RMI Zabuskey 1999

P1: APR/NBL/spd/mbg P2: NBL/plb/ary QC: NBL
November 17, 1998 18:37 Annual Reviews AR075-14

Annu. Rev. Fluid Mech. 1999. 31:495536
Copyright c 1999 by Annual Reviews. All rights reserved
VORTEX PARADIGM FOR
ACCELERATED INHOMOGENEOUS
FLOWS: Visiometrics for the
Rayleigh-Taylor and Richtmyer-Meshkov
Environments
Norman J. Zabusky
Laboratory for Visiometrics and Modeling, Department of Mechanical and Aerospace
Engineering, and Computer Aids for Industrial Productivity (CAIP) Center, Rutgers
University, Piscataway, New Jersey 08855; e-mail: nzabusky@caip.rutgers.edu
KEY WORDS: accelerated ows, vortex dynamics
ABSTRACT
We illustrate how cogent visiometrics can provide peak insights that lead to path-
ways for discovery through computer simulation. This process includes visualiz-
ing, quantifying, and tracking evolving coherent structure morphologies. We use
the vortex paradigm(Hawley &Zabusky 1989) to guide, interpret, and model phe-
nomena arising in numerical simulations of accelerated inhomogeneous ows,
e.g. Richtmyer-Meshkov shock-interface and shock-bubble environments and
Rayleigh-Taylor environments. Much of this work is available on the Internet at
the sites of my collaborators, A Kotelnikov, J Ray, and R Samtaney, at our Vizlab
URL, http://vizlab.rutgers.edu/vizlab.html.
The next El Dorado lies in the mathematical understanding of coherent structures
in weakly dissipative uids: the formation, evolution, and interaction of metastable,
vortex-like solutions of nonlinear partial differential equations. By juxtaposing ex-
perimental results with computer simulations, I hope to focus on essential analytical
directions.
(ZABUSKY1977)
495
0066-4189/99/0115-0495$08.00
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

496 ZABUSKY
1. INTRODUCTION
Nonlinear processes are usually unstable, chaotic, and turbulent. Often they
are unpredictable beyond early time eras, and laminar (or smooth) features
evolve into a multiplicity of space scales (sometimes with complex topologies)
and singular and near-singular space-time events. To acquire a deeper under-
standing of special phenomena, those that are considered by the community of
scientists to be the contemporary primary focus of attention, we nd it useful
to formulate low-degree-of-freedom models. These models are often derived
from insights gained by juxtaposing many-degree-of-freedom direct numerical
simulations (dns) with laboratory and eld experiments. If the latter experi-
ments are not available, then care must be taken to elucidate the accuracy and
content of the dns massive data sets, e.g. the convergence of the discretized and
regularized numerical simulations to solutions of the continuum formulation.
A practical goal is to obtain efcient numerical models that aid the prediction
and design of natural and man-made processes.
I will relate a few of my experiences with simulations of the Fermi-Pasta-
Ulam nonlinear lattice and the Korteweg de Vries (KdV) partial differential
equation in one space dimension that led to the discovery of the soliton (Zabusky
& Kruskal 1965; Zabusky 1981, 1984; Weissert 1997; Heyerhoff 1998). These
will illustrate the visiometric approach, namely how the use of well-chosen
visualizations and quantications aided in posing and answering key questions.
I will showthe robustness of this modus operandi by reviewing recent work that
emphasizes the vortex dynamics of accelerated inhomogeneous ows. In par-
ticular, I discuss progress on applying the vortex paradigm (Hawley & Zabusky
1989) to the Richtmyer-Meshkov (RM) and the Rayleigh-Taylor (RT) insta-
bility environments where inhomogeneous density ows are accelerated. The
vortex-related insights have led us recently to conjecture that some astrophys-
ical objects, called knots, that are seen in the vicinity of astrophysical jets and
supernova remnants may be vortex projectiles (Zabusky et al 1997; Zabusky
et al, submitted), as described below.
2. VISIOMETRICS AND REDUCED MODELING
2.1 Visiometrics
By visiometrics, I mean the visualization, quantication, and juxtaposition of
data sets (Bitz & Zabusky 1990). In quantication, I imply the extraction
and tracking of coherent structures above a low-lying and more random (less
coherent) background. These data sets are obtained from laboratory and eld
experiments, remote observations of natural phenomena, and computer simu-
lations. In this paper, we emphasize direct numerical simulations (dns) of the
compressible Euler equations of uid motion.
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

VORTEX PARADIGM AND VISIOMETRICS FOR RT & RM 497
By juxtaposition, I mean visualized and quantied comparisons among
overlaid data sets. These comparisons are of similar or different functions
at the same or different times from simulations of the same or linked math-
ematical models in a relevant domain of parameters. An example is an in-
compressible conguration that models a corresponding low-Mach-number
RM conguration. If one is tracking coherent structures, it is often useful to
be in a translating and/or rotating frame with only a subset (zoomed) of the
total space (Winkler et al 1987; Zabusky et al 1992, 1993; Silver & Zabusky
1993).
For nonlinear phenomena, one must be aware, particularly beyond early time
epochs in a simulation, of the effects due to specic forcing functions and noise
and the explicit and code-implicit dissipative and dispersive regularizations (e.g.
for very small scales the turbulence models, de-aliasing and ltering functions,
etc). This informationwill allowa validationof candidate mathematical models.
In essence, visiometrics is the way to mathematical metaphors, the mother lode
referred to in the opening quotation (Zabusky 1977).
2.1.1 VISUALIZATION By visualization, I mean the conversion of numerical
data into a variety of images. First, we preprocess or lter the data (to reduce
noise or size) and categorize it by choosing appropriate color maps, opacity
parameters, contour levels, etc. It can then be transformed to an appropri-
ate geometric model for display. Standard visualization techniques include
continuous-tone contour maps in two dimensions; volume rendering a function
in three dimensions [as it would appear when radiating or reecting light from
various sources (Upson & Keeler 1988), etc]; displaying isosurfaces (contours)
from connected polygons that bound regions of functions extracted by thresh-
olding (Lorensen & Cline 1987) or wavelets (Siegel & Weiss 1997, Farge et al
1998); drawing icons or hedgehogs (arrows) for vector elds and drawing lines
for streamlines, vector eld lines, etc.
A visualization community has arisen from diverse workers in computer
graphics, computer vision, computational geometry, image processing, mor-
phological mathematics, and applications people in every discipline (Hes-
selink 1988, Jahne 1991, Patrikalakis 1991). Many new journals, special jour-
nal issues, and conference proceedings report on environments such as AVS,
EXPLORER, IDL, etc. For example, see the proceedings of the ninth IEEE
Visualization Conference, which took place in October of 1998. Many of
these are now listed on Internet homepages. An excellent place to start is:
http://www.nas.nasa.gov/RNR/Visualization/annotatedURLs.html.
2.1.2 QUANTIFICATION Visualization alone is not enough. Intuition to guide
the development of new models requires the quantication of evolving mor-
phologies or structures. This process begins with structure (or region) isolation,
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

498 ZABUSKY
Table 1 Glossary for interacting uid dynamical coherent structures
Interact
Structure Move Come together Come apart
bubble mushroom advect accrete bifurcate
blast wave packet convect aggregate breakup
blob patch diffuse agglomerate burst
cloud point disperse align disassemble
critical pt. ring ow bind disrupt
depletion roll hop collapse nger
dipole separatrix intensify condense ssion
dislocation sheet migrate entangle phase-shift
eddy shock wave percolate ssion plow
lament soliton propagate focus reconnect
nger spike spike fold reect
gradient spiral stream fuse split
gyre stem swirl merge striate
hairpin striation transport pair strip
helix surface wind roll-up
hole tendril wind-about
interface triple-point
jet tube
knot vortex
layer whorl
extraction, and tracking. Each structure is in a data base with its diagnosed
properties as described below.
In Table 1, a list of some uid dynamical structures (or features or objects),
actions, andinteractions is presented. The move andinteract categories describe
how the objects may evolve. (Some of these verbs may serve as nouns in the
object column.)
Some diagnostic operations are:
1. Global quantication: measuring total quantities, some of which are invari-
ant in some region of parameter space. This includes energy, circulation,
enstrophy, palenstrophy, probability distribution functions (pdf), helicity,
drag force vector on an object, correlation functions, slopes and variances
of pdf and power density spectra, fractal dimension, moments of statistical
quantities (e.g. skewness, kurtosis, etc), etc.
2. Detailed quantication: Characterizing space-time topologies of coherent
structures, including evolution of critical and extremal points and surfaces
of scalar, vector, and tensor elds in appropriate moving frames (e.g. the rate
of strain tensor and its diagonal representation at every point in the eld,
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

or in projections onto a plane or interface, etc); space-time or parameter-
time projections of functions and their gradients, e.g. space-time diagrams
of transverse integrated vorticity (below); slopes, normals, curvatures, and
torsions of lines and surfaces; nite-time-interval particle trajectories and
Lyapounov exponents; low-order moment (e.g. ellipse and ellipsoidal rep-
resentations) skeletal descriptions of elongated objects; distances between
extrema (or centroids) and critical points; temporal variation of modal am-
plitudes and phases; wavelet extraction of coherent objects and a statistical
characterization of the remainder eld.
Consider the coherent vortex structure (CVS), a localized region (sheet, layer
or tube) of uid that persists for many periods of some relevant time interval.
We believe that in an optimal denition of a CVS, one will be able to track a
relatively compact object that contains a signicant content (for vortices, the
circulation or enstrophy) for the longest time interval. For example, when a
CVSis distorting due to the presence of a strong strain eld of nearby CVSs, one
should be able to track it until a complex interaction conguration among these
CVSs is formed and then afterward when another set of outgoing coherent
vortex states appears.
For a discussion of two-dimensional and three-dimensional feature identi-
cation, extraction, and tracking, see Carlbom et al (1992), Samtaney et al
(1994), and Silver & Wang (1996). An important aspect of segmentation is the
functional form of the variables available from a simulation that are used to
dene the object (Buntine & Pullin 1989, Villasenor & Vincent 1992, Banks
& Singer 1994, Silver 1995, Jeong et al 1997). In many of these denitions,
the features are dened with some sort of connectibility criteria that enable the
method to partition the data set into important regions and background.
2.2 Reduced Modeling and the Soliton
I propose a tentative denition: A reduced model is one in a hierarchy that
yields solutions to a cogent problem and minimizes an error-cost function.
The models in the hierarchy may be linked by means of mathematical asymp-
totics and cogent means within a narrow but physically useful parameter range
(which includes the evolutionary time interval). The error-cost function could
be the product (solution error) (computational cost). The total error is a
combination of several partial errors, including omission or inadequate rep-
resentation of realistic physical and chemical processes, roundoff, numerical
truncation, ltering (or regularization of small scales), etc. The computational
cost relates to algorithm (truncation accuracy and efciency), parallelization
process, and storage and retrieval procedures (for massive data sets that are
post-processed, etc).
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

500 ZABUSKY
For example, an explicit formula is usually good for steady, idealized pro-
cesses, e.g. the Kolmogoroff spectral indices for homogeneous, isotropic, forced,
incompressible turbulence. A set of linear simultaneous equations without and
with constraints may be the next level. This is followed in complexity by cou-
pled sets of linear ordinary, partial, or integro-differential equations (e.g. linear
stability theory) and single or coupled sets of nonlinear ordinary, partial, or
integro-differential equations.
2.2.1 SOLITONS IN DISCRETE AND CONTINUOUS MEDIA A paradigm for visio-
metrics and asymptotical modeling is the discovery of the soliton by Zabusky
& Kruskal (1965). The soliton and its remarkable properties could have been
discovered earlier, because certain key experimental observations on small-
but nite-amplitude shallow water waves were made by Scott Russell in the
1840s and were known to Korteweg and de Vries (KdV) in 1895, who using
asymptotic methods, derived the single nonlinear partial differential equation
u
t
+u
p
u
x
+
2
p
3
u
x
3
= 0, (1)
with p = 1 to model observed phenomena (Korteweg & de Vries 1895). They
found stationary isolated and periodic solutions. This (1 +1) nonlinear partial
differential equation models many situations, including the one-dimensional
cubic nonlinear lattice, as derived by M Kruskal (Zabusky 1995, Kruskal 1995)
to model the FPU simulations. Direct numerical simulations of the KdV equa-
tion were performed with Gary Deem in 1965 at Bell Labs in Whippany, New
Jersey, and the near recurrence of the solution to its initial state was visualized
cinematically (Deem et al 1975).
We studied the waveforms and space-time diagrams of the solution, u(x, t ),
and realized that the initial condition disassembled itself into localized coherent
structures that looked like the solitary wave solutions when they were well sep-
arated; we dubbed themsolitons. What we didnt realize was that this system
is mathematically integrable and the solitons are exact mathematical entities
whose number is determined by the parameters of the problem. This exact-
ness was realized when the key breakthrough was made by Gardner, Greene,
Kruskal, and Miura in 1967 (Gardner et al 1967), who transformed the nonlinear
evolution equation to the linear Schr odinger equation in one dimension.
The application of the soliton concept to a potentially successful commercial
venture, namely the issue of productivity of computational science, has taken
place recently. Ingenious experimentalists, materials scientists, theoreticians,
mathematicians, and simulation specialists built on the early ideas for optical
soliton properties of Hasegawa & Tappert (1973) [Hasegawa (1995), Hasimoto
& Kodama (1995), and Akhmediev & Ankiewicz (1997)], who both were at
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

Bell Labs in the early soliton days. Optical solitons, manifest as light pulses
of picosecond width, were rst observed in glass bers by Linn Mollenauer,
Roger Stolen, and James Gordon at Bell Laboratories (Stolen et al 1980) in
1980. However, in nature, solitons do decay, and many technical problems
had to be overcome. Recently, 20-gigabit/sec soliton signals were transmitted
error-free in a loop over a distance of 14,000 km.
To review,
1. We realized that a continuum model could characterize the nonlinear dy-
namics of a lattice.
2. Kruskal derived the KdVequation froma dispersively regularized nonlinear
wave equation after making a Galilean transformation to a reference frame
moving with the linear speed of the lattice.
3. We made simulations of the KdVwith parameters corresponding to the non-
linear lattice with progressive wave initial condition and obtained agreement
in the near-recurrence time.
4. We made simulations with two solitary waves isolated in a periodic domain
and observed nearly nondispersive passage with recurrence to the initial
waveforms and with a phase-shift.
5. We observed similar numerical phenomena for the modied KdV equation
( p = 2 above).
Aspects from mathematics and other elds are in Bullough & Caudrey 1995,
who review mathematical developments, Weissert (1997), who presents a thor-
ough history of FPU and nonlinear lattices and other low-dimension chaotic
nonlinear systems (from 1954 to 1974); and Heyerhoff (1998), who presents a
broad history of (1+1) nonlinear wave and related partial differential equations
(from 1834 to 1967).
3. ACCELERATED INHOMOGENEOUS FLOWS
AND THE VORTEX PARADIGM
3.1 Application Domains
There are many density-stratied classical and contemporary high-energy uid
environments where accelerated ows are paramount, including combustion
(Curran et al 1996); inertial connement (laser) fusion, or ICF (Lindl 1995);
and astrophysics, particularly supernova remnants (Wang & Robertson 1985,
Muller et al 1991, Chevalier et al 1992, Klein et al 1994, Stone &Norman 1994,
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

502 ZABUSKY
Chevalier & Blondin 1995, Xu & Stone 1995, Dohm-Palmer & Jones 1996).
The latter two domains were brought into focus at a March 1998 symposium
in Tucson, Arizona (Remington & Arnett 1999).
In ICF, one desires to conne an amount of nuclear reactive gaseous mixture,
such as D-T, until it is heated to a temperature where the nuclear reaction
releases enormous amounts of energy. One procedure advocated at present
is to illuminate with lasers a small spherical capsule containing gaseous DT
surrounded by layers of solid DT and a solid ablator. The short-duration laser
pulses deposit energy within a narrow layer, causing an ablative ejection of
material in the direction of the incoming radiation, while an intense shock
wave is driven toward the center of the pellet.
If everything were ideal, the process would be one-dimensional and the shock
would be transmitted and reected at the interfaces. However, despite the care
taken to obtain perfect radial symmetry during fabrication and illumination, the
interface between the two media is perturbed, breaking the one-dimensional
symmetry, and the dynamical processes become three-dimensional. Hence, in
this domain the goal is to understand and control the interfacial density per-
turbations that are the locations of vortex roll-up and subsequent turbulence
and mixing activity, a process that inhibits the elevation of temperature of the
volume fraction of the ignitable nuclear mixture.
3.2 Overview of Rayleigh-Taylor (RT)
and Richtmyer-Meshkov (RM) Environments
The contemporary study of accelerated incompressible ows at a density-
stratied interface began with the linear stability analysis of a harmonically
perturbed interface by Taylor (1950) and the corresponding experiments of
Lewis (1950) (undoubtedly brought on by wartime considerations). This con-
stant acceleration environment is called the Rayleigh-Taylor (RT) instability.
Lewis concluded that the agreement of his experiments with Taylors exponen-
tial growth held until the initial small sinusoidal amplitude grew to 0.4, where
is the initial wavelength. The RT linear instability, including viscosity and
magnetic elds, was elaborated in the 1950s and was comprehensively sum-
marized (Chandrasekhar 1961). The effects of diffusion were also examined
experimentally (Duff et al 1962).
The shock-induced impulsive acceleration of an interface with sinusoidal
perturbations was studied analytically by Richtmyer (1960), who used a linear
representation after shock passage and derived a simple formula that captures
the early time motion of the interface for a restricted range of Mach numbers
and density ratios; and experimentally by Meshkov (1969), whose results at
early times were in qualitative agreement with Richtmyers formula. For these
reasons, the shock-accelerated interface is called the Richtmyer-Meshkov (RM)
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

environment. A correspondingly important experimental study of a shock-
spherical bubble was made (Rudinger & Somers 1960), and attention was paid
to the emerging vortex rings. Figure 1 shows three primary focus geometries,
where (a) is half a wavelength of the sinusoidal perturbation, (b) is the planar
inclined interface, and (c) is representative of a symmetrical cylinder in two
dimensions or an axisymmetric sphere in three dimensions. The last assumes
symmetry about the lower boundary and thus represents the model used in our
Euler dns.
Sharp (1984) reviewed the linear RT instability in various geometries. He
briey mentions compressibility and the RM environment. He includes exam-
ples of numerical simulations of the evolving sinusoidal interface (somewhat
beyond the linear phase) and ad-hoc theories of bubble dynamics. Bubble is
a term associated with the smoother part of the undulating (e.g. sinusoidal or
sawtooth) interface that is evolving, as discussed below. He did not assess the
role of vorticity in either RT or RM environments.
3.2.1 LABORATORY EXPERIMENTS ON SHOCK-ACCELERATED INTERFACES
Since the early 1980s, many careful and innovative experiments were made by
Sturtevant, Haas, Meshkov, Zaitsev, and their collaborators (Haas & Sturtevant
1987, Brouillette &Sturtevant 1989, Aleshin et al 1993). For shock tubes, there
had been concern about the fabrication and properties during shock interaction
of the interfacial polymer membrane. Sturtevant (1987) gave a review of the
RM environment and showed many pioneering experimental results, includ-
ing Meshkovs 1969 experiment; a continuation to a more turbulent phase by
Andronov et al (1976); a reshock experiment by Zaitsev et al (1985) (where
the second shock arises from the reection of the incident shock from the
downstreamwall of the shock tube); the shock-sphere (heliumin air); shock in-
clined planar interface and double inclined interface (recently called a curtain).
Sturtevants experiments were carried to intermediate times, and he commented
about visual manifestations of vorticityin particular, interface roll-up.
3.2.2 NUMERICAL SIMULATIONS OF SHOCK-ACCELERATED INTERFACES: VOR-
TICITY DEPOSITION AND MODELING Although vorticity was displayed by
Chalmers et al (1988) and Picone & Boris (1988), and divergence by the for-
mer, the rst serious quantication of vorticity was given by Hawley &Zabusky
(1989). Their focus of attention was the single planar inclined interface geom-
etry; they showed the space-time integrated vorticity for positive and negative
Atwood numbers and introduced the vortex paradigm as a guide to understand-
ing and modeling. Comparisons of juxtaposed images, amplitudes, and spike
and bubble velocities by Yang et al (1992) with Sturtevants experiments on pla-
nar interfaces showed excellent agreement at early times. Quantitative estimates
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

504 ZABUSKY
Figure 1 Schematic of primary-focus two-dimensional physical domains andparameters inshock-
accelerated density-stratied interfaces prior to shock interaction. (a) Sinusoidal interface. (b)
Planar interface. (c) Circular or spherical-axisymmetric interface.
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

for circulation deposition on interfaces for a wide variety of parameters was
given by Samtaney & Zabusky (1992, 1994). Studies of sinusoidal interfaces
(Holmes et al 1995) also showed qualitative agreement with experiments.
At the 18th (International Symposium on Shock Waves) ISSW in 1991, the
RMenvironment was discussed. Rupert (1992) summarized research for planar
interfaces parallel to the shock front (i.e. not inclined) at various Mach numbers
up to 3.4. Several color images of vorticity and divergence were shown. She
remarked that the numerical results seem to differ from the observations by a
signicant amount. Zabusky et al (1992) summarized research for the inclined
interface and the shock cylinder at M=4. Many new effects were revealed that
are not present at lower Mach number. Furthermore, there was emphasis on
novel visiometrics, including quantifying total positive and negative circula-
tion in the domain and visualizing vorticity and integrated vorticity space-time
diagrams.
Since 1991, there has been a large increase in the number of RM papers in
the uid and plasma communities. This is particularly manifest in the quality
and quantity of the papers at the recent 1997 Marseilles meeting (Jourdan &
Houas 1997) of the sixth International Workshop on the Physics of Compress-
ible Turbulent Mixing (IWPCTM), the focus group for RM research. At these
MIX workshops, groups from around the world present analytical, experi-
mental, and computational results, e.g. a history of the work at the Russian
VNIIEF center was outlined (IIkaev et al 1997).
3.2.3 PRIMARY FOCUS Many papers have been written on the linear theory
and early time simulations of the RM environment and are available in the
reviews by Sharp (1984) and Rupert (1992). In this paper, our primary focus
will be on linear and nonlinear models and vortex-related phenomena up to
early-intermediate times, as discussed below. For example, in recent years,
it has been found convenient to quantify, at early times, the rate of change
of amplitude of the sinusoidally perturbed interface extrema, i.e. the spike
and bubble velocities and their difference, a(t). However, this low-order in-
formation, although essential for making low-order experiment-simulation and
different-code comparisons, provides little insight into the physics of vortex
deposition and its localization (or roll-up) phenomena and the emergence of
small-scale structures. Furthermore, the curvature, nearness, and magnitude of
the large gradients of the rolling density interface are important for deposition
of vorticity in re-acceleration and reshock experiments, as described below.
Thus we will not discuss the multimode-statistical (Ofer et al 1995) and tur-
bulent mixing aspects of interfacial region growth, a vital area of late-time three-
dimensional research in this eld, as discussed in Jourdan & Houas (1997).
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

506 ZABUSKY
4. EQUATIONS, NUMERICAL METHODS,
PARAMETERS, GEOMETRIES,
AND TIME EPOCHS
4.1 Euler Equations of Evolution
The Euler equation set that we solved in two dimensions is written in conser-
vation form as
U
t
+F(U)
x
+G(U)
y
= 0, (2)
where U = {, u, v, E, }
T
, F(U) = {u, u
2
+ p, uv, (E + p)u, u}
T
,
and G(U) = {v, uv, v
2
+p, (E +p)v, v}
T
. E is the total energy related
to the pressure, p, by p = ( 1)[E
1
2
(u
2
+ v
2
)]; and is the interface
tracking function. The boundary conditions are reecting in the y-direction and
inow/outow in the x-direction.
We have used the conservative level set formulation (Osher et al 1988) in
which a function (x, t ) denes the interfacial domain everywhere in the do-
main. We initialize (x, 0) = +1(1) in the incident (transmitted) gas. The
last PDE in the system of equations (2) governs (x, t ) and is coupled to the
other equations through the variable (x, t ), which is determined by linear
interpolation for a particular cell. Note this system is not strictly hyperbolic.
4.2 Numerical Methods
Many nite-difference numerical methods with xed and adaptive meshes have
been used. Comparisons among them for convergence, accuracy, etc, usually
involve the primitive variables, e.g. density and temperature, and almost never
anything related to rate of strain, vorticity, or divergence. Some examples in-
clude Kang et al (1994), Don & Quillen (1995), Quirk & Karni (1996), Lee
& Giltrud (1996), Ofgeneim et al (1995), and Holmes et al (1998). Investiga-
tors use, often without mention, various regularization and ltering techniques
to make the codes more robust. This includes the introduction of numerical
viscosities near strong shocks and shear layers as well as articial steepening
of contact discontinuities, etc. In the future it would be desirable to develop
standards for comparative tests so that we may enhance the believability of
simulation predictions.
Our results are fromtwo numerical methods. Both are second-order accurate
in space and time and include interface tracking. For the rst, we assume that:
the owis inviscid; the gases are perfect and in thermodynamic equilibrium; and
there are no chemical reactions between the gases. The method is generalization
of Godunovs method, and is similar to the Eulerian MUSCLalgorithm(Colella
1985), which is suitable for ows involving nonlinear wave interactions. The
solution at each time step is interpolated to give a piecewise linear distribution in
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

each grid zone. Van Leers monotonicity constraints (van Leer 1977) are used to
mitigate short-wavelength numerical oscillations. The linearized characteristic
equations are solved to get the solution at an intermediate time step. The ux
terms are then obtained at each cell boundary by solving the full nonlinear
one-dimensional Riemann problem (Smoller 1982).
In our implementation, we use (x, t ) to determine the volume fraction of the
gases in each cell where |(x, t )| < 1. We use alternating sweeps in the x- and
y-directions, which formally yields second-order accuracy (Strang 1968). We
do not employ articial viscosities or gradient augmenters for contact discon-
tinuities. The details of the numerical method are given in Samtaney (1993).
For the second method, we use the equilibrium ux method of Pullin (1980).
Because of its diffusive nature, we can simulate strong shock phenomena. The
code has been extended by Samtaney & Meiron (1997) to include an ideal
dissociating model of a gas.
4.3 Parameters
The classical geometries for the RT and RM are those amenable to a linear
stability analysis, namely a discontinuous density-stratied interface, perpen-
dicular to an external acceleration or the normal vector of a planar shock wave,
respectively. The interface is perturbed with a small-amplitude single harmonic
ka
0
1, where k = 2/ and a
0
is the initial amplitude prior to shock arrival,
as shown in Figure 1a for RM (where only a half wavelength is shown).
The density ratio across the interface prior to shock arrival is =
1
/
0
, or
alternatively, the Atwood number is A = (1)/(+1). At present, we do not
include a model for the interface, e.g. the thin polymer membrane that separates
the two gases in a shock tube. In fact, the initial density jump is usually over
three zones and causes grid-scale perturbations on the front, as discussed below.
Recent experiments with a sinusoidal interface include Zaytsev et al (1993) and
Aleshin et al (1993, 1996).
The excitations are the acceleration g (for RT) which may be a function
of time, or the shock (for RM) as measured by the normalized pressure jump,
S ( p
before
p
after
)/p
after
, or the Mach number, M, M
2
= 1 +( +1)/(2 ))
[S/(1 S)]. For compressible media, we have an ideal equation of state where
the specic heats in each medium are
0
and
1
.
4.4 Geometries
A more fundamental conguration is the inclined planar interface in Figure 1b,
because it allows an exact calculation of circulation deposition in many cases
(Henderson 1966, Sturtevant 1987, Henderson 1989, Samtaney & Zabusky,
1994) and because a self-similar solution can be found that includes vortex
roll-up and other fundamental vortex congurations in the vicinity where the
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

508 ZABUSKY
incident shock approaches the interface (Samtaney & Pullin 1996). Another
basic conguration, shown in Figure 1c, is the circular cylinder, or axisymmet-
rical spherical bubble, discussed below (Haas Sturtevant 1987, Jacobs 1992,
Bourne & Field 1992, Yamada et al 1996, Zabusky & Zeng 1998, Klein 1998).
Numerical simulations and laboratory experiments are also being conducted
in other congurations: the vertical interface with multiple small-amplitude si-
nusoidal modes or multiple piecewise linear sections (Budzinski et al 1994); the
planar layer of identically inclined interfaces or a sawtooth curtain (Sturtevant
1987, Zabusky & Yang 1989, Yang et al 1990); the sinusoidal curtain, or a
vertical layer where each interface is a single small-amplitude sinusoidal mode
(Rightley et al 1997); and the elliptical cylindrical bubble (Zabusky et al 1999).
Of course, one could exaggerate aspects of all the above geometries and ob-
tain a multitude of mushrooming interfaces. For example, one could increase
the amplitude or change the wavelength or phase of the various harmonics or
sawtooth segments (Aleshin et al 1993), ka
0
1.0. For example, for large-
amplitude single harmonic interfaces, one sees secondary pressure waves arise
as the shock traverses the interface, waves that were not present for very small
amplitudes.
In addition to geometrical complexity, one could consider the complexity
of excitation, e.g. multiple planar shocks arriving from both directions of the
shock tube (the reshock problem) as well as consider cylindrical or spherically
implodingshocks intwoandthree dimensions. Finally, all the above canbe stud-
ied in three dimensions. If the initial condition is controlled, then vortex-related
insights gathered fromtwo-dimensional dynamics will provide pathways to un-
derstand three-dimensional effects up to intermediate times. However, at later
times three-dimensional effects are essential for the turbulence and mixing that
arise.
4.5 Equations of Vorticity Evolution
If one takes the curl of the momentum equation for a compressible Navier-
Stokes uid, one obtains the vorticity evolution equation

t
=u + u ( u) +
1
2
( p)
+

2
( )
4
3
2
[ ( u)] +

+
_

_
1
2
3
( u)() +2(u) () +()
___
,
(3)
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

where we have assumed that the viscosity, , is variable. The last two lines show
that viscous effects may be important (Hewett & Madnia 1998), particularly
the third row at high-gradient interfaces.
In all discussions below, we omit viscosity. The remaining terms on the
right of Equation 3 are the usual advection term, which when combined with
the left side gives the material derivative; the vortex stretching term, which is
essential for discussions of three-dimensional turbulence and mixing (We will
ignore this term below, as all discussions are of two-dimensional and non-swirl
axisymmetric ows); the vortex dilatation term, which is important only for
highly compressible uids; and the baroclinic term, which is most essential in
our discussions.
Other important sources of vorticity that are not immediately evident from
this equation arise from curvature in shock waves, as discussed recently by
Kevlahan (Kevlahan 1997) and shock triple-point phenomena that arise from
Machreections (BenDor et al 1980, Hornung1985). Here, slip-lines, or vortex
sheets (in modern parlance), arise and are unstable and the source of turbulence.
In this context, the triple-point phenomenon was discussed by Colella et al
(1986) for the blast wave surface interaction and by Winkler et al (1987) and
Zabusky & Zeng (1998) for the shockaxisymmetric bubble interaction.
4.6 Time Epochs
For RT and RM, the evolution of a small-amplitude sinusoidal interface driven
by an external acceleration or the passage of a single shock wave, respectively,
is sketched in Figure 2 in the left and right columns, respectively. The cong-
urations, i.e. the direction of g or direction of incident shock, M or M,
and medium through which it moves (heavy =slow and light =fast) empha-
sizes the deposition of vorticity. The sign of the vorticity on the interface is
shown as ( + + ) to the right or ( ) to the left in all
panels and is obtained from the directions of p and g, as in-
dicated. The right-hand rule for motion induced is implied. Also indicated
is whether the panel shows phase reversal (lower panels) or not (upper pan-
els). The lower left RT panel has two branches, one for a constant acceleration
g =constant (which produces stable oscillations or gravity waves) and one
for g =0 (at right), which illustrates the roll-up following the impulsive RM-
like case. Note in all panels that the roll-up in the downward-moving spike
region is indicative of the localization of vorticity for a nite Atwood num-
ber, smaller than one. Also, the single roll-up domain (rather than many roll-
ups along the interface) is indicative of a smallbut niteinterfacial layer
thickness.
For the RM environment, we split the evolution into the four following time
epochs:
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

510 ZABUSKY
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

1. Very Early Time (VET). The interval when the incident shock wave com-
pletely traverses the interface, a distance of 2a
0
, for the rst time. For a
very small amplitude perturbation or a very small interface angle, the linear
theory applies.
2. Early Time (ET). The interval when secondary reected and transmitted
pressure or shock waves pass over the interface (e.g. less than four transit
times of a linear wave across L
T
, the width of the shock tube). Nonlinear
effects arise early in this interval due to the growing amplitude of the inter-
facial perturbation, and the magnitude of the growth rate has achieved its
maximum. During part of this interval, the interfacial evolution may be mod-
eled by nonlinear theories, as discussed below, or with an evolving vortex
sheet on a nearly incompressible, density-stratied interface.
3. Intermediate Time (IT). The interval following ET when the vortex sheet
or layer on the interface localizes or collapses into one or more CVSs. For
sinusoidal perturbations, one may conveniently begin this interval when
the contact discontinuity becomes a multivalued function of the transverse
distance y, as discussed below.
4. Late Time (LT). The very large interval beginning when the owis driven by
an array of compact vortices on the density-stratied interface and continu-
ing to late in the phase when a turbulent-like collection of vortex domains
merges, binds, splits, and mixes the interface.
The time of transition between these intervals depends specically on details
of the run, e.g. the strength of the shock or amplitude of the harmonic (or
alternatively, the inclination of the interface).
Figure 2 A schematic of the very-early and early time behavior for the classical RT and RM
congurations, illustrated in the left and right columns, respectively. The congurations, namely
the direction of g or direction of incident shock, IS, moving in the direction of M and medium
through which it moves (heavy =slow and light =fast) emphasizes the deposition of vorticity.
We see positive to the right and negative to the left (obtained from the directions of ,p, and g)
and a downward-moving spike in all panels. The right-hand rule, for motion induced, is implied.
Also indicated is amplitude phase reversal (lower panels) or not (upper panels). The RT panel
(b) has two branches: g =constant (which produces stable oscillations or gravity waves), and
g = 0 at right, which illustrates the roll-up following an impulse, as in the experiment of Jacobs
& Niederhaus (1997). The roll-up in the spike region is indicative of an Atwood number smaller
than one.
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

512 ZABUSKY
5. VISIOMETRIC METHODS AND CONVERGENCE
5.1 Visiometric Methods for RM Environments
As described in Bitz & Zabusky (1990), we visualize and quantify density;
dilatation, (div u) (to highlight wavefronts and shock waves); vorticity, =
e
z
u; numerical shadowgraph, (to juxtapose contact discontinuities
and shock fronts); and baroclinic generation,
2
( p). The Lapla-
cian (second derivative) of may oscillate and so is reserved for early time
analysis.
To obtain a global view of the ow, we project various functions to lower
dimensions and often present space-time diagrams. For example, we present
integrated positive, negative, and total vorticities (or circulations per unit length)
for planar or axisymmetric ows. For example, the projections
(x, t ) =
_

(x, r, t ) dr, (4)

where
= (1/2)( ||), are either the positive or negative values of the

z or components of vorticity, respectively. Similar diagrams are given for
integrations of vorticity along the shock direction, as seen in Figure 6, and also
the pressure on axis, p(x, 0, t ).
Figure 3 (see color section) (Figure 4.6 of Samtaney 1993) shows the inte-
grated total vorticity up to the late IT phase. The run is for a shock (M = 1.5)
interacting with a planar layer (Air/R22, = 3) at an angle of = 60
. (See
Yang et al 1992 at early time for Air/R22 and M = 1.2 and = 30
). Two
frames of reference are used: laboratory frame for 0 < t < 200; and a frame
translating uniformly with the velocity of the post-shock uid. (The latter
frame is chosen so the page is lled optimally with data.) The dominant fea-
tures are the smooth deposition of circulation at VET (the negative sloping line
over the black region); the localization or roll-up of vorticity at the lower wall
(at left), a dominant feature, particularly when examined in color; the array of
vortices that emerge from the initially deposited smooth layer as a result of a
Kelvin-Helmholtz instability, as indicated by the multitude of positively sloping
thin domains emerging at 300<t <500. Here, perturbations to the vortex layer
arise mainly from the mesh-related initialization of the density. The merger of
these vortices (VM) with the wall vortex and with each other is a characteris-
tic process in two-dimensional turbulence. Also seen are white regions around
the increasing stratied vortex at lower left. This is due to the entrainment
of the shear layer generated after shock transmission and the baroclinic gen-
eration of vorticity on the rolling interface, an important process yet to be
modeled.
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

Figure 3 y-integrated vorticity for a shock (M = 1.5) interacting with a planar layer (Air/R22,
= 3) at an angle of = 60
. Two frames of reference are used: laboratory frame for 0 < t <
200 and Galilean frame translating uniformly with the velocity of the post-shock uid, 0t > 200.
The dominant features are the deposition of a layer of vorticity (uniform black triangular region)
and emergence of lower wall vortex at left; array of vortices emerging from the layer as a result
of the Kelvin-Helmholtz instability ( positive sloping, thin, dark regions) and the merger of these
vortices (VM) with the wall vortex and with the neighboring vortex, a characteristic process in
two-dimensional turbulence. Also seen are white regions around the increasing stratied vortex,
mainly at lower left. This is due to the entrainment of the shear layer generated after shock
transmission and the further baroclinic generation of vorticity.
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

514 ZABUSKY
At any instant, the total circulation on the interface is calculated (Samtaney
& Zabusky 1994) as
(t ) =
D
(i, j, t )xy, (5)
where we use discretized representations
(i, j, t ) =
v(i +1, j, t ) v(i 1, j, t )
2x

u(i, j +1, t ) u(i, j 1, t )
2y
(6)
and where D = {(i, j, t ) such that |(i, j, t ) < 1 t h|}, and th is a chosen
threshold.
The domain of integration is chosen to highlight special phenomena. For
example, to examine the negative circulation on the interface and within the
bubble, we accumulate only those cells with
< 0 that coincide with the

location of
1
, the bubble uid.
5.2 Visiometrics of Circulation on Interfaces
and Within Elliptical Domains
We nowexamine several runs in detail and compare models for themin sections
below. We consider mostly the two-dimensional domain with a sinusoidal inter-
face, ka
0
= 0.1 separating air and R22 where = 3.0,
0
= 1.4,
1
= 1.172.
Panel c2 of Figure 2 provides heuristic sketches. The undisturbed incident
gas (0) is initialized with unit pressure and density, and the incident shock
of M = 1.5 is initialized by density and pressure jumps in accord with the
Rankine-Hugoniot conditions. The simulations were made with our Godunov
code, the domain was discretized with three meshes (800 40, 1600 80, and
3200 160), and a
0
corresponds to 4, 8, and 16 zones, respectively. It takes
4.5, 9.0, and 18.0 time units, respectively, for the M = 1.5 shock to cross the
interface, 2a
0
. At t = 0, the shock strikes the interface.
5.2.1 VISIOMETRICS AND CONVERGENCE OF INTERFACIAL CIRCULATION In
Figure 4 we see images of vorticity with the one mid-height density contour
superimposed, for an Air/R-22 interface with M = 1.5, ka
0
= 0.1, and reso-
lution 1600 80, shown at these times: (a) t = 113.2 (end of the VET phase);
Figure 4 Images of vorticity with the one mid-height density contour superimposed. For an
Air/R-22 interface with M = 1.5, = 3.0, ka
0
= 0.1, and resolution 1600 80. The time for
the shock to cross the interface, 2a
0
, or 16 grid points, is nine units. Shown at times (a) t = 113.2;
(b) t = 484.7, and (c) t = 981.3. Note in (a) the initial position of the incident shock with respect
to the interface (to scale) is shown in the box inserted at left and the present position of the slightly
curved transmitted shock is at right.
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

516 ZABUSKY
(b) t = 484.7 (end of the ET phase); and (c) t = 981.3 (early in the IT phase).
The actual shapes and relative positions of the initial condition (shock with
respect to the interface) is shown in the insertion at left in (a), and t = 0 occurs
when the shock rst strikes the interface. The present position of the slightly
curved transmitted shock is at right. In (b) the density contour rst becomes
multivalued (kinks), and in (c) the vorticity has localized and rolled up and
the density contour clearly shows the formation of a spike at left and a bubble
at right. Note the localization of the vorticity.
In Figure 5, we see temporal variations of circulation, obtained as described
above. The simulations were made with the Godunov code at three resolutions,
as mentioned above. Curves 2 and 3 are sufciently close for us to conclude
Figure 5 Convergence of interfacial circulation for an Air/R-22 interface with parameters as given
in Figure 4. Simulations with the Godunov code at resolutions (1) 800 40, (2) 1600 80, and
(3) 3200 160.
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

that the compressible vortex dynamics is captured with sufcient accuracy to
allow modeling during this time interval. The rst minimum of the negative
circulation corresponds to the time when the shock has traversed the interface,
2a
0
. The early undulations are due to reected shocks passing over the interface
(depositing rst positive and then negative vorticity, respectively.) The last two
crossing times are larger than the initial transit time because these reected
shock waves are weaker than the original shock and are moving across the
lateral extent of the shock tube.
5.2.2 QUANTIFICATION OF INTERFACIAL CIRCULATION In Figure 6, we see the
corresponding x-integrated vorticity plotted as a function of y at seven times.
The rolling and localization of the interfacial vorticity is manifest in that the
curve is narrowest in the interval between 4 and 5, 465 < t < 616, where the
interface become multivalued.
Figure 6 x-integrated vorticity across the interface as a function of y at seven times for an Air/R-
22 interface with parameters as given in Figure 4. The shock strikes the interface at t = 0 and
traverses it in nine units.
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

518 ZABUSKY
In Figures 7 and 8, we see results of quantifying the interfacial vorticity
with a quadratic form, ellipses tted to CVS at 30% of maximum vorticity.
In Figure 7 we see the evolution of aspect ratio and normalized major axis.
Curve 1 at M=1.5 corresponds to the previous quantications. Curve 2 at
M=3.0 shows that the larger the deposited circulation, the more rapidly does
the circulation localize, as with an incompressible layer (Baker &Shelly 1990).
Also, later in time, curve 2 grows slowly, possibly due to diffusion, and has a
superimposed quasiperiodic behavior due to the rotation of the localized vortex
domain. Figure 8 shows the evolution of normalized y-centroid position of
these ellipses toward the tip of the spike, a result seen in Figure 4c. To obtain
good models of interface dynamics we must elucidate and model the temporal
approach to one or more asymptotical states of centroid, major axis, and aspect
ratio, etc, as illustrated in these last gures.
6. ANALYSIS AND SIMULATION
6.1 Linear Stability
Because of mathematical simplicity, the growth of the small-amplitude single
harmonic perturbationof a two-dimensional interface, ka
0
1, has beena major
focus of attention in the 1990s. Richtmyer analyzed the temporal problem and
presented a simple scaling formula for the terminal or asymptotic growth rate
after shock passage,
a
I
ka
0
A
U, (7)
often called the impulse formula or impulse model. Here, k is the wave num-
ber of the perturbation, a
0
is related to the initial amplitude of the harmonic
perturbation to the interface or contact surface (as elaborated below), A
=
(
1)/(
+ 1) is the post-shock Atwood number (i.e.
1
/
0
), and
U is the speed of an interface after shock passage from medium zero to one
in a corresponding one dimensional interaction (if the interface is at rest prior
to the arrival of the shock). Richtmyer suggested using post-shock amplitudes
above, or a
0
= a
0+
for the case when the reected wave is a shock. Note
that we designate the pre-shock amplitude or initial interface perturbation as
a(0) a
0
. Recently, several investigators have claried and elaborated on the
linear theory (Fraley 1986, Mikaelian 1994a,b, Yang et al 1994, Wouchuk &
Nishihara 1997).
The work of Yang et al (1994) presented a comprehensive study of the linear
(or VET) RM environment for sinusoidal interfaces that separate regions of
different density and . [Note in their paper that the shock moves from region
2 to region 1 and that their interface growth rate is designated as a
0
(t ) while
ours is a(t ).] They introduced a new self-similar partial differential equation
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

Figure 7 Evolution of aspect ratio (solid) and normalized major axis ( ) of ellipses
tted to CVS at 30% of maximum vorticity. For an Air/R-22 interface with parameters as given in
Figure 4, and (1) M = 1.5 and (2) M = 3.0. The times for the shock to cross the interface are 9.0
and 4.5 units, respectively.
Figure 8 Evolution of normalized y-centroid position. For an Air/R-22 interface with parameters
as given in Figure 4, and (1) M = 1.5 and (2) M = 3.0. The times for the shock to cross the
interface are 9.0 and 4.5 units, respectively.
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

520 ZABUSKY
for the pressure as function of the variables (t, ) = (t, x/t ) (different from
the one used by Richtmyer) and solved it by consistent second-order accurate
numerical methods.
Yang et als Figures 11 and 12 show the linear terminal growth rate, a
0
()
vs incident shock strength, S = ( p
before
p
after
)/p
after
, for reected shocks
(fast/slow congurations) and reected rarefactions (slow/fast congurations),
respectively. As expected, for small Sand, the growthrate magnitudes increase
with increasing S and . However, they note without explanation that almost all
the curves are non-monotonic as S increases. Yang et al agreed with Richtmyer
that post-shock densities (on both sides of the interface) are to be used in the
impulsive formula. They also agreed with the consensus of investigators that
in the formula for a
I
: a
0
a
0+
, for a reected shock, a
0

1
2
(a
0
+ a
0+
),
for a reected rarefaction, where the last had been suggested by Meyer &
Blewett (1972). Careful comparisons of the terminal growth rate were made to
the impulsive model a
I
over a range of parameters for cases where the wave
reected fromthe interface was a rarefaction or a shock. These results are given
in Tables I and II and Figures 16 and 17 and show strong disagreement for
S > 0.8.
Yang et al conclude with the following comments:
1. The impulsive formula with parameters chosen as above agree well in the
weak incident shock limit.
2. The agreement between the impulsive model and the linear theory improves
as the adiabatic exponent s of the two uids increase while remaining
approximately equal. This is especially true for the reected rarefaction
case.
3. For strong incident shocks, large discrepancies between the linear theory and
impulsive model appear, especially when the s are substantially different
(see their Table 1).
6.2 Nonlinear Theory
Recently Velikovich & Dimonte (1996) (for A = 1) Zhang & Sohn (1997) and
Berning & Rubenchik (1998) presented formal perturbation calculations (fol-
lowing the work of Haan 1991) for the growth of a very small amplitude, single
harmonic perturbation to a vertical interface beyond early times, a nonlinear
theory. The latter two assumed incompressibility and irrotationality off the
interface and derive a Taylor series of four terms whose highest-order term in-
cludes the fourth harmonic multiplied by terms O(a
0
A
3
), etc. It is obvious from
the multivalued character of the evolving interface (due to voricity localization)
that this description is inadequate after a very short time.
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

The authors (Zhang & Sohn 1997) recognized the divergent nature of their
series, and because the overall growth rate decays at large times, they introduced
Pad e approximants and a heuristic inner-outer matching recipe. . . proposed
by Prandtl to obtain growth rates for the bubble and spike (their Equations 58
and 59) and total amplitude as
a
b
= a
li n
/D
1
+ a
li n
A
/D
2
, (8)
a
s
= + a
li n
/D
1
+ a
li n
A
/D
2
, (9)
a da/dt = (1/2)( a
s
a
b
) = a
li n
/D
1
, (10)
where
D
1
= 1 +a
0+
k +max
_
0, a
2
0+
k
2
A
2
+
1
2
_
2
, and
D
2
= 1 +2a
0+
k +(4/3)
_
3a
2
0+
k
2
A
2
+1
2
.
(11)
Here, A
is the Atwood number based on post-shock densities, = k a

li n
t , and
a
li n
a
0
is the time-varying solution obtained from their linear analysis. Note
that in the D
i
s we have used the post-shock amplitude, a
0+
, as described above
for the case of a reected shock.
Their growth rates are a product of the time-dependent linear growth rate,
a
li n
, multiplied with an approximant that has an asymptotic property of varying
like t
1
. However, the true asymptotic velocity is a constant, +0 (t
1
), because
a vortex dipole forms in the spike region. Thus, the Pad e process converts the
divergent Taylor series into a function that may be close to the comparable a
from the dns, over the ET range. However, at later times, particularly in the
spike region, the discrepancies are larger. Furthermore, their gures of smooth
waveforms of the interface do not showthe small-scale structures that truly arise
on very high-gradient interfacial layers. For additional comments on nonlinear
theories at ET, see Neuvazhayev & Parshukov (1997).
7. VORTEX PARADIGM MODELS
7.1 Overview
The deposition of vorticity in a two-dimensional inviscid uid is based on the
equation
D/Dt =
p
2
u, (12)
=
[Du/Dt +g]
u, (13)
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

522 ZABUSKY
which is obtained fromEquation 3, if the viscous terms and the vortex stretching
termare omitted. The rst (baroclinic term) is dominant in our discussions. For
RT, we replaced p/ by terms in the momentum equation to make explicit
the forcing by the external acceleration, which helps elucidate Figure 2.
Most RM models do not recognize the importance of the vortex view-
point. This situation is changing. Following Hawley & Zabusky (1989),
Samtaney & Zabusky (1994) derived formulas for circulation per unit length
deposited on discontinuous planar and curved interfaces. Samtaneys thesis
(1993) contains a multitude of simulations with visualizations and quantica-
tions. The early time vortex models discussed in Section 7 were rst presented
there. Jacobs et al (1995), Jacobs & Sheely (1996), and Rightley et al (1997)
interpreted experimental data fromcompressible shock layers (curtains) and im-
pulsively accelerated, falling tanks, respectively, with von Karmanlike point
vortex models of a nonstratied uid. Despite these simplications, they es-
timated sufcient parameters to achieve collapse of the data over a short time
interval. A more systematic approach up to intermediate times was made by
Ray et al (1996) using both incompressible vortex sheet and point models, and
is given below.
7.2 Caveats on Interfaces and Models in Two
and Three Dimensions
The problem of a two-dimensional vortex sheet in a homogeneous inviscid
medium (the Kelvin-Helmholtz instability) is ill-posed and gives rise to the
Moore singularity in a nite time (Hou 1995). This singular behavior is the
source of rapidly growing small-scale perturbations that roll up in numerical
computations. A comparison of a Navier-Stokes simulation with a regularized
single sheet of vorticity shows the limited capability of such single-sheet models
beyond a time of two rolls (Tryggvason et al 1991). With surface tension
this singularity is suppressed, but at large Weber number, other singularities
may arise (Hou et al 1997). Similarly, for an incompressible discontinuous-
density RT environment, vorticity is deposited on a sheet and the problem is ill-
posed (Baker et al 1993). Samtaney & Pullin (1996) and Samtaney (1997) have
reexamined the RM situation by comparing computer-generated self-similar
solutions with highly resolved evolutions and have found small-scale roll-up
phenomena in the latter, comparable to those shown in Figure 3. They note that
the small-scale structures appear sooner as the resolution increases, a result
to be expected, because the layer thickness is smaller ( x) and hence the
vorticity is larger, since circulation per unit length is conserved. In all further
discussions, we assume that the numerical solutions are regularized by pro-
cesses associated with grid-related truncation errors. For long-time turbulent
mixing calculations, the interfacial layer should be adequately resolved and
described by appropriate physical processes.
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

7.3 Modeling VET Circulation Deposition with a Local
Shock-Polar Analysis
7.3.1 INTRODUCTION The analysis assumes a frame of reference that is sta-
tionary with respect to the node where all the waves meet and both gases
are perfect and inviscid. The geometry and variables of interest are given in
Figure 9. The notation is the same as that used by Samtaney & Zabusky (1994).
Here, mm is the interface; i, r, t are the incident, reected, and transmitted
waves, respectively; s1 and s2 are the streamlines in the incident and transmit-
ted media;
0
and
1
are the deections of s1 due to i and r, respectively; and
b
is the deection of s2 due to t . p
0
and p
b
are the initial pressures in the inci-
dent and transmitted gases, respectively. Here p
1
, p
2
, and p
t
are (respectively)
the pressures behind the incident, reected, and transmitted waves. The free-
stream Mach number in front (behind) the incident, reected, and transmitted
shocks are M
0
(M
1
), M
1
(M
2
), and M
b
(M
t
), respectively.
0
and
b
are the ratio
of specic heats in the incident and transmitted gases, respectively.
Ashock polar is a graph of a deection angle vs a pressure ratio. It represents
the locus of nal states that can be connected to a given initial state for a
stationary oblique shock. For a planar interface, the angle between the incident
shock and the interface is constant. If is smaller than
cr
, all the three waves
are shocks and meet at a single node (called RRR). For very weak shock waves,
as is increased, the reected wave becomes a Mach line and then a centered
rarefaction fan (called RRE). As is increased further, the reected wave is not
a centered expansion fan and the refraction becomes anomalous. For stronger
Figure 9 Schematic of regular refraction, for three shocks at a fast-slow interface. Shown are
incident, i , reected, r, and transmitted, t , shocks intersecting at a node on the interface, m. The
subscript notation (before and after) in this gure is from Samtaney & Zabusky (1994) and differs
from that used in the present paper.
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

524 ZABUSKY
shocks, the irregular refraction (called MRR) is typically characterized by the
appearance of a Mach stem at the interface.
Samtaney & Zabusky (1994) show (in their Figure 3) that for a large range
of > 1,
cr
decreases rapidly from 90
in 1 < M < 1.5 and then approaches

a near-constant small negative slope when 1.5 < M < 4.0. The larger , the
larger the decrease in
cr
. For example, for Air-R22 ( = 3.0,
0
= 1.4,
b
=
1.172),
cr
= 54
at M = 4, and for He-Xe ( = 32.7,

0
= 1.667,
b
=
1.667),
cr
= 36
at M = 4.
1
7.3.2 EXACT CIRCULATION ON THE INTERFACE Samtaney & Zabusky (1994)
followed Henderson (1966) and obtained a 12th-order polynomial in p
2
. This
polynomial may have more than one real root or no real roots. We concern
ourselves with the physical root, which is the smallest real root.
They dened the circulation (vorticity deposition) per unit length of the
shocked interface as
d
ds
v
t
v
2
, (14)
where v
t
and v
2
are the velocities tangential to the shocked interface in the trans-
mitted gas and the twice-shocked incident gas, respectively. They renormalized
the circulation with respect to the original unshocked interface by multiplying
Equation 14 by the geometric factor ds
/ds = [cos / cos(

b
)], which ac-
counts for the instantaneous change in length of the interface due to the shock
passage

d
ds
= (v
t
v
2
)
cos
cos(
b
)
(15)
and obtained
=

1
2
0
sin
_
_
M
2
+
2
b
1
_
1 ( p
2
,
b
)
0
/
b
_
sin
2
_1
2
_
M
2
+
2
0
1
(1 ( p
2
/p
1
,
0
)( p
1
,
0
)) sin
2
_1
2
_
cos
cos(
b
)
,
(16)
where
2
i
= (
i
1)/(
i
+1) and (, ) (1 +
2
)( +
2
). Note that
sin in Equation 16 has the same sign as p.
1
For simplicity, we normalize by assuming p
b
= p
0
=
0
1 and
b
= initially. Thus,
the sound speed is
1
2
0
in the incident gas. In the sections that follow, we will normalize the
circulation on the interface by a length scale, which will be indicated. All the results will thus have
units of circulation per unit length, i.e. units of velocity. In order to convert to physical units, our
normalized circulation must be multiplied by the ratio of the physical sound speed to
1
2
0
.
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

In the above equation,
1
2
represents physically the ratio of sound speed be-
hind and ahead of a shock. Equation 16 may only be used for RRR at a planar
interface, and we refer to it as the exact shock polar result for
. Typically,
for a slow-fast interface, the shock refraction is of the RRE type for small
(Puckett et al 1991). For RRE, one can derive a corresponding equation for
circulation by replacing the shock jump conditions across the reected wave
with the Prandtl-Meyer expansion conditions. This is given in Samtaney et al
(1998).
Figure 10 shows the exact circulation,
, in the (M
1
,
1
2
) plane for =
1.4. The angle of the interface is = 15
, 30
, 45
, and 60
. The normalized
Figure 10 Circulation per unit length,
normalized by
in the M
1
,
1
plane. The angle
of the interfaces are (a) = 15
, (b) = 30
, (c) = 45
, and (d ) = 60
. The light gray

(or black) regions show where
is smaller (or larger) than 10% of
. The white spaces show

regions where the refraction is not RRR.
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

526 ZABUSKY
circulation is set to zero where the shock refraction is not RRR (white space
indicated by IR). The scaling lawholds exactly when the normalized circulation
is exactly equal to one. The 10% error boundaries (which bound the light grey
space) correspond to contour levels 1.1 and 0.9. For small , there are regions
where the scaling laws are within the specied error bounds up to very large M
at intermediate values of ; and up to very large at low to moderate values of
M. For instance, for 30
, the error is less than 10% for 1.0 < M 1.32

for all , and 5.8 32.6 for all M.
7.3.3 APPROXIMATE MODEL FOR CIRCULATION ON THE INTERFACE When the
shock polar equations do not yield a physical root (an irregular interaction),
we have no exact formula for velocity jump. However, if we expand
in
Equation 16 as a series in sin about sin = 0, we obtain results close to the
quantications of our numerical simulations:
1
sin +
3
sin
3
+ O(sin
5
), (17)
since the circulation on the interface is an odd function of sin . Here
1
=
/ sin |
sin =0
and
1
(the dominant term) and
3
are both real in 0
/2 and are given in Samtaney & Zabusky (1994).
7.3.4 SCALING LAW FOR CIRCULATION To simplify our discussions, we use
identical
s, or =
0
=
b
. Hence in the limit of large M and small , we
obtain
, where
=
2
1
2
+1
(1
1
2
) sin (1 + M
1
+2M
2
)(M 1). (18)
Thus, the circulation deposited per unit original length,
, scales as
1. (1 + M
1
+2M
2
)(M 1) in Mach number. This leads to linear behavior
in M for large M.
2. (1
1
2
)for the density ratio. Note that for large , circulation is essen-
tially independent of the density ratio. For , the circulation should
correspond to the slip velocity at a solid surface.
3. (sin ) for the inclination of the interface with the shock.
4.
1
2
/( +1) for the ratio of specic heats.
7.4 Reduced Vortex Models at Early
and Intermediate Times
Ray et al (1996) present three reduced models that assume a Biot-Savart formu-
lation, where the vorticity drives the interface.Two of them, the stratication-
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

modied Biot-Savart (SMBS) model and the line-vortex (LV) model, are pre-
sently limited in that they give accurate results over restricted time ranges.
Comparisons for a (the difference between the velocities of the x-extrema
[spike, y = 0, and bubble, y = /2] of the contact discontinuity) are given in
Figure 11a and b among the impulse model (horizontal solid line); linear anal-
ysis (described in Section 6.1); dns (
); and NBS (
);
SMBS () and LV (), as described below. Note that the
doubling in M yields an approximate factor of two in time in the appearance
of all phenomena (which is associated in the almost factor of two in deposited
circulation).
7.4.1 NUMERICAL BIOT-SAVART (NBS) For the numerical Biot-Savart, or NBS,
the interfacial vorticity and location are extracted from the dns and substituted
into the Biot-Savart formula. The comparison with dns for two runs, M= 1.5
and M= 3.0 for = 3.0, is shown in Figure 11a and b. The excellent quality of
the agreement at M= 1.5, after the waves have diminished, validates the code
andthe idea that the evolutionof vorticityis the cogent process for understanding
interface dynamics.
7.4.2 STRATIFICATION-MODIFIEDBIOT-SAVART(SMBS) MODEL Here we assume
that only the original rst harmonic is present and substitute into the equation
for x-component of velocity, u(x, y, t) to obtain
a =
(t )k(2
)
4( +1)
_

sin d
k
2
a
2
(1 cos )
2
+
2
, (19)
where = ky, (t ) is a piecewise linear approximation to the extracted time-
varying circulation, and where we have included (2
)/( + 1), a heuristic

factor obtained from linear theory (Chandrasekhar 1961), to account for the
stratied uid. The comparison for two runs, M=1.5 and M=3.0 for = 3.0,
is shown in Figures 11a and b by (). The model is valid in a small time
interval where a saturates, i.e. before the higher harmonics grow and couple to
each other.
7.4.3 LINEVORTEX(LV) MODELATINTERMEDIATEANDLATETIMES We assume
an array of point vortices of strength

, placed for simplicity at y = (2n +
1)/4, n = . . . 2, 1, 0, 1, 2, . . .. For the present, the circulation

is the
total circulation deposited after rst shock passage, a quantity well estimated
at low M by the methods above. From this, we obtain
a = (
3
k/2) cosh(ak) or sinh ka = (
3
/2)k(t t
s
) +sinh ka(t
s
).
(20)
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

528 ZABUSKY
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

The comparison for two runs, M= 1.5 and M= 3.0 for = 3.0, is shown in
Figure 11a and b by (). The model improves in accuracy at later times
when the dominant vortex is well localized and a(t ) is large, i.e. the interface
extrema are remote from the localized vortex centroid. See Jacobs 1995 for a
similar approach.
This model can easily be improved when we relax the simplication of a
point vortex to an elliptical vortex domain and also place the vorticies at their
true centroid, (x
c
, y
c
) (from a model yet to be obtained or from the data, e.g. as
previously shown for y
c
), in Figure 8 so that we obtain an asymptotic a which
approaches a constant.
8. SHOCK BUBBLES AND VORTEX
PROJECTILES (VPs)
The planar-shock bubble interaction is fundamental to many domains, but dif-
cult to understand and model, particularly at higher Mach numbers and later
times (Klein et al 1994, Zabusky & Zeng 1998, Klein 1998, personal commu-
nication). The separation and fragmentation of the initial interfacial vorticity
layer introduces complex uid dynamical processes.
We have introduced the concept of a translating vortex projectile (VP)
(Zabusky et al 1997, Zabusky & Zeng 1998) to synthesize a new understanding
of vortex dynamics in the vicinity of interacting nearby layers or structures of
opposite-signed vorticity, in particular, those deposited on contact surfaces or
high-gradient regions when a shock wave interacts with a spherical or elliptical
bubble (Zabusky et al 1999) that is heavier than ambient or in reacceleration
and reshock environments for RT and RM, respectively.
The VPs are coherent vortex structures (CVSs) of positive and negative or
dipolar circulation in two-dimensions or oppositely-directed, nearby tube-like
regions (e.g. a vortex ring) in three dimensions. These VPs move under their
own self induction, i.e. they have a dominant translational motion in a frame of
reference in which the mean ow is zero. The simplest planar and axisymmet-
ric examples for homogeneous incompressible ow include, respectively, the
dipole, a rectilinear pair of point vortices, (, ), separated by d (which trans-
lates with speed /(2d)) and the axisymmetric vortex ring [e.g. the limiting
formis Hills spherical vortex (Saffman 1992)]. Note that for three-dimensional
Figure 11 Evolution of circulation, 0.005 (dashed), a for the numerical simulation (dark solid )
and the models: Impulse model, a
I
(horizontal line); Linear analysis (of Section 6.1) ;
NBS (
); SMBS (); and LV, (). For Air/R-22, = 3.0, ka

0
= 0.1,
and resolution is 1600 80. (a) M = 1.5 and (b) M = 3.0.
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

530 ZABUSKY
homogeneous, incompressible uids, circular vortexrings are unstable (analysis
of Widnall and colleagues, 19711973, as described in Saffman 1992); ellipti-
cally deformed rings may break up into two ringlets in a nite time [Biot-Savart
simulations of Fernandez et al (1995)] and colliding coaxial rings will produce a
multitude of ringlets [experiments of Lim&Nickels (1992)]. The compressible
andstratiedanalogs of these andrelatedcongurations is yet tobe investigated.
These two- and three-dimensional VPs may arise through instability of some
stationaryor slowlyevolvingcongurations of positive andnegative circulation.
Couder and colleagues (Couder & Basdevant 1986), in pioneering experiments
and simulations on two-dimensional cylinder wakes, showed that perturbations
will induce strain elds that cause nearby opposite-signed vortices to link up
or bind and move together as a single dipolar VP, which is frequently manifest
as a mushroom or tadpole structure. They are an important component of
many geophysical ows (Fedorov & Ginsburg 1989). Note, in compressible
media, that isolated dipoles may move with variable speeds that depend upon
their Mach number and aspect ratio (Scheidegger 1997). Also, they have been
seen by Youngs (1994), who simulated a shock striking a randomly perturbed
two-dimensional interface and observed vortex pairs that propagate away from
the mixing zone.
The VP concept is generic in two and three dimensions and will play an
important role in interpreting reshock and reacceleration phenomena.
Also, in analogy with the geophysical environment and the CVS seen in
dns, I have conjectured that recent observations of explosion fragments
(Aschenbach et al 1995) or knots (ODell & Handron 1996) in the vicinity of
supernova remnants and planetary nebula could be VPs. To validate the conjec-
ture, observationalists must quantify the temporal variation of the translational
velocity, radius, and internal rotational motion associated with VPs.
9. SUMMARY
We have shown how a vortex viewpoint, enhanced by visiometrics, will enrich
understanding of phenomena and the accuracy and modeling of accelerated
inhomogeneous ows, e.g. RM processes up to intermediate times. Similarly,
this modus will facilitate the understanding and statistical modeling of simula-
tions of turbulent-like phenomena in two and three dimensions, which many re-
searchers performwith inviscid codes. However, their results are useful only for
large-scale and qualitative aspects, i.e. their intermediate and small-scale struc-
tures are corrupted beyond intermediate times. This follows because most codes
are second-order accurate away from shocks and because the inviscid problem
is ill posed for vortex sheets. In the future we need codes of higher-order
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

accuracy and interactive visiometric environments to handle the information
available from these massive data sets.
10. OMISSIONS
The author has tried to cite original papers, early reviews, and recent papers
in RT and RM, particularly those that emphasize the dynamics of vorticity.
Omissions from the list have been made because of space reasons or lack of
knowledge. I pay tribute to all, who in spite of valuable contributions, are not
mentioned in this brief paper. I hope they will send me their annotated works,
as I may have been unaware of them.
ACKNOWLEDGMENTS
The insights presented result from nearly a decade of work and collaborations
with many friends and colleagues. Some early insights in the shock interface
problem were obtained in joint work with K-H Winkler and P. Woodward in
1987 and John Hawley in 1989. Recent progress was made with X Yang, I-L
Chern, J Ray, S Zeng, AKotelnikov, and particularly RSamtaney. I amgrateful
to R Samtaney and W Rider for comments on the manuscript. J Ray assisted
in producing the manuscript. I am also grateful to other (at the time) graduate
students and staff in the VizGroup, including X Wang, M Gao, V Fernandez,
H-B Yao, P Mossip, W Townsend, and S Cooper.
In 1987 and 1989, collaborations on visiometrics began with F Bitz and
Prof. Deborah Silver, respectively. During the period 19891991, we had sup-
port from the National Science Foundation under Dr. Alvin Thaler (DMS-
89-01-900), and since 1993 our visiometric work has been supported by the
Department of Energy under Dr. Daniel Hitchcock and Dr. Fred Howes.
Computations were done on the Cray Y-MP and the CM-2 at the Pittsburgh
Supercomputer Center and on the CM-5 at NCSA. Recent computations were
done on the Cray T3D and T3E at the Pittsburgh Supercomputer Center, on the
IBM SP2 at the Cornell Theory Center, and the Origin2000 at NCSA. All these
centers are supported by the National Science Foundation. We also acknowl-
edge the support of the Center for Computer Aids for Industrial Productivity
(CAIP) at Rutgers University.
NOTE ADDED IN PROOF
There is a question of the asymptotic behavior of the growth rate for an ini-
tially small amplitude sinusoidal perturbation on an interface as discussed in
Section 6.2. The common belief of the community is that a varies as (1/t).
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

532 ZABUSKY
In the last paragraph of Section 6.2 we assert that it is = constant + 0 (t
1
).
A Kotelnikov, J Ray, and I have recently completed a study of this asymp-
totical process which shows that for an incompressible situation, following an
implusive acceleration, the total circulation on the interface decreases because
of baroclinic effects and that the vorticity localizes at a centroid, Y
c
(A, M)/L
T
which approaches monotonically to a constant that is 0.4 for 0.3 <A<0.9 and
M = 0 (incompressible) and decreases as M increases, as shown in Figure 8.
It is observed that the decrease in Y
T
is large due to nite M (compressibility)
at A = 0.5 and M = 3.0.
Visit the Annual Reviews home page at
http://www.AnnualReviews.org
Literature Cited
Akhmediev N, Ankiewicz A. 1997. Solitons:
Nonlinear Pulses and Beams. Chapman and
Hall.
Aleshin A, Chebotareva E, Krivets VV,
Lazareva E, Sergeev S, et al. 1996. Investi-
gation of evolution of interface after its inter-
action with shock waves. Tech. Rep., Int. Inst.
App. Physics High Technol. Rep. Nos. 02-96,
06-96
Aleshin A, Zaitsev S, Lazareva E. 1993. In-
vestigation of Richtmyer-Meshkov instabil-
ity. Res. J. Continuum Mech. 1:33
Andronov VA, Bakhrakh SM, Meshkov EE,
Mokhov VN,

Nikiforov VV, et al. 1976. Tur-
bulent mixing at a contact surface accelerated
by shock waves. Sov. Phys. JETP 44:424
Aschenbach B, Egger R, Trumper J. 1995. Dis-
covery of explosion fragments outside the
Vela supernova remnant shock-wave bound-
ary. Nature 373:58790
Baker G, Caisch R, Siegel M. 1993. Singu-
larity formation during the Rayleigh-Taylor
instability. J. Fluid Mech. 252:5178
Baker GR, Shelly M. 1990. On the connection
betweenthinvortexlayers andsheets. J. Fluid
Mech. 215:16194
Banks D, Singer B. 1994. Vortex tubes in tur-
bulent ows: identication, representation,
reconstruction. In Proc. Visualization 94
IEEE, pp. 13239. Computer Society
Ben Dor G, Takayama K, Kawauchi T. 1980.
The transitionfromregular toMachreection
in truly non-stationary ows. J. Fluid Mech.
100:14760
Berning M, Rubenchik AM. 1998. A weakly
nonlinear theory for the dynamical Rayleigh-
Taylor instability. Phys. Fluids 10:156487
Bitz F, Zabusky N. 1990. David and Visiomet-
rics: visualizing and quantifying evolving
amorphous objects. Comput Phys. Nov/Dec:
60314
Bourne N, Field JE. 1992. Shock-induced col-
lapse of single cavities in liquids. J. Fluid
Mech. 244:22540
Brouillette M, Sturtevant B. 1989. Richtmyer-
Meshkov instability at a continous interface.
Current Topics inShock Waves: 17thInterna-
tional Symposiumon Shock Waves and Shock
Tubes, AIP Conf. Proceedings, 208, ed. YW
Kim, pp. 28489
Budzinski JM, Benjamin R, Jacobs JW. 1994.
Inuence of initial conditions on the owpat-
terns of a shock accelerated thin uid layer.
Phys. Fluids 6:3510
Bullough RK, Caudrey PJ. 1995. Solitons and
the Korteweg-de Vries equation: integrable
systems in 18341995. See Hazewinkel et al
1995, 39:193228
Buntine JD, Pullin DI. 1989. Merger and can-
cellation of strained vortices. J. Fluid Mech.
205:26395
Carlbom I, Chakravarty I, Hsu W. 1992. Inte-
grating computer graphics, computer vision
and image processing in scientic applica-
tions. Comput. Graph. 26(1):816
Chalmers J, Hodson S, Winkler K-H, Wood-
ward P, Zabusky N. 1988. Shock-bubble in-
teractions: generation and evolution of vor-
ticity in two-dimensional supersonic ows.
Fluid Dyn. Res. 3:39294
Chandrasekhar S. 1961. Hydrodynamic and Hy-
dromagnetic Stability, Sec. 101, Eq. 29. Ox-
ford: Oxford Univ. Press
Chevalier RA, Blondin JM. 1995. Hydro-
dynamic instabilities in supernova rem-
nants: early radiative cooling. Astrophys. J.
444:31217
Chevalier RA, Blondin JM, Emmering RT.
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

1992. Hydrodynamic instabilities in super-
nova remnants: self-similar driven waves.
Astrophys. J. 392:11830
Colella P. 1985. A direct Eulerian MUSCL
scheme for gas dynamics. SIAM J. Sci. Stat.
Comput. 6:10417
Colella P, Ferguson RE, Glaz HM, Kuhl A.
1986. Mach reection from HE-driven blast
wave. In Dynamics of Explosions, ed. JR
Bowen, J-C Leyer, RI Soloukhin, pp. 388
421. AIAA, Vol. 106
Couder Y, Basdevant C. 1986. Experimental
and numerical study of vortex dipoles in two-
dimensional ows. J. Fluid Mech. 173:225
51
Curran E, Heiser W, Pratt D. 1996. Fluid phe-
nomena in scramjet combustion systems.
Annu. Rev. Fluid Mech. 28:32360
Deem G, Zabusky N, Kruskal M. 19651975.
Formation, propagation and interaction of
solitons in nonlinear dispersive media (nu-
merical solutions of differential equations de-
scribing wave motion in nonlinear dispersive
media. Bell Telephone Laboratories Film Li-
brary, Murray Hill, NJ
Dohm-Palmer RC, Jones TW. 1996. Young su-
pernova remnants in non-uniform media. As-
trophys J. 471:279
Don W-S, Quillen CB. 1995. Numerical simula-
tion of shock cylinder interactions. J. Comp.
Phys. 122:24465
Duff RE, Harlow F, Hirt C. 1962. Effects of dif-
fusion on interface instability between gases.
Phys. Fluids 5:41725
Farge M, Kevlahan N, Bardos C, Schneider K.
1998. Combining deterministic and statisti-
cal approaches to compute two dimensional
turbulent ows with walls. In Monte Verita
Lectures on Turbulence, ed. A Tsinober, A
Gyr. Birkhauser. In press
Fedorov KN, Ginsburg AI. 1989. Mushroom-
like currents (vortex dipoles): on the most
wide-spread forms of non-stationary cohe-
rent motion in the ocean. In Mesoscale/Syn-
optic Coherent Structures in Geophysical
Turbulence, pp. 115. Amsterdam: Elsevier
Feher A, Zabusky N. 1996. An interactive imag-
ing environment for scientic visualization
and quantication. Int. J. Imaging Syst. Tech-
nol. 7:12130
Fernandez V, Zabusky N, Gryanik V. 1995.
Vortex intensication and collapse of the
Lissajous-elliptic ring: single and multi-
lament Biot-Savart simulations and visio-
metrics. J. Fluid Mech. 299:2289331
Fraley G. 1986. Rayleigh-Taylor stability for a
normal shock wave-density discontinuity in-
teraction. Phys. Fluids 29:376
Gardner C, Greene J, Kruskal M, Miura R. 1967.
Method for solving the Kortewegde Vries
equation. Phys. Rev. Lett. 19:109597
Haas J, Sturtevant B. 1987. Interaction of weak
shock waves with cylindrical and spherical
gas inhomogeneities. J. Fluid Mech. 181:41
76
Hasegawa A. 1995. Optical solitons in commu-
nication: fromintegrability to controllability.
See Hazewinkel et al 1995, 39:8590
Hasegawa A, Tappert F. 1973. Transmission of
stationary, nonlinear optical pulses in dis-
persive dielectric bers. Appl. Phys. Lett.
23:17172
Hasimoto H, Kodama Y. 1995. Solitons in Op-
tical Communications. Oxford: Clarendon
Hawley J, Zabusky N. 1989. Vortex paradigm
for shock accelerated density stratied inter-
faces. Phys. Rev. Lett. 63:124144
Hazewinkel M, Capel HW, de Jager EM, eds.
1995. Acta Applicandae Mathematicae. In
Proc. KdV Symp., 1995, Amsterdam. Dor-
drecht, The Netherlands: Kluwer Academic
Henderson L. 1989. On the refraction of shock
waves. J. Fluid Mech. 198:365
Henderson LB. 1966. The refraction of a plane
shock wave at a gas interface. J. Fluid Mech.
26:60737
Hesselink L. 1988. Digital image processing in
ow visualization. Annu. Rev. Fluid Mech.
20:42185
Hewett JS, Madnia C. 1998. Flame-vortex inter-
action in a reacting vortex ring. Phys. Fluids
10:189205
Heyerhoff M. 1998. Die fr uhe Geschichte der
Solitonentheorie. In Arch. Hist. Exact Sci.
Berlin: Springer-Verlag. In press
Holmes RL, Dimonte G, Fryxell B, Gittings
ML, Grove JW, et al. 1998. Richtmyer-
Meshkovinstabilitygrowth: experiment sim-
ulation and theory. J. Fluid. Mech. In press
Holmes RL, Grove JW, Sharp DH. 1995. Nu-
merical investigation of Richtmyer-Meshkov
instabilityusingfront tracking. J. FluidMech.
301:5164
Hornung H. 1985. Regular and Mach reec-
tion of shock waves. Annu. Rev. Fluid Mech.
18:3358
HouT. 1995. Numerical solutions tofree bound-
ary problems. Acta Numerica, pp. 335415
Hou T, Lowengrub JS, Shelley M. 1997. The
long-time motion of vortex sheets with sur-
face tension. Phys. Fluids 9:193354
Ilkaev RI, Meshkov E, Nikiforov V. 1997. Hy-
drodynamic instabilities research at VNIIEF:
past, current andfuture. See Jourdan&Houas
1997, pp. 2037
Jacobs JW. 1992. Shock induced mixing of a
light-gas cylinder. J. Fluid. Mech. 234:629
49
Jacobs JW, Jenkins DG, Klein DL, Benjamin
RF. 1995. Nonlinear growth of the shock-
accelerated instability of a thin uid layer.
J. Fluid. Mech. 295:2342
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

534 ZABUSKY
Jacobs JW, Niederhaus CE. 1997. PLIF ow
visualization of single- and multi-mode in-
compressible Richtmyer-Meshkov instabil-
ity. See Jourdan & Houas 1997, pp. 21419
Jacobs JW, Sheely JM. 1996. Experimental
study of incompressible Richtmyer-Meshkov
instability. Phys. Fluids 8:40515
Jahne B. 1991. Digital Image Processing.
Berlin. Springer-Verlag
Jeong J, Hussain F, Schoppa W, Kim J. 1997.
Coherent structures near the wall in a turbu-
lent channel ow. J. Fluid Mech. 332:185
214
Jourdan G, Houas L, eds. 1997. Proc. 6th Int.
Workshop on Compressible Turbulent Mix-
ing, Marseilles, France, 1997. Imprimerie
Caractere
Kang H, Ostriker JP, Ryu RCD, Hernquist L,
Evrard AE, et al. 1994. A comparison of cos-
mological hydrodynamic codes. Astrophys. J.
430:83100
Kevlahan NK. 1997. The vorticity across a
shock in a non-uniform ow. J. Fluid Mech.
341:37184
Klein R. 1998. Three dimensional simulations
of the shock sphere interaction, private com-
munication. Astrophys. Suppl.
Klein R, McKee CF, Colella P. 1994. On the hy-
drodynamic interaction of shock waves with
interstellar clouds. I. Nonradiative shocks in
small clouds. Astrophys. J. 420:21336
Korteweg DJ, de Vries G. 1895. On the change
of form of long waves advancing in a rectan-
gular canal, and on a new type of long sta-
tionary waves. Phil. Mag. 39:42243
Kruskal MD. 1995. An ODE to a PDE: glo-
ries of the KdV equation. An appreciation
of the equation on its 100th birthday. See
Hazewinkel et al 1995, 39:15972
Lee CKB, Giltrud M. 1996. An evaluation of
two classes of shock hydrodynamic codes.
Shock Waves 5:27585
Lewis DJ. 1950. The instability of liquid sur-
faces when accelerated in a direction perpen-
dicular to their planes. II. Proc. R. Soc. A
202:81196
Lim TT, Nickels TB. 1992. Instability and re-
connection in the head-on collision of two
vortex rings. Nature 357:22527
Lindl J. 1995. Development of the indirect drive
approach to inertial connement fusion and
the target physics for ignition and gain. Phys.
Plasmas 2:39334024
Lorensen W, Cline H. 1987. Marching cubes:
a high resolution 3D surface construction al-
gorithm. Comput. Graphics 21:16370
Meshkov E. 1969. Instability of a shock wave
accelerated interface between two gases.
NASA Tech. Trans. TT-F-13074, Izv. Akad.
Nauk, SSSR, Mekh. Zhidk. Gaza
Meyer K, Blewett P. 1972. Numerical investi-
gation of the stability of a shock-accelerated
interface between two uids. Phys. Fluids
15:75359
Mikaelian K. 1994a. Oblique shocks and
the combined Rayleigh-Taylor, Kelvin-
Helmholtz, and Richtmyer-Meshkov insta-
bilities. Phys. Fluids 6:194345
Mikaelian KO. 1994b. Freeze-out and the ef-
fect of compressibility in the Richtmyer-
Meshkov instability. Phys. Fluids 6:356
68
Muller E, Fryxell B, Arnett D. 1991. Instability
and clumping in SN 1987a. Astron. Astro-
phys. 251:50514
Neuvazhayev VE, Parshukov IE. 1997. Non-
linear stage in the development of perturba-
tions at a shocked interface. See Jourdan &
Houas 1997, pp. 36873
ODell CR, Handron K. 1996. Cometary knots
in the helix nebula. Astron. J. 111:163039
Ofer D, Alon U, Hecht J, Shvarts D. 1995.
Power laws and similarity in Rayleigh-Taylor
and Richtmyer-Meshkov mixing fronts at all
density ratios. Phys. Rev. Lett. 74(4):53437
Ofgeneim D, Timofeev E, Voinovich P. 1995.
See Sturtevant et al 1995, pp. 53742
Osher S, Mulder W, Sethian J. 1988. Computing
interface motion in compressible gas dynam-
ics. J. Comp. Phys. 100:20928
Patrikalakis N, ed. 1991. Reconstructing and
Visualizing Models of Neuronal Dendrites.
Berlin: Springer-Verlag
Picone J, Boris JP. 1988. Vorticity generation by
shock propagation through bubbles in a gas.
J. Fluid Mech. 189:2351
Puckett E, Colella P, Henderson L. 1991. On the
refraction of shock waves at a slow-fast gas
interface. J. Fluid Mech. 224:127
Pullin DI. 1980. Direct simulation methods for
compressible ideal gas ows. J. Comp. Phys.
34:231
Quirk JJ, Karni S. 1996. On the dynamics of
a shock-bubble interaction. J. Fluid Mech.
318:12963
Ray J, Zabusky N, Samtaney R. 1995. Vor-
tex methods for early and intermediate
time Richtmyer-Meshkov environments. See
Sturtevant et al 1995, pp.
Remington B, Arnett D. 1999. Proc. 2nd Int.
Workshop Lab. Astrophys. Intense Lasers,
1998. In Astrophysical Supplement, ed. B
Remington, D Arnett. In press
Richtmyer R. 1960. Taylor instability in shock
acceleration of compressible uids. Com-
mun. Pure Appl. Math. XIII:297
Rightley PM, Vorobieff P, Benjamin RF. 1997.
Evolution of a shock-accelerated thin uid
layer. Phys. Fluids 9:1770
Rudinger G, Somers LM. 1960. Behavior of
small regions of different gases carried in ac-
celerated gas ows. J. Fluid Mech. 7:16176
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

Rupert V. 1992. Shock-interface interaction:
current research on the Richtmyer-Meshkov
problem. See Takayama 1992, pp. 8394
Saffman P. 1992. Vortex Dynamics. Cambridge
Univ. Press
Samtaney R. 1993. Vorticity in shock-accele-
rated density-stratied interfaces: an anal-
ytical and computational study. PhD thesis,
Rutgers Univ., New Brunswick, NJ. 219 pp.
Samtaney R. 1997. Computational methods for
self-similar solutions of the compressible Eu-
ler equations. J. Comp Phys. 132:32745
Samtaney R, Meiron DI. 1997. Hypervelocity
Richtmyer-Meshkov instability. Phys. Fluids
9:17831803
Samtaney R, Pullin DI. 1996. On initial-value
and self-similar solutions of the compressible
Euler equations. Phys. Fluids 8:265055
Samtaney R, Ray J, Zabusky NJ. 1998. Baro-
clinic circulation generation on shock accel-
erated slow/fast gas interfaces. Phys. Fluids
10:121730
Samtaney R, Silver D, Zabusky N, Cao J. 1994.
Visualizing features and tracking their evolu-
tion. IEEE Comp. 27(7):2027
Samtaney R, Zabusky NJ. 1992. Visiometrics
and reduced models for vorticity deposition
in shock interactions with heavy cylindrical
bubbles. CAIPTech. Rep. Rutgers Univ., New
Brunswick, NJ
Samtaney R, Zabusky N. 1994. Circulation
deposition on shock-accelerated planar and
curved density stratied interfaces: models
and scaling laws. J. Fluid Mech. 269:45
78
Scheidegger T. 1997. On compressibility effects
in two and three dimensional ows: vortex
dipoles and reconnection. PhD thesis, Rut-
gers Univ., New Brunswick, NJ
Sharp DH. 1984. An overview of the Rayleigh-
Taylor instability. Physica D 12:3
Siegel A, Weiss JB. 1997. A wavelet-packet
census algorithmfor calculating vortex statis-
tics. Phys. Fluids 9:198899
Silver D. 1995. Object-oriented visualization.
IEEE Comput. Graph. Appl. 15
Silver D, Wang X. 1996. Volume tracking. In
Proc. IEEE Visualization 96, San Francisco,
CA. pp. 15764
Silver D, Zabusky N. 1993. Quantifying visu-
alizations for reduced modeling in nonlinear
science: extracting structures from data sets.
J. Vis. Commun. Image Represent. 4:4661
Smoller J. 1982. Shock Waves and Reaction Dif-
fusion Equations. Springer-Verlag
Stolen R, Mollenauer L, Gordon J. 1980. Phys.
Rev. Lett. 45:1095
Stone JM, Norman ML. 1994. Numerical sim-
ulations of protostellar jets with nonequilib-
rium cooling. III. three-dimensional results.
Strang G. 1968. On the construction and com-
parison of difference schemes. SIAM J. Nu-
mer. Anal. 5:50617
Sturtevant B. 1987. Rayleigh-Taylor instability
in compressible uids. In Proc. Int. Symp.
Shock Tubes Waves, 16th, Aachen, ed. H
Gr onig, pp. 89100. Basel: VCH Verlag
Sturtevant B, Shepherd JE, Hornung H, eds.
1996. Shock waves. In Proc. Int. Symp. Shock
Waves, 20th, Pasadena, California, 1995, Vol
1. Singapore: World Scientic
Takayama K, ed. 1992. Shock waves. In Proc.
Int. Symp. Shock Waves, 18th, Sendai, Japan,
1991. New York: Springer-Verlag
Taylor G. 1950. The instabilityof liquidsurfaces
when accelerated in a direction perpendicu-
lar to their planes. I. Proc. R. Soc. A201:192
96
Tryggvason G, Dahm W, Sbeih K. 1991. Fine
structure of vortexsheet rollupbyviscous and
inviscid simulation. J. Fluids Eng. 113:3136
Upson C, Keeler M. 1988. V-BUFFER: vis-
ible volume rendering. Comput. Graph.
22(4):5964
van Leer B. 1977. Towards the ultimate con-
servative difference scheme. IV. A new ap-
proach to numerical convection. J. Comput.
Phys. 23:27699
Velikovich AL, Dimonte G. 1996. Nonlinear
perturbation theory of the incompressible
Richtmyer-Meshkov instability. Phys. Rev.
Lett. 76:3112
Villasenor J, Vincent A. 1992. An algorithm for
space recognition and time tracking of vor-
ticity tubes in turbulence. CVGIP: Image Un-
derst. 55:2735
Wang Y-W, Robertson JA. 1985. Late stages of
the Rayleigh-Taylor instability: a numerical
study in the context of accreting stars. Astro-
phys. J. 299:85108
Weissert T. 1997. The Genesis of Simulationand
Dynamics: Pursuing the Fermi-Pasta-Ulam
Problem. New York: Springer-Verlag
Winkler KH, Chalmers J, Hodson S, Woodward
P, Zabusky N. 1987. A numerical laboratory.
Phys. Today 40(10):2837
Wouchuk JG, Nishihara K. 1997. Asymptotic
growth in the linear Richtmyer-Meshkov in-
stability. Phys. Plasmas 4:102838
Xu J, Stone JM. 1995. The hydrodynamics of
shock-cloud interactions in three dimensions.
Yamada K, Kodama T, Takayama K. 1995.
Mechanisms of shock wave generation re-
sulting from the interaction between a shock
wave and a gas bubble in a liquid. See Sturte-
vant et al 1996, pp. 56974
Yang X, Chern IL, Zabusky N, Samtaney R,
Hawley J. 1992. Vorticity generation and evo-
lution in shock-accelerated density-stratied
interfaces. Phys. Fluids A 4:153140
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

536 ZABUSKY
Yang X, Zabusky NJ, Chern I-L. 1990. Break-
through via dipolar-vortex/jet formation in
shock-accelerated density-stratied layers.
Phys. Fluids A 2:89295
Yang Y, Zhang Q, Sharp D. 1994. Small ampli-
tude theory of Richtmyer-Meshkov instabil-
ity. Phys. Fluids 6(5):185673
Youngs DL. 1994. Numerical simulation of
mixing by Rayleigh-Taylor and Richtmyer-
Meshkov instabilities. Laser Part. Beams
12:72550
Zabusky N. 1977. Coherent structures in uid
dynamics. In The signicance of nonlinear-
ity in the natural sciences. pp. 145205. New
York: Plenum
Zabusky N. 1981. Computational synergetics
and mathematical innovation. J. Comp. Phys.
43:195249
Zabusky N. 1984. Computational synergetics.
Phys. Today 37:3646
Zabusky N, Kruskal M. 1965. Interaction of
Solitons in a collisionless plasma and the
recurrence of initial states. Phys. Rev. Lett.
15:140243
Zabusky N, Samtaney R, Yang X, Chern I-L.
1992. Vorticity in shock-accelerated density-
stratied interfaces: deposition and interme-
diate time evolution of coherent structures.
See Takayama 1992, pp. 1928
Zabusky N, Silver D, Pelz R, Vizgroup 93.
1993. Visiometrics, juxtaposition and mod-
eling. Phys. Today pp. 2431
Zabusky N, Yang X. 1989. Emerging vortex
structures and compressible turbulence fol-
lowing shock-layer interactions. In Annual
Research Report, pp. 217220. Princeton NJ
John von Neumann National Supercomputer
Center
Zabusky N, Zeng S, Ray J, Samtaney R. 1997.
Vortical projectiles from shock-bubble inter-
actions. See Jourdan &Houas 1997, pp. 539
44
Zabusky NJ. 1995. Coherent structure visio-
metrics: from the soliton to HEC. See
Hazewinkel et al 1995, 39:12732
Zabusky NJ, Zeng S. 1998. Shock cavity im-
plosion morphologies and vortical projectile
generation in axisymmetric shock-spherical
fast/slow bubble interactions. J. Fluid. Mech.
362:32746
Zaitsev SG, Lazareva EV, Chernukha VV,
Belyaev VM. 1985. Intensication of mix-
ing at the interface between media of differ-
ent densities upon passage of a shock wave
through it. Phys. Dokl. 30:579
Zaytsev S, Aleshin A, Lazareva E, Rozanov V,
Lebo I. 1993. Richtmyer-Meshkov instabil-
ity for two- and three-dimensional interfaces.
In Proc. Int. Workshop Phys. Compressible
Turbulent Mixing, 4th, Cambridge, UK, pp.
29196. London: Cambridge Univ. Press
Zhang Q, Sohn S-I. 1997. Nonlinear theory of
unstable uid mixing driven by shock wave.
Phys. Fluids 4:110624
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
Annual Review of Fluid Mechanics
Volume 31, 1999
CONTENTS
Linear and Nonlinear Models of Aniosotropic Turbulence, Claude
Cambon, J ulian F. Scott
1
Transport by Coherent Barotropic Vortices, Antonello Provenzale 55
Nuclear Magnetic Resonance as a Tool to Study Flow, Eiichi Fukushima 95
Computational Fluid Dynamics of Whole-Body Aircraft, Ramesh
Agarwal
125
Liquid and Vapor Flow in Superheated Rock, Andrew W. Woods 171
The Fluid Mechanics of Natural Ventilation, P. F. Linden 201
Flow Control with Noncircular J ets, E. J . Gutmark, F. F. Grinstein 239
Magnetohydrodynamics in Materials Processing, P. A. Davidson 273
Nonlinear Gravity and Capillary-Gravity Waves, Frdric Dias,
Christian Kharif
301
Fluid Coating on a Fiber, David Qur 347
Preconditioning Techniques in Fluid Dynamics, E. Turkel 385
A New View of Nonlinear Water Waves: The Hilbert Spectrum, Norden
E. Huang, Zheng Shen, Steven R. Long
417
Planetary-Entry Gas Dynamics, Peter A. Gnoffo 459
VORTEX PARADIGM FOR ACCELERATED INHOMOGENEOUS
FLOWS: Visiometrics for the Rayleigh-Taylor and Richtmyer-Meshkov
Environments, Norman J . Zabusky
495
Collapse, Symmetry Breaking, and Hysteresis in Swirling Flows,
Vladimir Shtern, Fazle Hussain
537
Direct Numerical Simulation of Free-Surface and Interfacial Flow, Ruben
Scardovelli, Stphane Zaleski
567
A
n
n
u
.

R
e
v
.

F
l
u
i
d

M
e
c
h
.

1
9
9
9
.
3
1
:
4
9
5
-
5
3
6
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

I
m
p
e
r
i
a
l

C
o
l
l
e
g
e

L
o
n
d
o
n

o
n

1
0
/
0
9
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

RMI Zabuskey 1999

Uploaded by

Copyright:

Available Formats

RMI Zabuskey 1999

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

RMI Zabuskey 1999

Uploaded by

Copyright:

Available Formats

P1: APR/NBL/spd/mbg P2: NBL/plb/ary QC: NBL

November 17, 1998 18:37 Annual Reviews AR075-14

(x, r, t ) dr, (4)

= (1/2)( ||), are either the positive or negative values of the

< 0 that coincide with the

+ 1) is the post-shock Atwood number (i.e.

is the Atwood number based on post-shock densities, = k a

in 1 < M < 1.5 and then approaches

at M = 4, and for He-Xe ( = 32.7,

/ds = [cos / cos(

. The light gray

is smaller (or larger) than 10% of

. The white spaces show

, the error is less than 10% for 1.0 < M 1.32

)/( + 1), a heuristic

); SMBS (); and LV, (). For Air/R-22, = 3.0, ka

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.