betaFeedback

WorksPhysics and Astronomy

Nanodroplet Condensation on Solid Surfaces

Abstract

This paper deals with the condensation of liquid droplets on hydrophobic and hydrophilic surfaces. A stochastic mesoscale model based on the theory of fluctuating hydrodynamics and the thermodynamics of a diffuse interface approach shows how direct simulation of the vapour-liquid transition from the nucleation process to droplet hydrodynamics can be achieved. Such simulations explain the role of wettability in filmwise and dropwise condensation regimes and the main limitations of classical nucleation theory.

Introduction

Condensation is of central interest due to the innumerable conditions where vapour/liquid phase change takes place, from planetary – e.g., clouds [1]– to human – e.g., water harvesting and desalination [2, 3] – and micro/nano scales – e.g., biotechnology applications, electronics cooling [4, 5, 6] and bionics/biomimetics [7, 8]. Fluid mechanics has long been involved in deciphering condensation [9, 10, 11, 12, 13], an effort partially hindered by the poor understanding of the onset of the phenomenon, namely the nucleation process [14].

The classical picture of condensation is known. Vapour transforms into liquid by decreasing the temperature below the saturation point or by increasing the pressure above the liquefaction threshold. However, this description is oversimplified because it does not take into account metastability. In reality, a vapour may remain in a metastable state (local free-energy minimum) for a long time well beyond its thermodynamic saturation limits. Here the granular nature of matter plays a central role: molecules tend to aggregate with a certain probability resulting in nanometric droplets. A rare event consisting of a large enough cluster —a critical droplet— requires a certain activation energy (nucleation barrier) to be exceeded, triggering the phase change. Associated with the atomistic features of the process, nucleation also involves the macroscopic hydrodynamics of the droplets, namely the exchange of latent heat and coalescence.

Macroscopic approaches currently do not account for thermal fluctuations and cannot describe droplet nucleation. These simulations are always initialized with preexisting nuclei which are only partially consistent with the thermodynamics of phase change. Consequently, it is impossible to predict whether the phase transformation begins within the vapour bulk (homogeneous nucleation) or in contact with a solid surface (heterogeneous nucleation) and whether droplets form in a film/dropwise mode. At the fundamental level, these different nucleation mechanisms are inherently connected with thermodynamic conditions and surface chemistry and proper modelling should take these aspects into account.

Currently, Molecular Dynamics (MD) is the universally trusted tool for such a task [15, 16, 17], but its computational cost limits its use to nanometric systems on the nanosecond time scales. The coupling with macroscopic hydrodynamics became feasible by exploiting a recent mesoscale model originally developed for the reverse liquid/vapour transformation [18, 19, 20, 21, 22].

The methodology is based on a diffuse interface description [23, 24] combined with Fluctuating Hydrodynamics (FH) [25]. FH is a powerful tool for assessing the influence of thermal fluctuations on macroscopic fluid dynamics [26, 27, 28, 29, 30, 31]. Properly accounting for thermal fluctuations (Einstein-Boltzmann distribution) allows for the spontaneous nucleation of liquid droplets, describing condensation from nucleation to hydrodynamics. Using a prototypal fluid, we analyse the dependency on the solid surface chemistry of the different (dropwise/filmwise) condensation modes, when the system temperature is initially uniform.

Mathematical Model

The proposed mesoscale model consists of the coupling between Landau-Lifshitz-Navier-Stokes equations and diffuse interface thermodynamics. Here the conservation equations for mass, momentum and energy are augmented with stochastic fluxes (δΣ\delta\boldsymbol{\Sigma}, δq\delta\boldsymbol{q}).

∂ρ∂t+∇⋅(ρv)=0,(1)\frac{\partial\rho}{\partial t}+\nabla\cdot(\rho\boldsymbol{v})=0, \tag*{(1)}
∂ρv∂t+∇⋅(ρv⊗v)=∇⋅Σ+δΣ,(2)\frac{\partial\rho\boldsymbol{v}}{\partial t}+\nabla\cdot(\rho\boldsymbol{v}\otimes\boldsymbol{v})=\nabla\cdot\boldsymbol{\Sigma}+\delta\boldsymbol{\Sigma}, \tag*{(2)}
∂E∂t+∇⋅(vE)=∇⋅[(Σ+δΣ)⋅v−q]+δq,(3)\frac{\partial E}{\partial t}+\nabla\cdot(\boldsymbol{v}E)=\nabla\cdot[(\boldsymbol{\Sigma}+\delta\boldsymbol{\Sigma})\cdot\boldsymbol{v}-\boldsymbol{q}]+\delta q, \tag*{(3)}

where ρ\rho is the mass density, v\boldsymbol{v} the velocity, and EE the total energy density (E=ub(ρ,θ)+1/2ρ∣v∣2+1/2λ∣∇ρ∣2E=u_b(\rho,\theta)+1/2\rho|\boldsymbol{v}|^2+1/2\lambda|\nabla\rho|^2, with ub(ρ,θ)u_b(\rho,\theta) the bulk internal energy density, θ\theta is the temperature, λ\lambda the capillary coefficient, and ∇\nabla the gradient operator). Σ\boldsymbol{\Sigma} and q\mathbf{q} are the deterministic stress tensor and energy flux as identified by non-equilibrium thermodynamics that include distributed capillary effects [32, 33] (see also [34, 35, 36] for extensions to multicomponent systems),

Σ=[−p+λ2∣∇ρ∣2+ρ∇⋅(λ∇ρ)]I−λ∇ρ⊗∇ρ+η1(∇u+∇uT)+η2∇⋅uI,(4)\boldsymbol{\Sigma}=\left[-p+\frac{\lambda}{2}|\nabla\rho|^2+\rho\nabla\cdot(\lambda\nabla\rho)\right]\mathbf{I}-\lambda\nabla\rho\otimes\nabla\rho+\eta_1(\nabla\mathbf{u}+\nabla\mathbf{u}^{\mathrm{T}})+\eta_2\nabla\cdot\mathbf{u}\mathbf{I}, \tag*{(4)}
q=λρ∇ρ∇⋅u−k∇θ,(5)\mathbf{q}=\lambda\rho\nabla\rho\nabla\cdot\mathbf{u}-k\nabla\theta, \tag*{(5)}

where p(ρ,θ)p(\rho,\theta) is the thermodynamic pressure, I\mathbf{I} the identity matrix, the transport coefficients η1(ρ,θ),η2(ρ,θ)\eta_1(\rho,\theta),\eta_2(\rho,\theta) are fluid’s viscosities, and k(ρ,θ)k(\rho,\theta) the thermal conductivity.

The fluctuation-dissipation balance provides the statistical properties of the Gaussian stochastic fluxes [20]

⟨δq(x^,t^)⟩=0⟨δΣ(x^,t^)⟩=0,\langle\delta\mathbf{q}(\hat{\mathbf{x}},\hat{t})\rangle=0\qquad\langle\delta\boldsymbol{\Sigma}(\hat{\mathbf{x}},\hat{t})\rangle=0,
⟨δΣ(x^,t^)⊗δΣ†(x~,t~)⟩=QΣδ(x^−x~)δ(t^−t~),\langle\delta\boldsymbol{\Sigma}(\hat{\mathbf{x}},\hat{t})\otimes\delta\boldsymbol{\Sigma}^{\dagger}(\tilde{\mathbf{x}},\tilde{t})\rangle=\mathbf{Q}^{\Sigma}\delta(\hat{\mathbf{x}}-\tilde{\mathbf{x}})\delta(\hat{t}-\tilde{t}),
⟨δq(x^,t^)⊗δq†(x~,t~)⟩=Qqδ(x^−x~)δ(t^−t~),(6)\langle\delta\mathbf{q}(\hat{\mathbf{x}},\hat{t})\otimes\delta\mathbf{q}^{\dagger}(\tilde{\mathbf{x}},\tilde{t})\rangle=\mathbf{Q}^{q}\delta(\hat{\mathbf{x}}-\tilde{\mathbf{x}})\delta(\hat{t}-\tilde{t}), \tag*{(6)}

where δ(x)\delta(\mathbf{x}) and δ(t)\delta(t) are the spatial and temporal Dirac delta functions, respectively, and

QαβνηΣ=2kBθ[η1(δανδβη+δαηδβν)+η2δαβδνη],Q^{\Sigma}_{\alpha\beta\nu\eta}=2k_{\mathrm{B}}\theta\left[\eta_1(\delta_{\alpha\nu}\delta_{\beta\eta}+\delta_{\alpha\eta}\delta_{\beta\nu})+\eta_2\delta_{\alpha\beta}\delta_{\nu\eta}\right],
Qαβq=2kBθ2kδαβ,(7)Q^{q}_{\alpha\beta}=2k_{\mathrm{B}}\theta^2 k\delta_{\alpha\beta}, \tag*{(7)}

with the subscript Greek indices denoting the Cartesian components of tensors, kBk_{\mathrm{B}} the Boltzmann constant, and δαβ\delta_{\alpha\beta} the Kronecker delta symbols.

To account for the fluid-solid interaction (i.e., the solid wettability), we have recently proposed a free-energy contribution [20] which enforces the following boundary condition for the mass density,

\lambda\nabla\rho\cdot\hat{\mathbf{n}}=-\frac{\mathrm{d}f_{\mathrm{w}}}{\mathrm{d}\rho}=\cos\phi\sqrt{2\lambda\left(\omega_{\mathrm{b}}(\rho,\theta)-\omega_{\mathrm{b}}(\rho_{\mathrm{V})\right)},} \tag*{(8)}

with ϕ\phi the Young contact angle, ωb(ρ,θ)=fb−ρ∂fb/∂ρ\omega_{\mathrm{b}}(\rho,\theta)=f_{\mathrm{b}}-\rho\partial f_{\mathrm{b}}/\partial\rho and fb(ρ,θ)f_{\mathrm{b}}(\rho,\theta) the Landau and Helmholtz bulk free-energy densities, respectively, and n^\hat{\mathbf{n}} the unit normal vector. It is worth stressing that the bulk energies ubu_{\mathrm{b}}, fbf_{\mathrm{b}}, ωb\omega_{\mathrm{b}} and the pressure pp are identified by a proper equation of state (EoS). The EoS also defines the surface tension [37, 38]. In the present case, we adopt the modified Benedict-Webb-Rubin equation for mimicking a Lennard-Jones (LJ) fluid [39].

Numerical Setup: Condensation simulations are based on the system of Eqs. (1), (2), (3). They are made dimensionless using the LJ reference quantities: σ=3.4×10−10 m\sigma=3.4\times10^{-10}\ \mathrm{m} as length, ϵ=1.65×10−21 J\epsilon=1.65\times10^{-21}\ \mathrm{J} as energy,

m=6.63×10−26 kgm=6.63\times10^{-26}\ \mathrm{kg} as mass, θr=kB/ϵ=119.56K\theta_{\mathrm{r}}=k_{\mathrm{B}}/\epsilon=119.56\mathrm{K} as temperature, vr=(ϵ/m)1/2=155.83m/sv_{\mathrm{r}}=(\epsilon/m)^{1/2}=155.83\mathrm{m/s} as velocity, and tr=σ/vr=2.15×10−12st_{\mathrm{r}}=\sigma/v_{\mathrm{r}}=2.15\times10^{-12}\mathrm{s} as time. To reproduce the surface tension of a LJ fluid and its temperature dependence, we set λ=5.224\lambda=5.224, with λr=(σ5ϵ)1/2/m2=1.52×10−18kgm7/s2\lambda_{\mathrm{r}}=(\sigma^5\epsilon)^{1/2}/m^2=1.52\times10^{-18}\mathrm{kgm^7/s^2} [40]. Simulations consist of homogeneous metastable vapour enclosed in a box with two flat solid surfaces parallel to the y−zy-z plane. The boundary conditions on these surfaces are ρv=0\rho_{\mathrm{v}}=0, θ=θw\theta=\theta_{\mathrm{w}}, with the contact angle ϕ\phi prescribed through Eq. (8), and periodicity in zz and yy directions.

The equations to be solved in a box of volume V=400×1000×1000V=400\times1000\times1000 are discretized on a uniform 40×100×10040\times100\times100 grid with the scheme proposed in [41] using a second-order Runge-Kutta scheme for time evolution (Δt=0.1\Delta t=0.1). This scheme was shown to preserve the fluctuation-dissipation balance at the discrete level [41, 42, 43, 44] and to guarantee converged statistics of thermal fluctuations [18, 20, 45].

Results and Discussion

The present simulation campaign investigates the dependence of condensation on the wettability of solid surfaces. The system is initialized with a uniform temperature θ=1.25\theta=1.25 and an initial vapour density ρV=0.16\rho_{\mathrm{V}}=0.16, changing the contact angle in the range 15∘15^\circ–95∘95^\circ. In these conditions, the vapour is metastable, hence condensation is expected to occur. Snapshots of the density field along the simulation for ϕ=30∘\phi=30^\circ are reported in Figure 1.

Snapshots of the density field for filmwise condensation

Figure 1. Filmwise condensation: Snapshots of the density field at the solid walls, ρV=0.16\rho_V = 0.16, ϕ=30∘\phi= 30^\circ. a) t = 600, b) t = 2000, c) t = 20000, and d) t = 150000. The isosurfaces at the critical density ρ=0.33\rho= 0.33, taken to discern liquid from vapour, are shown in green. A sketch of the simulation box setup is also illustrated.

Starting from a homogeneous metastable vapour, after a certain incubation time where sub-critical embryos are randomly formed and reabsorbed, droplet nucleation begins to take place, Figure 1.a). In a successive phase, the number and size of droplets are sufficiently large to give rise to a number of coalescence events, Figure 1.b) and Figure 1.c). Eventually, a liquid film is formed as the precursor of the well-known film condensation mode in macroscopic experiments, Figure 1.d). For ρV=0.16\rho_{\mathrm{V}}=0.16, filmwise condensation is observed for contact angles below 60∘60^\circ where walls are rapidly covered by many nanodroplets that coalesce to form a continuous film.

At larger contact angles, the phenomenology is different. Only a few droplets nucleate, see Figure 2 where ϕ=90∘\phi=90^\circ. After the condensation nuclei appear, Figure 2.a), individual droplets form, Figure 2.b), that have enough time to grow as distinct entities, Figure 2.c), to occasionally coalesce, Figure 2.d). Here we witness a sort of dropwise condensation mode at the nanoscale.

Snapshots of numerical simulations for dropwise condensation

Figure 2. Dropwise condensation: Snapshots of numerical simulations for the case ρV=0.16\rho_V = 0.16 and ϕ=90∘\phi= 90^\circ. a) t = 600, b) t = 2000, c) t = 20000, and d) t = 150000. The isosurfaces at ρ=ρcut=0.33\rho= \rho_{\mathrm{cut}} = 0.33 that delimits the liquid phase are shown in green and the density field at the wall is reported. A sketch of the simulation box setup is also illustrated.

The evolution of the number of droplets over time, Nd(t)N_{\mathrm{d}}(t), at various wall wettabilities, is illustrated in Figure 3. After the incubation time, the number of droplets initially increases, linearly in time during a first phase. Successively, the growth saturates and eventually, the number of detected drops steadily decreases. In part, some droplets are reabsorbed. This is due to the combined volume and total mass constraint which, af-

Evolution of the number of supercritical droplets

Figure 3. Evolution of the number of supercritical droplets (NdN_d) over time. Different contact angles (ϕ\phi) are shown.

ter a certain amount of fluid liquefies, forces part of it to re-evaporate to occupy all the available space. This phenomenon is partly attributable to coalescence events which, while maintaining the overall volume of the liquid phase, lead to a decreased number of larger droplets. This picture is also observed in MD simulations [46].

Figure 3 illustrates that the maximum number of formed droplets significantly decreases with increasing contact angle. Furthermore, following the peak, the subsequent decline in droplet count becomes more gradual.

In nucleation phenomena, the crucial observable is the nucleation rate JJ, related to the time history of the number of droplets, Nd(t)N_d(t). In heterogeneous nucleation, it is defined as the number of droplets formed per unit time and unit wall area. In most MD simulations, where single droplet nucleation occurs due to the small spatial extent of the domain, the average time needed to observe a droplet is directly estimated from repeated simulations in identical thermodynamic conditions. For systems as large as the one currently addressed (fraction of micrometers), multiple nucleation events are observed in a single simulation. In this case, JJ can be accurately determined from a single simulation [47]. As done in the context of cavitation [20], we adopt the same threshold method and measure JJ from the slope of the linear fit to the NdN_d-vs-time curve in the (stationary) nucleation phase where Nd∝tN_d \propto t, i.e. J=1/AwdNd(t)/dtJ = 1/A_w dN_d(t)/dt, with AwA_w the area of the condensing solid wall. Being Classical Nucleation Theory (CNT) the cornerstone of any discussion of (first-order) phase transition, it is natural to compare our estimates for the nucleation rates at different contact angles with those provided by CNT in the heterogeneous case JCNT=nV2/3(nV/nL)(1−cos⁡(ϕ))/22σ/(πmψ)exp⁡(−ΔΩ†/(kBθ))J_{\mathrm{CNT}} = n_V^{2/3}(n_V/n_L)(1-\cos(\phi))/2\sqrt{2\sigma/(\pi m\psi)}\exp(-\Delta\Omega^\dagger/(k_B\theta)), where nV/Ln_{V/L} is the number density of molecules in the vapour and liquid state, respectively, and ψ=(1+cos⁡(ϕ))2(2−cos⁡(ϕ))/4\psi= (1+\cos(\phi))^2(2-\cos(\phi))/4 is a geometric factor that takes into account the contact angle [48]. The comparison shows that CNT gives an acceptable prediction for ρV=0.16\rho_V = 0.16 only for contact angles near neutral wettability ϕ≃90∘\phi\simeq90^\circ.

By decreasing the contact angle, the CNT is still qualitatively accurate, i.e. a higher nucleation rate at smaller contact angles, but the difference with the numerical results is not negligible (orders of magnitude). This is not dissimilar to what is observed in homogeneous and heterogeneous condensation when comparing Molecular Dynamics results with CNT [49] and [50], respectively. Now, we focus on identifying the primary source of error in the estimation of the nucleation rate, J=J0exp⁡(−ΔΩ†/(kBθ))J = J_0 \exp(-\Delta\Omega^\dagger/(k_B\theta)), whether it originates from the pre-exponential factor J0J_0 or from the energy barrier ΔΩ†\Delta\Omega^\dagger associated with the nucleation process. When dealing with nanodrops/bubbles, the typical dimension of the liquid/vapor interface is comparable with that of the droplet/bubbles, hence sharp-interface predictions –as CNT– may fail [38, 51]. Therefore, to precisely identify ΔΩ†\Delta\Omega^\dagger, the string method [52] is employed on top of diffuse interface thermodynamics (ΔΩDI†\Delta\Omega^\dagger_{\mathrm{DI}}). This is expected to be the barrier that the vapour phase experiences in our simulations. Figure 4 compares our numerical results with CNT and CNT-DI where the actual energy barrier is ΔΩDI†\Delta\Omega^\dagger_{\mathrm{DI}} and JCNT−DI=J0exp⁡(−ΔΩDI†/(kBθ))J_{\mathrm{CNT-DI}} = J_0\exp(-\Delta\Omega^\dagger_{\mathrm{DI}}/(k_B\theta)), with J0=nV2/3(nV/nL)(1−cos⁡(ϕ))/22σ/(πmψ)J_0 = n_V^{2/3}(n_V/n_L)(1-\cos(\phi))/2\sqrt{2\sigma/(\pi m\psi)} the usual CNT pre-exponential factor.

Normalised free energy barrier and nucleation rates versus contact angle

Figure 4. Top panel: Normalised free energy barrier (ΔΩ†/kBθ)(\Delta\Omega^\dagger/k_{\mathrm{B}}\theta) vs contact angle (ϕ\phi). Dashed lines depict CNT, while symbols correspond to diffuse interface prediction. Bottom panel: Nucleation rates (JJ) vs contact angle (ϕ\phi). Numerical simulations (red diamonds), CNT coupled with diffuse interface (CNT-DI black squares), CNT (blue triangles), FH coupled with CNT (green circles).

As shown on the top panel of Figure 4, CNT overestimates the energy barrier over the entire range of observed contact angles at various degrees of metastability, in agreement with Molecular Dynamics simulations of heterogeneous condensation [50]. The nucleation rate JJ in the two approaches overlaps only for low contact angles, where the barrier goes to zero and the nucleation rate is dominated by J0J_0, which is the same in both models. As the contact angle grows, the contribution of J0J_0 becomes less and less important and the nucleation rate heavily depends on the free-energy barrier. It is also apparent that our numerical results, FH on the bottom panel of Figure 4, and the CNT-DI model are comparable only for more hydrophobic chemistry where the role of J0J_0 is not negligible. Hence, when the nucleation barriers are not so high, a significant part of the misalignment between CNT predictions and numerical data stems from a poor estimation of the factor J0J_0. From a physical perspective, the prefactor J0J_0 reflects both droplet diffusion in the vapor medium and the density of potential nucleation sites. In the homogeneous case, theoretical predictions [53] and numerical simulations [18] suggest that FH offers a comprehensive description of both the energetic barrier (ΔΩ†\Delta\Omega^\dagger) and the kinetic aspects (J0J_0) of nucleation. Therefore, given a numerically obtained nucleation rate JFHJ_{\mathrm{FH}} we extract the corresponding FH pre-exponential factor as J0FH=JFHexp⁡(ΔΩDI†/(kBθ))J_{0\mathrm{FH}} = J_{\mathrm{FH}}\exp(\Delta\Omega^\dagger_{\mathrm{DI}}/(k_B\theta)). Figure 5 shows J0FHJ_{0\mathrm{FH}} evaluated at various ϕ\phi for different initial vapour densities. For full context, we also report the nucleation rate resulting from the FH prefactor and CNT barrier, green circles in the bottom panel of Figure 4. In the latter case, JJ is underestimated due to the CNT higher barriers.

Numerical values of $J_0(\rho,\phi)$ versus $\phi$ for different densities

Figure 5. Numerical values of J0(ρ,ϕ)J_0(\rho,\phi) vs ϕ\phi for different densities. The two dashed lines represent the minimum and maximum CNT prediction J0J_0 for each ϕ\phi in the 0.155-0.17 density range.

For comparison, the range of classical nucleation theory (CNT) prefactors is also reported (region between the two dashed lines). Two distinct trends, associated with different condensation regimes, are evident. In the filmwise condensation regime JFHJ_{\mathrm{FH}} remains nearly constant and is largely independent of the degree of metastability. In contrast, during dropwise condensation, the prefactor J0J_0 exhibits dependence on both ϕ\phi and ρ\rho. This behavior can be qualitatively interpreted using Langer’s theory [53], originally formulated for homogeneous nucleation. According to this framework, the prefactor scales with the system volume divided by the cube of the correlation length. In DI thermodynamics, the correlation length increases with metastability and diverges near the spinodal limit [40]. Consequently, the prefactor is expected to decrease as metastability increases—an effect not captured by CNT. Although Langer’s theory has not been formally extended to heterogeneous nucleation, its core concepts plausibly apply to dropwise condensation, which closely mirrors the dynamics of the homogeneous case.

Conclusions

Condensation is a complex phenomenon rooted in atomistic processes but also involving macroscopic hydrodynamic scales. For this reason, it has traditionally been studied at separate levels: nucleation through molecular dynamics simulations and bulk fluid behavior through Navier-Stokes-based solvers. In this work, we adopt a holistic approach, describing the entire process using a mesoscale model that combines hydrodynamic equations with the inherent granularity of matter treated as stochastic processes. This approach offers new insights into the various observed condensation regimes. Moreover, the relatively low computational cost (∼103\sim10^3 core hours per simulation) paves the way for physically grounded, large-scale simulations of condensation in systems of significant technological relevance.

Acknowledgements

Funded by the European Union. Views and opinions expressed are however those of the authors only and do not necessarily reflect those of the European Union or the European Research Council Executive Agency. Neither the European Union nor the granting authority can be held responsible for them. This work is supported by an ERC grant (ERC-STG E-Nucl. Grant agreement ID: 101163330).

Declaration of Interests

The authors report no conflict of interest.

Data Availability

The data that support the findings of this study are openly available at [54].

References

  1. [1]C. Hoose, J. Kristjánsson, and S. Burrows, Environmental Research Letters 5, 024009 (2010).
  2. [2]K.-C. Park, P. Kim, A. Grinthal, N. He, D. Fox, J. C. Weaver, and J. Aizenberg, Nature 531, 78 (2016).
  3. [3]Y.-Q. Luo, L. Liu, L.-R. Zhao, J.-J. and X. Yao, Journal of Materials Chemistry A 10, 19348 (2022).
  4. [4]I. Wong, G. H. Teo, C. Neto, and S. C. Thickett, ACS applied materials & interfaces 7, 21562 (2015).
  5. [5]W. W. Grabowski and L.-P. Wang, Annual review of fluid mechanics 45, 293 (2013).
  6. [6]J. Cho and K. E. Goodson, Nature materials 14, 136 (2015).
  7. [7]C. Ma, L. Chen, L. Wang, W. Tong, C. Chu, Z. Yuan, C. Lv, and Q. Zheng, Nature Communications 13, 5381 (2022).
  8. [8]B. Shin, J. Ha, M. Lee, K. Park, G. H. Park, T. H. Choi, K.-J. Cho, and H.-Y. Kim, Science robotics 3, eaar2629 (2018).
  9. [9]H. Ganapathy, A. Shooshtari, K. Choo, S. Dessiatoun, M. Alshehhi, and M. Ohadi, International Journal of Heat and Mass Transfer 65, 62 (2013).
  10. [10]A. Orazzo and S. Tanguy, International Journal of Heat and Mass Transfer 129, 432 (2019).
  11. [11]B. El Fil, G. Kini, and S. Garimella, International Journal of Heat and Mass Transfer 160, 120172 (2020).
  12. [12]V. Rifert, V. Sereda, V. Gorin, P. Barabash, and A. Solomakha, Heat and mass transfer 56, 691 (2020).
  13. [13]A. Katselas, R. Parin, and C. Neto, Advanced Materials Interfaces 9, 2200246 (2022).
  14. [14]D. Lohse and A. Prosperetti, Proceedings of the National Academy of Sciences 113, 13549 (2016).
  15. [15]S. Ayuba, D. Suh, K. Nomura, T. Ebisuzaki, and K. Yasuoka, The Journal of chemical physics 149, 044504 (2018).
  16. [16]D. Niu and G. Tang, RSC advances 8, 24517 (2018).
  17. [17]D. T. Ranathunga, A. Shamir, X. Dai, and S. O. Nielsen, Langmuir 36, 7383 (2020).
  18. [18]M. Gallo, F. Magaletti, and C. M. Casciola, Phys. Rev. Fluids 3, 053604 (2018).
  19. [19]M. Gallo, F. Magaletti, and C. M. Casciola, Multiphase Flow: Theory and Applications , 347 (2018).
  20. [20]M. Gallo, F. Magaletti, and C. M. Casciola, Journal of Fluid Mechanics 906 (2021).
  21. [21]M. Gallo, F. Magaletti, A. Georgoulas, M. Marengo, J. De Coninck, and C. M. Casciola, Nature Communications 14, 6428 (2023).
  22. [22]M. Gallo and C. M. Casciola, International Journal of Multiphase Flow 179, 104924 (2024).
  23. [23]J. Van der Waals, Amsterdam (Sect. 1) 1, 1 (1893).
  24. [24]J. D. Van der Waals, Journal of Statistical Physics 20, 200 (1979).
  25. [25]L. Landau and E. Lifshitz, Course of theoretical physics 30 (1980).
  26. [26]A. Chaudhri, J. B. Bell, A. L. Garcia, and A. Donev, Physical Review E 90, 033014 (2014).
  27. [27]D. Bandak, N. Goldenfeld, A. A. Mailybaev, and G. Eyink, Physical Review E 105, 065113 (2022).
  28. [28]J. B. Bell, A. Nonaka, A. L. Garcia, and G. Eyink, Journal of Fluid Mechanics 939 (2022).
  29. [29]G. Eyink and A. Jafari, Physical Review Research 4, 023246 (2022).
  30. [30]B. Barker, J. B. Bell, and A. L. Garcia, Proceedings of the National Academy of Sciences 120, e2306088120 (2023).
  31. [31]G. Eyink and A. Jafari, in Annales Henri Poincaré, Vol. 25 (Springer, 2024) pp. 497–516.
  32. [32]D. Anderson, G. McFadden, and A. Wheeler, Annual Review of Fluid Mechanics 30, 139 (1998).
  33. [33]F. Magaletti, M. Gallo, L. Marino, and C. M. Casciola, International Journal of Multiphase Flow 84, 34 (2016).
  34. [34]J. Liu, G. Amberg, and M. Do-Quang, Physical Review E 93, 013121 (2016).
  35. [35]E. Benilov, Journal of Fluid Mechanics 954, A41 (2023).
  36. [36]S. Mukherjee and H. Gomez, Applied Physics Letters 117 (2020).
  37. [37]E. Benilov, Physical Review E 101, 042803 (2020).
  38. [38]F. Magaletti, M. Gallo, and C. M. Casciola, Scientific reports 11, 1 (2021).
  39. [39]J. K. Johnson, J. A. Zollweg, and K. E. Gubbins, Molecular Physics 78, 591 (1993).
  40. [40]M. Gallo, Physics of Fluids 34, 122011 (2022).
  41. [41]F. Balboa, J. B. Bell, R. Delgado-Buscalioni, A. Donev, T. G. Fai, B. E. Griffith, and C. S. Peskin, Multiscale Modeling & Simulation 10, 1369 (2012).
  42. [42]A. Donev, E. Vanden-Eijnden, A. Garcia, and J. Bell, Communications in Applied Mathematics and Computational Science 5, 149 (2010).
  43. [43]S. Delong, B. E. Griffith, E. Vanden-Eijnden, and A. Donev, Physical Review E 87, 033302 (2013).
  44. [44]F. Magaletti, M. Gallo, S. P. Perez, J. A. Carrillo, and S. Kalliadassis, Journal of Computational Physics 463, 111248 (2022).
  45. [45]M. Gallo, F. Magaletti, D. Cocco, and C. M. Casciola, Journal of Fluid Mechanics 883 (2020).
  46. [46]T. Kraska, The Journal of chemical physics 124 (2006).
  47. [47]K. Yasuoka and M. Matsumoto, The Journal of chemical physics 109, 8451 (1998).
  48. [48]T. Kimura and S. Maruyama, Microscale Thermophysical Engineering 6, 3 (2002).
  49. [49]J. Diemand, R. Angélil, K. Tanaka, and H. Tanaka, The Journal of chemical physics 139, 074309 (2013).
  50. [50]L. Inci and R. K. Bowles, The Journal of Chemical Physics 134 (2011).
  51. [51]G. Menzl, M. A. Gonzalez, P. Geiger, F. Caupin, J. L. Abascal, C. Valeriani, and C. Dellago, Proceedings of the National Academy of Sciences 113, 13582 (2016).
  52. [52]E. Weinan, W. Ren, and E. Vanden-Eijnden, Physical Review B 66, 052301 (2002).
  53. [53]J. Langer and L. Turski, Physical Review A 8, 3230 (1973).
  54. [54]M. Teodori, D. Abbondanza, M. Gallo, and C. M. Casciola, 10.5281/zenodo.15662993 (2025), DOI: 10.5281/zenodo.15662993.

Paper details

vlnc:W_yd25t2vy
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents