The Leidenfrost Effect at the Nanoscale Jhonatam Cordeiro Department of Industrial and Nanotechnology has been presenting successful applications in several fields, such as Systems Engineering, electronics, medicine, energy, and new materials. However, the high cost of investment in North Carolina A&T State University, facilities, equipment, and materials as well as the lack of some experimental analysis at 419 McNair Hall, the nanoscale can limit research in nanotechnology. The implementation of accurate 1601 East Market Street, computer models can alleviate this problem. This research investigates the Leidenfrost Greensboro, NC 27411 effect at the nanoscale using molecular dynamics (MDs) simulation. Models of water e-mail: [email protected] droplets with diameters of 4 nm and 10 nm were simulated over gold and silicon sub- 1 strates. To induce the Leidenfrost effect, droplets at 293 K were deposited on heated sub- Salil Desai strates at 373 K. As a baseline, simulations were run with substrates at room temperature Downloaded from http://asmedigitalcollection.asme.org/heattransfer/article-pdf/4/4/041001/5949632/jmnm_004_04_041001.pdf by guest on 27 September 2021 Department of Industrial and (293 K). Results show that for substrates at 293 K, the 4 nm droplet has higher position Systems Engineering, variability than the 10 nm droplets. In addition, for substrates at 373 K, the 4 nm droplets North Carolina A&T State University, have higher velocities than the 10 nm droplets. The wettability of the substrate also influ- 423 McNair Hall, ences the Leidenfrost effect. Droplets over the gold substrate, which has hydrophobic 1601 East Market Street, characteristics, have higher velocities as compared to droplets over silicon that has a Greensboro, NC 27411 hydrophilic behavior. Moreover, the Leidenfrost effect was observed at the boiling e-mail: [email protected] temperature of water (373 K) which is a significantly lower temperature than reported in previous experiments at the microscale. This research lays the foundation for investigating the fluid–structure interaction within several droplet based micro- and nano-manufacturing processes. [DOI: 10.1115/1.4034607] Introduction [10] used MD simulation to investigate the failure mechanism of Si molds and reduce defective outputs of nano-imprint lithography Nanotechnology has been presenting several applications (NIL) process. Chen et al. [11] used a hybrid approach of MD despite being a recent field. Applications include compact transis- simulation and experiments and proposed a highly sensitive sen- tors that make faster and energy efficient processors and memory sor to detect molecular conformation. Desai et al. [12] used MD chips [1], long-life battery cells [2], efficient drug-delivery sys- simulations to investigate the wettability of SiO and Si N as a tems [3], DNA sequencing chips [4], stronger structural materials 2 3 4 function of temperature. Borodin et al. [13] used MD simulations [5], superconducting materials [6], and others. These successful to investigate the properties of lithium batteries and minimize applications of components measuring less than 100 nm have interfacial resistance, improving battery safety and battery life. demonstrated the potential of nanotechnology and the need for a This research demonstrates the applicability of molecular better understanding of material properties at the nanoscale. dynamics to investigate the Leidenfrost effect at the nanoscale Working with nanotechnology often requires expensive materials, which is an underexplored phenomenon. Several direct-write specialized equipment, and state-of-the-art facilities. Besides the droplet based manufacturing processes involve the deposition of high cost of these resources, they also require long setup times, micro- and nano-scale droplets on heated substrates. These trained personnel, and complex procedures. The high cost of pro- include scalable inkjet [14,15], aerosol jet [16], electrohydrody- curement and operation of nanotechnology resources can restrict namic jet [17], and other atomized droplet processes where the the development in this field. To the best of our knowledge, there droplet deposition dynamics determines the morphology of the is no microscopy method that can observe droplet at the nano- printed feature. Thus, it is critical that the transport properties of scale, without influencing the measurements of the experiments. droplet be investigated in order to determine the fluid–structure One solution to help optimize the R&D in nanotechnology is to interaction between different substrate materials at elevated tem- use computer models to help optimize design and process parame- peratures. In addition, this research investigates the droplet move- ters of the experiments and products, hence reducing costs and ment on heated substrates, which has an impact on the final increasing design freedom. placement and stabilization shape of the droplet. The phenomena Traditional numerical simulation methods, such as finite ele- of water interfacing with a hot surface were first investigated by ments, predict material properties up to the submicron scale [7,8]. Leidenfrost in 1756 [18]. If a droplet of liquid is deposited over a However, they fail to make correct predictions for components surface around the boiling temperature of the liquid, the liquid smaller than several hundred nanometers [9]. A more accurate boils and evaporates rapidly. However, when the temperature of approach for computer modeling for the nanoscale is to use the surface is significantly higher than the boiling temperature of molecular dynamics models. In MDs models, each atom in the the liquid, the droplet levitates over its own vapor, greatly increas- system and their interaction with other atoms are represented. The ing the evaporation time of the droplet. Also, the thin layer of representation accuracy, freedom of design, and insights that can vapor avoids the nucleation of bubbles, preventing the droplet be gained by using MD modeling have attracted several research from boiling and making it evaporate slowly [19]. This research groups to investigate nanotechnology applications. Tada et al. investigates the Leidenfrost effect at the nanoscale based on varia- tions in droplet size, substrate temperature, and substrate material. 1Corresponding author. We track the droplet trajectory path via its centroid and its veloc- Contributed by the Manufacturing Engineering Division of ASME for publication ity overtime. The tools, parameters, and models used in this work in the JOURNAL OF MICRO- AND NANO-MANUFACTURING. Manuscript received May 30, 2016; final manuscript received August 22, 2016; published online October 10, 2016. can be extended to other materials and substrate geometries to Assoc. Editor: Rajiv Malhotra. enable other scenarios of the Leidenfrost effect to be studied. The Journal of Micro- and Nano-Manufacturing Copyright VC 2016 by ASME DECEMBER 2016, Vol. 4 / 041001-1 results of this research provide quantitative analysis on the role Table 1 Nonbonded interaction parameters of different process parameters that impact the Leidenfrost ˚ effect at the nanoscale regime. Further, they provide a basis to Atom type eijðkcal=molÞ rijðAÞ investigate several droplet based nano- and micro-manufacturing processes. Si Si À0.31 4.27 Au Au À0.039 1.8436 Methodology In this research, the nanoscale molecular dynamics (NAMDs) [20] source code was used in conjunction with virtual molecular dynamics (VMDs) software to model the materials in the system Table 2 Simulation configurations and visualize the simulation output. NAMD is robust for parallel computing [20,21] and open source and compatible with most Index Temperature (K) Substrate Droplet size (nm) force fields for commonly available CHARMM [22] and AMBER [23] packages. The MD simulations were executed on a graphical 1 293 Au 10 2 293 Au 4 Downloaded from http://asmedigitalcollection.asme.org/heattransfer/article-pdf/4/4/041001/5949632/jmnm_004_04_041001.pdf by guest on 27 September 2021 processing unit (GPU) processor (NVIDIA Tesla K40) with 3 293 Si 10 2880 CUDA cores on a 64-bit Linux based system to enhance the 4 293 Si 4 performance of the computations [24]. The NAMD source code 5 373 Au 10 uses the GPU for nonbonded force evaluation while the energy 6 373 Au 4 evaluation is done on the central processing unit [25]. 7 373 Si 10 In this work, MD models of 4 nm and 10 nm water droplets 8 373 Si 4 were used, containing 1108 and 17,267 molecules, respectively. Substrates of gold and silicon measuring 240 A˚ Â 240 A˚ Â 40 A˚ in dimensions were used. The water molecules had a TIP3P structure X and were modeled using VMD. All the force fields used in the angle 2 Uangle ¼ ki ðhi À h0iÞ (3) simulations are compatible with the CHARMM standard, and the bonds i format of potential energy function shown in Eq. (1) was used to X X represent the atomic interactions. The simulations were performed qiqj with a 2 fs integration time step. Van der Waals interactions were UCoulomb ¼ (4) 4pe0rij computed with a cutoff of 12 A˚ and switching function starting i j>i at 10 A˚ . The long-range electrostatic forces were computed "# X X 12 6 using a particle mesh Ewald (PME) summation method. A canoni- rij rij U ¼ 4e À (5) cal ensemble of conserved number of atoms, volume and tempera- vdW ij i j>i rij rij ture was used, and a Langevin thermostat was employed to control the temperature. The water droplets were modeled as The parameters for Eqs. (2)–(5) for water molecules (oxygen spheres and placed on top of the substrates, as illustrated in Fig. 1 and hydrogen) were taken from the CHARMM force fields [26]. The parameters for gold and Si substrates were adapted from U ¼ U þ U þ U (1) total bond angle nonbond Braun et al. [27] and Mayo et al. [28], respectively, as shown in Table 1. where Ubond and Uangle are the stretching and bending interac- The objective of the simulations was to analyze the Leidenfrost tions, as described in Eqs. (2) and (3), respectively. Unonbond is the effect which describes the fluid–substrate interactions at the nano- interaction between nonbonded pairs of atoms that correspond to scale.