The physics of aerobreakup. II.

Viscous
liquids
Cite as: Phys. Fluids 24, 022104 (2012); https://doi.org/10.1063/1.3680867
Submitted: 06 June 2011 • Accepted: 14 December 2011 • Published Online: 14 February 2012

T. G. Theofanous, V. V. Mitkin, C. L. Ng, et al.

ARTICLES YOU MAY BE INTERESTED IN

On the physics of aerobreakup

Physics of Fluids 20, 052103 (2008); https://doi.org/10.1063/1.2907989

Plane shock wave interaction with a cylindrical water column

Physics of Fluids 28, 056102 (2016); https://doi.org/10.1063/1.4948274

The physics of aerobreakup. III. Viscoelastic liquids

Physics of Fluids 25, 032101 (2013); https://doi.org/10.1063/1.4792712

Phys. Fluids 24, 022104 (2012); https://doi.org/10.1063/1.3680867 24, 022104

© 2012 American Institute of Physics.

 PHYSICS OF FLUIDS 24, 022104 (2012)

The physics of aerobreakup. II. Viscous liquids

T. G. Theofanous,a) V. V. Mitkin, C. L. Ng, C-H. Chang,
X. Deng, and S. Sushchikh
Chemical Engineering Department and Center for Risk Studies and Safety,
University of California, Santa Barbara, California 93106, USA
(Received 6 June 2011; accepted 14 December 2011; published online 14 February 2012;
publisher error corrected 16 February 2012)

We extend the work of Theofanous and Li [“On the physics of aerobreakup,” Phys.
Fluids 20, 052103 (2008)] on aerobreakup physics of water-like, low viscosity liquid
drops, to Newtonian liquids of any viscosity. The scope includes the full range of
aerodynamics from near incompressible to high Mach number flows. The key physics
of Rayleigh–Taylor piercing (RTP, first criticality) and of shear-induced entrainment
(SIE, second and terminal criticality) are verified and quantified by new viscosity-
and capillarity-based scalings for fluids of any viscosity. The relevance and predictive
power of linear stability analysis of the Rayleigh–Taylor and Kelvin–Helmholtz
problems (both including viscosity) is demonstrated for the RTP and the SIE regimes,
respectively. The advanced stages of breakup and of the resulting particle-clouds are
observed and clear definition and quantification of breakup times are offered.  C 2012

American Institute of Physics. [http://dx.doi.org/10.1063/1.3680867]

I. INTRODUCTION
The purpose of this paper is to extend the work of Theofanous and Li1 on aerobreakup physics
of water-like, low viscosity liquid drops, to Newtonian liquids of any viscosity. As in the previous
work, the scope includes the full range of aerodynamics from near incompressible to high Mach
number flows, including shock waves. The effect of viscosity, however little studied in the past, is
paramount, not only in retarding breakup but also in providing a new pathway to approaching the
operative basic physics of this complex, multiscale process.
Starting with Hinze,2 viscosity has been incorporated by means of the Ohnesorge number (Oh),
and the retardation of breakup noted above has been expressed by the increase of the critical Weber
number (Wec ) with increasing Oh, as in the well-known Brodkey correlation,3
μl ρg u 2g d0
W ec = W ec,Oh→0 (1 + 1.077 Oh 1.6 ) 10−2 < Oh < 2, Oh = , W e = ,
(ρl σ d0 )1/2 σ
(1)
where ug is the free-stream gas velocity, d the drop diameter, ρ g /ρ l the gas/liquid densities, σ the
surface tension coefficient, and μl liquid dynamic viscosity, and subscript 0 is to refer to an initial
value. According to Gelfand,4 the Oh exponent in (1) is 0.74, and the issue of extrapolating outside
the applicable range has been raised.5
Joseph et al.6 demonstrated that highly viscous drops could be broken by Mach-2 and Mach-3
shocks, but visualization was too limited to discern patterns and no attempt was made to investigate
critical conditions. Instead, this work was focused on what Harper et al.7 called the catastrophic
regime, which is now known to have been an artifact of the visualization method employed.1, 8
The first, and perhaps only theoretical consideration of viscosity effects on deformation as the
operative pathway to aerobreakup is credited to Hinze.9 Revisiting the subject a few years later, along
with his experimental data (in the range noted above), he concluded2 that breakup would cease at

a) Electronic mail: [email protected].

1070-6631/2012/24(2)/022104/39/$30.00 24, 022104-1 

C 2012 American Institute of Physics
022104-2 Theofanous et al. Phys. Fluids 24, 022104 (2012)

Oh ∼ 2. Rather, recently Hsiang and Faeth10 based on their own data taken over 10−2 < Oh < 500
placed this limit to Oh ∼ 4. In simple terms, the idea was9 that with slow-enough deformation, due to
viscosity, drop acceleration, and loss of relative velocity would eventually result in the drop missing
the window of opportunity for breakup. In this paper, we demonstrate by experiment, and by a theory
that accounts for viscosity effects on deformation and for finite-domain effects on Rayleigh–Taylor
instability, that this criticality is attainable at any viscosity.
The above refers to the first criticality. It embodies the formation of a bag, growing suddenly
and unstably on the windward face of a flattened drop. This is the first mode of Rayleigh–Taylor
piercing (RTP);1 a rather broad range of sub-regimes that include higher modes (cylindrical and
three-dimensional multiple bags), but with a common feature that breakup is governed by gas
penetrating and passing through the drop. A significant change in breakup pattern occurs when, with
increasing flow-dynamic pressure, the penetration process ceases as it gives way to a fragmentation
process occurring in the outer layers of the windward surface of the drop, in a peeling-like action
driven by the gas as it flows around. This shear-induced entrainment (SIE) has been identified as
the second and terminal criticality1 in low-viscous liquids, a conclusion that we will demonstrate to
carry over to liquids of any viscosity as well. As is the case for the first criticality, viscosity is found
to retard this process (higher dynamic pressures needed to attain the same breakup mechanism with
increasing viscosity), but otherwise the behaviors are very similar.
In both criticalities, a new kind of viscosity- and capillary-based scaling is necessary to cap-
ture the key physics,8 and this is addressed in the course of presenting the experimental data
(Secs. IV and V A). At a more basic level, and with the help of direct numerical simulations, we
show that the whole range of behaviors pivots around a dominant role of viscous Kelvin–Helmholtz
(K-H)1 instability (Sec. V B). (The essential role of viscosity in real-fluid layers has been pointed
out by Drazin and Reid,27 Yih,28 and Boomkamp and Miesen29 among many others and we include
viscosity in our treatment here. Following this clarification, the term K-H instability will be under-
stood to convey this sense.) Only when K-H instabilities are screened by the drop length scale, are
Rayleigh–Taylor (R-T) instabilities able to manifest, again screened by the drop length scale, but
incorporating increases due to cross-stream deformation (Sec. IV), and on much longer time scales.
Additional insights are provided by consideration of breakup times (Sec. VI), and brief illustration
of resulting particle-cloud dynamics and particle size patterns within these clouds (Sec. VII). We
begin with a summary of the experimental and numerical methods employed in this work (Sec. II
and Sec. III, respectively).

II. EXPERIMENTAL METHODS

The facilities and methods utilized in this study are an outgrowth of those detailed in our previous
work with the ALPHA facility1, 11 —a pulse supersonic wind tunnel, designed to operate from mildly
sub-atmospheric to rarefied-gas conditions. In the present work, we have employed ASOS—a shock
tube, designed to operate from mildly pressurized to gas-dynamic pressures of up to 2 MPa. The
expansion section of ASOS is exactly of the same geometry and construction as the ALPHA flow
channel (200 × 200 mm2 in flow cross sectional area), except for being specially engineered to
withstand the dynamic loadings at the facility-design envelop with fully transparent (polycarbonate)
construction. This design envelope was defined as the release of 10 MPa helium gas pressure from
the driver section of the shock tube. It yields a shock Mach number of MS = 3.5, and a gas-dynamic
pressure of Pd = 2 MPa. The 2-m-long driver section ensures that uniform flow conditions exist over
the time periods (∼8/3 ms with N2 /He in the driver) necessary to attain and observe break up with
highly viscous liquids as well as with visco-elastic liquids.12–14 The 6.5-m-long expansion section
allows for a 4.6-m-long flow-development section ahead of the drop-injection port. The 2-m-long
test section allows the attainment of near-equilibrium clouds, and improved resolutions have been
obtained by the installation of optical-quality windows at four locations selected on the basis of
extensive testing with the fully transparent test section.
The shock tube is supported in a horizontal orientation, it is operated by a double-diaphragm
bursting arrangement, and it is made to vent into a large chamber with controlled and filtered venting
to the outside (for testing with liquids that need to be fully contained). These features are similar
022104-3 The physics of aerobreakup. II. Viscous liquids Phys. Fluids 24, 022104 (2012)

FIG. 1. (Color online) The experimentally accessible regions in the ALPHA and the ASOS facilities.

to ALPHA, and so is the technology for injecting drops of various liquids and timing the sequence
of operations in an experimental run. The operating gas-dynamic ranges of the two facilities are
illustrated in Figure 1. In ALPHA, a similar dynamic pressure as in ASOS yields a much lower
Reynolds number and this affects flow separation, and the drag coefficient, with attendant changes in
drop response (Sec. IV); however, when ALPHA is run as a shock tube (at similar static pressures), the
results are indistinguishable to those obtained in ASOS (for example Figures 3 and 4 of Theofanous
and Li1 ).
Tests reported herein were conducted with a series of silicon oils (SO) of varying viscosities,
glycerol (G), water (W), and tri-butyl phosphate (TBP), an extremely low vapor pressure liquid
with surface tension (∼28 mN/m) similar to that of the silicon oils (∼21 mN/m), and a viscosity
(4 cSt) similar to that of water (1 cSt). We have computed upper bounds of drop-surface heat up by
ignoring internal liquid motions (Appendix A) and concluded that for all practical purposes in data
analysis these properties can be taken at the above-quoted values. In the case of glycerol special
care is taken to minimize water contamination (affecting both surface tension and viscosity of the
outer drop layers) from atmospheric humidity. Initial drop sizes were in the range of 1.6 to 3.5 mm.
The run ID is provided in the following form: Run [Fluid-Name, M = xx, Reg = xx, We = xx,
Oh = xx, Camera (view)] where M is the flow Mach number (flow speed divided by the speed
of sound at free-stream conditions), and Reg is the gas Reynolds number based on free-stream
conditions (Reg = d0 ug ρ g /μg ). Unless otherwise noted (as for ALPHA or MuSiC+ runs) this ID
refers to ASOS runs. The scope of the experimental program is summarized in Table I. In this table,
MS is the shock Mach number (shock speed divided by the speed of sound ahead of the shock). The
viscous and inertia timescales, tV+0 = (νl /a02 )1/3 [ν l is the liquid kinematic viscosity, and a0 is the
initial value of drop acceleration] and t I+ = ud0g (ρl /ρg )1/2 , are to define corresponding scaled times
TI = t/t I+ and TV 0 = t/tV+0 as needed in Secs. IV–VI.
Our overall strategy in experimentation consists of two main thrusts: (a) vary fluids and dynamic
pressure in such a manner as to focus on the onset of the two criticalities (RTP and SIE) over as broad
an Oh-range as possible, and (b) vary the dynamic pressure at three specific values of the Oh so as
to scrutinize the detailed structure of regime transitions as well as behaviors at highly supercritical
SIE conditions. These values were selected (2 × 10−3 , 2 × 10−2 , and ∼2) so as to span the inviscid,
and viscous extremes—as detailed in Secs. III and IV, the transition between them being found at
2 × 10−1 < Oh < 2 and 2 × 10−2 < Oh < 2 × 10−1 for RTP and SIE regimes, respectively.
The principal measurement task is that of visualization, and the essential advance in this regard
is the introduction of laser-induced fluorescence (LIF) as the means for illumination.1 Applied in
conjunction with exposure times of ∼20 ns (copper vapor laser, Oxford lasers, model LS20-50) and
022104-4 Theofanous et al. Phys. Fluids 24, 022104 (2012)

TABLE I. Experiment ranges employed and values of the key characteristic times.

Liquid, cSt We Oh MS Reg tI + , s tV0 + , s

SO1 1.5 × 105 0.0044 2.3 1.5 × 105 7 × 10−5 3.3 × 10−6
W1 60 − 3 × 104 2 × 10−3 1.1 − 2.1 6 × 103 −105 10−4 − 10−3 3.5 × 10−6 −
2.3 × 10−4
TBP4 12 − 1.5 × 105 2 × 10−2 1.02 − 2.5 103 − 2 × 105 7 × 10−5 − 3.4 × 10−6 −
4 × 10−3 1.6 × 10−3
SO5 170 3 × 10−2 1.1 4 × 103 10−3 3.4 × 10−4
SO10 310 5 × 10−2 1.1 6 × 103 10−3 2.8 × 10−4
SO20 330 10−1 1.1 6 × 103 10−3 3.2 × 10−4
SO50 300 − 2 × 103 2 10−2 1.1 − 1.2 6 × 103 − 4 × 10−4 − 1.5 × 10−4 −
2 × 104 10−3 5.0 × 10−4
SO350 2 × 103 − 1.8 1.2 − 2.0 1 × 104 − 7 × 10−5 − 2.0 × 10−5 −
5 × 104 8 × 104 4 × 10−4 2.8 × 10−4
SO500 320 − 5 × 104 2.5–2.7 1.1 − 2.0 6 × 103 − 7 × 10−5 − 2.3 × 10−5 −
8 × 104 8 × 10−4 1.0 × 10−3
G575 57 − 7 × 104 1.5–1.9 1.1 − 2.7 4 × 103 − 7 × 10−5 − 1.7 × 10−5 −
2 × 105 10−3 2.1 × 10−3
G935 8.0 × 104 2.59 2.7 2.23 × 105 6.5 × 10−5 1.2 × 10−5
SO1000 180 − 2 × 105 4.8–5.3 1.1 − 2.7 5 × 103 − 4 × 10−5 − 1.3 × 10−5 −
1 × 105 10−3 1.9 × 10−3
SO2300 103 12 1.2 104 4 × 10−4 6.8 × 10−4
SO4350 103 − 2 × 103 23 1.2 − 1.3 104 − 2 × 104 3 × 10−4 − 4.8 × 10−4 −
4 × 10−4 7.2 × 10−4
SO5000 103 − 2 × 105 27 1.2 − 2.7 104 − 2 × 105 4 × 10−5 − 2.2 × 10−5 −
5 × 10−4 1.0 × 10−3
SO10000 2 × 103 − 105 50 1.3 − 2.7 2 × 104 − 105 4 × 10−5 − 2.7 × 10−5 −
2 × 10−4 7.0 × 10−4
SO30000 2 × 103 − 105 1.5 × 102 1.7 − 2.7 5 × 104 − 105 4 × 10−5 − 4.0 × 10−5 −
10−4 1.6 × 10−4
SO60000 5 × 104 − 105 3.2 × 102 2.1 − 2.4 8 × 104 − 105 5 × 10−5 − 6.5 × 10−5 −
6 × 10−5 1.0 × 10−4
SO100000 5 × 104 − 105 5.4 × 102 2.1 − 2.7 8 × 104 − 105 4 × 10−5 − 5.7 × 10−5 −
6 × 10−5 1.0 × 10−4

two 0.5 megapixel high speed, digital cameras (Vision Research, model Phantom 7.3), it revealed
for the first time the interfacial features and internal structures of the breakup zone of low-viscosity
liquid drops. In the present work, these techniques are augmented by the addition of an advanced
1 megapixel high speed digital camera (Vision Research, model Phantom V12), of several single-
frame 11 megapixel digital cameras (Luminera, model Lw11059M), and ∼5 ns exposure times
provided by a single, 100-MW-pulse from a Nd:YAG laser (New Wave Research, model Tempest
10 Hz). As a result of these enhancements, and with appropriate light splitting, expansion and
collimating optics, and a range of lenses and magnification accessories, we have achieved resolutions
of up to ∼200 pixels/mm at object-speeds of 100 m/s. The timing of the single-shot visualizations
could be adjusted to within 1 μs, which along with the highly reproducible experiment-outcomes
provide the equivalent of a 1 MHz video at 22 megapixels frames (two stacked cameras) and
unlimited total capacity of record. This capability allowed us to revisit the low-viscous work with
an enhanced level of clarity and detail, especially in the rather complex configurations found in the
advanced stages of breakup; this work is included here.
The several schema of visualization, along with related nomenclature can be seen in Figure 2.
The zero position (P0) is where the initial drop-shock interaction takes place, and all downstream
positions are coded by the distance from this point in centimeters (P10, P20, P65). Data acquisition
is based on a multi-scale visualization strategy applied within a series of repeat runs (the drop can be
positioned at the time of shock arrival within 1 mm in the center of the flow channel). First, the overall
features in the initial-interaction region and of the particle-clouds at various downstream positions
022104-5 The physics of aerobreakup. II. Viscous liquids Phys. Fluids 24, 022104 (2012)

FIG. 2. (Color online) Illustration (top view) of the various visualization arrangements as indicated by the numbers in
parentheses: (a) 1 shadow imaging, 2 oblique photography (22◦ ), and 3 LIF imaging (22◦ for ASOS runs, 30◦ in ALPHA).
(b) 1 LIF imaging and 2 shadow imaging with LIF screen. HS1/HS2 denote high speed video camera 1/2; S1/S2/S3 are for
single-frame camera 1/2/3. Oblique views introduce no image distortion. Applicable visualization is included in run definition
by “camera (view)”.
022104-6 Theofanous et al. Phys. Fluids 24, 022104 (2012)

FIG. 3. (Color online) (Left) Shock visualization S1(2): the scale indicates 1 mm. The shot was taken 496 μs after shock
arrival to a pressure transducer located 421 mm upstream of the drop. According to the shock speed (858 m/s) measured
by the travel time between two transducers located 50 and 101 cm downstream, the drop exposure time to flow was 5.4 μs.
(Right) Numerical simulation result at the same time instant. The bow shock is at the position 0.34 d0 , which compares to
the value 0.35 d0 predicted. The shock-intersection point on the back is at a distance of 0.34 d0 which compares to the value
0.33 d0 predicted. Also evident in the shadow image is the projection of the curved shocks on the back, showing as a blurry
line. Noticeable is also the trace of fine mist emanating from the equatorial region. Run [W1, M = 1.14, Reg = 2.1 × 105 ,
We = 5.0 × 104 , Oh = 2.2 × 10−3 , S1(2)].

are recorded at medium resolution (4–50 pixels/mm) using simultaneously the two Phantom 7.3 and
Phantom V12 cameras, synchronized with the copper vapor laser at a pulse frequency of 20 kHz,
or only the Phantom V12 at 42 kHz. At the downstream positions, complete clouds are digitally
reconstructed by splicing frames according to overlapping portions of images, which also provides
individual particle velocities. Then using these overall results, fine features are explored using the
Luminera cameras with the Nd:YAG laser, at resolutions that vary from 50 pixels/mm (to capture
near complete clouds), up to the maximum of 200 pixels/mm (to capture interfacial features and
particles down to 5 μm). At this maximum resolution the depth of view narrows to ∼1/2 mm, and
this allows the selecting of slices through the mixing zone in well-defined planes parallel to the flow.
These techniques are employed in a complementary fashion with LIF as well as specular reflection
and shadow imaging, which attain new utility at the space–time resolutions achieved in this work.
Of special interest for its enhanced clarity (including visualization of the shock wave and hence
the exact timing of the shots) is Schema 2 in Figure 2(b), using the LIF-screen for shadow imaging.
As illustrated in Figure 3, the timing of a YAG shot can be determined with an accuracy of better
than ±1 μs.

III. NUMERICAL SIMULATION METHODS

Direct numerical simulations are used here for the purpose of probing details/effects of the
gas dynamics and interfacial-morphology features not available otherwise. For example, on the gas
dynamics of such problems, starting with Hinze2 and Harper et al.,7 shear stresses were thought to
be of no consequence and were, erroneously as it turned out,1 ignored. On the other hand, while
the boundary layer stripping idea of Taylor15 and Ranger and Nicholls16 is based on interfacial
shear, it has only been addressed for steady-state incompressible flow and only very roughly. Indeed,
experimental work even for the basic spherical shape has been limited to addressing only the total-
drag coefficient17 and even so missing the transonic region. By contrast, aerobreakup includes an
amazingly rich variety of shapes and interfacial morphologies found with various liquid viscosities
and gas-flow Mach and Reynolds numbers.8
At a yet higher level of complexity, the gas-dynamics couple to evolving interfacial mor-
phologies; namely, interfacial instabilities which in this problem are the nonlinear outcome of the
superposition of R-T instabilities (predominantly active in the forward stagnation region of the
022104-7 The physics of aerobreakup. II. Viscous liquids Phys. Fluids 24, 022104 (2012)

flow), K-H instabilities (of increasing potential away from the stagnation region), and of the mean,
shear-induced liquid flow at the outer layers of the drop.1, 8 Especially important in this regard is
to determine the onset of the apparent bifurcation between RTP and SIE seen experimentally here
with liquids of any viscosity; namely, the long-wave phenomena (drop deformation, R-T waves
comparable to the drop diameter), being overtaken by short-wave length, parallel-flow instabilities
(K-H waves, and sheet/filament ejections followed by capillary breakup and entrainment). Interfacial
shear and emerging K-H waves that enhance this gas–liquid interfacial coupling are deemed to be
the critical factors in this transition.
Our intention is to provide some initial perspectives of what can be done to further understand-
ing in the above two directions by use of numerical simulations. We employ for these purposes
two recently developed codes: (1) MuSiC+ —a compressible Navier–Stokes equations solver with
a sharp-interface treatment18, 19 (Appendix A) and (2) AROS—a Chebyshev collocation solver of
the linear stability (Orr–Sommerfeld) problem in parallel flows of fluids with a sharp or a diffuse
interface35, 36 (Appendix B). In combined use, the MuSiC+ code provides the shear stress distribu-
tions and viscous boundary layer details around a drop, which are then used as input to AROS to
predict viscous K-H dispersion graphs for comparison with wave-structures seen in the experiments.
In addition, MuSiC+ results are used to interpret, by means of the aerodynamics, other key features
of interfacial morphologies and drop-deformation patterns.

IV. LONG-WAVE PHENOMENA

As noted above, under long-wave phenomena, we consider drop shape changes (deformation),
associated drag variations, and appearance and growth of some small number of R-T waves on
the windward face of an aerodynamically flattened drop (Figure 4, also Figure 14 of Theofanous
and Li1 ). Coupling of shear stresses at the gas–liquid interface in this regime are insignificant, and
entrainment is completely absent. Deformation results from principally irrotational motions within
the drop, which are driven by normal stresses on the drop surface. However, shear stresses in the gas
phase couple to the drop shape and along with compressibility define the gas-dynamics (for example
position of the flow separation) responsible for the normal stress distribution.8 Deformation and
associated drag coefficients tie the drop motion to the gas flow, and the resulting drop acceleration
provides the body-force field for R-T instability. All such drop-internal motions are resisted by
viscous stresses (within the liquid drop), which depending on the value of the Oh may or may not be
significant. We have correspondingly the inviscid1, 8 (Oh < 2 · 10−1 ) and viscous (Oh > 2) regimes,
the primary focus in this paper being on the latter.

FIG. 4. Base mode of RTP with a viscous drop. t(ms) = 1.33; 1.37. Run [SO1000, M = 0.3, Reg = 1.3 × 104 , We = 1.5
× 103 , Oh = 5.3, S1(2)] (enhanced online) [URL: http://dx.doi.org/10.1063/1.3680867.1].
022104-8 Theofanous et al. Phys. Fluids 24, 022104 (2012)

The key point is that in the viscous regime, surface tension and shear stresses on the liquid are
negligible. Characteristic length and time scales can accordingly be obtained as8
 2 1/3  ν 1/3
+ ν
tV+ = 2
l
lV = l , (2)
a a
where a is the drop (center-of-mass) acceleration. Note that the acceleration embodies the gas-
dynamic pressure gradients acting on deformation.8 These in turn inherently reflect variations due
to compressibility (Mach number effects).
The above scaling groups arise also from a R-T instability analysis on viscous liquid layers of
finite depth as developed by Mikaelian22 (see also Chandrasekhar23 and Theofanous8 ). Adapted to
our present interest the dispersion relation is
   1/2
K V2
N V = −K V2 + K V4 + K V 1 − tanh(H V K V ) , (3)
Oh 2 DV 0
where the growth factor (NV ), wave number (KV ), initial drop diameter (DV0 ), and the flattened-drop
thickness (HV ) are shown in scaled form according to (2). Similarly to the inviscid case, in the first
mode of RTP we must have8
DV K V∗ = π, (4)
where the * indicates the peak-growth wave number in (3). We thus have the means to express the
first criticality in the viscous regime in terms of the local-instant parameters; namely, acceleration
and deformed-drop diameter. The latter is related to the thickness of a flattened drop and its initial
diameter (by conservation), which allows the system (3) and (4) to be closed.
We can continue analytically by noting that when Oh2 DV0 > 10 (viscous regime, K V∗ far away
from the capillary cutoff wave number), the peak-growth wave number becomes only a function of
HV , and that to an excellent approximation this function can be written as8
−1/4
K V∗ ∼ 0.9 HV for HV < 2. (5)
It then follows that at criticality we must have
DV 0 D 2 ∼ 4.6, (6)
or alternatively
PV,c D 8 ∼ 130, (7)
where D is the deformed drop cross-stream diameter normalized by the initial drop diameter, and
PV, c is the critical dynamic pressure (Pd, c ) scaled by a characteristic viscous liquid-dynamic pressure
(PV+ ) defined by

PV+ ≡ 1/2ρl u +
2
V ≡ 1/2ρl (νl /d0 )2 . (8)
In the order-of-magnitude estimations of (6) and (7), we have assumed8 (also Appendix A) that the
drag coefficient C D ∼ 1, which is also applicable to the disk shape. We can now see that RTP is
inextricably connected to deformation. We can also see the quantitative role of acceleration (also
tied to deformation to a significant degree) as the driver for breakup, and how liquid inertia and
viscosity enter as retarding parameters.
The above ideas are tested against our data in Figures 5 and 6. According to these results,
the fastest-growing mode (K V∗ ) of the R-T instability is selected8 in a temporal coincidence of
the drop-flattening process (the DV , which increases with time), and of the R-T instability process
(K V∗ , which remains almost a constant), until condition (4) is met, as illustrated in Figure 7. The
increase in DV with time incorporates increases in D (Figure 5) as well as in acceleration (through
the viscous length scale). These increases in D and a have a compensating role in K V∗ . This can be
seen in K V∗ ∼ D 1/2 a −1/12 , which follows from (2) and (5), and an estimate of HV based on volume
conservation. Moreover, we can see that there are no practical limits to the first criticality due to
viscosity.
022104-9 The physics of aerobreakup. II. Viscous liquids Phys. Fluids 24, 022104 (2012)

FIG. 5. (Color online) Deformation dynamics in the viscous regime of RTP. The scaling of each data point is based on
instantaneous acceleration deduced from the center-of-mass motion. This figure includes all near-criticality data on the
viscous branch of Figure 9.

(a) (b)

(c)

FIG. 6. (Color online) Theoretical interpretation of the first criticality in the viscous RTP regime according to (5)–(7). The
dashed and solid lines in (a) represent the 90% and 95% bounds from the peak growth factors. The figures include all near-
but above-critical data on the viscous branch of Figure 9 for which there was enough information to estimate accelerations
and deformations as the onset of instability is approached; for accurate determination of acceleration this requires visual data
with an intact drop in both cameras (P0 and P20).
022104-10 Theofanous et al. Phys. Fluids 24, 022104 (2012)

FIG. 7. (Color online) The approach to viscous RTP criticality according to (4). Run [SO4350, M = 0.37, Reg = 1.73
× 104 , We = 2.5 × 103 , Oh = 23.2, HS1(1)].

These results also demonstrate the important role of finite thickness on the onset of instability.
For an infinitely thick viscous liquid (HV > 10; note that (5) gives very closely the exact asymptotic
value of 0.5 at HV = 10) in place of (5) and (6), we have respectively,

K V∗ = 0.5 and DV = DV 0 D = 2π, (9)

which conflicts with the data in Figure 6. This also shows that it would require about six times greater
acceleration for criticality than found experimentally (at the same viscosity and same deformation).
From Figure 5, we have an empirical measure of the timing for deformation, which provides the
basis for simplified modeling of the overall breakup process even under transient flow conditions.
Specifically, marching out in time, from the deformation at each time step one can obtain the drag
and with the actual relative velocity at that time one can compute the acceleration, thus being able to
check for instability against criteria (6) or (7). The alternative means to prediction is direct numerical
simulation (DNS), which as of late is becoming feasible.8
In any case, correct, in-detail prediction of RTP requires great attention to the aerodynamics,
which can only be captured by DNS, and even so with great difficulty, because of the extreme grid-
resolution demands for the viscous boundary layer (Appendix A). Even though they only contribute
a small fraction of the total drag, shear stresses control separation and reattachment phenomena, and
thus indirectly influence the distribution of pressure forces over the drop surface. As a consequence,
the indirect influence on total drag is great. The effects of drop shapes during the transient, and
of the Mach number have been discussed previously;8, 11 here we provide another, perhaps more
tangible, demonstration by means of direct comparison between ALPHA and ASOS tests at the same
We = 75 (Figure 8).
In more direct terms, the criticality status of a given liquid in a flow condition can be judged
from the data plotted in the classical coordinates (Figure 9, which includes the data of Hanson et al.30
reduced by us for this purpose). Considering the vast difference in the experimental ranges covered
in respective databases, the result

W ec = 12 (1 + 4/3 Oh 1.5 ) 10−2 < Oh < 550, (10)

is remarkably close to the extrapolation of (1). We can see the transition from inviscid to viscous
behavior to extend over an one-decade region of the Ohnesorge number, 2 · 10−1 < Oh < 2 · 100 ,
with the corresponding extremes of critical Weber numbers being 13.3 and 57.2.
022104-11 The physics of aerobreakup. II. Viscous liquids Phys. Fluids 24, 022104 (2012)

FIG. 8. (Color online) Comparison of deformations between a Mach 3, low pressure run in ALPHA (c), and a near-ambient
pressure, low-Mach ASOS run (a) and (b) at similar We; however, the Reg s, are vastly different. Images are shown at the same
scaled times TI = 0.32 and 0.51 (upper and lower rows). The length scales indicate 1 mm. (d) and (e) show the pressure fields
and streamlines from numerical simulations of the ASOS and ALPHA runs at respective (M − Reg ) conditions. (f) is from a
numerical simulation in which the drop was allowed to respond to the aerodynamics. The general pattern expected in ideal in-
compressible flow has been classically evoked to estimate drop shapes and deformation histories, but this is correct only at very
restrictive conditions in (M − Reg ) space that allow surface tension forces to compensate for asymmetries.11 Most commonly,
as in the present illustration, this balancing effect cannot occur, and in conjunction with stronger flow-separation phenomena8
the muffin-like pattern results. This describes also the manner in which the SIE regime is approached from below. It turns
out that maximum cross-stream diameter is not affected, and it scales in the usual manner by TI (for example, Figure 4(a) in
Ref. 8). (a) and (b) Run [TBP4, M = 0.08, Reg = 4.2 × 103 , We = 88, Oh = 1.6 × 10−2 , HS2/HS1(3)] (c) ALPHA Run
[TBP4, M = 3, Reg = 4.4 × 102 , We = 75, Oh = 1.3 × 10−2 , HS1(3)] (f) MuSiC Run [G575, Ms = 1.2, Reg = 1.57 × 104 ,
We = 520, Oh = 1.8].
022104-12 Theofanous et al. Phys. Fluids 24, 022104 (2012)

FIG. 9. (Color online) The first mode of RTP criticality shown in the classical coordinates. Only data near the criticality line
are shown. Note that there is an error in Table IV of Hanson (1963). The K for SO10 should be divided by 10, but properties
in his Table I are correct. The parameters for this plot were recalculated accordingly.

This rather smooth transition region to the viscous branch appears as a dramatic increase
when seen in semi-log coordinates, as plotted for example in Figure 2 of the review paper by
Pilch and Erdman.31 With the data available at the time reaching only up to Oh ∼ 1, they proceed to
falsely speculate that “Drop breakup at higher Ohnesorge numbers is progressively more difficult and
ultimately impossible (in a practical sense).” Moreover, it appears that being unaware of this transition
region, these authors, along with the original investigators (Wolfe and Andersen32 ), misinterpreted
test results on the effect of viscosity on breakup. Pilch and Erdman’s statement, “Wolfe and Andersen
(1964) found that large drop viscosity (Oh > 0.1) delayed initiation of breakup without altering the
observed breakup mechanism,” derives from Wolfe and Andersen’s: “As shown in Figure 14, a
change in viscosity of more than two orders of magnitude does not appreciably alter the mode of
breakup,” which refers to tests with 0.47-cP and 170-cP silicone oils. With the help of Figures 9 and
18 (that depicts the onset of SIE), we can readily demonstrate that the breakup behavior could not
have been the same, and that in fact it was not the same. Based on test conditions (ug = 38.7 m/s,
d0 = 2.1 mm, ρ g ∼ 1.3 kg/m3 ) and the (μl , σ , ρ l ) equal to (4.7 × 10−4 , 17.5 × 10−3 , 758) and
(1.68 × 10−1 , 23.5 × 10−3 , 966) for the GE FX(96)0.65 and GE FX(96)200 liquids, respectively,
we obtain WeL = 234, WeH = 174, OhL = 2.8 × 10−3 , OhH = 0.75, where the subscripts L/H
(“Light”/“Heavy”) stand for the 0.65/200 liquids, respectively. Then from Figures 9 and 18, we can
see that the Heavy liquid is above the 1st but well below the 2nd criticality, and indeed the visual
images in Figure 14 of Wolfe and Anderson show no evidence of entrainment. On the contrary, the
Light liquid is found well above even the 2nd criticality, as indeed is seen to be the case (entrainment)
in their Figure 14.
Now it can be readily seen that on the viscous branch of (10), the role of surface tension is slight,
as indeed must be the case according to the theoretical development given above. Accordingly, a
more-to-the-point version of (10), for the viscous region, obtained by rearrangement,
 1/2
μl ρg
Reg,c = 4 Oh −1/4 Oh > 2, (11)
μg ρl
shows a surface tension dependence to the 1/8th power. The viscous-branch criticality data are shown
in coordinates suggested by (11) in Figure 10.
Salient data features in the viscous regime include:
(a) Near criticality, the approach to RTP may not be monotonic (online video link to Figure 4, and
Figure 11).
022104-13 The physics of aerobreakup. II. Viscous liquids Phys. Fluids 24, 022104 (2012)

FIG. 10. (Color online) The criticality data from the viscous branch of Figure 9 on the coordinates suggested by (11).

(b) Above criticality, with increasing dynamic pressure, increasingly higher modes of RTP
are observed (Figure 12, also Figure 14 of Theofanous and Li1 and Figure 9 of
Theofanous et al.37 ).
(c) As expected, with increasing viscosity the time needed for the rim to break up (by Rayleigh
instability) increases significantly (Figure 13).
(d) Drop displacements prior to appearance of instability increase significantly with increasing
viscosity, and special care is needed to assure accurate determination of criticality. This could
be the reason for the “limit” (Oh ∼ 4) on breakup stated in Faeth’s work as noted above. In
any case, their equipment was not up to the task—flow-limited to We < 103 , according to (10)
they would not be able to achieve breakup for Oh > 6. Conversely, with ASOS we found that
we nearly overestimated the criticality line on the upper end of the viscous branch, if not for
caring to observe with a second camera up to 40 cm downstream of the initial drop position.
(e) At near-critical conditions on the viscous branch, the viscous timescale is numerically quite
close (within a factor of 3) to the inertial timescale11 (t I+ ). This follows directly from the

FIG. 11. (Color online) The rebound-approach to the first criticality with viscous liquids (Oh = 5.2) at flow conditions near
the onset.
022104-14 Theofanous et al. Phys. Fluids 24, 022104 (2012)

FIG. 12. Illustration of the second and third symmetric RTP modes obtained with increasing dynamic pressure past the first
criticality. (a) Run [SO1000; M = 0.59, Reg = 3.6 × 104 , We = 9.2 × 103 , Oh = 5.2, HS1(1)]. (b) ALPHA Run [W1, M
= 0.03, Reg = 5.2 × 103 , We = 30, Oh = 2 × 10−3 , HS1(1)] (enhanced online) [URL: http://dx.doi.org/10.1063/1.3680867.2]
[URL: http://dx.doi.org/10.1063/1.3680867.3].

definitions of the two timescales and the use of (10):

 1/6
t I+  2 2 1/3 W ec
 1/3 ug − U
+ ∼ 1.3 C D D ψ 2
∼ 2 C D D2ψ 2 Oh −1/12 , ψ= ,
tV Oh ug

where U is the drop velocity.

V. SHORT-WAVE PHENOMENA
The onset of the second criticality is marked by shear-layer ejections at the equator. Up to and
in the immediate neighborhood of this transition, defined by the appearance of the first localized
(asymmetric) traces of detached liquid from the drop’s main body, the deformation is rather sym-
metrical, and the process is still dominated by high-modes of RTP. At modestly higher velocities,
the ejections become azimuthally symmetric, and slight interfacial undulations appear, while at the
same time a deformation asymmetry develops, in the flow direction, and the entrainment phenomena
take on a dominant character (Figure 14). The extending sheets quickly break up to yield threads
and drops in a process reminiscent of Taylor’s15 boundary-layer stripping model (separation of the
viscous boundary-layer at the equator, in a smooth stream-wise fashion). However, the above-noted
deformation-asymmetry, and the surface waves, contrast with assumptions made by Taylor and in
subsequent versions of this model.

FIG. 13. Rayleigh (capillary) breakup of the rim in base-mode of RTP, and its delay with highly viscous liquids. (a) ALPHA
Run [W1, M = 0.017, Reg = 3.3 × 103 , We = 12, Oh = 2 × 10−3 , HS1(1)]. TI = 3.1. (b) Run [SO1000, M = 0.14, Reg
= 6.5 × 103 , We = 325, Oh = 4.85, S1(2)]. TI = 12, P20. (enhanced online) [URL: http://dx.doi.org/10.1063/1.3680867.4]
[URL: http://dx.doi.org/10.1063/1.3680867.5].
022104-15 The physics of aerobreakup. II. Viscous liquids Phys. Fluids 24, 022104 (2012)

FIG. 14. The near-onset SIE. In this and the next three figures, the reference length scales indicate 1 mm, and the short,
horizontal line-marks indicate the diameter of the original drop. t = 0.4, 0.95, 1.2, and 1.7 ms. Run [W1, M = 0.16, Reg
= 1.2 × 104 , We = 210, Oh = 2.1 × 10−3 , HS2(3)/HS1(3)] (enhanced online) [URL: http://dx.doi.org/10.1063/1.3680867.6].
022104-16 Theofanous et al. Phys. Fluids 24, 022104 (2012)

FIG. 15. Early interfacial instabilities in the SIE regime. (a): Run [TBP4, M = 0.31, Reg = 1.6 × 104 , We = 1.6 × 103 , Oh
= 1.8 × 10−2 , S2/S1(1)]. t(μs) = 80, TI = 0.2. (b): Run [G575, M = 1.26, Reg = 1.6 × 105 , We = 5.41 × 104 , Oh = 1.9,
S2/S1(1)]. t(μs) = 12, TI = 0.24. (c): Run [G935, M = 1.28, Reg = 2.23 × 105 , We = 8.0 × 104 , Oh = 2.6, S3/S2/S1(1)].
t(μs) = 32, TI = 0.44.

At modestly higher velocities (as detailed in Sec. V B), the behavior alters to one dominated
by short interfacial waves that initiate promptly behind the shock, at some distance downstream
from the forward stagnation point. These waves initially occupy a band extending over a polar
angle of roughly 30◦ < θ < 60◦ , but spread out in time to 90◦ and beyond (Figure 15). Over the
spherical segment ahead of this band, the drop surface remains perfectly smooth. Also smooth, at
the very early stages, is the downstream region reaching out some distance beyond the equator.
Associated entrainment phenomena are also prompt, but the pattern quickly increases in complexity,
to a sheet-like flow with a significant radial component, accompanied by fragmentation of this sheet
by three-dimensional waves into coarse pieces and simultaneously by fragmentation of these pieces
into fines (Figure 16). Convected downstream, these fragments create a cylindrical “curtain” around
an empty space behind the coherent portion of the drop; that is, the portion that diminishes in mass
but remains intact until it becomes thin enough to break through (Figure 17). In previous work,
shadow imaging concealed these internal structures; instead, and erroneously, external features,
such as the overall envelope of the gas–liquid mixture, were taken to characterize the cross-stream
diameter and acceleration of the parent drop. Significantly, in the fully developed SIE regime, the
cross-stream deformation of the parent drop is negligible (as detailed in Sec. VI).
Previously established1 for Oh  1 conditions, these SIE phenomena are now seen to be
pervasive (in fact they are further clarified due to the larger length scales involved) at all liquid
viscosities—only the above-noted sequential manifestation occurs at higher dynamic pressures
with increasing Oh. Similarly, the absence of the catastrophic regime1 is confirmed for all liquid
viscosities. In the remaining of this section, we focus on the scaling laws that govern the onset of this
022104-17 The physics of aerobreakup. II. Viscous liquids Phys. Fluids 24, 022104 (2012)

FIG. 16. The fully developed SIE regime. (a) Run [TBP4, M = 0.31, Reg = 1.6 × 104 , We = 1.6 × 103 , Oh = 1.8 × 10−2 ,
S2/S1(1)]. t(μs) = 150, TI = 0.38. (b) Run [G575, M = 1.26, Reg = 1.6 × 105 , We = 5.4 × 104 , Oh = 1.9, S2/S1(1)]. t(μs)
= 35, TI = 0.7. (c) Run [G935, M = 1.28, Reg = 2.2 × 105 , We = 8.0 × 104 , Oh = 2.6, S3/S2/S1(1)]. t(μs) = 59, TI = 0.81.

second criticality and on the nature of the interfacial instabilities observed. In Sec. VI, we address
the advanced-breakup patterns and associated timings.

A. Scaling laws of the second criticality

Depicted in the classical coordinates (Figure 18), our data on the onset of SIE show that on the
viscous branch we have Wec ∝Oh. Thus, as expected, and unlike (the viscous branch of) the first
criticality, surface tension plays an important role here. A simple rearrangement suggested by the
above proportionality then yields,

Pd,c
PV C,c ∼ 2.4 · 103 for 2 · 10−1 < Oh < 102 , where PV C.c = ,
PV+C

σ (12)
PV+C = 1/2ρl u + + +
V u C and u C = .
ρl d0

The characteristic viscous velocity (u + +

V ), we have seen above. The capillary velocity (u C ) can be
thought of as providing the limiting balance between the thrust of a liquid eddy (or wave) within
the drop and the surface tension force that restrains its detachment. This viscous-capillary liquid-
dynamic pressure scales well the retarding effect of viscosity on the second criticality (Figure 19).
Equivalently, but perhaps in a physically more tangible interpretation, (12) can be expressed in terms
022104-18 Theofanous et al. Phys. Fluids 24, 022104 (2012)

FIG. 17. Advanced stages of the SIE regime. (a) Run [W1, M = 0.56, Reg = 4.7 × 104 , We = 3.4 × 103 , Oh = 2.5 ×
10−3 , S2/S1(1)]. t(μs) = 191, TI = 0.72. (b) Run [W1, M = 0.59, Reg = 6.7 × 104 , We = 5.2 × 103 , Oh = 2.5 × 10−3 ,
HS2/HS1(3)]. t(μs) = 150, TI = 0.81. (c): Run [G935, M = 1.28, Reg = 2.2 × 105 , We = 8.0 × 104 , Oh = 2.6, S3/ S2/S1(1)].
t(μs) = 84, TI = 1.18.

FIG. 18. (Color online) The SIE criticality in the classical coordinates.
022104-19 The physics of aerobreakup. II. Viscous liquids Phys. Fluids 24, 022104 (2012)

FIG. 19. (Color online) Viscous-capillary scaling of the SIE criticality. Symbols are same as Figure 18.

of the maximum shear stress (τ m , using the numerical simulations in Appendix A) as

τm,c
τV C,c ∼ 16 for 2 · 10−1 < Oh with τV C.c = . (13)
PV+C

Similarly, in the inviscid regime, we have W ec ∼ 102 , which yields

τm,c + +2
τC,c ∼ 2 for Oh < 2 · 10−2 with τC.c = + and PC = /2 ρl u C .
1
(14)
PC

It may well be that the much larger scaled value to critical shear stress on the viscous branch is
reflective of the relevance of an integral value of stress in this case rather than the maximum.
Near transition, the entrainment rate is not sufficient for the process to bifurcate decisively to
SIE, and a significant participation of RTP, along with the characteristic cross-steam deformation,
is seen. On the other hand, for SIE to remain effective, the relative velocity must remain above the
threshold defined by (13). At high-Oh conditions, the rate of breakup may be small enough to allow
drop acceleration to a degree sufficient to violate this threshold prior to complete breakup. In such
a case, the breakup regime should be expected to revert back to RTP, as indeed is found in our
experiments (Figure 20). It can be seen that even at the highest dynamic pressures used in this set of
experiments (∼1 MPa), a drop cannot be fully fragmented in SIE when Oh > 50.

B. Interfacial instabilities
When dynamic pressure exceeds the critical value for the onset of SIE by about a factor of
∼2 (Figure 21(a)), the drop surface becomes rippled promptly upon passage of the shock, and the
SIE proceeds without the kind of cross-stream distortion found at lower flow speeds. The resulting
interfacial roughness enhances the shear force on the drop, which gives rise to a fully developed SIE
as detailed in Sec. VI. Here we address the initial stages of this rippling, including the dominant
wave lengths and the overall morphologies such as quiescent bands, sharp local ejections, and an
apparent cave-in of the backside, all appearing simultaneously at different regions on the drop surface
(Figure 21, also Figure 15(c)). All these phenomena occur before any significant deformation, so
we use MuSiC+ simulations for flow over a sphere, at the relevant Reg and M conditions to probe
the aerodynamics (Appendix A). There is no intent to provide a comprehensive treatment here, but
rather to highlight the underlying causes of the main morphological features and the pivotal role of
viscous K-H instability as it emerged from this line of inquiry.
022104-20 Theofanous et al. Phys. Fluids 24, 022104 (2012)

FIG. 20. The effect of diminishing dynamic pressure on the drop over the relatively long breakup times found with high-Oh
liquids. The dynamic pressure trajectories are based on drop velocities deduced from displacement histories. (a) 1: Run
[SO10000, M = 0.52, Reg = 2.8 × 104 , We = 6.1 × 103 , Oh = 54, HS1(1)]. (a) 2: Run [SO10000, M = 0.6, Reg = 3.7
× 104 , We = 9.7 × 103 , Oh = 53, HS1(1)]. (b) 1: Run [SO10000, M = 0.96, Reg = 8.1 × 104 , We = 4.5 × 104 , Oh = 53,
HS1(1)]. (b) 2: Run [SO10000, M = 1.24, Reg = 1.3 × 105 , We = 1.3 × 105 , Oh = 53, HS1(1)].

1. The key early morphologies

All main features of interest are found in the TBP test result shown in Figure 21(b). The
respective MuSiC+ simulation is shown in Figure 22 and the inset shows the nomenclature used to
refer to the various regions of interest on the drop surface. The flow is seen to attach at θ ∼ 30◦ , and
to separate at θ ∼ 90◦ . Within this band (A-B), we can see the waves created by shear. Both ahead
(S-A) and behind (B-C) region, the surface remains perfectly smooth, as expected in light of the
absence of significant shear action. The shear stress distribution on the back of the drop is unsteady
and with local unsteady structures due to intense vortices that merge and break up, establishing a
localized low pressure region on the outside and a so-induced radial flow on the inside11 ; these are
responsible for the flattening with cave-in (D) (also Figure 15(c)), bulging (C-E), and entrainment
(E) phenomena.
022104-21 The physics of aerobreakup. II. Viscous liquids Phys. Fluids 24, 022104 (2012)

FIG. 21. (Color online) (a) The 3rd criticality is marked by K-H waves. All runs with S2(1)/S1(1). (b) Run [TBP4,
M = 0.56, Reg = 3.7 × 104 , We = 7.0 × 103 , Oh = 1.8 × 10−2 ]. t = 42 μs, TI = 0.23. (c) Run [TBP4, M = 1.25,
Reg = 1.55 × 105 , We = 1.23 × 105 , Oh = 1.8 × 10−2 ]. t = 8 μs, TI = 0.18. (d) Run [SO10, M = 0.51, Reg = 3.2 × 104 ,
We = 6.9 × 103 , Oh = 5 × 10−2 ]. t = 26 μs, TI = 0.13. (e) Run [G575, M = 1.26, Reg = 1.6 × 105 , We = 5.41 × 104 , Oh
= 1.9]. t = 10 μs, TI = 0.2 (f) Run [SO1000, M = 1.3, Reg = 1.7 × 105 , We = 1.9 × 105 , Oh = 4.9]. t = 5 μs, TI = 0.12.
022104-22 Theofanous et al. Phys. Fluids 24, 022104 (2012)

FIG. 21. (Continued.)

Region (S-A) would be expected to be R-T unstable, but in these, and all other images (at
SIE conditions), at all times, none is found. Theofanous and Li1 attribute this to stretching at the
forward-stagnation flow area induced by the shear in region (B-C). In the following, we show that
this intense stretching is triggered by viscous K-H instability at a dynamic pressure that corresponds
to what we have defined above as the onset of rippling. We also show that once established, K-H
022104-23 The physics of aerobreakup. II. Viscous liquids Phys. Fluids 24, 022104 (2012)

FIG. 22. (Color online) The detailed aerodynamics around a solid sphere at a flow condition corresponding to the run in
Figure 21(b) (enhanced online) [URL: http://dx.doi.org/10.1063/1.3680867.7].

completely overshadows the R-T process. The mollifying effect of stretching on R-T instability adds
to this principal dominance of K-H.

2. The viscous K-H instability

We proceed in two steps: (a) establish the method of computation and predictability of K-H
waves, and (b) employ the method to assess the relative roles of K-H and R-T instabilities over a
range of flow conditions appropriate for generalizations. As an illustration on item (a) we consider
the SO10 run shown in Figure 21(d); the analysis and conclusions reached are typical. The respective
MuSiC+ simulation is reflected in Figure 23, which includes a sample of profiles within the viscous
boundary layer. The relevant parameters are summarized in Table III (Appendix A), which also
informs the parameter space for viscous K-H instability analysis, namely the Reynolds and Weber
numbers (Reδ , Weδ ), both based on the boundary layer thickness (δ g ) and free-stream conditions.
We can see that variations are mild enough for the incompressible, isothermal, planar flow analysis
of the AROS code to be applicable. The viscous boundary layer in the liquid, as developed under the
applicable shear stress (τ ), is obtained from the exact result for a flat, constant-shear configuration,
μl u l,t  y
= 1 + er f (η) + π −1/2 η−1 exp −η2 η= √ y < 0, (15)
yτ 4νl t
where ul, t is the tangential liquid velocity in the interfacial region. In AROS, both (gas and liquid) base
flows are approximated by error function velocity profiles that match very closely the computed
ones (Appendix B). Juxtaposition of the measured wave lengths (Figure 21(d)) against the so-
computed dispersion graph (Figure 24) reveals that the waves seen are predictable by viscous K-H
instability at the (self-selected) peak-growth wave-number. In the time frame of this comparison,
022104-24 Theofanous et al. Phys. Fluids 24, 022104 (2012)

FIG. 23. (Color online) The detailed aerodynamics around a solid sphere at a flow condition corresponding to
the run in Figure 21(d). The boundary layer detail is shown for a polar angle of 45◦ (enhanced online) [URL:
http://dx.doi.org/10.1063/1.3680867.8].

liquid convection is negligible; interfacial velocity is ui ∼ 1 m/s (Appendix A) and net displacement
is less than 26 μm. According to the predicted growth factor (0.4 μs−1 ) and the timing involved
(26 μs) the amplification factor is 1600, which suggests initial wave amplitude of much less than
1 μm. This is reasonable for such a surface-tension-stabilized initial configuration. Whereas it is
impossible to measure or predict the initial amplitude of the surface disturbances, this deduced
value of the amplification factor provides an adequate surrogate for addressing item (b) of this
discussion.
To this end, we consider a typical mm-scale TBP drop subjected to flows that yield Weber
numbers of 75, 750, and 7500. For each dynamic pressure, the viscous K-H instability is resolved by
utilizing results of MuSiC+ simulations of the gas dynamics to define the necessary AROS input in
the manner described just above. The R-T instability is derived via the corresponding accelerations
of a sphere, where the drag coefficient variation with Mach number is taken into account. For TBP
(Oh = 0.02), viscous effects are insignificant for R-T, but crucial for K-H. The dispersion plots are
displayed in Figure 25. Wave numbers and growth factors of K-H instability are consistently greater
than those of R-T instability by more than an order of magnitude. Thus, just as in the case of RTP,

FIG. 24. (Color online) Measured wavelengths (run of Figure 21(d)) shown against predictions of viscous K-H instability
(Appendix B). The two bands represent two independent measurements, one emphasizing integral wave bands and the other
focusing on local wave structures. Variation in prediction due to uncertainty ranges of Reδ and Weδ shown in Table III are
negligible for purposes of the comparison addressed here.
022104-25 The physics of aerobreakup. II. Viscous liquids Phys. Fluids 24, 022104 (2012)

FIG. 25. (Color online) Linear-stability results for K-H (solid lines) and R-T (dashed lines) instabilities driven by free-steam
velocities and accelerations corresponding to Weber numbers of 75 (1), 750 (2), and 7500 (3).

there is a cutoff that is defined by the drop diameter (Figure 26). At dynamic pressures (equivalently
We) below this cut-off, K-H instability cannot appear, and this allows the time necessary for the
drop to deform and the much-slower R-T instability to develop. At dynamic pressures above this
cut-off, K-H instability develops rapidly (Figure 27), the so-induced roughness further enhances the
shear-driven flow, which then by continuity creates a drop-internal stagnation flow that meets the gas
stagnation on the outside. Besides, R-T instability being too slow to develop at such flow conditions,
the stretching further contributes to keeping this area mollified and free of any instability all the way
to the end. Thus the physical demarcation between RTP and SIE is predicted from the drop length
scale screening on the peak-growth wave of viscous K-H instability: λ*/d0 ≤ 0.2. It turns out that
when this condition is satisfied, so does the condition for growth shown in Figure 27.

FIG. 26. (Color online) Peak-growth waves of R-T and K-H instabilities as functions of the Weber number. The upper bound
on permissible waves defines the lower-bound We-range for this criticality (∼100–350) in agreement with experiment. It is
seen that R-T is completely out of range. It can only manifest if K-H cannot be activated (We < 100), and only after given
enough time and enough drop deformation to meet criticality condition (4).
022104-26 Theofanous et al. Phys. Fluids 24, 022104 (2012)

FIG. 27. (Color online) The peak growth factors of K-H (upper solid line) and R-T (dashed line) instabilities at times 20%
of TI (α = 0.2). The lower threshold, taken as 4–7, defines the potential range of We for this criticality (100–600), which is in
order-of-magnitude agreement with experiment. The time sensitivity on this range is shown by the K-H result at a time 10%
of TI (lower solid line, α = 0.1).

3. The sonic effects

Another remarkable feature that appears in the early nonlinear wave-growth regime is the
development of one of the K-H waves into a large trough (Figure 21(e)). At later times in
some of the high Oh cases, the effect is a dramatic cutting off, or splitting the drop as seen in
Figures 15(b) and 15(c). This typically is at a polar angle of ∼60◦ . The MuSiC+ simulation for
the run in Figure 21(e) is shown in Figure 28. The point made is that at the sonic condition
(M = 1), the pressure coefficient on the waves becomes singularly large, as known to be the case24
in compressible flow over a fixed, wavy interface defined by x = sin (αx), with → 0,
p − p∞ 2 α sin(αx) p − p∞ 2 α cos(αx)
=− √ M <1 =− √ M > 1. (16)
1
2
ρu 2
g 1 − M2 1
2
ρu 2
g M2 − 1
Note that in subsonic flow the pressure field is in-phase with the waves; therefore, it is instability
augmenting. In supersonic flow, the pressure is out-of-phase with the waves and thus it is stabilizing.

VI. BREAKUP TIMES

Previous work, hampered with visualization issues, could address the timing of breakup only in
a rough, overall manner; and only for low-viscous liquids (Oh  1). An original estimate,16, 25, 33
tb
Tb ≡ ∼ 5, (17)
t I+
intended to represent the time required for a drop to be transformed into a diffuse mist, became en-
trenched in the veil of low-resolution shadow imaging. Engel33 covered 1.3 < MS < 1.7, Ranger and
Nicholls16 1.5 < MS < 3.5, and Reinecke and Waldman25 3 < MS < 11, all with water drops. Even
though this is all that could be determined (time to transform the drop to a diffuse mist) in shadow
graphic imaging, Pilch and Erdman produced an estimate of “primary breakup times” which they de-
fine as “the time when a coherent residual drop ceases to exist,” while at the same time acknowledging
that this determination “. . . is very difficult to achieve when the drop Weber number exceeds about
350 because mist formed by wave crests completely obscures the event.” Upon closer examination
we find that their Figure 4 misrepresents the data of these original works as well as the findings
022104-27 The physics of aerobreakup. II. Viscous liquids Phys. Fluids 24, 022104 (2012)

FIG. 28. (Color online) The detailed aerodynamics around a solid sphere at a flow condition corresponding to the run in
Figure 21(e) (enhanced online) [URL: http://dx.doi.org/10.1063/1.3680867.9].

and intents of the original investigators:

(1) It can be readily seen that from Engel they plot what Engel defined as the “time to produce a
mist cone having a width-to-length ratio of 1 to 3.” This cone shape refers to the downstream-
convected mist and as seen in Engel’s Figure 10 (frame 2), nothing can be said about the
state of fragmentation of the main portion of the mass for which Engel states: “the remaining
portion of the drop has taken on a distinct quarter-moon shape”).
(2) The “data” attributed to Reinecke and Waldman can be found no-where, and in fact they are
contradicted in the several publications produced from this work.25, 26, 34 Take for example the
point shown at W e ∼ 7 × 104 . According to the original authors, in the stripping mode, the
complete-breakup time is 3.5, while for what they call catastrophic mode using their formula
one obtains 2.5, not the 1.2 value plotted by Pilch. For the data-points themselves taken at MS
= (3, 6, 11), or We = (6 × 104 , 3 × 105 , 7 × 105 ) respectively, Reinecke and Waldman25 report
Tb = (2.5, 2, 1.4) [however, with significant issues raised1 for the MS = 11 case]. In their final
022104-28 Theofanous et al. Phys. Fluids 24, 022104 (2012)

report,34 we find values 3.5–6, and even for MS = 12.6, they quote Tb = 1.9. It should also
be noted that the reproductions of the x-ray mass-contour maps found in these publications
is so poor that there is no way for a reader to challenge the original author’s estimates of
the progression of breakup as Pilch and Erdman have apparently done. Reinecke’ s results
themselves are discussed in Sec. VI B.
(3) Ranger and Nicholls, following up Taylor’s boundary layer stripping idea envisioned a gradual
depletion of the drop mass all the way to extinction, and they were the first to suggest (1) as a
measure of extinction. This is not represented in Pilch and Erdman’s Figure 4.

It was first thought that the timing of (17) is consistent with the time necessary to remove all
the liquid mass on the basis of Taylor’s boundary layer stripping theory. While controversial from
the very beginning,1 in light of present evidence this model can be seen to be utterly implausible. To
wit: (a) at the conceptual level, the model is missing the K-H instabilities discussed just above, and
the nonlinear development of these instabilities discussed further below, and (b) in implementation,
the cross-stream diameter was erroneously thought to encompass the already fragmented zone; in
fact as noted above and as demonstrated further below, in the fully developed SIE regime, the cross-
stream diameter of the coherent portion of the drop remains essentially unchanged and with it the
model-predicted fragmentation rate is nearly an order of magnitude too slow in comparison to (17)
(Ref. 1). Similarly, it was thought that the breakup time in the catastrophic regime was consistent
with the theory of Harper et al. (the theory provided the basis for a Bond number scaling used to
represent the experimental data25, 26 ), but this too turned out to be a confluence of missteps in theory
and experiments.1
Depending on the regime of breakup, we define respective measures of timing. In RTP, it is the
time required for waves to penetrate the flattened drop. This is taken as the point of complete loss of
bodily coherence and is denoted by subscript LC. In SIE, it is the time it takes for the entrainment
to reduce the volume of the coherent portion of the drop to nil. This is denoted by the subscript CA
(for complete annihilation). In both cases, we also have a second measure, as the time required that
all breakup processes cease (terminal breakup). But this requires detail consideration of the cloud
dynamics and it is left for future treatment.
Salient features of the type of analyses pursued towards achieving these ends include the defor-
mation dynamics of the coherent portion of the drop, and the detailed flow patterns of the detached
liquid masses. In both respects, we make complementary use of the full array of visualization tools
at our disposal, starting in each condition with high speed video and refocusing at selected time-
instances by the YAG-LIF method. This quantification is for all Newtonian liquids irrespective of
viscosity.

A. Timings of breakup processes in RTP

As we have seen already, the RTP regime is tied intimately to the deformation process that
precedes the onset of the actual instability. Once initiated, the instability develops in a time scale
that is much shorter than the timescale of deformation, which therefore can be taken as a measure
of the breakup time (Figure 5 for Oh > 1 and Figure 4(a) of Theofanous8 for Oh < 1). The TLC RT P

can be defined on this basis yielding,

RT P
TV,LC ∼ 7 − 10 Oh > 1, (18)

RT P
TI,LC ∼1 Oh < 1. (19)

Near criticality, the approach to RTP involves a rebound (Figures 11 and 29), but at higher values
of the dynamic pressure this disappears and the timing reaches the asymptote (18) as summarized
in Figure 30. Note that the acceleration enters these estimates to the 1/3rd power, so rather than
evolving the acceleration in time, it may be adequate that (18) is applied with the initial value.
022104-29 The physics of aerobreakup. II. Viscous liquids Phys. Fluids 24, 022104 (2012)

FIG. 29. Typical approach to RTP at a dynamic pressure near the critical value. (a) The rebound. The short, horizontal
line-marks indicate the length scale of the original drop. (b) The particle cloud at a position P65 downstream. The length
scale indicates 1 cm. Run [SO500, M = 0.15, Reg = 6.1 × 103 , We = 330, Oh = 2.6, HS1(1)] (enhanced online) [URL:
http://dx.doi.org/10.1063/1.3680867.10].

B. Timings of breakup processes in SIE

The quantification of SIE for low-Oh liquids is considerably more laborious but the results turn
out to be rather simple. The timing of complete annihilation is obtained quantitatively by estimating
the amount of liquid mass in the coherent portion of the drop from repeat runs with YAG-LIF as
illustrated in Figures 31 and 32. The results are then checked by complementary visualizations
using a combination of shadow and specular-reflection photography as illustrated in Figure 33. The

FIG. 30. (Color online) Timing of loss-of-coherence on the viscous branch of RTP. Only data with monotonic approach to
RTP are shown. Scaling based on initial acceleration.
022104-30 Theofanous et al. Phys. Fluids 24, 022104 (2012)

FIG. 31. (Color online) Core-drop outlines at selected instances (indicated in μs) during the breakup process in SIE. The
relative spatial positioning is made for legibility. The flow is from the right. The base flattens out almost immediately as
the remaining drop is eroded over the upstream-facing surface (initially a hemisphere, eventually evolving to a sequence of
shallow spherical segments with increasing radii of curvature). All runs with S2(1)/S1(1). (a) Run [TBP4, M = 0.31, Reg
= 1.6 × 104 , We = 1.5 × 103 , Oh = 1.8 × 10−2 ]. (b) Run [TBP4, M = 0.56, Reg = 3.7 × 104 , We = 7.0 × 103 , Oh = 1.8
× 10−2 ]. (c) Run [TBP4, M = 1.01, Reg = 1.07 × 105 , We = 5.0 × 104 , Oh = 1.8 × 10−2 ]. (d) Run [TBP4, M = 1.2, Reg
= 1.56 × 105 , We = 1.23 × 105 , Oh = 1.8 × 10−2 ].

FIG. 32. The fraction of drop mass lost by SIE for the runs in Figure 31.
022104-31 The physics of aerobreakup. II. Viscous liquids Phys. Fluids 24, 022104 (2012)

FIG. 33. Sample high speed visualizations made to complement the YAG-LIF data in identifying the core-drop boundaries.
The white outline in (a) is to demarcate the coherent portion of the drop. The short, horizontal line-marks indicate the
length-scale of the original drop. Run: [W1, M = 0.32, Reg = 2.2 × 104 , We = 7.8 × 102 , Oh = 2.4 × 10−3 , HS2(2)/HS1(1)]
(enhanced online) [URL: http://dx.doi.org/10.1063/1.3680867.11] [URL: http://dx.doi.org/10.1063/1.3680867.12].

asymptotic results for We > 1.5 × 103 can be simply represented in terms of the inertia-scaled time,

A ∼ 0.6
SI E
TI,C Oh < 10−1 , W e > 1.5 × 103 . (20)
Below this asymptotic range, we find a We-dependence as illustrated in Figure 34. For the
HS1(3)/HS2(3) data in this figure, the upper estimate of breakup was taken as the time corre-
sponding to the first frame where no drop remnant could be seen, and the lower estimate was that
of the previous frame. These data seem to anchor reasonably well on the considerably more precise
S2(1)/S1(1) data at the two ends of the range; however, they are probably an overestimate in the
middle region.

FIG. 34. Variation of breakup time in SIE as function of the Weber number. At the high-end of this We-range we can see
agreement with the TBP4 results presented above. The intermediate data obtained from analysis of HS visualizations are
only very approximate, and likely overestimates. Runs [W1, Oh = 2.4 × 10−3 , HS1(3)/HS2(3) or S2(1)/S1(1) as indicated].
022104-32 Theofanous et al. Phys. Fluids 24, 022104 (2012)

FIG. 35. (Color online) Core shapes and mass losses as in Figures 31 and 32, but for the viscous branch of SIE. Run [G935,
M = 1.28, Reg = 2.23 × 105 , We = 8.0 × 105 , Oh = 2.59, S2(1)/S1(1)].

It is interesting to compare (20) with the results of Reinecke and co-workers.25, 26, 34 They used
x-ray attenuation to determine mass distributions at selected instances of the fragmentation process
and concluded: “It was therefore postulated that two modes of breakup are important at high Mach
numbers; first, a gradual stripping mode in which material is removed continuously, followed by
a catastrophic mode in which the drop is rapidly torn apart” and “Transition from the stripping
mode of breakup to catastrophic breakup occurs early in the tests at shock Mach number of 11,
at intermediate time at Mach 6, and at late time (if at all) at Mach 3.” For stripping they give
Tb = 3.5 and for catastrophic the “theoretically motivated” correlation Tb = 45 · W e−1/4 . We can
now see that for stripping, they overestimated the time, probably because their method could not
“record” a fragmented mass that has not been adequately dispersed. On the other hand, while as
discussed in detail in part I of the present series1 and further demonstrated in the present paper the
idea of a catastrophic mode derived from false theory and a “made-to-fit” experimental mirage, we
can expect that the above-noted delay would diminish with increasing Weber number. Indeed, at
We = 106 , the above estimate yields Tb = 1.4, which means that their data show an approach to (20)
as they deviate significantly from (17).
With increasing liquid viscosity the length scales increase and the data reduction process
becomes considerably simpler. The results (Figure 35 and Figure 36) covering the upper end of the
We-range investigated specifically for this purpose can be summarized by

A ∼2
SI E
TI,C 10−1 < Oh < 10 We W ec . (21)

VII. PARTICLE-CLOUD DYNAMICS

As we have seen above, in the RTP the breakup occurs in two more or less discrete events, the
bursting of the bag (or bags) and the disintegration of the rim (s). This is reflected in a strongly
bimodal particle-size distribution, and also in a spatial stratification of scales in the final particle-
cloud constitution—a cloud of fines leading the way, and a cloud of coarser particles that follow with
a time delay that corresponds to that needed for disintegration of the rim(s) (Figure 37). There is also
a slight delay due to the extra time it takes for the coarser particles to accelerate to the free-stream
gas velocity. By contrast, in SIE the breakup process is continuous and multi-stage. It is continuous
in the sense that fine-scale particles form and entrain from the very initial moments of the interaction
with the shock wave up to the very end where breakup ceases (Figure 38). It is multi-stage in that
this fine-scale fragmentation originates both from the body of the original drop as well as from
any coarse fragments obtained during the primary disintegration process. As a consequence, the
particle-size distribution is polydisperse, and the spatial distribution of scales undergoes continuous
development, as the finer particles attain higher velocities faster.
022104-33 The physics of aerobreakup. II. Viscous liquids Phys. Fluids 24, 022104 (2012)

FIG. 36. Timing of complete annihilation in the viscous branch of the SIE regime at the upper end of the Weber numbers
investigated for complete breakup. In increasing order of Oh, the Weber numbers (in thousands) are: 3, 2, 75, 46, 80, 46, and
146. The plot incorporates the result of Figure 35 with a series of HS(1) runs as in Figure 34.

VIII. CONCLUSIONS
(1) The principal modes of breakup of Newtonian liquids are dominated by two criticalities:
Rayleigh–Taylor piercing and shear-induced entrainment. Both consist of inviscid and viscous
branches with critical conditions that obey the key physics of Rayleigh–Taylor instability and
viscous Kelvin–Helmholtz instability, respectively. Taylor’s boundary-layer stripping is seen
but only in the initial moments of SIE, and only in the neighborhood of the transition from
RTP to SIE. Harper’s (and Joseph’s) catastrophic breakup is completely absent.
(2) Both R-T and K-H instabilities originate by self-selecting the fastest-growing linear wave, and
therefore they are a priori predictable. Accordingly, the onset of RTP is also predictable on
basic grounds—one-half of a peak-growth wavelength fitting the diameter of the deformed
drop. When Oh > 2 surface tension becomes irrelevant and the loss of relative velocity over
timescales required for deformation to develop is important for a quantitative description

FIG. 37. A fragmented G575 drop passing-by at P10. The size of the biggest particles in the image is ∼650 μm. Note
the fine threads that connect the large fragments; due to the high viscosity, capillary breakup is delayed relative to that
seen with low viscous liquids (where this breakup is essentially instantaneous). Run [G575, M = 0.26, Reg = 1.25 × 104 ,
We = 390, Oh = 1.9, HS1(1)] (enhanced online) [URL: http://dx.doi.org/10.1063/1.3680867.13].
022104-34 Theofanous et al. Phys. Fluids 24, 022104 (2012)

FIG. 38. A fragmented TBP drop passing-by at P20. Particle sizes are in the range ∼20 to 350 μm (samples are shown in
the two zoomed areas). Run [TBP4, M = 0.15, Reg = 6.6 × 103 , We = 260, Oh = 1.8 × 10−2 , S1(2)].

of the RTP criticality. Deformation in this regime scales with a viscous timescale based on
acceleration and liquid kinematic viscosity.
(3) The SIE criticality is considerably more complex in that it is strongly modified by nonlinear
waves that break and entrain. On the viscous branch of this criticality the maximum shear stress
(found at ∼60◦ ) scales with a characteristic liquid-dynamic pressure based on the product of
the characteristic viscous and capillary velocities. On the inviscid branch, the characteristic
liquid-dynamic pressure is based only on the capillary velocity.
(4) At a basic level, SIE is seen to originate by viscous K-H waves, and this criticality is screened
by the drop length scale (λ*/ d0 ≤ 0.2)—the condition for growth is then automatically satisfied.
These kinds of analyses, juxtaposed with R-T instability analyses, show that K-H instability
always dominates (basically the acceleration obtained on a drop in a given flow is always
insufficient for placing R-T instability in a competitive advantage). Thus RTP can be seen
as the result of R-T instability left to act alone, and in concert with drop deformation, both
occurring at much longer time scales (of order t I+ or tV+ as applicable).
(5) Present visualization approach allowed quantification of the coherent portion of the drop
during the breakup transient and thereby the definition/quantification of a breakup time that
is an objective, physically defined measure of this rate process. In both the inviscid and the
viscous branches of the RTP, this complete-loss-of-coherence time is tied to the deformation
processes and subject to respective scaling laws. In the SIE regime, we have an empirical
quantification in terms of the characteristic inertial-time (t I+ ). Particle clouds can be visualized
clearly enough at all scales necessary to understand the overall, as well as the fine-internal
structures, and to obtain particle-size distributions.

ACKNOWLEDGMENTS
This work was supported by the Joint Science and Technology Office, Defense Threat Reduction
Agency (JSTO/DTRA), and the National Ground Intelligence Center (NGIC). We are grateful to Dr.
Richard Babarsky (NGIC) for his encouragement, cooperation, and support all through this work
from the very beginning. The lead construction engineer for the ALPHA and ASOS facilities was
Mr. T. Salmassi, and the lead experimentalists were (successively) Dr. G. J. Li, Dr. C. L. Ng, and
Dr. V. V. Mitkin. The lead code developer of MuSiC+ code is Dr. C-H. Chang. The early stage of
this research program was with the able participation of Dr. T. N. Dinh.

APPENDIX A: SOME KEY FEATURES OF THE AERODYNAMICS

AROUND SOLID SPHERES
We use the MuSiC+ code to probe some key details of the aerodynamics at conditions that
span the Re − M space found in our experiments (Figure 39). In particular, we are interested in
022104-35 The physics of aerobreakup. II. Viscous liquids Phys. Fluids 24, 022104 (2012)

FIG. 39. (Color online) The Reg − M space of the seven simulations presented in this appendix. The Reynolds number is
defined on free-stream quantities and a sphere diameter of 1.7 mm.

the following three aspects: (a) the potential extent of aerodynamic heating within the time period
relevant to the onset of instabilities (this is taken as TI = 0.2), (b) mapping dynamic pressure onto
maximum shear stress, as needed for better understanding of the SIE criticality, and (c) quantifying
the gas and liquid boundary layers as needed for the K-H linear stability analysis presented in
Sec. V (again, the relevant time scale is TI = 0.2). Each computational run was initiated by a shock
wave introduced at the inlet of the flow domain, initially found at atmospheric pressure. With a
nominally 1.7-mm sphere, as a typical length scale in the experiments, this defines the problem
completely. The transient response is over in about 3 μs, so here we focus on the steady state.
The code is an in-house computational platform that provides adaptive mesh refinement and cut-
cell capabilities for a finite volume method. Originally designed to simulate compressible interfacial
flows, with fully resolved, sharp interfaces, and employing the exact jump conditions at a free
interface, the code can be used also as a general purpose compressible Navier–Stokes solver.
Viscous dissipation and heat conduction to the interface are included in the present calculations. The
computation is second-order accurate in space and fourth-order accurate in time. The local resolution
is enhanced by refining in an adaptive manner the mesh according to pressure/density/vorticity
gradients and distance to a material interface. The whole procedure is benchmarked by even stricter
resolutions based on conformal meshes applied to the interface. Such are the solutions for flow
over a solid sphere presented in this appendix. In particular, the 360 × 200 conformal mesh is
concentrated around the sphere wall with the minimal grid size of 1/4000th the sphere diameter. All
calculations were performed with an isothermal sphere boundary, set at the ambient temperature,
a gas viscosity varying with temperature according to Sutherland’s law, and thermal conductivity
obtained from a constant value, 0.7, of the Prandtl number. The potential increases in interfacial
temperatures (Ti ) were estimated from the computed heat fluxes (qi ) by applying equation analogous
to (15) and the results are summarized in Table II. The simplification of constant wall temperature

TABLE II. Interfacial temperature bounds due to aerodynamic heating over the experimental range.

M ρ g (kg/m3 ) ug (m/s) d (mm) t I+ (μs) qi , 45◦ /0◦ (MW/m2 ) Ti , 45◦ /0◦ (K)a

0.31 1.60 116.5 1.81 384.2 0.0997/0.122 302/303

1.3 4.25 706.6 1.7 36.5 7.589/10.09 356/374

a The exposure time is taken as TI = 0.2. The drag coefficient experiments referred to below were conducted with copper and
iron spheres, and for these the change of surface temperature is much less.
022104-36 Theofanous et al. Phys. Fluids 24, 022104 (2012)

FIG. 40. Prediction of the classical Bailey and Hiatt17 data. The contribution of shear to the total drag diminishes in a
near-linear fashion from ∼9% to about 2% as the flow Mach umber increases from 0.3 to 1.3.

in computing the gas dynamics is indeed justified. In regards to surface tension, aerodynamic-
heating effects are negligible; at the high end of the Mach number range considered they are
less than 15%.
A sensitive measure of simulation fidelity is capturing flow separation as in turn it affects the
value of the drag coefficient. Figure 40 shows that our simulations meet this test, except for the
transonic region 0.6 < M < 0.8, which, because of the violence and complexity of flow separation,
could not be computed with the same degree of reliability over back side of the sphere and over
the complete transient. Interestingly enough, there are no experimental data in this region, which
in fact was interpolated even in the comprehensive experimental test matrix on Bailey and Hiatt.17
The 15% error in the M = 0.31 case reflects the sensitivity of the simulation to flow separation
at such flow conditions. A further benchmark of these numerical simulations can be found in
Figure 3.
The maximum shear stress is typically at 60◦ , and its variation with dynamic pressure exhibits
the effects of flow compressibility, with subsonic, and supersonic regions bridged smoothly by the
transonic one (Figure 41). As noted above, the transonic region could not be resolved in the complete
steady state (to get reliable drag coefficients for example), but the results shown in Figure 41 concern
the frontal area of the sphere and for this purpose they are deemed reliable. And so are the results
of the boundary layer thickness (Figure 42). Some details of the viscous boundary layer needed for
the stability analysis are listed in Table III.

TABLE III. Boundinga of the parameter space for viscous Kelvin–Helmholtz instability in the free shear layer corresponding
to the test of Figure 21(d) and the simulation of Figure 23.

Position μg , 10−5 (Pa · s) ρg (kg/m3 ) u g (m/s) δ g (μm) δ l b (μm) u i (m/s) Reδ Weδ

30◦ 2.06 2.31 142.3 5.2 23.1 1.03 103 14.3

45◦ 2.04 2.74 206.8 6.2 23.9 1.24 144 28.9
60◦ 2.01 2.08 256.9 7.9 24.2 1.2 210 51.6

a The other parameters of the problem, the liquid-to-gas density and viscosity ratios have extremely large values, so that the
K-H stability is already in the asymptotic regime in this regard.
b δ is the viscous boundary layer thickness defined in the same way as δ .
l g
022104-37 The physics of aerobreakup. II. Viscous liquids Phys. Fluids 24, 022104 (2012)

FIG. 41. The relation between dynamic pressure and maximum shear stress for the simulations marked in Figure 39.

APPENDIX B: LINEAR STABILITY ANALYSIS FOR FREE-SHEAR-LAYER FLOW

We consider a gas–liquid, two-layer flow in two dimensions such as that illustrated in
Figure 43. A standard normal-mode derivation, carried out by means of the linearized vorticity
equations for disturbances in the stream functions, leads to the well-known Orr-Sommerfeld system
of equations,27, 28

ϕ iv − 2κ 2 ϕ ii + κ 4 ϕ = iκ Reδ [(Ug − c)(ϕ ii − κ 2 ϕ) − Ugii ϕ]

ψ iv − 2κ 2 ψ ii + κ 4 ψ = iκ(Reδ r/m)[(Ul − c)(ψ ii − κ 2 ψ) − Ulii ψ].

In the above, ϕ i and ψ i are the ith derivatives of the amplitudes in the disturbance stream functions,
 and , defined for the gas and liquid flow-domains respectively as,

(x, y, t) = ϕ(y) exp [iκ(x − ct)] (x, y, t) = ψ(y) exp [iκ(x − ct)] ,

FIG. 42. The viscous boundary layer thickness dependence on the flow Mach number for the simulations in Figure 39.
Values are given at a polar angle of 45◦ , and the outer “edge” taken at 85% of the free-stream velocity.
022104-38 Theofanous et al. Phys. Fluids 24, 022104 (2012)

FIG. 43. (Color online) Base-flow velocity profiles (not to scale) and related nomenclature. On a reference frame moving
with the interface Ug∗ = u g − u i and Ul∗ = −u i . The gas and the liquid boundary layers, obtained from MuSiC+ simulations
and (15), respectively, are well approximated by error functions that match both the shear stress at the interface and the
free-stream velocities.

and Ug (y)/ Ul (y) are the gas/liquid mean (base-flow) velocities. The wavelength of the disturbance
is λ = 2π /κ, and the complex wave-speed is c, the real part of which, cR , is the phase speed, and the
imaginary, cI , gives the growth factor of wave κ as n = κcI . It can be seen that the problem depends
on the Reδ = ρ g ug δ g /μg , and the property ratios m = μl /μg , r = ρl /ρg . The boundary conditions
at the interface bring in the fourth parameter, W eδ = ρg u 2g δg /σ ,

ϕ=ψ
c(ϕ i − ψ i ) = ϕ(Uli − Ugi )
 ii 
c ϕ + κ ϕ + Ugii ϕ = mc ψ ii + κ 2 ψ + mUlii ψ
2

−iκ Reδ (cϕ  + ϕUg ) − ϕ  + 3κ 2 ϕ i + iκ Reδ r (cψ i + ψUli ) + m(ψ iii −3κ 2 ψ i ) = iκ 3 Reδ ϕ/cW eδ .

The above system of equations is transformed into a linear eigenvalue problem in terms of Chebyshev
polynomials using the collocation method,20 and solved by the QZ algorithm.21 The outside bound-
aries of the domain are taken so that the solution becomes independent of this selection (typically
5 times the boundary layer thickness), and the collocation grid is refined so as to resolve gradients
by using the virtual-interface method.20 At the outer boundaries, the stream function disturbances
and their derivatives are set to zero. At a virtual interface, the same set of equations as above apply
but with the same property values and same velocity gradients on both sides. In the calculations
reported here we use two layers of grids in each phase and 30 nodes for each layer; this was found to
be quite adequate for convergence. The numerical tool (AROS code) has been extensively verified
with previously published results.35, 36
1 T. G. Theofanous and G. J. Li, “On the physics of aerobreakup,” Phys. Fluids 20, 052103 (2008).
2 J. O. Hinze, “Fundamentals of the hydrodynamic mechanism of splitting in dispersion processes,” AIChE J. 1, 289 (1955).
3 R. S. Brodkey, The Phenomena of Fluid Motions (Addison Wesley, Reading, MA, 1967).
4 B. E. Gelfand, “Droplet breakup phenomena in flows with velocity lag,” Prog. Energy Combust. Sci. 22, 201 (1996).
5 J. F. Heagy and S. D. Kramer, Reentry breakup of thickened chemical agents, Institute for Defense Analysis Paper:

IDA-3446 (1998).
6 D. D. Joseph, G. S. Beavers, and T. Funada, “Rayleigh–Taylor instability of viscoelastic drops at high Weber numbers,”

J. Fluid Mech. 453, 109 (2002).

7 E. Y. Harper, G. W. Grube, and I. Chang, “On the breakup of accelerating liquid drops,” J. Fluid Mech. 52, 565 (1972).
8 T. G. Theofanous, “Aerobreakup of Newtonian and viscoelastic liquids,” Annu. Rev. Fluid Mech. 43, 661 (2011).
9 J. O. Hinze, “Critical speeds and sizes of liquid globules,” Appl. Sci. Res., Sect. A 1, 273 (1949).
022104-39 The physics of aerobreakup. II. Viscous liquids Phys. Fluids 24, 022104 (2012)

10 L. P. Hsiang and G. M. Faeth, “Drop deformation and breakup due to shock-wave and steady disturbances,” Int. J.
Multiphase Flow 21, 545 (1995).
11 T. G. Theofanous, G. J. Li, T. N. Dinh, and C. H. Chang, “Aerobreakup in disturbed subsonic and supersonic flow fields,”

J. Fluid Mech. 593, 131 (2007).

12 V. V. Mitkin,, A. N. Rozhkov, and T. G. Theofanous, “Pulse jets, rims and elastic-liquid sheets: Rheology at high strain

rates and rupture criteria,” AIP Conf. Proc. 1027, 1126 (2008).
13 C.-L. Ng and T. G. Theofanous, “Modes of aero-breakup with visco-elastic liquids. Pulse jets, rims and elastic-liquid

sheets: Rheology at high strain rates,” AIP Conf. Proc. 1027, 183 (2008).
14 D. Kivotides, V. V. Mitkin, and T. G. Theofanous, “Mesoscopic dynamics of polymer chains in high strain rate extensional

flows,” J. Non-Newtonian Fluid Mech. 161 (1–3), 69 (2009).

15 G. I. Taylor, “The shape and acceleration of a drop in a high speed air stream,” in The Scientific Papers of G.I. Taylor,

edited by G. K. Batchelor (Cambridge University Press, Cambridge, 1949).

16 A. A. Ranger and J. A. Nicholls, “Aerodynamic shattering of liquid drops,” AIAA J. 7, 285 (1969).
17 A. B. Bailey and J. Hiatt, “Sphere drag coefficients for a broad range of Mach and Reynolds numbers,” AIAA J. 10, 1436

(1972).
18 R. R. Nourgaliev, M. S. Liou, and T. G. Theofanous, “Numerical prediction of interfacial instabilities: Sharp interface

method (SIM),” J. Comput. Phys. 227, 3940 (2008).

19 C. H. Chang, X. Deng, and T. G. Theofanous, “Numerical prediction of interfacial instabilities: The sharp-interface method

for compressible flows,” Paper No. AIAA 2011-3834. Proceedings of the 20th AIAA Computational Fluid Dynamics
Conference, 27–30 June 2011, Honolulu, Hawaii (American Institute of Aeronautics and Astronautics, Reston, Virginia,
2011).
20 P. A. M. Boomkamp, B. J. Boersma, R. H. M. Miesen, and G. V. Beijnon, “A Chebyshev collocation method for solving

two-phase flow stability problems,” J. Comput. Phys. 132, 1424 (1997).

21 G. Golub and C. Van Loan, Matrix Computations (Johns Hopkins University Press, Baltimore, 1996).
22 K. O. Mikaelian, “Rayleigh–Taylor instability in finite-thickness fluids with viscosity and surface tension,” Phys. Rev. E

54, 3676–80 (1996).

23 Chanrdashekar, Hydrodynamic and Hydromagnetic Stability (Dover, New York, 1981).
24 H. W. Liepmann and A. Roshko, Elements of Gas-Dynamics (Wiley, New York, 1957).
25 W. G. Reinecke and G. D. Waldman, The study of drop breaking behind strong shocks with applications to flight, Report

No. SAMCO-TR-70-142, AVCO Systems Division Report (1970).

26 G. D. Waldman, W. G. Reinecke, and D. C. Glenn, “Raindrop breakup in shock layer of a high-speed vehicle,” AIAA J.

10, 1200 (1972).

27 P. G. Drazin and W. H. Reid, Hydrodynamic Stability (Cambridge University Press, Cambridge, 1981).
28 C. S. Yih, “Instability due to viscosity stratification,” J. Fluid Mech. 27(2), 337 (1967).
29 P. A. M. Boomkamp and R. H. M. Miesen, “Classification of instabilites in parallel flow,” Int. J. Multiphase Flow 22

(Suppl.), 67 (1996).
30 A. R. Hanson, E. G. Domich, and H. S. Adams, “Shock tube investigation of the breakup of drops by airblasts,” Phys.

Fluids 6, 1070 (1963).

31 M. Pilch and C. A. Erdman, “Use of breakup time data and velocity history data to predict the maximum size of stable

fragments for accelaration-induced breakup of a liquid drop,” Int. J. Multiphase Flow 13, 741 (1987).
32 H. E. Wolfe and W. H. Andersen, Kinetics, mechanism, and resultant droplet sizes of the aerodynamic breakup of liquid

drops. Aerojet General Corporation, Downey Plant, California, Report No. 0395-04(18)SP (1964).
33 O. Engel, “Fragmentation of waterdrops in the zone behind an air shock,” J. Res. Natl. Bur. Stand. 60, 3, 245 (1958).
34 W. G. Reinecke, G. D. Waldman, W. L. McKay, and M. B. Ziering, Shock layer shattering of water drops and ice crystals

in reentry flight. AVCO Systems Division Report. AFML-TR-75-71, DTIC No. ADB006836 (1975).
35 T. G. Theofanous, R. R. Nourgaliev, and B. Khomami, “Short communication: An experimental/theoretical investigation

of interfacial instabilities in superposed pressure-driven channel flow of Newtonian and well characterized viscoelastic
fluids - Part I. Linear stability and encapsulation effects,” by B. Khomami and K. C. Su, J. Non-Newtonian Fluid Mech.
143, 131 (2007).
36 T. G. Theofanous, R. R. Nourgaliev, and S. Wiri, “Brief Communication: Direct numerical simulations of two-layer

viscosity-stratified flow, by Qing Cao, Kausik Sarkar, Ajay K. Prasad, Int. J. Multiphase Flow (2004) 30, 1485–1508,” Int.
J. Multiphase Flow 33, 789 (2007).
37 T. G. Theofanous, G. J. Li, and T. N. Dinh, “Aerobreakup in rarefied supersonic gas flows,” Trans. ASME 126, 516 (2004).