IPhO 1984 · Theory problem 2

Seiching of lake water

In certain lakes there is a strange phenomenon called “seiching” which is an oscillation of the water. Lakes in which you can see this phenomenon are normally long compared with the depth and also narrow. It is natural to see waves in a lake but not something like the seiching, where the entire water volume oscillates, like the coffee in a cup that you carry to a waiting guest.

Figure 1

In order to create a model of the seiching we look at water in a rectangular container. The length of the container is \(L\) and the depth of the water is \(h\). Assume that the surface of the water to begin with makes a small angle with the horizontal. The seiching will then start, and we assume that the water surface continues to be plane but oscillates around an axis in the horizontal plane and located in the middle of the container.

Create a model of the movement of the water and derive a formula for the oscillation period \(T\). The starting conditions are given in figure above. Assume that \(\xi \ll h\). The table below shows experimental oscillation periods for different water depths in two containers of different lengths. Check in some reasonable way how well the formula that you have derived agrees with the experimental data. Give your opinion on the quality of your model.

Table 1. \(L = 479\ \mathrm{mm}\)

\(h/\mathrm{mm}\) 30 50 69 88 107 124 142
\(T/\mathrm{s}\) 1.78 1.40 1.18 1.08 1.00 0.91 0.82

Table 2. \(L = 143\ \mathrm{mm}\)

\(h/\mathrm{mm}\) 31 38 58 67 124
\(T/\mathrm{s}\) 0.52 0.52 0.43 0.35 0.28

The graph below shows results from measurements in lake Vättern in Sweden. This lake has a length of 123 km and a mean depth of 50 m. What is the time scale in the graph?

Figure 2

The water surface level in Bastudalen (northern end of lake Vättern) and Jönköping (southern end).

Official English text of the problem.

Answers

Problem 2 has no numbered subparts or printed point values. Its requests are answered in their printed order.

  • 2 — Motion and period: A simple model using the centre of mass and hydrostatic end-wall forces gives harmonic oscillations with \[ \boxed{T_0=\frac{\pi L}{\sqrt{3gh}}.} \] An energy model accounting for the variation of horizontal flow speed gives the refined estimate \[ \boxed{T_1=\frac{2\pi L}{\sqrt{10gh}} \simeq1.987\,\frac{L}{\sqrt{gh}}.} \]

  • 2 — Experimental comparison and quality: \(T_0\) underestimates the measured periods by \(10.0\)–\(15.2\%\) in the longer container and \(8.6\)–\(20.0\%\) in the shorter one. The model captures the approximate scaling \(T\propto L/\sqrt h\). The energy estimate agrees more closely: its largest discrepancies are \(7.1\%\) and \(12.4\%\), respectively. Both models have limited validity when the depth becomes comparable with the length.

  • 2 — Lake Vättern: \(T_0\simeq2.8\ \mathrm h\), while \(T_1\simeq3.1\ \mathrm h\). Thus the graph shows approximately three hours per oscillation, or two hours per horizontal tick interval, with about a day and a half represented overall.

Solution

2 — Model of the motion and oscillation period

Let the container have width \(b\), and let the water density be \(\rho\). Measure \(x\) horizontally from the middle of the container. Write the surface height above the bottom as

\[ H(x,t)=h+\eta(x,t),\qquad \eta(x,t)=\frac{2q(t)x}{L}, \qquad -\frac L2\leq x\leq\frac L2. \]

Thus the depths at the left and right ends are \(h-q\) and \(h+q\), respectively. Initially \(q=\xi\).

We neglect viscosity, damping and surface tension. For a long, shallow container, vertical accelerations can initially be neglected and the pressure approximated as hydrostatic. Small amplitude, \(|q|\ll h\), allows a linear oscillation model.

The total mass is

\[ M=\rho bLh, \]

since the volume above the mean surface equals the missing volume below it. The horizontal coordinate \(X\) of the centre of mass is

\[ \begin{aligned} X &=\frac{\rho b}{M} \int_{-L/2}^{L/2}xH(x,t)\,dx\\ &=\frac{1}{Lh}\frac{2q}{L} \int_{-L/2}^{L/2}x^2\,dx\\ &=\boxed{\frac{Lq}{6h}}. \end{aligned} \]

The hydrostatic force exerted by an end wall on water of depth \(H\) has magnitude

\[ b\int_0^H \rho g(H-z)\,dz =\frac12\rho gbH^2. \]

Consequently, the net horizontal force on the water is

\[ \begin{aligned} F_x &=\frac12\rho gb\left[(h-q)^2-(h+q)^2\right]\\ &=-2\rho gbhq. \end{aligned} \]

The force is restoring: when the water is higher on the right, it acts towards the left. Newton’s second law gives

\[ M\ddot X=F_x, \]

or

\[ \rho bLh\frac{L}{6h}\ddot q=-2\rho gbhq. \]

Therefore,

\[ \boxed{\ddot q+\frac{12gh}{L^2}q=0}. \]

This is simple harmonic motion, with

\[ \omega_0=\frac{\sqrt{12gh}}{L}, \qquad \boxed{T_0=\frac{2\pi}{\omega_0} =\frac{\pi L}{\sqrt{3gh}}}. \]

If the illustrated configuration is released from rest,

\[ q(t)=\xi\cos(\omega_0t), \qquad \eta(x,t)=\frac{2\xi x}{L}\cos(\omega_0t). \]

The two ends therefore oscillate in opposite phase. The period is independent of the small amplitude, the width and the density.

An energy estimate accounting for the flow speed

The planar surface is an approximation to the actual motion. We can assess the effect of resolving the internal flow by constructing an energy model.

Assume that the horizontal velocity \(u(x,t)\) is approximately uniform over the depth. Conservation of volume, linearized about the depth \(h\), requires

\[ \frac{\partial\eta}{\partial t} +h\frac{\partial u}{\partial x}=0. \]

Using \(\eta=2qx/L\) and imposing zero horizontal flow through both end walls gives

\[ \boxed{ u(x,t)=\frac{\dot q}{Lh} \left(\frac{L^2}{4}-x^2\right). } \]

The horizontal speed is greatest at the middle and vanishes at the walls. In the shallow-water approximation, the kinetic energy of vertical motion is small compared with that of horizontal motion. Thus

\[ \begin{aligned} K &=\frac12\rho bh \int_{-L/2}^{L/2}u^2\,dx\\ &=\boxed{\frac{\rho bL^3}{60h}\dot q^{\,2}}. \end{aligned} \]

The increase in gravitational potential energy above that of a horizontal surface is

\[ \begin{aligned} U &=\frac{\rho gb}{2} \int_{-L/2}^{L/2}\left[(h+\eta)^2-h^2\right]dx\\ &=\frac{\rho gb}{2}\int_{-L/2}^{L/2}\eta^2\,dx\\ &=\boxed{\frac{\rho gbL}{6}q^2}, \end{aligned} \]

where the term proportional to \(\int\eta\,dx\) vanishes by volume conservation.

Conserving \(K+U\) now gives

\[ \ddot q+\frac{10gh}{L^2}q=0, \]

and hence

\[ \boxed{ T_1=\frac{2\pi L}{\sqrt{10gh}} =\sqrt{\frac65}\,T_0. } \]

This estimate is about \(9.5\%\) longer than the simple centre-of-mass result.

The differing coefficients reflect the approximate descriptions of the motion: the first model uses hydrostatic end forces, while the second uses the energy of a volume-conserving trial flow. The actual free surface generally departs from a plane, so neither coefficient should be regarded as an exact hydrodynamic result.

2 — Comparison with the measurements and quality of the model

Using \(g=9.81\ \mathrm{m\,s^{-2}}\), the predicted periods are:

For \(L=479\ \mathrm{mm}\):

\(h\) (mm) Measured \(T\) (s) \(T_0\) (s) \(T_1\) (s)
30 1.78 1.60 1.75
50 1.40 1.24 1.36
69 1.18 1.06 1.16
88 1.08 0.935 1.02
107 1.00 0.848 0.929
124 0.91 0.788 0.863
142 0.82 0.736 0.806

For \(L=143\ \mathrm{mm}\):

\(h\) (mm) Measured \(T\) (s) \(T_0\) (s) \(T_1\) (s)
31 0.52 0.470 0.515
38 0.52 0.425 0.465
58 0.43 0.344 0.377
67 0.35 0.320 0.350
124 0.28 0.235 0.258

For a quantitative comparison, define the relative discrepancy as

\[ \varepsilon= \frac{T_{\mathrm{model}}-T_{\mathrm{measured}}} {T_{\mathrm{measured}}}. \]

The simple model gives periods systematically below the observations: by \(10.0\)–\(15.2\%\) in the longer container and \(8.6\)–\(20.0\%\) in the shorter one. The energy estimate reduces the maximum absolute discrepancies to \(7.1\%\) and \(12.4\%\).

The approximate scaling can also be checked through the dimensionless quantity

\[ C=\frac{T\sqrt{gh}}{L}. \]

The experimental values range from approximately \(2.02\) to \(2.14\) in the longer container and \(1.98\) to \(2.27\) in the shorter one. These are reasonably close to a common constant. The two model predictions are

\[ C_0=\frac{\pi}{\sqrt3}\simeq1.814, \qquad C_1=\frac{2\pi}{\sqrt{10}}\simeq1.987. \]

Thus the models provide a useful description of the scale and parameter dependence, and resolving the flow speed improves agreement with these data.

Their validity nevertheless has limits. The ratios \(h/L\) range from \(0.063\) to \(0.30\) in the longer container and from \(0.22\) to \(0.87\) in the shorter one. In particular, the deepest measurements in the shorter container fall outside a reliable shallow-water approximation. Vertical motion and departures from the assumed planar surface then matter. No experimental uncertainties are supplied, so the table establishes numerical agreement rather than a statistical validation.

2 — Time scale of the lake measurements

For lake Vättern,

\[ L=1.23\times10^5\ \mathrm m, \qquad h=50\ \mathrm m, \]

giving

\[ \sqrt{gh}\simeq22.1\ \mathrm{m\,s^{-1}}. \]

The simple model predicts

\[ T_0= \frac{\pi(1.23\times10^5)} {\sqrt{3(9.81)(50)}} \simeq1.01\times10^4\ \mathrm s =\boxed{2.8\ \mathrm h}. \]

The energy estimate gives

\[ T_1= \frac{2\pi(1.23\times10^5)} {\sqrt{10(9.81)(50)}} \simeq1.10\times10^4\ \mathrm s =\boxed{3.1\ \mathrm h}. \]

Accordingly, successive peaks in either trace are separated by approximately three hours.

In the reproduced graph, a complete oscillation occupies approximately \(1.5\) horizontal tick intervals. Therefore,

\[ \text{time per tick interval} \approx\frac{3\ \mathrm h}{1.5} \approx\boxed{2\ \mathrm h}. \]

There are roughly twelve oscillations across the record, corresponding to approximately \(36\ \mathrm h\), or a day and a half. The two traces are approximately in opposite phase, as expected for oscillations at opposite ends of the lake.

The solution was checked against the official solution.