You are on page 1of 147

CHAPTER ONE INTRODUCTION

1.1 He II Thermal Counterflow


Liquid helium can exist in one of two liquid phases separated by a second-order phase transition, called the -transition. Above the -transition, liquid helium is referred as He I, or normal helium, and behaves like a classical Navier-Stokes fluid obeying conventional fluid dynamic and thermal hydraulic models. Below the -transition, liquid helium enters the superfluid state, called He II. One unique fluid dynamic phenomenon existing in He II is the thermal counterflow, which can be understood in terms of the relative motion of two interpenetrating fluid components in the two-fluid model developed by Tisza [1] and Landau [2]. As the dominant heat transfer mode in He II as well as an ideal system for exploring the dynamics of superfluid turbulence, thermal counterflow has been studied extensively since its discovery in 1950s, focusing on two major aspects: 1) hydrodynamic characteristics of the two fluid components in He II, and the interaction between them under dynamic conditions; and 2) fundamentals of superfluid turbulence, including the generation and development of superfluid turbulence, and its effects on the hydrodynamics of the two fluid components [3-9]. However, due to the restrictions of measurement technology, past research mainly rely on the measurements of pressure and temperature gradients to study the heat transport and hydrodynamics of thermal counterflow, while direct measurements of the velocity fields of the two fluid components remain unattainable. Furthermore, although the numerical studies are able to

provide some detailed information about the velocity fields [10, 11], these results have to be verified by actual fluid dynamics measurements.

1.2 Second Sound and Heat Diffusion


Also unique to the fluid system of He II are the processes of second sound and non-linear heat diffusion. Second sound, which is basically a wave motion of entropy driven by a temperature perturbation, was predicted by Tisza [1] and Landau [2] based on the two-fluid model, and first discovered by Peshkov in 1944 [12]. Later in 1952, the nonlinearity theory of second sound was developed by Khalatnikov, who concluded that the second sound wave would become nonlinear and eventually form a thermal shock when the temperature perturbation is finite [31]. Heat diffusion is generally induced by the applying of a heat source, and develops when He II enters the turbulent state. Due to the existence of mutual friction between the two fluid components of He II in turbulent state, the heat diffusion equation of He II displays a unique non-linearity similar to the ordinary diffusion equation derived from Fouriers law of conduction. The generation and development of both second sound and heat diffusion are intrinsically associated with the dynamics of superfluid turbulence. Therefore, the study of these two processes can help to further understand the fundamentals of He II in turbulent state. In previous research, second sound shock and heat diffusion have been studied only through the measurements of the heat pulse induced temperature profiles. While it is predicted that transient thermal counterflow may be induced by these two processes, to date actual measurements of the corresponding velocity field have never been performed. Undoubtedly, such measurements will provide a new means of studying the propagation of second sound and heat diffusion, as well as the superfluid turbulence associated with them.

1.3 PIV Technique


Particle Image Velocimetry (PIV) is one of the most important achievements of flow diagnostic technologies in experimental fluid mechanics. It originated from a series of particle-imaging techniques developed in 1980s, such as Particle Tracking Velocimetry (PTV) and Laser Speckle Velocimetry (LSV) [13]. The basic principle of PIV is quite simple: the fluid motion is visualized by seeding micron-sized tracer particles and illuminating the flow field with a laser light sheet. Two images of the particles in the sheet are recorded in a short time interval. From their positions at the two instances of time, the particle displacements are calculated, and thus the velocity field of the flow can be inferred. Compared to other flow measurement techniques, PIV has many advanced features, such as non-intrusive and instantaneous whole field measurement, high accuracy and high spatial resolution. Due to these features, PIV has been applied in most areas of fluid mechanics and aerodynamics research. But, in contrast to its vast and broad applications to conventional fluids, such as water and gases, there is very little in the literature concerning the application of PIV to cryogenic fluids, especially to liquid helium [14, 15].

1.3 Research Objectives


This study attempts to apply the PIV technique to the unique fluid system of He II. As an entirely new technique for the research of He II, PIV would, for the first time, make it available to measure the spatial structure of the thermal counterflow velocity field induced by steady heating, or by second sound and heat diffusion. These detailed

fluid dynamics measurements should enable us to further understand the hydrodynamics of the two fluid components of He II and also the dynamics of quantized superfluid turbulence. In this study, three specific research objectives need to be accomplished. The first objective is to implant the PIV technique to liquid helium. Due to the exceptional physical properties of liquid helium, i.e. low density and extremely small viscosity, selection of tracer particles is tough work, which requires a careful study of the particle dynamics in liquid helium [16]. The critical parameters for particle selection must be determined. Based on those parameters, appropriate tracer particles for liquid helium must be selected from a variety of candidate particles. Also, the unique experimental environment, low-temperature and vacuum, of liquid helium requires the development of a new seeding method to properly introduce the tracers particles into liquid helium. The second objective is to conduct PIV measurements of thermal counterflow at steady state. In particular, the normal fluid velocity field at different applied heat fluxes and bath temperatures must be measured to reveal the spatial structures of thermal counterflow. Measurements of normal fluid velocity profile as a function of the heat flux and bath temperature are also required in order to compare it with the results predicted by thermal counterflow theory. Some research questions need to be answered, including how the tracer particles interact with the two fluid components in He II, and how the interaction will influence on the motion of particles. The third objective is to study the propagation of second sound shock and heat diffusion by measuring the induced transient thermal counterflow field. To accomplish this goal, the instantaneous velocity fields need to be measured with a high temporal resolution. Also, the velocity profiles versus time must be measured and compared with the temperature profile to evaluate the applicability of PIV technique in this case.

This dissertation is organized in five major parts: chapter 2 is the background of liquid helium, which consists of the two phases of liquid helium, He I and He II, the twofluid model, hydrodynamics associated with the two-fluid model in laminar and turbulent state, basics of thermal counterflow, and the two transient heat transfer processes, second sound and heat diffusion, in He II; chapter 3 is a brief introduction to the PIV technique, in which the principle and features of PIV are discussed, and several important technical aspects of PIV, including the selection of tracer particles, image acquisition, and correlation process, are reviewed; in chapter 4, the experimental techniques associated with the application of PIV to liquid helium is presented. Challenges for applying PIV to liquid helium are discussed. The particle dynamics in a fluid is thoroughly studied in the context of three different kinds of flows. The tracking characteristics of a variety of tracer particles are discussed to evaluate their potential application in liquid helium, and the particle seeding technologies are discussed. At the end of this chapter is some detailed information about the PIV experimental set-up, including the laser, camera, timing control unit and programs, optical system, cryostat, and the counterflow channel; chapter 5 is the results and discussions, in which the PIV measurements of steady thermal counterflow as well as the transient thermal counterflow induced by second sound shock and heat diffusion are presented and discussed; chapter 6 is the conclusions and some suggestions for potential improvements and future experiments.

CHAPTER TWO BACKGROUND OF LIQUID HELIUM

2.1 Liquid Helium


Helium is the most difficult of all the permanent gases to liquefy due to the weakness of attractive forces between helium molecules. The fluid has a critical temperature of 5.2 K and a normal boiling point of 4.21 K. The first successful liquefaction was accomplished by Kamerlingh Onnes in 1908. Helium is the only element that exists in the liquid state at absolute zero temperature, and as a result liquid helium has received extensive research as an ideal system for the study of quantum fluids. Figure 2.1 shows the P-T phase diagram of helium. From it, several unique characteristics of helium can be noted [17, Chap. 3]: a) The solid state is not obtainable at any temperature unless an external pressure in excess of 2.5 MPa is applied. This is due to the large zero point energy of the helium molecule, which causes liquid as the lowest entropy state. b) There is no triple point of coexistence between liquid, vapor and solid because the solid state can exist only under a certain pressure. c) Liquid helium can exist in two very different phases, normal fluid phase (He I) and superfluid phase (He II). The line that separates these two phases is termed the line, which was adopted because the specific heat near the transition is discontinuous and has the shape of the Greek letter . The -transition between He I and He II is classified as a second-order phase transition, which means it has a discontinuous slope in the

temperature dependence of the entropy and there is no latent heat resulted from the transition. The physical significance of -transition is that the two liquid phases cannot coexist in equilibrium. The -transition temperature, T , is 2.176 K at saturated vapor pressure and decreases gradually along the line with increasing pressure until it intersects with the solid phase.

Figure 2.1 4 He phase diagram [17].


P P

The two liquid phases of helium, He I and He II, show substantially different behaviors. He I is essentially a classical Navier-Stokes fluid with associated static and thermodynamic properties, and obeys the hydrodynamics described by conventional

Navier-Stokes equations. However, He II, referred as superfluid helium, displays many exceptional physical and transport properties, and thus will be discussed in detail below.

2.2 He II as a Quantum Fluid


Below -transition, liquid helium enters the He II phase and displays several unique transport properties, such as a very high thermal conductivity, many orders of magnitude larger than that of other liquids and even pure metals, and an ideal inviscid flow behavior discovered by Allen and Misener in 1938. In their experiment (Fig. 2.2 a), Allen and Misener determined that the viscosity of He II is vanishingly small (on the order of 10-12 Pas) through measuring the He II flow through a capillary channel [18].
P P

However, another measurement done by Keesom, who used a rotating cylinder viscometer (Figure 2.2 b) to measure the viscosity of He II, yielded a finite viscosity of the order of 10 5 Pas, which is very close to that of He I [19]. To interpret this so called viscosity paradox, a two-fluid model of He II has been developed by Tisza [1] and also by Landau in a slightly different form [2]. According to this model, He II consists of two interpenetrating fluid components, the normal fluid component and the superfluid component. The normal fluid component behaves like the ordinary Navier-Stokes fluid with a density n , viscosity n and specific entropy s n . In contrast, the superfluid has only the density s but no entropy ( s s = 0 ) and viscosity ( s = 0 ). In the two-fluid model, the above viscosity paradox can be explained as follows. For Allen and Miseners experiment, the normal fluid component was held back by its viscous interaction with the capillary wall so that only the superfluid component can flow through the narrow capillary. Thus the measured

viscosity is that of the superfluid component s = 0 , so the flow is inviscid. For the damping experiment of a rotating cylinder, however, the normal fluid component interacts with the rotating disc and produces the damping force. Therefore, the viscosity measured in this experiment is actually the normal fluid viscosity n , which is similar in magnitude to that of He I.

(a)

(b)

Figure 2.2 Two different methods of measuring the viscosity of He II (a) flow through a capillary channel [18]; (b) damping of a rotating cylinder [19].

Two-fluid model of He II
U

From the two-fluid model, the density and total mass flow rate of He II can be written as,

= n + s
v v v v J = v = n v n + s v s

(2.1) (2.2)

where is the bulk density, v is the bulk flow velocity, vn and v s are the velocities of normal fluid and superfluid components respectively. While is nearly constant (145 kg/m3) at saturated pressure, the normal fluid
P P

and superfluid densities, n and s are strongly temperature dependent


5.6

n T = T

and

s T = 1 T

5 .6

for T T

(2.3)

Figure 2.3 shows the relative ratio of normal fluid and superfluid density to the bulk density of He II at different temperatures.

Figure 2.3 Ratio of normal fluid and superfluid density in He II [17].

Since the velocity fields of the normal fluid and superfluid components are independent of each other, the momentum equations of He II should be written for the

10

two components separately. For the superfluid component, the flow is driven by the gradient of chemical potential rather than the pressure gradient, so one can get
r Dv s = Dt

(2.4)

where is given by = s T + 1

(2.5)

Here, P is the fluid pressure. Combining equations (2.4) and (2.5), the momentum equation of superfluid can be written as,
r Dv s = s P + s sT Dt

(2.6)

To compare with classical fluids, the term superfluid pressure is introduced as Ps = s . Then, equation (2.6) can be rewritten in the form of
r Dv s s = Ps Dt

(2.7)

This equation has exactly the same form as the momentum equation for nondissipative Euler fluid, indicating that superfluid itself can be treated as an ideal Euler fluid. For the normal fluid component, since it possesses viscosity like an ordinary Navier-Stokes fluid, its momentum equation is identical to the classical Navier-Stokes

11

equation as below,
r Dv n r n = Pn + n 2 v n Dt

(2.8)

where Pn is termed as normal fluid pressure and Pn + Ps = P

(2.9)

Substituting equation (2.9) into equation (2.6), the momentum equation of normal fluid component can be finally written as:

r Dv n r n = n P s sT + n 2 v n Dt

(2.10)

From equations (2.7) and (2.10), He II can be viewed as a mixture of an ideal Euler fluid and a classical Navier-Stokes fluid.

2.3 He II in Turbulent State


As we know from the two-fluid model, there is no dissipative term for the superfluid component of He II. Therefore, it can be predicated that, when a vessel initially at rest is filled with He II and driven to rotate, the superfluid would stay at rest and only the normal fluid component could be brought into rotation by the viscous force. This predication has been verified by Andronikashvilis experiment [20], in which the normal fluid density n measured by a group of oscillating discs is in agreement with the

12

value obtained through second sound measurement (Figure 2.4a). However, another experiment done by Osborne [21] (Figure 2.4b), shows that the superfluid does take part in the rotation. This rotation paradox can be interpreted by the turbulent theory of He II, which is first developed by Hall and Vinen based on a series of experiments of measuring the rotation of He II under different conditions [22-27].

(a)

(b)

Figure 2.4 Rotation paradox arising from two different experiments (a) Measurement of n by Andronikashvili [20]; (b) Rotating bucket by Osborne [21].

According to Vinen, the superfluid component may begin to rotate when the relative velocity between the normal fluid and superfluid reaches a certain value, termed as the critical relative velocity Vrc . Above this critical velocity, He II enters the turbulent

13

state, and a mutual friction force between the two components of He II will be generated, driving the superfluid to rotate. This mutual friction force, which was first introduced by Gorter and Mellink [28] to explain the heat transport data, is further described by Vinen as the scattering of normal fluid thermal excitation by a tangled mass of quantized vortex lines in superfluid. In an effort to determine the mutual friction force, the force per unit length of the vortex line f is first given based on the classical turbulent interactive mechanism,
s n nVr 2
r r

f =

(2.11)

where Vr = v n v s is the relative velocity between superfluid and normal fluid. Then the next step is to determine the vortex line length per unit volume, termed vortex line density. Assuming a homogeneous distribution of vortex lines that at any time have a density L and an average interline spacing l = L1 / 2 , the vortex line density will grow with the expansion of vortex rings due to the relative velocity between the normal fluid and superfluid turbulent, and the growth rate is given by
B n 3/ 2 dL L Vr = 1 2 dt g

(2.12)

where 1 and B are both empirical parameters. At the same, the line density will also decrease due to the breakdown of interacting vortex lines into small rings. Those small rings contract in size and eventually dissipate their energy as thermal excitation. The rate of decay can be approximately described by

14

L2 dL = 2k 2 dt d

(2.13)

where 2 is another empirical parameter, and k =

h ( h the Plancks constant and m the m

mass of a helium atom) is the quantum of circulation. Suppose the growth and decay processes are independent, the steady state vortex line density, L0 , can be obtained by setting these two rates equal,

dL dL = dt g dt d

(2.14)

Thus L0 can be written as,

L0 = Vr

(2.15a)

where,

B n 1 k 2

(2.15b)

Combining equation (2.11) and (2.15), the mutual friction force per unit volume is given by
r r r Fns = f L0 = A s nVr2 ( v n v s )

(2.16)

15

where A is called the Gorter-Mellink coefficient, which can be determined empirically by experiments.
r Considering the existence of a mutual friction force Fns between the normal fluid

and superfluid components, the momentum equations of He II introduced before (equations (2.6) and (2.10)) can be rewritten as,

r r Dv s s = s P + s sT Fns Dt r r Dv n r n = n P s sT + n 2 v n + Fns Dt
Also, the equation of mass conservation is given by

(2.17)

(2.18)

r + ( v ) = 0 t
and the equation of entropy conservation is
r ( s ) + ( sv n ) = 0 t

(2.19)

(2.20)

The above group of equations (2.17-2.20) is termed the Navier-Stokes equations for He II in turbulent state. Since there are two fluid components in He II, each component should display its own form of turbulence. As the normal fluid component behaves like a classical viscous

16

fluid, its transition to the turbulent state occurs when the Reynolds number exceeds an ordinary critical Reynolds number of
dv nc

Re( v nc ) =

1200 to 2000

(2.21)

where d is the hydraulic diameter of the channel. This equation yields a critical velocity v nc , called normal fluid critical velocity, above which the normal fluid component enters the turbulent state. For the superfluid component, the turbulence that takes the form of tangled vortex lines occurs when a superfluid critical velocity v sc is reached. According to the quantum mechanics theory, v sc can be estimated as [17, Chapt. 4]
10h 10 2 (cm/s) md d

v sc

(2.22)

Actual experimental measurements of v sc follow a d 1 / 4 law as


v sc d 1 / 4 (cm/s)

(2.23)

where d is in units of centimeters. When this empirical equation is applied to a flow channel with diameter on the order of 1 cm, the estimated superfluid critical velocity is about 10 mm/s, which is very small and can be easily reached. For the thermal counterflow that will be discussed next, the applied heat flux corresponding to this critical velocity is only about 2.6 kW/m2 at 1.80 K. In case that the superfluid velocity is
P P

above v sc , the superfluid component can no longer be regarded as an inviscid fluid.

17

2.4. Thermal Counterflow in He II


Consider a flow channel that connects two He II reservoirs (as shown in Figure 2.5), when a steady heat is applied to one end of the channel, there exists a temperature difference T between the two ends. Since only the normal fluid component has the entropy and can carry the heat flow, it will move away from the heat source (left reservoir) to the right reservoir and then give up the heat. At the same time, the superfluid component must counter-flow from right to left to conserve the mass. When it arrives at the left reservoir, part of the superfluid component will be converted to normal fluid by absorbing heat, making the local densities of the two components obey the relationships given by equations (2.1) and (2.3). Thus, a relative counterflow between the normal fluid and superfluid components is established, and this internal convection process is termed thermal counterflow.

Figure 2.5 He II thermal counterflow [29].

Since only the normal fluid can carry the entropy, the heat transport equation for thermal counterflow takes the form,

18

av q = sTv n

(2.24)

where q is the applied heat flux, and v nav represents the spatially averaged normal fluid velocity over the cross section perpendicular to the heat flow. Also, the zero net mass flow condition requires that
av s v sav + n v n = 0

(2.25)

Combining equations (2.24) and (2.25), the averaged relative velocity Vr can be written as

av Vrav = v n v sav =

q s sT

(2.26)

When Vrav < Vrc , the thermal counterflow is in laminar state, and there is no mutual friction force between the normal fluid and superfluid components. Considering the simple case of steady state, equations (2.6) and (2.10) can be combined, yielding

r P = n 2 v n
P s

(2.27)

T =

(2.28)

Equation (2.28) is called Londons equation. Equation (2.27) is equivalent to the Poiseuille equation in classical fluid dynamics, indicating that the normal fluid velocity

19

will display a parabolic profile in laminar state. Given a channel of constant cross section, equation (2.27) can be further solved to relate the normal fluid velocity v n to the pressure gradient [17],
n v n
d2

P =

(2.29)

where is a numerical constant: = 12 for parallel plates or large aspect ratio rectangular cross section, and = 32 for circular tubes. Combining equations (2.29), (2.28) and (2.24), the heat conduction equation for thermal counterflow in laminar state can be obtained as below,
d 2 2 s 2T

q=

(2.30)

which indicates that the temperature gradient is a linear function of the applied heat flux
& & (the region of Q < Q2 as shown in Figure 2.6).

When Vrav Vrc , thermal counterflow is in turbulent state, and the mutual friction force must be taken into account. In this case, equations (2.17) and (2.18) are applied. For the case of steady state, these equations reduce to

r P = n 2 v n
P Fns + s s s

(2.31)

T =

(2.32)

20

Figure 2.6 Temperature and pressure difference data from Tough [3].

It can be seen that the first term in equation (2.32) has the same form as in equation (2.28), while the second term represents an additional temperature gradient,
T = Fns , due to the mutual friction force. Replacing Fns by equations (2.16) and s s

(2.26), the temperature gradient T in turbulent state can be finally written as,
n
A n q3 s3 s 4T 3

T =

d 2 2 s 2T

(2.33)

& & which is a nonlinear function of the applied heat flux (the region of Q Q3 as shown in

Figure 2.6). The second term in this equation, due to its dependence on cubic power of

21

the heat flux, will dominate the temperature gradient for moderate heat flux or in a fully developed turbulent state. In this case, equation (2.33) can be reduced to T = f (T ) q 3

(2.34)

where

f (T ) = A n /( s3 s 4T 3 )

(2.35)

is a temperature- and pressure-dependent parameter. Equation (2.34) is called the GorterMellink relation, and f 1 (T ) can be looked at as the effective thermal conductivity of He II.

2.5 Transient Processes in He II


Generally, transient heat transfer processes in He II are initiated by an impulse or a stepwise heat input. Unlike the steady state heat transfer process that only depends on one characteristic parameter, applied heat flux q , transient heat transfer processes also depend on the characteristic time s . Based on these two parameters, transient heat transfer processes in He II can be classified into several different regimes [30]. As shown in Figure 2.7, the applied heat flux in regime I is very small, where
q < qcr and qcr 0.01 W/cm2. In this case, the heat transfer process for any s can be
P P

simply described by equations (2.27), (2.28) and (2.30). For an applied heat flux in the range of qcr q < 1.0 W/cm2, the transient processes is divided into two regimes, II and
P P

III, separated by the line V, for which

22

Figure 2.7 Diagram of transient heat transfer regimes in He II in the coordinates ts: time of applying heat and q: heat flux [30].
B B

v = a(T ) q 3 / 2

(2.36)

where v is the characteristic development time of superfluid turbulence, and a(T ) is a temperature-dependent coefficient. In regime III, where s v the superfluid turbulence is fully developed, and the transient heat transfer process can be characterized by second sound attenuation and/or heat diffusion. Typically, in the upper part of regime III, where

s >> v , the heat transfer process reaches the steady state and thus can be simply

23

described by equation (2.32). In regime II, where s < v , the superfluid turbulence is not fully developed, and the heat is transported in second sound shock and/or the heat diffusion mode. When q 1.0 W/cm2, three different regimes are possible: in regime IV,
P P

where s is relative small, second sound attenuation is the main characteristic of the heat transfer; while in regime VI, where s is large, film boiling heat transfer may occur. Below, the two main transient heat transfer processes, second sound and heat diffusion are discussed.

2.5.1 Second Sound As a result of having two fluid components, He II is able to transmit more than one type of sound. In addition to the ordinary or first sound, there is a process known as second sound, which can be described as the propagation of entropy in wave motion. Differences between first and second sound are illustrated in Figure 2.8. It can be seen that first sound is driven by a pressure perturbation, and during its propagation, the local fluid density oscillates but the local temperature remains constant; while second sound is driven by a temperature perturbation, and during its propagation, the local density remains constant but the local temperature oscillates out of phase due to the variation of relative concentration of normal and superfluid components. When the amplitude of temperature perturbation T is very small, the linear theory of wave motion is applied, and an amplitude-independent second sound velocity is given by [17],
s Ts 2 n cp

c20 =

(2.37)

24

Figure 2.8 Illustration of (a) first sound and (b) second sound. The portion of n and s represents the relative concentration of normal and superfluid component [29].

where c p is the constant pressure specific heat. Though this equation shows that the second sound velocity is temperature dependent, it is almost constant at about 20 m/s for bath temperatures between 1.0 and 2.0 K. When the temperature perturbation is finite, the second sound displays nonlinearity. In this case, the second sound velocity is amplitude-dependent and given by [31] c2 = c20 [1 + B(T / T )] where B is called the nonlinear steepening coefficient (2.38)

25

3 c20 c p ln B = T T p T

(2.39)

The nonlinear steepening coefficient B is strongly temperature-dependent, and, as shown in Figure 2.9, can be both positive and negative [32]. When it is positive, the nonlinearity causes the points with higher amplitude on a pulse to travel faster than those with lower amplitude, and thus makes an initially rectangular pulse evolve into a frontsteepened shock wave with an expansion fan at the tail (Figure 2.10 (a)). In contrast, when it is negative, the nonlinearity causes a back-steepened shock wave with an expansion fan at the front (Figure 2.10 (b)). The second sound shock velocity u s can be obtained by averaging the velocities at the front and tail of the shock,
1 u s = c20 1 + B (T / T ) 2

(2.40)

From this expression, the shock Mach number M can be determined as,

M=

1 u = 1 + B(T / T ) 2 c20

(2.41)

One unique characteristic of second sound is that it primarily induces thermal counterflow between the normal fluid and superfluid. Turner suggested that this process only takes a very short time, one microsecond or less [33]. For the induced counterflow, equation (2.26) gives the relative velocity between the two components. As discussed before, when the applied heat flux q is small and the relative velocity is less than the

26

critical relative velocity Vrc , there is no mutual friction and the counterflow is in laminar state. In this case, the heat flux that can be transported by second sound shock is approximately equal to the applied heat flux, and the shock wave generated by a rectangular heat pulse has a trapezoidal temperature versus time profile (waveforms 1 and 2 in Figure 2.11).

Figure 2.9 Experimental measurement results of the nonlinear steepening coefficient B . The solid line represents the value given by equation (2.39) [32].

(a)

(b) time t
Figure 2.10 Development of second sound shock from a rectangular pulse (the arrow shows the movement of pulse); (a) front-steepened shock; (b) back-steepened shock.

27

Figure 2.11 Deformation of second sound shock waves ( q p : applied heat flux; t H : pulse heating time; bath temperature is 1.70 K and the distance from heater is 30 mm) [34].

As q is increased and the relative velocity is larger than Vrc , the quantized vortices in the superfluid start to develop, and the mutual friction force between the normal fluid and superfluid is generated. The second sound shock wave then interacts with the quantized vortices, causing the shock waves to deform. It can be seen from the waveforms 3 and 4 in Figure 2.11, that the deformation is characterized by the tilting of the flat top and an elongated shock tail. In this case, the shock amplitude, i.e. T , continues to increase with increasing heat flux. However, the heat flux that can be transported by second sound shock is no longer equal to the applied heat flux, and the ratio of transported to input heat flux decreases markedly with increasing q [33]. A further increasing of heat flux q makes the shock wave eventually reach the limiting profile (waveforms 5 and 6 in Figure 2.11). It can be seen that the whole waveform is much shorter than the initial heat pulse, and the two forms almost coincide with each other irrespective of different heating times. As long as the limiting profile is

28

attained, further increasing the heat flux cannot generate a stronger shock, or higher T , and, in some cases, may even cause T to decrease. In summary, for the cases of waveforms 3, 4, 5 and 6, only a fraction of the applied heat flux can be transported by second sound shock. The rest of the energy accumulates in a thin thermal boundary layer, and is then transported by the heat diffusion process that will be discussed next [33, 34].

2.5.2 Heat Diffusion in He II Heat diffusion is the other important transient heat transport phenomenon in He II. It may occur in many cases once the turbulent state is established. The two common cases that have been broadly analyzed and studied are stepwise heating and pulse heating. For the stepwise heating problem, the characteristic time s is defined as the time from the heater is on to the moment at which the measurement is conducted. When

s > v , where v is defined by equation (2.36), the superfluid turbulence is fully


developed, and there exists a mutual friction force between the normal fluid and superfluid. In this case, the heat transfer process can be described by the Gorter-Mellink relation, yielding a non-linear heat diffusion equation,
1 T = T C p t f (T )
1/ 3

(2.42)

Compared to the ordinary diffusion equation,

T 1 2 = T , based on classical Fouriers t

law of conduction, this heat diffusion equation is quite difficult to be solved analytically

29

because of the existence of 1/3 power. While numerical methods are the most popular way to treat this equation, an elegant approach has been developed by Dresner to give a similarity solution under different boundary conditions [35, 36]. From equation (2.42), the resulting temperature profile during the process of heat diffusion is a function of elapsed time. Figure 2.12 shows a simple one-dimensional case in which a stepwise heating is applied at the bottom of a channel filled with He II. The top of the channel is open to helium bath. At the initial stage of heat diffusion (point A), the temperature gradient at the bottom is much steeper than that at the top. As time increases, the temperature profile develops along the channel and finally reaches the steady state (point B), where the temperature gradient is approximately constant along the channel, and can be determined by equation (2.34). Heat diffusion may also be generated by a heat pulse when the applied heat energy is greater than that can be transported by the second sound. Figure 2.13 shows an example of the temperature profile measurement after a heat pulse is applied (applied heat flux is q = 40 W/cm2 and pulse duration is 1 ms) [34]. It can be seen that the first
P P

temperature rise with a sharp front is due to the arrival of second sound shock, while the second temperature rise results from the heat diffusion. Clearly, compared to second sound, the development of heat diffusion is much slower. However, the energy transported by heat diffusion, which is proportional to the area under the temperature profile, is much larger than that can be transported by second sound. From this viewpoint, in the case that either the applied pulse heat flux is very large or the pulse duration is very long, the transient heat transfer process is dominated by the heat diffusion. In contrast to the heat diffusion under stepwise heating that can be simply described by equation (2.42), heat diffusion induced by pulse heating may become more complicated. When the duration of a heat pulse w are much larger than the characteristic

30

Figure 2.12 Development of one-dimensional temperature profile during heat diffusion.

Figure 2.13 Profiles of temperature rise due to pulse heating (Tb=1.7 K, q=40 W/cm2, pulse duration=1 ms; z represents the distance away from the heater) [34].
B B P P

31

development time for superfluid turbulence, v , the density of the quantized vortex line, L , would have reached its equilibrium value (equation (2.15)) with respect to the relative velocity, Vr . Thus, the Gorter-Mellink relation (equation (2.34)) is valid and the heat diffusion can still be described by equation (2.42) [37]. However, when w is comparable to or shorter than v , the vortex line density may not always reach the fully developed equilibrium value. In this case, the vortex developing process is fully coupled with the hydrodynamic processes [38]. Regarding the heat diffusion, the Gorter-Mellink relation based on a steady state assumption may not be valid, and therefore the application of equation (2.42) should be somewhat limited.

32

CHAPTER THREE BASICS OF PIV TECHNIQUE

3.1 Introduction to PIV


Figure 3.1 shows a typical experimental set-up of a PIV system [39], which consists of a laser, light sheet optics, image capture device (either a regular camera to record the images on photographic film or a CCD camera to record the images on CCD sensor), and a synchronizer to control the camera and laser. Firstly, small tracer particles are seeded into the flow field. Then the seeded flow field is illuminated twice by two pulses of laser sheet separated by a certain time delay t . The light scattered by the tracer particles is recorded and two successive images are captured. Each of the images is then subdivided into an array of small size region, called interrogation window, and for each of such windows, a numerical correlation algorithm (either auto-correlation or crosscorrelation) is applied to statistically determine the local displacement vector of particles between the first the second illumination. It is assumed that all particles within one interrogation window have moved homogeneously between the two illuminations. Further assuming that the tracer particles move with the local flow velocity, the velocity vectors in the whole flow field can be obtained by dividing the particle displacement by
t .

Compared with traditional flow diagnostic techniques, such as pressure tubes or hot wires, PIV technique has the following features: 1) Non-intrusive and indirect velocity measurement. PIV works nonintrusively by employing an optical signal as the probe. This allows the application of PIV in high speed flows with shocks or in measuring boundary layers, where the flow may be disturbed by

33

Figure 3.1 A typical PIV experimental set-up [39].

the presence of other probes. Also, PIV measures the velocity of a fluid element indirectly by measuring the velocity of tracer particles within the flow [39]. 2) Statistical method. The density of particle images in PIV is mediated to be medium so that the images of individual particles can be detected while it is no longer possible to identify particle pairs by visual inspection of the image. Therefore, rather than tracking individual particles, the PIV technique follows a group of particles through statistical correlation of sampled numbers of the image field. 3) Whole field technique. Depending on the size of imaged flow field and interrogation window, the PIV technique allows the velocity information at hundreds or thousands of points be extracted out of the images. This is a very unique feature of PIV

34

compared with other velocity measurement techniques, such as Laser Doppler Velocimetry (LDV) that only allows the measurement of velocity at a single point. Also, this feature results in a high spatial resolution of PIV, and makes it possible to detect the very fine spatial structure even in unsteady flow fields. However, as a trade-off of the high spatial resolution, the temporal resolution of PIV is limited due to technical restrictions, such as the frame rate of CCD camera and the firing rate of laser. As a widely used quantitative flow diagnostic tool, many aspects of the PIV technique, including the selection and seeding of tracer particles, image recording, image processing, correlation analysis, improvement of accuracy, error correction and so on, have been extensively studied [40-45]. Below, only several important technical aspects of PIV are selected for discussion.

3.2 Tracer Particles


It is clear from the principle of PIV that it relies on the scattering of light from tracer particles to determine the displacement of particles within a certain time delay. Since PIV indirectly measures the flow velocity by means of measuring the particle velocity, the traceability of particles has to be checked in order to avoid significant discrepancies between the fluid and particle velocity. Basically, when the density of tracer particles is not the same as that of the fluid, the influence of gravitational force will cause a primary error in the PIV measurements. In order to reduce the error, the tracer particles must be as small as possible to faithfully follow the fluid flow. On the other hand, if the tracer particles are neutrally buoyant in the fluid, namely the density of particles is equal to that of the fluid, it seems that the particle size is not limited by its ability to track the flow. However, in this case, one more aspect of the fluid dynamic

35

properties of tracer particles, the relaxation time s , has to be considered. According to Raffel [39], the relaxation time is a critical measure for the tendency of particles to attain velocity equilibrium with the fluid, which is given by
2 pd p s = 18 f

(3.1)

where p is particle density, d p the diameter of particle, and f the viscosity of fluid. Since a large relaxation time may cause a delay for the particles to reach the flow velocity, and a continued phase lag when the fluid velocity is fluctuating, it has to be limited. Thus, the particle size has to be limited even for the neutrally buoyant particles because the relaxation time depends on the particle density rather than the density difference between particles and fluid. This issue becomes more important when PIV is applied to measure the turbulent or unsteady flows with large velocity fluctuations. As described above, the tracking characteristics of particles require that the particle size should be very small to ensure good tracking of the fluid motion. On the other hand, the optical characteristics of particles require that the particle diameter should not be too small so that there is sufficient light scattered from the particles for image acquisition. The light scattering capability of particles can be measured as the SNR (signal-to-noise ratio) of the scatted light. According to Mies scattering theory, the SNR is a complex function of the power, wavelength and frequency of the light, and the diameter and refractive index of the particles [39]. For simplicity, Melling gave a convenient measure of the light scattering capability, the scattering cross section As , which is defined as the ratio of the total scattered power to the laser intensity [40]. The variation of As as a function of the particle diameter and laser wavelength is shown in Figure 3.2. It can be seen that larger particles generally scatter more light and give higher

36

SNR; when the particle diameter is less than 1 m (~ 2 for green light with = 532 nm) the scattering capability decreases drastically with the particle diameter; for particles of diameter from 1 to 10 m, they have a good SNR and are appropriate for PIV experiments. In addition to the scattering capability or SNR, it is also required that the particle size should be as uniform as possible because the excessive intensity of light scattering from larger particles will induces a high background noise to the light scattering from the smaller particle. In summary, selecting the appropriate tracer particles for PIV applications is really a case of compromise between the tracking and optical characteristics of particles. Optimum tracer particles should be small enough to faithfully follow the flow and large enough to scatter sufficient light intensity.

Figure 3.2 Scattering cross section as a function of the particle size [40].

37

3.3 Particle Image and Image Acquisition


When the seeded flow field is illuminated by a laser sheet, the light scattering from particles is captured by the camera, and an image is produced on the recording medium. The image diameter of a particle is given by Adrian as [41],
2 d i = ( M 2 d p + d s2 )1 / 2

(3.2)

where d i is the image diameter, d p the particle size, M the magnification number of the camera, and d s is the point response function resulting from diffraction of the lens,
d s = 2.44(1 + M ) f #

(3.3)

Here f

is the f -number of the lens, and the wavelength of light.

The image size of particles has a great deal to do with the accuracy of PIV measurements. Westerwell, Dabiri & Gharib show that the uncertainly of using 2-pixel large particle image is half the uncertainty of using 4-pixel large particle image when the 32 by 32 pixels interrogation window is used [46]. The image size also has an effect on estimating the centroid of correlaton peak. It is suggested that the error of peak estimation is minimal when the image size is on the order of the size of pixel [47]. In addition, Raffel et al. asserted that, for a 32 by 32 pixels interrogation window, the optimum particle image size should be around 2.2 pixels to achieve a minimal uncertainty [39]. The two successive particle images acquired from two laser shots separated by time t can be recorded either on two individual image frame (single-exposed mode) or on a single image frame (double-exposed mode). The double-exposed mode has an

38

inherent disadvantage called directional ambiguity, which means one cannot determine which particle image is from the first exposure and which one is from the second exposure. However, this mode is superior for high-speed flow because a small exposure interval t can be easily obtained since there is no need to switch the recording medium. For the single-exposed mode, the directional information is preserved since the sequence of the frames is known. But it requires the exposure interval t not to exceed the frame rate of cameras (for CCD camera, the typical maximum rate is 30 frames/second). Because of the large t , the application of single-exposed mode is limited and only suitable for the measurements of low-speed flow. To overcome this limitation, a so-called frame straddling technique was developed. This technique used a high-performance progressive-scan-interline CCD chip as the sensor. The chip consists of an array of photosensitive cells and an equal number of storage cells. After the first laser pulse is triggered, the first image is acquired and immediately transferred from the photosensitive cells to the storage cells. Later, when the second laser pulse is triggered, the photosensitive cells are available to store the second image. In this case, the storage cells contain the first image and the photosensitive cells contain the second image. Then both of them are transferred sequentially from the camera to the computer for permanent storage. Using this technique, the exposure interval t can be reduced to less than 1 microsecond. As a result, the application of single-exposed mode is able to be extended for measuring very high-speed flow.

3.4 Correlation Process


PIV images are typically processed by subdivision into an array of overlapping interrogation windows. For each window, a correlation process is performed to produce a table of correlation values over a range of displacements. The overall displacement of

39

particles in the window is represented by a peak in this correlation table [45]. Corresponding to the two image acquisition modes, there are two different correlation methods, auto-correlation and cross-correlation. For images acquired in double-exposed mode, auto-correlation is applied. An auto-correlation function is defined by
ACOR(dx, dy ) =

I .W .

I ( x, y) I ( x + dx, y + dy)dxdy

(3.4)

where (dx, dy ) is the displacement vector, I ( x, y ) is the luminous intensity distribution of the interrogation window. The overall displacement of particles is determined by (dx0 , dy0 ) that makes the function ACOR(dx, dy ) reach its peak value. For auto-correlation, a primary correlation error is from the loss of image pairs because particles move out of the interrogation window. When the particle displacement increases, this so called velocity bias becomes worse, and the auto-correlation peak becomes smaller and less likely to be determined. In order to reduce the correlation error, one can either increase the seeding density of particles or increase the size of interrogation window. As the number of particles increases, the probability of obtaining an accurate measure of the displacement of a set of particles increases [45]. It has been demonstrated by Adrian that the number of spurious vectors that appears in PIV data drop dramatically as particle seeding density increases to an average of about 10 particle images per interrogation window [41]. However, the seeding density cannot be too high otherwise the characteristics of the flow being measured will be altered. An alternative is to increase the interrogation window size, which will increase the number of tracer particles in an interrogation window without increasing the seeding density. In this case, the velocity bias associated with particles entering and exiting the interrogation window during the exposure time interval t is also reduced. But one drawback of increasing the

40

interrogation window size is that the spatial resolution of PIV measurements decreases. For auto-correlation, it is suggested that the optimal interrogation window size should be about six times of the local particle image displacement [48]. When images are acquired in single-exposed mode, cross-correlation is applied. A cross-correlation function is defined by
CCOR (dx, dy ) =

I .W .

I ( x, y) I
1

( x + dx, y + dy )dxdy

(3.5)

where I 1 ( x, y ) is the luminous intensity distribution of the first image and I 2 ( x, y ) is the intensity distribution of the second image. The statistical overall displacement of the particles is determined by a vector originating from the geometrical center of the first interrogation window to the point where the CCOR(dx, dy ) function reaches its peak value. There exist a rule for standard cross-correlation processing, which means the particle displacements between two frames must be less than length of the interrogation window. Limited by the rule, the spatial resolution of cross-correlation is at least four times of the particle displacements. To improve the resolution, a sub-region shifting technique has been developed, where the second interrogation window is spatially shifted with respect to the first interrogation window by an amount equal to the mean flow displacement between the two exposures. Using this technique, most of the particles within the first interrogation window will also be in the second interrogation window, making it possible to use an interrogation window much smaller than that prescribed by the rule. In addition to the image shifting technology, another way to improve the spatial resolution, called adaptive cross-correlation, has been developed by Adrian [41], who suggested that cross-correlation can be performed between a small interrogation window on the first image and a larger interrogation window on the second

41

image. Just like the image-shifting, this technique also ensures that most of the particles within the first interrogation windows will be in the second interrogation windows. For both the auto-correlation and cross-correlation process, the vital step is to find the position of correlation peak in sub-pixel level. This step may greatly affect the accuracy of correlation results. Typically, correlation results without a special peakfinding scheme are accurate within +/- half pixel. In contrast, when a sophisticated peakfinding scheme is applied, the accuracy may be improved up to 0.01 pixel. Many different peak-finding schemes have been developed. Among them, the Gaussian threepoint curve fit scheme developed by Westweel is claimed to produce the least uncertainty because the cross-correlation peak actually displays a Gaussian intensity profile [43]. Above is a brief summary of the basics of PIV technique. Those technical aspects that have been discussed, including tracer particle selection, image acquisition, and correlation process, are generally applicable to most of the PIV systems. But, when applying the PIV technique to the unique fluid system of liquid helium, some more and new aspects should also be considered, which are going to be discussed in the next chapter.

42

CHAPTER FOUR EXPERIMENTAL TECHNIQUES

4.1 Challenges for Applying PIV to Liquid Helium


Flow visualization studies provide micro-scale details of the flow field under study, and when applied to liquid helium may help to further understand the fluid dynamics of this unique fluid system. The first attempt to visualize liquid helium fluid dynamics was reported by Chopra and Brown, who used suspended tracer particles to observe and measure acoustic streaming in He II [50]. Since then, there have been a number of visualization studies in liquid helium, including the qualitative visualization of Taylor-Couette flow in He II [51] and a series of quantitative studies of He II thermal counterflow jet using Laser Doppler Velocimetry (LDV) technique [52-54]. Compared to these techniques that have been employed (either direct observation or LDV), PIV is a more advanced flow diagnostic technique, having the ability to make quantitative wholefield velocity measurements with high accuracy and spatial resolution. Using PIV for visualization studies of liquid helium allows one to directly measure the velocity distribution, and thus more thoroughly explore the fluid dynamic and heat transport processes on small scales. Despite the fact that PIV has been an established technique and extensively applied for thermal-fluid studies of conventional fluids, such as water, air and various gases, the technique has not been previously applied to liquid helium experiments due to the following challenges: 1) The unique experimental environment of liquid helium, namely extremely low temperature and in some cases low pressure, make it difficult to conduct PIV

43

measurements. As discussed before, PIV is a nonintrusive and indirect measurement technique, which relies on tracking the motion of tracer particles to acquire the velocity of flow. So, tracer particles play a critical role in PIV. While there is a wide variety of tracer particles available for PIV measurements of conventional flows, many of them, such as gas bubbles and oil drops, are not applicable for liquid helium because of the experimental environment. 2) Liquid helium is one of the lowest density condensed fluids with a density about 1/8th of that of water at its normal boiling point ( T = 4.2 K), increasing by about 15% as the temperature is lowered to 2.2 K. In the superfluid state (He II phase), the density of saturated liquid helium is nearly constant at 145 kg/m3. This low density
P P

makes it difficult to find neutrally buoyant tracer particle for liquid helium. Also, it is more difficult to find tracer particles to accurately follow the flow because of the extremely low dynamic viscosity of liquid helium, which is around 310-6 Ns/m2 for He
P P P P

I at 4.2 K and 1.410-6 Ns/m2 for He II at 1.8 K.


P P P P

3) The seeding of particles has been assumed as a trivial issue for PIV experiments on conventional fluids. However, this task is not so easily accomplished for liquid helium. Due to its unique experimental environment, most of the traditional particle generation and seeding technologies that involve in using aerosol generator, atomizer, or Laskin nozzle, are no longer applicable. In addition, the particles have an increased tendency for coagulation in liquid helium. This places a stringent requirement on particle seeding. Because of these challenges, two major issues, particle selection and particle seeding, have to be addressed in order to successfully apply PIV to liquid helium.

44

4.2 Particle Dynamics in a Fluid


As indicated before, the selection of appropriate tracer particles is two-fold: on one hand the particles must be small enough to faithfully follow the flow; while on the other hand, they must be large enough to scatter sufficient light for image acquisition. Since it has already been pointed out that particles with diameter larger than 1 m are generally able to generate enough SNR for PIV measurements, selecting particles in this size range would primarily depends on their tracking characteristics, which can be described by two parameters, slip velocity and relaxation time. For an in-depth view of these two parameters, the particle dynamics in a fluid must be studied. Before the discussion of particle dynamics, for PIV measurements, one should first consider the effect of particle concentration. As we know, from the viewpoint of correlation process, the particle concentration should be as high as possible because more particles in an interrogation window considerably reduces the velocity bias resulting from particles entering or leaving the interrogation window. However, from the viewpoint of the flow field itself, the particle concentration cannot be too high otherwise the particles would interact with each other and affect the characteristics of the flow. Thus, the particle concentration must be in a certain range to achieve accurate correlation results without altering the flow field. Adrian has suggested an ideal particle concentration at about 10 particles per interrogation window [41]. Also, a more practical formula to approximate the seeding concentration, C s , required for PIV measurements has been determined by Gray [55], 24M 2 z d I2

Cs

(4.1)

45

where M is the magnification number of the camera, z the thickness of laser sheet, and
d I the length of a square interrogation window. These parameters can vary depending on

the applied PIV system. In our liquid helium PIV experimental set-up that will be discussed later, a 48 48 pixels interrogation window is used with the pixel size of 6.7 m, giving the length of interrogation window d I = 0.32 mm. The camera has a magnification number of M = 0.455 and the thickness of laser sheet is 2 mm. From equation (4.1), the desirable particle concentration is around 7.7109 particles/m3. If we
P P P P

further assume that the particles are dispersed homogeneously within the flow field, the average distance d s at which the particles are separated with each other can be simply estimated from the relation that,
3

1 + 1 = C s d s

(4.2)

From this equation, the average particle separation at the desirable particle concentration is about 0.5 mm. For the tracer particles with mean diameter d p less than 10 m, the ratio of d s / d p is greater than 50, which is large enough to ensure that both the
interaction between particles and their influence on flow field are negligible. For tracer particles of significantly larger diameter than 10 m, however, the interaction between particles may influence the flow field unless the particle concentration is reduced from the optimum value given by equation (4.1). Assuming that the influence of particles on the flow and the interaction between them can be ignored, the movement of a spherical particle with a diameter of d p and density of p in a fluid with density f and dynamic viscosity f along the direction of

46

the gravitational field can be described by the following one-dimensional partial differential equation [56],
3 d P P dv P

dt

2 d p C D f (v p v f

3 d P g( f P ) +

3 dp f

dv f dt

12

3 dp f

d (v f v p ) dt

(4.3) where v p is the particle velocity, v f the fluid velocity, and C D is the viscous drag coefficient. This equation is known as the Basset-Boussinesq-Oseen (BBO) equation. The left side gives the acceleration force acting on the particle. On the right side of the equation, the first term represents the viscous drag force; the second term is the body force due to the gravitational field and is assumed to be in the direction of the flow; the acceleration of the fluid leads to a pressure gradient in the vicinity of the particle and hence generate an additional force given by the third term; and the acceleration force of the virtual mass, which is equal to half the fluid mass displaced by the particle, is accounted for in the fourth term. In this equation, we have neglected the so-called Basset history term, which takes into account the unsteadiness of the flow field. This is because, for the dynamics of solid particles in liquid flow, the effect of Basset term need not always be significant [40]. When the particle Reynolds number Re p is very small ( Re p << 1 ), the viscous drag coefficient C D can be expressed by the Stokesdrag law,
24 Re p

CD =

(4.4)

where Re p is defined as,

47

Re p =

f d p (v f v p ) f

(4.5)

Equation (4.4) is a good approximation for most conventional flows (water and gas flows) due to the combination of extremely small particle size and moderate viscosity of the fluids. However, for liquid helium flows, the particle Reynolds number given by equation (4.5) may not meet the requirement of Re p << 1 because the dynamic viscosity of liquid helium is orders of magnitude smaller ( f ~ 10-6 Ns/m2). When Re p becomes
P P P P

finite, the viscous drag coefficient is modified as [57],


24 Re p

CD =

(4.6)

where accounts for the deviation from the Stokes drag law when the particle Reynolds number becomes finite. A commonly used expression for is

= (1 + 0.15 Re 0.687 ) p
Substituting for C D , equation (4.3) can be rewritten as,
18 f d
2 p

(4.7)

dv p dt

(v f v p ) + g ( f p ) +

dv 3 f f 2 dt

(4.8)

1 where = p + f . Depending on the state of the flow field, this equation can be 2 solved differently to describe the motion of particles in the fluid.

48

4.2.1 Motion of Particles in a Steady Flow

Considering a simple condition that the flow is in steady state, i.e. equation (4.8) can be simplified as,
d (v p v f ) dt

dv f dt

= 0,

18 f
2 dp

(v f v p ) + g ( f p )

(4.9)

Since is also a function of (v f v p ) , this equation is a nonlinear partial differential equation and may be solved numerically. If we consider the case that the particle velocity also reaches steady state, which means
dv p dt

= 0 , the ultimate velocity difference between the particle and the fluid,

termed slip velocity, can be solved for as,


2 gd p ( p f )

vslip =

18 f

(4.10)

When the particle density is not equal to that of the fluid, the particle velocity deviates from the actual fluid velocity. Since the PIV technique actually measures the velocity of particles rather than that of fluid, this slip velocity principally determines the uncertainty of PIV measurements. Also, the slip velocity represents the terminal velocity that a particle can reach in a stationary fluid, termed the settling velocity v s . From equation (4.10), measuring the settling velocity of particles in a fluid provides an easy way to indirectly estimate the particle size [58].

49

When the particle Reynolds number is very small, 1 , equation (4.9) can be solved analytically, giving the velocity difference as a function of time
2 gd p ( p f )

v f vp =

18 f

t + C1 exp s

(4.11)

where s is the relaxation time and in this case is


2 d p s = 18 f

(4.12)

and constant C1 in equation (4.11) can be determined using the initial condition of the particle. Assuming the particle velocity is v pi when injected into the flow, the solution of particle motion can be written as
t (v f v p ) = vslip 1 exp s t + (v f v pi ) exp s

(4.13)

From this equation, the velocity difference between particle and fluid is changing with time. The relaxation time describes the rate at which the particle catches up with the flow. When t >> s , the effect of initial particle velocity is negligible, and the velocity difference is equal to the slip velocity given by equation (4.10). For the specific case that the particle is neutrally buoyant in the fluid, i.e. p = f , the slip velocity between particle and fluid is zero. And the relaxation time becomes s =
2 pd p . 12 f

50

When the deviation from Stokes drag law, equation (4.7), has to be considered, there is no analytical solution for equation (4.9). But in this case the relaxation time can be approximated as,
2 d p s = 18 f

(4.14)

Since is always greater than 1, equation (4.12) actually gives a conservative estimation of the relaxation time when the particle Reynolds number is finite. As the same, a conservative estimation of the slip velocity can be derived from equation (4.10), which is
2 g ( p f )d p

v slip =

18 f

(4.15)

4.2.2 Motion of Particles in an Accelerating Flow

Now consider the situation with the particle initially at rest while the fluid undergoes an accelerating flow described by, v f (t ) = a t (4.16)

where a is the acceleration rate. The initial condition of the flow and the particle is v f (t = 0) = v p (t = 0) = 0 (4.17)

51

Substituting equation (4.16) into equation (4.8), one can get a t

dv p dt

vp =

f p 3 f g+ a 2

(4.18)

where s is either from equation (4.12) or (4.14) depending on the particle Reynolds number. The solution of this equation consists of two parts: a homogeneous solution v p ,h from the equation

dv p ,h dt

v p ,h = 0

(4.19)

and a particular solution v p , p . Equation (4.19) can be easily solved as,

t v p ,h = C1 exp s

(4.20)

where C1 is a constant. And the particular solution v p , p is assumed to have the following form, v p, p = A t + B (4.21)

Substituting this equation into equation (4.18), constants A and B can be determined as,
A=a

(4.22)

52

3 f f p B = 2 1a +

g s

(4.23)

Then the analytical solution of equation (4.18) can be written as, 3 f f p + a t + 1a + g s 2

t v p (t ) = C1 exp s

(4.24)

Now, C1 can be determined using the initial condition, equation (4.17). The final solution becomes, 3 f f p t 1a + v p = g s 1 exp 2 s + a t

(4.25)

When t >> s , the slip velocity between particle and fluid in an accelerating flow is,
3 f v f v p = vslip + 1 2 a s

(4.26)

where vslip is given by either equation (4.10) or (4.15) depending on the particle Reynolds number. Compared with the results for particle in a steady flow, the particle in an accelerating flow has the same relaxation time, but there exists an additional slip velocity between the fluid and particle resulting from the acceleration of flow. Usually, (a s ) is small because s is small, so the additional slip velocity can be neglected. But for some

53

flows that are associated with abrupt velocity changes, such as oblique shock in transonic flow and second sound shock induced counterflow that will be discussed later, a is a large number so that the additional slip velocity must be considered. It should be noted that these results can also be applied to decelerating flows.

4.2.3 Motion of Particles in Turbulent Flows

An extensive discussion of equation (4.8) and its solution for turbulent flow at finite particle Reynolds number has been given by Mei [57]. Here, we will just consider a simplified case, in which the turbulent flow is described as a time-averaged velocity u plus sinusoidal velocity fluctuations, v f = u + U 0 sin( t ) where U 0 and are the amplitude and frequency of the velocity fluctuation respectively. Based on the previous discussion, equation (4.8) can be rewritten as, 3 f dv f vslip 2 dt s (4.27)

dv p dt

vp t

vf

(4.28)

Same as equation (4.18), this equation can be solved analytically by finding the homogeneous and particular solution. In this case, the homogeneous solution is the same as given by equation (4.20), but the particular solution takes the following form, v p , p = C 2 sin( t + ) + C3 (4.29)

54

Substituting this solution into equation (4.28), we can get the constants C 2 , C3 and . Also, the constant C1 in equation (4.20) can be determined by applying the initial condition of particle, v p (t = 0) = 0 . The results are summarized as below,

t v p (t ) = C1 exp s

+ C 2 sin ( t + ) + C3

(4.30a)

C1 = v slip u

tg U0 1 St tg

(4.30b)

C2 =

1 + (tg ) 2 U0 1 St tg

(4.30c)

C3 = u vslip

(4.30d)

tg = St

1 a 1 + a St 2

(4.30e)

where the Stokes number St represents the ratio of particle relaxation time to the characteristic time of the fluctuating flow, St = s and a is given by (4.31)

55

a=

3 f 2

(4.32)

When the particle density is much larger than the fluid density, i.e. p >> f , a is very close to zero. In this case, the solution (4.30a-e) can be simplified as,
U 0 sin( t + )
1 + St 2 t U St + 0 2 + vslip u exp 1 + St s

v p (t ) = (u vslip ) +

(4.33)

and the phase lag can be determined as, tg = St (4.34)

From equation (4.33), when t >> s , the particle velocity in a turbulent flow also consists of two parts: a time-average velocity and a sinusoidal velocity fluctuation. Comparing to the fluid velocity given by equation (4.27), the velocity difference between the particle and fluid comes from two sources: one is a time-independent regular slip velocity vslip as given by equation (4.10) or (4.15), and the other is the difference between their fluctuating amplitudes, which is usually measured as an amplitude error given by,

U0

U0 1 1 + St 2 = 1 U0 1 + St 2

(4.35)

56

For high frequency fluctuating flows, i.e. a large , the amplitude error as well as the phase lag may becomes very large. To ensure that the particle is still able to accurately track the fluctuating behavior of turbulent flow, it must have an extremely small size, which makes the relaxation time s very small. Another special case is when the particle density is the same as that of the fluid, i.e. p = f . From equation (4.32), we can get a = 1 . Substituting into equations (4.30ae), the particle velocity is t v p (t ) = u 1 exp s + U 0 sin( t )

(4.36)

It can be seen that both the amplitude error and the phase lag are zero in this case. The only velocity difference exists in the time-averaged velocity, but it is negligible when t >> s . Therefore, when PIV is used for measuring the turbulent flow, it is highly recommended to use the neutrally buoyant particles.

4.3 Selection of Tracer Particles


In order to make successful PIV measurements, one critical task is to select the appropriate tracer particles. Generally, the particles are required to have a small slip velocity and a short relaxation time. A quantitative criterion to determine how small the slip velocity should be can be set as,
vslip v f

(4.37)

57

where is an arbitrary proportionality factor typically of order a few percent. For high velocity flows, it is relatively easy to meet the above criterion. However, when the flow velocity is low, this criterion may not be satisfied unless there is a very small slip velocity. In this case, the slip velocity must be taken into account when doing quantitative analysis of PIV measurements. For steady flows, the relaxation time does not affect the accuracy of measurements, but it has to be considered for determining how long it takes the particle to acquire the flow velocity. For transient or turbulent flows, the relaxation time becomes more important since it also affects the measurement accuracy in addition to just the response time (see equations (4.26) and (4.35)). Usually, it is required that the particle relaxation time be much smaller than the characteristic time constant of the transient or turbulent flows. To select tracer particles for liquid helium, the previous discussion of particle dynamics in a fluid is applied. For normal helium or He I, the above theory can be applied because He I behaves like a classical Navier-Stokes fluid with associated static and dynamic properties. However, for superfluid helium or He II, the inviscid nature of the superfluid component of He II may bring into question the applicability of previous discussion in this unique fluid system. To a certain extent this issue is addressed in the dissertation research. However, there is evidence that the drag coefficient on a solid sphere in He II obeys classical correlations, indicating that a sphere interacts with He II flow as if it were a classical fluid [59]. Therefore, just for the purpose of selecting potential tracer particles, we assume that the motion of particles in He II follows the same dynamics equation as in classical flows, so that all the previous analysis results of particle dynamics are also valid for He II. Since only the normal fluid component of He II has viscosity, the fluid viscosity f in previous discussion should be replaced with the normal fluid viscosity of He II, n . At the same time, one would expect for He II flows that only the normal fluid velocity could be measured with PIV technique.

58

From equations (4.10), the slip velocity becomes zero if the tracer particles have the same density as the fluid. However, the very low density of liquid helium makes it hard to find neutrally buoyant particles. One practical approach is to select the particles with density very close to that of liquid helium. This can reduce the slip velocity, but it does not help to reduce the relaxation time, which is given by equation (4.14). From this viewpoint, a better approach is to select particles with extremely small particles size, which would make both the slip velocity and relaxation time small.

Table 4.1 Tracking Characteristics of Tracer Particles in Liquid Helium


He I (4.2 K) Type of particles Hollow glass spheres Supplier He II (1.8 K)

p
(kg/m3)
P P

dp
(m) 30 120 20 100 8 12

v slip (mm/s)
5.7 (3.9*)
P P

(ms)

vslip (mm/s)
P

(ms)

3M PQ TSI

160 160 200 200 1100 1100

3.7 (3.4*)
P P P

5.5 (3.1*)
P P P P

9.0 (8.1*)
P P P P

90.4 (20.0*) 5.4 (4.0 )


P P

59.4 (53.8*) 2.0 (1.8 )


P P

88.7 (13.8*) 9.2 (5.1 )


P P

143.2 (129.8*)
P P

4.7 (4.2 )
P P P P

135 (27.6*)
P P

48.6 (44.1*)
P P

230 (26.8*) 25.6 (13.8 )


P P

117 (106*)
P P

11.3 (8.6 )
P P

1.4 (1.25 )
P P P

3.2 (2.9 )
P P

25.5 (15.7*)
P

3.1 (2.8*)
P P

57.6 (22.5*)
P P

7.2 (6.5*)
P P

Polymer microspheres Solid deuterium Solid neon Solid H2/D2


B B B B

Bangs Laboratory

1100 206 1150 140

1.7 10 10 10

0.51 (0.42*)
P P

0.06 (0.06*)
P P

1.2 (1.0*)
P P

0.15 (0.13*)
P P

1.5 (1.26*)
P P

0.50 (0.45*)
P P P

2.5 (2.0*)
P P

1.2 (1.1)
*
P P

18.6 (12.6 )
P

* *
P P

2.3 (2.0 )
P P

42.1 (19.0 )
*
P P

5.2 (4.7*)
P P

0.27 (0.18 )

0.38 (0.34 )

0.2 (0.11 )
P P

0.9 (0.82*)
P P

Note: The values with a superscript * are calculated from equations (4.10) and (4.14), where the effect of

is considered. The values without * are from equations (4.12) and (4.15), where is regarded as 1.

59

We have tried both the above approaches, and many different kinds of particles have been selected, including a variety of commercially available solid particles as well as solid particles generated by freezing liquids or gases. The particle tracking characteristics, i.e. slip velocity and relaxation time, in liquid helium (He I and He II) have been calculated and listed in Table 4.1. In the following, potential uses of these particles for tracking liquid helium flows are discussed.

4.3.1 Hollow Glass Spheres

Hollow glass spheres are manufactured in a variety of sizes and densities and can be obtained from a number of commercial vendors. 3M, Inc. and PQ, Inc. supply particles with a nominal density close to that of liquid helium; however, in that case the individual spheres have a very wide size range varying from 20 m to over 100 m. Some issues that need to be considered if these particles are to be used for tracking liquid helium flows include: a) Resulting from the large size distribution, only a very small portion of the particles have neutral density. Thus, most of the particles cannot be suspended in liquid helium. This problem can be reduced in its significance by using sieves to narrow the particle size distribution. However, this approach represents a significant handling effort. b) The relatively large particle size and wide size distribution make it very difficult, if not impossible, to conduct quantitative measurements of flows with low velocity. The slip velocity calculated for these particles ranges from about 5 to over 100 mm/s depending on the particle size. Therefore, to achieve accurate PIV measurements the fluid velocity needs to be in excess of a few meters per second. Many experiments of interest have considerably smaller velocities. For example, the normal fluid velocity in thermal counterflow is usually less than 100 mm/s for moderate applied heat flux.

60

c) The relaxation time for these particles is of order 1 to 100 ms. For steady flows, this relaxation time may not affect the measurements. But for turbulent or transient flows, the relaxation time is too large, and would result in additional slip velocity as well as a large phase lag and amplitude error when the flow velocity is fluctuating at moderate frequencies. Therefore, these particles are unacceptable for tracking very turbulent flows and other transient phenomena. Murakamis group was the first to use hollow glass spheres for the visualization and Laser Doppler Velocimetry (LDV) measurements of a He II thermal counterflow jet [52]. Similar particles has also been used in our preliminary experiments of using PIV to measure He II thermal counterflow [15]. Both sets of experiments achieved a moderate level of success although quantitative comparison was not fully realized. Hollow glass spheres of smaller size (8-12 m) but higher density (mean density of 1100 kg/m3) can be obtained from TSI, Inc. Compared to hollow glass spheres with
P P

low density, these particles have a much narrower size distribution and a smaller diameter. The slip velocity for these particles is in the range of 20 to 50 mm/s, which is still too high for tracking low-speed flows in liquid helium. However, due to the narrow size distribution, the slip velocity can be determined within a certain range. As indicated before, the slip velocity can be actually measured by means of measuring the particle settling velocity in stationary liquid helium (see equation (4.10)). Thus, we are able to use the measured slip velocity to correct the PIV measurement results in the presence of flow. The relaxation time for these smaller hollow glass spheres is on the order of several milliseconds, which is much better than that of low-density hollow glass spheres. This makes these particles suitable for the measurement of steady flows, and turbulent flows with moderate velocity change rate.

61

4.3.2 Polymer Micro-Spheres

Polymer micro-spheres are one of the most popularly used tracer particles and have been applied extensively in flow measurements, ranging from ultra-high speed flows like transonic airflow [60] down to super-low speed flows like fluidics flow in MEMS (micro-electro-mechanical system) [61]. These particles have a density very close to that of water (mean density of 1100 kg/m3) and a size range from several microns to
P P

sub-microns. As tracer particles, the most attractive property is their extremely narrow size distribution, normally with a standard deviation less than 5%. Since the probability that particles will stick together is a strong function of the
2 particle diameter ( d p ), the small particle size of polymer particles causes an inherent

problem associated with them aggregation. A common way to prevent aggregation is to store and supply these particles in the form of suspension in water. This approach works well if the particles are used for conventional fluids, such as water, air, and gases. But, when used in liquid helium, it is unsuitable because the extremely low temperature of liquid helium (below 4.2 K) would make water or any content of moisture immediately freeze, causing the particles to stick together. Though some techniques can be applied to absorb or evaporate the water, these techniques require very careful handling. Therefore, polymer particles supplied in the form of dry powder are preferred. There are hundreds of types of polymer micro-spheres available from various sources so it is a time-consuming work to find an appropriate one. We selected the polymer micro-spheres from Bangs Laboratories mostly because they have a very small mean diameter of 1.7 m and are supplied in the form of dry powder instead of suspension. From Table 4.1, these particles have a computed slip velocity of around 1 mm/s, making them applicable for both low-speed thermal counterflow in He II and highspeed forced flows in liquid helium. Also, the relaxation time is on the order of 0.1 ms,

62

which is small enough to allow them to track the transient and turbulent flows in liquid helium.

4.3.3 Solidified particles

Solid particles made from mixtures of hydrogen and deuterium can achieve near neutral density in liquid helium and be made small enough to support visualization experiments. As is discussed below, there are a number of methods to produce these particles but all techniques involve either condensing from a gas mixture or freezing a fine liquid spray into a liquid helium bath. Depending on the technique in application, the particle diameter can range from sub-micron to millimeters. Compared to commercially available solid particles, solidified particles have four distinct advantages: 1) The particle size and size distribution can be controlled by varying experimental parameters, such as the pressure difference between gas source and liquid helium bath, the diameter of injection nozzle and the distance between nozzle and liquid helium level; 2) The density of particles can be adjusted by using different gases or varying the mixture ratio of different gases; 3) Particles can be supplied continuously, making it easy to control and adjust the seeding concentration; and 4) The particles can be easily removed from the experiment by returning to room temperature and evacuating the system. The principal disadvantage of these particles is their difficulty in generation. Because of the wide range of particle parameters available through this approach it is difficult to quantify their tracking characteristics. Solid hydrogen, deuterium, and neon particles have been produced in our laboratory by vapor condensation and liquid spray injection. The neon particles have a relatively high density (1150 kg/m3) compared
P P

to that of liquid helium and thus a larger slip velocity (> 10 mm/s) and relaxation time (about 5 ms). For solid deuterium particles or H2-D2 mixtures, since their density is close
B B B B

63

to that of liquid helium, the very small slip velocity makes them suitable for measuring thermal counterflow and other low-speed helium flows, and the relaxation time of less than 1 ms makes them applicable for transient and turbulent flow measurements. From the above discussion, it is concluded that: of the commercially available solid particles, polymer micro-spheres with a mean diameter of 1.7 m and specific gravity 1.1 are good candidates for tracking low-speed flows in liquid helium because of their small size; solidified hydrogen/deuterium (H2-D2) particles show potential to have
B B B B

the best tracking fidelity and fastest response due to their small size as well as density very close to that of liquid helium; hollow glass spheres are somewhat less useful mostly due to their larger size, but they may be applicable for the measurement of steady flows with relative high velocity, such as liquid helium forced flow.

4.4 Tracer Particle Seeding


Selecting the appropriate tracer particles is only part of the process for seeding flows. After that, the tracer particles need to be seeded into the flow field. A general requirement for particle seeding is that the particles should be introduced into the flow in sufficient and stable concentration, with a spatially uniform distribution. In addition, the particle coagulation should be minimized because it tends to produce a lower particle concentration and larger particle size, both of which cause erroneous results due to correlation error and large slip velocity. For conventional fluids, the issue of particle seeding has been well addressed, and a variety of technologies have been developed to seed the particles into liquid or gas flows [40]. But the unique experimental environment of liquid helium makes many of these technologies unsuitable. For example, one commonly used technique to seed solid particles is through the atomization of suspensions. As shown in Figure 4.1, the high

64

speed air jets are created by small nozzles submerged in the suspensions consisting of solid particles and suspending liquid (usually water). Due to the shear stress induced by the air jets, small droplets are generated. The droplets first formed by the atomization process are polydisperse, but the aerosol of solid particles will be monodisperse after evaporation of the suspending liquid if no droplet contains more than one solid particle, which can be attained by using a dilute suspension. This technique has been widely used and accepted as a standard [56]. However, when it is applied for liquid helium, the droplets that have not been fully evaporated will freeze immediately after they come in contact with the very cold liquid helium. Even if the droplets can be fully evaporated, the vapor as well as the air that carries the solid particles will condense and then freeze quickly after their injection into liquid helium. In both cases, larger particles are generated in addition to the solid particles to be seeded.

Figure 4.1 Laskin nozzle used for seeding solid particles [40].

65

So, considering the challenges associated with the unique experimental environment of liquid helium, some new methods have to be developed for the seeding of particles.

4.4.1 Seeding of Solid Particles

As indicated before, the solid particles to be used for PIV measurements usually have a very small particle size ranging from sub-micron up to over 100 microns. When these particles are supplied in the form of powder, it is inevitable that the small gaps between them will trap air and moisture from ambient environment. Therefore, before they are seeded into liquid helium, one important thing is to remove the moisture content and also the trapped air. Otherwise, they will freeze when the particles are introduced into the cold liquid helium, resulting in severe coagulation, larger particle size, and irregular particle shape. To attain this special requirement, we have developed a simple seeding method based on the well-established fluidized bed technique [62]. As shown in Figure 4.2, the main part of this solid particle seeder is a vertical cylinder, which is used not only to remove the trapped air and moisture but also to disperse particles. The whole particle seeder is separated by a ball valve into two parts: the top part is a 200 mm long tube connected to a vacuum pump through valve #2, and the bottom part is simply a fluidized bed with a porous ceramic plate (maximum pore size of 6 m) to aerosolize the seeding particles. A heater is submerged in the particles, and a thermometer is used to monitor the heating temperature. A high-pressure pure helium gas (100 kPa at gage pressure) is supplied from the bottom through control valve #1. Another needle valve #3 is placed before the exit, which is extended to the inside of liquid helium bath through an injecting tube.

66

Figure 4.2 A simple particle seeder based on the fluidized bed technique [62].

Before the seeding, particles are first heated over 100 C for more than eight hours to evaporate the moisture content of particles. During this heating process, valve #2 and the ball valve are both open for pumping out the evaporated vapor. Following the heating process, a purging process is performed to remove the air trapped between particles. Valve #2 is closed and valve #1 is open to make the particles fluidized by the high-pressure helium gas. After about one minute of fluidization, the pressure across the porous ceramic plate is balanced, and some particles suspended in the upper part will fall down on the ceramic plate due to gravity. During this fluidization process, the trapped air can be partially forced out of the particles by helium gas. The ball valve is then closed and valve #2 is open to pump out the air. After the upper part is pumped down below 3 kPa, valve #2 is closed and the ball valve is open for another fluidization process. This

67

purging process is repeated more than ten times to make sure that there is no more air trapped between particles. After that, the ball valve is closed while valve #1 is kept open to maintain a helium gas environment inside the seeder. When seeding is needed, needle valve #3 is opened so that particles will be carried into liquid helium by the pure helium gas.

(a)

(b)

Figure 4.3 Seeding results for two different kinds of tracer particles: (a) hollow glass sphere from 3M, Inc.; (b) polymer particles from Bangs Laboratories, Inc.

This method has been successfully applied for seeding hollow glass spheres in liquid helium. Figure 4.3 (a) shows one example of the seeding results, where the hollow glass spheres from 3M, Inc. with a size range from 30 to 120 m were dispersed into a flow channel through the helium vapor phase. It can be seen that a high particle concentration is achieved, which is sufficient for conducting PIV correlation process. To further evaluate the seeding performance, the settling velocity of seeded particles was

68

measured using PIV technique, giving a range from 2 to 16 mm/s. Based on this settling velocity, the size of seeded particles estimated from equation (4.10) is from 40 to 108 m, which is consistent with the size range and particle density provided by the supplier. This indicates that the seeding particles are dispersed very well and there is no severe coagulation of particles during the seeding process. Using this method, we have also tried to seed polymer particles into liquid helium. But the result shows an unsuccessful seeding with insufficient and unstable particle concentration and also severe particle coagulation, which is primarily due to the limitation of conventional fluidized beds and a higher tendency for these smaller particles to aggregate together. In an effort to improve the stability of the particle seeding and to reduce the particle coagulation, a two-phase fluidized bed can be applied. As illustrated in Figure 4.4, the bed contains a mixture of large size beads with diameter from 100 to 200 m and the small size particles to be seeded. The upwards gas velocity is set between the gravitational settling velocities of the powder particles and the beads, so that the beads are fluidized but not entrained while the powder particles can move freely in the volume between the beads. In this case, the continuous friction between the beads can significantly reduce aggregation of the power particles, and the resulted particle concentration is more stable. To apply this two-phase fluidized bed technique to the seeding of polymer particles, two kinds of particles have been put inside the seeder shown in Figure 4.2: one is the polymer particles from Bangs Laboratories with a mean diameter of 1.7 m and a specific gravity of 1.1, and the other is solid glass spheres from PQ Corporation with a means diameter of 156 m and a specific gravity of 2.5. After undergoing the purging process as described previously, the pressure head of pure helium gas is carefully adjusted so that the solid glass spheres are fluidized while the polymer particles are seeded into liquid helium. Figure 4.3 (b) shows a typical seeding result by using this twophase fluidized bed method. Compared to the result shown in (a), the particle

69

concentration in this case is higher, and also it has a quite uniform spatial distribution, which is very helpful to reduce the uncertainty of PIV correlation process.

Figure 4.4 Two-phase fluidized bed [40].

4.4.2. Generation and Seeding of Solidified Particles

For particles solidified from gases or liquids, the seeding process is combined with the generation process. When the particles are generated, they have also been seeded into liquid helium. Therefore, the following discussion is mainly focus on the generation of solidified particles. The method of injecting a mixture of gasses into liquid helium to produce small solid particles for visualization experiments was first demonstrated by Chopra and Brown [50]. In their experiment, a roughly 50-50 mixture by volume of hydrogen and deuterium gas was injected into liquid helium through an 8 mm diameter tube heated by a spiral

70

heater attached on the outside surface. Using this method, they were able to generate neutrally buoyant H2-D2 particles of diameter less than 1 mm. Murakami used the same
B B B B

method to generate solid H2-D2 particles from a premixed H2-D2 gaseous mixture [52]. In
B B B B B B B B

their experiment, the injection tube was heated up to about 130 K, and the gas solidified when it interacts with cold helium vapor and contacts with liquid helium. In the initial stage, the particle size was very small (around 1 m). After passing through a wire screen, they then grew to about an order of 100 m in diameter due to the partial solidification and coagulation of particles. Gordon and Frossati have developed a similar method to produce solid particles from deuterium gas. To prevent the particles from sticking together before they entered liquid helium, the deuterium gas is highly diluted using pure helium gas (mix ratio D2/He
B B

from 1:20 to 1:1000) [63]. Also, in order to generate very fine particles, a small orifice (about 0.75 mm) at the bottom of the injection tube was used to provide high gas velocity. Using this helium jet technique, solid deuterium particles of diameter less than 100 nm have been continuously produced under a pressure difference of 1-2 kPa between the gas source and helium bath. Obviously, the generated particles are too small to scatter sufficient light of image acquisition. In order to apply this technique to PIV in our experiments, some modifications have been made. The deuterium and helium gas were premixed at a ratio of 1:50 by volume, and the gas mixture was injected directly into liquid helium through a 0.5 mm diameter injection tube. The result shows that the solid deuterium particles generated in this case have a diameter from 1 to 10 m, which is large enough to be used for PIV measurements. But it should be noted that the concentration of seeded particles is not stable. Thus, in some cases, the particle concentration is not high enough for a successful PIV experiment. In addition to being generated directly from the gaseous phase, solidified particles can also be produced by atomization of liquid [58]. In this process, the gas is first liquefied by the cold helium vapor, and stored inside a supply reservoir. Then pressurized

71

pure helium gas is used to inject the liquid into liquid helium through a nozzle. When exiting the nozzle, the liquid stream is ruptured and disintegrates into very fine individual droplets by the high velocity gas flow. After these droplets encounter the cold helium vapor or liquid helium, they are solidified and dispersed into liquid helium simultaneously. One advantage of using this method is that the particle size and concentration can be readily controlled by adjusting the nozzle diameter and regulating the driving pressure head. A seeding concentration as high as about 109-1010 particles/m3 can be achieved by
P P P P P P

atomization at room temperature [40]. However, for liquid helium, the concentration can be much smaller due to helium vapor counter flow and coagulation of particles during the solidifying process. One inherent limitation of atomization is the production of polydisperse spray, i.e. the droplets at the core of the jet are typically larger than those near the sides [64]. This results in a wide size distribution, which may greatly affect the accuracy of PIV measurements. To overcome this problem, an impactor (shown in Figure 4.1) that is able to catch the larger droplets with higher momentum can be placed at a certain distance from the injecting nozzle. Thus, only droplets with smaller size can be solidified and seeded into liquid helium.

4.5 PIV Experimental Set-up


To conduct He II thermal counterflow experiments using the PIV technique, it is required that the experimental set-up consists of two sub-systems: one is the PIV system operated at room temperature, the other is the low temperature system which includes an optical cryostat and a thermal counterflow channel. Figure 4.5 shows a schematic diagram and a picture of the experimental set-up. It can be seen that the PIV system includes five major components, i.e. a laser, beam expanding optics, a CCD camera, a

72

Cryostat Laser heads Optical windows Beam dump Optics Computer

CCD camera Timing control unit


(a)

Timing unit

Cryostat

Window Camera Optics

Laser heads
(b) Figure 4.5 PIV experimental set-up: (a) schematic diagram; (b) the real system.

73

synchronizer for timing control and a high-performance computer for data acquisition and analysis. The characteristics of each component, and also the cryostat and counterflow channel are discussed in detail below.

1. The Laser
U

Lasers are usually used as the light sources for PIV measurements because of their ability to emit monochromatic light with high energy density, which can easily be bundled into thin light sheet for illuminating and recording the tracer particles without chromatic aberrations [39]. There are two main categories of lasers: continuous and pulsed, both of which have been widely applied to PIV experiments with conventional fluids. For liquid helium experiments, the heating by absorbed light may considerable affect the characteristics of the flow under study, and thus must be minimized. Continuous lasers are less desirable as they provide a steady background heat loading to the experiment. Pulsed lasers can be classified into single pulse lasers like the ruby laser and repetitive lasers, such as copper vapor or Nd:YAG laser. Since single pulse lasers can only be operated at a very low repetition rate, usually they are also not suitable for PIV experiments. A compact Nd:YAG laser system with dual laser heads from New Wave Research, Inc. is used in the present experiment. It is designed to supply a highly stable, pulsed green light (wavelength = 532 nm) source with a pulse duration of 3-5 ns and a maximum repetition rate at 15 Hz. The maximum emitting energy is 120 mJ per pulse, but it can be adjusted by setting the attenuator. Since the heating by absorbed light must be minimized for liquid helium, the attenuator is set at around 100 during our experiment. From Figure 4.6, the emitting energy per pulse in this case is only about 2 mJ, which is very low but is enough for recording the tracer particles. One attractive feature of this laser system is its dual head configuration, which allows each head to be fired

74

independently. Thus, the inter-pulse separation between the first and the second laser pulses can be freely set from less than 1 microsecond up to many milliseconds. Also, the laser can be operated in both internal triggering and external triggering mode. In our experiment, the external triggering mode is used because it allows customized control over the timing of the laser.

Figure 4.6 Laser output vs. attenuator reading.

2. The Camera
U

The camera used in our PIV system is a sharpVISION 1300-DE digital camera
P P

from IDT, Inc., which features a high-performance progressive scan interline CCD sensor. It has a high spatial resolution at 1300 (H) 1030 (V) pixels, with a pixel size of 6.7 6.7 m. The data are transferred through an IEEE-1394 interface with a high transfer rate at 400 Mb/sec, which allows fast image acquisition. There are two operation

75

modes for this CCD camera. When operated in trigger mode, each image is acquired from two exposures, one from the first laser pulse and one from the second laser pulse. Because the two exposures are recorded in one single image, the directional ambiguity exists for this mode. When operated in double-exposure mode, each image is acquired from only one exposure. The laser we used is dual head, which means two laser pulses are generated for each laser firing, therefore, in double-exposure mode, a pair of images are acquired every time the laser is fired. As the time-sequence of these two images are known, there is no directional ambiguity in this case. In both the trigger and doubleexposure mode, an external trigger input is required, and also the camera needs to be synchronized with the laser.

3. The Synchronizer
U

For timing control of the camera and the laser, a synchronizer is needed to generate and output the trigger signals. In our PIV system, the task of synchronization is accomplished by a timing control unit and a laser/camera trigger control program. The trigger signals that are used to control the laser and camera are listed in Table 4.2. The flash lamp and Q-switch signals are used to control the Nd:YAG laser, which is operated in external triggering mode in this case. Since the laser is dual head, two pairs of trigger signals are required, with A for the first laser pulse and B for the second laser pulse. Also, there are two TTL signals for triggering camera, which are both used only when conducting 3-D PIV. In our case, only one of them is needed as 2-D PIV is conducted. With the trigger control program, the timing unit supports three modes of operation: internal, external and synchronous. During our steady state measurements of thermal counterflow, the internal mode is used. Shown in Figure 4.7 is the control panel and timing diagram for the internal mode. Under this mode of operation, an internal clock generates all of the required timing signals. The laser and camera triggering signals are

76

derived from this clock. The control panel is used to set the desired rate for laser operation (Laser Frequency), the separation between laser pulses (Pulse Separation), the required delay between the flash lamp and the Q-switch triggers (Q-Switch Delay A
and B), and the camera frequency (Camera Freq Divider). In this program, the laser

frequency can be set from 0 to the maximum rate at 15 Hz, while the pulse separation and the Q-switch trigger delay may be adjusted from less than 1 microsecond to 10 milliseconds. The timing diagram below the panel shows that the camera is triggered synchronously and phase locked with the laser when the Camera Freq Divider is set as 1. Additional control allows the camera to be triggered at a sub-harmonic frequency.

Table 4.2 Trigger Signals Used for Timing Control of Laser/Camera


Signal TRG IN TRG OUT CAMERA A CAMERA B FLASH LAMP A FLASH LAMP B Q-SWITCH A Q-SWITCH B Description External trigger signal input Trigger output Camera A trigger Camera B trigger Flash lamp A trigger Flash lamp B trigger Q-Switch A trigger Q-Switch B trigger Timing Reference T0
B B

Level TTL High-True TTL High-True TTL High-True

T0
B B

T0
B B

T0
B B

TTL High-True
B

T0
B

TTL High-True
B

T0
B

TTL High-True
B

T0 + configurable delay
B

TTL High-True TTL High-True

T0 + configurable delay
B B

77

Figure 4.7 Timing control panel and timing diagram for internal mode.

Figure 4.8 shows the control panel and timing diagram for the synchronous mode. In this mode, a TTL compatible external drive signal is required at TRG IN. After a certain time delay set by the TRG Delay, this TRG IN signal initiates a gate with an adjustable length, which can be set by the Delay Cycles of the period of the laser rate. During the high state of the gate, the flash lamp of the laser operates at the frequency set by the control panel in the internal mode (Figure 4.7). But, the camera and Q-switch

78

triggers only follow the last pulse of this gated signal. The timing diagram below the panel shows the configuration of three delayed cycles.

Figure 4.8 Timing control panel and timing diagram for synchronous mode.

For the transient measurement of second sound and heat diffusion, the synchronous control mode is used because of its ability to set the timing for the heater and the PIV data acquisition. Figure 4.9 shows the control scheme. A rectangular pulse

79

generated from the pulse generator goes two ways: one is to the heater through a highspeed power supplier, and the other is to the Trigger in input of the timing control unit. By setting the Delay Cycles to one, and the TRG Delay as tT, the delay between the
B B

start of pulse heating from the heater to the acquisition of the first image (Q-SWITCH signal) is equal to 100 s (the time needed for set up the triggering circuit) plus the Qswitch delay (set by the user) and plus tT. To check this time delay, a synchronized
B B

signal from the power amplifier and the Camera out signal are both fed into a HP universal counter to make an in-situ measurement.

t0
B

HP 8116 A Pulse Generator

NF 4015 Power Amplifier

t0 t0
B B

Second Sound Heater

t0
B

Trigger in Camera out

t0+t1+tT
B B B B B

HP Universal Counter

Nd:YAG Dual Head Laser

t0+t1+t2+tT
B B B B B B B

Q-Switch out Flash Lamp out Timing control unit

t0+t1+tT
B B B B B

Figure 4.9 Synchronization of second sound pulse and PIV acquisition (t1=100 s, t2 is the Q-Switch delay set by the users, and tT is the trigger delay).
B B B B B B

80

4. The Optical System


U

The beam emitted out of the laser has a diameter of 5 mm, and cannot be directly used for PIV without being transformed into a sheet of light through light sheet optics. In our case, two spherical and one cylindrical lens have been used for this purpose. As shown in Figure 4.10, by first arranging the converging and diverging spherical lens, the beam diameter decreases from 5 mm to about 1 mm. It then passes through a cylindrical lens, which makes the beam expand in one direction to form a light sheet while maintaining the light thickness in the perpendicular direction as constant of about 1 mm.

Figure 4.10 Generation of a light sheet with constant thickness.

In addition to the optics for generation of the light sheet, there is another optical system, i.e. the camera lens. In PIV, there are two critical requirements for the optical setup of the camera: one is that the image size of particles should be larger than or on the order of the pixel size, which is 6.7 m for the CCD camera used; the other is that the

81

depth of field should be close to the thickness of light sheet, which is about 1 mm as given above. In an effort to meet these two requirements, the distance between the light sheet and the camera lens is set as d o = 160 mm, and the f -number is set as f # = 5.6 . As the optical lens of the CCD camera has a focus length f = 50 mm, the magnification number
M under this configuration is given by

M=

f = 0.455 do f

(4.38)

With M known, the image size of the particles can be calculated using equations (3.2) and (3.3). Since the diameter d s resulting from the diffraction of the lens is around 10.6 m in this case, for particles with diameter d p less than 10 m, the image size is primarily determined by d s at around 10 m; and, for particles with diameter larger that 50 m, the image size is primarily determined by the magnification number and the actually particle size. It can be seen that the image size in both cases is larger than the pixel size. From the optical point of view, when the camera lens is set, only one object plane can be in focus. Thus a point source which is not in that plane will have a blurred image with larger size. To reduce this effect, the depth of field is defined as the maximum distance of the object to the object plane, which means a sharp image can be attained when the particles are within the depth of field. Gaussian optics calculation gives the depth of field z as [43],
2

1+ M # 2 z = 2.44 (f ) M

(4.39)

82

Substituting M , f # , and the wavelength = 532 m into this equation, the depth of field for our optical set-up of camera is z 0.4 mm. This gives an optimum light sheet thickness of 0.8 mm, which is close to the actual thickness of about 1 mm. Due to the use of above optical systems, a calibration procedure is needed every time the system is set up or re-configured. The calibration image of an object with its physical dimensions known is first captured by the CCD camera. After that, a calibration program is used, which allows the four handles of a rectangle to be moved and positioned as close as possible to the borders of the objects image. By specifying the pixel size of the CCD sensor, physical size of the object, and focal length of the camera lens, a magnification factor that converts the pixels into millimeters is computed. Using this magnification factor, the particle displacements in the image plane which have the units of pixels can be automatically converted into the physical displacements of particles in the object plane having the units of millimeters. A very high accuracy of calibration can be achieved when the camera is perfectly aligned in the horizontal plane.

5. The Optical Cryostat


U

To conduct PIV experiments in liquid helium, an optical cryostat with optical windows on three sides is required: one side is for laser sheet in, one for laser out, and the third for camera to capture images. To ensure a good transmission of the laser light, optical windows made of fused silica are used. One big challenge associated with optical cryostat is the sealing of glass window on the stainless steel flange because the total thermal contraction of glass, from room temperature to liquid helium temperature, is at least one order of magnitude smaller than that of most metal. This issue becomes more challenging in our case, where the optical window is very large (63 mm in diameter) and thick (6 mm in thickness). After trying

83

many different designs, a final design as reported in reference [65] was used and gains a success. The window-flange sealing using this design survived more than ten cycles of cooling-down and warm-up between room temperature and 1.6 K. There is no leak for superfluid helium, and the optical windows remains intact.

6. The Counterflow Channel


U

The PIV measurement of thermal counterflow was conducted in a counterflow channel. As shown in Figure 4.11, the channel is made of 6.5 mm thick G-10 plate with a cross section area of 43 mm in the y direction and 19.5 mm in the x direction. The total length of the channel is about 200 mm. This parallel-sided configuration with long length is desired for maintaining the one-dimension character of thermal counterflow flow. The whole channel is submerged in He II bath, with the top end open to the bath and the lower end enclosed by a 12 mm thick G-10 substrate. As G-10 is an epoxy resin and glass textile composite and has a very small thermal conductivity compared to liquid helium, it can be assumed that the heat leak though the channel walls and the substrate during the counterflow measurements is negligible. To make PIV measurements, three optical windows are placed on three sides of the channel. Each of the two windows on the side walls has a size of 50 mm (z) 2 mm (y), and are placed at the central line of wall. The 2 mm in the y direction is chosen because the light sheet is only 1 mm thick. On the front wall is a 50 mm (z) 20 mm (x) optical window. But the actual view field that can be recorded by the CCD camera is only about 20 mm (z) 20 mm (x). All these windows are glued to the G-10 channel using Stycast 2850 with Catalyst 24LV. To reduce the reflection of light, the inner surface of the channel is coated with a thin layer of graphite. An injecting tube with 1.0 mm inner diameter and 1.5 mm outer diameter is inserted inside the channel to introduce the tracer particles. After the particles are injected inside the channel, it is assumed that the normal

84

fluid will drive them to flow upward, and images will be captured when they passing through the view field.

Figure 4.11 Schematic of the counterflow channel (T1 is a thermometer used to monitor the bath temperature).

85

A customized Kapton heater from Electro-Flex Heat, Inc. is bonded to the G-10
P P

substrate and used to generate the thermal counterflow or second sound shock. The total resistance of the heater is 157 ohms at room temperature and 150.5 ohms at liquid helium temperature. The heating element is made of nichrome, covered with a 0.02 mm thick insulation layer made of Kapton. As shown in Figure 4.12, the heating element is sandwiched between the insulation layer and the substrate. For this configuration, it is important to determine what fraction of the applied heat flux will enter the liquid helium and thus generate the thermal counterflow. For the steady state case, we can get the following equations to describe the heat balance:

q1 =

k1 (T1 T2 ) = hk (T2 T0 ) x1

(4.40)

q2 =

k2 (T1 T3 ) = hk (T3 T0 ) x 2

(4.41)

q1 + q2 = q

(4.42)

where q is the total applied heat flux from the heating part, q1 the heat flux transported through the insulation layer, q2 the heat flux transported through the substrate, and hk is the Kapitza conductance, which is considered the same for the upper and lower surfaces and for the purpose of calculation given by
hk = 0.9T03

(kW/m2K)
P P

(4.43)

86

Channel wall Helium inside the channel T0


B

G-10 substrate Helium bath T0


B

x2

q2

Heating element

T1
B

T3
B

Figure 4.12 Heat transfer model of the heater in steady state.

Assume the bath temperature is T0 = 1.80 K, the Kapitza conductance is estimated at


hk = 5250 W/m K. Since the thermal conductivities of Kapton and G-10 are strongly
P P

temperature-dependent, the effective thermal conductivities k1 = T k1 dT and k 2 = T k 2 dT


2 3

T1

x1
B

Kapton insulation

T2
B

q1
B

T1

in the corresponding temperature range are used in equations (4.40) and (4.41). After simple manipulations, the ratio of q1 / q2 can be calculated as,
hk x 2 k2 q1 = hk x1 q2 1+ k1 1+

(4.44)

87

Table 4.3 Thermal Conductivities of Mylar and G-10 at Different Temperatures

(Unit: W / m K , Data source: CRYOCOMP from Cryodata, Inc.) 2.0 K Mylar G-10 3.0 K 4.0 K 5.0 K 6.0 K 7.0 K 8.0 K 9.0 K 10.0 K 0.105

0.0026 0.0048 0.0074 0.0105 0.014 0.0225 0.0406 0.057

0.0177 0.0217 0.0258 0.030

0.0702 0.0808 0.0892 0.0959 0.101

The data for thermal conductivities of Kapton are unavailable at low temperatures, so the thermal conductivities of a similar material, Mylar, were used for the purpose of estimation. Table 4.3 shows the thermal conductivities of Mylar and G-10 at different temperatures, from which k1 and k 2 can be estimated as 0.0219 and 0.0883 W/mK respectively. Substituting x1 = 0.02 mm and x 2 = 12 mm into equation (4.44), one can get that
q1 123 . This indicates that under steady state condition over 99 percent q2

of the applied heat flux is transported through the insulation layer to generate the thermal counterflow. The same heater is also used to generate thermal pulse for transient experiments. In this case, the thermal response time of the heater is an important issue, and must be considered. As shown in Figure 4.13, the thermal response time is defined as how long it takes the applied heat flux q to be transported through the insulation layer. This can be simply modeled by a one-dimensional heat diffusion equation as,
T T k = c p x x t

(4.45)

88

where and c p are the density and specific heat of the Kapton, and k is the thermal conductivity, which is significantly temperature-dependent. The boundary and initial conditions can be described as, Initial condition:
T ( x, t = 0) = T0

(4.46)

Left boundary condition:

T x

=q
x =0

(4.47)

Right boundary condition:

T x

= hk [T ( x = L, t ) T0 ]
x=L

(4.48)

L
B

He II

x
Figure 4.13 Transient heat transfer model of the heater.

A numerical method is used to solve this heat diffusion problem. When the bath temperature T0 is 1.80 K and the applied heat flux q is 11.2 kW/m2, the thermal response
P P

time is calculated to be 350 microseconds. In another case, when the bath temperature is

89

1.60 K and the applied heat flux is 44.8 kW/m2, the calculated thermal response time is
P P

about 600 microseconds. Since the duration of the thermal pulse used in one of our second sound experiments is one millisecond long, the slow response of the heater must be taken into account when analyzing the results.

90

CHAPTER FIVE RESULTS AND DISCUSSIONS

5.1 Steady State Measurements of Thermal Counterflow


5.1.1 Data Acquisition and Processing

Our first PIV experiment consisted of measuring thermal counterflow in steady state. In this experiment, the polymer particles from Bangs Laboratories were used as tracer particles. When the particles are introduced into the counterflow channel, the heater is turned on. After a time delay of around 30 seconds for low applied heat flux and 5 seconds for high applied heat flux, the data acquisition starts. This time delay is much larger than the characteristic development time of superfluid turbulence v given by equation (2.36). So, it can be assumed that the time delay is long enough for the thermal counterflow to fully develop and reach the steady state. During the process of data acquisition, the CCD camera and the laser are operated in internal mode as shown in Figure 4.7, and their frequencies are both set at 5 Hz. The double exposure mode is used for image acquisition, which means a pair of images is needed for performing the cross-correlation process to attain the velocity field. The time separation between the two images, i.e. the pulse separation between the two laser heads (see Figure 4.7), is a critical parameter for the PIV experiment, and should be set based on the flow velocity to be measured. For the steady state thermal counterflow experiment, the normal fluid velocity induced by an applied heat flux q can derived from equation (2.24) as,

91

av vn =

q sT

(5.1)

where v nav represents the averaged velocity over the cross section of the flow channel. For a moderate applied flux of around 6 kW/m2, when the bath temperature T is 1.80 K, the
P P

normal fluid velocity is around 42.8 mm/s. In the image plane, the particle image will move at a velocity of 42.8 M = 19.5 mm/s. To attain the requirement that the displacement of particle image within the pulse separation should be larger than the particle image size (about 10 m for the polymer particles), a time separation larger than 0.5 ms is needed. On the other hand, it is also suggested that the displacement of particle image should not exceed 1/4 of the interrogation window size, which in our case is about 320 m. This requires that the time separation should not be larger than 4 ms. If we also consider that the normal fluid velocities to be measured has a dynamic range from about 10 to over 100 mm/s, a pulse separation of 1 ms is used when the velocity is higher than 50 mm/s and 2 ms is used when the velocity lower than 50 mm/s. Figure 5.1 shows a typical image acquired by the CCD camera. During each data acquisition process, which means a set bath temperature and applied heat flux, a total of 25 pairs of such images are collected within a time frame of about 5 seconds. For each pair of image, a cross-correlation is applied to determine the displacement of particles. In this process, a 48 48 pixels interrogation window is used. The result of velocity field is presented using Tecplot in the form of vector field.
P P

Figure 5.2 shows an example of the velocity field from analyzing only one pair of images. The bath temperature is 1.68 K and the applied heat flux is 2.26 kW/m2. The
P P

contour bar shows the magnitude of the measured particle velocity, which in this case is represented by the vertical velocity. It can be seen that the flow field is not uniform and the velocity has a dynamic range from 4 to 17 mm/s. Since only one pair of images is

92

used, it is believed that this range of variation may result from inhomogeneous particle seeding and some correlation errors. To reduce these effects, the 25 pairs of images are analyzed, and the results are combined to produce an averaged particle velocity field as shown in Figure 5.3. Compared with the result from only one pair of images, the averaged velocity field shows a more uniform velocity distribution. The contour bar indicates that the particle velocity has a narrower range from about 6 to 12 mm/s. In most of the flow field, the particle velocity displays a mean value of about 9 mm/s. A total of 400 (20 20) velocity vectors have been obtained with a spatial resolution of around 1 mm. Thus we are able to study many details of the velocity field. It can be seen that the direction of the particle velocity is almost uniformly upward with a horizontal velocity component near zero. This confirms our assertion that the particles will track the movement of normal fluid in thermal counterflow. But, since the magnitude of the particle velocity has a variation from 6 to 12 mm/s, further processing of the data is needed for quantitative analysis of the results.

Figure 5.1 An example of acquired images (two bright lines represent the channel wall).

93

Figure 5.2 Particle velocity field from the analysis of one pair of images.

Figure 5.3 Averaged particle velocity field from up to 25 pairs of images.

94

For the 400 velocity vectors obtained from Figure 5.3, the magnitudes of vertical velocity component can be extracted and the variation is shown in Figure 5.4. By specifying a cut-off range so that more than 90% of the total data points lie within the range, it can be seen that the particle velocity in the whole counterflow field has a variation from 6.5 to about 11.5 mm/s and the averaged velocity is about 9 mm/s. Using this method, the measured particle velocity can be determined as 9 2.5 mm/s. It should be noted that here the 2.5 represents the spatial variation of the measured particle velocity rather than the error of the PIV measurement. This kind of data processing is reasonable because the measured particle velocity is to be compared with the theoretical value of normal fluid velocity, v nav , which is also spatially averaged.

20 18 Measured particle velocity 16 14

Velocity (mm/s)

12 10 8

V = 9 mm/s
6 4 2 0 -10

-8

-6

-4

-2

10

Relative position
Figure 5.4 Variation of measured particle velocities in the counterflow field.

95

The error of PIV results depend on two parameters, the time separation between two images, and the displacement of particles within that time. Since the time separation is set electronically by the software, the error in it can be regarded as zero. As discussed before, the error of particle displacement depends on the accuracy of applied peak finding scheme. The PIV analysis software we used can find the position of correlation peak with an accuracy of 0.05 pixel. Thus, the relative error of measured particle velocity, v , can be written as,
0.05a P 1 t p M v p

v =

(5.2)

where a p is the pixel size, t p the pulse separation set during the experiment, M the magnification number, and v p the particle velocity. For the PIV system used in this experiment, a p = 6.7 m and M = 0.455 . When the pulse separation is set as 1 ms, the relative error is about 7% when the particle velocity is 10 mm/s, and only about 1% when the particle velocity is up to 70 mm/s. Compared to spatial variation discussed above, this error is relatively small. Therefore, for all the PIV results presented thereafter, only the effect of spatial variation is considered.

5.1.2 Results and Discussions

Figure 5.5 shows the particle velocity fields induced by different applied heat flux. The bath temperature is 1.68 K. In Figure 5.5 (a), when the applied heat flux is very small, the particle velocity displays a little circular motion at the center of the flow field, which may because the particle velocity is so small and the displacement of particle

96

image is less than the particle image size. At higher heat fluxes, as shown in (b d), the movement of particles is approximately parallel with no considerable circulation shown. Thus, it can be inferred that the normal fluid velocity in thermal counterflow is also approximately parallel and uniform across the cross-section of the flow channel. There is no evidence showing that this velocity profile will change when the applied heat flux is increased, but the amplitude of particle velocity does show an increase with the increasing of applied heat flux, from about 10 mm/s when q = 2.26 kW/m2 up to about 70
P P

mm/s when q = 13.7 kW/m2.


P P

Using the data processing method discussed above, the spatially averaged particle velocity is plotted versus applied heat flux and compared with the theoretical value of v nav given by equation (5.1). Figure 5.6 shows the result at bath temperature of 1.68 K. One can see that the particle velocity is approximately a linear function of the applied heat flux. But the comparison between the measured particle velocity and the theoretical normal fluid velocity shows a substantial discrepancy between them. As indicated before, this discrepancy may result from the slip velocity between the particles and the normal fluid. To clarify this problem, we have measured the slip velocity by means of measuring the settling velocity of the tracer particles in stationary flow field, i.e. no applied heat flux and no generated thermal counterflow. This result is shown in Figure 5.7. It can be seen that the measured slip velocity is around 6 mm/s, which is much larger than the calculated value of about 1 mm/s. This difference is quite reasonable if considering that the particles dispersed in the flow field are normally several times larger than the nominal particle size for very fine particles with size around or less than one micron. But, adding the measured slip velocity to the particle velocity, the result shown in Figure 5.6 still shows a significant discrepancy with the theoretical normal fluid velocity.

97

(a) q = 2.26 kW/m2


P P

(b) q = 4.16 kW/m2


P P

(c) q = 8.10 kW/m2


P P

(d) q = 13.7 kW/m2


P P

Figure 5.5 Change of particle velocity field with applied heat fluxes ( the bath temperature is 1.68 K).

98

Figure 5.6 Comparison between particle velocity and normal fluid velocity at 1.68 K.

Figure 5.7 Measured settling velocity of polymer tracer particles in He II.

99

To explore whether this discrepancy is temperature dependent, more measurements at different bath temperatures have been conducted, and the results are listed in Table 5.1-5.4. Based on these results, the relation between the particle velocity and the theoretical normal fluid velocity is shown in Figure 5.8. It can be seen that the discrepancy is present at all the temperatures. A similar behavior has also been reported by Nakano and Murakami, who observed that the normal fluid velocity measured by Laser Doppler Velocimetry (LDV) is much smaller than the theoretical value [53]. They further suggested that the discrepancy could be explained by the following relation,
n

v n ,m v n ,theory

(5.3)

which means the ratio of measured normal fluid velocity to the theoretical one is equal to the ratio of normal fluid density to the bulk density of He II. Since the normal fluid density is strongly temperature dependent (see equation (2.3)), this relation results in a significant temperature dependence of
v n ,m v n ,theory

. However, Figure 5.8 indicates that in our

experiments the ratio is most likely temperature independent. Further analysis shows that
v n ,m v n ,theory

has a variation from 0.4 to 0.6 and can be roughly estimated as a constant at

around 0.5. The reason that equation (5.3) is not applied to our case is probably because the relation is based on the counterflow jet measurement in which the normal fluid and superfluid components move in the same direction while in this experiment the two components move in opposite direction.

100

Figure 5.8 Comparisons between the measured particle velocities and theoretical normal fluid velocities at different temperatures ( The experimental data are based on the results listed in Table 5.1 to Table 5.4, v n ,t represents the theoretical normal fluid velocities given by equation (5.1), v p ,a is the adjusted particle velocity by adding the 6 mm/s slip velocity, the solid line represents where v p ,a = v n ,t ).

101

Table 5.1 Comparisons between the Measured Particle Velocity and the Theoretical Normal Fluid Velocity at Different Temperatures ( T = 1.62 K and T = 1.65 K)

T = 1.62 K
q (kW/m )
P P

T = 1.65 K
vp,a
B B

vn,t
B

vp,m
B B

v (mm/s)
4.21 7.61 9.42 13.22 15.92 18.33 23.82 28.82 31.43 41.03 45.44 49.83 53.04 57.04 71.83 71.76 73.95 81.34 -

q (kW/m )
P P

vn,t
B B

vp,m
B B

vp,a
B B

v (mm/s)
6.81 10.01 11.22 14.92 17.13 19.63 22.23 27.33 34.04 35.63 38.15 47.64 55.84 62.63 58.24 67.23 79.94 82.84 84.94

(mm/s)
16.2 20.6 25.4 30.7 36.4 42.8 49.8 56.8 64.4 73.0 81.4 91.3 101.0 111.0 121.8 133.7 144.9 157.3 -

(mm/s)
61 71 102 11.52 14.52 18.53 202 222 273 263 304 35.53 424 484 443 566 655 704 -

(mm/s)
121 131 162 17.52 20.52 24.53 262 282 333 323 364 41.53 484 544 503 626 715 764 -

(mm/s)
17.8 22.5 27.2 32.4 38.1 44.1 50.2 56.8 65.0 73.1 80.1 90.1 97.8 107.6 117.2 128.2 138.9 150.8 162.9

(mm/s)
51 6.51 102 11.52 153 18.53 223 23.53 254 31.53 365 36.54 364 393 534 553 534 624 724

(mm/s)
111 12.51 162 17.52 213 24.53 283 29.53 314 37.53 425 42.54 424 453 594 613 594 684 784

1.16 1.48 1.82 2.20 2.61 3.07 3.57 4.07 4.62 5.24 5.84 6.54 7.24 7.95 8.73 9.59 10.38 11.3 -

1.44 1.82 2.20 2.62 3.08 3.57 4.06 4.60 5.26 5.91 6.48 7.29 7.91 8.71 9.48 10.37 11.24 12.20 13.2

Note: q is the applied heat flux, v n ,t the theoretical value of normal fluid velocity, v p ,m the measured particle velocity, v p ,a the particle velocity plus the slip velocity, and v = v n ,t v p , a .

102

Table 5.2 Comparisons between the Measured Particle Velocity and the Theoretical Normal Fluid Velocity at Different Temperatures ( T = 1.68 K and T = 1.70 K)

T = 1.68 K
q (kW/m )
P P

T = 1.70 K
vp,a
B B

vn,t
B

vp,m
B B

v (mm/s)
4.52 6.02 8.82 10.32 12.52 18.22 21.02 21.71 24.22 24.93 31.53 37.43 36.23 45.03 46.23 54.43 64.93 67.03 73.32 75.13

q (kW/m )
P P

vn,t
B B

vp,m
B B

vp,a
B B

v (mm/s)
2.71 4.22 6.31 9.41 12.32 13.92 16.82 21.11 25.52 28.43 33.22 41.12 43.14 47.04 49.64 50.84 59.74 64.66 71.24 -

(mm/s)
13.0 16.5 20.3 24.8 29.0 34.2 39.5 45.7 52.2 58.4 66.5 73.4 82.2 89.0 99.2 107.4 116.9 127.0 137.3 150.1

(mm/s)
2.52 4.52 5.52 8.52 10.52 102 12.52 181 222 27.53 293 303 403 383 473 473 463 543 582 693

(mm/s)
8.52 10.52 11.52 14.52 16.52 162 18.52 241 282 33.53 353 363 463 443 533 533 523 603 642 753

(mm/s)
11.7 14.7 18.3 22.4 26.3 30.9 35.8 41.1 47.0 52.9 59.2 66.1 73.1 81.0 88.6 96.8 105.2 114.6 124.2 -

(mm/s)
31 4.52 61 71 82 112 132 141 15.52 18.53 202 192 244 284 334 404 39.54 446 474 -

(mm/s)
91 10.52 121 131 142 172 192 201 21.52 24.53 262 252 304 344 394 464 45.54 506 534 -

1.18 1.50 1.85 2.26 2.64 3.11 3.60 4.16 4.75 5.31 6.05 6.68 7.49 8.10 9.03 9.78 10.64 11.56 12.50 13.66

1.15 1.44 1.80 2.20 2.59 3.04 3.52 4.04 4.62 5.20 5.82 6.50 7.19 7.97 8.72 9.52 10.35 11.27 12.21 -

Note: q is the applied heat flux, v n ,t the theoretical value of normal fluid velocity, v p ,m the measured particle velocity, v p ,a the particle velocity plus the slip velocity, and v = v n ,t v p , a .

103

Table 5.3 Comparisons between the Measured Particle Velocity and the Theoretical Normal Fluid Velocity at Different Temperatures ( T = 1.77 K and T = 1.80 K)

T = 1.77 K
q (kW/m )
P P

T = 1.80 K
vp,a
B B

vn,t
B

vp,m
B B

v (mm/s)
5.3 7.7 13.0 15.6 17.5 18.2 19.7 26.1 27.9 32.5 38.2 40.4 -

q (kW/m )
P P

vn,t
B B

vp,m
B B

vp,a
B B

v (mm/s)
2.81 3.82 5.01 6.91 8.02 7.53 10.13 15.63 16.72 19.03 21.95 22.93 20.83 26.14 28.23 35.24 36.23 39.14 44.74 -

(mm/s)
17.3 20.7 24.1 28.0 32.1 36.5 41.2 46.7 51.5 57.1 62.9 69.1 75.5 82.2 89.2 96.4 -

(mm/s)
6 7 9 10.5 13 17 21 25 29 37 45 50 -

(mm/s)
12 13 15 16.5 19 23 27 31 35 43 51 56 -

(mm/s)
10.3 12.8 15.5 18.4 21.5 25.0 28.6 32.6 36.7 41.0 45.9 50.9 55.8 61.6 67.2 73.2 79.2 85.6 92.7 -

(mm/s)
1.51 32 4.51 5.51 7.52 11.53 12.53 113 142 163 185 223 293 29.54 333 324 373 40.54 424 -

(mm/s)
7.51 92 10.51 11.51 13.52 17.53 18.53 173 202 223 245 283 353 35.54 393 384 433 46.54 484 -

2.20 2.64 3.08 3.57 4.10 4.66 5.26 5.97 6.58 7.29 8.03 8.82 9.64 10.49 11.38 12.31 -

1.47 1.82 2.20 2.62 3.06 3.56 4.07 4.64 5.23 5.83 6.53 7.24 7.94 8.77 9.56 10.42 11.27 12.18 13.19 -

Note: q is the applied heat flux, v n ,t the theoretical value of normal fluid velocity, v p ,m the measured particle velocity, v p ,a the particle velocity plus the slip velocity, and v = v n ,t v p , a .

104

Table 5.4 Comparisons between the Measured Particle Velocity and the Theoretical Normal Fluid Velocity at Different Temperatures ( T = 1.90 K and T = 2.0 K)

T = 1.90 K
q (kW/m )
P P

T = 2.0 K
vp,a
B B

vn,t
B

vp,m
B B

v (mm/s)
1.32 1.02 6.22 7.12 7.24 7.64 12.12 14.03 15.93 17.64 19.93 20.64 22.44 23.64 25.13 26.64 -

q (kW/m )
P P

vn,t
B B

vp,m
B B

vp,a
B B

v (mm/s)
-2.02 0.62 2.02 2.61 3.02 6.82 7.42 6.14 12.52 9.82 14.32 20.92 17.82 -

(mm/s)
7.3 9.0 10.9 15.2 17.6 20.2 23.1 29.1 32.5 35.9 39.6 43.4 47.6 51.9 56.1 60.6 65.6 -

(mm/s)
02 22 32 4.52 74 9.54 112 12.53 143 164 17.53 214 23.54 26.54 29.53 334 -

(mm/s)
62 82 92 10.52 134 15.54 172 18.53 203 224 23.53 274 29.54 32.54 35.53 394 -

(mm/s)
11.0 12.6 14.5 16.6 21.0 23.3 25.9 28.6 31.5 34.3 37.3 40.4 43.8 -

(mm/s)
72 62 6.52 81 122 10.52 12.52 16.54 132 18.52 172 13.52 202 -

(mm/s)
132 122 12.52 141 182 16.52 18.52 22.54 192 24.52 232 19.52 262 -

1.47 1.81 2.19 3.05 3.54 4.05 4.64 5.85 6.53 7.22 7.97 8.73 9.56 10.42 11.28 12.18 13.19 -

3.07 3.52 4.06 4.62 5.86 6.51 7.22 7.99 8.78 9.56 10.42 11.28 12.23 -

Note: q is the applied heat flux, v n ,t the theoretical value of normal fluid velocity, v p ,m the measured particle velocity, v p ,a the particle velocity plus the slip velocity, and v = v n ,t v p , a .

105

In an effort to find out what exactly causes the discrepancy between particle velocity and theoretical normal fluid velocity, analysis has been done from the following three different points of view.

1. Slip velocity between the particles and normal fluid.


U

Though it is shown in Figure 5.5 that adding slip velocity to the measured particle velocity does not reduce the discrepancy between the measured particle velocity and theoretical normal fluid velocity, this issue should be further addressed. Slip velocity may result in the discrepancy is because in our experiment the normal fluid velocity is upward and in opposite direction with the slip velocity. Thus, a simple way to clarify the effect of slip velocity is to turn the counterflow channel shown in Figure 4.13 upside down. In this case, the normal fluid components moves downward, and slip velocity would make the measured particle velocity higher than the normal fluid velocity if the discrepancy is due to slip velocity. However, the results shown in Figure 5.9 indicate that the particle velocity measured when the channel is upside down is only a little higher than that measured when the channel is upward, but is still much smaller than the theoretical normal fluid velocity. So it may be concluded that slip velocity is not the primary reason for the discrepancy between the particle and the normal fluid velocities.

2. The effect of turbulent flow.


U

According to equations (2.21) and (2.23), the critical normal fluid velocity v n ,c and superfluid velocity v s ,c in our experiment are about 0.67 and 7.8 mm/s respectively. For most of our measurements listed in Table 5.1 to 5.4, the normal fluid velocity and superfluid velocity are greater than these critical values, indicating that both the normal fluid and the superfluid components are in the turbulent state. When tracer particles are used to track the turbulent flow, equations (4.33) and (4.35) show that, in addition to the

106

Figure 5.9 Measured particle velocities for different channel configurations. Top: 1.80 K, bottom: 1.62 K. : channel upside down, : channel upward, solid line represents the theoretical normal fluid velocity.

107

slip velocity, there exists an amplitude error which may cause a velocity difference between the particles and the fluid. Assuming that the velocity discrepancy shown in Figure 5.8 is due to the amplitude error, one can get
1 1 + St 2

= 1

0.5

(5.4)

which gives a Stokes number of St = 3 . To verify this assumption, we have used the hollow glass spheres from TSI, Inc. as the tracer particles to do the same experiment. From equation (4.31), a conservative estimation of the Stokes number with these particles is about 4 3 (we assume that the frequency of turbulence does not change, and the polymer particle size is about 4 m because of coagulation). Substituting this number back to equation (4.35), the amplitude error for the hollow glass spheres particles is about 0.87, which means the ratio
v p ,a v n ,t

would be about 0.13, a much larger velocity

discrepancy than that for polymer particles. But Figure 5.10 shows that the measured particle velocity of the TSI particles is only a little smaller than that of the polymer particles, which is due to the different slip velocities they have. Since the velocity discrepancy does not become much higher when larger particles are used, the turbulent fluctuation can not be the cause of the observed discrepancy.

108

Figure 5.10 Measured particle velocities with different tracer particles. Top: 1.80 K, bottom: 1.62 K. : TSI particles, : polymer particles, Solid line represents the theoretical normal fluid velocity.

109

3. Additional force from the superfluid component.


U

In two-fluid model of He II, the superfluid component is inviscid. Therefore, we have assumed that the tracer particles only interact with the normal fluid component, and for thermal counterflow the particles only track the motion of normal fluid. This issue becomes more complicated when thermal counterflow transfers He II into the turbulent state. It has been argued that, in the presence of quantized vortex lines in the superfluid component, the vortex lines may trap the particles and affect their motion [14]. In addition, it has also been reported that, in pure superfluid flow where the normal fluid velocity is zero, the tracer particles actually moved in the same direction with the superfluid component and thus track the superfluid velocity [66]. These studies strongly suggest that the tracer particles may also interact with the superfluid component resulting in an additional force on the tracer particles. For thermal counterflow, the additional force which is in the opposite direction to the viscous force from the normal fluid component would slow down the particle velocity, and thus should be taken into account when applying the particle dynamics equation, i.e. equation (4.3). There is no exact form of the additional force because the physical mechanism associated with it is still not clear, but an immature study suggests that it is mostly likely induced by the interaction between the particles and the quantized vortex line [67]. According to this assertion, two assumptions can be made about the force: 1) it depends on the superfluid density, and 2) it is proportional to the probability that the particles interact with the vortex lines (this probability should depend on the particle size and the vortex line density). Combining these assumptions with the experiment results shown in Table 5.1 to 5.4, a semi-empirical correlation given by equation (5.5) below has been developed,
Fs = s v s v n d p

(5.5)

110

where Fs represents the additional force, is a constant that can be determined by experiment, s the superfluid density and d p the particle diameter. Adding this force, equation (4.3) becomes,
3 d P P dv P

dt

2 d p C D f (v p v f

3 d P g ( f P ) Fs +

3 dp f

dv f dt

12

3 dp f

d (v f v p ) dt

(5.6)

Considering a simple case that the steady state is reached, i.e.

dv p dt

dv f dt

= 0 , and

= 1 , the velocity difference between the particles and the normal fluid component can

be written as,
v f v p = v slip + v add

(5.7a)

v slip =

2 dpg

18 n

(5.7b)

v add =

3 n d p

Fs

(5.7c)

We can see that the total velocity difference (v f v p ) consists of two parts, one is the same as the slip velocity given by equation (4.15), and the other is an additional velocity difference due to the force from the superfluid component. For thermal counterflow, v add is attained by substituting equations (2.26) and (5.5) into equation (5.7c),

111

v add =

1 q 3 n sT

(5.8)

Thus, the additional velocity difference at a certain bath temperature should be a linear function of the applied heat flux. This relation is in consistent with the experiment results shown in Figure 5.11.

Figure 5.11 Relation between the additional velocity difference and applied heat flux ( v add is the same as v given by Table 5.1 to 5.4).

112

T In addition, the entropy s in equation (5.8) can be given by s = s T

, where

5.6

s is the entropy at the -temperature and a constant. So the linear coefficient of v add / q

should have a strong temperature dependence given below,


1 T 6 .6 3 n sT

(5.9)

To compare this relation with the experiment result, the slopes of each fitting line in Figure 5.11 have been calculated and plotted versus bath temperatures. As shown in Figure 5.12, the experiment data displays an approximate temperature dependence of
T 6.7 , which is in reasonable agreement with that predicated by equation (5.9). This

shows further evidence that equation (5.5) is the correct form for the superfluid force.

Figure 5.12 Temperature dependence of linear coefficient v add / q .

113

Furthermore, equation (5.8) predicts no particle size dependence of the additional velocity difference. Since the normal fluid and superfluid components are still in opposite direction when the channel is upside down, v add should also not depend on the configuration of the channel. Both of these predications are consistent with the experimental results shown in Figure 5.9 and 5.10. The above discussion strongly suggests that the velocity discrepancy between the particles and the normal fluid is caused by an additional force resulting from the interaction between particles and the superfluid component. But, it should be emphasized that the existence of this force is still a hypothesis. Also, equation (5.5) is semi-empirical, and its validity needs more experimental data to be verified.

5.2 PIV Measurements of Second Sound and Heat Diffusion


5.2.1 Principle of Measurement

Second sound propagation and thermal shock have been studied extensively [30, 32-34, 68-75], but all the previous research has focus on measuring the temperature profile induced by heat pulse and the associated second sound attenuation (see Figure 2.11). As discussed before, a unique characteristic of second sound is that it can induce thermal counterflow. This allows us to study the second sound shock by measuring the thermal counterflow induced by it. Shown in Figure 5.13 is the counterflow channel the same as we used in steady state measurement. When a heat pulse is applied from the heater, second sound shock is generated and propagates upward along the channel. Simultaneously, an accompanying thermal counterflow is induced by the second sound, with a normal fluid velocity given

114

by equation (5.1). At time t 0 , the second sound shock is just generated and is very close to the heater. Since the bottom line of the view field is about 155 mm above the heater surface, the normal fluid velocity in the view field is zero. When t t a , where
t a = H / a 2 represents the time it takes the second sound shock front to arrive at the

view field, the normal fluid in the view field will accelerate to a finite velocity. Later, after the second sound shock tail passes the view field, the induced counterflow will recover to the initial quiescent state, and the normal fluid velocity in the view field should return to zero. Thus, the normal fluid velocity profile can be used to measure the propagation of second sound shock, as is normally done with the temperature profiles. When using this method, one should keep in mind that the induced normal fluid velocity, which depends on the applied heat flux, is different from the second sound shock velocity, which can be easily estimated as a 2 = H / t a .

View field

Second sound shock

Shock velocity
H

Normal fluid velocity

Heat pulse t=0 t = ta


B

Figure 5.13 Measurement of second sound using the PIV technique.

115

The transient thermal counterflow induced by a second sound shock can be measured using the PIV technique. However, one limitation of PIV for transient measurements is its low temporal resolution due to the technical restrictions of the frame rate of the CCD camera, as well as the repetition rate of the laser. For the current PIV system described before, the Nd:YAG pulse laser only has a maximum repetition rate of 15 Hz, and the CCD camera must be synchronized with the laser. So the minimum time interval between two consecutive pairs of images is 66.7 milliseconds, although the time separation between the first and the second image of each pair can be set as low as 1 microsecond because the laser has dual head. Considering that the second sound velocity is around 20 m/s, and that the view field is about 20 mm long, it takes only about one millisecond for the second sound shock to traverse the view field. Thus, we are unable to measure the propagation of second sound shock when the laser and camera are operated in repetitive mode, i.e. internal mode as shown in Figure 4.7. To overcome this limitation, the synchronous control mode described in Figure 4.8 and 4.9 was used. For each set trigger time delay tT, the laser was fired once (two
B B

laser pulses are generated with a set pulse separation), and only one pair of images was acquired. Thus, a series of runs with different tT is needed to complete the measurement
B B

for one set of generating parameters, i.e. bath temperature, amplitude and width of the heat pulse. The feasibility of this method is based on the assumption that when the generating parameters are fixed, the temperature versus time profiles of the second sound shock wave are indistinguishable from one run to the next [33]. To avoid any buildup of superfluid turbulence during the experiment, one should wait at least three minutes between each run. This waiting time is required for the quantized vortices induced from previous heat pulse to fully dissipate. Since tT can be set with an increment of 50
B B

microseconds, this method allows us to measure the propagation of second sound with a temporal resolution of 50 microseconds.

116

Heat diffusion in the counterflow channel can also be measured by measuring the temporal evolution of the normal fluid velocity profile. From equation (2.24) and (2.34), the normal fluid velocity induced by the heat diffusion can be written as,
1/ 3

1 Tz v = Ts f (T )
av n

(5.10)

indicating that the normal fluid velocity will change with the local temperature gradient
Tz . As shown in Figure 5.14, at the beginning of the heat diffusion, the temperature

profile along the channel is like curve A, so the normal fluid velocity at the view field is almost zero. With the evolution of time, the normal fluid velocity increases due to the increasing of local temperature gradient. This transient process develops until the steady state (curve B) is reached. In this case, the temperature gradient along the channel is approximately constant, and the normal fluid velocity will also reach a steady state value.

Figure 5.14 Measurement of heat diffusion using the PIV technique.

117

5.2.2 Results and Discussions

Figure 5.15 shows the PIV flow field measurements of the second sound induced transient thermal counterflow. The applied pulse heat flux is 44.8 kW/m2 and the pulse
P P

width is 1 millisecond. At time t = 3.2 ms after the heater was on (shown in (a)), the entire flow field shows a small downward particle velocity (< 10 mm/s) which is due to the settling of the polymer particles which have higher density than that of He II. At time t = 6.2 ms, (b) shows a similar motion of particles as in (a). This indicates that the second sound shock front has yet to arrive at the view field. Shown in (c) is the particle velocity measured at t = 7.4 ms. At the bottom of the view field, there are upward particle velocities distributed at around 35 mm/s, which clearly indicates the arrival of second sound shock front. Since the shock front has yet to arrive at the top of the view field, there is still no apparent upward particle velocity in that region. At t = 8.2 ms (shown in (d)), the particle velocities at the bottom have increased up to 70 mm/s, and the shock front arrives at the top of the view field, resulting in upward particle velocities distributed around 30 mm/s. As time goes on, the shock continues to move up the counterflow channel. At t = 9.1 ms, the particle velocity at the top of the view field increases to about 70 mm/s, while at the bottom the particle velocities have decreased from its maximum of 70 mm/s at 8.2 ms to about 40 mm/s. This indicates that the bottom of the view field is in the expansion fan region of the second sound shock. In that region, the heat flux q s that can be carried by a second sound shock is given by
q s = c p a 2 T

(5.11)

118

where T is the temperature rise due to second sound shock. Since T gradually decreases in the expansion fan region (see Figure 2.10), q s also decreases. According to equation (5.1), this causes the normal fluid velocity to decrease, as observed by the particle velocity. At t = 9.8 ms as shown in (f), the expansion fan arrives at the top of the view fields, resulting in a decrease of particle velocities from 70 mm/s at 9.1 ms to about 30 mm/s. At the bottom, there is no apparent upward particle velocity, showing that the second sound shock tail has passed this region, and the induced counterflow starts to dissipate. Later, at t = 11.1 ms (g), the entire flow field displays a random motion, indicating that the tail of second sound shock has passed the top of the view field. As shown in (h), further observation of the flow field after 11.1 ms indicates that the random motion of particles will last for a quite long time on the order of hundreds of milliseconds. Since the polymer particles in stationary He II should display a downward velocity due to their settling as shown in Figure (5.7), this observation makes us believe that the random motion may be associated with the residual superfluid turbulence induced by second sound shock. During the dissipation process of the residual turbulence, which may take hundreds of milliseconds, the additional force from superfluid component that we discussed before will prevent the tracer particles from settling. From the measured particle velocity profiles in (c-f), it can be inferred that the second sound induced normal fluid velocity is approximately parallel and uniform across the cross-section of the channel. This result is in support of Turners assertion that the shock front is planar except for minor distortions near the channel walls, and thus the velocity field induced is parallel and uniform across the entire channel [33].

119

(a) t = 3.2 ms

(b) t = 6.2 ms

(c) t = 7.4 ms

(d) t = 8.2 ms

120

(e) t = 9.1 ms

(f) t = 9.8 ms

(g) t = 11.1 ms

(h) t = 100 ms

Figure 5.15 Transient counterflow field induced by second sound shock (bath temperature: 1.61 K, applied heat flux: 44.8 kW/m2, width of heat pulse tp = 1 ms).
P P B B

121

Shown in Figure 5.16 below is the transient counterflow field induced by a heat pulse with applied heat flux of 11.2 kW/m2 and width of 999 ms. For this long pulse, the
P P

flow field at t = 2.2 ms and 6.2 ms is similar to that for 1 ms long pulse. Some circulation in the particle velocity is due to the injection. At these points, the second sound shock has not arrived at the view field, and there is no induced counterflow. At t = 7.4 ms, the second sound shock front arrives at the bottom of the view field, resulting in an apparent upward particle velocity at about 20 mm/s, while at the top there is no apparent upward particle velocity. At t = 8.2 ms, the shock front arrives at the top of the view field, and the particle velocity at the bottom increases from 20 mm/s to about 35 mm/s. As time goes on, the particle velocity continues increasing until a relatively steady velocity of about 43 mm/s is reached at about t = 9.2 ms. As shown in (e), the particle velocities in the entire view field are approximated the same, which is quite different from that induced by the 1 ms long pulse (shown in Figure 5.15 (e)). This is because the pulse is very long, and the expansion fan has not arrived at the view field. The steady particle velocity lasts for approximately 6 ms to t = 15.2 ms. As shown in (f), the particle velocities in the entire view field are around 45 mm/s. After that, the particle velocity starts to drop, indicating the expansion fan arrives at the view field. At t = 18.2 ms, the particle velocity decreases from 45 mm/s to about 30 mm/s. This process continues until t = 25.2 ms. As shown in (h), particle velocity displays a random motion due to the residual superfluid turbulence, indicating that the second sound shock tail has passed over the view field. Thus, the second sound shock generated by this 999 ms long heat pulse only lasts about (25.2 7.4) = 17.8 ms. As discussed in Chap. 2.5.1, this is because the critical energy that can be transported by the second sound has been reached. After the second sound pulse passes, the rest of the energy is transported in heat diffusion, which is clearly shown in (i) at t = 50 ms. This heat diffusion process lasts for a long time, causing the

122

particle velocity to increase slowly. For example, at t = 200 ms which is the longest time we can go in this experiment, the particle velocity has increased from 25 mm/s at t = 50 ms to about 40 mm/s.

(a) t = 2.2 ms

(b) t = 6.2 ms

(c) t = 7.4 ms

(d) t = 8.2 ms

123

(e) t = 9.2 ms

(f) t = 15.2 ms

(g) t = 18.2 ms

(h) t = 25.2 ms

124

(i) t = 50 ms

(j) t = 200 ms

Figure 5.16 Transient counterflow field induced by a long heat pulse (bath temperature: 1.61 K, applied heat flux: 11.2 kW/m2, width of heat plus tp: 999 ms).
P P B B

To make a more quantitative study of the propagation of second sound shock and heat diffusion, the profiles of particle velocity versus time are plotted based on the transient measurements discussed above. Figure 5.17 displays the measured particle velocity, which is the averaged particle velocity along the bottom line of the view field, for the two heat pulses, 11.2 kW/m2 for 999 ms and 44.8 kW/m2 for 1 ms. From the
P P P P

velocity profiles, the arrival of the second sound shock front is clearly registered by the increasing particle velocity. For both pulses, the arrival times are almost the same at around 7.3 milliseconds after the heater was turned on. Since the distance from the heater to the bottom of the view field H is around 155 mm (see Figure 5.13), this yields an estimated shock velocity of 21.2 m/s, or about 4.5% higher than the accepted value of

125

20.3 m/s at 1.61 K. Careful analysis shows that this inconsistency is primarily caused by the measurement technique we used. For PIV, the velocity is derived by measuring the displacement of particles within a certain time interval, thus the measured velocity is actually a temporally averaged velocity. In this experiment, the pulse separation is set at one millisecond. This means the measured velocity at, for example, 7.3 ms in Figure 5.17 is actually the temporally averaged velocity between 7.3 ms and 8.3 ms. So the actual time at which the particle velocity starts to arise, i.e. the arrival time of second sound shock, may be greater than 7.3 milliseconds. Certainly, reducing the pulse separation would produce a more accurate velocity versus time profile, but it is not applicable in this case due to the low velocity of induced counterflow field. This slow response behavior is one limitation of using PIV to measure the second sound. For heat diffusion measurement, this limitation may have little effect because the heat diffusion develops much more slowly than second sound and the velocity change within one millisecond is relatively small. According to Turner [33], the counterflow velocity field could be induced within one microsecond or less after the second sound shock front arrives, therefore the rising slope of induced normal fluid velocity should be extremely sharp. However, Figure 5.17 shows that the duration of particle velocity increase is on the order of milliseconds. This result may be due to three factors: First is the slow response of PIV because large pulse separation is used, which is just discussed above. Second is the response time of tracer particles. From Table 4.1, the polymer particles used in this experiment has a calculated response time of 0.13 ms. But, if considering that the actually slip velocity measured in He II is about 6 times larger than the calculated one due to particle coagulation, the actual response time of these particles can be estimated as about 0.8 ms. Third is the thermal response time of the heater, which is about 0.6 ms as discussed in Chap. 4.5. Because of these factors, we are not able to infer the shape and thickness of the leading edge of second sound induced normal fluid velocity.

126

For the 1 ms long pulse shown in Figure 5.17, the induced particle velocity, after reaching the maximum of about 70 mm/s around 8.3 ms, rapidly drops to almost zero at 11 ms. By contrast, for the 999 milliseconds long heat pulse, the particle velocity increases until a relatively steady velocity of about 43 mm/s is reached after about 9.2 ms. Compared to the theoretical normal fluid velocity, which is about 162.8 mm/s from equation (5.1), this steady velocity is much lower, which, as we know, may be due to the additional force from the superfluid component. With measured slip velocity of 6 mm/s taken into account, the ratio of measured particle velocity to the theoretical normal fluid velocity in this case is about 0.3, much less than the ratio of about 0.5 from the steady state measurement of thermal counterflow (see Figure 5.8). This result brings into question the applicability of equation (5.5) for the transient state. As we assumed, the additional force Fs should depend on the vortex line density in superfluid component. In equation (5.5), the vortex line density is accounted by the relation v n v s L1 / 2 derived from equation (2.15). Since equation (2.15) is applied only when the vortex line density reaches steady state, equation (5.5) many be not valid for transient counterflow induced by second sound. To overcome this limitation, a general form for Fs can be re-defined as,
d p s 1 / 2 L 1/ 2

Fs =

(5.12)

which is the same as equation (5.5) at steady state. Thus, the smaller ratio of particle velocity to theoretical normal fluid velocity can be attributed to that the superfluid turbulence, i.e. vortex line density, induced by second sound shock is much stronger than that generated in steady state thermal counterflow. In addition, there is one more possible reason for this behavior. From Figure 5.17, the acceleration of normal fluid may be very high in the second sound induced counterflow field. According to equation (4.26), this

127

acceleration will cause an additional velocity difference between the particles and the normal fluid.

80 70 q = 4.48 W/cm , dt = 1 ms 2 q = 1.12 W/cm , dt = 999 ms


2

Measured particle velocity (mm/s)

60 50 40 30 20 10 0 0 2 4 6 8 10 12 14 16

Time after the heater was on (ms)

Figure 5.17 Measured particle velocity profiles at 1.61 K.

128

Shown in Figure 5.18 below is the entire trace of particle velocity for the 999 ms pulse. It is apparent that the steady velocity only lasted for approximately 6 milliseconds, and after that started to drop because the critical energy flux is reached. From this result, the critical energy flux can be estimated to be about 67 J/m2 at 1.61 K (11.2 kW/m2 6
P P P P

ms), which is quite reasonable compared to 75 J/m2 at 1.7 K given in [76]. The heat
P P

diffusion, which is represented by the second particle velocity rise after the second sound has passed, is also shown in this figure. Clearly, the development of heat diffusion is very slow, and it may take hundreds of milliseconds to reach steady state. Compared to the profiles of temperature rise induced by heat pulse as shown in Figure 2.13, the profile of particle velocity shown in Figure 5.18 displays a similar characteristic. This is a strong support that PIV has significant potential for the study of second sound shock, counterflow, and quantized superfluid turbulence.

129

50

40

Measured particle velocity (mm/s)

30

20

10

20

40

60

80

100

120

140

160

180

200

Time after heater was on (ms)

Figure 5.18 Entire trace of particle velocity profile induced by the 999 milliseconds long heat pulse with applied heat flux of 11.2 kW/m2. The bath temperature is 1.61 K.
P P

130

CHAPTER SIX CONCLUSIONS

6.1 Application of PIV to Liquid Helium


The challenges for applying PIV to liquid helium, which are mainly due to its unique physical properties and experimental environment, have been analyzed. These challenges place stringent requirements on particle selection and seeding. To select the appropriate tracer particles for liquid helium, particle dynamics in a fluid is carefully studied, as well as the motion of particles in three different types of flows, steady, accelerating, and turbulent flows. Two critical parameters for selecting tracer particles, slip velocity and relaxation time, have been discussed in the context of different tracer particles. Among all the particles that we have tried to use for liquid helium, including both commercially available solid particles and particles generated by freezing liquid or gases, the polymer micro-spheres with mean diameter of 1.7 m and density of 1100 kg/m3 show a calculated slip velocity of 1.2 mm/s and relaxation time of 0.15 ms, both of
P P

which are small enough for tracking low speed flows, as well as transient and turbulent flow fields, in liquid helium. These particles have been used in our PIV experiments of measuring the thermal counterflow generated by a steady heat flux, and also the transient thermal counterflow induced by second sound shock and heat diffusion, both yielding a great success. Though solidified H2-D2 particles show potential to have good tracking
B B B B

fidelity and fast response due to their small size and density close to that of liquid helium, the generation of these particles requires special apparatus and extra care. The unique experimental environment of liquid helium makes many seeding technologies used for conventional fluids inapplicable. Thus, particle seeding is no longer

131

a trivial problem. In an effort to introduce the solid particles into liquid helium, a modified seeding method based on the fluidized-bed technology has been developed. The seeding results show that this method works well for hollow glass spheres because of their large particle size. For polymer particles that have very small size and thus higher tendency to coagulate, a two-phase fluidized bed technique has to be adopted in order to reduce the particle aggregation and achieve stable dispersion with more uniform spatial distribution.

6.2 Steady Thermal Counterflow


The thermal counterflow is generated in a rectangular channel by a heat current, and its velocity field under steady state condition has been measured using the PIV technique. Qualitative analysis shows that the measured particle velocity is in the same direction as that of the normal fluid velocity. It is inferred from the distribution of particle velocity that the normal fluid velocity in thermal counterflow is approximately parallel and uniform across the cross-section of the flow channel. And the velocity profile does not change apparently with the applied heat flux in a wide range from about 1.1 kW/m2
P P

up to 13 kW/m2.
P P

Comparisons between the measured particle velocity and theoretical normal fluid velocity show a significant discrepancy even when corrected for the slip velocity between the tracer particles and normal fluid. Furthermore, results at the bath temperatures from 1.61 K to 2.0 K indicate that the ratio of measured particle velocity to the theoretical normal fluid velocity temperatures,
v p ,a v n ,t v p ,a v n ,t

is most likely temperature independent. At all those

is approximately 0.5. The velocity discrepancy has been analyzed

132

from three different points of view: slip velocity, effect of turbulent flow, and additional force from the superfluid component. By turning the counterflow channel upside down and conducting the experiment again, the measured particle velocity is only a little larger than that when the channel is upward, but is still much smaller than the normal fluid velocity, indicating that the velocity discrepancy is not mainly due to the slip velocity. And the experimental result with different tracer particles, which are quite larger than polymer micro-spheres, also shows that the effect of turbulent flow is not the primary reason. Based on previous research, we suggest that there may exist an additional force
Fs due to the interaction of tracer particles with the quantized vortex lines in the

superfluid component. A semi-empirical correlation for this force is introduced. By adding it to the particle dynamics equation, an additional velocity difference v add is discovered as a result of the additional force. The analytical solution of v add indicates that it must be a linear function of the applied heat flux q and also independent of particle size, both of which have been verified by the experiment data. Furthermore, the linear coefficient of v add / q from the experiment data displays a temperature dependence of
T 6.7 , which is in reasonable agreement with the predicted temperature dependence.

These results strongly suggest that the velocity discrepancy between the particles and the normal fluid is primarily caused by an additional force resulting from the interaction between particles and the superfluid component.

6.3 Second Sound and Heat Diffusion Measurements


The propagation of second sound shock and heat diffusion in He II has been successfully measured by probing the induced transient thermal counterflow field with

133

PIV. The detailed information of second sound induced normal fluid velocity has been unveiled, in qualitative agreement with the assertion that the shock front is planar and thus the velocity induced is parallel and uniform across the entire channel. We have analyzed the transient counterflow field in a time sequence for two different heat pulses, 11.2 kW/m2 for 999 ms and 44.8 kW/m2 for 1 ms. Results from the 1 ms long pulse show
P P P P

clearly the arrival of shock front, effect of expansion fan, passage of shock tail, and the dissipation of induced superfluid turbulence. For the 999 ms long pulse, in addition to the arrival and passage of second sound, the onset and development of heat diffusion is also visualized. In particular, the particle velocity profile of the 999 ms heat pulse has been studied quantitatively. For this pulse, a relatively steady particle velocity of about 43 mm/s is reached around 2 ms after the arrival of the shock front. Compared to the particle velocity measured in steady thermal counterflow, this result is much smaller. A further analysis of this result suggests that the additional force Fs may be modified as a function of the vortex line density to ensure its applicability for both steady and transient state. The particle velocity stays at the steady state for only about 6 ms and then starts to drop. This gives a critical energy flux of about 67 J/m2 at 1.61 K, which is quite reasonable
P P

compared to a reported value of 75 J/m at 1.7 K.


P P

6.4 Suggestions for Potential Improvements


The study presented in this dissertation is the first of a kind experiment to apply the PIV technique to the research of He II. It has opened a new field in the study of thermal counterflow, superfluid turbulence, second sound and heat diffusion. As discussed above, the PIV experiments of thermal counterflow in both steady and transient state have been successful in exploring the details of induced velocity field that has never

134

been done before. But, some new issues have also been brought out, which require further improvements of current experiments to be clarified. These improvements can be summarized in the following aspects. 1) Though the heat transfer analysis of the heater shows that in steady state over 99% of the applied heat flux is transported by the thermal counterflow through the channel, it would be better to measure the actual transported heat flux q . According to equations (2.30) and (2.34), this can be done by placing two or three thermometers at different locations along the channel and measuring the temperature gradient T . Since
T could be very small for low applied heat flux, it is recommended to use an AC

resistance bridge to directly measure the temperature difference rather than measuring the temperatures at each location separately. This measurement would provide a stronger support to the existence of the additional force from superfluid component because q is a critical parameter to determine the theoretical normal fluid velocity v n ,t . 2) For the transient experiment of second sound shock, one important improvement would be the integration of velocity measurement with the temperature measurement. As discussed in Chap. 2.5.1, the conventional method of studying second sound is by measuring the temperature profile, and this method has been used successfully in exploring the superfluid turbulence. Also, the PIV experimental results discussed in Chap 5.2 show that the velocity profile displays a similar characteristic as the temperature profile. If these two profiles can be measured and studied together, the results will provide a new and significant insight into the dynamics of quantum turbulence. 3) The technical limitations of laser and camera used in current PIV system make it impossible to measure the propagation of second sound and heat diffusion in continuous mode. To overcome this problem, a high speed PIV system featuring a camera with CMOS sensor and a continuous laser may be used. With an operating rate as

135

high as 5 kHz, the system is able to measure the transient counterflow field in continuous mode with a temporal resolution of 2 ms. This would considerably save the measurement work and, more important, makes the integration of temperature and velocity measurements easier to be realized. For steady state measurement, the high-speed system would allow many more image pairs to be acquired so that the averaged particle velocity field would be more stable and have less spatial variation. 4) A new heater with improved thermal response is desirable for the transient experiment of second sound and heat diffusion. The heat transfer analysis shows that the current heater has a slow thermal response time of about 0.35 ms, which is one of the reasons for the rounding of the leading edge of velocity profiles. Also, the slow thermal response makes it hard, if not impossible, to determine how much heat flux has been transported at the initial stage of the heat pulse, a critical issue in the study of second sound. The film heater as described in [76] would be a good candidate for this new heater. It has a thin film of nichrome (80% Ni/20% Cr by weight) deposited by physical vapor deposition (PVD) onto a fused silica substrate, and the reported thermal response time is only 190 ns. 5) Different tracer particles should be tried more to check if they have better tracking characteristics, i.e. smaller slip velocity and faster response time, than the polymer particles. From the discussion in Chap. 4.2, it is preferred that the tracer particles have the same density with that of the fluid when turbulent flow is measured. Thus, the hydrogen-deuterium (H2-D2) particles are one of the best choices. But more work is
B B B B

needed to reduce the particle size down to about 1 m and develop an efficient way to disperse the H2-D2 particles into liquid helium with high concentration and uniform
B B B B

spatial distribution. Due to the temporal limits, the above improvements cannot be made for this dissertation. But they are highly recommended to be incorporated into future

136

experiments. It is expected they will be very helpful to further our understanding of superfluid helium.

137

REFERENCES

1. 2. 3.

L. Tisza, The -transition explained, Nature 141 (1938) 643-644 & 913. L. D. Landau, The theory of superfluidity of helium II, J. Phys. (U.R.S.S.) 5 (1941) 71. R. K. Childers and J. T. Tough, Helium II thermal counterflow: Temperature- and pressure-difference data and analysis in terms of the Vinen theory, Phys. Rev. B 13 (1976) 1040. C. Piotrowski and J. T. Tough, Helium II thermal counterflow: The effect of the vortex line drift velocity on the analysis of temperature difference data, Phys. Rev. B 17 (1978) 1474. D. R. Ladner, R. K. Childers and J. T. Tough, Helium II thermal counterflow at large heat currents, Phys. Rev. B 13 (1976) 2918. J. D. Henberger and J. T. Tough, Uniformity of the temperature gradient in the turbulent counterflow of He II, Phys. Rev. B 25 (1982) 3123. S. S. Courts and J. T. Tough, Transition to superfluid turbulence in two-fluid flow in He II, Phys. Rev. B 38 (1988) 74. J. T. Tough, in Progress in Low Temperature Physics, Vol. VIII, North Holland Publishing (1982) Ch. 3.
U U

4.

5. 6. 7. 8. 9.

K. P. Martin and J. T. Tough, Evolution of superfluid turbulence in thermal counterflow, Phys. Rev. B 27 (1983) 2788.

10. T. Kitamura, K. Sheramizu et. al, A numerical model on transient, two-dimensional flow and heat transfer in He II, Cryogenics 37 (1997) 1. 11. T. Suekane M. Sekiguchi et. al, Heat transfer and flow of He II in narrow channels, Cryogenics 45 (2003) 125. 12. V. P. Peshkov, Determination of the velocity of propagation of the second sound in helium II, J. Phys. (U.S.S.R.) 10 (1946) 389.

138

13. R. J. Adrian, Particle-imaging techniques for experimental fluid mechanics, Ann. Rev. Fluid Mech. 23 (1991) 261. 14. R. J. Donnelly, et al, The use of Particle Image Velocimetry in the study of turbulence in liquid helium, J. Low Temp. Phys. 126 (2002) 327. 15. D. Celik, T. Zhang and S. W. Van Sciver, Application of PIV to counterflow in He II, Advances in Cryogenic Engineering 47B (2002) 1372. 16. T. Zhang, D. Celik and S. W. Van Sciver, Tracer particles for application to PIV studies of liquid helium, J. Low Temp. Phys. 134 (2004) 985. 17. S. W. Van Sciver, Helium Cryogenics, Plenum Press, New York (1986).
U U

18. J. F. Allen and A. D. Misener, The fountain effect, Nature 141 (1938) 243. 19. W. H. Keesom and G. E. MacWood, Physica 5 (1938) 737. 20. E. L. Andronikashvili, Zh. Esksp Theor. Fiz. 16 (1946) 780; 18 (1948) 424. 21. D. V. Osborne, The rotation of liquid helium II, Proc. Phys. Soc. A 63 (1950) 909. 22. H. E. Hall and W. F. Vinen, The rotation of liquid helium II: I. Experiments on the propagation of second sound in uniformly rotating helium II, Proc. R. Soc. A 238 (1956) 204. 23. H. E., Hall and W. F. Vinen, The rotation of liquid helium II: II. The theory of mutual friction in uniformly rotating helium II, Proc. R. Soc. A 238 (1956) 215. 24. W. F. Vinen, Mutual friction in a heat current in liquid helium II: I. Experiments on steady heat currents, Proc. R. Soc. A 240 (1957) 114. 25. W. F. Vinen, Mutual friction in a heat current in liquid helium II: II. Experiments on transient effects, Proc. R. Soc. A 240 (1957) 128. 26. W. F. Vinen, Mutual friction in a heat current in liquid helium II: III. Theory of mutual friction, Proc. R. Soc. A 242 (1957) 493.

139

27. W. F. Vinen, Mutual friction in a heat current in liquid helium II: IV. Critical heat currents in wide channel, Proc. R. Soc. A 243 (1958) 400. 28. C. J. Gorter and J. H. Mellink, On the irreversible processes in liquid helium II, Physica XV (1949) 285. 29. K. R. Atkins, Liquid Helium, Cambridge University Press, Cambridge (1959).
U U

30. S. K. Nemirovskii and A. N. Tsoi, Transient thermal and hydrodynamic processes in superfluid helium, Cryogenics 29 (1989) 985. 31. I. M. Khalatnikov, An introduction to the theory of superfluidity, Benjamin, New York (1965).
U U

32. A. J. Dessler and W. M. Fairbank, Amplitude dependence of the velocity of seond sound, Phys. Rev. 104 (1956) 6. 33. T. N. Turner, Using second-sound shock waves to probe the intrinsic critical velocity of liquid helium II, Phys. Fluids 26 (1983) 3227. 34. T. Shimazaki and M. Murakami, Second sound wave heat transfer, thermal boundary layer formation and boiling; highly transient heat transport phenomena in He II, Cryogenics 35 (1995) 645. 35. L. Dresner, Transient heat transfer in superfluid helium: I, Advances in Cryogenic Engineering 27 (1982) 411. 36. L. Dresner, Transient heat transfer in superfluid helium: II, Advances in Cryogenic Engineering 29 (1984) 323. 37. S. K. Nemirovskii, L. P. Kondaurova and A. Ja. Baltsevich, Heat transfer in He II with stepwise switching on of heater, Cryogenics 32 (1992) 47. 38. T, Iida, M. Murakami and etc., Visualization study on the thermo-hydrodynamic phenomena induced by pulsative heating in He II by the use of a laser holographic interferometer, Cryogenics 36 (1996) 943. 39. M. Raffel, C. Willert and J. Kompenhans, Particle Image Velocimetry a practical guide, Springer-Verlag, Heidelberg, (1998).
U U

140

40. A. Melling, Tracer particles and seeding for Particle Image Velocimetry, Meas. Sci. Technol. 8 (1997) 1406. 41. R. J. Adrian, Particle-imaging techniques fro experimental fluid mechanics, Annu. Rev. Fluid Mech. 23 (1991) 261. 42. R. D. Keane, R. J. Adrian and Y. Zhang, Super-resolution particle imaging velocimetry, Meas. Sci. Technol. 6 (1995) 755. 43. M. Stanislas and J. C. Monnier, Practical aspects of image recording in particle image velocimetry, Meas. Sci. Technol. 8 (1997) 1417. 44. D. D. Udrea, P. J. Bryantston-Cross and etc., Two sup-pixel processing algorithms for high accuracy particle centre estimation in low seeding density particle image velocimetry, Optics & Laser Technology 28 (1996) 389. 45. D. P. Hart, PIV error correction, Exp. Fluids 29 (2000) 13. 46. J. Westerweel, D. Dabiri and M. Gharib, The effect of a discrete window offset on the accuracy of cross-correlation analysis of PIV recordings, Exp. Fluids 23 (1997) 20. 47. M. P. Wernet and A. Pline, Particle displacement tracking technique and camera lower bound error in centroid estimates from CCD image, Exp. Fluids 15 (1993) 295. 48. H. Huang, An extension of digital PIV-processing to double-exposed images, Exp. Fluids 24 (1998) 364. 49. J. Westweel, Digital particle image velocimetry theory and application, Ph. D. thesis, Technical University of Delft (1993). 50. K. L. Chopra and J. B. Brown, Suspension of particles in liquid helium, Phys. Rev. 108 (1957) 157. 51. F. Bielert and G. Stramm, Visualization of Taylor-Couette flow in superfluid helium, Cryogenics 33 (1993) 938. 52. M. Murakami and N. Ichikawa, Flow visualization study of thermal counterflow jet in He II, Cryogenics 29 (1989) 438.

141

53. A. Nakano and M. Murakami, Velocity measurements of He II thermal counterflow jet accompanied by second sound Helmholtz oscillation, Cryogenics 34 (1994) 179. 54. A. Nakano and M. Murakami, Flow structure of thermal counterflow jet in He II, Cryogenics 34 (1994) 991. 55. As reported in M. I. Yorquez-Ramirez and G. R. Duursma, Insights into the instantaneous freeboard flow above a bubbling fluidised bed, Powder Technology 116 (2001) 76. 56. H. E. Albrecht, M. Borys, N. Damaschke and C. Tropea, Laser Doppler and Phase Doppler measurement techniques, Springer-Verlag, Heidelberg (1998) Chap. 13.
U U

57. R. Mei, Velocity fidelity of flow tracer particles, Exp. Fluids 22 (1996) 1. 58. D. Celik and S. W. Van Sciver, Tracer particle generation in superfluid helium through cryogenic liquid injection for particle image velocimetry (PIV) applications, Exp. Therm. Fluid Sci. 26 (2002) 971. 59. M. R. Smith, Y. S. Choi, and S. W. Van Sciver, in Quantized Vortex Dynamics and Superfluid Turbulence, Newton Institute, Cambridge, UK, August 14-18, 2000, Springer-Verlag, (2001) 51.
U U

60. P. J. Bryanston-Cross, et al, The application of digital particle image velocimetry (DPIV) to transonic flows, Prog. Aerospace Sci. 31 (1995) 273. 61. J. G. Santiago, S. T. Wereley, C. D. Meinhart, D. J. Beebe and R. J. Adrian, A particle image velocimetry system for microfluidics , Exp. Fluids 25 (1998) 316. 62. T. Zhang, D. Celik and S. W. Van Sciver, A way of seeding solid particles into He II for PIV studies, in Proceedings of 19th International Cryogenic Engineering Conference, ed. G Gistau-Baguer and P. Seyfert, Narosa Pub. (2003) 755.
U UP UP U

63. R. E. Boltnev, G. Frossati, E. B. Gordon etc., Embedding impurities into liquid helium, J. Low Temp. Phys. 127 (2002) 245. 64. A. H. Lefebvre, Atomization and Sprays, Hemisphere Publishing, NY (1989) pp. 106-112.
U U

142

65. D. Celik, S. W. Van Sciver, Large size optical windows for superfluid helium applications, Cryogenics 42 (2002) 547. 66. D. Y. Chung and P. R. Critchlow, Motion of suspended particles in turbulent superflow of liquid helium II, Phys. Rev. Letters 14 (1965) 892. 67. S. W. Van Sciver, vnsciver@magnet.fsu.edu (private communication).
HTU UTH

68.

S. K. Nemirovskii, Propagation of heat pulses generating quantized vortices in He II, Sov. Phys. JETP 64 (1986) 803.

69. L. S. Goldner, G. Ahlers and R. Mehrotra, Quantitative studies of nonlinear second sound in superfluid 4He, Phys. Rev. B 43 (1991) 12861.
P P

70. L. P. Kondaurova, S. K. Nemirovskii and M. V. Nedoboiko, Mutual influence of quantum vortices and heat pulses in superfluid helium, Low Temp. Phys. 25 (1999) 475. 71. J. R. Torczynski, On the interaction of second sound shock waves and vorticity in superfluid helium, Phys. Fluids 27 (1984) 2636. 72. D. K. Hilton and S. W. Van Sciver, Techniques for the detection of second sound shock pulses and induced quantum turbulence in He II, Cryogenics 41 (2001) 347. 73. V. B. Efimov, G. V. Kolmakov and etc., Propagation of short nonlinear secondsound pulses through He II in one- and three-dimensional geometry, Low Temp. Phys. 24 (1998) 81. 74. W. Fiszdon, Z. Paradzynski and G. Stamm, The evolution of axisymmetric rectangular second-sound (heat) pulses in superfluid helium, Phys. Fluids A 1 (1989) 881. 75. V. B. Efimov, G.V. Kolmakov and etc., Generation of the second and first sound waves by a pulses heater in He II under pressure, J. Low Temp. Phys. 119 (2000) 309. 76. D. K. Hilton, Growth and decay of quantum turbulence induced by second sound shock pulses in helium II, Ph. D. Dissertation, Department of Physics, Florida State University (2002).
U U

143

BIOGRAPHICAL SKETCH
Tao Zhang

Professional Experience
U

Graduate Research Assistant January 1999 April 2004 Cryogenics Laboratory National High Magnetic Field Laboratory Florida State University, Tallahassee, Florida

Conducted general experimental research on the heat and mass transfer and fluid mechanics at low temperatures. Studied the application of new flow diagnostic technology to the research of superfluid helium and superfluid turbulence. Developed new measuring techniques and devices for cryogenic experimentation. Designed the thermal system for cryogenic experiments, and studied the application of cryogenic cooling technology to large-scale systems.
Graduate Research Assistant September 1997 November 1998 Cryogenics Laboratory Department of Cryogenics and Refrigeration Engineering Shanghai Jiao Tong University, Shanghai, China

Conducted research on physical properties and heat transfer characteristics of various materials at low temperatures. Studied the applications of new cryogenic cooling system and cryo-cooler technology. Investigated the heat transfer processes at solid interfaces in liquid nitrogen temperature range. Supervised the thesis work of three undergraduate students in the laboratory.
Graduate Teaching Assistant September 1996 August 1997 Department of Cryogenics and Refrigeration Engineering Shanghai Jiao Tong University, Shanghai, China

Co-taught Cryogenic Engineering and Cryogenic and Vacuum Technology to over 50 undergraduate students, and instructed them on the cryogenic experiments. Graded the assignments, and assisted with student assessment.

144

Educational Background
U

Doctor of Philosophy January 1999 April 2004 Department of Mechanical Engineering Florida State University, Tallahassee, Florida

Degree Awarded: April 2004


Master of Science September 1996 November 1998 Department of Cryogenics and Refrigeration Engineering Shanghai Jiao Tong University, Shanghai, China

Degree Awarded: March 1999


Bachelor of Science September 1992 July 1996 Department of Energy Engineering Shanghai Jiao Tong University, Shanghai, China

Degree Awarded: July 1996

Publications
U

Zhang, T. and Van Sciver, S. W., Use of the Particle Image Velocimetry (PIV) Technique to Study Propagation of Second Sound Shock in Superfluid Helium, Submitted to Phy. Rev. Letters (2004). Zhang, T., Celik, D. and Van Sciver, S.W., Tracer Particles for Application to PIV Studies of Liquid Helium. J. Low Temp. Phys. 134(3/4) (2004) 985-1000. Zhang, T. and Van Sciver, S.W., Measurements of He II Thermal Counterflow Using PIV Technique. Advances in Cryogenic Engineering 49, AIP Press, New York (article in press). Zhang, T., Celik, D. and Van Sciver, S.W., A Way of Seeding Solid Particles into He II for PIV Studies, in Proceedings of 19th International Cryogenic Engineering Conference, ed. G Gistau-Baguer and P. Seyfert, Narosa Pub., New Delhi (2003) 755-758.
U UP UP U

145

Zhang, T., Smith, M.R. and Van Sciver, S.W., A Helium II Heat Transfer Level Meter, in Proceedings of the ASME-Heat Transfer Division, Vol.1 (2000) 215-219.
U U

Zhang, T. and Xu, L., Comparison and Analysis of Theoretical Models in the Research of Thermal Contact Conductance. Cryogenics and Superconductivity (in Chinese) 26(2) (1998) 58-64. Zhang, T. and Xu, L., Experimental Research on the Thermal Contact Resistance between Cu-Cu in Vacuum and Low Temperature. Cryogenics (in Chinese) 108(2) (1999) 19-26. Celik, D., Zhang, T. and Van Sciver, S.W., Application of PIV to Counterflow in He II. Advances in Cryogenic Engineering, 47B, AIP Press, New York, (2002) 1372-1379. Smith, M.R., Zhang, T. and Xiang, Y., Experimental Investigation of the Thermal Resistance in Niobium Samples for Superconducting RF Cavities. Advances in Cryogenic Engineering, 45A, Plenum Press, New York, (2000) 899-904. Xu, L., Zhang, T. and Xiong, W., The Thermal Contact Resistance between the Solid Interface at Low Temperature and Vacuum, in Cryogenics and Refrigeration Proceedings of ICCR98, International Academic Publishers, Beijing, (1998) 321-324.
U U

Xu, L. and Zhang, T., The Effect of Contact Interfaces on Thermal Contact Resistance at Low Temperature and Vacuum. Vacuum and Cryogenics (in Chinese) 4(1) (1998) 1-4. Xu, L. and Zhang, T., Using the Double Heat-Flux Meter Method to Measure the Thermal Contact Resistance of Solid Material at Low Temperature and Vacuum. Cryogenics (in Chinese) 110(4), (1999) 185-189. Celik, D., Hilton, D.K., and Zhang, T., Helium II Level Measurement Techniques. Cryogenics 41 (2001) 355-366. Xu, L., Xiong, W. and Zhang, T., The Research on Thermal Expansion Characteristics of the Nonflammable Phenolic Foam Insulation Material, in Cryogenics and Refrigeration Proceedings of ICCR98, International Academic Publishers, Beijing, (1998) 474-477.
U U

Xiong, W., Xu, L. and Zhang, T., Comparison and Analysis of Pulse Tube Refrigerator and Thermal Separator. Cryogenics (in Chinese) 105(5) (1998) 45-51. Xu, L., Xiong, W. and Zhang, T., The Applications of Helium Refrigeration in Space Refrigeration Technology. Cryogenics (in Chinese), 101(1) (1998) 1-6.

146

Xu, L., Yu, Z.F., Cao, M.L., Xiong W., Zhang, T. and Zhao, L.P., Measurements of the Expansion Coefficients of Several Sealing Materials at Low Temperature. Cryogenics (in Chinese) 102(2) (1998) 28-31. Xu, L., Yang, J., Xu, J.M., Li, S.M., Xiong W. and Zhang, T., The Effect of Solid Interfaces on Thermal Contact Resistance at Low Temperature. Advances in Cryogenic Engineering 43B, Plenum Press, New York (1998) 1369-1375.

147

You might also like