumination of ≈10 Sun and air pressure of ≈50 Pa. B) Density plot of the areal density that can levitate as a function of pressure and plate width (assuming rectangular plates with 2:1 aspect ratio) at illumination of 10 Sun. C) Density plot of the maximum payload as a function of pressure and plate side length (assuming nanocardboard plate with areal density of 1 g m−2). The crossed circle shows the pressure and size of the payload experiment shown in panel (A), which had a mass of 0.33 mg in agreement with the theoretical prediction. as 10 mm even though we replaced the solid glass substrate levitation (the glass vacuum chamber we used allowed optical with a very sparse metal mesh (84% open area) to eliminate the access from all sides but did not allow vacuum pressures below air cushion effect. Instead of the air cushion, the lift force was ≈10 Pa). The total time for the levitation was typically less than generated by the jet of air going through the plate at a flow- 1 s as the plates would typically move horizontally and escape through velocity of >1 m s−1 (Figure 3B). As described in the from the light beam created by the LED array. Supporting Information, the lift force can be predicted by calcu- Finally, we explored the payload capability of the nanocard- lating the drag force for thin plates in a certain moving frame board plates at reduced pressure conditions by lifting and of reference. Alternatively, the ability of a plate to levitate can carrying small silicon rings. The typical rings had an inner be predicted by comparing the terminal velocity of a freefalling radius of 950 µm, an outer radius of 1000 µm, and a height plate to the flow-through velocity, i.e., the average velocity of the of 200 µm, corresponding to a mass of ≈0.3 mg. Figure 4A air pumped through the plate by thermal transpiration. shows the plate with the silicon ring on top of it as well as a Figure 3C shows a comparison of the photophoretic lift force set of sequential screenshots from a test flight at a pressure and a plate’s weight as a function of pressure for an optimized of ≈50 Pa. In Figure 4A-ii, the nanocardboard plate lifted off plate measuring 0.5 cm × 1 cm. The lift force exceeds the from the metal mesh, then drifting toward the right side of the weight in a wide pressure range from 0.2 to 400 Pa, which frame (Figure 4A-iii). In this test, the mass of the plate alone agrees with our experimental results for plates of this size and was ≈0.1 mg, while the payload weighed ≈0.33 mg. geometry. We observed no levitation at pressures above 250 Pa, Figure 4B shows theoretical predictions for the maximum while pressures between 8 and 250 Pa resulted in reliable areal density of the plates that can be levitated as a function

Adv. Mater. 2020, 32, 1906878 1906878 (5 of 7) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.advancedsciencenews.com www.advmat.de of their side length and the ambient pressure. At the standard for a new optimized design, which features even more channels sea-level atmospheric pressure (≈100 kPa), even sub-millimeter per unit area, since the absorption is no longer as dependent on plates need to have an areal density of less than 0.1 g m−2, the geometric fill factor. The corresponding increase in the lift which is below the lightest plates we manufactured (≈1 g m−2), force would enable much larger payloads, as well as levitation at explaining why we did not observe mid-air levitation at the the regular sea-level pressure (≈100 kPa). Optical metamaterials atmospheric pressure. However, at reduced pressures, the lift can also be used to facilitate the control of the flight.[41] Future force can exceed the weight of the plate by many orders of research will focus on such improvements in the lift force as magnitude. Moreover, photophoretic propulsion always works well as effective control of the flight. better as the devices get smaller, as higher areal densities can be levitated as plate size is decreased (Figure 4B), in contrast to conventional propulsion methods. Supporting Information Figure 4C shows the theoretical predictions for the payloads that can be levitated by our plates. For example, a 3 cm × 6 cm Supporting Information is available from the Wiley Online Library or plate has the theoretical ability to lift as much as 2 mg at a pres- from the author. sure of ≈10 Pa and an optical flux of 1 W cm−2 (10 Sun). The model’s predictions agree with the experiments we performed, such as the one shown in Figure 4A. Much progress has recently Acknowledgements been reported in creating such lightweight payloads for “smart [33] The authors thank the staff of the Singh Center for Nanotechnology, dust” technology, with multiple demonstrations of electronic Nanoscale Characterization Facility and Scanning and Local Probe systems with cubic-millimeter dimensions and few milligrams facility at the University of Pennsylvania, which are partly funded by the mass, such as fully functional sensing platforms[34,35] and aero- NSF National Nanotechnology Coordinated Infrastructure Program, space applications.[36] Recent research on ultrathin electronics under grant NNCI-1542153. Special thanks to Dr. Hiromichi Yamamoto, and solar cells[37–39] provides another avenue to self-powered Dr. Gerald Lopez, Meredith Metzler, and David Jones. This work ultralight payloads with multiple functionalities. was supported in part by the United States Department of Education Graduate Assistance in Areas of National Need (GAANN) Fellowship The levitation and payload capabilities of our plates peak in (J.C.), the National Science Foundation under CBET-1845933, and the 1–100 Pa range, which is ideal for applications in the upper the School of Engineering and Applied Science at the University of atmosphere of Earth or on Mars. For example, in the Earth’s Pennsylvania. mesosphere, which corresponds to altitudes from 50 to 80 km, the air is too thin for conventional airplanes or balloons but too thick for satellites, so measurements can be performed for only Conflict of Interest a few minutes at a time during the short flight of a research rocket. However, the range of ambient pressures at these The authors declare no conflict of interest. altitudes (1–100 Pa) is nearly optimal for our plates’ payload capabilities. With further optimization, the photophoretically levitated microflyers could stay aloft for extended periods of Author Contributions time by using the Sun’s natural light throughout the daytime, or even indefinitely if the microflyer system is designed to ascend The experimental samples were fabricated by J.C. Experimental testing and characterization was performed by J.C., C.S., and M.A. with during the day and descend the same vertical distance at night. support from S.N. Theoretical models were developed by J.C., I.B., Microflyers based on photophoretic propulsion would enable and H.H. Numerical simulations were performed by C.S. and M.N. The new ways to study the mesosphere using microsensor payloads project was conceived and directed by I.B. The manuscript was written that weigh only a few milligrams. Similarly, the Martian atmos- by J.C. and I.B. with contributions from all authors. phere has a pressure of about 600 Pa at the surface, making photophoretic microflyers an excellent fit for exploring Mars and other celestial bodies with similarly thin atmospheres. The Keywords microflyers could work alongside current terrestrial rovers to study the Martian atmosphere and provide a cheaper, small- atomic layer deposition, carbon nanotubes, mechanical metamaterials, photophoretic levitation scale alternative to the Mars helicopter.[40] While, here, we have demonstrated the proof of concept of Received: October 19, 2019 photophoretic propulsion for macroscopic structures, the lift Revised: December 30, 2019 force can be greatly increased in the future by further opti- Published online: February 20, 2020 mizing the geometry and using advanced optical absorbers. By placing optical absorbers with a large optical cross section, such as nanoparticles, at the entrances of the channels, local [1] D. Floreano, R. J. Wood, Nature 2015, 521, 460. temperature gradients can be enhanced by at least an order of [2] J. Sirohi, Proc. SPIE 2013, 8686, 86860U. magnitude. Because the speed of the thermal-transpiration jet [3] R. Bogue, Ind. Rob. 2010, 37, 17. is proportional to a local temperature gradient, resonant optical [4] M. Piccoli, M. Yim, in Proc. of 2017 IEEE Int. Conf. on Robotics and absorbers film can lead to much more efficient levitation than Automation (ICRA), IEEE, Piscataway, NJ, USA 2017, p. 3328. the carbon nanotube films we used as uniform, broadband [5] K. Y. Ma, P. Chirarattananon, S. B. Fuller, R. J. Wood, Science 2013, absorbers here. The use of localized absorbers would also allow 340, 603.

[6] N. T. Jafferis, E. F. Helbling, M. Karpelson, R. J. Wood, Nature 2019, [28] N. K. Gupta, Y. B. Gianchandani, in Proc. of 2010 IEEE 23rd Int. Conf. 570, 491. on Micro Electro-Mechanical Systems (MEMS), IEEE, Piscataway, NJ, [7] L. Niccolai, M. Bassetto, A. A. Quarta, G. Mengali, Prog. Aeronaut. USA 2010, p. 1095. Sci. 2019, 106, 1. [29] A. Seungdo, N. K. Gupta, Y. B. Gianchandani, J. Microelectromech. [8] W. Shyy, H. Aono, C. Kang, H. Liu, An Introduction to Flapping Wing Syst. 2014, 23, 406. Aerodynamics, Cambridge University Press, Cambridge, UK 2013. [30] Y. Sone, Kinetic Theory and Fluid Dynamics, Birkhäuser, Boston, MA, [9] A. Ketsdever, N. Gimelshein, S. Gimelshein, N. Selden, Vacuum USA 2002. 2012, 86, 1644. [31] C. Lin, S. M. Nicaise, D. E. Lilley, J. Cortes, P. Jiao, J. Singh, [10] A. Einstein, Z. Phys. 1924, 27, 1. M. Azadi, G. G. Lopez, M. Metzler, P. K. Purohit, I. Bargatin, Nat. [11] M. Knudsen, The Kinetic Theory of Gases: Some Modern Aspects, Commun. 2018, 9, 4442. Methuen, London, UK 1950. [32] N. T. Tong, J. Colloid Interface Sci. 1975, 51, 143. [12] G. Wurm, O. Krauss, Phys. Rev. Lett. 2006, 96, 134301. [33] B. Warneke, M. Last, B. Liebowitz, K. S. J. Pister, Computer 2001, [13] V. G. Shvedov, C. Hnatovsky, A. V. Rode, W. Krolikowski, Opt. 34, 44. Express 2011, 19, 17350. [34] Y. Lee, S. Bang, I. Lee, Y. Kim, G. Kim, M. H. Ghaed, P. Pannuto, [14] S. Dong, Y. Liu, J. Aerosol Sci. 2017, 113, 114. P. Dutta, D. Sylvester, D. Blaauw, IEEE J. Solid-State Circuits 2013, [15] G. M. Hidy, J. R. Brock, J. Geophys. Res. 1967, 72, 455. 48, 229. [16] C. Orr Jr., E. Y. H. Keng, J. Atmos. Sci. 1964, 21, 475. [35] G. Chen, H. Ghaed, R. U. Haque, M. Wieckowski, Y. Kim, G. Kim, [17] R. F. Pueschel, S. Verma, H. Rohatschek, G. V. Ferry, N. Boiadjieva, D. Fick, D. Kim, M. Seok, K. Wise, D. Blaauw, Proc. 2011 IEEE Int. S. D. Howard, A. W. Strawa, J. Geophys. Res.: Atmos. 2000, 105, 3727. Solid-State Circuits Conf., IEEE, Piscataway, NJ, USA 2011, p. 310. [18] D. Keith, Proc. Natl. Acad. Sci. USA 2010, 107, 16428. [36] L. Niccolai, M. Bassetto, A. A. Quarta, G. Mengali, Prog. Aerosp. Sci. [19] H. Rohatschek, J. Aerosol Sci. 1996, 27, 467. 2019, 106, 1. [20] C. Loesche, T. Husmann, J. Aerosol Sci. 2016, 102, 55. [37] H. Zhang, L. Xiang, Y. Yang, M. Xiao, J. Han, L. Ding, Z. Zhang, [21] D. E. Smalley, E. Nygaard, K. Squire, J. Van Wagoner, J. Rasmussen, Y. Hu, L. M. Peng, ACS Nano 2018, 12, 2773. S. Gneiting, K. Qaderi, J. Goodsell, W. Rogers, M. Lindsey, [38] E. Song, Q. Guo, G. Huang, B. Jia, Y. Mei, E. Song, ACS Appl. Mater. K. Costner, Nature 2018, 553, 486. Interfaces 2017, 9, 12171. [22] D. E. H. Jones, Nature 1998, 392, 337. [39] D. Jariwala, A. R. Davoyan, J. Wong, H. A. Atwater, ACS Photonics [23] D. E. H. Jones, Nature 1998, 395, 120. 2017, 4, 2962. [24] J. D. Sanchez, P. A. Garbuz, A. D. Zonenberg, US 8794930, 2012. [40] W. J. K. Koning, E. A. Romander, W. Johnson, Low Reynolds Number [25] G. Karniadakis, A Beskok, N. Aluru, Microflows and Nanoflows: Airfoil Evaluation for the Mars Helicopter Rotor, https://ntrs.nasa. Fundamentals and Simulation, Springer, New York 2005. gov/archive/nasa/casi.ntrs.nasa.gov/20180008702.pdf (accessed: [26] F. Sharipov, V. Seleznev, J. Phys. Chem. Ref. Data 1998, 27, 657. December 2019). [27] A. Lotfian, E. Roohi, J. Fluid Mech. 2019, 860, 544. [41] O. Ilic, H. A. Atwater, Nat. Photonics 2019, 13, 289.