IPhO 2021 · Theory problem 1

Planetary Physics

This problem consists of two independent problems related to the interior of planets. The effects of the surface curvature of the planets can be neglected. You might need to use the formula

\[ (1+x)^\varepsilon\approx 1+\varepsilon x,\quad \text{when }|x|\ll 1. \tag{1} \]

Part A. Mid-ocean ridge (5.0 points)

Consider a large vessel of water that is situated in a uniform gravitational field with free-fall acceleration \(g\). Two vertical rectangular plates parallel to each other are fitted into the vessel so that the vertical edges of the plates are in a tight gap-less contact with the vertical walls of the vessel. Length \(h\) of each plate is immersed in water (Fig. 1). The width of the plates along the \(y\)-axis is \(w\), water density is \(\rho_0\).

Figure 1

Figure 1. Parallel plates in water.

Oil of density \(\rho_{\mathrm{oil}}\) (\(\rho_{\mathrm{oil}}<\rho_0\)) is poured into the space between the plates until the lower level of the oil has reached the lower edges of the plates. Assume that plates and vessel edges are high enough for oil not to overflow them. Surface tension and mixing of fluids can be neglected.

A.1 — 0.8pt

What is the \(x\)-component of the net force \(F_x\) acting on the right plate (magnitude and direction)?

Fig. 2 shows a cross-section of a mid-ocean ridge. It consists of overlaying layers of mantle, crust and ocean water. The mantle is composed of rocks that we assume can flow in geological timescales and therefore, in this problem will be treated as a fluid. The thickness of the crust is much smaller than the characteristic length scale in the \(x\)-direction, hence, the crust behaves as a freely bendable plate. To high accuracy, such a ridge can be modeled as a two-dimensional system, without any variation of variables along the \(y\)-axis, which is perpendicular to the plane of Fig. 2. Assume that the ridge length \(L\) along the \(y\)-axis is much larger than any other length introduced in this problem.

At the centre of the ridge the thickness of the crust is assumed to be zero. As the horizontal distance \(x\) from the centre increases, the crust gets thicker and approaches a constant thickness \(D\) as \(x\to\infty\). Correspondingly, the ocean floor subsides by a vertical height \(h\) below the top of the ridge O, which we define as the origin of our coordinate system (see Fig. 2). Water density \(\rho_0\) and temperature \(T_0\) can be assumed to be constant in space and time. The same can be assumed for mantle density \(\rho_1\) and its temperature \(T_1\). The temperature of the crust \(T\) is also constant in time but can depend on position.

It is known that, to high accuracy, the crustal material expands linearly with temperature \(T\). Since water and mantle temperatures are assumed to be constant, it is convenient to use a rescaled version of the thermal expansion coefficient. Then \(l(T)=l_1[1-k_l(T_1-T)/(T_1-T_0)]\), where \(l\) is the length of a piece of crustal material, \(l_1\) is its length at temperature \(T_1\), and \(k_l\) is the rescaled thermal expansion coefficient, which can be assumed to be constant.

Figure 2

Figure 2. Mid-ocean ridge. Note that the \(z\)-axis is pointing downwards.

A.2 — 0.6pt

Assuming that the crust is isotropic, find how its density \(\rho\) depends on its temperature \(T\). Assuming that \(|k_l|\ll 1\), write your answer in the approximate form

\[ \rho(T)\approx \rho_1\left[1+k\frac{T_1-T}{T_1-T_0}\right], \tag{2} \]

where terms of order \(k_l^2\) and higher are neglected. Then, identify constant \(k\).

It is known that \(k>0\). Also, thermal conductivity of the crust \(\kappa\) can be assumed to be constant. As a consequence, very far away from the ridge axis the temperature of the crust depends linearly with depth.

A.3 — 1.1pt

By assuming that mantle and water each behave like an incompressible fluid at hydrostatic equilibrium, express the far-distance crust thickness \(D\) in terms of \(h\), \(\rho_0\), \(\rho_1\), and \(k\). Any motion of the material can be neglected.

A.4 — 1.6pt

Find, to the leading order in \(k\), the net horizontal force \(F\) acting on the right half (\(x>0\)) of the crust in terms of \(\rho_0\), \(\rho_1\), \(h\), \(L\), \(k\) and \(g\).

Suppose that crust is thermally isolated from the rest of the Earth. As a result of heat conduction, the temperatures of the upper and lower surfaces of the crust are going to get closer to each other until the crust reaches thermal equilibrium. Specific heat of the crust is \(c\) and can be assumed to be constant.

A.5 — 0.9pt

By using dimensional analysis or order-of-magnitude analysis, estimate the characteristic time \(\tau\) in which the difference between the upper and lower surface temperatures of the crust far away from the ridge axis is going to approach zero. You can assume that \(\tau\) does not depend on the two initial surface temperatures of the crust.

Part B. Seismic waves in a stratified medium (5.0 points)

Suppose that a short earthquake happens at the surface of some planet. The seismic waves can be assumed to originate from a line source situated at \(z=x=0\), where \(x\) is the horizontal coordinate and \(z\) is the depth below the surface (Fig. 3). The seismic wave source can be assumed to be much longer than any other length considered in this question.

As a result of the earthquake, a uniform flux of the so-called longitudinal P waves is emitted along all the directions in the \(x\)-\(z\) plane that have positive component along the \(z\)-axis. Since the wave theory in a solid is generally complicated, in this problem we neglect all the other waves emitted by the earthquake. The crust of the planet is stratified so that the P-wave speed \(v\) depends on depth \(z\) according to \(v=v_0(1+z/z_0)\), where \(v_0\) is the speed at the surface and \(z_0\) is a known positive constant.

Figure 3

Figure 3. Coordinate system used in part B.

B.1 — 1.5pt

Consider a single ray emitted by the earthquake that makes an initial angle \(0<\theta_0<\pi/2\) with the \(z\)-axis and travels in the \(x\)-\(z\) plane. What is the horizontal coordinate \(x_1(\theta_0)\ne 0\) at which this ray can be detected at the surface of the planet? It is known that the ray path is an arc of a circle. Write your answer in the form \(x_1(\theta_0)=A\cot(b\theta_0)\), where \(A\) and \(b\) are constants to be found.

If you were unable to find \(A\) and \(b\), in the following questions you can use the result \(x_1(\theta_0)=A\cot(b\theta_0)\) as given. Suppose that total energy per unit length of the source released as P waves into the crust during the earthquake is \(E\). Assume that waves are completely absorbed when they reach the surface of the planet from below.

B.2 — 1.5pt

Find how the energy density per unit area \(\varepsilon(x)\) absorbed by the surface depends on the distance along the surface \(x\). Sketch the plot of \(\varepsilon(x)\).

From now on, assume that the waves are instead fully reflected when reaching the surface. Imagine a device positioned at \(z=x=0\) that has the same geometry as the previously considered earthquake source. The device is capable of emitting P waves in a freely chosen angular distribution. We make the device emit a signal with a narrow range of emission angles. In particular, the initial angle the signal makes with the vertical belongs to the interval \([\theta_0-\tfrac12\delta\theta_0,\theta_0+\tfrac12\delta\theta_0]\), where \(0<\theta_0<\pi/2\), \(\delta\theta_0\ll 1\) and \(\delta\theta_0\ll\theta_0\).

B.3 — 2.0pt

At what distance \(x_{\max}\) along the surface from the source is the furthest point that the signal does not reach? Write your answer in terms of \(\theta_0\), \(\delta\theta_0\) and other constants given above.

Official English text of the problem.

Answers

  • A.1 \[ \boxed{F_x=\frac{\rho_0gwh^2}{2} \left(\frac{\rho_0}{\rho_{\mathrm{oil}}}-1\right)} \] The force points in the \(+x\) direction, away from the oil.

  • A.2 \[ \boxed{\rho(T)=\rho_1 \left[1-k_l\frac{T_1-T}{T_1-T_0}\right]^{-3} \approx\rho_1\left[1+3k_l\frac{T_1-T}{T_1-T_0}\right], \qquad k=3k_l.} \]

  • A.3 \[ \boxed{D=\frac{2(\rho_1-\rho_0)h}{\rho_1k}} \] within the first-order density approximation.

  • A.4 The total horizontal force exerted by the fluids is, to leading order, \[ \boxed{F_x\simeq \frac{2gLh^2(\rho_1-\rho_0)^2}{\rho_1k^2},} \] directed away from the ridge, along \(+x\).

  • A.5 \[ \boxed{\tau\sim\frac{\rho_1cD^2}{\kappa} =\frac{4c(\rho_1-\rho_0)^2h^2}{\kappa\rho_1k^2}.} \] This is an order-of-magnitude estimate.

  • B.1 \[ \boxed{x_1(\theta_0)=2z_0\cot\theta_0,\qquad A=2z_0,\qquad b=1.} \]

  • B.2 \[ \boxed{\varepsilon(x)= \frac{2Ez_0}{\pi(x^2+4z_0^2)}.} \] The graph is even, peaks at \(\varepsilon(0)=E/(2\pi z_0)\), reaches half its maximum at \(x=\pm2z_0\), and approaches zero as \(x^{-2}\).

  • B.3 In the narrow-band limit, \[ \boxed{x_{\max}\simeq \frac{2z_0\cos^2\theta_0}{\delta\theta_0}.} \] This is the outer boundary of the last unreached interval along the beam’s propagation direction. An exact expression accounting for the integer number of reflections is derived below.

Solution

A.1

Let \(H\) be the height of the oil column above the lower edges of the plates. At those edges, the oil and water pressures must agree. With atmospheric pressure subtracted,

\[ \rho_{\mathrm{oil}}gH=\rho_0gh, \qquad H=\frac{\rho_0}{\rho_{\mathrm{oil}}}h. \]

A fluid column of height \(a\) and density \(\rho\) exerts a horizontal force

\[ w\int_0^a \rho g(a-u)\,du=\frac12\rho gwa^2 \]

on a vertical plate, where \(u\) measures height above the column’s bottom.

The oil pushes the right plate towards \(+x\); the water pushes it towards \(-x\). Atmospheric pressure cancels between the two sides. Including the oil above the exterior water level gives

\[ F_x=\frac{gw}{2} \left(\rho_{\mathrm{oil}}H^2-\rho_0h^2\right). \]

Substituting \(H\),

\[ \boxed{F_x= \frac{\rho_0gwh^2}{2} \left(\frac{\rho_0}{\rho_{\mathrm{oil}}}-1\right)>0.} \]

Thus the plate is pushed outwards.

A.2

Define the dimensionless temperature difference

\[ q=\frac{T_1-T}{T_1-T_0}. \]

Isotropic expansion changes all three linear dimensions by the same factor. Consequently,

\[ \frac{V(T)}{V(T_1)}=(1-k_lq)^3. \]

Mass is conserved, and the reference density at \(T_1\) is \(\rho_1\), so

\[ \rho(T)=\rho_1(1-k_lq)^{-3}. \]

Using the given expansion and neglecting terms of order \(k_l^2\),

\[ \rho(T)\approx\rho_1(1+3k_lq). \]

Comparison with equation (2) gives

\[ \boxed{k=3k_l.} \]

A.3

Use the downward-pointing \(z\)-axis of Fig. 2. Far from the ridge, the crust extends from \(z=h\) to \(z=h+D\). Its temperature is

\[ T(z)=T_0+(T_1-T_0)\frac{z-h}{D}, \]

and therefore its density is

\[ \rho(z)\approx \rho_1\left[1+k\left(1-\frac{z-h}{D}\right)\right]. \]

Its mean density is

\[ \overline{\rho} =\frac1D\int_h^{h+D}\rho(z)\,dz =\rho_1\left(1+\frac{k}{2}\right). \]

Let \(p_O\) be the common fluid pressure at the ridge crest, where the crust has zero thickness. Hydrostatic equilibrium gives the pressures immediately above and below the distant crust:

\[ p_{\mathrm{top}}=p_O+\rho_0gh, \]

\[ p_{\mathrm{bottom}}=p_O+\rho_1g(h+D). \]

Vertical force balance on a distant horizontal column of crust requires

\[ p_{\mathrm{bottom}}-p_{\mathrm{top}} =\overline{\rho}\,gD. \]

Thus

\[ \rho_1(h+D)-\rho_0h =\rho_1D\left(1+\frac{k}{2}\right). \]

After cancelling the common term \(\rho_1D\),

\[ (\rho_1-\rho_0)h=\frac12\rho_1kD, \]

so

\[ \boxed{D=\frac{2(\rho_1-\rho_0)h}{\rho_1k}.} \]

A.4

We calculate the pressure forces on the actual water–crust and mantle–crust interfaces.

Write the upper and lower crust surfaces as \(z=z_t(x)\) and \(z=z_b(x)\). Their endpoint values are

\[ z_t(0)=z_b(0)=0, \qquad z_t(\infty)=h, \qquad z_b(\infty)=h+D. \]

The fluid pressure fields are

\[ p_w(z)=p_O+\rho_0gz, \qquad p_m(z)=p_O+\rho_1gz. \]

For an interface element of horizontal length \(dx\) and width \(L\), the horizontal component of pressure force contains the surface slope. The upper fluid pushes towards negative \(x\), while the lower fluid pushes towards positive \(x\):

\[ dF_x =L\left[ p_m(z_b)\frac{dz_b}{dx} -p_w(z_t)\frac{dz_t}{dx} \right]dx. \]

The factors from the inclined surface area and the unit normal cancel, so this expression does not require an additional small-slope approximation. Integrating,

\[ \begin{aligned} F_x &=L\int_0^{h+D}(p_O+\rho_1gz)\,dz -L\int_0^h(p_O+\rho_0gz)\,dz\\ &=p_OLD+\frac{gL}{2} \left[\rho_1(h+D)^2-\rho_0h^2\right]. \end{aligned} \]

Equivalently,

\[ F_x= \frac12\rho_1gLD^2 +(p_O+\rho_1gh)LD +\frac12(\rho_1-\rho_0)gLh^2. \]

From A.3, \(D\propto k^{-1}\) at fixed \(h\). With the background pressure fixed, these three contributions are respectively of orders \(k^{-2}\), \(k^{-1}\), and \(k^0\). Hence the leading term is

\[ F_x\simeq\frac12\rho_1gLD^2. \]

Substituting the expression for \(D\),

\[ \boxed{ F_x\simeq \frac{2gLh^2(\rho_1-\rho_0)^2}{\rho_1k^2}. } \]

It points along \(+x\), away from the ridge. This is the total force exerted by the fluids; no lithostatic force on an artificial cut through the distant crust has been subtracted.

A.5

To leading order in \(k\), the crust has volumetric heat capacity \(\rho_1c\), giving thermal diffusivity

\[ \chi=\frac{\kappa}{\rho_1c}. \]

Far from the ridge, the relevant distance over which heat must diffuse is \(D\). The heat equation,

\[ \frac{\partial T}{\partial t} =\chi\frac{\partial^2T}{\partial z^2}, \]

therefore gives the scale estimate

\[ \frac{\Delta T}{\tau}\sim \chi\frac{\Delta T}{D^2}. \]

The temperature difference cancels, as required:

\[ \boxed{\tau\sim\frac{D^2}{\chi} =\frac{\rho_1cD^2}{\kappa}.} \]

Using A.3,

\[ \boxed{ \tau\sim \frac{4c(\rho_1-\rho_0)^2h^2} {\kappa\rho_1k^2}. } \]

Dimensional analysis does not determine a numerical prefactor. The surface temperature difference approaches zero asymptotically rather than vanishing at a finite time.

B.1

In a medium whose wave speed depends only on depth, Snell’s law preserves the horizontal slowness:

\[ \frac{\sin\theta(z)}{v(z)} =\frac{\sin\theta_0}{v_0}, \]

where \(\theta(z)\) is the local angle with the vertical. Since

\[ v(z)=v_0\frac{z+z_0}{z_0}, \]

we obtain

\[ \sin\theta(z) =\frac{z+z_0}{z_0}\sin\theta_0. \]

Define

\[ R=\frac{z_0}{\sin\theta_0}. \]

On the descending part of the ray,

\[ \frac{dx}{dz}=\tan\theta =\frac{z+z_0}{\sqrt{R^2-(z+z_0)^2}}. \]

Integration with \(x=0\) at \(z=0\) gives

\[ x=\sqrt{R^2-z_0^2} -\sqrt{R^2-(z+z_0)^2}. \]

Thus the trajectory belongs to the circle

\[ (x-x_c)^2+(z+z_0)^2=R^2, \]

whose centre has coordinates

\[ x_c=\sqrt{R^2-z_0^2}=z_0\cot\theta_0, \qquad z_c=-z_0. \]

The ray turns at the deepest point of this circle and returns symmetrically to the surface. Its two intersections with \(z=0\) are \(x=0\) and \(x=2x_c\). Therefore

\[ \boxed{x_1(\theta_0)=2z_0\cot\theta_0,} \]

and

\[ \boxed{A=2z_0,\qquad b=1.} \]

The surface speed \(v_0\) affects travel times but cancels from the ray geometry.

B.2

The downward half-plane spans an angular range \(\pi\). Uniform emission therefore assigns energy per unit source length

\[ dE=\frac{E}{\pi}\,d\theta \]

to an angular interval \(d\theta\).

For rays returning at \(x>0\),

\[ x=2z_0\cot\theta, \qquad \left|\frac{dx}{d\theta}\right| =2z_0\csc^2\theta. \]

Energy conservation between the emitted angular interval and its receiving surface strip gives

\[ \varepsilon(x)\,dx =\frac{E}{\pi}|d\theta|. \]

Hence

\[ \varepsilon(x) =\frac{E}{2\pi z_0}\sin^2\theta. \]

Using

\[ \cot\theta=\frac{x}{2z_0}, \qquad \sin^2\theta=\frac{1}{1+x^2/(4z_0^2)}, \]

we find

\[ \boxed{\varepsilon(x) =\frac{2Ez_0}{\pi(x^2+4z_0^2)}.} \]

Left-going emission gives the same expression for \(x<0\). Its normalization is

\[ \int_{-\infty}^{\infty}\varepsilon(x)\,dx =\frac{E}{\pi} \left[\arctan\frac{x}{2z_0}\right]_{-\infty}^{\infty} =E. \]

For the requested sketch: the curve is smooth and symmetric about \(x=0\), with a single maximum \(E/(2\pi z_0)\). At \(x=\pm2z_0\) its height is \(E/(4\pi z_0)\), so its full width at half maximum is \(4z_0\). It remains positive at every finite \(x\), and its tails satisfy

\[ \varepsilon(x)\sim\frac{2Ez_0}{\pi x^2} \qquad (|x|\gg z_0). \]

Since \(E\) is energy per unit source length, \(\varepsilon\) has units of energy per unit area.

B.3

At a specular surface reflection, the vertical component of the ray velocity reverses while its horizontal component is preserved. Each ray therefore repeats the same arc, translated horizontally. Its surface encounters occur at

\[ x_n(\theta)=n\ell(\theta), \qquad \ell(\theta)=2z_0\cot\theta, \qquad n=1,2,\ldots. \]

Let

\[ \theta_\pm=\theta_0\pm\frac{\delta\theta_0}{2}. \]

For a band lying entirely within \(0<\theta<\pi/2\), the shortest and longest distances between successive surface encounters are

\[ \ell_{\min}=2z_0\cot\theta_+, \qquad \ell_{\max}=2z_0\cot\theta_-. \]

Because the beam fills the angular interval, the surface interval reached at the \(n\)-th encounter is

\[ I_n=[n\ell_{\min},\,n\ell_{\max}]. \]

A gap exists between \(I_n\) and \(I_{n+1}\) precisely when

\[ n\ell_{\max}<(n+1)\ell_{\min}, \]

or

\[ n(\ell_{\max}-\ell_{\min})<\ell_{\min}. \]

Put \(\Delta\ell=\ell_{\max}-\ell_{\min}\). Once \(n\Delta\ell\ge\ell_{\min}\), adjacent intervals overlap, and they continue to overlap for every larger \(n\). Consequently, the exact outer boundary of the last gap is

\[ \boxed{ x_{\max} =\ell_{\min} \left\lceil\frac{\ell_{\min}}{\Delta\ell}\right\rceil. } \]

The ceiling accounts for the integer number of surface encounters. The boundary itself is reached; unreached points approach it from below.

For the requested narrow-band approximation,

\[ \ell_0=2z_0\cot\theta_0, \]

and

\[ \Delta\ell \simeq \left|\frac{d\ell}{d\theta}\right|_{\theta_0} \delta\theta_0 = 2z_0\csc^2\theta_0\,\delta\theta_0. \]

At fixed \(\theta_0\), a sufficiently narrow band requires many reflections before the intervals overlap. To leading order, the integer rounding can then be neglected:

\[ x_{\max}\simeq\frac{\ell_0^2}{\Delta\ell}. \]

Therefore

\[ \boxed{ x_{\max}\simeq \frac{(2z_0\cot\theta_0)^2} {2z_0\csc^2\theta_0\,\delta\theta_0} = \frac{2z_0\cos^2\theta_0}{\delta\theta_0}. } \]

The solution was checked against the official solution.