1: \documentclass[12pt,preprint]{aastex}
2:
3: %
4: % BEGIN PREAMBLE
5: %
6: \renewcommand{\b}[1]{\boldsymbol{#1}}
7: \newcommand{\pard}[2]{\frac{\partial #1}{\partial #2}}
8: \newcommand{\unit}[1]{\,{\rm #1}}
9: % Units
10: \newcommand{\cm}{\unit{cm}}
11: \newcommand{\G}{\unit{G}}
12: \newcommand{\kG}{\unit{kG}}
13: \newcommand{\rpm}{\unit{rpm}}
14: \newcommand{\s}{\unit{s}}
15: % Special symbols
16: \newcommand{\ab}{{\rm a}}
17: \newcommand{\bp}{{\rm b}}
18: \newcommand{\bB}{\b{B}}
19: \newcommand{\cs}{c_{\rm s}}
20: \newcommand{\half}{{\textstyle\frac{1}{2}}}
21: \newcommand{\bO}{\b{\Omega}}
22: \newcommand{\Rm}{Re_{\rm m}}
23: % Environments and macros
24: \newcommand{\remark}[1]{{\bf #1}}
25: %
26: % END PREAMBLE
27: %
28:
29: \newcommand{\noun}[1]{\textsc{#1}}
30: %% Bold symbol macro for standard LaTeX users
31: \providecommand{\boldsymbol}[1]{\mbox{\boldmath $#1$}}
32:
33: %% Because html converters don't know tabularnewline
34: \providecommand{\tabularnewline}{\\}
35: %\usepackage{bm}% bold math
36:
37: %\usepackage{babel}
38: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% End of user commands
39: \shorttitle{2D convective instability simulations}
40: \shortauthors{W. Liu \emph{et al.}}
41:
42: \begin{document}
43:
44: \title{Noise-Sustained Convective Instability in a Magnetized Taylor-Couette Flow}
45:
46: \author{Wei Liu\altaffilmark{1}}
47:
48: \altaffiltext{1}{Current address: Theoretical Division, Los Alamos National Laboratory, Los Alamos, NM, USA 87545}
49:
50: \affil{Center for Magnetic Self-Organization in Laboratory and Astrophysical Plasma, Princeton Plasma Physics Laboratory, Princeton, NJ, USA 08543}
51: \email{wliu@lanl.gov}
52:
53: %\author{Jeremy Goodman}
54:
55: %\affil{Princeton University Observatory, Princeton, NJ, USA 08544}
56:
57: %\author{Hantao Ji}
58:
59: %\affil{Center for Magnetic Self-Organization in Laboratory and Astrophysical Plasma, Princeton Plasma Physics Laboratory, Princeton University, P.O. Box
60: %451, Princeton, NJ, USA 08543 }
61:
62: \begin{abstract}
63: The helical magnetorotational instability of the magnetized Taylor-Couette flow is studied numerically in a finite cylinder. A distant upstream insulating boundary is shown to stabilize the convective instability entirely while reducing the growth rate of the absolute instability. The reduction is less severe with larger height. After modeling the boundary conditions properly, the wave patterns observed in the experiment turn out to be a noise-sustained convective instability. After the source of the noise resulted from unstable Ekman and Stewartson layers is switched off, a slowly-decaying inertial oscillation is observed in the simulation. We reach the conclusion that the experiments completed to date have not yet reached the regime of absolute instability.
64:
65: \end{abstract}
66: \keywords{accretion, accretion disk---instability---(magnetohydrodynamics:) MHD
67: ---methods: numerical}
68:
69: %\maketitle
70: \section{Introduction}
71: The magnetorotational instability (MRI) is probably the main source of
72: turbulence and accretion in sufficiently ionized astrophysical disks
73: \citep{bh98}. Due to this crucial role in astrophysics, substantial efforts have been spent worldwide to observe MRI in a laboratory setting \citep{jgk01,gj02,npc02,sisan04,vils06}, but MRI has never been conclusively demonstrated in the laboratory.
74:
75: Most experiments have been done in cylindrical geometry with a background flow that approximates the ideal Couette rotating profile:
76: \begin{equation}
77: \label{couette}
78: \Omega=a+b/r^{2}
79: \end{equation}
80: where $a=(\Omega_{2}r_{2}^{2}-\Omega_{1}r_{1}^{2})/(r_{2}^{2}-r_{1}^{2})$
81: and $b=r_{1}^{2}r_{2}^{2}(\Omega_{1}-\Omega_{2})/(r_{2}^{2}-r_{1}^{2})$, $\Omega_1$ and $\Omega_2$ are the rotation speed of the inner and outer cylinder and $r_1$ and $r_2$ are the radius of the inner and outer cylinder, respectively (see Fig.~\ref{boundary}).
82: For axially periodic or infinite magnetized Taylor-Couette flow, MRI-like modes have been shown theoretically to grow at much reduced magnetic Reynolds number $\Rm\equiv\Omega_1r_1(r_2-r_1)/\eta$ and Lundquist number $S\equiv V_{A,0}r_1(r_2-r_1)/\eta$ in the presence of a combination of axial and current-free toroidal field %
83: \begin{equation}
84: \label{eq:backgroundfield}
85: \b{B}^{0}=B_z^{0}\left(\b{e}_z+\beta r_1/r \b{e}_\varphi\right)
86: \end{equation}
87: than the standard MRI (SMRI) with purely axial magnetic field \citep{hr05,rhss05}. Here the cylindrical coordinates $(r,\varphi,z)$ are used. $B_z^{0}$ and
88: $\beta$ are constants. The Alfv\'en speed is defined as $V_{A,0}\equiv B_z^{0}/\sqrt{4\pi\rho}$. $\eta$ and $\rho$ are the magnetic diffusivity and density of the fluid, respectively (see Fig.~\ref{boundary}).
89:
90: The Potsdam ROssendorf Magnetic Instability Experiment (PROMISE) group claimed to have observed this kind of helical MRI (HMRI) experimentally \citep{sgg06,rhsgg06,sgg07}. However we have shown that the wave pattern observed in PROMISE is not a global instability, but rather a transient disturbance somehow excited by the Ekman circulation and then transiently amplified as it propagates along the background axial Poynting flux with nonzero group and phase velocities, but is then absorbed once it reaches the jet formed at midheight between two neighboring Ekman cells \citep{lgj07}. PROMISE group have accordingly updated the experimental facility to PROMISE II to allow for two split rings at both endcaps: the inner ring attached to the inner cylinder and outer ring attached to the outer cylinder. If the width of the inner ring is chosen appropriately $\sim0.4(r_2-r_1)$, the magnetized Ekman circulation could be significantly reduced, therefore removing one of the possible disturbance sources, \emph{i.e.} the unsteady jet \citep{sj07}.
91:
92: As with other examples in the literatires, such as drifting dynamo waves \citep{tpk98, ptk00}, it is of vital importance to distinguish absolute instability from convective instability in a traveling wave experiment like PROMISE. It is also an essential ingredient of the threshold prediction for the Riga dynamo \citep{ggglps08}. For a traveling wave the positivity of the growth rate implies only an amplification of the perturbation as it moves downstream. In one case, despite the movement of the wave packet, the perturbation increases without limit in the course of time at any point fixed in space; this kind of instability with respect to any infinitesimal perturbations will be called \emph{absolute instability}. In the other case, the packet is carried away so swiftly that at any point fixed in space the perturbation tends to zero as $t\rightarrow\infty$; this kind will be called \emph{convective instability} \citep{ll87} (see the details of \S\ref{packet}). For PROMISE II, it appears that under the experimental conditions the second kind occurs. A recent preprint also highlights the importance of the distinction between absolute and convective instabilities in the context of the HMRI \citep{pg08}.
93:
94: In a Taylor-Couette experiment bounded by insulating endcaps, \citet{tpk98} have pointed out that without any external disturbances except a small initial disturbance needed as a seed for the instability, the distant upstream insulating boundary acts as an ``absorbing" boundary while the characteristics of the downstream endcap is unimportant. Due to this absorption the convective unstable state cannot be sustained by a uniform driving force, therefore this unstable mode eventually decays \citep{tpk98}. This driving force is not the noise mentioned before, but the power to drive the instability, which in the usual Taylor-Couette experiments can be quantified by the magnetic Reynolds number $\Rm$. This conclusion has been rigorously demonstrated in the very resistive limit in \S II.C of \citet{lghj06} using a perturbative approach and \S II.D of \citet{lghj06} using a modifed WKB analysis, showing that the insulating endcap entirely stabilizes the HMRI mode, which is a convective unstable mode given the parameters of the PROMISE experiment.
95:
96: The absorbing boundary is essential to the development, regardless of how distant it may be. The larger height only defers the time when we have to wait for the boundary-induced dissipation to dominate \citep{tpk98}. On the other hand, if $\Rm$ exceeds a higher threshold $Re_{\rm m,f}$, the driving force of the system overcome the dissipation and a globally unstable mode appears \citep{tpk98}. Therefore in a bounded system the unstable mode appears at $Re_{\rm m,f}$ rather than $Re_{\rm m,c}$, where $Re_{\rm m,c}$ is the critical magnetic Reynolds number for the onset of the convective unstable mode without the ``absorbing" boundary. \citet{tpk98} has showed that in the presence of an ``absorbing" boundary and large $h$, a global unstable mode appears when
97: \[
98: \Rm\geqslant Re_{\rm m,f}\equiv Re_{\rm m,a}+O(h^{-2})\,,
99: \]
100: where $Re_{\rm m,a}$ is the critical magnetic Reynolds number corresponding to the onset of the absolute instability without the ``absorbing" boundary.
101:
102: This has raised a big obstacle for people to observe absolutely unstable HMRI in the laboratory. The advantage of HMRI itself, \emph{i.e.}, unstable with a low critical Reynolds number ($3$ orders lower than the SMRI) conflicts with the necessarily high threshold of the onset of an absolute HMRI mode, \emph{i.e.}, excited at a reasonably high critical magnetic Reynolds number, thus high Reynolds number, which would result in much more severe end-effects than people had expected. Moreover the fact that the critical Lundquist number must usually increase together with the magnetic Reynolds number and high ratio of toroidal-to-poloidal magnetic field requirement ($\beta>1$) would even worsen the situation.
103:
104: We also find that by nonlinear numerical simulation the insulating endcap reduces the growth rate of the absolute instability somewhat. The higher the height $h$ is, the less the growth rate is reduced (Table~\ref{reduce}).
105:
106: In a typical experiment, the experiment is, however, highly likely affected by small external noise either from a physical cause or experimental imperfection such as the misalignment of the cylinders. If the system is convectively unstable, \emph{i.e.}, disturbances grow as they move downstream, noise would sustain structures in the system even if no global mode is unstable \citep{dr87,ptk00}. In the present paper, we show by numerical simulations that the perturbations from the unstable magnetized residual Ekman layer and Stewartson layer at the upper endcap would play the role of ``noise" generator, though this perturbation level is reduced with increasing axial magnetic field \citep{lw08a}. What is observed in PROMISE II turns out to be a noise-sustained convective traveling wave, not the absolute unstable mode.
107:
108:
109: This paper is organized as follows:
110: \S\ref{packet} presents the wave packet analysis in a unbounded cylinder, which is the basis of the following sections.
111: %In Section~\ref{influence}, we summary our results of the influence of the distant insulating endcap upon the instability in the point of view of the distinction between absolute and convective instability and show how the height of the cylinder affects them respectively in a finite cylinder by numerical simulations.
112: We report the nonlinear simulation results with \emph{partially} conducting boundary conditions of PROMISE II experiment in \S\ref{noise}. The final conclusions and implications to the HMRI experiments are given in \S\ref{discussion}.
113:
114:
115: \section{Wave Packet Analysis in an Unbounded Cylinder}\label{packet}
116: %We have the ideal Couette rotating profile:
117: %\begin{equation}
118: %\label{couette}
119: %\Omega=a+b/r^{2}
120: %\end{equation}
121: %where $a=(\Omega_{2}r_{2}^{2}-\Omega_{1}r_{1}^{2})/(r_{2}^{2}-r_{1}^{2})$
122: %and $b=r_{1}^{2}r_{2}^{2}(\Omega_{1}-\Omega_{2})/(r_{2}^{2}-r_{1}^{2})$, where $\Omega_1$ and $\Omega_2$ are the rotation speed of the inner and outer cylinder respectively (see Fig.~\ref{boundary}).
123: Assuming a cylinder of infinite height $h$, $k_{z}$ is a continuous variable.
124: Let the gap width be fixed and finite, so $k_{r}\cong\pi/(r_{2}-r_{1})$. We define the total wavenumber $K=\sqrt{k_{r}^{2}+k_{z}^{2}}$ and the growth rate $\gamma$.
125:
126: Since the fast growing mode is the dominant mode, here we focus on waves with vertical wavenumber $k_z$ close to that of the fastest growing mode, $k_z^0$. The range of values of $k_z$ lies near the point for which $\gamma(k_z)$ is a maximum, \emph{i.e.} $d\gamma/d k_z=0$ at $k_z=k_z^0$ [as seen from Fig.~\ref{fig:global} (a)]. Let a slight perturbation occurs near the middle of the flow $(z\sim0)$ in the format of a wave packet as follows:
127: \begin{equation}
128: \label{wavepacket}
129: B_{r}(z,t=0)=b_{0}\exp\left(-\frac{z^{2}}{2L^{2}}\right)\exp(ik_{z}^{0}z)\, ,
130: \end{equation}
131: where we have used the envelope $\exp(-z^{2}/2L^{2})$ to confine the perturbation around the central part of the cylinder, where $L\sim O(h)$. In the course of time, the components for which $\gamma(k_z)>0$ will be amplified, while the remainder will be damped. The amplified wave packet thus formed will also be carried downstream with a velocity equal to the group velocity $d\omega/dk_z$ of the packet, where $\omega=\mathbb{R}\omega+i\gamma$ and $\mathbb{R}\omega$ is the real part of the frequency; since we are now considering waves whose wave numbers lies in a small range near the point where $d\gamma/dk_z=0$, the quantity
132: \begin{equation}
133: \label{group_velocity}
134: V_g=d\omega/dk_z \cong d(\mathbb{R}\omega)dk_z
135: \end{equation}
136: is real, and is therefore the actual velocity of propagation of the packet. This downstream displacement of the perturbations is very important, and causes the complications of absolute instability \emph{v.s.} convective instability.
137:
138: We can approximate the dispersion relation like (Fig.~\ref{fig:global}):
139: \begin{eqnarray}
140: \label{approx1}
141: \mathbb{R}\omega=\mathbb{R}\omega(k_{z})=\kappa\frac{k_{z}}{K}\,; \\
142: \label{approx2}
143: \gamma=\gamma(k_{z})=\gamma^{0}-\frac{\sigma}{2}(k_{z}-k_{z}^{0})^{2}\,,
144: \end{eqnarray}
145: in which $\kappa^{2}=\frac{1}{r^{3}}\frac{d}{dr}(r^{2}\Omega)^{2}=4(1+Ro)\Omega^{2}$ and $Ro\equiv1/2d\ln \Omega/d\ln r=a/\Omega-1$ is the Rossby number. We know $\gamma=0$ when $k_{z}=0$. Thus
146: $\sigma=2\gamma^{0}/k_{z}^{02}$. And in order to simplify the derivation, we assume $K\approx constant $ from now on (though this is not a good approximation, we can get some insightful results from this simple approximation). From Eq.~\ref{approx1}, we get
147: $V_g=\kappa/K$.
148:
149: At later time $t>0$
150: \begin{eqnarray}
151: \label{eqn5}
152: \widetilde{B_{r}}(k_{z},t)&=\widetilde{B_{r}}(k_{z},0)\exp(\gamma(k_{z})t+i\mathbb{R}\omega(k_{z})t) \nonumber\\
153: &=\widetilde{B_{r}}(k_{z},0)\exp\{[\gamma^{0}-\frac{\sigma}{2}(k_{z}-k_{z}^{0})^{2}]t+i\kappa\frac{k_{z}}{K}t\}\,,
154: \end{eqnarray}
155: if we define $D=\sqrt{L^{2}+\sigma t}$, the result can be expressed as:
156: \begin{equation}
157: \label{eqn20}
158: B_{r}(z,t)=b_{0}\frac{L}{D}\exp(\gamma^{0}t)\exp\left[-\frac{(z+V_gt)^{2}}{2D^{2}}\right]\exp[ik_{z}^{0}(z+V_gt)]\,.
159: \end{equation}
160: In Eq.~\ref{eqn20}, As $t\rightarrow0$, Eq.~\ref{eqn20} can be simplified as:
161: \begin{equation}
162: \label{eqn21}
163: B_{r}(z,t)=b_{0}\exp(\gamma^{0}t)\exp[ik_{z}^{0}(z+V_gt)]\,,
164: \end{equation}
165: which is a ``transient" growing phase.
166: As $t\rightarrow\infty$,
167: \begin{equation}
168: \label{convective}
169: B_{r}(z,t)=b_{0}\frac{L}{\sqrt{\sigma t}}\exp\left[\left(\gamma^{0}-\frac{V_g^2}{2\sigma}\right)t\right]\exp[ik_{z}^{0}(z+V_gt)]
170: \end{equation}
171:
172: Obviously, If $\gamma^0<\gamma_{\rm a}=V_g^2/2\sigma$, we will get \emph{convective instability}, that is, it starts with a transiently growing phase (Eq.~\ref{eqn21}), followed by a phase asymptotically decaying to zero (Eq.~\ref{convective}).
173: If $\gamma^0>\gamma_{\rm a}$, we will get \emph{absolute instability}.
174:
175: \section{Noise-Sustained Convective Instability in PROMISE II Experiment}\label{noise}
176: In order to reduce the undesirable effects induced by the endcaps and also the accompanying hydromagnetic asymmetries, \citet{sj07} have proposed to split both endcaps into two rings which are attached to both cylinders and found that if the width of the inner ring is chosen to be $0.4D$ (see Fig.~\ref{boundary}), where $D=r_2-r_1$ is the gap between the inner and outer cylinder, the magnetic energy in term of $b_{\varphi}$, where $b_{\varphi}$ is the perturbed azimuthal magnetic field, is minimized. Therefore the magnetized Ekman circulation is significantly reduced, leading to a satisfactory ideal Couette state (Eq.~\ref{couette}) in the bulk flow. PROMISE has been accordingly updated to PROMISE II adopting this idea.
177:
178: While we have confirmed their conclusions (Fig.~\ref{noekman}) (Please note that in \citet{sj07}, this conclusion is derived with $\beta=0$, \emph{i.e.}, no background toroidal magnetic field, while our simulation results show that this conclusion is also valid with nonzero $\beta$), here we report nonlinear simulations with the ZEUS-MP
179: 2.0 code \citep{hnf06}, which is a
180: time-explicit, compressible, astrophysical ideal MHD parallel 3D code,
181: to which we have added viscosity, resistivity (with subcycling to reduce
182: the cost of the induction equation), and \emph{partially} conducting boundary conditions \citep{lgj07}, for axisymmetric flows in cylindrical coordinates
183: $(r,\varphi,z)$. It has been demonstrated that the finite conductivity ($\eta_{\rm Cu}= 1.335\times10^{2}\unit{cm^{2}s^{-1}}$) and thickness of the
184: copper vessel are important,
185: and this noticeably improves agreement with the measurements compared
186: to previous much simplified boundary condition \citep{lgj07}. Please note that in this paper $\mu=\Omega_2/\Omega_1=0.26$, rather than $\mu=0.27$ reported in previous work.
187: The parameters of PROMISE II as reported in or inferred from
188: \citet{sgg08} are used: gallium density
189: $\rho=6.35\unit{g\;cm^{-3}}$, magnetic diffusivity $\eta=2.43\times10^{3}\unit{cm^{2}\;s^{-1}}$, magnetic Prandtl
190: number $Pr_{\rm m}\equiv\nu/\eta=1.40\times10^{-6}$; Reynolds number $Re\equiv
191: \Omega_{1}r_{1}(r_{2}-r_{1})/\nu=1775$; axial current
192: $I_{z}=6000\unit{A}$; toroidal-coil currents $I_\varphi=0, 50, 75,
193: 120\unit{A}$; and dimensions as in Fig.~\ref{boundary}.
194:
195: For comparison, we start with purely hydrodynamic (unmagnetized) simulations (Fig.~\ref{hydro}). From Fig.~\ref{hydro} (a), after splitting the endcaps into two rings, the two big Ekman cells are divided into four smaller cells and localized near the endcaps. Compared to the simulation results of PROMISE \citep{lgj07}, there is not an flapping ``jet" near the mid-plane as in the usual Ekman circulations. This removes the possible noise from this unsteadiness. However from Fig.~\ref{hydro} (b), there are some perturbations near both endcaps, which supply the possible sources of noise in the system. These perturbations are resulted from unstable Ekman layer and Stewartson layer \citep{lw08a}. The magnitude of this noise is around $\pm0.2\unit{mm\;s^{-1}}$. As we will see later (Fig.~\ref{wave}), this unsteadiness is reduced by increasing axial magnetic field \citep{gp71,lw08a}.
196:
197: Figure.~\ref{wave} displays vertical velocities
198: near the outer cylinder in simulations corresponding
199: to the experimental runs of \citet{sgg08} for several values
200: of the toroidal current, $I_\varphi$.
201: A wave pattern very similar to that in the
202: experimental data \citep{sgg08} is seen. Since now there is no jet, the traveling wave is propagating to the bottom endcap and absorbed there while in the old PROMISE experiment, the traveling wave disappears at the jet \citep{lgj07}.
203: %The agreement is remarkably good (Table~\ref{table1}). The system costs around $200\unit{s}$ for the wave pattern to appear in rough accordance with the experimental data with $\mu=0.27$ (The experimental data about this with $\mu=0.26$ is not available yet).
204: We also notice that the perturbation near the upper endcap weakens with strong axial magnetic field. This could be explained by a more stable magnetized residual Ekman layer and Stewartson layer \citep{lw08a}. Both the weakening of the noise sources and disappearance of the amplifying mechanism leads to a rather steady state with $I_{\varphi}=120\unit{A}$.
205:
206: It is highly possible that there is much noise in the real experiment due to some experimental imperfection such as misalignment and in the numerical simulation such as numerical noise. Also the noise could result from physical causes such as the unsteady Ekman layer or Stewartson layer. These noises would cause a noise-sustained convective instability in the system as in \citet{ptk00}. The continuous impulse from the noise sources would have the system always in the state of ``transiently growing" phase (Eq.~\ref{eqn21}). This results in similar wave patterns as the ones from the primary instability without noise, which are observed in PROMISE and PROMISE II experiments and simulations \citep{lgj07}. The noise-induced wave pattern is always susceptible to noise-induced disruption as discussed by \citet{dr87}. That is exactly what we found here and in \citet{lgj07}. We can see this point more clearly by following \citet{lgj07}: performing
207: a simulation that begins with the experimental boundary conditions until
208: the traveling waves are well established, and then switches abruptly
209: to ideal-Couette endcaps (Fig.~\ref{decay}). After the switch, the traveling waves disappear after one axial propagation time and slowly decaying inertial oscillations (asymptotically to zero) result. The main difference in results between \citet{lgj07} and the present simulation are: (1) there is no jet, thus the traveling waves are absorbed near the bottom endcap both before and after the switch; (2) there is no change of wave speed associated with the switch since the background state does not change much before and after the switch. We reach the conclusion that even after the endcaps are split into two rings as in PROMISE II, the wave patterns observed in the experiment are not global instability, but rather noised-sustained convective instability. The similarity between these ``inertial-oscillation-induced wave" after the switch and the earlier noise-sustained ``MRI waves" in the simulation or ``MRI-type waves" observed in the various versions of PROMISE stems from the physical nature of HMRI that HMRI is a weakly destabilized inertial oscillation \citep{lghj06}. More importantly, this similarity supports our conclusion in another aspect: the frequency and wave number selection mechanism for a noise-sustained structure is determined by a linear mechanism, thus resembling the properties of the primary instability.
210:
211: %We have also performed linear calculations by adapting a code developed by \citet{gj02} to allow for a helical field (Table~\ref{table1}). Vertical periodicity is assumed, but the radial equations are
212: %solved directly by finite differences with \emph{perfectly} conducting
213: %boundary conditions \citep{lghj06}. The excellent agreement between this linear calculation and nonlinear simulation stems from the fact that in the bulk flow nonlinear simulation has the same basic state, ideal Couette state as the linear calculation. More importantly, the agreement between the linear calculation, nonlinear simulation and the experiment supports our hypothesis in another aspect: the frequency, wave number selection mechanism for a noise-sustained structure is determined by a linear mechanism, thus resembling the properties of the primary instability. In contrast in the globally unstable regime, a nonlinear eigenvalue problem selects the frequency \citep{ptk00}.
214:
215:
216:
217: \section{Discussion}\label{discussion}
218: In this paper, nonlinear simulations of the helical magnetorotational instability in a magnetized Taylor-Couette flow are performed. The geometry mimics PROMISE II experiment with endcaps split into two rings. The partially conducting boundary condition introduced in \citet{lgj07} is used. The waves patters change with applied magnetic field as in the experiment. However via numerical tests, we find that the wave patterns observed in PROMISE II experiment are not due to a global instability, but rather a noise-sustained convective instability.
219:
220: The importance of the distinction between absolute and convective instability in a bounded system with broken reflection symmetry is discussed. The addition of the toroidal magnetic field breaks the axial symmetry of the system. In such cases, the effects of distant upstream insulating boundaries on the absolute instability differs remarkably from the ones on the convective instability. The insulating endcap would only reduce the growth rate of the absolute instability, but would stabilize the convective instability entirely, however distant it may be. For the absolute instability, the more distant insulating endcap would less reduce the growth rate, while for the convective instability the more distant endcap would only have the system wait longer for the dissipation due to the ``absorption" boundary to dominate. These discoveries cast great obstacles for people to observe the helical magnetorotational instability in the laboratory: An absolute HMRI is needed to observe the global unstable mode in the experiment.
221:
222: Unfortunately it is not easy to derive the critical magnetic Reynolds number $Re_{\rm m,f}$ of the absolute HMRI analytically in a bounded system. However we can get a rough estimate of $Re_{\rm m,a}$, \emph{i.e.}, the critical magnetic Reynolds number of the absolute HMRI in an unbounded system, by wave packet analysis (\S\ref{packet}) and the approximate dispersion relation from Fig.~\ref{fig:global}. From Fig.~\ref{fig:global}, we derive the group velocity $V_g\sim1.08\cm\s^{-1}$, $\gamma^0\sim0.31\s^{-1}$, $k_z^0\sim0.52\cm^{-1}$ and $\sigma\sim2.29\cm^2\s^{-1}$. Therefore $\gamma^0-V_g^2/2\sigma\sim0.05\s^{-1}>0$, which corresponds to an absolute HMRI instability with $Re_{\rm m, a}\sim0.07$. We therefore conjecture that $Re_{\rm m,f}\equiv Re_{\rm m,a}+O(h^{-2})\gtrsim0.07$ in PROMISE II. The critical magnetic Reynolds number is somehow one order of magnitude lower than the standard MRI, but still requires Reynolds number $Re\sim10^5$. Therefore we need to rotate the cylinder typically with more than one hundred $\unit{rpm}$. Such rotation rates are of course achievable, however with such a Reynolds number the advantage of HMRI with much lower Reynolds number, thus much lower end-effects, is not so great as people had expected. Moreover in most HMRI unstable modes $\beta>1$ is preferred, this suggests a toroidal magnetic field typically at $\sim1,000\G$, which requires axial currents $>10^4\unit{A}$ inside the inner cylinder. This is a big engineering challenge in itself. The technical constraints prevent us to try to find the threshold by nonlinear numerical simulations such as: (1)~the current code can not afford large Reynolds number $(\sim10^{5})$, which is required for HMRI to enter the absolutely unstable regime; (2)~from the global linear calculation, the HMRI mode is stabilized if an artificially low Reynolds number like $\sim10^3$, which could be afforded by the current code, is employed.
223:
224: The non-axisymmetric $m=1$ modes are observed in the experiments \citep{sgg06,rhsgg06,sgg07}. Unfortunately since the simulations presented in this paper are all axisymmtric, the possibility to study this important mode is excluded. The extension of the current work to 3D will be the subject of the future study. Ruediger et al have already done some excellent work on this issue and found that given PROMISE parameters the nonaxisymmetric HMRI modes are always harder to be excited than the symmetric mode \citep{rhss05,rs08}.
225:
226: \acknowledgments
227: The author would like to thank Jeremy Goodman and Hantao Ji for their very inspiring discussion and constructive comments. The author would also like to thank James Stone for the advice on the ZEUS
228: code, Stephen Jardin for the advice to implement fully insulating boundary conditions and Frank Stefani for pointing out the distinction between the convective instability and absolute instability in a bounded Taylor-Couette experiment at 2007 APS-DPP annual meeting.
229: This work was supported by the US Department of Energy, NASA
230: under grants ATP03-0084-0106 and APRA04-0000-0152, the
231: National Science Foundation under grant AST-0205903.
232:
233: %\bibliographystyle{apj}
234: %\bibliography{thesis}
235:
236: \begin{thebibliography}{29}
237: \expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi
238:
239: \bibitem[{Balbus \& Hawley(1998)}]{bh98}
240: Balbus, S. \& Hawley, J. 1998, Rev. Mod. Phys., 70, 1
241:
242: \bibitem[{Deissler(1987)}]{dr87}
243: Deissler, R. 1987, Physica D, 25, 233
244:
245: \bibitem[{Gailitis {et~al.}(2008)Gailitis, Gerbeth, Gundrum, Lielausis,
246: Platacis, \& Stefani}]{ggglps08}
247: Gailitis, A., Gerbeth, G., Gundrum, T., Lielausis, O., Platacis, E., \&
248: Stefani, F. 2008, Comtes Rendus Physique, 9, 721
249:
250: \bibitem[{Gilman(1971)}]{gp71}
251: Gilman, P. 1971, Phys. Fluids, 14, 7
252:
253: \bibitem[{Goodman \& Ji(2002)}]{gj02}
254: Goodman, J. \& Ji, H. 2002, J. Fluid Mech., 462, 365
255:
256: \bibitem[{Hayes {et~al.}(2006)Hayes, Norman, Fiedler, Bordner, Li, Clark,
257: ud~Doula, \& Low.}]{hnf06}
258: Hayes, J.~C., Norman, M.~L., Fiedler, R.~A., Bordner, J.~O., Li, P.~S., Clark,
259: S.~E., ud~Doula, A., \& Low., M.-M.~M. 2006, Astrophys, J. Suppl., 165, 188
260:
261: \bibitem[{Hollerbach \& R{\"{u}}diger(2005)}]{hr05}
262: Hollerbach, R. \& R{\"{u}}diger, G. 2005, Phys. Rev. Lett., 95, 124501
263:
264: \bibitem[{Ji {et~al.}(2001)Ji, Goodman, \& Kageyama}]{jgk01}
265: Ji, H., Goodman, J., \& Kageyama, A. 2001, Mon. Not. R. Astron. Soc., 325, L1
266:
267: \bibitem[{Landau \& Lifshitz(1987)}]{ll87}
268: Landau, L.~D. \& Lifshitz, E.~M. 1987, Fluid Mechanics (Butterworth Heinemann)
269:
270: \bibitem[{Liu(2008{\natexlab{a}})}]{lw08a}
271: Liu, W. 2008{\natexlab{a}}, Phys. Rev. E, 77, 056314
272:
273: \bibitem[{Liu(2008{\natexlab{b}})}]{lw08b}
274: ---. 2008{\natexlab{b}}, Astrophys. J., 684, 515
275:
276: \bibitem[{Liu {et~al.}(2006{\natexlab{a}})Liu, Goodman, Herron, \& Ji}]{lghj06}
277: Liu, W., Goodman, J., Herron, I., \& Ji, H. 2006{\natexlab{a}}, Phys. Rev. E,
278: 74, 056302
279:
280: \bibitem[{Liu {et~al.}(2006{\natexlab{b}})Liu, Goodman, \& Ji}]{lgj06}
281: Liu, W., Goodman, J., \& Ji, H. 2006{\natexlab{b}}, Astrophys. J., 643, 306
282:
283: \bibitem[{Liu {et~al.}(2007)Liu, Goodman, \& Ji}]{lgj07}
284: ---. 2007, Phys. Rev. E., 76, 016310
285:
286: \bibitem[{Noguchi {et~al.}(2002)Noguchi, Pariev, Colgate, Beckley, \&
287: Nordhaus}]{npc02}
288: Noguchi, K., Pariev, V.~I., Colgate, S.~A., Beckley, H.~F., \& Nordhaus, J.
289: 2002, Astrophys. J., 575, 1151
290:
291: \bibitem[{Priede \& Gerbeth(2008)}]{pg08}
292: Priede, J. \& Gerbeth, G. 2008, http://arxiv.org/pdf/0810.0386
293:
294: \bibitem[{Proctor {et~al.}(2000)Proctor, Tobias, \& Knobloch}]{ptk00}
295: Proctor, M., Tobias, S., \& Knobloch, E. 2000, Physica D, 145, 191
296:
297: \bibitem[{R{\"{u}}diger {et~al.}(2005)R{\"{u}}diger, Hollerbach, Schultz, \&
298: Shalybkov}]{rhss05}
299: R{\"{u}}diger, G., Hollerbach, R., Schultz, M., \& Shalybkov, D. 2005, Astron.
300: Nachr., 326, 409
301:
302: \bibitem[{R{\"{u}}diger {et~al.}(2006)R{\"{u}}diger, Hollerbach, Stefani,
303: Gundrum, Gerbeth, \& Rosner}]{rhsgg06}
304: R{\"{u}}diger, G., Hollerbach, R., Stefani, F., Gundrum, T., Gerbeth, G., \&
305: Rosner, R. 2006, Astrophys. J., 649, L145
306:
307: \bibitem[{R{\"{u}}diger \& Schultz(2008)}]{rs08}
308: R{\"{u}}diger, G. \& Schultz, M. 2008, Astron. Nachr., 329, 659
309:
310: \bibitem[{{Sisan} {et~al.}(2004){Sisan}, {Mujica}, {Tillotson}, {Huang},
311: {Dorland}, {Hassam}, {Antonsen}, \& {Lathrop}}]{sisan04}
312: {Sisan}, D.~R., {Mujica}, N., {Tillotson}, W.~A., {Huang}, Y., {Dorland}, W.,
313: {Hassam}, A.~B., {Antonsen}, T.~M., \& {Lathrop}, D.~P. 2004, Phys. Rev.
314: Lett., 93, 114502
315:
316: \bibitem[{Stefani {et~al.}(2008)Stefani, Gailitis, \& Gerbeth}]{sgg08}
317: Stefani, F., Gailitis, A., \& Gerbeth, G. 2008, http://arxiv.org/pdf/0807.0299
318:
319: \bibitem[{Stefani {et~al.}(2006)Stefani, Gundrum, Gerbeth, R{\"{u}}diger,
320: Schultz, Szklarski, \& Hollerbach}]{sgg06}
321: Stefani, F., Gundrum, T., Gerbeth, G., R{\"{u}}diger, G., Schultz, M.,
322: Szklarski, J., \& Hollerbach, R. 2006, Phys. Rev. Lett., 97, 184502
323:
324: \bibitem[{Stefani {et~al.}(2007)Stefani, Gundrum, Gerbeth, R{\"{u}}diger,
325: Szklarski, \& Hollerbach}]{sgg07}
326: Stefani, F., Gundrum, T., Gerbeth, G., R{\"{u}}diger, G., Szklarski, J., \&
327: Hollerbach, R. 2007, New J. Phys., 8, 295
328:
329: \bibitem[{Stone \& Norman(1992{\natexlab{a}})}]{sn921}
330: Stone, J. \& Norman, M. 1992{\natexlab{a}}, ApJS., 80, 753
331:
332: \bibitem[{Stone \& Norman(1992{\natexlab{b}})}]{sn922}
333: ---. 1992{\natexlab{b}}, ApJS., 80, 791
334:
335: \bibitem[{Szklarski(2007)}]{sj07}
336: Szklarski, J. 2007, Astron. Nachr., 328, 499
337:
338: \bibitem[{Tobias {et~al.}(1998)Tobias, Proctor, \& Knobloch}]{tpk98}
339: Tobias, S., Proctor, M., \& Knobloch, E. 1998, Physica D, 113, 43
340:
341: \bibitem[{Velikhov {et~al.}(2006)Velikhov, Ivanov, Lakhin, \&
342: Serebrennikov}]{vils06}
343: Velikhov, E.~P., Ivanov, A.~A., Lakhin, V.~P., \& Serebrennikov, K.~S. 2006,
344: Physics Letters A., 356, 357
345:
346: \end{thebibliography}
347:
348:
349:
350: \clearpage
351:
352:
353: %
354: \begin{table}[!htp]
355: \begin{center}
356: \begin{tabular}{c|c|c|c|c}
357:
358: \hline
359: $\Omega_1 h/V_{A,0}$&
360: 20.3 &
361: 40.6 &
362: 81.2 &
363: periodic \tabularnewline
364: \hline
365: \hline
366: Growth Rate $\gamma\;\unit{s^{-1}}$&
367: 0.27&
368: 0.58&
369: 0.82&
370: 1.06\tabularnewline
371: \hline
372:
373: \end{tabular}
374: \caption{\label{reduce} Influence of the height $h$ upon the growth rate $\gamma$ of the absolute instability in a bounded cylinder. $r_{1}=7.1\cm$, $r_{2}=20.3\cm$,
375: $\Omega_{1}=400\rpm$, $\Omega_{2}=53.3\rpm$, $B_{z}=500\G$,
376: $B_\theta(r_1)=1\kG$, the height $h=27.9\;\unit{cm}$, $55.8\;\unit{cm}$ and $111.6\;\unit{cm}$; the material properties are based on
377: gallium: $\eta\approx2000\cm^2\s^{-1}$ and $\rho\approx6\unit{g\;cm^{-3}}$, which give $\Rm=2$ and $S=2.7$, no explicit viscosity present. The simulations are performed using a modified version of the astrophysical code ZEUS2D \citep{sn921,sn922,lgj06,lw08b}. The boundary conditions adopt the one introduced in \S II.D of \citet{lghj06}. Please note that no-slip boundary conditions are employed on all applicable boundaries and ideal Couette state (Eq.~\ref{couette}) is enforced at both endcaps in order to remove the Ekman circulation and possible disturbances induced by this boundary layer effect. The one labeled ``periodic" uses vertically periodic boundary conditions with periodicity length $h=27.9\;\unit{cm}$.}
378: \end{center}
379: \end{table}
380:
381: %
382: % \begin{table}[!htp]
383: % \begin{center}
384: %\begin{tabular}{|c|c|c|c|c|}
385:
386: %\hline
387: %&
388: %Linear Calculation &
389: %Experiment &
390: %Nonlinear Simulation \tabularnewline
391: %\hline
392: %\hline
393: %$f_{\rm wave}/f_{1}$&
394: %$0.145$&
395: %$\sim0.15$&
396: %0.15\tabularnewline
397: %\hline
398: %$\lambda_{\rm wave}$~[cm] &
399: %$10$&
400: %$\sim 11$&
401: %$\sim 11$\tabularnewline
402: %\hline
403: %$v_{p}$[$\unit{mm\;s^{-1}}$]&
404: %0.87&
405: %$\sim 0.9$&
406: %$0.89$\tabularnewline
407: %\hline
408: %$A$[$\unit{mm\;s^{-1}}$]&
409: %unavailable&
410: %$\sim 0.5$&
411: %$\sim 0.8$\tabularnewline
412: %\hline
413: %\end{tabular}
414: %\caption{\label{table1} Comparison of results for
415: %the frequency, wavelength, axial phase speed, and amplitude
416: %obtained from linear calculation, nonlinear simulation and experiment
417: %for the case $I_\varphi=75\unit{A}$. $f_1\equiv\Omega_1/2\pi$ is rotation frequency
418: %of inner cylinder. Note that the linear calculation is done with \emph{perfectly} conducting boundary condition \citep{lghj06}. }
419: %\end{center}
420: %\end{table}
421:
422: \clearpage
423:
424: %
425: \begin{figure}[!htp]
426: \begin{center}
427: \epsscale{0.4}
428: \plotone{f1.eps}
429: \caption{
430: Computational domain for simulations of PROMISE II experiment. Region
431: (I): Inner copper cylinder, angular velocity $\Omega_{1}$.
432: (II): outer copper cylinder,
433: $\Omega_{2}$. (III): liquid gallium; (IV):
434: vacuum. Thick dashed line: insulating inner ring, corotating with the inner cylinder. Thick dash-dot line: insulating outer ring, corotating with the outer cylinder. The junction of these two rings lies at 40\% of the gap ($D=r_2-r_1$) between the inner and outer cylinder \citep{sj07}. Dimensions:
435: $r_{1}=4.0\unit{cm}$; $r_{2}=8.0\unit{cm}$; $h=40.0\unit{cm}$;
436: $d_{wI}=1.0\unit{cm}$; $d_{wII}=1.5\unit{cm}$;
437: $\Omega_{1}/2\pi=3.6\unit{rpm}$; $\Omega_{2}/2\pi=0.936\unit{rpm}$. Note that $\mu=\Omega_2/\Omega_1=0.26$, rather than $\mu=0.27$ used in previous work \citep{sgg06,rhsgg06,lgj07,sgg07}.
438: The exact configuration of the toroidal coils being unavailable to us,
439: six coils (black rectangles) with dimensions as shown were used,
440: with $67$ turns in the two coils nearest the midplane and $72$ in the rest.
441: Currents $I_\varphi$ were adjusted to reproduce the reported %\cite{sgg06}
442: Hartmann numbers $Ha\equiv B_{z}^{0}r_{1}/\sqrt{\rho\mu_{0}\eta\nu}$.
443: \label{boundary} }
444:
445: \end{center}
446: \end{figure}
447:
448: %
449: \begin{figure}[!htp]
450: \epsscale{1.0}
451: \plotone{f2.eps}
452: \caption{\label{fig:global} (a) Growth Rate $\gamma$; (b) Real Frequency $\mathbb{R}\omega$. *~ Linear Calculation, -~ Approximation by Eq.~\ref{approx1} and Eq.~\ref{approx2}. $r_{1}=4.0\cm$, $r_{2}=8.0\cm$,
453: $\Omega_{1}=101.25\rpm$, $\Omega_{2}=26.325\rpm$, $B_{z}=220.5\G$,
454: $\beta=4.0$; the material properties are based on
455: gallium: $\eta=2.43\times10^{3}\cm^2\s^{-1}$, $\nu=3.4\times10^{-3}\cm^2\s^{-1}$ and $\rho=6.35\unit{g\;cm^{-3}}$. The calculations are performed using a code \citep{gj02} adapted to allow for a helical field. Vertical periodicity is assumed, but the radial equations are
456: solved directly by finite differences with perfectly conducting
457: boundary conditions (\S II.B of \citet{lghj06}).}
458: \end{figure}
459:
460: %
461: \begin{figure}[!htp]
462: \begin{center}
463: \epsscale{0.8}
464: \plotone{f3.eps}
465: \caption{~Azimuthal velocity \emph{v.s.} Radius $r$. $Re=1775$, $\beta=3.81$ and $I_{\varphi}=75\unit{A}$. Solid line,~ideal Couette state; $+,~1.31\;\unit{cm}$;
466: $*$,~$2.72\;\unit{cm}$; $\Box,~13.95\;\unit{cm}$. \label{noekman} }
467:
468: \end{center}
469: \end{figure}
470:
471: %
472: \begin{figure}[!htp]
473: \epsscale{1.0}
474: \plottwo{f4a.eps}{f4b.eps}
475: \caption{\label{hydro} Purely hydrodynamic (unmagnetized) simulations. \emph{Left}: Time-averaged poloidal flow stream function $\Psi$; \emph{Right}:~(color) Axial velocities $[\unit{mm\;s^{-1}}]$ versus time and
476: depth sampled at $r=6.5\;\unit{cm}$, for
477: the parameters of the PROMISE II experiment without any magnetic field. Note height increases upward from the bottom endcap.
478: No-slip velocity boundary conditions are imposed at the rigidly rotating
479: endcaps. The steady part of the resulting Ekman circulation is
480: suppressed in right panel by subtracting the time average at each height. }
481: \end{figure}
482:
483: %
484: \begin{figure}[!htp]
485: \begin{center}
486: \epsscale{1.0}
487: \plotone{f5.eps}
488: \caption{
489: (color). Axial velocities $[\unit{mm\;s^{-1}}]$ versus time and
490: depth sampled at $r=6.5\cm$, for
491: the parameters of the PROMISE II experiment %\cite{sgg06}
492: with toroidal currents $I_\varphi$ as marked.
493: No-slip velocity boundary conditions are imposed at the rigidly rotating
494: endcaps. The steady part of the resulting Ekman circulation is
495: suppressed in these plots by subtracting the time average at each height.
496: The waves appear to be
497: absorbed near the bottom endcap.
498: \label{wave} }
499:
500: \end{center}
501: \end{figure}
502:
503: %
504: \begin{figure}[!htp]
505: \begin{center}
506: \epsscale{0.8}
507: \plotone{f6.eps}
508: \caption{
509: (color). An extended version of the case
510: $I_\varphi=75\unit{A}$ shown in Fig.~\ref{wave} but without
511: subtraction of the time average. After $t=360\unit{s}$,
512: the no-slip boundary condition at both endcaps is switched to an
513: ideal Couette profile (Eq.~\ref{couette}). A slowly decayed inertial oscillation is resulted.
514: \label{decay} }
515:
516: \end{center}
517: \end{figure}
518:
519: \end{document}