1: % ****** Start of file apssamp.tex ******
2: %
3: % This file is part of the APS files in the REVTeX 4 distribution.
4: % Version 4.0 of REVTeX, August 2001
5: %
6: % Copyright (c) 2001 The American Physical Society.
7: %
8: % See the REVTeX 4 README file for restrictions and more information.
9: %
10: % TeX'ing this file requires that you have AMS-LaTeX 2.0 installed
11: % as well as the rest of the prerequisites for REVTeX 4.0
12: %
13: % See the REVTeX 4 README file
14: % It also requires running BibTeX. The commands are as follows:
15: %
16: % 1) latex apssamp.tex
17: % 2) bibtex apssamp
18: % 3) latex apssamp.tex
19: % 4) latex apssamp.tex
20: %
21: \documentclass[twocolumn,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
22: %\documentclass[preprint,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
23: %%\documentclass[preprint,preprintnumbers,amsmath,amssymb]{revtex4}
24:
25: % Some other (several out of many) possibilities
26: %\documentclass[preprint,aps]{revtex4}
27: %\documentclass[preprint,aps,draft]{revtex4}
28: %\documentclass[prb]{revtex4}% Physical Review B
29:
30: \usepackage{graphicx}% Include figure files
31: \usepackage{dcolumn}% Align table columns on decimal point
32: \usepackage{bm}% bold math
33: \newcommand{\sech}{\mathrm{sech}}
34: \newcommand{\Ri}{\mathrm{Ri}}
35: \newcommand{\Rea}{\mathrm{Re}}
36: \newcommand{\Ima}{\mathrm{Im}}
37:
38: \topmargin -15pt
39: %\nofiles
40:
41: \begin{document}
42:
43: %\preprint{APS/123-QED}
44:
45: %\title{On the stability of stratified shear flows}% Force line breaks with \\
46: \title{Marginally unstable Holmboe modes}
47:
48: \author{Alexandros Alexakis}
49: % \altaffiliation[Also at ]{Physics Department, XYZ University.}
50: %Lines break automatically or can be forced with \\
51: %\author{Second Author}%
52: % \email{Second.Author@institution.edu}
53: \affiliation{Laboratoire Cassiop\'ee, Observatoire de la C\^ote d'Azur, BP 4229, Nice Cedex 04, France}
54: \affiliation{National Center for Atmospheric Research, Boulder, Colorado, 80305 USA}
55: %\altaffiliation{The National Center for Atmospheric Research is supported
56: %by the National Science Foundation}
57:
58: \date{\today}% It is always \today, today,
59: % but any date may be explicitly specified
60:
61: \begin{abstract}
62: Marginally unstable Holmboe modes for smooth density and velocity
63: profiles are studied. For a large family of flows and stratification
64: that exhibit Holmboe instability, we show that the modes with phase
65: velocity equal to the maximum or the minimum velocity of the shear are
66: marginally unstable.
67: This allows us to determine the critical value of the control
68: parameter $R$ (expressing the ratio of the velocity variation length
69: scale to the density variation length scale) that Holmboe instability appears $R_{crit}=2$.
70: We then examine systems for which the parameter $R$ is very close to this
71: critical value $R_{crit}$.
72: For this case we derive an analytical
73: expression for the dispersion relation of the complex phase speed
74: $c(k)$ in the unstable region. The growth rate and the
75: width of the region of unstable wave numbers has a very strong
76: (exponential) dependence on the deviation of $R$ from the critical value.
77: Two specific examples are examined and the implications of the results
78: are discussed.
79: \end{abstract}
80:
81: %\pacs{Valid PACS appear here}% PACS, the Physics and Astronomy
82: % Classification Scheme.
83: %\keywords{Suggested keywords}%Use showkeys class option if keyword
84: %display desired
85: \keywords{hydrodynamics --- instabilities --- waves}
86:
87: \maketitle
88:
89: \section{Introduction}
90: \label{Intro}
91:
92: Holmboe instability in stratified shear flows appears in a variety
93: of physical contexts such as in astrophysics,
94: the Earth's atmosphere and oceanography
95: \cite{Rosner01,Alexakis04a,Pawlak97,Farmer83,Pettre91,Armi88,Oguz90,Sargent87,Yoshida98}.
96: Although the typical
97: growth rate is smaller than the one of Kelvin-Helmholtz instability
98: it is present for arbitrarily large values of the global Richardson
99: number making Holmboe instability a good candidate for the
100: generation of turbulence and mixing in many physical scenarios.
101:
102:
103: What distinguishes Holmboe from the Kelvin-Helmholtz instability is
104: that unlike the later instability the Holmboe unstable modes have
105: non-zero phase velocity that depends on the wavenumber ({\it i.e.} traveling
106: dispersive modes). It was first identified by Holmboe \cite{Holmboe62}
107: %(see also \cite{Taylor})
108: in a simplified model of a
109: continuous piece-wise linear velocity profile and a step-function
110: density profile. Several authors have expanded Holmboe's theoretical
111: work \cite{Lawrence91,Caulfield94,Haigh99,Ortiz02} by considering different
112: stratification and velocity profiles that do not include the
113: simplifying symmetries Holmboe used in his model. Hazel
114: \cite{Hazel72} and more recently Smyth and Peltier \cite{Smyth89}
115: and Alexakis \cite{Alexakis05} have shown that Holmboe's results
116: hold for smooth density and velocity profiles as long as the length
117: scale of the density variation is sufficiently smaller than the
118: length scale of the velocity variation. Furthermore, effects of
119: viscosity and diffusivity \citep{Nishida87,Smyth90}, non-linear
120: evolution \cite{Smyth88,Smyth91,Sutherland94,Alexakis04c} and mixing properties
121: \cite{Smyth03} of the Holmboe instability have also been
122: investigated. The predictions of Holmboe have also been tested
123: experimentally. Browand and Winant \cite{Browand73} first performed
124: shear flow experiments in a stratified environment under
125: conditions for which Holmboe's instabilities are present. Their
126: investigation has been extended further by more recent experiments
127: \citep{Koop76,Lawrence91,Pawlak99,Pouliquen01,Zhu01,Hogg03,Yonemitsu96,Caulfield95}.
128:
129: Although the understanding of Holmboe instability has progressed a
130: lot since the time of Holmboe, there are still basic theoretical
131: questions that still remain unanswered, even in the linear theory.
132: Most of the work for the linear stage of the instability has been
133: based on the Taylor-Goldstein equation (see \cite{Drazin81}), which describes linear
134: normal modes of a parallel shear flow in a stratified, inviscid,
135: non-diffusive, Boussinesq fluid:
136: %
137: \begin{equation}
138: \frac{d^2\phi}{dy^2}
139: - \left[ k^2 + \frac{U''}{U-c} -\frac{J(y)}{(U-c)^2} \right]\phi=0,
140: \label{TG}
141: \end{equation}
142: where $\phi(y)$ is the complex amplitude of the stream function for
143: a normal mode with real wavenumber $k$.
144: $c$ is the complex phase velocity.
145: %(see \cite{soAndSo} for the effects of barocliniciy).
146: Im$\{c\}>0$ implies instability with growth rate given by
147: $\zeta=k{\mathrm Im}\{c\}$. $U(y)$ is the unperturbed velocity in the $x$
148: direction. $J(y)=-g\rho'/\rho$ is the squared Brunt-V\"ais\"ala
149: frequency where $\rho$ is the unperturbed density stratification and
150: $g$ is the acceleration of gravity. Prime on the unperturbed
151: quantities indicates differentiation with respect to $y$. Equation
152: (\ref{TG}) together with the boundary conditions $\phi \to 0$ for
153: $y\to \pm \infty$, forms an eigenvalue problem for the complex
154: eigenvalue $c$.
155: %
156: Here we just note a few known results for the Taylor-Goldstein equation.
157: %
158: If $c$ is real and in the range of
159: $U$ there is a height $y_c$, at which $U(y_c)=c$. At this height
160: $y_c$, called the critical height, equation (\ref{TG}) has a regular
161: singular point.
162: %
163: For some conditions unstable modes exist with the real part of the
164: phase velocity within the range of $U$. The phase speed of these
165: modes satisfies Howard's semi-circle theorem $|c-1/2
166: (\sup\{U\}+\inf\{U\}) |<1/2\,|\sup\{U\}-\inf\{U\}|$.
167: %
168: If these unstable
169: modes exist the Miles-Howard theorem \cite{Howard61} guarantees that
170: somewhere in the flow the local Richardson number defined by:
171: \begin{equation}
172: \mathrm{Ri}(y)=\frac{J(y)}{[U'(y)]^{\,2}}
173: \end{equation}
174: must be smaller than 1/4.
175:
176:
177: \begin{figure}
178: \includegraphics[width=8cm]{FIG_01.eps}
179: \caption{\label{fig1} Stability diagram for the Hazel model for
180: $R=3$. Panel (a) shows the Kelvin-Helmholtz unstable regions with
181: dark gray, and Holmboe unstable regions with light gray. Panel (b)
182: shows the same instability diagram for larger values of $J_0$.}
183: \end{figure}
184: %Equation \ref{TG} has been shown to give rise to both
185: %Kelvin-Helmholtz and Holmboe unstable modes for a variety of
186: %velocity and density profiles.
187: A typical example used in many studies assumes a velocity profile
188: given by $U(y)=\tanh(y)$ and the squared Br\"unt-Vaisala frequency
189: being given by $J=J_0\cosh(Ry)^{-2}$. Where $J_0$ is the global
190: Richardson number usually defined as $J_0\equiv \mathrm{Ri}(0)$. The
191: case of $R=1$ was examined by Miles \cite{Miles63}
192: %and Howard \cite{Howard61}
193: analytically and numerically by Hazel
194: \cite{Hazel72}. It exhibits only Kelvin-Helmholtz instability for
195: the wavenumbers that satisfy $k(1-k)>J_0$.
196: Hazel \cite{Hazel72} also examined numerically the case $R=5$
197: where it was shown
198: that along with the Kelvin-Helmholtz unstable region there is also a
199: stripe (in a $J_0-k$ diagram) of unstable Holmboe modes. As an
200: example the instability region for $R=3$ is shown in figure
201: \ref{fig1}.
202: %
203: Hazel observed that if $R>2$ there is always a height at which the
204: local Richardson number is smaller than 1/4. Based on this
205: observation Hazel conjectured that $R=2$ is the critical value of
206: $R$ above which the Holmboe instability appears. Later careful
207: numerical examination by Smyth \cite{Smyth89} found unstable Holmboe
208: modes to appears only for values of $R$ larger than $R>2.4$. More
209: recently Alexakis \cite{Alexakis05} showed that the instability can
210: be found for smaller values of $R$ up to $R=2.2$ making the
211: conjecture by Hazel still plausible. More specifically Alexakis
212: \cite{Alexakis05} showed (numerically) for the examined shear and
213: density profiles that the left instability boundary of the Holmboe
214: instability region (see figure \ref{fig1}a) is composed of marginally unstable modes with
215: phase velocity equal to the maximum or the minimum of the shear
216: velocity. Such a condition has been known to hold for
217: smooth velocity and discontinuous density profiles \cite{Alexakis02,Alexakis04b,Churilov05}.
218: Finding these marginally unstable modes corresponds to
219: solving a Schr\"ondiger problem for a particle in a potential well:
220: \begin{equation}
221: \frac{d^2\phi}{dy^2} - \left[ k^2 + V_c(y) \right]\phi=0,
222: \label{Shrondiger}
223: \end{equation}
224: where
225: \begin{equation}
226: \label{potential}
227: V_c(y)=\frac{U''}{U-c} -\frac{J(y)}{(U-c)^2}
228: \end{equation}
229: and $c$ is taken to be
230: $c=U_{\max/\min}$ the maximum or minimum velocity of the shear
231: layer. The right boundary of the Holmboe unstable region on the other hand
232: (see figure \ref{fig1}a) is composed of singular modes with phase velocity
233: within the range of the shear velocity.
234: These modes can be determined by imposing the condition that the
235: solution close to the critical height can be expanded in terms of
236: only one of the two corresponding Frobenius solutions. Furthermore
237: it was shown that for sufficiently large $J_0$ more than one
238: instability stripe exists. These new instability stripes are related
239: with the higher internal gravity modes of the unforced system.
240: Figure \ref{fig1}b shows the instability region for $R=3$ and $J_0$ up to
241: $80$. We note that different unstable Holmboe modes have been found
242: experimentally in \cite{Caulfield95} that were then interpreted in
243: terms of the multi-layer model of \cite{Caulfield94}.
244:
245: The understanding however of the linear part of Holmboe instability
246: for smooth shear and density profiles still remains conjectural and
247: most of the results are based on numerical calculations and do not therefore
248: constitute proofs. We try to address some of these issues in the present work.
249: %
250: In the next section we prove for a general class of velocity
251: profiles that the modes that have phase velocity equal to the
252: maximum/minimum velocity of the shear are marginally unstable.
253: %
254: In section \ref{marginalR} we examine the case for which the
255: parameter $R$ is slightly larger than it's critical value, and the
256: dispersion relation inside the instability region is derived based
257: on an asymptotic expansion.
258: %
259: In section \ref{Examples} we test these results for specific shear and density
260: profiles.
261: %
262: A summary of the results and final conclusions are in the last
263: section. To help the reader, a table \ref{table} with the
264: definitions of all the basic symbols used in this paper has been
265: added.
266:
267:
268: \begin{table}
269: \label{table}
270: \begin{center}
271: \begin{tabular}{|l|l|}
272: \hline
273: Symbol & Definition \\
274: \hline
275: $c$ & Complex phase velocity \\
276: $\zeta$ & $=k{\mathrm{Im}}\{c\}$ the growth rate\\
277: $k$ & Wavenumber \\
278: $\phi(y)$ & Stream function \\
279: $k_0$ & Wavenumber for which $c(k_0)=U_{\infty}$ \\
280: $\phi_0$ & Stream function for the $c(k_0)=U_{\infty}$ mode\\
281: $y_c$ & Location of the critical layer $U(y_c)=c$\\
282: $J_0$ & Global Richardson Number \\
283: $U_\infty,U_*,\alpha$ & Coefficients that appear in the large $y$\\
284: & behavior of $U(y)\simeq U_\infty-U^*e^{-\alpha y}$ \\
285: $\beta,J_*$ & Coefficients that appear in the large $y$\\
286: & behavior of $J(y)\simeq J^* e^{-\beta y}$ \\
287: $\varphi_\infty$,
288: $\lambda$ & Coefficients that appear in the large $y$\\
289: & behavior of $\phi_0(y)\simeq \varphi_\infty e^{-\lambda y}$ \\
290: $J_\infty$, $\sigma$,
291: $R$, $q$ & $J^*/(U^*\alpha)^2$, $\,\,U^*/U_\infty$, $\,\,\alpha/\beta$, $\,\,k/\alpha$ \\
292: $\delta$ & $=R-2$\\
293: $\tilde{J}$ & =$J_\infty(\epsilon c_1/\sigma)^\delta$\\
294: & local Richardson number at $y=y_c$\\
295: $\mu$ & $=1/2-\sqrt{1/4-\tilde{J}}$ \\
296: $\lambda$ & $=\sqrt{q^2 +1 -\tilde{J}}$\\
297: $F(a,b,d,s)$ & Hypergeometric Function \\
298: \hline
299: \end{tabular}
300: \caption{Table of used symbols}
301: \end{center}
302: \end{table}
303:
304:
305: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
306: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
307: \section{Marginal wavenumber}
308: \label{marginalK}
309:
310: In this section we examine modes with phase velocity equal to the
311: maximum/minimum velocity of the flow, and show under what conditions
312: these modes constitute a stability boundary. We start by considering
313: an infinite shear layer specified by the monotonic velocity profile
314: $U(y)$ that has the asymptotic values $U(\pm\infty)=U_{\pm\infty}$.
315: Since the system is Galilean invariant with no loss of generality we
316: can set $U_{+\infty}=-U_{-\infty}=U_{\infty}$. More precisely we
317: will assume that the asymptotic behavior of $U(y)$ for $y\to+\infty$
318: is going to be given by
319: \[
320: U(y) \simeq U_{\infty} - U_* e^{-\alpha y}.
321: \]
322: We further assume that the layer is stably stratified with $J(y)>0$
323: having asymptotic behavior $J(y) \simeq J_* e^{ - \beta y} $ for
324: $y\to+\infty$. In what follows we are going to concentrate only on
325: the modes with phase velocity close to $c=U_{\infty}$; the results
326: can easily be reproduced for the $c=U_{-\infty}$ modes by following
327: the same arguments.
328: To simplify the problem we will
329: non-dimentionalize the equations using the maximum velocity $U_{\infty}$ and the
330: length-scale $\alpha^{-1}$ (essentially setting $U_\infty=1$ and
331: $\alpha=1$). The resulting non-dimensional control parameters for
332: our system are
333: %
334: the asymptotic Richardson number:$J_{\infty}\equiv
335: J_*/(U_*\alpha)^2$,
336: %
337: the ratio of the two velocities:$\sigma \equiv U_*/U_{\infty}$,
338: %
339: and the ratio of the two length scales: $R \equiv \beta/\alpha$. The
340: wavenumber becomes $q=k/\alpha$ and $c$ is measured in units of
341: $U_\infty$.
342:
343: Finally we assume that a solution $\phi_0(y)$ of the Schr\"ondiger
344: problem described in equation \ref{Shrondiger} for the wavenumber $q_0=k_0/\alpha$ exists.
345: Clearly if $R>2$ and $c=1$ the asymptotic behavior of $V_c$ for
346: large $y$ is $V_c(y)\simeq 1$ and $\phi_0(y)$ behaves as
347: $\phi \sim \varphi_\infty e^{-\lambda y}$ with
348: $\lambda=\sqrt{q^2+1}$. If however we have $R = 2$ then $V_c(y)\simeq 1-J_{\infty}$ for
349: $y\to\infty$ and $\lambda = \sqrt{q^2+1-J_{\infty}}$.
350: For abbreviation we will denote for both cases
351: \[\lambda = \sqrt{q^2+1-\tilde{J}}\]
352: where $\tilde{J}$ (that will be defined precisely later on) takes the values
353: $\tilde{J} \simeq 0$ when $R>2$ and $\tilde{J}=J_\infty$ when $R=2$.
354: As discussed in \cite{Alexakis05}, no solution exists that satisfies
355: the boundary conditions for the Schr\"ondiger problem described in
356: equation \ref{Shrondiger} if $R < 2$ since it corresponds in finding
357: bounded eigenstates in an unbounded potential well.
358:
359: Our aim in this section is to find how $c$ changes from the
360: value 1 as we increase $q$ from the value $q_0=k_0/\alpha$. We
361: proceed by carrying out a regular asymptotic expansion by letting
362: $q=q_0+\epsilon q_1$ and $c=1-\epsilon c_1+ \dots $ with
363: $\epsilon\ll1$ and $c_1$ in general complex.
364: However as we deviate from the $c=1$ solution the
365: behavior of the potential $V_c(y)$ drastically changes
366: ($\mathcal{O}(1)$ change) in the large $y$ region and only slightly
367: (linearly with respect to the change in $c$) for
368: $y\simeq{\mathcal{O}}(1)$. Figure \ref{fig2} illustrates this
369: change.
370: \begin{figure}
371: \includegraphics[width=8cm]{FIG_02.eps}
372: \caption{\label{fig2} The potential $V_c$ for the Hazel model with
373: $R=3$ and $J_0=1$ for three values of $c$, $c=1$ (solid line),
374: $c=1+\epsilon$ (dashed line), $c=1-\epsilon$ (dotted line), where
375: $0<\epsilon\ll1$. For $c=1$ $V_c$ asymptotes to the value $V_c\simeq
376: \alpha^2$. This behavior drastically changes when $c\ne1$. }
377: \end{figure}
378: This implies that two different expansions are needed, one for $y$ being of ${\mathcal O}(1)$
379: and one for large $y$.
380:
381: \subsection{Local solution: $y={\mathcal{O}}(1)$}
382:
383: We begin with the local solution and expand $\phi$ as
384: $\phi=\phi_0+\epsilon \phi_1 +\dots\,$ . For $y={\mathcal{O}}(1)$ at
385: first order we obtain:
386: \begin{equation}
387: \frac{d^2\phi_1}{dy^2}-\left[q_0^2 + V_0(y) \right]\phi_1 = \left[2
388: q_1q_0- c_1 V_1(y) \right]\phi_0
389: \end{equation}
390: where $V_0=V_c$ is given in equation \ref{potential} for $c=1$ and
391: \[
392: V_1(y)=\frac{\partial}{\partial c}
393: \left[
394: \frac{U''}{U-c} -\frac{J(y)}{(U-c)^2} \right]
395: \Big{|}_{c=1}
396: \]
397: \[
398: =\frac{U''}{(U-1)^2} -\frac{2J(y)}{(U-1)^3}.
399: \]
400: The solution of this inhomogeneous equation can be found using the
401: Wronskian to obtain:
402: %\[ \phi_1=\]
403: %\begin{equation}
404: %\phi_0(y) \int_{0}^y {\phi^{-2}_0(y')}
405: %\int_{-\infty}^{y'}[2k_1k_0-c_1V_1(y'')]\phi^{2}_0(y'')dy''dy'
406: %\end{equation}
407: %
408: \begin{equation}
409: \phi_1= \phi_0(y) \int_{0}^y \frac
410: {\int_{-\infty}^{y'}[2q_1q_0-c_1V_1(y'')]\phi^{2}_0(y'')dy''}{\phi^{2}_0(y')}dy'
411: \end{equation}
412: %
413: Where we have used as normalization condition $\phi(0)=\phi_0(0)$.
414: Clearly this solution satisfies the boundary condition for $y\to
415: -\infty$. For $y\to +\infty$, by performing the integrations we
416: obtain:
417: %\begin{equation}
418: %\begin{array}{l}
419: %\phi_1 \simeq \\
420: %\phi_0(y) \int_{0}^y \phi^{-2}_0(y')\left[
421: %\int_{-\infty}^{+\infty}[2k_1k_0-c_1V_1(y'')]\phi^{2}_0(y'')dy''
422: %\right]dy'
423: %\nonumber\\
424: %\nonumber\\
425: %\simeq
426: %\frac{2k_1k_0I_1-c_1I_2}{\varphi_\infty} e^{\lambda y}
427: %\end{array}
428: %\end{equation}
429: \begin{equation} \phi_1 \simeq
430: \frac{2q_1q_0I_1-c_1I_2}{\varphi_\infty} e^{\lambda y}
431: +{\mathcal{O}}(e^{-\lambda y})
432: \end{equation}
433: where $ I_1=\int_{-\infty}^{+\infty}\phi^{2}_0dy>0$ and
434: $I_2=\int_{-\infty}^{+\infty}V_1\phi^{2}_0dy.$
435: %Note that $|I_2|<\infty$ for $R>2$, but $|I_2|<\infty$ only if $\lambda>1/2$ for
436: %$R=2$.
437: %
438: So the large $y$ behavior of $\phi$ based on the local solution
439: is given by:
440: \begin{equation}
441: \phi \simeq \varphi_\infty e^{-\lambda y} +
442: \epsilon \frac{2q_1q_0I_1-c_1I_2}{\varphi_\infty} e^{\lambda y}
443: + \dots \,\,\,.
444: \end{equation}
445:
446: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
447: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
448: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
449: \subsection{Far away solution: $y={\mathcal O}(\ln[1/\epsilon])$}
450:
451: To capture the large $y$ behavior we need to make the change of
452: variables $\tilde{y}=y-y_c$ where $y_c=-\ln(\epsilon c_1/\sigma)$ is
453: the location of the singularity determined by $U(y_c)=c$. (Note that
454: for $c_1$ complex $\tilde{y}$ does not coincide with the real $y$ axis.) The
455: Taylor-Goldstein equation \ref{TG} then reads:
456: \begin{equation}
457: \frac{d^2\phi}{d\tilde{y}^2}-\left[q^2 - \frac{1}{ e^{\tilde{y}}-1}
458: - \frac{J_\infty \cdot
459: \left(\frac{\epsilon c_1}{\sigma}\right)^{\delta} \cdot
460: \big( e^{-\tilde{y}}\big)^{\delta}}
461: { (e^{\tilde{y}}-1)^2}
462: \right]\phi=0
463: \end{equation}
464: where $\delta=R-2$ and
465: only the leading terms have been kept.
466: %exponentially small terms have been dropped out.
467: Introducing the variable $s=e^{-\tilde{y}}$ we obtain:
468: \begin{equation}
469: \label{Faraway1} s^2\frac{d^2}{ds^2}\phi + s\frac{d}{ds}\phi -
470: \left[q^2 - \frac{s}{1-s} - \frac{ \tilde{J} s^2\,\,s^{\delta} } {
471: (1-s)^2} \right]\phi=0,
472: \end{equation}
473: where $\tilde{J}=J_\infty \cdot
474: \left({\epsilon c_1}/{\sigma}\right)^{\delta}$
475: gives the Richardson number at the critical height. Note that if $R=2$ ($i.e.$ $\delta=0$)
476: then $\tilde{J}=J_\infty= {\mathcal{O}}(1)$. If however $R>2$ then
477: $\tilde{J}\ll1$ and the term in the brackets proportional to
478: $\tilde{J}$ is small and can be neglected everywhere except close to
479: the singularity $s=1$. To deal with this small singular term we can
480: write for $s$ close to 1: $ s^{\delta}\simeq 1-\delta (1-s) +\dots $
481: and keep the leading term. That way the principal term inside the
482: brackets is always kept for all values of $s$ and our solution will
483: be correct to first order for all values of $R\ge2$ by solving:
484: \begin{equation}
485: s^2\frac{d^2}{ds^2}\phi + s\frac{d}{ds}\phi - \left[q^2 -
486: \frac{s}{1-s} - \frac{ \tilde{J} s^2 } { (1-s)^2} \right]\phi=0.
487: \label{Faraway}
488: \end{equation}
489: %Equation \ref{Faraway} is valid for all values of $\beta\le 2\alpha$ up to first order.
490: To deal with the singularities at $s=0$ and $s=1$ we
491: make the substitution $\phi = s^{q}(1-s)^{\mu} h(s)$ with
492: $\mu=1/2-\sqrt{1/4-\tilde{J}}$. We then obtain:
493: \[s(1-s)\frac{d^2}{ds^2}h + [(2q+1)-(2\mu+2q+1)s] \frac{d}{ds}h +\]
494: \begin{equation}
495: [1-\mu-(q+1)]h=0,
496: \label{hyperg}
497: \end{equation}
498: the solution of which is the Hypergeometric function
499: $h(s)=F(a,b,d;s)$ with:
500: \[{ a} =(\mu+q)+\sqrt{q^2+1-\tilde{J}},\]
501: \[{ b} =(\mu+q)-\sqrt{q^2+1-\tilde{J}},\]
502: and
503: \[{d} =(2q + 1).\]
504: Note that $q+\mu-a=-\lambda$ and $q+\mu-b=+\lambda$. Some basic
505: properties of the Hypergeometric function are given in appendix A,
506: here we give just some of the resulting asymptotic behavior of $\phi$:
507: \begin{eqnarray}
508: \lim_{s\to 0 } \phi & \simeq & s^{q} = e^{-qy} \\
509: \lim_{s\to +\infty} \phi & \simeq &
510: \frac{\Gamma(d)\Gamma(b-a)}{\Gamma(b)\Gamma(d-a)}s^{q}(-s)^{\mu-a} \nonumber \\
511: &+&
512: \frac{\Gamma(d)\Gamma(a-b)}{\Gamma(a)\Gamma(d-b)}s^{q}(-s)^{\mu-b}.
513: \end{eqnarray}
514: Returning to the $y$ variable and up to a normalization factor $A$
515: we have for $y\ll y_c$ that
516: \begin{equation}
517: \phi\simeq A\left[e^{-\lambda y} +
518: (-\epsilon c_1)^{2\lambda}
519: \frac{\Gamma(a)\Gamma(d-b)\Gamma(-2\lambda)}
520: {\Gamma(b)\Gamma(d-a)\Gamma( 2\lambda)}
521: e^{\lambda y}\right].
522: \end{equation}
523:
524: %%%%%%%%%%%%%%%%%%%%%
525: \subsection{Matching}
526: Matching the exponentially decreasing terms of the
527: local and the far-away
528: solution we obtain: $ A=\varphi_\infty$,
529: and from the exponentially increasing terms we have:
530: \begin{equation}
531: \epsilon \frac{2q_1q_0I_1-c_1I_2}{\varphi_\infty}=
532: \varphi_\infty (-\epsilon c_1)^{2\lambda}
533: \frac{\Gamma(a)\Gamma(d-b)\Gamma(-2\lambda)}
534: {\Gamma(b)\Gamma(d-a)\Gamma( 2\lambda)}.
535: \end{equation}
536: We can solve the equation above iteratively by letting
537: $\epsilon c_1=\epsilon c'_1+\epsilon^{2\lambda}c'_2+\dots\,\,\, $. To first order we
538: obtain:
539: \begin{equation}
540: c'_1=\frac{2q_1q_0I_1}{I_2},
541: \label{c1}
542: \end{equation}
543: which gives the first correction to the phase speed and determines
544: if the real part of phase speed is increasing or decreasing with the wavenumber.
545: If for example $I_2>0$ (which will be the case in the examples that follow)
546: then $c$ is decreasing with $q$ and the correction $c_1'$
547: is positive for positive $q_1$ and negative for negative $q_1$.
548: The opposite holds if $I_2<0$.
549: %%
550: From now on we will assume that $I_2>0$ which is the physically relevant
551: case (Re$\{c(k)\}$ being a decreasing function of $k$) if however there is
552: a velocity and density profile such that $I_2<0$ the same results will hold but
553: for the opposite direction in $q$ ($i.e.$ wavenumbers smaller than $q_0$ will be unstable
554: and wavenumbers larger than $q_0$ will be stable).
555: %%
556: The $c_1'$ correction however is real, and contains no information
557: about the growth rate. At the next order we have
558: \begin{equation}
559: -c'_2I_2/\varphi_{_\infty} = \varphi_\infty (-c'_1)^{2\lambda}
560: \frac{\Gamma(a)\Gamma(d-b)\Gamma(-2\lambda)}
561: {\Gamma(b)\Gamma(d-a)\Gamma( 2\lambda)}.
562: \label{expan1}
563: \end{equation}
564: This correction is much smaller but contains the first order
565: correction of the imaginary part of $c$. The dispersion relation of
566: $c$ for $q$ close to $q_0$ can then be written in terms of $q$ as:
567: \[
568: c=1 - \frac{2q_0I_1}{I_2} (q-q_0) \nonumber + \qquad\]
569: \begin{equation}
570: \frac{ {\varphi_{_\infty}}^2}{I_2} \left(\frac{-2k_0I_1}{I_2}
571: (q-q_0) \right)^{2\lambda}
572: \frac{\Gamma(a)\Gamma(d-b)\Gamma(-2\lambda)}
573: {\Gamma(b)\Gamma(d-a)\Gamma( 2\lambda)}
574: \label{disp}
575: \end{equation}
576: Special care is needed to interpret the term $(-c_1')^{2\lambda}$
577: for $c_1'$ is given by equation \ref{c1}.
578: When $q_1<0$, $c_1'$ is negative and the term $(-c_1')^{2\lambda}$ is real, this
579: corresponds to the case that $c$ becomes larger than the shear velocity
580: and no critical layer is formed.
581: The Howard semi-circle theorem then
582: guaranties stability.
583: %If for example $I_2>0$ then $c$ is increasing as we decrease $q$,
584: %as a result
585: This proves that
586: wavenumbers slightly smaller than $q_0$ are stable.
587: When $q_1>0$, $c_1'$ is positive and $(-c_1')^{2\lambda}$ becomes a complex number
588: that can take different values depending on whether the minus sign is
589: interpreted as $e^{i\pi}$ or $e^{-i\pi}$. The choice depends on the
590: location of the singularity on the complex plane when we integrate the
591: Taylor Goldstein equation \ref{TG}.
592: %
593: If Im$\{c_1\} >0 $ then $(-c_1)^{2\lambda}$ should be interpreted as $|c_1|^{2\lambda}\,\,
594: e^{ -i 2\lambda \pi}$ because the integration is going over the singularity.
595: %
596: If Im$\{c_1\} <0 $ then $(-c_1)^{2\lambda}$ should be interpreted as $|c_1|^{2\lambda}\,\,
597: e^{ +i 2\lambda \pi}$ because the integration is going under the singularity.
598: %and as $|c_1|^{2\lambda}\,\, e^{- i 2\lambda \pi}$ if
599: %Im$\{c_1\}<0$.
600: %
601: Here we arrive at an important point in our derivation: the sign of
602: the imaginary part of $c$ based on equation \ref{disp} depends on
603: the original assumption about the sign of Im$\{c\}$ when we
604: integrate across the singularity.
605: Thus, in order for the matching to be successful
606: we need to verify that the original assumption about the sign of
607: Im$\{c_1\}$ is consistent with the final result.
608: If we assume that Im$\{c_1\}>0$ then from \ref{expan1} we have that
609: \begin{equation}
610: 0< {\mathrm{Im}}\{c_1\}=
611: \sin(2 \lambda \pi)
612: |c_1|^{2\lambda} \frac{\varphi_\infty^2}{I_2}
613: \frac{\Gamma(a)\Gamma(d-b)\Gamma(-2\lambda)}
614: {\Gamma(b)\Gamma(d-a)\Gamma( 2\lambda)}
615: \label{sign}
616: \end{equation}
617: where Im$\{(c_1')^{2\lambda}\}$ is written as $-\sin(2 \lambda \pi)
618: |c_1|^{2\lambda}$ as previously discussed.
619: The matching is successful \underline{only if} the sign of
620: the right hand side ($r.h.s.$) of equation \ref{sign} is positive as originally
621: assumed and only then is the dispersion relation \ref{disp} valid.
622: (We arrive at the same condition if we initially
623: assume that Im$\{c\}<0$). It is shown in appendix B that for $R> 2
624: $ ({\it i.e.} $\tilde{J}\simeq0$) the $r.h.s.$ of \ref{sign} is always positive and the matching is
625: successful. For the special case however that $R= 2$ (i.e. $\tilde{J} = J_\infty$)
626: the matching is not always successful
627: because the product $\Gamma(b)\Gamma(d-a)$ that appears in equation \ref{sign} can change sign
628: depending on the value of $\tilde{J}$. In particular it is shown in
629: the appendix that if $\tilde{J} > 2q/(2q+1)^2$ the $r.h.s.$ of
630: equation \ref{sign} is negative and thus we end up with a
631: contradiction. Therefore in the $R=2$ case we have shown instability only if
632: \begin{equation}
633: \label{cond} \tilde{J} < 2q/(2q+1)^2.
634: \end{equation}
635: The unsuccessful matching when the condition \ref{cond} is not satisfied
636: suggests that the modes with $q>q_0$ are part of the continuous spectrum
637: of the Taylor Goldstein equation. This kind of modes
638: have a discontinuity
639: of the first derivative of $\phi$ at the critical layer
640: and have been studied before in the literature \cite{Banks76,Balmforth95}.
641: We need to emphasize here that the lack of instability at this order does not imply
642: stability. Non-zero growth rate of smaller order can still exist and
643: therefore the above result should be interpreted only as a
644: sufficient condition for instability.
645:
646: To summarize we have shown that if $R>2$ the modes with phase velocity equal to the maximum
647: phase velocity of the shear (when they exist, and for density and velocity profiles that
648: satisfy the conditions stated at the beginning of this section) are marginally unstable:
649: wavenumbers with $q<q_0$ are stable and wavenumbers with $q>q_0$ are unstable.
650: If $R=2$ these modes are marginally unstable only if the condition \ref{cond} is further
651: satisfied and stable (to the examined order) otherwise.
652:
653: \section{Marginal R}
654: \label{marginalR}
655:
656: In the last sectioned we showed marginal instability when the wavenumber $q$
657: is varied from the critical value $q_0$.
658: However the wavenumber is not a
659: control parameter in a system. We would like therefore to examine a
660: system for which one of the control parameters ($J_0$ or $R$) is
661: close to the critical value for which the instability begins. Since
662: Holmboe instability is present for arbitrary large values of $J_0$
663: the only other control parameter left is $R$. We consider therefore
664: a case for which $R=2 +\delta$ with $0<\delta\ll1$ and the $c=1$
665: solution ($\phi_0,q_0)$ is known with $q_0$ such that $J_\infty>
666: 2q_0/(2q_0+1)^2$ so that the $R=2$ case gives no instability at the
667: examined order. We make a small variation in $q=q_0+\epsilon q_1$
668: and $c=1-\epsilon c_1$ with the exact relation between $\delta$ and
669: $\epsilon$ still undetermined. At this stage we assume that
670: $\epsilon$ is sufficiently smaller than $\delta$ so that the
671: procedure in the previous section is still valid and then gradually
672: increase its value until the approximations in the previous section
673: start to fail. As we increase the value of $\epsilon$ the most
674: sensitive term (in $\epsilon$) that will be affected first, is the
675: term proportional to $\tilde{J}=J_\infty \cdot \left({\epsilon
676: c_1}/\sigma\right)^{\delta}$ in the equation \ref{Faraway} for
677: which $\epsilon$ is raised to the smallest appearing power. Note
678: that if $\epsilon \ll \exp[-1/\delta]$ then $\tilde{J}\ll1$ and the
679: results of the previous section are still valid. If however
680: $\epsilon \sim {\mathcal O}(\exp[-1/\delta])$ then
681: $\tilde{J}\sim{\mathcal O}(1)$.
682: %The resulting equation is still \ref{Faraway} and
683: %since $s^\delta$ in equation \ref{Faraway1} can be
684: %written as $s^\delta\simeq 1+\delta\ln(s)$.
685: Following the same steps as in the previous section
686: we end up in the dispersion relation given by equation \ref{disp}
687: but like in the $R=2$ case $\tilde{J}$ cannot be treated as a small
688: parameter.
689:
690: The difference from the $\delta={\mathcal O}(1)$ case will therefore
691: appear when we try to determine the sign of the $r.h.s.$ of equation
692: \ref{sign}. To have successful matching we need to satisfy the
693: condition \ref{cond}. Since $\tilde{J}$ is finite the condition
694: $\tilde{J} < 2q/(2q+1)^2$ that also appears in the $R=2$ case could
695: be violated. To capture the whole unstable region we define
696: $\epsilon$ such that $ J_\infty\epsilon^{\delta}=2q_0/(1+2q_0)^2 $
697: or
698: \begin{equation}
699: \label{epsilon}
700: \epsilon=\left[\frac{2q_0/J_\infty}{(1+2q_0)^2}\right]^{1/\delta}\ll1.
701: \end{equation}
702: Note that the term inside the brackets is always smaller than one.
703: For such a choice the condition \ref{cond} for instability reads
704: \[
705: \tilde{J}=J_\infty \cdot
706: \left(\frac{\epsilon c_1}{\sigma} \right)^{\delta}=
707: \]
708: \[
709: \left[\frac{2q_0}{(1+2q_0)^2}\right] [1+\delta \ln(c_1/\sigma)+
710: \mathcal{O}(\delta^2)]
711: \]
712: \begin{equation}
713: < \left[\frac{2q_0}{(1+2q_0)^2}\right] + \mathcal{O}(\epsilon)
714: \end{equation}
715: or $c_1/\sigma<1$. Already at this stage it can be seen that we have
716: instability only if $c_1= 2q_1q_0I_1/I_2<\sigma$ and therefore
717: the instability is confined in the region of wave numbers
718: \begin{equation}
719: q_0 < q < q_0+\Delta q
720: \end{equation}
721: where $\Delta q= \epsilon \sigma I_2/2q_0 I_1$.
722: Therefore, the second
723: instability boundary for the Holmboe instability
724: is given by $q+\Delta q$.
725: To get the full
726: dispersion relation in this asymptotic limit we need to expand in
727: terms of $\delta$ the product $\Gamma(d-a)\Gamma(b)$ that appears in
728: equation \ref{disp} since this is the term that can change sign
729: depending on the value of $\tilde{J}$. This is done in Appendix B
730: and the resulting growth rate inside the instability region
731: to the first non-zero order becomes:
732: \begin{equation}
733: \label{growth} \zeta=q_0{\mathrm{Im}}\{c\}=-{\delta C_1q_0}
734: |q_0-q|^{2\lambda}\ln\left(\frac{2(q-q_0)q_0I_1}{\epsilon
735: I_2\sigma}\right)
736: \end{equation}
737: where $C_1>0$ is an ${\mathcal O}(1)$ quantity and is given in equation \ref{const}.
738: The maximum of the growth rate is obtained for $q-q_0=\epsilon e^{-1/2\lambda}I_2\sigma/(2q_0I1)$
739: with the growth rate being given by:
740: \begin{equation}
741: \max{[\zeta]}=\delta \epsilon^{2\lambda} \frac{C_1q_0\lambda}{2\lambda e} \left[ \frac{I_2 \sigma}{q_0I_1}\right]^{2\lambda}
742: \end{equation}
743: Therefore the growth rate scales like $\epsilon^{2\lambda}$
744: and the width of the instability region scales like $\Delta q\sim \epsilon$.
745: In terms of $\delta$ these relations are given by
746: $\zeta \sim \delta e^{-2\lambda\gamma/\delta}$ and $\Delta q\sim e^{-\gamma/\delta}$
747: where $\gamma$ is a positive constant.
748: This very strong dependence with $\delta$ suggests that both $\zeta$
749: and $\Delta q$ decrease very rapidly as $\delta$ becomes smaller.
750: This can explain the difficulty numerical codes have,
751: when attempting to calculate
752: growth rate for values of $R$ very close to $R=2$.
753:
754:
755: \section{Examples}
756: \label{Examples}
757:
758: In the previous sections we showed some general results for the
759: Holmboe unstable modes. In this section we examine some specific
760: examples often used in the literature to model Holmboe's
761: instability. We begin with the model introduced by Hazel that we
762: briefly mentioned in the introduction. The model assumes a velocity
763: profile given by $U(y)= \tanh(y)$ and a density stratification
764: determined by $-g\rho'/\rho=J_0\cosh^{-2}(Ry)$. Based on the
765: definitions given in section \ref{marginalK} we have that $\alpha=2$,
766: $\beta=2R$, $U_*=2$ and $J_*=4J_0$. The resulting non-dimensional
767: quantities are $J_\infty=J_0/4$, $\sigma=2$, $q=k/2$ and $R$ has the
768: same meaning.
769: %
770: This model satisfies all the conditions that are stated in section \ref{marginalK}
771: therefore for $R>2$ the modes with $c(k)=\pm 1$ are marginally unstable.
772: %
773: %Based on the results of section \ref{marginalK} for
774: %$R>2$ the modes with $c(k)=\pm 1$ are marginally unstable.
775: %This model satisfies all the conditions that are stated in section \ref{marginalK}
776: %and therefore we have instability if $R>2$.
777: %
778: Furthermore for the case $R=2$ we will have instability only if the
779: condition \ref{cond} is satisfied, or in the units of this example
780: if $J_0/4<k/(k+1)^2$. A simple numerical integration shows that this
781: is not the case for this profile (see figure \ref{fig3}).
782: \begin{figure}
783: \includegraphics[width=8cm]{FIG_03.eps}
784: \caption{\label{fig3} The $c=1$ solutions for the Hazel model $J_0(k)$
785: for $R=2$ (solid line),
786: the condition from equation \ref{cond} $J_0<4k/(k+1)^2$ (dashed line).}
787: \end{figure}
788: Therefore, the $R=2$ is stable (to the examined order) and
789: is the critical value beyond which the Holmboe
790: instability begins. The imaginary part of $c(k)$ for this profile for the case
791: that $R=2.1$ and $J_0=1.2$ is shown if figure \ref{fig4} where the
792: numerical result is compared with the asymptotic expansion of
793: equation \ref{growth}. Although $\delta=0.1$ is not very small
794: there is satisfactory agreement (a 20\% difference)
795: between the asymptotic and the numerical result.
796: It is worth mentioning that it is very hard to find a range of values of $\delta$
797: that both the asymptotic result is valid and Im$\{c\}$ is large enough
798: to be captured by a numerical code. Note that by decreasing the value of $\delta$
799: from $0.1$ to $0.05$ has resulted in a drop of Im$\{c\}$ by three orders of magnitude.
800: \begin{figure}
801: \includegraphics[width=8cm]{FIG_04.eps}
802: \caption{\label{fig4}
803: The Im$\{c\}$ for the Hazel model for $R=2+\delta$ with $\delta=0.1,\,0.75,\, 0.05$
804: and $J_0=1.2$.
805: The diamonds indicate the results from numerical integration for the $\delta=0.1$ case. }
806: \end{figure}
807:
808:
809: A second family of flows we will examine assumes a velocity profile
810: given by $U(y)=\tanh(y)$ as in the Hazel model and the density
811: stratification being determined by:
812: \[-g\rho'/\rho=\frac{J_0}{\cosh^{2R}(y)}.\]
813: The advantage of this profile is that there are analytic solutions
814: for the Kelvin-Helmholtz stability boundaries $J_0(k)$ for the cases
815: that $R=0,1$ and $2$. In addition there is an analytic solution $J_0(k)$
816: for the modes $k$ for which $c(k)=1$ for the $R=2$ case. The
817: $R=0$ case was examined by \cite{Drazin58} where it was shown that the
818: Kelvin Helmholtz unstable modes satisfy $J_0<k^2(1-k^2)$. The $R=1$ case
819: (that reduces to the $R=1$ case of the Hazel model) was investigated by
820: \cite{Miles63}
821: where it was shown that the Kelvin Helmholtz unstable modes satisfy
822: $J_0<k(1-k)$. The $R=2$ case has not been investigated before
823: (to the author's knowledge). One can show
824: following the same methods used for the $R=0,1$ cases \cite{Drazin58,Miles63,Howard63}
825: that the $c(k)=0$ modes satisfy:
826: \[
827: J_0=\frac{k(1-k)(2+k)(3+k)}{4(k+1)^2}
828: \]
829: with $\phi(y)=[1-\tanh(y)^2]^{k/2}\cdot[\tanh(y)]^{1/4-\sqrt{1/4-J_0}}$
830: and provide the Kelvin-Helmholtz instability boundary.
831: $c(k)=1$ modes on the other hand that are of interest for the Holmboe
832: instability satisfy
833: \[J_0=\frac{k(3+2k)}{(k+1)^2}\]
834: for $k<1$.
835: The stream-function $\phi$
836: for these modes is
837: given by $\phi=[1+\tanh(y)]^{k/2} \cdot [1-\tanh(y)]^{\sqrt{k^2/4+1-J_0}}$.
838: The Kelvin-Helmholtz stability
839: boundaries for the three cases $R=0,1,2$ along with the $c=1$
840: solutions for the $R=2$ case are shown in figure \ref{fig5}.
841: \begin{figure}
842: \includegraphics[width=8cm]{FIG_05.eps}
843: \caption{\label{fig5} The $c=1$,and $c=0$ solutions for the model
844: with density stratification given by $J(y)=J_0\cosh^{-2R}(y)$
845: for $R=0$ (dotted line), $R=1$ (dotted line), $R=2$ (solid line).}
846: \end{figure}
847: For this example we also have that $\sigma=2$ and $q=k/2$ but
848: $J_\infty=2^{2R-4}J_0$.
849: The $J_0(k)$ relation for the $c=1$ solutions does not
850: satisfy the criterion \ref{cond} that now reads
851: $J_0 < k/(k+1)^2$, thus the $R=2$ case is stable (to the examined order) and is the
852: critical value above which the Holmboe instability begins.
853: Because $J_\infty$ is four times bigger than in the Hazel model (for the same $J_0$)
854: the resulting growth rate is smaller by a factor of $4^{-2\lambda/\delta}$
855: that is close to $10^{-14}$ for the $\delta=0.1$ case.
856: Figure \ref{fig6} shows the growth rate based on the asymptotic expansion \ref{growth}.
857: No numerical results could be obtained for this case for values of $\delta$ smaller than
858: $\delta \le 0.1$ that would justify a comparison with the asymptotic expansion.
859: \begin{figure}
860: \includegraphics[width=8cm]{FIG_06.eps}
861: \caption{\label{fig6}
862: The Im$\{c\}$ for the $J(y)=J_0\cosh^{-2R}(y)$ model for $J_0=1.2$ and $R=2+\delta$
863: with $\delta=0.1$ (solid line) and $\delta=0.075$ (dashed line).}
864: \end{figure}
865: This example, when compared with one of Hazel, clearly demonstrates the sensitivity of the resulting growth rate
866: to the large $y$ asymptotic behavior of $J(y)$ and $U(y)$: a change by a factor of $4$ in $J_\infty$
867: resulted in 14 orders of magnitude difference in the growth rate.
868:
869: \section{Conclusions}
870: In this paper we have examined analytically Holmboe's instability
871: for smooth density and velocity profiles. We have shown
872: for a large family of flows that the
873: modes with phase velocity equal to the maximum or minimum of the
874: unperturbed velocity profile when they exist and if the parameter
875: $R$ is above the critical value ($R_{crit}=2$) they constitute a
876: stability boundary. This result confirms the results obtained
877: numerically in \cite{Alexakis05} where the fact that the
878: $c=U_{\max/\min}$ modes are marginally unstable was only conjectured
879: based on physical arguments. It is also the first time shown
880: analytically that the value of $R=2$ for the Hazel model is the
881: critical value $R_{crit}$ above which the Holmboe instability begins.
882:
883: For the case that the parameter $R$ is only slightly larger than
884: it's critical value $R_{crit}=2$, the dispersion relation $c(k)$ was
885: obtained based on an asymptotic expansion. For this marginally
886: unstable flow the growth rate $\zeta$ as well as the width
887: of the instability stripe $\Delta q$ have a very strong dependence on the
888: deviation of $R$ from it's critical value. In particular the
889: growth rate $\zeta$ and the width of the instability $\Delta
890: k=$ scale as $\exp[-2\lambda \gamma/(R-R_{crit})]$ and $\exp[-\gamma/(R-R_{crit})]$ respectively
891: (for some positive constant $\gamma$). For this reason the
892: numerical investigations performed in the past
893: \cite{Smyth89,Alexakis05} were not able to capture the instability
894: for values of $R$ very close to $R_{crit}$.
895:
896: We believe also that the present results go beyond the clarifying of
897: a mathematical detail in the literature. They clearly demonstrate
898: the nature of the Holmboe instability in a quantitative way.
899: The critical layer for
900: Holmboe unstable modes appears at large heights where the shear
901: rate can overcome stratification and the behavior mode will strongly
902: depend on the properties of the shear at this height. The critical
903: layer is then coupled to the gravity wave modes that
904: their properties are determined by $y={\mathcal{O}}(1)$ quantities
905: and are fast enough to travel with the
906: same velocity as the velocity at the critical layer height. It is
907: this coupling that gives rise to the instability, and it is restricted
908: only to the wave numbers that result in a phase speed that is smaller than
909: the maximum velocity of the shear but big enough so that the critical layer
910: is at a large enough height so that $Ri(y)<1/4$.
911: Therefore in
912: any experimental setup, precise measurements of the
913: velocity and density stratification are needed in
914: both large and small heights in order to make
915: comparisons of the measured growth rate and the theoretical
916: predictions.
917:
918: Finally we believe that the results given in this paper can provide a basis
919: for further numerical and analytical investigations such as an examination of the
920: weakly non-linear theory where a small non-linearity is taken into account in order to
921: examine the long time evolution of an unstable mode beyond the linear stage.
922:
923:
924: \begin{acknowledgments}
925: Most part of this work was done while the author was at ASP post-doc
926: at the National Center for Atmospheric Research (NCAR) and the
927: support is greatly acknowledged. Present support by the
928: Observatory of Nice (Observatoire de la C\^ote d'Azur)
929: that helped the author finish this work is also acknowledged.
930: \end{acknowledgments}
931:
932: \appendix
933: \section{The Hyper-geometric equation: Basic Properties}
934:
935: The Hypergeometric equation is:
936: \begin{equation}
937: z(1-z)\frac{d^2f}{dz^2} + [d-(a+b+1)z]\frac{df}{dz}-abf=0.
938: \end{equation}
939: The solution that remains finite as $z\to0$ is the Hypergeometric
940: function: $f=F(a,b,c;z)$. For the normalization condition we are
941: using we have the following limits:
942: \begin{eqnarray*}
943: \lim_{z\to0} F(a,b,c;z)&= &1/\Gamma(d)\\
944: \lim_{z\to1} F(a,b,c;z)&\simeq&
945: \frac{\Gamma(d)\Gamma(d-a-b)}{\Gamma(d-a)\Gamma(d-b)}\\
946: &+ &\frac{\Gamma(d)\Gamma(a+b-d)}{\Gamma(a)\Gamma(b)}(1-z)^{d-a-b}\\
947: \lim_{z\to+\infty}F(a,b,c;z)&\simeq&
948: \frac{\Gamma(d)\Gamma(b-a)}{\Gamma(b)\Gamma(d-a)}(-z)^{-a}\\
949: &- & \frac{\Gamma(d)\Gamma(a-b)}{\Gamma(a)\Gamma(d-b)}(-z)^{-b}
950: \end{eqnarray*}
951: provided that $d\ne0,-1,-2,\dots$ and $a-b$ is not an integer.
952:
953:
954: \section{The sign of the instability term}
955:
956: To determine whether we have successful matching or not we need to
957: find the sign of the imaginary part in the dispersion relation
958: \ref{disp}. We examine each term separately. Clearly $\Gamma(a)$,
959: $\Gamma(2\lambda)$ and $\Gamma(d-b)$ are all positive factors since
960: the argument of the $\Gamma$-function is positive. The factor
961: $\Gamma(-2\lambda)$ is changing sign every time $2\lambda$ is an
962: integer. However its product with $\sin(2 \lambda \pi)$ is always
963: remains negative. The factors $\Gamma(b)$ and $\Gamma(d-a)$ however
964: can change sign depending on value of $\tilde{J}$. Using the
965: expressions for $a,b,d$ one can show that $-1\le b \le 0$ if
966: $\tilde{J} \le (2q)/(2q+1)^2$ or if $2q \le 1$,
967: and positive otherwise. Similarly we have that $-1\le d-a < 0$ if
968: $\tilde{J} > (2q)/(2q+1)^2$ and $2q < 1$ and
969: non-negative otherwise. Combining these two inequalities we can
970: determine the sign of the product
971: \[\Gamma(b)\Gamma(d-a) \le 0 {\mathrm{\, if\, and\, only\, if \,}}
972: \tilde{J} \le (2q)/(2q+1)^2. \]
973: The result in \ref{sign} then follows.
974:
975: To find the dispersion relation for the small $\delta$ and
976: $\epsilon$ given by \ref{epsilon} we need to find an expression for
977: the term $\Gamma(b)\Gamma(d-a)$. Substituting the choice of
978: $\epsilon$ given by \ref{epsilon} in the expression for $b$ and
979: $d-a$
980: and using $\tilde{J}=J_\infty (\epsilon c_1/\sigma)^\delta\simeq
981: J_\infty (\epsilon)^\delta (1+\delta \ln(c_1/\sigma))$
982: we have that to first order in $\delta$ if $2q_0>1$:
983: \[
984: b\simeq \frac{1}{2} \delta J_\infty\epsilon^\delta \ln(c_1/\sigma)
985: \left[
986: \frac{1}{\sqrt{1/4-J_\infty\epsilon^\delta}}+\frac{1}{ \sqrt{1+q_0-J_\infty\epsilon^\delta}}
987: \right]
988: \]
989: and $b={\mathcal{O}}(1)$ if $2q_0<1$.
990: Similarly,
991: \[
992: d-a\simeq
993: \frac{1}{2} \delta J_\infty\epsilon^\delta \ln(c_1/\sigma)
994: \left[
995: \frac{1}{\sqrt{1/4-J_\infty\epsilon^\delta}}+\frac{1}{ \sqrt{1+q_0-J_\infty\epsilon^\delta}}
996: \right]
997: \]
998: if $2q<1$ and $d-a={\mathcal{O}}(1)$ if $2q>1$. Using the
999: $\Gamma$-function property $\Gamma(\delta)=\Gamma(1+\delta)/\delta$
1000: we can write the dispersion relation for $q_0 < q< q_0 + \epsilon
1001: \sigma I_2/k_0 I_1$ as:
1002: \[
1003: c=1-2(q-q_0)q_0I_1/I_2+
1004: \delta C_1 (q_0-q)^{2\lambda}\ln\left(\frac{2(q-q_0)q_0I_1}{I_2\sigma}\right)
1005: \]
1006: where
1007: \[
1008: C_1=\frac{J_\infty \epsilon^\delta {\varphi_{_\infty}}^2}{2I_2}
1009: \left(\frac{2q_0I_1}{I_2}\right)^{2\lambda}
1010: \frac{\sin(2\lambda)\Gamma(a)\Gamma(d-b)\Gamma(-2\lambda)}
1011: {\Gamma(w)\Gamma( 2\lambda)} \times
1012: \]
1013: \begin{equation}
1014: \label{const}
1015: \left[
1016: \frac{1}{\sqrt{1/4-J_\infty\epsilon^\delta}}+\frac{1}{ \sqrt{1+q_0-J_\infty\epsilon^\delta}}
1017: \right]
1018: \end{equation}
1019: with $w=b$ if $2q_0<1$ and $w=d-a$ if $2q_0>1$.
1020:
1021:
1022: %\appendix
1023: %\section{Appendixes}
1024: %This signals that all following section commands refer to appendixes
1025: %instead of regular sections.
1026: \begin{thebibliography}{}
1027:
1028:
1029:
1030: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1031: %APL ASTRO
1032: \bibitem[Rosner, Alexakis, Young, Truran \& Hillebrand (2002)]{Rosner01}
1033: { R. Rosner, A. Alexakis, Y. Young, J. Truran \&
1034: W. Hillebrand}, ``On the C/O enrichment of novae ejecta".
1035: {Astrophys. J.\/} {\bf 562}, L177--L179 (2002).
1036:
1037: \bibitem[Alexakis (2004)]{Alexakis04a}
1038: { A. Alexakis, A.C. Calder, A. Heger, E.F. Brown, L.J. Dursi, J.W. Truran, R. Rosner, D.Q. Lamb, F.X. Timmes,
1039: B. Fryxell, M. Zingale, P.M. Ricker, K. Olson},
1040: ``On heavy element enrichment in classical novae"
1041: {Astrophys. J. \/} {\bf 603}, 931--937 (2004).
1042:
1043:
1044: \bibitem[ Farmer \& Freeland (1983)]{Farmer83}
1045: {D. M. Farmer \& H. Freeland},
1046: ``The physical oceanography of fjords,''
1047: {Prog. Oceanogr.\/} {\bf 12}, 147 (1983).
1048:
1049:
1050: \bibitem[Pawlak \& Armi (1997)]{Pawlak97}
1051: {G. Pawlak \& L. Armi}, ``Hydraulics of two-layer arrested wedge flows,''
1052: {J. Hydraul. Res.\/} {\bf 35}, 603 (1997).
1053:
1054: %APL GEO
1055: \bibitem[Armi \& Farmer (1988)]{Armi88}
1056: { L. Armi \& D.M. Farmer,}
1057: ``The flow of Mediterranean water through the Strait of Gibraltar".
1058: { Prog. Oceanogr. \/} {\bf 21}, 1--105 (1988).
1059:
1060: %APL GEO
1061: \bibitem[Oguz, Ozsoy, Latif, Sur, \& Unluata (1990)]{Oguz90}
1062: {T. Oguz, E. Ozsoy, M.A. Latif, H.I. Sur \& U. Unluata,}
1063: ``Modeling of hydraulically controlled exchanged flow in the Bosphorous Strait".
1064: { J. Phys. Oceanogr. \/} {\bf 20}, 945--965 (1990).
1065:
1066:
1067: %APL GEO
1068: \bibitem[Sargent \& Jirka (1987)]{Sargent87}
1069: { F.E. Sargent \& G.H. Jirka,}
1070: ``Experiments on saline wedge".
1071: { J. Hydraulic Eng. \/} {\bf 113}, 1307--1324 (1987).
1072:
1073: %APL GEO
1074: \bibitem[Yoshida, Ohtani, Nishida \& Linden (1998)]{Yoshida98}
1075: { S. Yoshida, M. Ohtani, S. Nishida \& P.F. Linden,}
1076: ``Mixing processes in a highly stratified river".
1077: {\it Physical Processes in Lakes and Oceans}, edited by J. Imberger
1078: (American Geophysical Union, Washington DC, 1998)
1079:
1080: \bibitem[Pettre \& Andre (1991)]{Pettre91}
1081: {P. Pettre \& J. C. Andre},
1082: ``Surface-pressure change through Loewe's phenomena
1083: and katabatic flow jumps. Study of two cases in Adelie Land Antartica,''
1084: {J. Atmos. Sci.\/} {\bf 48}, 557--571 (1991).
1085:
1086:
1087: \bibitem[Holmboe (1962)]{Holmboe62}
1088: {J. Holmboe,}
1089: ``On the behavior of symmetric waves in stratified shear layers".
1090: { Geophys. Publ. \/} {\bf 24}, 67--113 (1962).
1091:
1092: %LIN EXP
1093: \bibitem[Lawrence, Browand \& Redecopp (1991)]{Lawrence91}
1094: {G.A. Lawrence, F.K. Browand \& L.G. Redecopp,}
1095: ``The stability of a sheared density interface".
1096: { Phys. Fluids-A.\/} {\bf 3}, 2360--2370 (1991).
1097:
1098:
1099: %LTH
1100: \bibitem[Caulfield (1994)]{Caulfield94}
1101: {C.P. Caulfield,}
1102: ``Multiple linear instability of layered stratified shear flow".
1103: {J. Fluid Mech. \/} {\bf 258}, 255--285 (1994).
1104:
1105: %LTH
1106: \bibitem[Haigh \& Lawrence (1999)]{Haigh99}
1107: {S.P. Haigh \& G.A. Lawrence,}
1108: ``Symmetric and non-symmetric Holmboe instabilities in an inviscid flow".
1109: { Phys. Fluids \/} {\bf 11}, 1459--1468 (1999).
1110:
1111: %LTH
1112: \bibitem[Ortiz, Chomaz\& Loiseleux (2002)]{Ortiz02}
1113: {S. Ortiz, J.M. Chomaz \& T. Loiseleux}
1114: ``Spatial Holmboe instability".
1115: {Phys. Fluids \/} {\bf 14}, 2585-2597 (2002).
1116:
1117: %LTH
1118: \bibitem[Hazel (1972)]{Hazel72}
1119: {S.P. Hazel,}
1120: ``Numerical studies of the stability of inviscid shear flows".
1121: { J. Fluid Mech. \/} {\bf 51}, 3261--3280 (1972).
1122:
1123: %LTH
1124: \bibitem[Smyth \& Peltier (1989)]{Smyth89}
1125: { W.D. Smyth, \& W.R. Peltier,}
1126: ``The transition between Kelvin-Helmholtz and Holmboe instability;
1127: An investigation of the over-reflection hypothesis".
1128: { J. Atmos. Sci. \/} {\bf 46}, 3698--3720 (1989).
1129:
1130:
1131: %LTH
1132: \bibitem[Alexakis (2005)]{Alexakis05}
1133: { A. Alexakis}
1134: ``On Holmboe's instability for smooth shear and density profiles".
1135: { Phys. Fluids \/} {\bf 17}, 084103 (2005).
1136:
1137:
1138:
1139: %LTH
1140: \bibitem[Nishida \& Yoshida (1987)]{Nishida87}
1141: { S. Nishida \& S. Yoshida},
1142: ``Stability and eigenfunctions of disturbances in stratified two layer shear flow".
1143: {\it Proc. Third Intl. Symp. on Stratified Flows, Pasadena, California, 3-5 February 1987} pp.28--34
1144:
1145: %LIN
1146: \bibitem[Smyth, \& Peltier (1990)]{Smyth90}
1147: {W.D. Smyth \& W.R. Peltier,}
1148: ``Three-dimensional primary instabilities of a stratified dissipative, parallel flow".
1149: { Geophys. Astrophys. Fluid Dyn. \/} {\bf 52}, 249--261 (1990).
1150:
1151: %SIM
1152: \bibitem[Smyth, Klaasen \& Peltier (1988)]{Smyth88}
1153: {W.D. Smyth, G.P. Klaasen \& W.R. Peltier,}
1154: ``Finite amplitude Holmboe waves,".
1155: { Geophys. Astrophys. Fluid Dyn.. \/} {\bf 43}, 181--222 (1988).
1156:
1157: %SIM
1158: \bibitem[Smyth, \& Peltier (1991)]{Smyth91}
1159: {W.D. Smyth \& W.R. Peltier,} ``Instability and transition in finite
1160: amplitude Kelvin-Helmholtz and Holmboe waves". { J. Fluid Mech. \/}
1161: {\bf 228}, 387--415 (1991).
1162:
1163:
1164: %SIM
1165: \bibitem[Sutherland, Caulfield \& Peltier (1994)]{Sutherland94}
1166: {B.R. Sutherland, C.P. Caulfield \& W.R. Peltier, }
1167: ``Internal gravity generation and hydrodynamic instability".
1168: { J. Atmos. Sci. \/} {\bf 51}, 3261--3280 (1994).
1169:
1170:
1171: \bibitem[Alexakis (2004)]{Alexakis04c}
1172: {A. Alexakis, A.C. Calder, L.J. Dursi, R. Rosner, J.W. Truran, B. Fryxell, M. Zingale, F.X. Timmes, K. Olson, P. Ricker}
1173: ``On the nonlinear evolution of wind-driven gravity waves"
1174: {Phys. Fluids\/} {\bf 16}, 3256--3268 (2004).
1175:
1176:
1177: \bibitem[Smyth, \& Winters (2003)]{Smyth03}
1178: {W.D. Smyth \& K.B. Winters,}
1179: ``Turbulence and mixing in Holmboe waves".
1180: { J. Phys. Oceanogr.. \/} {\bf 33}, 694--711 (2003).
1181:
1182: %EXP
1183: \bibitem[Browand \& Winant (1973)]{Browand73}
1184: {F.K. Browand \& C.D. Winant}
1185: ``Laboratory observations of shear layer instability in a stratified fluid."
1186: { Boundary Layer Met. \/} {\bf 5}, 67--77 (1973).
1187:
1188: %EXP
1189: \bibitem[Koop (1976)]{Koop76}
1190: {C.G. Koop,} ``Instability and turbulence in a stratified shear
1191: layer". { Tech. Rep. USCAE {\bf 134} Department of Aerospace
1192: Engineering University of South California.} (1976)
1193:
1194: %EXP
1195: \bibitem[Pawlak (1999)]{Pawlak99}
1196: {G. Pawlak \& L. Armi},
1197: ``Vortex dynamics in a spatially accelerating shear layer,''
1198: {J. Fluid Mech.\/} {\bf 376}, 1-- (1999).
1199:
1200: %EXP
1201: \bibitem[Pouliquen, Chomaz \& Huerre (1994)]{Pouliquen01}
1202: {O. Pouliquen, J.M. Chomaz \& P. Huerre,}
1203: ``Propagating Holmboe waves at the interface between two immiscible fluids".
1204: { J. Fluid Mech.\/} {\bf 266}, 277--409 (1994).
1205:
1206: %EXP
1207: \bibitem[Zhu \& Lawrence (2001)]{Zhu01}
1208: {D.Z. Zhu \& G.A. Lawrence,}
1209: ``Holmboe's instability in exchange flows".
1210: { J. Fluid Mech.\/} {\bf 429}, 391--409 (2001).
1211:
1212: %EXP
1213: \bibitem[Hogg \& Ivey (2003)]{Hogg03}
1214: {A.M. Hogg \& G.N. Ivey,}
1215: ``The Kelvin-Helmholtz to Holmboe instability transition in stratified exchange flows".
1216: { J. Fluid Mech.\/} {\bf 477}, 339--362 (2003).
1217:
1218: %EXP
1219: \bibitem[Yonemitsu, Swaters, Rajaratnam \& Lawrence (1996)]{Yonemitsu96}
1220: {N. Yonemitsu, G.E. Swaters,N. Rajaratnam \& G. A. Lawrence}
1221: ``Shear instabilities in arrested salt wedge flows".
1222: { Dyn. Atm. Oceans\/} {\bf 24}, 173--182 (1996).
1223:
1224:
1225: %EXP
1226: \bibitem[Caulfield (1995)]{Caulfield95}
1227: {C.P. Caulfield, W.R. Peltier, S. Yoshida \& M. Ohtani}
1228: ``An experimental investigation of the instability of a shear flow with multilayered dencity stratification".
1229: {Phys. Fluids\/}{\bf 7},3028--3041 (1995).
1230:
1231: %EXP
1232: \bibitem[ Drazin \& Reid (1981)]{Drazin81}
1233: {P.G. Drazin \& W.H. Reid}
1234: ``{\it Hydrodynamic Stability"}
1235: Cambridge University Press (1981)
1236:
1237: %LTH
1238: \bibitem[Howard (1961)]{Howard61}
1239: { L.N. Howard,}
1240: ``A note on a paper of John Miles".
1241: { J. Fluid Mech.\/} {\bf 10}, 509--512 (1961).
1242:
1243: \bibitem[Miles (1963)]{Miles63}
1244: {J. Miles}
1245: ``On the stability of heterogeneous shear flow Part 2".
1246: { J. Fluid Mech.\/} {\bf 16}, 209--227 (1963).
1247:
1248: %LTH
1249: \bibitem[Alexakis, Young \& Rosner (2002)]{Alexakis02}
1250: { A. Alexakis, Y. Young \& R. Rosner} ``Shear instability of
1251: fluid interfaces: a linear analysis". { Phys. Rev. E.\/} {\bf 65},
1252: 26313 (2002).
1253:
1254: %LTH
1255: \bibitem[Alexakis, Young \& Rosner (2004)]{Alexakis04b}
1256: { A. Alexakis, Y. Young \& R. Rosner,} ``Weakly non-linear analysis
1257: of wind driven gravity waves". { J. Fluid Mech.\/} {\bf 503},
1258: 171--200 (2004).
1259:
1260:
1261:
1262: %LTH
1263: \bibitem[Churilov (2005)]{Churilov05}
1264: {S.M. Churilov} ``Stability analysis of stratified shear
1265: flows with a monotonic velocity profile without inflection
1266: points"
1267: {J. Fluid Mech. \/}{\bf 539}, 25--55 (2005).
1268:
1269: %LTH
1270: \bibitem[Banks, Drazin \& Zaturska (1976)]{Banks76}
1271: {W.H.H. Banks, P.G. Drazin \& M.B. Zaturska}
1272: ``On the normal modes of parallel flow of inviscid stratified fluid".
1273: { J. Fluid Mech.\/} {\bf 75}, 149--171 (1976).
1274:
1275:
1276: %LTH
1277: \bibitem[Balmforth (1995)]{Balmforth95}
1278: {N.J. Balmforth \& P.J. Morrison}
1279: ``Normal modes and continuous spectra"
1280: {Annals of the New York Academy of Science\/} {\bf 773} 80--94 (1995)
1281:
1282: \bibitem[Drazin (1958)]{Drazin58}
1283: {P.G. Drazin}
1284: ``The stability of a shear layer in an unbounded heterogeneous inviscid fluid".
1285: {J. Fluid Mech.\/}{\bf 4},214--224 (1958).
1286:
1287: \bibitem[Howard (1963)]{Howard63}
1288: { L.N. Howard,}
1289: ``Neutral curves and stability boundaries in stratified shear flow".
1290: { J. Fluid Mech.\/} {\bf 16}, 333--342 (1963).
1291:
1292:
1293: %%%LTH
1294: %\bibitem[Helmholtz(1868)]{Helmholtz68}
1295: %{ H. Helmholtz,}
1296: %``Ueber discontinuirliche Fl\"{u}ssigkeitsbewegungen".
1297: %{ Wissenschaftliche Abhandlungen\/} {\bf 3}, 146 (1868).
1298:
1299: %%%LTH
1300: %\bibitem[Lord Kelvin(1910)]{Kelvin10}
1301: %{ Lord Kelvin},
1302: %``Influence of wind and capillarity on waves in water supposed frictionless".
1303: %{ Mathematical and Physical papers {\bf iv} Hydrodynamics and general Dynamics.\/} 76, (1910).
1304:
1305: %LTH
1306:
1307: \end{thebibliography}
1308:
1309: \newpage
1310: %
1311: %\begin{figure}
1312: %\includegraphics[width=8cm]{FIG_01.eps}
1313: %\caption{\label{fig1} The local Richardson number $Ri(y)$ normalized by $J_0$
1314: %for different values of the parameter $R$}
1315: %\end{figure}
1316:
1317:
1318: \end{document}
1319: %
1320: % ****** End of file apssamp.tex ******
1321: