betaFeedback

WorksPhysics and Astronomy

The influence of surface tension in thin-film hydrodynamics: gravity free planar hydraulic jumps

Abstract

Hydraulic jumps in thin films are traditionally explained through gravity-driven shallow-water theory, with surface tension assumed to play only a secondary role via Laplace pressure. Recent experiments, however, suggest that surface tension can be the primary mechanism. In this work we develop a theoretical framework for surface tension driven hydraulic jumps in planar thin-film flows. Starting from the full interfacial stress conditions, we show that the deviatoric component of the normal stress enters at leading order and fundamentally alters the balance. A dominant-balance analysis in the zero-gravity limit yields parameter-free governing equations, which admit a similarity solution for the velocity profile. Depth-averaged momentum conservation then reveals a singularity at unit Weber number, interpreted as the criterion for hydraulic control. This singularity is regularised by a non-trivial pressure gradient at the jump. This work establishes the theoretical basis for surface-tension-driven hydraulic jumps, providing analytical predictions for the jump location and structure.

Introduction

Recently, the role of surface tension in hydrodynamics—particularly for kitchen-sink–scale hydraulic jumps—has been debated Bhagat et al. [2] (2020); Bhagat & Linden (2020); Duchesne & Limat [8]. These jumps, documented since Leonardo da Vinci (Marusic & Broomhall [11]), are classically attributed to gravity. In large-scale open-channel flows, the depth-averaged inviscid (Saint–Venant) equations are hyperbolic, and a hydraulic jump can be interpreted as a shock-like transition connecting supercritical and subcritical states, with criticality associated with the long-wave gravity-wave speed. In the radial setting, however, Bohr et al. showed that the ideal (inviscid) shallow-water equations do not allow a determination of the jump position (Bohr et al. [5]).

Applying this wave-propagation argument to viscous thin films is theoretically ambiguous. The viscous shallow-water equations used to describe these flows are not hyperbolic;

Appendix A

Planar hydraulic-jump geometry

Figure 1. Planar hydraulic-jump geometry.

instead, they contain a parabolic contribution arising from wall shear stress, which plays a key role in setting the dynamics [14, 5]. Consequently, the system does not admit the same characteristic-based notion of information propagation as in the inviscid theory. Nevertheless, Jannes & Rousseaux [9] provided experimental evidence supporting the supercritical/subcritical interpretation in a thin film using silicone oil. Via Mach-cone measurements, they demonstrated unidirectional propagation of surface disturbances in the fast-moving film and identified the hydraulic jump as a sharp boundary separating super- and subcritical regimes. Physically, this is consistent with viscous stresses decelerating the flow while the film thickens: the long-wave gravity-wave speed increases whereas the mean speed decreases, driving the Froude number toward criticality and leading to choking and the jump. A complementary view, developed by Kurihara [10]; ); Tani [13], attributes the jump to boundary-layer separation driven by an adverse gravitational pressure gradient that builds up as the film thickens.

Bhagat et al. [2] challenged the classical picture and proposed a surface-tension–driven mechanism. They argued that hydraulic jumps may be controlled by surface tension, gravity, or a combination of both, and suggested a composite criticality condition based on an energy balance,

We−1+Fr−2=1,(1)\mathrm{We}^{-1}+\mathrm{Fr}^{-2}=1, \tag*{(1)}

where the local Froude and Weber numbers are defined as

Fr2=2u‾2gh,We=ρu‾2hσ.\mathrm{Fr}^{2}=\frac{2\overline{u}^{2}}{gh}, \qquad\mathrm{We}=\frac{\rho\overline{u}^{2}h}{\sigma}.

Here uu is the velocity, hh is the local film thickness, ρ\rho is the density, and σ\sigma is the surface tension. In thin films they further argued that gravity can be subdominant (Fr≫1\mathrm{Fr}\gg1), so that surface tension provides the dominant control.

While much of the debate has centred on circular jumps, the fundamental balance between surface tension and gravity remains unresolved. To separate physical mechanisms from geometric effects, we consider steady, planar, thin-film channel flow. Under conventional shallow-water scalings—normalising the pressure either by a dynamic scale (kinetic-energy density) or by hydrostatics—the leading-order balances are hydrostatic in the wall-normal direction and viscous–gravity in the streamwise direction [1, 7, 12]. For nearly flat films, this would suggest that surface tension does not enter at leading order.

We challenge that inference. We show that, beyond the Laplace pressure, the deviatoric normal stress can enter the leading-order balance and, in a zero-gravity thin-film regime, control the dynamics. We first analyse the zero-gravity limit and develop a theoretical framework for surface-tension–driven hydraulic control and jumps, and then validate the predictions against numerical simulations.

The paper is organised as follows. § 3 interprets criticality via a capillary–gravity wave argument. § 4 states the governing equations and boundary conditions, derives the interfacial conditions without a priori assumptions, and scales them to identify the leading-order terms. § 5 develops the asymptotics: a dominant-balance analysis identifies the characteristic scales and yields a parameter-free system, from which a similarity solution and a depth-integrated momentum equation are obtained. The latter exhibits a singularity at a unit effective local Weber number, which is regularised by pressure. § 7 summarises the findings.

Hydraulic jump as a zero-mode standing wave

We frame the steady hydraulic jump as a wave-pinning problem: a location where a surface wave propagating against the flow is held stationary by the background current. In this sense, the jump may be viewed as a standing wave in the laboratory frame (a zero-mode, ω=0\omega= 0) of the gravity-capillary dispersion relation. Using silicone oil as the working fluid, Jannes & Rousseaux [9] performed Mach cone experiments and identified subcritical and supercritical regions separated by the jump, consistent with a blocking point where the background flow first matches the relevant long-wave propagation speed.

For small free-surface disturbances on a layer of depth hh with a uniform current UU in the streamwise direction, the Doppler-shifted dispersion relation is

(ω−kU)2=(gk+σρk3)tanh⁡(kh),Ω:=ω−kU,(2)(\omega-kU)^2 = \left(gk+\frac{\sigma}{\rho}k^3\right)\tanh(kh), \qquad\Omega:= \omega-kU, \tag*{(2)}

where ω\omega and kk are the laboratory-frame frequency and wavenumber, and Ω\Omega is the intrinsic (fluid-frame) frequency.

Gravity-induced jump: setting σ=0\sigma= 0

In the absence of surface tension and in the shallow-water limit (kh≪1kh \ll1), (2) reduces to (ω−kU)2≃gh k2(\omega-kU)^2 \simeq gh\,k^2, i.e. a non-dispersive relation with phase and group speeds equal to c=ghc=\sqrt{gh}. A steady (zero-mode) standing wave requires ω=0\omega=0, hence Ω=−kU\Omega=-kU and U=∣Ω/k∣=c=ghU=|\Omega/k|=c=\sqrt{gh}, which yields the classical criterion

U2=gh⟺Fr2=1,(3)U^2 = gh \qquad\Longleftrightarrow\qquad\mathrm{Fr}^2 = 1, \tag*{(3)}

with Fr:=U/gh\mathrm{Fr}:=U/\sqrt{gh}.

Surface-tension (capillary)-induced jump: setting g=0g = 0

The capillary part of (2) is dispersive: the intrinsic phase speed cp0(k):=Ω/kc_{\mathrm{p}0}(k):=\Omega/k increases with kk. In deep water (kh≫1kh\gg1) this scales as cp0∼(σ/ρ)kc_{\mathrm{p}0}\sim\sqrt{(\sigma/\rho)k} and therefore becomes unbounded as k→∞k\to\infty. If one considers arbitrarily short capillary ripples, this appears to imply that for any given mean flow UU there always exists a sufficiently short wave that can propagate upstream, so the flow would never be truly “supercritical” with respect to all capillary modes.

The resolution is that hydraulic control concerns depth-averaged communication of mass and momentum. Only long waves (in the shallow-water sense) have a substantial depth-averaged communication of mass and momentum. integrated signature; very short capillary ripples are confined to a thin interfacial layer and are strongly damped by viscosity. Accordingly, we restrict attention to the long-wave band

S={k:0<kh≤ψ},ψ=O(1),(4)S=\{k:0<kh\le\psi\},\qquad\psi=O(1), \tag*{(4)}

where the dimensionless cutoff ψ\psi serves as an effective upper limit for hydraulically relevant waves. Besides the formal shallow-water validity (kh≲ψSWkh\lesssim\psi_{\mathrm{SW}}), practical limits arise from (i) excitation: the background inhomogeneity of length LL injects little energy at k≫1/Lk\gg1/L (so ψexc∼h/L\psi_{\mathrm{exc}}\sim h/L), and (ii) dissipation: high-kk waves are strongly damped (viscous decay γ∼2νk2\gamma\sim2\nu k^2), giving an attenuation length ℓdamp≈cg/(2νk2)\ell_{\mathrm{damp}}\approx c_g/(2\nu k^2) that falls rapidly with kk. Modes with ℓdamp\ell_{\mathrm{damp}} shorter than a few wavelengths do not contribute to a steady state. We therefore take

ψ=min⁡{ψSW,ψexc,ψdiss}\psi=\min\{\psi_{\mathrm{SW}},\psi_{\mathrm{exc}},\psi_{\mathrm{diss}}\}

and treat it as an O(1)O(1) constant to be fixed by the flow configuration. Typically ψexc\psi_{\mathrm{exc}} and ψdiss\psi_{\mathrm{diss}} exceed unity (see Appendix § 1), so the effective cutoff is often set by the shallow-water limit and we may take ψ=ψSW\psi=\psi_{\mathrm{SW}}.

Isolating capillarity by setting g=0g=0, the dispersion relation is

Ω2=(σρk3)tanh⁡(kh),cp0(k)=Ωk=σρktanh⁡(kh).(5)\Omega^2=\left(\frac{\sigma}{\rho}k^3\right)\tanh(kh),\qquad c_{p0}(k)=\frac{\Omega}{k}=\sqrt{\frac{\sigma}{\rho}k\tanh(kh)}. \tag*{(5)}

A zero-mode (standing wave in the laboratory frame) is obtained by imposing ω=0\omega=0, which gives Ω=−kU\Omega=-kU and therefore the phase-speed matching condition

U=cp0(k)⟺U2=σρktanh⁡(kh).(6)U=c_{p0}(k)\quad\Longleftrightarrow\quad U^2=\frac{\sigma}{\rho}k\tanh(kh). \tag*{(6)}

Within the admissible long-wave band SS, cp0(k)c_{p0}(k) monotonically increases with kk, so the last steady long wave to survive is the fastest admissible long wave at the band edge kmax⁡=ψ/hk_{\max}=\psi/h. Evaluating (6) at kmax⁡k_{\max} yields the edge-of-band threshold

Ucrit2=σρψhtanh⁡ψ,(g=0, steady zero-mode).(7)U_{\mathrm{crit}}^2=\frac{\sigma}{\rho}\frac{\psi}{h}\tanh\psi,\qquad(g=0,\ \text{steady zero-mode}). \tag*{(7)}

We define the local wave-Weber number based on the band-edge wavenumber kmax⁡=ψ/hk_{\max}=\psi/h,

We:=ρU2hσψ2,(8)\mathrm{We}:=\frac{\rho U^2h}{\sigma\psi^2}, \tag*{(8)}

for which (7) becomes

Wecrit(steady)=tanh⁡ψψ.(9)\mathrm{We}_{\mathrm{crit}}^{(\mathrm{steady})}=\frac{\tanh\psi}{\psi}. \tag*{(9)}

Thus, at g=0g=0, a steady capillary-controlled jump occurs when UU equals the phase speed of the fastest admissible long capillary wave (k=ψ/hk=\psi/h).

Gravity-capillary induced jump (completed picture)

Combining gravity and surface tension in the steady setting, we use the zero-mode condition ω=0\omega=0 and the admissible long-wave band 0<kh≤ψ0<kh\le\psi to obtain

Ucrit2=(ghψ+σρψh)tanh⁡ψ⇒1Fr2+1We=ψtanh⁡ψ≈1,(10)U_{\mathrm{crit}}^2=\left(\frac{gh}{\psi}+\frac{\sigma}{\rho}\frac{\psi}{h}\right)\tanh\psi \quad\Rightarrow\quad \frac{1}{\mathrm{Fr}^2}+\frac{1}{\mathrm{We}}=\frac{\psi}{\tanh\psi}\approx1, \tag*{(10)}

where Fr:=U/gh\mathrm{Fr} := U/\sqrt{gh} and We\mathrm{We} is defined in (8). Criticality corresponds to satisfying (10). Rearranging (10) yields

Fr2≈1+1Bo,(11)\mathrm{Fr}^{2} \approx1 + \frac{1}{\mathrm{Bo}}, \tag*{(11)}

with the Bond number defined as

Bo:=ρgh2σψ2=h2ψ2lc2,lc:=σρg.(12)\mathrm{Bo} := \frac{\rho g h^{2}}{\sigma\psi^{2}} = \frac{h^{2}}{\psi^{2}l_{c}^{2}}, \qquad l_{c} := \sqrt{\frac{\sigma}{\rho g}}. \tag*{(12)}

Thus, for Bo≫1\mathrm{Bo} \gg1 (thick films, h≫ψlch \gg\psi l_{c}) the jump is gravity-controlled with Fr≈1\mathrm{Fr} \approx1; whereas for Bo≪1\mathrm{Bo} \ll1 (thin films) the capillary term sets the threshold, which is equivalently captured by We≈1\mathrm{We} \approx1.

This standing wave viewpoint shows that capillarity can, in principle, set hydraulic control even within an inviscid wave-propagation framework.

Theory

We derived the composite criterion for the thin-film hydraulic jump by interpreting the jump as a steady (‘zero-mode’) standing wave. We now examine the role of surface tension in such films—a point of recent debate— using momentum equation and quantify the conditions under which capillarity, rather than gravity, sets the control.

We consider a 2-D, steady, incompressible, Newtonian thin film with constant properties. Gravity is set to zero in this section to expose the mechanism cleanly. Cartesian coordinates are used with xx streamwise and zz wall-normal; the velocity is u=(u,w)\mathbf{u} = (u,w). The volumetric flux per unit span is

Q=∫0h(x)u(x,z) dz.(13)Q = \int_{0}^{h(x)} u(x,z)\,\mathrm{d}z. \tag*{(13)}

The steady incompressible governing equations (in conservative form) are

∂xu+∂zw=0,(14)\partial_x u + \partial_z w = 0, \tag*{(14)}
u∂x(u)+w∂z(u)=−1ρ∂xp+ν(∂xxu+∂zzu),(15)u\partial_x(u) + w\partial_z(u) = -\frac{1}{\rho}\partial_x p + \nu(\partial_{xx}u + \partial_{zz}u), \tag*{(15)}
∂x(uw)+∂z(w2)=−1ρ∂zp+ν(∂xxw+∂zzw),(16)\partial_x(uw) + \partial_z(w^2) = -\frac{1}{\rho}\partial_z p + \nu(\partial_{xx}w + \partial_{zz}w), \tag*{(16)}

where ν\nu is the kinematic viscosity.

Interface geometry and boundary conditions

We define the free surface by J(x,z)=z−h(x)=0J(x,z) = z - h(x) = 0 [6]. The unit normal and tangent to the free surface are

n=11+h′2[−h′1],t=11+h′2[1h′],(17)\mathbf{n} = \frac{1}{\sqrt{1+h'^2}}\begin{bmatrix}-h'\\1\end{bmatrix}, \qquad\mathbf{t} = \frac{1}{\sqrt{1+h'^2}}\begin{bmatrix}1\\h'\end{bmatrix}, \tag*{(17)}

and the curvature is

κ=∇⋅n=−h′′(1+h′2)3/2.\kappa= \nabla\cdot\mathbf{n} = -\frac{h''}{(1+h'^2)^{3/2}}.

We denote the rate-of-strain tensor by

eij=12(∂iuj+∂jui)e_{ij} = \frac{1}{2}(\partial_i u_j + \partial_j u_i)

and the Cauchy stress by

T=−pI+2μe\mathbf{T}=-p\mathbf{I}+2\mu\mathbf{e}

with the dynamic viscosity μ=ρν\mu=\rho\nu. We apply no slip and no penetration condition at the wall (z=0z=0):

at the wall: (z=0z=0) u=0u=0; w=0w=0. (3.6)

We assume a clean interface, and negligible gas shear and hence the boundary condition at the free surface z=h(x)z=h(x). The boundary conditions are,

kinematic: w=ush′w=u_{\mathrm{s}}h^{\prime}, where, us(x):=u(x,h)u_{\mathrm{s}}(x):=u(x,h), (3.7a)

zero tangential stress: t⋅e⋅n=0\mathbf{t}\cdot\mathbf{e}\cdot\mathbf{n}=0, (3.7b)

normal stress with surface tension: −p+2μn⋅e⋅n=−σκ-p+2\mu\mathbf{n}\cdot\mathbf{e}\cdot\mathbf{n}=-\sigma\kappa. (3.7c)

For an analytical solution we resolve (3.7b) and (3.7c) further.

Zero tangential stress

Resolved along x−zx-z using (17), (3.7b) becomes

[1−h′21+h′2(∂zu+∂xw)+2h′1+h′2(∂zw−∂xu)]z=h=0.(18)\left[\frac{1-h^{\prime2}}{1+h^{\prime2}}(\partial_z u+\partial_x w)+\frac{2h^{\prime}}{1+h^{\prime2}}(\partial_z w-\partial_x u)\right]_{z=h}=0. \tag*{(18)}

Note that the streamwise axis xx is generally not aligned with the free-surface tangent. The zero–tangential–stress condition applies along the surface tangent t\mathbf{t}, not along xx. Consequently, its projection onto the xx-direction does not enforce ∂zu=0\partial_z u=0 at z=hz=h; a small residual streamwise shear is permitted. In the long-wave regime (∣h′∣=O(ϵ)|h^{\prime}|=O(\epsilon)), this shear is O(ϵ2)O(\epsilon^2) and is negligible at leading order. For completeness—and to keep the analysis explicit—we retain all terms below and discard the O(ϵ2)O(\epsilon^2) contribution only when taking the leading-order limit. Solving (18) for the streamwise shear at the free surface (∂zu+∂xw)(\partial_z u+\partial_x w) gives

(∂zu+∂xw)∣z=h=−2h′1−h′2(∂zw−∂xu)=4h′1−h′2∂xu(using continuity ∂zw=−∂xu).(19)\begin{aligned} \left.(\partial_z u+\partial_x w)\right|_{z=h}=-\frac{2h^{\prime}}{1-h^{\prime2}}(\partial_z w-\partial_x u) \\ &=\frac{4h^{\prime}}{1-h^{\prime2}}\partial_x u\qquad\text{(using continuity }\partial_z w=-\partial_x u\text{)}. \tag*{(19)} \end{aligned}

Equations (18)–(19) are exact for any slope ∣h′∣<1|h^{\prime}|<1 (the long-wave regime).

Normal stress condition

The deviatoric component of normal stress is resolves to

nieijnj=11+h′2[(h′)2∂xu−h′(∂zu+∂xw)+∂zw]z=h,(20)n_i e_{ij}n_j=\frac{1}{1+h^{\prime2}}\left[(h^{\prime})^2\partial_x u-h^{\prime}(\partial_z u+\partial_x w)+\partial_z w\right]_{z=h}, \tag*{(20)}

which, using (19) and continuity, reduces to

nieijnj=−1+h′21−h′2∂xu∣z=h=1+h′21−h′2∂zw∣z=h.(21)n_i e_{ij}n_j=-\frac{1+h^{\prime2}}{1-h^{\prime2}}\left.\partial_x u\right|_{z=h}=\frac{1+h^{\prime2}}{1-h^{\prime2}}\left.\partial_z w\right|_{z=h}. \tag*{(21)}

For nearly flat films (∣h′∣≪1|h^{\prime}|\ll1) this simplifies to

nieijnj∣z=h→∂zw∣z=h=−∂xu∣z=h.(22)\left.n_i e_{ij}n_j\right|_{z=h}\to\left.\partial_z w\right|_{z=h}=-\left.\partial_x u\right|_{z=h}. \tag*{(22)}

Thus, even at leading order, the deviatoric component of normal stress contribute to the normal stress balance; only in the inviscid limit μ→0\mu\to0 removes this contribution, leaving Laplace pressure only.

Scaling

Here we will first carry out a generic nondimensionalisation and later identify the characteristic scales by a dominant-balance argument. Introduce characteristic scales

x=Xx∗,z=δz∗,h=δh∗,u=υu∗,w=δυXw∗.x = Xx^*, \qquad z = \delta z^*, \qquad h = \delta h^*, \qquad u = \upsilon u^*, \qquad w = \frac{\delta\upsilon}{X}w^*.

where δ\delta is the characteristic film thickness, XX the streamwise length scale, and υ\upsilon the velocity scale, and the aspect ratio

ϵ≡δX≪1.\epsilon\equiv\frac{\delta}{X} \ll1.

In zero gravity free surface flow, pressure should be naturally be scaled with capillary pressure; with κ∼δ/X2\kappa\sim\delta/X^2 we set:

p=σδX2p∗.p = \frac{\sigma\delta}{X^2}p^*.

Define the Reynolds and Weber numbers

Re=υδν,We=ρυ2δσ.Re = \frac{\upsilon\delta}{\nu}, \qquad We = \frac{\rho\upsilon^2\delta}{\sigma}.

For notational convenience, we now drop the asterisks; all variables below are dimensionless unless stated otherwise. Dropping asterisks, the nondimensional equations are

∂xu+∂zw=0,(23)\partial_x u + \partial_z w = 0, \tag*{(23)}
∂x(u2)+∂z(uw)=−ϵ2We∂xp+ϵRe∂xxu+1ϵRe∂zzu,(24)\partial_x(u^2) + \partial_z(uw) = -\frac{\epsilon^2}{We}\partial_x p + \frac{\epsilon}{Re}\partial_{xx}u + \frac{1}{\epsilon Re}\partial_{zz}u, \tag*{(24)}
∂x(uw)+∂z(w2)=−1We∂zp+ϵRe∂xxw+1ϵRe∂zzw.(25)\partial_x(uw) + \partial_z(w^2) = -\frac{1}{We}\partial_z p + \frac{\epsilon}{Re}\partial_{xx}w + \frac{1}{\epsilon Re}\partial_{zz}w. \tag*{(25)}

The (dimensionless) flux is

∫0hu dz=F,F≡Qυδ.(26)\int_0^h u\,\mathrm{d}z = F, \qquad F \equiv\frac{Q}{\upsilon\delta}. \tag*{(26)}

Scaled boundary condition

At the wall, we have u=0u = 0; w=0w = 0, and the kinematic condition at free surface is w=ush′w = u_s h'. The resolved tangential condition (18) yields, to O(ϵ2)O(\epsilon^2),

∂zu=−ϵ2[∂xw+2h′1−ϵ2h′2(∂zw−∂xu)]z=h,(27)\partial_z u = -\epsilon^2\left[\partial_x w + \frac{2h'}{1-\epsilon^2 h'^2}(\partial_z w-\partial_x u)\right]_{z=h}, \tag*{(27)}

so the streamwise free-surface shear is O(ϵ2)O(\epsilon^2). The normal-stress condition gives

−pWe+2ϵRe∂zw=h′′We+O(ϵ2)(z=h).(28)-\frac{p}{We} + \frac{2}{\epsilon Re}\partial_z w = \frac{h''}{We} + O(\epsilon^2) \qquad(z=h). \tag*{(28)}

Writing the free-surface conditions in this resolved, scaled form exposes the hierarchy in ϵ\epsilon, ReRe, WeWe and allows direct substitution into depth-integrated balances.

Asymptotic analysis for the critical regime

Critical-regime scales (capillary control)

We now determine the characteristic scales (X,δ,ν)(X,\delta,\nu) by dominant balances: 1) mass conservation, 2) inertia–viscous balance, and 3) capillary control, appropriate to the critical regime:

mass conservation:Q∼νδ,(29)\text{mass conservation:}\qquad Q \sim\nu\delta, \tag*{(29)}
inertia–viscous balance:u ∂xu∼ν ∂zzu⇒νδ2∼νX,(30)\text{inertia–viscous balance:}\qquad u\,\partial_x u \sim\nu\,\partial_{zz}u \quad\Rightarrow\quad\nu\delta^2 \sim\nu X, \tag*{(30)}
capillary control:We=O(1)⇒ρν2δ∼σ.(31)\text{capillary control:}\qquad\mathrm{We}=O(1) \quad\Rightarrow\quad\rho\nu^2\delta\sim\sigma. \tag*{(31)}

Solving (4.1)–(4.3) for (X,δ,ν)(X,\delta,\nu) yields the unique scales

X=Q3ρσν,δ=Q2ρσ,ν=σρQ.(32)X=\frac{Q^3\rho}{\sigma\nu},\qquad\delta=\frac{Q^2\rho}{\sigma},\qquad\nu=\frac{\sigma}{\rho Q}. \tag*{(32)}

Hence,

Re=Qν=ϵ−1soϵRe=1,andWe=1,(33)\mathrm{Re}=\frac{Q}{\nu}=\epsilon^{-1}\quad\text{so}\quad\epsilon\mathrm{Re}=1,\quad\text{and}\quad\mathrm{We}=1, \tag*{(33)}

Parameter-free leading-order system

With (33) inserted in (3.13)–(3.15) and retaining the dominant long-wave terms and ignoring O(ϵ2)O(\epsilon^2) terms, the governing equations reduce to

∂xu+∂zw=0,(34)\partial_x u+\partial_z w=0, \tag*{(34)}
u ∂xu+w ∂zu=∂zzu,(35)u\,\partial_xu+w\,\partial_zu=\partial_{zz}u, \tag*{(35)}
∂x(uw)+∂z(w2)=−∂zp+∂zzw,(36)\partial_x(uw)+\partial_z(w^2)=-\partial_zp+\partial_{zz}w, \tag*{(36)}

the dimensionless flux reduces to

∫0hu dz=1,F≡Qνδ=1,(37)\int_0^h u\,\mathrm{d}z=1,\qquad F\equiv\frac{Q}{\nu\delta}=1, \tag*{(37)}

and the boundary conditions simplifies to

u=w=0(z=0),(38)u=w=0\qquad(z=0), \tag*{(38)}
w=ush′(z=h),(39)w=u_s h'\qquad(z=h), \tag*{(39)}
∂zu=0(z=h),(40)\partial_z u=0\qquad(z=h), \tag*{(40)}
−p+2∂zw=h′′(z=h).(41)-p+2\partial_z w=h''\qquad(z=h). \tag*{(41)}

Similarity solution and depth-averaged relations

Guided by the leading order streamwise momentum equation (35) and the boundary conditions (38) and (39), we adopt a planar similarity profile,

u(x,z)=us(x)f(η),η=zh(x).(42)u(x,z)=u_s(x)f(\eta),\qquad\eta=\frac{z}{h(x)}. \tag*{(42)}

with f(0)=0f(0)=0, f(1)=1f(1)=1, f′(1)=0f'(1)=0. The flux constraint gives

∫0hu dz=1⇒ushI1=1,I1:=∫01f(η) dη.(43)\int_0^h u\,\mathrm{d}z=1\quad\Rightarrow\quad u_s h I_1=1,\qquad I_1:=\int_0^1 f(\eta)\,\mathrm{d}\eta. \tag*{(43)}

From the flux constraint (43) we have ush=constantu_s h = \text{constant}. Using continuity (34) with the ansatz (42) then yields,

w(x,z)=us(x)h′(x)ηf(η).(44)w(x,z)=u_s(x)h'(x)\eta f(\eta). \tag*{(44)}

Substituting (42)–(44) into the streamwise momentum equation (35) and using (ush=1I1=const)(u_s h = \frac{1}{I_1}=\text{const}) yields a separated equation

h2dusdx=f′′(η)f(η)2:=−32c2(45)h^2\frac{\mathrm{d}u_s}{\mathrm{d}x}=\frac{f''(\eta)}{f(\eta)^2}:=-\frac{3}{2}c^2 \tag*{(45)}

Note f′′(η)≤0f''(\eta)\leq0 as the shear stress is maximum at the plate, and f′(1)=0f'(1)=0, and the right-hand side is a constant by separation. Equation (45) gives the same shape function as obtained by Watson [14] for an axisymmetric boundary layer flow, and the shape function can be obtained by setting,

f′′(η)f2(η)=−32c2⟹f′2=c2(1−f3)(46)\frac{f''(\eta)}{f^2(\eta)}=-\frac{3}{2}c^2 \qquad\Longrightarrow\qquad f'^2=c^2(1-f^3) \tag*{(46)}

Equation (46) can be integrated to obtain the shape factor. Since f′≥0f'\geq0, and f(1)=1f(1)=1, we can get c=1.402c=1.402, similarly I1=0.615I_1=0.615 are the same as Watson [14] due to same shape ODE. Using ush=1/I1u_s h=1/I_1, differentiating and combining with (45) gives a constant free-surface slope

ushdhdx=constant=32c2(47)u_s h\frac{\mathrm{d}h}{\mathrm{d}x}=\text{constant}=\frac{3}{2}c^2 \tag*{(47)}
⟹dhdx=I132c2≈1.8(48)\Longrightarrow\qquad\frac{\mathrm{d}h}{\mathrm{d}x}=I_1\frac{3}{2}c^2\approx1.8 \tag*{(48)}

the leading order streamwise momentum equation gives a constant slope for the free surface.

Depth-integrated wall-normal momentum

We integrate (36) across the film to obtain

ddx∫0huw dz=p(0)+h′′−2∂zw∣z=h+[∂zw]0h.(49)\frac{\mathrm{d}}{\mathrm{d}x}\int_0^h uw\,\mathrm{d}z=p(0)+h''-2\left.\partial_z w\right|_{z=h}+\left[\partial_z w\right]_0^h. \tag*{(49)}

Substituting equations (42) and (44) for velocity, we find

∫0huw dz=us2hh′I2,I2:=∫01ηf(η)2 dη≈0.339,(50)\int_0^h uw\,\mathrm{d}z=u_s^2hh'I_2,\qquad I_2:=\int_0^1\eta f(\eta)^2\,\mathrm{d}\eta\approx0.339, \tag*{(50)}
∂zw∣z=h=ush′h,∂zw∣z=0=0.(51)\left.\partial_z w\right|_{z=h}=\frac{u_sh'}{h},\qquad\left.\partial_z w\right|_{z=0}=0. \tag*{(51)}

Using us=1/(I1h)u_s=1/(I_1h) from (43) gives

ddx∫0huw dz=I2I12ddx(h′h)=I2I12(h′′h−h′2h2),(52)\frac{\mathrm{d}}{\mathrm{d}x}\int_0^h uw\,\mathrm{d}z=\frac{I_2}{I_1^2}\frac{\mathrm{d}}{\mathrm{d}x}\left(\frac{h'}{h}\right)=\frac{I_2}{I_1^2}\left(\frac{h''}{h}-\frac{h'^2}{h^2}\right), \tag*{(52)}
∂zw∣z=h=h′I1h2.(53)\left.\partial_z w\right|_{z=h}=\frac{h'}{I_1h^2}. \tag*{(53)}

Substituting into (49) and writing P0:=p(0)P_0 := p(0) yields the film-height ODE

h′′=I2I12h′2h2−h′I1h2+P0I2I121h−1(54)h'' = \frac{\dfrac{I_2}{I_1^2}\dfrac{h'^2}{h^2}-\dfrac{h'}{I_1h^2}+P_0}{\dfrac{I_2}{I_1^2}\dfrac{1}{h}-1} \tag*{(54)}

Equation (54) shows an apparent singularity when I2I121h=1\dfrac{I_2}{I_1^2}\dfrac{1}{h}=1 however, this singularity must be regularised by the wall pressure P0P_0, a Lagrange multiplier enforcing incompressibility. We revisit the scaled form of the denominator, which can be written as,

I2I121h∗≡I2μus∗2h∗We≡I2ρus2hσ.(55)\frac{I_2}{I_1^2}\frac{1}{h^*}\equiv I_2\mu u_s^{*2}h^*\mathrm{We}\equiv I_2\frac{\rho u_s^2h}{\sigma}. \tag*{(55)}

I2ρus2hσI_2\frac{\rho u_s^2h}{\sigma} can also be defined as the local Weber number. Confirming the wave argument; equation (55) shows the shock at the location where the local weber number reaches unity. Hence, we divide the flow that is connected by a jump or a shock into two regimes,

I2I121h∗={>1,Supercritical<1,Subcritical(56)\frac{I_2}{I_1^2}\frac{1}{h^*}= \begin{cases} >1, & \text{Supercritical}\\ <1, & \text{Subcritical} \end{cases} \tag*{(56)}

Comparison with the standing wave result

In equation (8), we set UU as the average velocity of the film. We compare Equation (8) with (55) to find the value of ψ\psi to get

ψ=I1I2≈1.(57)\psi=\frac{I_1}{\sqrt{I_2}}\approx1. \tag*{(57)}

Results

Solving the film-height evolution equation (54) requires knowledge of the pressure field. Our leading-order analysis demonstrated that the streamwise pressure gradient ∂xp\partial_x p is O(ϵ2)O(\epsilon^2); thus, within the slowly varying thin-film region, the pressure may be approximated as constant. However, this assumption holds only in the outer region and breaks down near the jump, where pressure gradients become significant to regularise the flow. Consequently, we distinguish between the outer free-surface flow, where p≈constp\approx\mathrm{const}, and the inner jump region, where the pressure P0P_0 effectively varies to enforce continuity.

With H≡h′\mathcal{H}\equiv h', the depth-integrated model reduces to a dynamical system:

h′=H,H′=N(h,H)D(h),(58)h'=\mathcal{H},\qquad\mathcal{H}'=\frac{N(h,\mathcal{H})}{D(h)}, \tag*{(58)}
N(h,H)=I2I12H2h2−1I1Hh2+P0,(59)N(h,\mathcal{H})=\frac{I_2}{I_1^2}\frac{\mathcal{H}^2}{h^2}-\frac{1}{I_1}\frac{\mathcal{H}}{h^2}+P_0, \tag*{(59)}
D(h)=I2I121h−1.(60)D(h)=\frac{I_2}{I_1^2}\frac{1}{h}-1. \tag*{(60)}

The phase-space structure can be determined by the nullclines obtained by setting N(h,H)=0N(h,\mathcal{H})=0 and D(h)=0D(h)=0. One boundary condition is provided by the similarity slope (48), while the second depends on the initial film height.

The system (5.1)–(5.3) exhibits a critical singularity when the denominator D(h)D(h) vanishes, corresponding to the critical Weber number condition derived in § 5.1. The apparent singularity gives an impression of a breakdown of the model or a shock. However, in the full viscous system, the pressure is not merely a passive variable but acts as a degree of freedom. Specifically, the wall pressure P0P_0 (and its gradient) functions as the Lagrange multiplier that enforces the incompressibility constraint.

We hypothesise that when the flow approaches the singularity (We→1We \to1), the constraint of mass conservation forces the pressure gradient to adjust, thereby regularising the equation. This adjustment allows the flow to negotiate the transition from the supercritical thin film (We>1We > 1) to the subcritical downstream layer (We<1We < 1) through a smooth, albeit rapid, variation in depth—the hydraulic jump. Thus, while the depth-averaged scaling predicts a singularity at We=1We = 1, the complete pressure field resolves it, ensuring a continuous physical solution.

Conclusions

We have presented a theoretical framework for planar hydraulic jumps where surface tension acts as the primary control mechanism. By analysing the gravity-capillary dispersion relation, we showed that even for inviscid fluids, capillarity can in principle set hydraulic control, resulting in the hydraulic jump.

We then analysed the viscous shallow-water equations in the zero-gravity limit and demonstrated that the deviatoric component of normal stress plays a leading-order role in the thin-film momentum balance—a contribution often neglected in classical shallow-water theory.

Our dominant-balance analysis identifies a critical singularity where the local Weber number reaches unity (We=1We = 1), analogous to the Froude number condition in gravity-driven flows. We have shown that this singularity corresponds to the location where the mean flow velocity matches the speed of the fastest admissible capillary wave (the zero-mode), providing a robust criterion for hydraulic control. Furthermore, our analysis suggests that the apparent singularity in the depth-averaged equations is regularised by the pressure field, which acts as a Lagrange multiplier to enforce continuity across the jump. These results establish that surface tension alone is sufficient to pin a hydraulic jump in thin films, challenging the classical gravity-centric view.

Acknowledgements

I would like to express my sincere gratitude to Prof. Paul Linden and Dr. Amir Atufi for their stimulating discussions and suggestions.

Viscous cutoff (order-of-magnitude)

Very short capillary waves are attenuated by viscosity. Balancing the decay rate ∼kν2\sim k_\nu^2 with the capillary frequency ω∼[(σ/ρ)k3]1/2\omega\sim[(\sigma/\rho)k^3]^{1/2} gives kν∼σ/(ρν2)k_\nu\sim\sigma/(\rho\nu^2). For water (σ≈0.072 N m−1\sigma\approx0.072\ \mathrm{N\,m^{-1}}, ρ≈103 kg m−3\rho\approx10^3\ \mathrm{kg\,m^{-3}}, ν≈10−6 m2 s−1\nu\approx10^{-6}\ \mathrm{m^2\,s^{-1}}) this yields kν∼7×107 m−1k_\nu\sim7 \times10^7\ \mathrm{m^{-1}} (λ∼90 nm\lambda\sim90\ \mathrm{nm}), far beyond the long-wave range and irrelevant to the control criterion.

References

  1. [1]Balmforth, Neil J & Mandre, S 2004 Dynamics of roll waves. Journal of Fluid Mechanics 514, 1–33.
  2. [2]Bhagat, Rajesh K, Jha, NK, Linden, PF & Wilson, D Ian 2018 On the origin of the circular hydraulic jump in a thin liquid film. Journal of Fluid Mechanics 851, R5.
  3. [5]Bohr, Tomas, Dimon, Peter & Putkaradze, Vakhtang 1993 Shallow-water approach to the circular hydraulic jump. Journal of Fluid Mechanics 254, 635–648.DOI
  4. [6]Bush, John WM & Aristoff, Jeffrey M 2003 The influence of surface tension on the circular hydraulic jump. Journal of Fluid Mechanics 489, 229–238.DOI
  5. [7]Dhar, Minmoy, Das, Gargi & Prasanta Kumar 2020 Planar hydraulic jumps in thin film flow. Journal of Fluid Mechanics 884, A11.DOI
  6. [8]Duchesne, Alexis & Limat, Laurent 2022 Circular hydraulic jumps: where does surface tension matter? Journal of Fluid Mechanics 937, R2.DOI
  7. [9]Jannes, Gil & Rousseaux, Germain 2012 The circular jump as a hydrodynamic white hole. arXiv preprint arXiv:1203.6505 .arxiv.org/abs/1203.6505
  8. [10]Kurihara, M 1946 On hydraulic jumps. Proceedings of the Report of the Research Institute for Fluid Engineering, Kyusyu Imperial University 3 (2), 11–33.
  9. [11]Marusic, Ivan & Broomhall, Susan 2021 Leonardo da vinci and fluid mechanics. Annual Review of Fluid Mechanics 53 (1), 1–25.DOI
  10. [12]Razis, Dimitrios, Kanelopoulos, Giorgos & van der Weele, Ko 2021 Continuous hydraulic jumps in laminar channel flow. Journal of Fluid Mechanics 915, A8.DOI
  11. [13]Tani, Itiro 1949 Water jump in the boundary layer. Journal of the Physical Society of Japan 4 (4-6), 212–215.DOI
  12. [14]Watson, EJ 1964 The radial spread of a liquid jet over a horizontal plane. Journal of Fluid Mechanics 20 (3), 481–499.DOI

Paper details

vlnc:W_zeezp2k7
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