Nonsimilar solutions of the viscous shallow water equations governing weak hydraulic jumps

Ratul Dasgupta and Rama Govindarajan Citation: Phys. Fluids 22, 112108 (2010); doi: 10.1063/1.3488009 View online: http://dx.doi.org/10.1063/1.3488009 View Table of Contents: http://pof.aip.org/resource/1/PHFLE6/v22/i11 Published by the American Institute of Physics.

Related Articles
Slip boundary for fluid flow at rough solid surfaces Appl. Phys. Lett. 100, 074102 (2012) Wangs shrinking cylinder problem with suction near a stagnation point Phys. Fluids 23, 083102 (2011) Sidewall and thermal boundary condition effects on the evolution of longitudinal rolls in Rayleigh-BnardPoiseuille convection Phys. Fluids 23, 084101 (2011) Optimal streaks in a FalknerSkan boundary layer Phys. Fluids 23, 024104 (2011) A formula for the wall-amplified added mass coefficient for a solid sphere in normal approach to a wall and its application for such motion at low Reynolds number Phys. Fluids 22, 123303 (2010)

Additional information on Phys. Fluids

Journal Homepage: http://pof.aip.org/ Journal Information: http://pof.aip.org/about/about_the_journal Top downloads: http://pof.aip.org/features/most_downloaded Information for Authors: http://pof.aip.org/authors

Downloaded 14 Mar 2012 to 203.200.35.12. Redistribution subject to AIP license or copyright; see http://pof.aip.org/about/rights_and_permissions

PHYSICS OF FLUIDS 22, 112108 2010

Nonsimilar solutions of the viscous shallow water equations governing weak hydraulic jumps
Ratul Dasgupta and Rama Govindarajana
Engineering Mechanics Unit, Jawaharlal Nehru Centre for Advanced Scientic Research, Jakkur, Bangalore 560064, India

Received 9 March 2010; accepted 9 August 2010; published online 4 November 2010 The steady viscous shallow water equations are often used for the study of hydraulic jumps. We cast these as a single parametric ordinary differential equation with global continuity as a constraint. The solution provides both the local velocity prole and the downstream evolution of the lm height. Moreover this is an exact approach, in contrast with existing approaches which encounter a closure problem and need modeling. There is only one solution which is supercritical initially. This shows a jumplike behavior at a Froude number close to unity, in consonance with predictions of inviscid theory. At low Froude number, it is shown that two solutions are possible, one with a separated prole and one without. Flow downstream of a real hydraulic jump must switch to the second solution, calling into question the validity of the shallow water approach in resolving the region of the switch. A series solution of the velocity prole shows that the rst correction to a streamwise-varying parabolic prole is a quartic term. Circular and planar solutions are qualitatively similar. 2010 American Institute of Physics. doi:10.1063/1.3488009
I. INTRODUCTION

A layer of uid owing horizontally at high speed over a solid often displays an abrupt increase in height, known as a hydraulic jump. In inviscid theory this phenomenon is analogous to a gas-dynamic shock. Just as a shock results when ow transitions from supersonic to subsonic through the speed of sound, a hydraulic jump is said to arise when a lm ow decelerates from supercritical Froude number Fr 1 to subcritical Fr 1 through the small-amplitude surfacegravity wave speed gh for shallow water. The basis for this analogy lies in the transformation of the inviscid unsteady shallow water equations into their gas dynamics counterpart. This transformation was rst shown for the one-dimensional case by Riabouchinsky1 and in two dimensions by Refs. 2 and 3. Other early work of note on the analogy is in Refs. 47, and experiments on hydraulic jumps aimed at understanding shocks may be found in Refs. 811. The reader is also referred to a useful pedagogical summary in Ref. 12 and in Ref. 13. The shock-hydraulic jump analogy is strictly valid only in the inviscid limit. Inviscid shallow water theory however cannot explain why should a hydraulic jump necessarily occur: it only offers two possible solutions for lm thickness at a given spatial location, one subcritical and the other supercritical, but no transition between the two is predicted. Thus all that we may obtain from inviscid theory is that if a jump was to be assumed at a given location, the Froude number must go through unity there. In order to explain a hydraulic jump, it has therefore traditionally been considered important to include the effects of viscosity. Moreover the lm is thin, and in the typical experiment the near-wall vorticity diffuses across the entire thickness much upstream of the jump,14 so
a

Electronic mail: rama@jncasr.ac.in.

viscous effects are likely to inuence the jump dynamics substantially. The study of viscous effects dates back to TaniKurihara15,16 who included a boundary layer type term in the shallow water equations. These equations are then integrated vertically to derive an evolution equation for the lm height, with an assumed self-similar velocity prole. This vertical averaging approach has frequently been used since,14,1720 although what results is an unreal description of the height prole, in the form of a weird turning around instead of a jump.21 The hope that the inclusion of viscous effects by this simplistic approach would rectify matters completely was thus belied. Despite this, the approach has proved useful, for example in estimating the location of standing jumps. In particular, the shock tting technique of Ref. 21 provides a way of transitioning from one arm of the solution to another, thus avoiding the region of turn around. We begin in Sec. II by discussing the origin of this unrealistic behavior. The process of vertical averaging leads to a well-known closure problem. In the standard approach, closure is obtained by the assumption of a self-similar parabolic velocity prole. In a planar geometry this assumption leads to two solutions up to a certain streamwise distance, and none beyond it, while in a circular geometry, we obtain a height prole which is a spiral. There is no immediate resemblance between the viscous spirals and the inviscid supercritical and subcritical solutions. Since these spirals have attracted a lot of attention, we dwell briey in the same section on them, and show how they are related to the inviscid solutions. The main contribution of the paper follows, in Sec. IV. Here we propose a direct approach where, without any assumptions other than those described by the adjectives boundary layer and shallow water , we express the boundary layer shallow water equations as a parametric ordinary differential equation. We refer to this latter equation as the BLSWE. There are several advantages to this ap 2010 American Institute of Physics

1070-6631/2010/22 11 /112108/8/$30.00

22, 112108-1

Downloaded 14 Mar 2012 to 203.200.35.12. Redistribution subject to AIP license or copyright; see http://pof.aip.org/about/rights_and_permissions

112108-2

R. Dasgupta and R. Govindarajan

Phys. Fluids 22, 112108 2010

proach, beginning with the absence of modeling. An appealing feature is that, using global continuity as an additional constraint, the same equation may be used to yield both the velocity prole as a function of the wall-normal coordinate y, as well as the variation of the lm height h in the downstream coordinate x. The solutions are local. At high Froude number, a single solution is obtained, with an increasing lm height, ensuring that we move inexorably toward a Froude number of unity. We then record a denite change in slope resembling a jump in the neighborhood of Fr= 1. This nding in a viscous case emphasizes the relevance of the inviscid Froude number criterion. Downstream of the jump, we obtain two solutions, one of which displays a separated prole. There are only a small number of studies of the boundary layer shallow water equations which are able to obtain realistic height proles. The two most relevant to the present discussion are i the recent work of Refs. 14 and 22, who obtain closure by modeling the velocity prole as a cubic whose coefcients change downstream, rather than a parabolic with constant coefcients. For practical purposes, this model is sufcient. To compare it to the exact solution, we obtain a polynomial expansion of the BLSWE and show that these do not admit cubic proles. For closure, thus, a good model would be a quartic polynomial with no cubic term. In a region of very low Froude number, we will see that a parabolic self-similar prole will sufce. ii The numerical solution of the boundary layer shallow water equations by Ref. 23, who imposed a downstream boundary condition, and found this condition to be important. We compare our results to Ref. 23 and show that there must be a switch from one solution to another in the numerical treatment, which the boundary layer shallow water equations cannot capture. We also compare our results with the experimental results of Ref. 24 and those obtained by the vertical averaging approach of Ref. 14. The existence of multiple solutions downstream of the jump has not been pointed out earlier and it is our surmise that this is a key element in the reattachment of the ow that occurs dowsntream of the jump. Finally, it is to be observed that the shallow water equations restrict us to weak jumps. The location of a jump is dened for a weak jump. A study of strong jumps where streamline curvature causes dispersive effects to become important can be found in a forthcoming study where we derive a low-order equation describing the immediate vicinity of the jump and compare with our numerical simulations.
II. GOVERNING EQUATIONS

and g is the acceleration due to gravity. The volumetric ow rate Q dictates global continuity according to
hx hr

Q=
0

udy,

and

Q=2 r
0

udy,

for planar and circular geometries, respectively. The radial coordinate r is to be used in the latter in the place of the variable x. Note also the different dimensions of Q. Equation 1 follows from NavierStokes equations under the assumption that streamwise variations are small, i.e., h 1, and u/ x u / y. As a consequence, the shear stress boundary condition appears in its simplied form and nonhydrostatic contributions to pressure are neglected. Note that the condition on h 1 restricts us to weak hydraulic jumps. Surfacetension is known to have an inuence on the jump-shape and the upstream height prole. see, e.g., Ref. 25 . It was found in Ref. 25 that the wavy prole upstream of a circular jump can be a result of viscous-inviscid interaction inuenced by surface-tension. Similar effects are also found by us in an ongoing preliminary NavierStokes computation. However for the sake of simplicity the effects of surface-tension are neglected in the present study.

III. THE CLOSURE PROBLEM AND ITS TRADITIONAL RESOLUTION

The closure problem, in a situation where there is none to begin with, owes its origin to the vertical averaging procedure. Integrating Eq. 1 vertically, we obtain d u2 = ghh dx u y ,
y=0

hx 2 where u2 0 u x , y dy. We now have more unknowns than equations, unless we can relate u2 to Q / h. To do this, we must resort to modeling, e.g., by using self-similar, or Pohlhausen proles.16,22 We will see that neither is representative of the actual velocity prole. In the case of planar geometry, with a parabolic selfsimilar velocity prole, the vertically averaged shallow water equations yield19

15/Re dh = . dx 6 5/Fr2

The steady viscous shallow water equations under the boundary layer approximation16,23 are given by u u u 1 + v = gh + x y u y = 0,
y=h x 2

u , y2 1

u y=0 = v y=0 = 0.

Here u and v are the respective velocity components in the coordinates x and y, h dh / dx, is the kinematic viscosity,

The Froude number Fr Q / gh3 1/2 and the Reynolds number Re Q / . While the inviscid limit provides two solutions of constant height at any streamwise location, the viscous Eq. 4 affords an analytical solution19 which turns around at a nite value xmax of the downstream distance, so that there are two solutions for h below xmax and none beyond. This is not realistic, especially downstream, but its connection with the inviscid solution is visible at low x. The counterpart of Eq. 4 Refs. 16 and 21 in a circular geometry is more interesting,

Downloaded 14 Mar 2012 to 203.200.35.12. Redistribution subject to AIP license or copyright; see http://pof.aip.org/about/rights_and_permissions

112108-3

Nonsimilar solutions

Phys. Fluids 22, 112108 2010

1.5

5 0

1 0.5

10

15

The spiral-like solutions are not consistent with the shallow water approximation of small h , as there exist locations where h . As mentioned above, Refs. 14 and 22 proposed a model for resolving the issue. In the next section, with no recourse to modeling, we recast the boundary layer shallow water equations as a parametric ordinary differential equation and introduce a solution procedure. The evolution of the lm height is predetermined here and we are able to obtain realistic jumplike transitions. We thus argue that our procedure, rather than vertical averaging, would be preferred. We also discuss how the downstream conditions affect the description.
IV. THE BOUNDARY LAYER SHALLOW WATER EQUATIONS A. In a planar geometry

FIG. 1. Color online The rolling away of the critical point as the initial Reynolds number Re0 increases from 10 for the leftmost spiral in powers of 10 up to 104 for the righmost spiral. Initial Froude Fr0 = 2. Last curve and inset: inviscid solution.

h(r)

h(r)

5 h r Re0 r dh = . 5 dr 2 3 1 rh 6 Fr2 0

To rewrite Eq. 1 in a form conducive for further analysis, we rst use incompressibility to replace the velocity components in terms of the streamfunction . This is then nondimensionalized as = Qf , , = y , hx d = dx , h x Re

For the circular case, the local Reynolds number is redened as Re0 = Qh0 / r2 and the local Froude number as Fr2 0 0 = Q2 / 4 2r2gh3. Here r and h are nondimensional, r0 and h0 0 0 being scales at some upstream location. It is known16,21 that any solution of this equation spirals into a critical point rc , hc where the slope dh / dr is indeterminate of the form 0/0 .16,21 For inviscid ow, Eq. 5 gives r2 = 1 ch 5h3/3 Fr2 0
2

so that u = Uf , v = U h f f , and U Q / h. For convenience the independent coordinate in the slope h of the interface is retained as the dimensional x, while is used everywhere else. Equation 1 is then f f 0, h Re = 0, 1 f2 = f f Fr2 f 0, = 0, f f 1, f , 9 = 0. Not only must the above equation satisfy the three boundary conditions specied, but to conserve the global mass ow rate, we must also have f 1 , = 1. This provides a constraint using which we may determine h . Referred to above and hereafter as BLSWE, Eq. 9 is the boundary layer shallow water equation in coordinates which lend it to solution as successive ordinary differential equations in and , and to simplication at various limits. First, notice that the Reynolds number, which is constant for a given ow, merely causes a rescaling of x, and can be scaled out of Eq. 9 . Second, since the additional constraint of global mass balance must be satised, h Re is not a free parameter. In fact, given a Froude number, this quantity is completely determined, since only select values of h Re will satisfy f 1 , = 1. Thus, given a starting Froude number, Eq. 9 may be solved not only to obtain the velocity prole at a given streamwise location, but to evolve the height prole of the lm downstream. Since we obtain local velocity proles and slopes, we may march either downstream or upstream to get the height prole. Let us now discuss the high and low Froude number limits. For Fr 1, the rst term within the square bracket in Eq. 9 may be dropped. We have seen that h Re depends on the Froude number alone. With the Froude number no longer present in the equation, h Re in this limit is a constant.

For each c these are U-shaped curves21 whose upper and lower arms are subcritical and supercritical solutions, respectively see inset in Fig. 1 . With increasing viscosity, the lower arm rolls in at decreasing radial locations to give rise to spirals. The viscous upper arm looks like the inviscid one up to a certain radius, beyond which the height drops to negative values very rapidly, consistent with the arguments of Ref. 21. This radius tends to as Re0 . We conclude this section by presenting an analytical solution in the neighborhood of the critical point. This is an appealing feature of a well-known equation but pertains to an unphysical solution. Linearizing Eq. 5 about the critical point, we have dh s 2 , = dr 3s + 2 7

h with s /, = r / rc 1, and = h / hc 1. This may be solved h r r to give = r K 3s + s + 2

2

exp

3 23

tan1

1 + 6s 23

where K is the constant of integration. This is a modied logarithmic spiral, which collapses very sharply into the critical point.

Downloaded 14 Mar 2012 to 203.200.35.12. Redistribution subject to AIP license or copyright; see http://pof.aip.org/about/rights_and_permissions

112108-4

R. Dasgupta and R. Govindarajan

Phys. Fluids 22, 112108 2010

15
200

10

h/hi

hRe
Fr=1

5 0 -5

100

0 0

20

x/(Re hi)

40

60

80

100

0.1

Fr

10

FIG. 2. Color online Variation of lm height h with downstream distance of the P solution of Eq. 9 . Notice the well-known Ref. 22 linear increase in height upstream of the jump. At x 22, Fr 1, and downstream, Fr 1. hi is the lm height at the inlet.

FIG. 3. Color online The slope of the lm height as a function of the local Froude number. Thick line: P solution of Eq. 9 . Thick dashes: N solution for Fr 0.71. The squares are the analytical solution 10 at low Fr. The line with symbols is obtained from Higueras simulations, Fig. 2 of Ref. 23, for his case S = 2, where S is the inverse of Froude number. At high Fr, h Re 1.8138 in agreement with Ref. 23. The thin lines are obtained by setting the right hand side of Eq. 9 to zero.

Setting the right hand side of Eq. 9 to zero would reduce it to the similarity equation of Watson.23,26 Upon solving this and imposing the mass ux condition, it is found that h Re= 1.8138 see Fig. 3 , as in the gravity-free solution of Watson. Thus the height increases linearly at high Fr, qualitatively consistent with experimental ndings,19,22 and the Froude number Fr= Q / g1/2h3/2 decreases as x3/2 with the downstream distance. Thus starting at large Froude number, the solution inevitably evolves toward Fr= 1, where a jump is seen. At very low Froude numbers on the other hand, the second term in the square bracket in Eq. 9 is negligible compared to the rst. Setting the right hand side to zero, we may obtain another self-similar solution in the form of a parabolic velocity prole, since the equation reduces to f = h Re , Fr2 h Re = 3 Fr2, Fr 0. 10

The expression for h Re is obtained by integrating the rst equation above in and using the boundary conditions. Thus the height shows a slight downstream decrease at low Froude numbers. The complete solution to Eq. 9 is now obtained at a given streamwise location. The approach is based on the fact that the BLSWE is derived by neglecting the nonhydrostatic pressure terms, which amounts to neglecting second and higher-order derivatives in and retaining only the rst derivative. This means we may treat the right hand side as a , since for exfunction of alone, i.e., f f f f . If ample, f being negligible implies that f = f were known, then Eq. 9 would be just an inhomogeneous ordinary differential equation in . Since we do not know it a priori we obtain iteratively. The solution procedure uses three embedded iterative loops I1 to I3, and the outer, as discussed below. most loop I3 is designed to obtain As a rst approximation at the beginning of the computation, we set to 0 at every . The middle loop I2 is meant for arriving at the correct interface slope h . We assume an h Re and begin at the innermost loop I1. At = 0, we specify the starting Froude number Fr0. We guess a slope for the velocity prole, f at the wall, and integrate the equation up to the

lm surface. By the NewtonRaphson technique, we iterate our guess for f 0 , until the stress-free boundary condition f 1 , = 0 is satised at the lm surface. We next enter loop I2, where we obtain, again by NewtonRaphson technique, the correct h Re which ensures the satisfaction of global mass conservation. Loop I3 is approached differently. We go to an incrementally downstream location + , use the same right hand side , and repeat loops I1 and I2. With the two neighboring velocity proles, we compute a new , which is usually a better approximation than the previous guess. Repeating this procedure until the right hand side converges forms loop I3. We now have a local solution, as well as a knowledge of the downstream variation of the height prole, which enables us to proceed downstream or upstream as we wish by performing a quadrature in . The entire height prole, as well as the velocity proles at each location, may be obtained to excellent accuracy in a few seconds on a small computer. At high Froude numbers only one solution is obtained, which is denoted here by P since it is of positive h , as seen in Fig. 2. The slope of the height goes through a sudden increase in the neighborhood of Fr= 1, which is the critical value for a jump in the inviscid case. Further, the unphysical turning around of the height prole is absent. Downstream of the jump, in the regime of Fr 0.71, Eq. 9 admits two solutions. In addition to the P solution, we have an N solution, labeled thus to denote that the slope of the height prole is negative in this case. These two are shown in Fig. 3 in terms of h Re versus the Froude number. We now examine which of these solutions will be manifested. Up to Fr= 0.71, the P solution is the only possibility. It appears at rst glance that the P solution would be sufcient at lower Froude as well. However, at low Froude numbers, the P solution constitutes a highly separated velocity prole. The separated region at a Froude number of 0.73 is seen in Fig. 4 to be quite large. As the Froude is further decreased, i.e., as one moves downstream, the separated region becomes larger and larger. The P solution is however

Downloaded 14 Mar 2012 to 203.200.35.12. Redistribution subject to AIP license or copyright; see http://pof.aip.org/about/rights_and_permissions

112108-5

Nonsimilar solutions

Phys. Fluids 22, 112108 2010

1 0.75

4 3
14.4 1.53 0.93 0.73 0.62N

0.25

f(0,)

0.5

2 1 0 -1 -2 0 1

0 0 0.5 1 1.5 2 2.5

FIG. 4. Color online Velocity proles for various Froude numbers, as listed. The curve for Fr= 0.62 is taken from the N solution and is practically parabolic. Note the difference between the parabolic prole and the selfsimilar one at high Froude.

Fr

not separated at Froude numbers above 0.8, i.e., at the jump and upstream of it, indicating a connection between the jump and ow separation. Proles at various Froude numbers are shown in the same gure. Figure 4 also underlines that the assumption of self-similar velocity proles is only reasonable for Froude numbers above 1.5 or so. The N solution on the other hand is seen to be unseparated. It is better and better approximated by a parabolic prole as Fr decreases. Figure 5 shows the shear stress at the wall of the P and N solutions as functions of the Froude number. The rapidly increasing severity of separation in the P solution below Fr= 0.8 is evident. Given their inexional nature, proles with large separation are likely to be unstable. We therefore surmise that at some location downstream of the jump, the P solution becomes untenable, and the ow switches to the N. This is in analogy with the FalknerSkan equations for adverse pressure-gradient boundary layers, which displays, up to a certain value of the pressure-gradient, two such solutions, one separated and very unstable, and the other not, and much less unstable. For the present solutions, a stability study is underway, and initial results indicate that the P solution is extremely unstable at low Froude numbers. Given the highly nonparallel nature of the ow, that study is cumbersome and will be presented separately. The predictions above are consistent with experimental observations, where the ow usually reattaches downstream.14,23,24,27,28 Moreover, downstream of the reattachment point, the slope of the height prole is usually small and negative. The Froude number thus rises again downstream although slowly, as per the behavior of solution N. We compare our results with those obtained in the numerical simulations of Ref. 23. Shown by symbols in Figs. 3 and 5 are the slope of the height prole and the wall-shear stresses that we derive from Ref. 23 as functions of Fr. The extreme right of the curve of Higuera corresponds to the most upstream location, and x increases monotonically as one moves along the line with symbols from this point. For the P solution, the height is a monotonically increasing function of x, and so Fr= Q / g1/2h3/2 decreases monotonically. Higueras curve with increasing x thus traces the P solution from right to left. For the N solution on the other hand, h 0, so the simulated curve turns around and traces it from left to right for increasing x. Notice in Fig. 3 that downstream of the jump, i.e., at

FIG. 5. Color online Wall-normal derivative of streamwise velocity at the wall, f 0 , , as a function of local Froude number. Solid line: P solution. Dashes: N solution. The vertical line at Fr= 1 is provided to guide the eye. It is seen that the P solution ow separates downstream of Fr= 1. The line with symbols is extracted from Ref. 23, Fig. 2, case S = 2.

Fr 1, the solution of Ref. 23 transitions from the P solution to the N solution. The agreement with Higuera in the upstream and downstream regions is a check of our prediction that the slope for a given solution depends on the Froude number alone. It also supports our argument that away from the immediate neighborhood of the jump, the slope may be obtained locally. There is a sudden rise in the slope of the height prole near Fr= 1 in the numerical results too, but the quantitative behavior is different. The switching from the N to the P solution and the detailed behavior in the region of this switching cannot be captured by the BLSWE. The local solution of the BLSWE thus fails in the immediate vicinity of the jump, whereas the numerical procedure of Ref. 23 is able to go through, presumably because of numerically introduced effective higher-order derivatives which are barred in the BLSWE. The question now is whether the complete physics in the immediate vicinity of even a weak jump is contained in the shallow water theory. The present work indicates a negative answer, but does not constitute a proof, since it is possible that more solutions of the BLSWE exist. Further evidence from elaborately designed computations is needed and is being obtained. The divergence of the N solution as it reaches Fr 0.7 in Fig. 3 is consistent with the singularity in the downstream boundary condition of Refs. 14 and 23, where the local Froude number is around 1. Note that both the analytical and the numerical solution reach a Froude number between 0.7 and 0.8 at the most downstream location, where the height rapidly decreases, and its slope appears to diverge in the negative direction. To our knowledge, this is the rst time the existence of multiple solutions downstream of the jump is reported, along with its consequences, including the nding that a self-similar parabolic prole is a good assumption at very low Froude numbers. To estimate how much the right hand side affects the answers, a solution with the right hand side set to zero is shown in the same gure. Except for minor differences in the vicinity of Fr= 1, we nd good agreement. Note that since h Re and Fr vary with x, a neglect of the right hand side does not imply that the solution is self-similar. Thus, al-

Downloaded 14 Mar 2012 to 203.200.35.12. Redistribution subject to AIP license or copyright; see http://pof.aip.org/about/rights_and_permissions

112108-6

R. Dasgupta and R. Govindarajan

Phys. Fluids 22, 112108 2010

6 5 2D 50 10 1

f (0,)

hRe

4 3

1
2D 50 10 1

2 1 0 5 10

-1

0.6 0.8 1 1.2 1.4 1.6 1.8 2

Fr

Fr

FIG. 6. Color online The slope of the lm height as a function of the local Froude number in the circular case. The legend indicates the initial radial location at a Froude number of 100.

FIG. 7. Color online The shear stress at the wall in circular geometry. Flow separation is slightly delayed as compared to the planar case.

f though we have no need to take recourse to it, the solution upstream and downstream of the jump may be obtained to good approximation by a homogeneous ordinary differential equation. Thus, although the BLSWE is a partial differential equation, it is in effect a parametric ordinary differential equation, providing local solutions for a given Fr. Since the solution for a given Froude number is completely specied, this means in particular that we do not need boundary conditions in the streamwise direction. Now, it is well-known that downstream conditions can affect the location of the jump and other behavior upstream, so we must discuss how this may affect present predictions. In most experiments, an obstacle of a certain height H is placed at some downstream location L, which determines the maximum Froude number possible there as Frm = Q / g1/2H3/2 , since the height of the uid interface there must be at least H. In simulations too, Ref. 23 sees an effect of the prescribed downstream condition. If one could make the low Froude number assumption everywhere downstream of the jump, and we impose Fr j = 1 at the jump location x j, we would have, from Eq. 10 , L xj Q g Re 8/3 FrL 1 , 12
2/3 1/3

h Re h 2 f = Re f f 2 + Re h + Fr r

f .

12

There is an additional parameter now, namely, the radial location for a given Froude number. We x the upstream Froude number as 100 and examine the solution for different upstream radii, scaled by the height there. The solutions for the slope of the interface and the wall-shear stress are shown in Figs. 6 and 7, respectively. The N solution for the slope h is not shown since, given that Fr is small and r is large, it looks very much like the planar N solution. The overall behavior of the P solution too is qualitatively the same as the planar case. Some nonmonotonic behavior is possible for the P solution in the circular case when the starting radius is small. This is evident when the Froude number is large, where Re h + h / r = 1.8, so at r = h / 1.8 Re we must have h = 0, i.e., a minimum in the height. Upstream of this radial location the height is a decreasing function of the radius.
V. COMPARISON WITH SELF-SIMILAR, POHLHAUSEN AND EXPERIMENTS

11

where FrL Frm. Since the velocity prole is parabolic in the low Froude number limit, Eq. 11 is unsurprisingly just the scaling of Refs. 19 and 22. If the obstacle height H is large, Eq. 11 would demand a long distance between x j and L, so the jump may be pushed upstream if L is inadequate. When L is long enough and H is not too intrusive, we expect the present predictions to hold, i.e., the downstream conditions not to have a signicant effect. In this case, if the downstream solution involves a sharp turning around, the N solution predicts an FrL 0.7.
B. In a circular geometry

It is instructive to compare predictions obtained from the vertically averaged equations with the self-similar assumptions to the present, more realistic, ones. We return to Eq. 4 and compare it with Eq. 9 . In the limit of high Fr, this equation gives h Re= 5 / 2 whereas the planar BLSWE 9 predicts a value of 1.813. Thus both approaches predict a linear height prole far upstream, with different slopes. Similarly Eq. 5 for the axisymmetric case is rewritten as 5 h dh 2 Re r = , 5 dr 1 6 Fr2

13

The viscous shallow water equation under the boundary layer approximation for the circular geometry, with the same boundary conditions as in Eq. 9 and d = dr / h r , = y / h r is

which gives Re h + h / r = 5 / 2 for high Fr. This quantity is again 1.813 in the circular BLSWE 12 . In the other limit of low Froude number and large radius, both equations give h Re= 3 Fr2, similar to the planar case of Eq. 10 . In a region of low Froude number and large radius, moreover, a parabolic prole is a very good assumption. Note that the limiting case corresponds to the N

Downloaded 14 Mar 2012 to 203.200.35.12. Redistribution subject to AIP license or copyright; see http://pof.aip.org/about/rights_and_permissions

112108-7

Nonsimilar solutions

Phys. Fluids 22, 112108 2010

20 15

Craik et al. 1981 P solution Tani 1949, Bohr et al. 1993 N solution

h/hi

10 5

equations, but only their moments.14 While the averaged equations admit cubic proles with variable coefcients, we will show below that the governing equations themselves do not admit a cubic term in its solution. Writing the streamfunction in the BLSWE as a power series in whose coefcients are functions of , as f , =
j=0

aj

x / (Rei hi)

substituting into the BLSWE 9 , and equating coefcient of each power of , we have a0 = a1 = 0, a3 = h Re , 6 Fr2 a4 = 0.

FIG. 8. Color online Comparison of the P and N solutions of Eq. 12 with the experiments of Ref. 24 and the vertically averaged solution of Refs. 16 and 21. The experimental data extracted from Craik are for the case Q = 18 ml/ s obtained with dyed water as the working uid. The dashed line represents the vertically averaged solution obtained from Eq. 13 .

The remaining coefcients may be obtained from the recursion relation solution whereas in our numerical procedure we are able to obtain a P solution as well for quite small Froude numbers. However, in a real ow, the Froude number may never become too small, since as one proceeds well downstream of a jump, the Froude number increases again. Far downstream, we have seen that both the spiral solution and the present one will give qualitatively the same behavior: of a sudden and sharp decrease in height at some radial location. In Fig. 8 the solution of the circular BLSWE 12 is compared to the experimental observations of Ref. 24. The solution from the vertical averaging procedure is shown as well. The initial condition for both equations was obtained from the experimental data in the gure. The solid line and the dashed-dotted line are our P and N solutions, respectively. It is seen that the height prole is reasonably close to the experimentally observed one both upstream and downstream of the jump. However, some deviation is seen in the vicinity of the jump. This is to be expected since dispersive and surface-tension effects, both of which are neglected here, can signicantly inuence the shape of the jump. Note that the present solution does better than the spiral in this region. Downstream, our N solution does a good job of matching the experimental prole but unlike in the experimental prole, no smooth transition between the P and N solution is achieved. In addition, since we cannot predict where the transition should occur, we have chosen a location which matches well with the experiment. Note that the N solution in Fig. 8 has a very small negative slope. We note that the same experimental data of Ref. 24 was compared by Ref. 25 to their theory. That comparison focused on the precise shape of the jump itself with the jump height as input parameters. The present study on the other hand focuses on the regions away from the jump. In a forthcoming study of strong jumps and the near-jump region, we compare with the work of Ref. 25 in greater detail. We have mentioned that Refs. 14 and 22 assume the with velocity proles to be cubic polynomials in streamwise-varying coefcients. Due to this assumption, the solution obtained does not directly satisfy the governing n n 1 n 2 an
n3

=
p=2

pa p n p 1 anp1 anp1h Re p + 1 anp2 .

All an for n 5 may be written in terms of the hitherto undetermined coefcient a2 and h Re/ Fr2. The global mass balance may now be used to obtain a2. Note that since a4 = 0, the BLSWE does not admit a cubic term in the velocity prole.

VI. CONCLUSION

The BLSWE is cast as a parametric ordinary differential equation at a given streamwise location and a novel solution method is proposed. The only parameters are the local Froude number Fr and a product h Re of the interface slope and the Reynolds number. The second parameter is completely dependent on the rst and may be obtained by imposing global continuity. This enables us to obtain the velocity prole at a given Froude number, and having the obtained the interface slope, march either downstream or upstream to obtain the height prole for a weak hydraulic jump. The jump is shown to occur at Fr 1 even in the viscous case. Downstream of the jump, for Fr 0.71, two solutions are obtained and we show evidence from numerical simulations of Ref. 23 to support our surmise that there is a switch from one solution to the other at a Froude number below 1. Instability of the separated prole is suggested as a reason for this switch. The exact nature of the switch and the behavior in its vicinity is beyond the reach of the BLSWE. We also compare our results with experiments and nd good agreement upstream and downstream. It is also seen that our solution improves over the vertically averaged model. In a numerical study being concluded, dispersive effects are seen to be important in this region.

Downloaded 14 Mar 2012 to 203.200.35.12. Redistribution subject to AIP license or copyright; see http://pof.aip.org/about/rights_and_permissions

112108-8

R. Dasgupta and R. Govindarajan

Phys. Fluids 22, 112108 2010 pp. 446778, see Sec. E on shallow-water waves, as of August 2010, freely available online at http://coe.berkeley.edu/SurfaceWaves/. 13 J. J. Stoker, Water Waves: The Mathematical Theory with Applications Wiley-Interscience, New York, 1957 , Chap. 1. 14 S. Watanabe, V. Putkaradze, and T. Bohr, Integral methods for shallow free-surface ows with separation, J. Fluid Mech. 480, 233 2003 . 15 M. Kurihara, On hydraulic jumps, Proceedings of the Rep. of the Research Institute for Fluid Engg., Kyusyu Imperial University, 1946, Vol. 3, pp. 1133. 16 I. Tani, Water jump in the boundary layer, J. Phys. Soc. Jpn. 4, 212 1949 . 17 J. H. Arakeri and A. J. Rao, On radial lm ow on a horizontal surface and the circular hydraulic jump, J. Indian Inst. Sci. 76, 73 1996 . 18 J. W. M. Bush and J. M. Aristoff, The inuence of surface tension on the circular hydraulic jump, J. Fluid Mech. 489, 229 2003 . 19 S. B. Singha, J. K. Bhattacharya, and A. K. Ray, Hydraulic jump in one-dimensional ow, Eur. Phys. J. B 48, 417 2005 . 20 A. Kasimov, A stationary circular hydraulic jump, the limits of its existence and its gasdynamic analogue, J. Fluid Mech. 601, 189 2008 . 21 T. Bohr, P. Dimon, and V. Putkaradze, Shallow-water approach to the circular hydraulic jump, J. Fluid Mech. 254, 635 1993 . 22 D. Bonn, A. Anderson, and T. Bohr, Hydraulic jumps in a channel, J. Fluid Mech. 618, 71 2009 . 23 F. J. Higuera, The hydraulic jump in a viscous laminar ow, J. Fluid Mech. 274, 69 1994 . 24 A. D. D. Craik, R. C. Latham, M. J. Fawkes, and P. W. F. Gribbon, The circular hydraulic jump, J. Fluid Mech. 112, 347 1981 . 25 R. I. Bowles and F. T. Smith, The standing hydraulic jump: Theory, computations and comparisons with experiments, J. Fluid Mech. 242, 145 1992 . 26 E. J. Watson, The radial spread of a liquid jet over a horizontal plane, J. Fluid Mech. 20, 481 1964 . 27 V. E. Nakoryakov, B. G. Pokusaev, and E. N. Troyan, Impingement of an axisymmetric liquid jet on a barrier, Int. J. Heat Mass Transfer 21, 1175 1978 . 28 F. J. Higuera, The circular hydraulic jump, Phys. Fluids 9, 1476 1997 .

ACKNOWLEDGMENTS

We express our gratitude to Tomas Bohr for his useful comments. We also thank the referees for their comments which helped us improve this paper.
D. Riabouchinsky, Sur lanalogie Hydraulic des Movements dun Fluid Compressible, C. R. Acad. Sci. 195, 998 1932 . 2 E. Preiswerke, Applications of the methods of gas dynamics to water ows with free surfaces: Part I, NACA TM Report No. 934, 1938. 3 E. Preiswerke, Applications of the methods of gas dynamics to water ows with free surfaces: Part II, NACA TM Report No. 935, 1940. 4 H. A. Einstein and E. G. Baird, Progress Report of the Analogy Between Surface Shock Waves on Liquids and Shocks in Compressible Gases California Institute of Technology, Pasadena, CA, 1946 . 5 A. H. Shapiro, An Appraisal of the Hydraulic Analogue to Gas Dynamics Massachusetts Institute of Technology, Guided Missiles Program, Cambridge, MA, 1949 . 6 F. R. Gilmore, M. S. Plesset, and H. E. J. Crossley, Jr., The analogy between hydraulic jumps in liquids and shock waves in gases, Appl. Phys. Berlin 21, 243 1950 . 7 R. F. Harleman Donald, Studies on the validity of the hydraulic analogy to supersonic ow, Massachusetts Institute of Technology AF Technical Report No. 5985, 1950. 8 A. T. Ippen, Gas-wave analogies in open channel ow, Proceedings of the Second Hydraulics Conference, Bulletin 27, Studies in Engineering, University of Iowa, 1943. 9 L. Solomon, A hydraulic analogue study of the Hartmann oscillator phenomenon, J. Fluid Mech. 28, 261 1967 . 10 J. S. Rao, V. V. R. Rao, and V. Sheshadri, Hydraulic analogy for isentropic ow through a nozzle, Def. Sci. J. 33, 97 1983 . 11 M. Carbonaro and V. Van der Haegen, Hydraulic Analogy to Supersonic FlowLab Notes, Euroavia Symposium, Von Karman Institute for Fluid Dynamics, Brussels, Belgium, November 2002, pp. 19. 12 J. V. Wehausen and E. V. Laitone, in Surface Waves, Encyclopedia of Physics Vol. 9, edited by S. Flugge Springer-Verlag, Berlin, 1960 ,
1

Downloaded 14 Mar 2012 to 203.200.35.12. Redistribution subject to AIP license or copyright; see http://pof.aip.org/about/rights_and_permissions