Until now we have worked with the equations for time-dependent flow phenomena. We will now turn to effects which are mathematically very similar, yet of a different physical nature. We will assume a fluidic system as depicted in Fig. 18.10. A fluid is driven by a pressure drop from a reservoir. Depending on the shape of the container, a stable velocity profile will have developed. We will assume this velocity to be constant vav. The fluid is driven into the capillary by the pressure drop. Initially the fluid will have a constant velocity profile which will gradually converge to the Hagen-Poiseuille flow profile given by Eq. 16.36 as
The flow in the regime of conversion from the constant velocity profile to the fully-developed steady-state profile is referred to as entrance flow. Over a certain length zchar the profile slowly changes until it is stabilized to the steady-state profile.
This is another example of a transient and steady-state problem. Here the transient state is not timedependent, but length-dependent on z. We therefore expect a solution in the form of
In the following, we will derive the fluid mechanics of the transient flow for the Hagen-Poiseuille flow. As you will see, the method of solving these problems is identical to the method of solving time-dependent transient problems.
Continuity. As you can see, we have chosen the velocity vav because it allows us to satisfy the continuity, i.e., the amount of fluid transported by the constant velocity profile is identical to the amount of fluid transported by the parabolic flow profile. This is how we initially derived vav (see section 16.3.4).
General Version of the Navier-Stokes Equation. The first thing we need is the modified Navier-Stokes equation. We neglect changes with respect to time, as the entrance effects are not time-dependent, but only dependent on z, which is why we can set . The flow is again driven by a pressure along the z-axis and there are no volume forces involved. Taken together, these assumptions will simplify Eq. 11.40 to
Eq. 18.65 is the Navier-Stokes equation that we need to solve. It is in general form, i.e., it can also be applied for systems which are not radially symmetric.
Neglecting the Convection Term.Eq. 18.65 is still a rather complex equation and we would have a hard time finding a general solution to it. We will therefore make another assumption that is in line with the assumption made for Poiseuille flow. We will consider only flows at a very low Reynolds number Re, i.e., we neglect momentum transport by convection. In most microfluidic systems, this is a valid assumption and it will make our problem significantly easier to solve. For low Reynolds numbers we can drop the entire left side of Eq. 18.65 and are left with
This equation should look familiar - it is the fundamental equation for pressure-driven Poiseuille flow (see Eq. 15.14) when neglecting effects of the volume forces.
Cylindrical Coordinates. We now convert Eq. 18.66 into cylindrical coordinates. For this we need the Laplace (see Eq. 7.99) operator in cylindrical coordinates that once applied to Eq. 18.66, results in the following equation
Please note that we have used the scalar Laplacian as we have converted the velocity vector to the scalar vz, because there is no flow in radial and azimuthal direction, thus vφ = vr = 0. This reduces the vector notation to a single equation of vz. There are no changes of vz in azimuthal direction, but now we need to consider changes along the z-axis. This is in contrast to the fully developed Hagen-Poiseuille flow which we derived in section 16.3 where we assumed a fully developed flow profile and thus neglected changes along the z-axis, i.e., . See section 16.3.2 for details on this assumption. Taking together we simplify Eq. 18.67 to
Eq. 18.68 is as far as we can go with the simplifications. This is the equation that we will need to solve.
We now apply our solution consisting of a steady state and transient solution as given by Eq. 18.64 to Eq. 18.68. In doing so we find
At this point we note that the steady-state solution is not a function of z therefore . Furthermore we already know the steady-state solution to the Poiseuille flow that we derived in section 16.3 and that is given by Eq. 18.63. For Eq. 18.69 we need the derivative with respect to the radius that is given by
We therefore rewrite Eq. 18.69 to
From Eq. 16.34 we know that
which simplifies Eq. 18.70
where we now, once again, have a homogeneous PDE. Please note that this is expected, as the steady-state solution must satisfy the Navier-Stokes equation (see Eq. 18.65) and this solution is effectively derived from this inhomogeneous PDE. This procedure is identical to the procedure we have taken when deriving the solution to the time-dependent transient problem in section 18.3. We now proceed to solving Eq. 18.71 again with a substitution and separation of variables approach.
For Eq. 18.71 we assume a solution of the form
which after inserting into Eq. 18.71 yields the following equation
which yields the two ODEs
The first ODE will look familiar by now if rewritten to
which is a Bessel differential equation according to Eq. 3.44 with and ν = 0 with the solution according to Eq. 3.46
Obviously, we also need to apply the boundary conditions which are given by
where λi is the ith root of the Bessel function Jν (see section 3.2.4.3). We therefore find the general solution for A (r) by
The general solution for Z (z) from Eq. 18.73 can be simply taken from Tab. 8.2 as
We have one initial condition that will solve for one of the unknowns of Eq. 18.75. The second we can derive by stating that we expect the effects of z on the flow profile to diminish as z increases. This is due to the fact that the flow profile will eventually become fully-developed and thus independent of z. As for z → ∞ the first term of Eq. 18.75 will become large and we know that in order for the solution to be physically meaningful.
We now have our general solution given by
for which we obviously still have to determine the missing constants ci. These we can derive by the initial conditions.
We know that for z = 0 the overall velocity profile must be equal to the constant vav. Therefore
from which we derive
We now have to compare coefficients in order to derive ci.
In order to find ci we have to perform a Fourier-Bessel series expansions (see section 4.4). We know that the coefficients are determined by Eq. 4.51 as
Fortunately we have already derived the two terms of the integral during an introductory example in section 4.4.2.1. Using Eq. 4.58 we can immediately write down the first integral as
Using Eq. 4.57 we can write down the integral for the second term as
Summing up Eq. 18.80 and Eq. 18.81 we find
Here we can apply the recurrence relation for expression the Bessel function of order 2, as Bessel functions of lower order according to Eq. 3.60 from which we find for ν = 1
in which case we can rewrite Eq. 18.82 to
where we have used the fact that λi are the roots of J0 and thus J0 (λi) = 0. We can now write down the missing coefficients ci according to Eq. 18.79 as
Eq. 18.84 gives us the coefficients we are seeking.
Combining Eq. 18.84 and Eq. 18.76 gives us the transient solution as
Combining Eq. 18.65, Eq. 18.63, and Eq. 18.85 gives us the overall solution of the velocity profile of the transient Hagen-Poiseuille entry profile as
Visualization.Fig. 18.11 visualizes Eq. 18.86 as the normalized velocity profile for different values of z. As you can see, for z = 0 the profile starts out as a constant velocity profile with values vav and gradually evolves into the parabolic profile known from Hagen-Poiseuille flow. For there is no change in the profile anymore and the velocity profile can be considered as being fully-developed. Therefore we can assume that a fully developed profile will be obtained for .
As stated initially, we have made one major simplification during the derivation of Eq. 18.86: We assumed the Reynolds number to be so small that we could neglect all momentum flow due to convection. Obviously, this is only true for a microfluidic system at very low flow speeds. Obviously, we are interested in the question if our solution can also be used for predicting the entrance flow in systems with higher Reynolds number. Referring to Eq. 18.65, we would retain the term on the left-hand side of the equation when taking into account momentum transport by convection. If this term remains in the equation, momentum is transported into and out of our control volume via convection. This in turn will make the momentum diffusion less relevant as it operates on longer time scales. This will significantly reduce the speed at which the steady-state profile evolves. This in turn will increase the characteristic length zchar after which the stable steady-state solution is found. This means that our assumption is a rather optimistic one. In systems with higher Reynolds numbers, the profile will develop significantly after a distance . In section 33.3 we will use a numerical scheme to directly compare the influence of momentum convection on the length and time scales it takes a flow profile to fully establish.
Qualitative Estimation. As stated, evaluating Eq. 18.65 with the convection analytically is very challenging. Therefore one usually defaults to either using experimental data or numerical evaluation in order to find a quantitative correlation between the Reynolds number and zchar. There are several commonly used equations for estimating zchar, one of the most commonly used ones being
Please note that this equation is empirical - it is derived from experimental data in which zchar is determined for flow fields with different Reynolds numbers thus establishing a correlation.
The derivation of the transient velocity profile during entrance flow was derived for a circular cross-section and thus based on the Hagen-Poiseuille flow. Obviously, the mathematics is identical for other cross-sections. In this case, Eq. 18.66 will serve as a basis which will, after inserting the combined steady-state and transient solution, yield a PDE just as it did for all transient/steady-state problems we discussed in this section. Solving this PDE for a given initial condition will result in the velocity profiles sought.
In this section we have studied time- and space-dependent effects during accelerating and deceleration of fluids, as well as during entrance flow. As we have seen, the time-dependent equations can be derived from the modified Navier-Stokes equation, taking into account the time-dependent term on the left-hand side, whereas the spacedependent equations can be derived from the modified Navier-Stokes equation, taking into account the changes along the x-axis. In all cases, we assumed the solution to the velocity distribution to consist of a time- or spacedependent transient solution and the time- or space-independent steady-state solution. The latter we have already derived while studying steady-state fluid flows in section 15. In all cases, we ended up with a homogeneous PDE which we solved by a substitution and separation of variables approach.
For the time-dependent transient solution, we then found a time-independent steady-state term and a time-dependent exponential term which reflects the time-dependency of the flow profiles. For the space-dependent transient solutions, we found a space-independent steady-state term and a space-dependent transient term, which reflects the space-dependency of the flow profile. For the time-dependent transient solutions, we found that the profiles can be scaled to a characteristic timescale which depends, as we expected initially, on the momentum diffusion properties of the fluid and on the length scale of the system. For the space-dependent transient solutions, we found a term which was purely dependent on the variable x. In all cases, we were able to derive dimensionless flow profiles from which we derived realistic flow profiles for chosen sets of geometric parameters.
As you can see, transient problems occur regularly in fluid mechanics, but they are not complicated to solve and the methods of solving these equations is always the same. It may be conveniently extended to other profiles and cross-sections that we did not study in this section, e.g., the planar infinitesimally extended channel.