7. Virtual mass

Summary

  • Subject: (Fluid) particles feels their masses heavier than the actual.

  • Main conclusion: Particle mass is virtually \(\rho_{P} + C_{VM} \rho\), where \(\rho\) is the density of surrounding fluid, and \(C_{VM} = 1/2\) for sphere.

  • Key idea Particles must accelerate surroundings, thereby the accelerating motion gives an inertial force.

  • Reference: Landau and Lifshitz [LL87] (Chap. 1)

In this section, a solid sphere moving in an incompressible inviscid fluid is considered. Therefore, let us introduce the velocity potential to express the fluid velocity \(\mathbf{v}\):

(7.1)\[ \mathbf{v} = \nabla \phi\]

Substituting this expression into the continuity equation

(7.2)\[ \nabla \cdot \mathbf{v} = 0\]

yields the Laplace equation of \(\phi\):

(7.3)\[ \nabla^{2} \phi = 0\]
_images/cvm.png

Fig. 7.1 Spherical body in infinite fluid

The center of the sphere is set at the origin of the spherical coordinates. At positions far from the sphere, we assume that \(\phi\) is a function of \(r\). \(\phi_{0} = 1/r\) obviously satisfies the Laplace equation:

(7.4)\[ \nabla^{2} \phi_{0} \rightarrow \frac{1}{r^{2}} \frac{d}{dr} r^{2} \frac{d \phi_{0}}{dr} = \frac{1}{r^{2}} \frac{d}{dr} r^{2} \left( - \frac{1}{r^{2}} \right) = 0\]

Derivatives of \(\phi_{0}\), can also be solution of the Laplace equation since

(7.5)\[ \nabla^{2} \frac{\partial^{n} \phi_{0}}{\partial x_{i_{1}} \partial x_{i_{2}} \cdots \partial x_{i_{n}}} = \frac{\partial^{n} }{\partial x_{i_{1}} \partial x_{i_{2}} \cdots \partial x_{i_{n}}} \nabla^{2} \phi_{0} = 0\]

where \(n \ge 1\). The velocity component \(\mathbf{v}^{(1)}\) for \(- a \phi_{0}\) is given by

(7.6)\[ \mathbf{v}^{(1)} = \nabla \left( - \frac{\alpha}{r} \right) = \frac{\partial r}{\partial x_{i}} \frac{d}{dr} \left( - \frac{\alpha}{r} \right) = \alpha \frac{x_{i}}{r} \frac{1}{r^{2}} = \frac{\alpha x_{i}}{r^{3}}\]

where \(a\) is a constant. The mass flow rate through the surface of a large sphere \(S_{\infty}\) set at the origin is

(7.7)\[\begin{split} \iint_{S_{\infty}} \rho \mathbf{v}^{(1)} \cdot \mathbf{n} dS &= \iint_{S_{\infty}} \rho \mathbf{v}^{(1)} \cdot \mathbf{e}_{r} dS \rightarrow \rho \alpha \iint_{S_{\infty}} \frac{x_{i} x_{i}}{r^{4}} dS = \rho \alpha \iint_{S_{\infty}} \frac{1}{r^{2}} dS \\ &= 4 \pi \rho \alpha\end{split}\]

where \(\mathbf{n}\) is the outward unit normal to \(S_{\infty}\). However, this must vanish because of the incompressibility; therefore, \(\alpha = 0\). The next candidate is \(\nabla \phi_{0}\). Let us therefore set \(\mathbf{A} \cdot \nabla \phi_{0}\), where \(\mathbf{A}\) is a constant vector. The vector component for this velocity potential is given by

(7.8)\[\begin{split}\begin{split} \mathbf{v}^{(2)} &= \nabla ( \mathbf{A} \cdot \nabla \phi_{0} ) \rightarrow \frac{\partial }{\partial x_{i}} \left( A_{j} \frac{\partial }{\partial x_{j}} \frac{1}{r} \right) = \frac{\partial }{\partial x_{i}} \left( A_{j} \frac{\partial r}{\partial x_{j}} \frac{d}{dr} \frac{1}{r} \right) = A_{j} \frac{\partial }{\partial x_{i}} \left( \frac{x_{j}}{r} \frac{d}{dr} \frac{1}{r} \right) = - A_{j} \frac{\partial }{\partial x_{i}} \left( \frac{x_{j}}{r^{3}} \right) \\ &= - A_{j} \left( \frac{\delta_{ij}}{r^{3}} - x_{j} \frac{3 x_{i}}{r^{5}} \right) = \frac{3 A_{j} x_{j} x_{i}}{r^{5}} - \frac{A_{i}}{r^{3}} \rightarrow \frac{3 ( \mathbf{A} \cdot \mathbf{e}_{r} ) \mathbf{e}_{r} - \mathbf{A}}{r^{3}} \end{split}\end{split}\]

In the far field, velocity components given by the velocity potentials of higher order negligibly small compared with \(\mathbf{v}^{(2)}\). Therefore, in the following, we may write \(\mathbf{v}\) instead of \(\mathbf{v}^{(2)}\).

The velocity component normal to the solid sphere \(v_{r}~(= \mathbf{v} \cdot \mathbf{e}_{r})\) must be equal to that of the sphere \(u_{r}~(= \mathbf{u} \cdot \mathbf{e}_{r})\), i.e. the boundary condition \(( \mathbf{v} - \mathbf{u} ) \cdot \mathbf{e}_{r} = 0\).

(7.9)\[\begin{split}\begin{split} &\left. \left( \frac{3 A_{j} x_{j} x_{i}}{r^{5}} - \frac{A_{i}}{r^{3}} - u_{i} \right) \frac{x_{i}}{r} \right|_{r=a} = 0 \rightarrow \left. \frac{2 A_{i} x_{i} }{r^{4}} - u_{i} \frac{x_{i}}{r} \right|_{r=a} = 0 \rightarrow \left. \left( \frac{2 A_{i}}{r^{3}} - u_{i} \right ) \frac{x_{i}}{r} \right|_{r=a} = 0 \\ &\rightarrow A_{i} = \frac{a^{3} u_{i}}{2} \end{split}\end{split}\]

The total kinetic energy \(E\) of the fluid is given by

(7.10)\[ E = \iiint_{V_{\infty}-V} \frac{\rho v^{2}}{2} dV\]

where \(V_{\infty}\) is the fluid volume of a sphere of large size, while its radius will be taken as infinity later to cover the whole system. \(V\) is the volume of the sphere, \(V = 4 \pi a^{3} / 3\). \(( \mathbf{v} - \mathbf{u} ) \cdot ( \mathbf{v} + \mathbf{u} ) = v^{2} - u^{2}\). Therefore, \(v^{2} = u^{2} + ( \mathbf{v} - \mathbf{u} ) \cdot ( \mathbf{v} + \mathbf{u} )\). The integral of the \(v\) square is therefore rewritten as

(7.11)\[ \iiint_{V_{\infty}-V} \left\{ u^{2} + ( \mathbf{v} - \mathbf{u} ) \cdot ( \mathbf{v} + \mathbf{u} ) \right\} dV = \iiint_{V_{\infty}-V} u^{2} dV + \iiint_{V_{\infty}-V} ( \mathbf{v} - \mathbf{u} ) \cdot ( \mathbf{v} + \mathbf{u} ) dV\]

The first term can be immediately integrated as

(7.12)\[ \iiint_{V_{\infty}-V} u^{2} dV = u^{2} \iiint_{V_{\infty}-V} dV = u^{2} \left( V_{\infty} - V \right)\]

For the second term, by writing \(\mathbf{v} = \nabla \phi\) and \(\mathbf{u} = \nabla ( \mathbf{u} \cdot \mathbf{r} )\), we have

(7.13)\[ \mathbf{v} + \mathbf{u} = \nabla \phi + \nabla ( \mathbf{u} \cdot \mathbf{r} ) = \nabla ( \phi + \mathbf{u} \cdot \mathbf{r} )\]

Then,

(7.14)\[\begin{split}\begin{split} ( \mathbf{v} - \mathbf{u} ) \cdot ( \mathbf{v} + \mathbf{u} ) &= ( \mathbf{v} - \mathbf{u} ) \cdot \nabla ( \phi + \mathbf{u} \cdot \mathbf{r} ) = \nabla \cdot ( \mathbf{v} - \mathbf{u} ) ( \phi + \mathbf{u} \cdot \mathbf{r} ) - ( \phi + \mathbf{u} \cdot \mathbf{r} ) \nabla \cdot ( \mathbf{v} - \mathbf{u} ) \\ &= \nabla \cdot ( \mathbf{v} - \mathbf{u} ) ( \phi + \mathbf{u} \cdot \mathbf{r} ) \end{split}\end{split}\]

The second term in the third term vanishes since \(\nabla \cdot \mathbf{v} = \nabla \cdot \mathbf{u} = 0\). Applying the divergence theorem yields

(7.15)\[\begin{split}\begin{split} &\iiint_{V_{\infty}-V} u^{2} dV + \iiint_{V_{\infty}-V} ( \mathbf{v} - \mathbf{u} ) \cdot ( \mathbf{v} + \mathbf{u} ) dV \\ &= u^{2} \left( V_{\infty} - V \right) + \iiint_{V_{\infty}-V} \nabla \cdot ( \mathbf{v} - \mathbf{u} ) ( \phi + \mathbf{u} \cdot \mathbf{r} ) dV \\ &= u^{2} \left( V_{\infty} - V \right) + \iint_{S_{\infty}} ( \mathbf{v} - \mathbf{u} ) ( \phi + \mathbf{u} \cdot \mathbf{r} ) \cdot \mathbf{e}_{r} dS - \iint_{S} ( \mathbf{v} - \mathbf{u} ) ( \phi + \mathbf{u} \cdot \mathbf{r} ) \cdot \mathbf{e}_{r} dS \end{split}\end{split}\]

where \(S_{\infty}\) is the surface surrounding \(V_{\infty}\). However, \(( \mathbf{v} - \mathbf{u} ) \cdot \mathbf{e}_{r} = 0\) on \(S\).

\[\begin{equation*} \begin{split} u^{2} \left( V_{\infty} - V \right) + \iint_{S_{\infty}} ( \mathbf{v} - \mathbf{u} ) ( \phi + \mathbf{u} \cdot \mathbf{r} ) \cdot \mathbf{e}_{r} dS \end{split} \end{equation*}\]

Expanding the second term gives

(7.16)\[\begin{split}\begin{split} \iint_{S_{\infty}} ( \mathbf{v} - \mathbf{u} ) ( \phi + \mathbf{u} \cdot \mathbf{r} ) \cdot \mathbf{e}_{r} dS &= \iint_{S_{\infty}} \left( \frac{3 ( \mathbf{A} \cdot \mathbf{e}_{r} ) \mathbf{e}_{r} - \mathbf{A}}{r^{3}} - \mathbf{u} \right) \left( \mathbf{A} \cdot \nabla \phi_{0} + \mathbf{u} \cdot \mathbf{r} \right) \cdot \mathbf{e}_{r} dS \\ &\rightarrow \iint_{S_{\infty}} \left( \frac{3 A_{j} x_{j} x_{i} - A_{i} r^{2}}{r^{5}} - u_{i} \right) \left( A_{k} \frac{\partial}{\partial x_{k}} \frac{1}{r} + u_{k} x_{k} \right) \frac{x_{i}}{r} dS \\ &= \iint_{S_{\infty}} \left( \frac{2 A_{j} x_{j}}{r^{3}} - u_{j} x_{j} \right) \left( - \frac{A_{k} x_{k}}{r^{3}} + u_{k} x_{k} \right) \frac{ dS }{r} \\ &= \iint_{S_{\infty}} \left( - \frac{2 ( A_{j} x_{j} )^{2}}{r^{3}} + 3 A_{j} x_{j} u_{k} x_{k} - r^{3} ( u_{j} x_{j} )^{2} \right) \frac{dS}{r^{4}} \end{split}\end{split}\]

Since the radius \(a_{\infty}\) of the sphere \(S_{\infty}\) is large, the first term of the integrand is negligible in comparison with the other two terms, so that,

(7.17)\[\begin{split}\begin{split} \iint_{S_{\infty}} \left( - \frac{2 ( A_{j} x_{j} )^{2}}{r^{3}} + 3 A_{j} x_{j} u_{k} x_{k} - r^{3} ( u_{j} x_{j} )^{2} \right) \frac{dS}{r^{4}} &\rightarrow \iint_{S_{\infty}} \left( 3 A_{j} x_{j} u_{k} x_{k} - r^{3} ( u_{j} x_{j} )^{2} \right) \frac{dS}{r^{4}} \\ &= \frac{3 A_{j} u_{k} - a_{\infty}^{3} u_{j} u_{k}}{a_{\infty}^{2}} \iint_{S_{\infty}} \frac{x_{j} x_{k}}{r^{2}} dS \\ \end{split}\end{split}\]

The integrand of the surface integral is the dyadic \(\mathbf{e}_{r} \mathbf{e}_{r}\).

(7.18)\[\begin{split} \iint_{S_{\infty}} \frac{x_{j} x_{k}}{r^{2}} dS = \iint_{S_{\infty}} \mathbf{e}_{r} \mathbf{e}_{r} dS \end{split}\]

For the \(xx\) component,

(7.19)\[\begin{split} \int_{0}^{2 \pi} \int_{0}^{\pi} ( \sin \theta \cos \varphi )^{2} a_{\infty}^{2} \sin \theta d\theta d\varphi = a_{\infty}^{2} \int_{0}^{\pi} \sin^{3} \theta d\theta \int_{0}^{2 \pi} \cos^{2} \varphi d\varphi = \cdot \frac{4 \pi a_{\infty}^{2}}{3} \end{split}\]

For the \(xy\) component,

(7.20)\[\begin{split} \int_{0}^{2 \pi} \int_{0}^{\pi} ( \sin \theta \cos \varphi ) ( \sin \theta \sin \varphi ) a_{\infty}^{2} \sin \theta d\theta d\varphi = a_{\infty}^{2} \int_{0}^{\pi} \sin^{3} \theta d\theta \int_{0}^{2 \pi} \cos \varphi \sin \varphi d\varphi = 0 \end{split}\]

In summary,

(7.21)\[\begin{split} \iint_{S_{\infty}} \mathbf{e}_{r} \mathbf{e}_{r} dS \rightarrow \frac{4 \pi a_{\infty}^{2}}{3} \delta_{jk} \end{split}\]

Therefore,

(7.22)\[\begin{split} \frac{3 A_{j} u_{k} - a_{\infty}^{3} u_{j} u_{k}}{a_{\infty}^{2}} \iint_{S_{\infty}} \frac{x_{j} x_{k}}{r^{2}} dS = 4 \pi A_{j} u_{j} - \frac{ 4 \pi a_{\infty}^{3} }{3} u_{j} u_{j} = 4 \pi A_{j} u_{j} - V_{\infty} u_{j} u_{j} \end{split}\]

Thus,

(7.23)\[\begin{split} E = \iiint_{V_{\infty}-V} \frac{\rho v^{2}}{2} dV = \frac{\rho}{2} \left( 4 \pi A_{j} u_{j} - V u^{2} \right) = \frac{\rho}{2} ( 4 \pi A_{j} - V u_{j} ) u_{j} \end{split}\]

As we have confirmed (Eq. (7.9)), \(\mathbf{A}\) is proportional to \(\mathbf{u}\). Therefore, we may write

(7.24)\[ P_{j} = m_{ij} u_{i} = 4 \pi \rho A_{j} - \rho V u_{j}\]

where \(\mathbf{P}\) is the total momentum of the fluid, and the symmetric tensor \(m_{ij}\) is referred to as the induced-mass tensor. Thus, the total kinetic energy of the fluid is expressed as

(7.25)\[ E = \frac{1}{2} m_{ij} u_{i} u_{j}\]

By substituting Eq. (7.9) into Eq. (7.24) we obtain

(7.26)\[ m_{ij} u_{i} = ( 2 \pi \rho a^{3} - \rho V ) u_{j}\]

Hence,

(7.27)\[ m_{ij}= \rho \frac{2 \pi a^{3}}{3} \delta_{ij} = \frac{1}{2} \rho V \delta_{ij}\]

For the sphere, the induced mass is the half of the fluid mass removed by the sphere.

Consider a spherical particle oscillating in a fluid under the action of an external force \(\mathbf{f}\), which is the source of the change in the total momentum in the system, i.e., the momentum of the fluid and that of the particle. Therefore,

(7.28)\[ \frac{d ( M \mathbf{u} + \mathbf{P} )}{dt} = \mathbf{f}\]

where \(M\) is the mass of the particle. Substituting Eq. (7.24) yields

(7.29)\[ \left( M \delta_{ij} + m_{ij} \right) \frac{d u_{j}}{dt} = f_{i}\]

We then consider a spherical particle in a fluid under oscillation. If this particle was the fluid, the momentum of this volume is \(\rho V \mathbf{v}\) and the force acting on the fluid particle of volume \(V\) is given by

(7.30)\[ \rho V \frac{d v_{i}}{dt}\]

where the particle is assumed to be much smaller than the length scale for the spatial change in \(\mathbf{v}\). This volume is actually the particle and may have velocity \(\mathbf{u}\) different from the fluid velocity \(\mathbf{v}\). We therefore need to account for the force due to the relative motion \(\mathbf{u} - \mathbf{v}\):

(7.31)\[ - m_{ij} \frac{d (u_{j} - v_{j})}{dt} \]

where the negative sign is added since this is a force acting on the particle as the reaction.

When the relative velocity \(\mathbf{u} - \mathbf{v}\) increases, the force acts on the sphere so as to retard the relative velocity.

The equation of motion of the particle is therefore given by

(7.32)\[ M \frac{d u_{i}}{dt} = \rho V \frac{d v_{i}}{dt} - m_{ij} \frac{d (u_{j} - v_{j})}{dt} \]

The second term on the R.H.S. is the so-called virtual (added) mass force. By making use of Eq. (7.27), we have

(7.33)\[ M \frac{d u_{i}}{dt} = \rho V \frac{d v_{i}}{dt} - \frac{1}{2} \rho V \frac{d (u_{i} - v_{i})}{dt} \]

Denoting the particle density as \(\rho_{P}~(=M/V)\), one can write

(7.34)\[ \left( \rho_{P} + \rho C_{VM} \right) \frac{d u_{i}}{dt} = \rho \left( 1 + C_{VM} \right) \frac{d v_{i}}{dt} \]

where \(C_{VM} = 1/2\) is the virtual mass coefficient of the sphere.

The virtual mass coefficient is a function of particle shape. According to Tomiyama [Tom04], for an oblate spheroild moving vertically [Lam45],

(7.35)\[\begin{split}C_{VM} = \left( \begin{array}{ccc} C_{VM}^{H} &0 &0 \\ 0 &C_{VM}^{H} &0 \\ 0 &0 &C_{VM}^{V} \end{array} \right)\end{split}\]

where

\[C_{VM}^{V} = \chi^{2} \frac{\sec^{-1} \chi - \sqrt{\chi^2 - 1}}{\sqrt{\chi^2 - 1} - \chi^2 \sec^{-1} \chi} \]

and

\[C_{VM}^{H} = \frac{\chi^2 \sec^{-1} \chi - \sqrt{\chi^2 - 1}}{(2 \chi^2 - 1) \sqrt{\chi^2 - 1} - \chi^2 \sec^{-1} \chi} \]

For a prolate ellipsoids, see Tomiyama [Tom04].

_images/cvm1.png

Fig. 7.2 Virtual mass coefficients, \(C_{VM}^{V}\) and \(C_{VM}^{H}\), of ellipsoidal body.