You are on page 1of 20

An enstrophy-based linear and nonlinear receptivity theory

Aditi Sengupta, V. K. Suman, Tapan K. Sengupta, and Swagata Bhaumik

Citation: Physics of Fluids 30, 054106 (2018); doi: 10.1063/1.5029560


View online: https://doi.org/10.1063/1.5029560
View Table of Contents: http://aip.scitation.org/toc/phf/30/5
Published by the American Institute of Physics

Articles you may be interested in


Non-linear instability analysis of the two-dimensional Navier-Stokes equation: The Taylor-Green vortex
problem
Physics of Fluids 30, 054105 (2018); 10.1063/1.5024765

Shed vortex structure and phase-averaged velocity statistics in symmetric/asymmetric turbulent flat plate
wakes
Physics of Fluids 30, 055104 (2018); 10.1063/1.5025946

Boundary layers at the interface of two different shear flows


Physics of Fluids 30, 053604 (2018); 10.1063/1.5025721

Three-dimensional numerical investigation of vortex-induced vibration of a rotating circular cylinder in uniform


flow
Physics of Fluids 30, 053602 (2018); 10.1063/1.5025238

Vortical structures and development of laminar flow over convergent-divergent riblets


Physics of Fluids 30, 051901 (2018); 10.1063/1.5027522

Analysis of disturbances in a hypersonic boundary layer on a cone with heating/cooling of the nose tip
Physics of Fluids 30, 054103 (2018); 10.1063/1.5024025
PHYSICS OF FLUIDS 30, 054106 (2018)

An enstrophy-based linear and nonlinear receptivity theory


Aditi Sengupta,1 V. K. Suman,2,3 Tapan K. Sengupta,3,a) and Swagata Bhaumik4
1 Department of Engineering, University of Cambridge, Cambridge, United Kingdom
2 Scientist,
National Aerospace Laboratory Bangalore, Karnataka 560 017, India
3 High Performance Computing Laboratory, Department of Aerospace Engineering,

IIT Kanpur, Kanpur 208 016, India


4 Department of Mechanical Engineering, IIT Jammu, Jammu 181221, India

(Received 15 March 2018; accepted 2 May 2018; published online 25 May 2018)

In the present research, a new theory of instability based on enstrophy is presented for incompressible
flows. Explaining instability through enstrophy is counter-intuitive, as it has been usually associ-
ated with dissipation for the Navier-Stokes equation (NSE). This developed theory is valid for both
linear and nonlinear stages of disturbance growth. A previously developed nonlinear theory of incom-
pressible flow instability based on total mechanical energy described in the work of Sengupta et al.
[“Vortex-induced instability of an incompressible wall-bounded shear layer,” J. Fluid Mech. 493,
277–286 (2003)] is used to compare with the present enstrophy based theory. The developed equa-
tions for disturbance enstrophy and disturbance mechanical energy are derived from NSE without
any simplifying assumptions, as compared to other classical linear/nonlinear theories. The theory is
tested for bypass transition caused by free stream convecting vortex over a zero pressure gradient
boundary layer. We explain the creation of smaller scales in the flow by a cascade of enstrophy, which
creates rotationality, in general inhomogeneous flows. Linear and nonlinear versions of the theory
help explain the vortex-induced instability problem under consideration. Published by AIP Publishing.
https://doi.org/10.1063/1.5029560

I. INTRODUCTION three-dimensional roughness effects in internal and external


flows.27,37 In some of these flows, disturbances grow tempo-
One of the principal concerns of transition to turbu-
rally due to the presence of inflection point(s) in the velocity
lence is the route by which external ambient perturbations are
profile. Plane Poiseuille flow profile has no inflection point,
internalized via receptivity, which in turn triggers instabili-
and linear spatial theory predicts a critical Reynolds number
ties in the fluid flow. Routes of transition to turbulence have
of 5772. However, this flow has been experimentally observed
been primarily classified into two main types:31 (i) the clas-
to become turbulent at a sub-critical Reynolds numbers of
sical route, wherein the onset is marked by the presence of
around 1000.8 In Fig. 3.1 presented by Morkovin,31 specific
Tollmien-Schlichting (TS) waves, as in the case of zero pres-
mechanisms of bypass transition are not identified. Instead
sure gradient boundary layer flow,43 and (ii) all other mech-
it has been hinted as being caused by nonlinear, nonparallel
anisms as bypass routes, for example, those caused by free
effect(s) and/or some altogether unknown mechanisms. In this
stream excitations,21,41,65,66 which do not exhibit TS waves.
context, one such mechanism for bypass transition based on
The classical route has been studied exhaustively,14,44 while
vortex-induced instability has been identified theoretically49
the latter is the subject of many studies for different internal
and experimentally,26 for zero pressure gradient boundary
and external flows.42,44 Some researchers have defined bypass
layer. The same mechanism is also identified to cause bypass
transition as similar to the secondary instability and break-
transition for the flow at the leading edge of an aircraft swept
down process in boundary layer natural transition noted for
wing, on the attachment-line plane.46 The physical mecha-
the classical route.66 The narrow definition of bypass tran-
nism for this bypass transition has also been explained with the
sition as being caused by free stream excitation29 is also
help of an instability theory based on disturbance mechanical
strictly not correct, as researchers have been able to create TS
energy (DME) derived from the incompressible Navier-Stokes
waves by vibrating a ribbon in the free stream,25 which can be
equation (NSE) without any underlying assumptions.49 One
explained by linear theory governed by the Orr-Sommerfeld
of the successful applications of this energy-based receptiv-
equation (OSE).44,48
ity theory has been in showing the presence of DME waves
The above generic definition of Morkovin31 has also been
in the work of Sengupta et al.52 for classical instability
used by other researchers4,44 to identify bypass routes. There
described by the OSE for zero pressure gradient boundary
are flows which undergo transition without TS waves, like
layer.
flow past bluff bodies, Couette and pipe flows,31 leading edge
While solving the OSE and NSE for localized harmonic
contamination on swept wings,46 and two-dimensional and
wall excitation of a zero pressure gradient boundary layer
flow, one notices a three-element response field: (a) a local
a) Electronic mail: tksen@iitk.ac.in solution—which is in the immediate neighbourhood of the

1070-6631/2018/30(5)/054106/19/$30.00 30, 054106-1 Published by AIP Publishing.


054106-2 Sengupta et al. Phys. Fluids 30, 054106 (2018)

exciter; (b) an asymptotic solution comprising of TS waves; experimental observations.26 While the experiments display
and (c) the spatio-temporal wave-front (STWF),2,47,51 which the 2D nature of the primary instability, it is essential to
convects downstream of the former two elements. While DME compute such a flow using 3D formulation of the NSE to
waves can trace the disturbance field clearly,44,52 showing the trace the transition to turbulence. While ( , !)-formulation is
asymptotic solution and the STWF, these are smoother, as com- known for its superior accuracy, it cannot be directly extended
pared to the corresponding elements obtained by solving the for 3D simulations, and we instead use a vorticity-velocity
OSE and NSE. Thus, it is preferred to identify vortical struc- formulation, which is used here for both 2D and 3D flow
tures by nonlinear diagnostic tools, which do not remove the fields.
high wavenumber components. In this context, an appropriate Using velocity-vorticity formulation for simulating natu-
identifier of vortical structures is enstrophy [vorticity squared, ral 3D transition caused by wall excitation from the receptivity
(! ~ denoted here as ⌦1 . Although it has been pointed out
~ · !)], of the laminar flow to the fully developed stage of turbulent
that diffusion may cause the vorticity to increase or decrease,38 flow,2,3 the authors have shown excellent agreement with the
there seems to be confusion regarding all the roles of enstro- experiment of Klebanoff et al.23 In the present research, we
phy, including growth and decay of disturbances. Doering and have employed the 2D formulation and associated numerical
Gibbon,10 however, have shown for periodic 2D flows that methods. The results are used to demonstrate the relevance of
the total enstrophy (||!||22 = s s ⌦1 dx dy) integrated over the the instability theory based on ⌦d to explain the physical mech-
periodic domain is given by anism of bypass transition, while validating the numerical
D results with the designed receptivity experiment.26
||!||22 = ⌫||!||22 , (1) Ever since the experimental verification of TS waves,43
Dt
the classical route has become the main tool to predict transi-
D
where ⌫ is the kinematic viscosity and Dt is the substantive tion for flows past streamlined bodies. The experiment was
derivative. The negative right-hand side of a nominally positive successful after many attempts, and this was realized only
quantity indicates that total enstrophy must decay. However, with the creation of an ultra-quiet wind tunnel. In this facility,
to study the growth of disturbances over an equilibrium state monochromatic time harmonic excitation inside the bound-
given by, ⌦m , one needs to investigate the evolution of dis- ary layer created a TS wave-packet, while acoustic distur-
turbance enstrophy (⌦d ), where ⌦1 = ⌦m + ✏ 1 ⌦d . In terms bances from outside were not effective in creating the same.
of equilibrium and disturbance vorticity fields (!! !
m and !d , Thus, experimental verification of the transition process is
respectively), one can define far removed from the natural transition in many respects.
The major distinction comes from the fact that in a natu-
⌦ = !! · !! and ⌦ = 2!! · !!.
m m m d m d
ral setting, the boundary layer is excited simultaneously by
In addition, ⌦d can also have an additional lower order pos- multiple frequencies of vortical and/or entropic and/or acous-
itive contribution coming from !!d · !!d . Such a study for the tic disturbances. As a consequence, the disturbance field is
evolution of ⌦d has not been performed before. a combination of multiple responses for multiple frequencies,
Instead, the implication of Eq. (1) developed for homo- modulating each other. One such convolution has been demon-
geneous 2D flow has been extrapolated to inhomogeneous 3D strated by considering multi-periodic excitation in the work of
flows, and it has been claimed that the primary role of enstro- Sengupta et al.51 for Blasius boundary layer, with constituent
phy is related to viscous dissipation, as the right-hand side TS waves being all unstable, yet the composite response field
originates from diffusion term and its negative sign indicates superposed from solutions of the OSE remains subdued for a
dissipation of total enstrophy.12,13,22,67 However, it has been long streamwise stretch. This is explained to be due to wave
shown using enstrophy transport equation (developed from cancellation by phase shift among the constituents. Thus, TS
vorticity transport equation) that total enstrophy can grow for waves can be noted in extremely controlled experiments by
inhomogeneous flows. The idea that enstrophy transport equa- monochromatic excitation of zero pressure gradient boundary
tion can be used to trace rotationality was shown for flow inside layers.
a rectangular cavity.55 Here, a detailed analysis is made to iden- The experiment performed for vortex-induced instabil-
tify the criteria for growth of ⌦d and quantify the growth rate, ity26 was designed to test the hypothesis that a distant vortex
for an inhomogeneous flow experiencing vortex-induced insta- can induce a small longitudinal adverse pressure gradient to
bility, by developing an evolution equation for ⌦d . To develop destabilize a wall-bounded flow, an idea mooted to study
this equation, we have focused mainly on a 2D flow. For 3D dependence of critical Reynolds number on free-stream turbu-
flows, vortex stretching can additionally contribute to growth lence (FST).60 Monin and Yaglom,30 in discussing this idea,
of enstrophy,38,55 and hence a 2D flow will provide a more crit- noted furthermore that the change in critical Reynolds number
ical limit. Additionally, we also provide 3D simulation results by a small longitudinal adverse pressure gradient is due to a
of the same flow in Appendix B, to show the centrality of 2D sequence of unsteady separations, presumably created by vor-
nature of vortex-induced instability during the early primary tices embedded in the free-stream turbulence. An assumption
and secondary stages. implicit in this is that the effect is connected to the gener-
The 2D computations of vortex-induced instability have ation of longitudinal pressure gradient by the disturbances,
been performed for the free stream convecting vortex prob- leading to the formation of individual spots.30 The study of
lem49 using stream function ( )-vorticity (!) formulation. The vortex-induced instability is thus a kernel experiment to check
flow field undergoing this instability displays unsteady sepa- the unit process of how a distant vortex can affect the zero
rations, resulting in strong vortical structures, corroborating pressure gradient boundary layer.
054106-3 Sengupta et al. Phys. Fluids 30, 054106 (2018)

This model experiment and computational study will According to these authors, “it is possible to understand
help one understand the role of FST in causing transition such behavior by studying the redistribution of the total
that is initiated by vortex-induced instability without creat- mechanical energy of the flow.” Instead of characterizing tur-
ing TS waves.26,49 A shear layer can interact with a finite- bulence by TKE, one must include the roles of fluctuating
core vortex convecting at a distance far outside the boundary pressure, as the unsteadiness during bypass transition is due
layer leading to unsteady separation on the wall. The prob- to the shear noise term in the Poisson equation for the static
lem of vortex-induced instability has been noted as one of pressure.31 This prompted Sengupta et al.49 to develop equa-
the most important unsolved problems of fluid dynamics.11 tions for the mechanical energy from incompressible NSE,
This problem can have many variations, but we only con- for both the equilibrium and the disturbance flow fields. In
sider the case where the primary instability is triggered by the following, we derive the equation for mechanical energy
a convecting vortex over the zero pressure gradient bound- of incompressible fluid flows. In the NSE, we can use the
ary layer.26,49 The same physical mechanism is seen in other following vector identity for the convective acceleration term:
examples of unsteady flow separation9 and near-wall eddy ! ! 1 ! ! !
formation in turbulent boundary layers.5 Authors have also ( V · r) V = r(| V | 2 ) V ⇥ (r ⇥ V )
2
discussed formation of hairpin vortices in the near-wall region
of turbulent flows.39,57 Another study considered the scenario to yield the NSE as
where a vortex placed above a plane wall caused the vortex to !
@V 1 ! 2 ! ! rp !
move and thereby create a thin unsteady boundary layer over + r(| V | ) V ⇥ ! = + ⌫r2 V . (2)
@t 2 ⇢
the wall.35,36
In a designed experiment,26,49 the authors modeled the Furthermore, one can alter the diffusion term with the
unit process for effects of FST, by convecting a vortex at con- following identity:
stant height and speed over a zero pressure gradient boundary ! ! !
r2 V = r(r · V ) r ⇥ (r ⇥ V )
layer. The key feature of the experiment is that the ensu-
ing instability is sub-critical with respect to linear instability so that one gets the rotational form of the NSE as
(as governed by the OSE) and is due to unsteady separation. ! ✓ p |!
@V ! ! V |2 ◆
Linear theories to study vortex-dominated flows do not exist, V ⇥! = r + ⌫(r ⇥ !
!). (3)
and hence a nonlinear receptivity theory for incompressible @t ⇢ 2
flow has been developed using DME,49 briefly described in Define the total mechanical energy as
the next section, Sec. II. Here, we develop another enstrophy- !
based theory of instability for incompressible flows emphasiz- p | V |2
E= +
ing on the creation of rotationality and small scales in flows, ⇢ 2
as created by enstrophy cascade. The NSE is alternatively written with E as
The paper is formatted as follows. We provide a brief !
description of receptivity theory of flow based on total mechan- @V ! ! !
V ⇥ ! = rE + ⌫r2 V . (4)
ical energy in Sec. II. In Sec. III, we describe the numerical @t
simulation of a bypass transition caused by free stream con- Taking a divergence and making incompressibility
vecting vortex over a zero pressure gradient boundary layer. assumption for the equation above, one gets the governing
In the same section and in Appendix B, numerical details Poisson equation for E as
including the formulation and boundary conditions are !
described. In Sec. IV, we describe the instability mechanism r2 E = r · ( V ⇥ !
!). (5)
with the help of numerical results and the mechanical energy
Further using the identity,
based instability theory developed in Sec. II. In Sec. V, we
! ! !
provide an explanation for the developed instability theory r · (V ⇥ !
!) = !! · (r ⇥ V ) V · (r ⇥ !
!),
based on enstrophy transport equation. We also describe its
connection to the creation of smaller scales by cascade of one finally gets the governing equation for the total mechanical
enstrophy at different scales. In Sec. VI, we develop the new energy as
!
theory of instability based on enstrophy directly from the r2 E = !
! ·!! V · (r ⇥ ! !). (6)
NSE to explain creation of rotationality in any inhomoge- We note that in the absence of body force, the expression
neous flow. We also describe the linearized version of the of E is identical to Bernoulli’s head often used to study invis-
theory in this section for the vortex-induced instability prob- cid, irrotational flows, as the energy equation. Thus, the present
lem under consideration. The paper closes with a summary derivation for the boundary value problem of E is a general-
and conclusion. ization of the same, in the context of NSE. Despite the fact
that no simplifying assumption is invoked, apart from enforc-
ing incompressibility assumption (making the velocity field
II. NONLINEAR INSTABILITY THEORY BASED
solenoidal or divergence free), the contribution to E comes
ON MECHANICAL ENERGY
essentially from the convective acceleration and pressure gra-
Landahl and Mollo-Christensen24 noted that, to under- dient term, with the local acceleration and the viscous actions
stand turbulence, one must consider growth of total mechan- do not play any role in this derivation. This can be, therefore,
ical energy and not just the turbulent kinetic energy (TKE). viewed as generalization of existing linear inviscid theories.
054106-4 Sengupta et al. Phys. Fluids 30, 054106 (2018)

The distribution of E is seen to be forced by the enstrophy vortex-induced instability problem; the same formulation can
(!
! ·! !). Thus, rotationality appearing in the flow by viscous be used for both 2D and 3D flows.
diffusion or other sources like shock waves generating entropy
gradient (as stated in Crocco’s theorem relating vorticity with III. NUMERICAL SIMULATION OF VORTEX-INDUCED
entropy gradient 54,56 ) will affect the distribution of E, through INSTABILITY
its forcing by enstrophy.
Without solving Eq. (6), one can qualitatively make state- Vortex-induced instability problem is used here as a typ-
ments about the solution, by the sign of the right-hand side of ical case to demonstrate bypass transition caused by a con-
the Poisson equation.58 It indicates the presence of a source or vecting vortex in the free stream. Bhaumik and Sengupta
a sink of E, with a negative/positive sign, respectively. Thus, have developed a velocity-vorticity formulation in a stag-
it may appear that the enstrophy stabilizes E in Eq. (6). For gered grid to solve the 3D receptivity problem from excitation
nonlinear evolution of E, destabilization of the equilibrium of laminar flow to a fully developed turbulent flow stage.2,3
state can be studied, if one expresses E as E = E m + ✏E d , with In Appendix B, we show some preliminary results for the
subscripts m and d signifying the equilibrium and disturbance vortex-induced instability problem solved numerically for a
components, respectively. If one is interested in the growth 3D perturbation field, for the same case of 2D perturbation
of small perturbation applied to the equilibrium state, then caused by convecting vortex in the free stream. This was the
the parameter ✏ represents a small value, and one can apply scenario of the receptivity experiment reported26 and here the
the same splitting scheme for all the quantities appearing in intention is to show that the flow field remains 2D, even when
Eq. (6). The equation for disturbance energy, E d , is obtained we solve 3D NSE using the velocity-vorticity formulation,
by subtracting the equation for the equilibrium state from the using identical numerical methods. Here, the same formulation
equation for the instantaneous state as is used for the 2D flow field shown in Fig. 1, for the receptivity
of zero pressure gradient boundary layer to convected vortex
! in the free stream. The time-dependent boundary forcing by
r2 Ed = ⌦d + ✏!!d ·!! d V m · (r ⇥ ! !d)
the convecting vortex excites the physical instability of the
! !
V d · (r ⇥ !
! m ) ✏ V d · (r ⇥ !
! d ). (7) flow. The equilibrium flow is calculated by solving the NSE,
without the external forcing. The origin of the co-ordinate
Existence of waves for E d caused by spatially localized system is placed at the leading edge of the plate, whereas
excitation for a boundary layer has been reported,51,52 by tak- the computational domain starts slightly ahead of it. A vor-
ing the equilibrium state as the Blasius profile and solving tex with counter-clockwise circulation ( ) translates from left
the linearized version of Eq. (7). In this approach, growth of to right with a constant speed, c, at a constant height, H, above
primary disturbances is due to interactions of velocity and vor- the plate. This free stream vortex convecting far outside the
ticity fields acting as source terms on the right-hand side of shear layer requires an image vortex below the plate, to ensure
Eqs. (6) and (7). Interestingly, the lead term on the right-hand zero wall-normal velocity boundary condition, as shown in the
side is nothing but the disturbance enstrophy ⌦d = 2! ! m ·!
!d. figure.
Thus, as noted before, ⌦d can have both positive and nega-
tive signs, with the negative sign implying a source of DME. A. Velocity-vorticity formulations for 2D flows
Thus, the major issue is about how the energy is initially The vorticity transport equation is found relevant for the
exchanged from the equilibrium to the disturbance field and solution of the NSE, which avoids the pressure-velocity cou-
this is clearly brought out by the first term on the right-hand pling issue, while computing the vorticity field directly. The
side of Eq. (7), which indicates an interaction between equi- corresponding velocity field is obtained by solving the veloc-
librium and disturbance vorticity fields or the disturbance ity Poisson equation. Here, the unknowns are velocity com-
enstrophy. ponents, u, 3, and the out-of-plane vorticity component, !.
The NSE is a consequence of conservation of translational For 2D flows, the non-dimensional governing equations for
momentum, and viscous terms are absent in Eqs. (5) and (6). !
( V , !)-formulation in Cartesian co-ordinate are as given in
Despite this, instability can arise for E d , due to forcing from the prior studies.7,16,33,64
right-hand side by enstrophy and vorticity gradient. However,
we know that the vorticity field and hence the enstrophy even in
the linear form is governed by the OSE. In hindsight, one can
show the correlation between Rayleigh’s inviscid instability
equation and the OSE to be in the same vein, as one would
relate E d with ⌦d . Thus, to study the nonlinear instability of
fluid flow, there is a specific need to develop an evolution
equation for ⌦d , over and above the one, that is provided by
the equation for DME. This is the motivation for the present
research.
To develop such a theory for ⌦d , we revisit the vortex-
induced instability problem through which the governing
FIG. 1. Computational domain used in studying vortex-induced instability
equation for DME has been developed.26,49 Here, the numer-
! by an isolated convecting (at a speed, c) vortex (of circulation, ) in the free
ical simulation is reported using ( V ,! !)-formulation for the stream over a flat plate supporting a zero pressure gradient boundary layer.
054106-5 Sengupta et al. Phys. Fluids 30, 054106 (2018)

Here, we solve the governing equation in the transformed where d is the core diameter of the vortex of strength
plane by a high accuracy compact scheme to discretize con- , convecting at a constant height, H. The convection
vection terms. Details about the numerical methods, grids speed of the vortex is given by c, and hence the dis-
used, etc., are given in Appendix A. The governing vorticity placement effect of the finite core vortex is determined
transport and velocity Poisson equations in the transformed by the relative speed, (U 1 c), and the circulation
(⇠, ⌘)-plane (with ⇠ along streamwise direction and ⌘ in the effect is given by the last term on the right-hand side.
wall-normal direction) are given as Also, the time dependence of the boundary condition
@! 1 @ 1 @ is a function of the translated co-ordinate, x̄ = x0 ct,
+ (u!) + (v!) with x 0 indicating the initial position of the vortex.
@t h1 @⇠ h2 @⌘
" ! # (ii) At the wall, no-slip condition is imposed on both the
1 1 @ h2 @! @ ✓ h1 @! ◆ components of velocity. Wall vorticity is computed
= + , (8)
Re h1 h2 @⇠ h1 @⇠ @⌘ h2 @⌘ based on its kinematic definition.
" # (iii) In the segment AO of Fig. 1, ! and 3 are zero due to
@ ✓ h2 @u ◆ @ ✓ h1 @u ◆ @! symmetry. The boundary condition on u is fixed by the
+ = , (9)
@⇠ h1 @⇠ @⌘ h2 @⌘ @⌘ condition given by @u/@y = 0.
" ✓ # (iv) At the far-field boundary: ! = 0 and (u, 3) are calculated
@ h2 @v ◆ @ ✓ h1 @v ◆ @!
+ = , (10) by the Biot-Savart law caused by the free stream vortex
@⇠ h1 @⇠ @⌘ h2 @⌘ @⇠ and its image.
where h1 and hq2 are the scale factors
q of the transformation (v) At the outflow, ! is calculated using the Sommerfeld
given by h1 = x⇠ + y⇠ and h2 = x⌘2 + y⌘2 . The divergence
2 2 boundary condition given by
! @! @!
of the velocity field (Dv = r · V ) in the (⇠, ⌘)-plane is given + Uc = 0,
as " # @t @x
1 @(h1 v) @(h2 u)
Dv = + . (11) where U c is chosen as U 1 . The condition used on u is @ 2 u/@x 2
h1 h2 @⌘ @⇠ = 0, whereas the condition for 3 is given by @3/@x = ! + @u/@y.
For wall excitation problems, after solving Eq. (9) for u, In the experimental study,26 it is shown that a vortex with
the 3-component is obtained by integrating the right-hand side anti-clockwise circulation creates instability ahead of it, due
of Eq. (11) from the solenoidality condition given as to the imposed adverse pressure gradient by the convecting
⌅ " # free stream vortex. Boundary conditions created by the finite
1 ⌘ @(h2 u)
v(⇠, ⌘) = v(⇠, 0) d⌘. (12) core translating vortex is shown in Fig. 2, indicating the dis-
h1 0 @⇠
placement and circulatory components of the imposed stream
The far-field boundary condition on 3 velocity is satis- function on the far-field boundary in frame (i). The associated
fied asymptotically and hence, Eq. (12) is suitable for exter- induced pressure gradient, at the edge of the shear layer, is
nal flows, while numerically ensuring solenoidality. For the shown in frame (ii). It will be shown from the numerical solu-
present investigation of instability caused by free stream exci- tion of the NSE that unsteady separation occurs at a lower value
tation, this approach is not applicable. The alternative is given of adverse pressure gradient than that is necessary for steady
in the next Subsection III B. The above non-dimensionalized separation44 as shown by the dotted line in frame (iv). The
equations have been obtained with a length scale and the imposed pressure gradient at the edge of the undisturbed shear
free-stream speed of the oncoming flow as the velocity scale. layers is indicated here by plotting the Falkner-Skan parame-
From these two scales, the time scale is constructed and all ter, m, plotted as functions of space and time in the bottom two
computational results are in non-dimensional units. frames of Fig. 2. This figure should be contrasted with Figs.
4.8(a) and 4.8(b) in the work of Sengupta,44 where the induced
B. Boundary conditions adverse pressure gradient by the free stream convecting vortex
of strengths = 14 and 30 indicated values of m which were
The boundary conditions used in solving this problem are far in excess of what is needed for steady separation. Thus,
listed as follows: for vortex-induced instability, it is not merely the strength of
(i) At the inflow (segment AB of Fig. 1), uniform inlet veloc- adverse pressure gradient that triggers unsteady separation, but
ity U 1 is imposed along with the contribution imposed the time duration over which the adverse pressure gradient acts
by the free stream vortex and its image system. A corre- matters more.
sponding condition is applied on 3-velocity component,
as computed by the free stream inviscid vortex and its
C. Numerical results and discussion
image. The induced stream function created by the finite
core translating vortex in the inviscid part of the flow is To solve the governing equations in the computational
given as domain of Fig. 1, the parameters are similar to those in the
" # experiments of Lim et al.,26 except for the strength of the
(y H)(d/2)2 (y + H)(d/2)2 convecting vortex, which cannot be measured and is treated
1 = U1 y (U1 c) 2 +
x̄ + (y H)2 x̄ 2 + (y + H)2 here as a parameter of the problem. In solving the problem
x̄ 2 + (y + H)2 computationally, we define a length scale, L, based on which
+ Ln 2 ,
4⇡ x̄ + (y H)2 the Reynolds number is 105 . The computational domain in
054106-6 Sengupta et al. Phys. Fluids 30, 054106 (2018)

FIG. 2. Imposed disturbances at y = 1.0


by a convecting vortex at x c = 5 and
H = 2: (i) Disturbance stream function
versus x caused by circulatory and dis-
placement effects by the free stream vor-
tex moving at c = 0.3. (ii) Induced pres-
sure gradient for the case of frame (i) for
the indicated propagation speeds. (iii)
The pressure gradient shown as a func-
tion of time for various convecting free
stream vortex cases. (iv) Variation of
Falkner-Skan parameter m with x shown
at two representative times for the case
of c = 0.3 at t = 9.5 for plotting m.

the streamwise direction is 0.05  x ⇤  20 and includes the frame at t = 20, the free stream vortex is just downstream of
leading edge of the plate. The height of the computational the leading edge of the plate. The vortex with anti-clockwise
domain is taken as ymax ⇤ = 0.75, with the wall resolution circulation lifts up the boundary layer ahead of it. At the
given by y = 3.688 ⇥ 10 4 , which is half the wall resolu- foot of the vortex (outside the boundary layer), the induced
tion taken in the work of Sengupta et al.49 For the present velocity is maximum. Thus, upstream of the vortex one would
computations, the strength of the convecting free stream vor- notice a favourable pressure gradient, while in the downstream
tex is taken as = 2.0 (as compared to 9 taken in the work of direction, the induced velocity reduces with x, i.e., an adverse
Sengupta et al.49 ). This is important for the experimental and pressure gradient is created. The free stream vortex convects
computational study of receptivity, where we need as small at a constant height and its instantaneous streamwise location
an excitation as possible. The Reynolds number based on dis- is shown by an arrowhead. For farther downstream location
placement thickness at the outflow of the undisturbed flow is of the vortex, the boundary layer ahead of it experiences sus-
2432.44, and thus the domain is more than five times than that tained adverse pressure gradient, and in frame (b) of Fig. 3,
was taken in the work of Sengupta et al.49 The flow computed unsteady separation bubbles are noted near x = 4. Downstream
in the latter reference remained sub-critical over the com- of these bubbles, the flow over the flat plate experiences addi-
putational domain, while the present domain is significantly tional adverse pressure gradient due to the bubbles. This is
longer. readily seen in frames (c) and (d) at t = 28 and 30, respectively,
The equilibrium solution is obtained first by solving the with increase in the number of unsteady separation bubbles,
NSE, without the convecting free stream vortex. Once this all of which convect downstream. The fact that a single con-
is obtained, the free stream vortex is started with constant vecting free stream vortex, far out in the inviscid part of the
convection speed (c = 0.3) from the initial location: (x = 5, flow, can cause a bypass transition is noted in frame (e) of
y = 2). The flow is started impulsively with the initial location Fig. 3. Such unsteady separation bubbles during bypass tran-
of the vortex being ahead of the leading edge. In comput- sition were conjectured to be caused as a result of buffeting
!
ing the flow by ( V , !)-formulation, computations have been of the boundary layer by FST vortices.30 The present unit-
performed with a time step of t = 8 ⇥ 10 5 . process provides the physical mechanism involved in bypass
In Fig. 3, identical stream function contours are plotted at transition. Here, no model is required, with the effects gov-
indicated times to show physically important events. For the erned solely by the NSE. For example, in the work of Obabko
054106-7 Sengupta et al. Phys. Fluids 30, 054106 (2018)

FIG. 3. Stream function contours plotted in the computational domain at the FIG. 4. Vorticity contours plotted in the computational domain at the indi-
indicated times. Arrowheads at the top show the streamwise location of the cated times, as in Fig. 3. Same contour values are plotted in all the frames.
convecting vortex. Arrowheads at the top show the streamwise location of the convecting vortex.

IV. THE INSTABILITY MECHANISM


and Cassel,34 the primary vortex is taken as a Batchelor vortex
VIA GROWTH OF DME
moving at a constant speed along the wall, as the model to
study unsteady flow evolution. The experimental and accompanying computational
The unsteady separations and vortical structures on the results display the existence of a receptivity mechanism inside
wall are consequences of free stream convecting vortex, as the shear layer as a consequence of a single vortex migrating
shown in Fig. 4, at the same time instants for the stream func- in the free stream at a constant speed.26,49 The role of various
tion contours in Fig. 3. It is already noted that zero pressure gra- parameters responsible for this instability is characterized by
dient boundary layer does not exhibit receptivity to free stream the redistribution of E m to E d in the flow. The equation for E
acoustic excitation.43 Here, we have shown strong receptivity for incompressible flows has been derived in Eq. (6), which
of the zero pressure gradient boundary layer to free stream shows the distribution at any time instant to depend on the
inviscid vortical excitation, as also reported before.25,26,49 An enstrophy and interaction between the velocity field with the
explanation of this has been provided44,48 with the help of palinstrophy (which is half the curl of vorticity). The contribu-
a spatial linear instability theory described by the OSE. It tion of total enstrophy on E, as noted on the right-hand side, is
is shown that boundary layers support disturbances created positive, i.e., stabilizing. However, as one views the evolution
inside or outside a shear layer by the wall and free stream of E d given in Eq. (7), one notices the corresponding term to
modes of the OSE, respectively. When free stream-modes are be given by ⌦d . As ⌦d can be either positive or negative, the
excited (as we have here), those in turn cause the wall-mode to corresponding E d can display growth or decay of DME. The
be excited by a coupling mechanism that ensures homogeneous growth is indicated when the equilibrium and disturbance vor-
boundary conditions at the wall. ticities are of opposite sign, triggering a transfer of energy from
The genesis and growth of the primary bubble and appear- equilibrium to disturbance flow. At the same time, the second
ance of subsequent separation bubbles are due to the vortex- term of Eq. (7) indicates that the palinstrophy can interact with
induced instability, the mechanics of which is discussed next. the velocity field in causing instability, when this is a negative
First, we explain it with the help of DME, followed by the quantity. Here, Eq. (7) has been used to explain the complete
proposed disturbance enstrophy transport equation. nonlinear evolution of DME.
054106-8 Sengupta et al. Phys. Fluids 30, 054106 (2018)

In Fig. 5, the distributed sources of E d are plotted as originating from the leading edge terminates before the down-
negative contours. At t = 20, there are two sites from where stream spike. The present analysis based on the right-hand side
instability originates—one starting from the leading edge and of Eq. (7) clearly reveals the physical mechanism of the insta-
the other starting downstream of leading edge on the plate. It bility, as compared to the information from stream function
is seen that the leading-edge instability continues to remain and vorticity contours.
outside the boundary layer due to the shear sheltering effect,20
and despite its larger extent, it does not contribute to insta-
bility inside the shear layer. The major instability inside the V. ENSTROPHY TRANSPORT EQUATION: A NEW
shear layer originates near x = 2.4 (for t = 20), as is also seen APPROACH TO NONLINEAR RECEPTIVITY THEORY
in the vorticity contours in Fig. 4. Disturbance energy struc- We have introduced the nonlinear receptivity/instability
tures from these two sites remain distinct till t = 25. In the theory based on DME, derived by taking the divergence of
figure, for the frame at t = 28, it is seen that these two sources the NSE, which is the conservation equation for translational
of E d interact, as is also evident from the bottom three frames momentum. In fluid mechanics, one does not consider the con-
in Fig. 5, where the spike forming at the downstream site at the servation equation for rotational momentum. This is partially
wall interacts with the vortical structure originating from the addressed by the vorticity transport equation (VTE) obtained
leading edge that remains sheltered from the shear layer. After by taking a curl of the NSE, and information about rotation-
this time and downstream of this location, these two sources ality in a flow is provided by the measure of the enstrophy.
merge together. Interestingly, as we have remarked earlier, the turbulence lit-
In stream function and vorticity contour plots, the spike erature alludes enstrophy to provide a measure of dissipa-
is evident near x = 4.5 at t = 28 in the form of a secondary tion.10,13,22,67 Doering and Gibbon10 studied the enstrophy
bubble. Thus, it is important to include the leading edge in transport equation (ETE) for 2D periodic flows and obtained
the computations, otherwise one would compute the unim- the evolution of integrated enstrophy of the full domain in
peded spike stage.34,57 However beyond t = 28, the instability Eq. (1). This shows that viscous diffusion destructs enstrophy
in the full domain, for 2D periodic flows. However for the
most generic cases of 3D inhomogeneous flows, a more gen-
eral ETE has been developed to discuss the role of diffusive
terms in fluid flows.38,55 The true role of diffusion remained
problematic, as it was considered to be strictly stabilizing
flows by damping disturbances, an idea behind the proposal
for developing Rayleigh’s stability equation.44 Unfortunately,
such inviscid studies could not even explain instability of zero
pressure gradient boundary layer. But, the same flow was sub-
sequently investigated by solving the OSE.18,40,62 However,
the governing equations for E and E d arise entirely from local
and convective acceleration terms (in the absence of body
force).
However, if diffusion is viewed at any instant for any
point in the flow, then the effects of diffusion are not strictly
dissipative, as has been explained.38,55 When one looks at
time-averaged TKE globally, effects of diffusion are again
seen as dissipative.28,61 As shown in Eq. (4.34) in Mathieu
and Scott,28 time-average of the diffusion term in the NSE
manifests as a combination of (i) a strictly dissipation term
and (ii) another viscous transfer term. The viscous trans-
fer term does not contribute, when integrated over the flow
domain, due to divergence theorem. This diffusive term is
though dissipative for homogeneous turbulence. It is fur-
ther noted that the viscous transfer term is negligible at
high Reynolds numbers, except within the thin viscous lay-
ers very near solid surfaces. While on the other hand, the
dissipative term is of crucial importance to turbulence ener-
getics everywhere. Similar observations are made in Sec. 3.3
in Tennekes and Lumley,61 with respect to time-averaged
TKE.
To understand the role of diffusion in creating rotationality
for inhomogeneous flows, an evolution equation is developed
for enstrophy and its higher powers,55 as a point property.
FIG. 5. Contours of the right-hand side of the equation for DME, Eq. (7).
The negative values indicating disturbance energy sources are plotted as dark This explains the roles of diffusion, dissipation, and creation
blue and green contours. of rotationality progressively to smaller scales.
054106-9 Sengupta et al. Phys. Fluids 30, 054106 (2018)

A. Enstrophy transport equation equation:


" #
The role of ⌦1 in fluid flows is similar to kinetic energy D⌦2 1 1 2
= 2Re r ⌦2 (r⌦1 )2 ⌦1 (r!)2 . (16)
describing the translational motion in fluid flows. While vortic- Dt 2
ity describes rotationality, enstrophy unambiguously describes
Noting further that D⌦ Dt
2
= 2⌦1 D⌦Dt , one can rewrite
1
the energy expended by the system in creating and sustaining
Eq. (16) as the ETE, i.e., an evolution equation for ⌦1 . Mul-
rotationality. In all flows, physical instabilities take an equilib-
tiplying the above equation with ⌦2 and simplifying, one can
rium state to another one and in the process, the energy of the
obtain the transport equation for ⌦3 , which can also be used
system is redistributed into rotational and translation degrees
to write the ETE involving ⌦1 , ⌦2 , and ⌦3 . This process can
of freedom. We explain instabilities and pattern formations,
be generalized to obtain the transport equation for ⌦n as
with the help of enstrophy transport equation (ETE) derived " #
from the non-dimensional VTE in tensor notation given for D⌦n 1
= 2Re 1 r2 ⌦n (r⌦n 1 )2 C , (17)
3D flows by Dt 2
@!i @!i @ui 1 @ 2 !i where
k 1*
⌦j +/(r⌦k )2
+ uj = !j + , (13) n 2
X n 1
Y
@t @xj @xj Re @xj @xj C= 2n. (18)
where subscripts, i, j = 1, 2 and 3, represent Cartesian axes P
k=0 , j=k+1 - Q
and repeated index implies summation. Taking a dot product and ⌦0 = ! with indicating summation over all k’s and
of Eq. (13) with !i we obtain ETE as indicating the product of all the jth elements.
! ! Also, the substantive derivative of ⌦n can be written and
@⌦1 @⌦1 @ui 1 @ 2 ⌦1 2 @!i @!i simplified as
+ uj 2!i !j = . ✓ D⌦ ◆
@t @xj @xj Re @xj @xj Re @xj @xj D⌦n n 1
= 2⌦n 1
(14) Dt Dt
D⌦n 2
The third term on the left-hand side (LHS) is due to vor- = 2⌦n 1 (2⌦n 2 ) (19)
Dt
tex stretching [corresponding to the first term on the right-hand ✓Yn 1 ◆ D⌦
1
side (RHS) of Eq. (13)], which is absent for 2D flows. The dif- = 2n 1 ⌦j ,
j=1 Dt
fusion of !i gives rise to RHS terms in Eq. (14). The first term
of RHS shows diffusion of ⌦1 and the second term represents which can be further simplified to give
strictly a loss or dissipation for the transport of ⌦1 . The present D⌦n D⌦1
= 2 n 1 ⌦1 , (20)
study views ⌦1 as a point property in the flow and is different Dt Dt
from the usage in the work of Doering and Gibbon,10 where where = 1 2n 1 . Using the above relations, one can rewrite
the enstrophy is defined over the full domain for the simpli- Eq. (17) as the ETE given by
fication of homogeneous and periodic problems. Focusing on
1 " #
2D flows with the spanwise component of vorticity (!), ETE D⌦1 Re ⌦1 1 2 2
= r ⌦ n (r⌦ n 1 ) C . (21)
can be written in vector form as Dt 2n 2 2
" #
D⌦1 2 1 2 2
One notes that while writing the transport equation for
= r ⌦1 (r!) . (15) ⌦n , the diffusion term from the transport equation for ⌦n 1
Dt Re 2
contributes two terms: one of which is strictly dissipative
The first term on RHS of Eq. (15) is missing from Eq. (1). (dependent on ⌦n 2 ) and the other as a diffusion term for
By contrast, for inhomogeneous flows, this term of Eq. (15) ⌦n 1 . The diffusion term involving ⌦n 1 can be furthermore
can be either positive or negative. Therefore, the diffusion term expressed by two terms involving a strictly dissipative term
in ETE can create or destroy rotationality depending upon with ⌦n 1 and another diffusive term involving ⌦n 2 . This pro-
the sign of RHS in Eq. (15). As ⌦1 > 0, positive RHS indi- cess can cascade indefinitely in Eq. (21), for increasing n with
cates the diffusion to cause instability. Negative RHS acts as the leading term as a diffusion term and the rest as strictly dis-
a sink of ⌦1 . This provides a mechanism of creating rota- sipative. Higher order moments of ⌦1 will contribute more for
tionality at different scales by diffusion and is distinctly dif- higher wavenumbers, implying that the order of even moments
ferent form the concept of creating smaller scales by vortex of ⌦1 will be restricted by the energy supplied to the flow. For
stretching, as the dominant mechanism of generating small 3D flows as well, the RHS of Eq. (21) is present as the forc-
eddies for 3D flows. We note that the role of diffusion in ing term. However in this case, the vortex stretching term is
creating new length scales is ubiquitous for both 2D and 3D retained. The ETE for 3D flow is the same, as given by Eq. (14).
flows. Following a similar approach, as in deriving the transport equa-
tion for ⌦n for 2D flows, the transport equation for ⌦n can be
B. Enstrophy cascade for general inhomogeneous derived for 3D flows as
flows " #
D⌦n 1 1 2 2
= 2Re r ⌦n (r⌦n 1 ) C
To further investigate effects of diffusion in Eq. (15) at Dt 2
multiple scales, one derives transport equations for higher n 1
Y !
powers of ⌦1 . Multiplying Eq. (15) with ⌦1 and defining n @ui
+2 ⌦k ! i ! j , (22)
n 1
⌦n = ⌦21 , one obtains for ⌦2 the following transport k=1
@xj
054106-10 Sengupta et al. Phys. Fluids 30, 054106 (2018)

where the expression for C is the same as in Eq. (18). Using


Eq. (20), one can rewrite the ETE for 3D flows as
1 " #
D⌦1 Re ⌦1 r2 ⌦n @ui
= (r⌦n 1 )2 C + 2!i !j .
Dt 2n 2 2 @xj
(23)

One also notes that the diffusion term gives rise to enstro-
phy cascade for both 2D and 3D flows, for which the contribu-
tion at higher wavenumbers depends upon the value of n, which
should be decided by the energy supplied to the fluid dynam-
ical system. However in 3D flows, vortex stretching is also
present, which provides an additional mechanism of energy
redistribution process. This indicates that in 3D flows, genera-
tion of different scales of vorticity is due to enstrophy cascade
via the stretching and diffusion terms and energy cascade is
due to the vortex stretching implicit in convection process.
In 2D flows, it is only the diffusion term which gives rise to
enstrophy (and hence vorticity) at different scales.
It is emphasized that physically the role of diffusion for
inhomogeneous flows is not strictly dissipative, as is the case
for homogeneous turbulent flows. By developing the ETE,
in terms of higher even moments of ⌦1 , we identified the
index n in this equation, which is fixed from total energy
imparted to create a flow. This view of how the smallest scale is
fixed is entirely different from the conventional logic used for
the dissipation of kinetic energy to heat, even for isothermal
flows.
In Fig. 6, we note the growth rate of total enstrophy
(D⌦1 /Dt)-contours for the problem of vortex-induced instabil-
ity, whose results are shown in Figs. 3 and 4 for the streamfunc- FIG. 6. The growth rate of total enstrophy as given in Eq. (15) for the vortex-
induced instability problem defined in Fig. 1 (for which the physical variables
tion and vorticity. The corresponding E d evolution is shown are shown in Figs. 3 and 4) is shown for the identical time instants. The marked
by contour plots of RHS of Eq. (7) in Fig. 5. The times at regions are for the instability of enstrophy, i.e., where rotationality increases
which the total enstrophy growth shown in Fig. 6 are identi- with time at the indicated time instants.
cal to the times shown earlier in Figs. 3–5. The spatial scales
shown in enstrophy growth rates are much more refined than VI. ENSTROPHY BASED THEORY OF INSTABILITY:
those seen for E d . Another aspect that draws our attention CREATION OF ROTATIONALITY
is the clear presence of wall- and free stream-modes at early
times.44,48 At t = 20, one notices growth of rotationality orig- If ⌦1 is written as a sum of the equilibrium and disturbance
inating from the leading edge of the plate (due to Goldstein values as ⌦1 = ⌦m + ✏ 1 ⌦d , then the linearized growth rate of
singularity) that remains at the edge of the boundary layer very ⌦d can be evaluated retaining the terms of order ✏ 1 . However,
prominently. The wall-mode is in nascent stage in the region if the primary variables are represented as ! ! =! ! m + ✏ 2!
!d
! ! !
1  x  2 of the flat plate. Both these modes are noted more and V = V m + ✏ 2 V d , then the growth or decay rate of ⌦1 can
prominently in the frame at t = 25 and the growing wall-mode be written as
causes a bulge in the growth region of the free stream mode. D⌦1 @ ! !
In a short time interval, one notices severe interaction between = (⌦m + ✏ 1 ⌦d ) + ( V m + ✏ 2 V d ) · r(⌦m + ✏ 1 ⌦d ) = RHS.
Dt @t
these two modes, as noted in the frame for t = 28, which is
That is, from the above, one can rewrite it as
noted more in the vorticity contours of Fig. 4 than in the cor-
responding figure for DME in Fig. 5. Thus, it is apparent that D⌦m @⌦d ! !
+ ✏1 + ✏ 1 V m · r⌦d + ✏ 2 V d · r⌦m
the signature of vortex-induced instability is more evident in Dt @t
the rotational part of the NSE, i.e., in disturbance ETE, than !
+ ✏ 1 ✏ 2 V d · r⌦d = RHS.
in the equation for E d . Also visible is the presence of much
finer wall-normal structures in Fig. 6, during early interactions The order ✏ 1 terms of the left-hand side of the above are
between wall and free stream modes. Such events are equally simply nothing but the substantive derivative of ⌦d and thus,
"⇢
visible in the frame at t = 30. In the bottom frames of Figs. 5 D⌦d @uid @uim @uim
and 6, one can notice more energetic events in E d , as compared = !im !jm + !im !jd + !id !jm
Dt @xj @xj @xj
to those in (D⌦1 /Dt)-contours. The latter still grows in the 2 ✓ ◆ ✓ ◆ #
location over the plate which is exactly beneath the convecting 1 @ ⌦d 2 @!im @!id
+ . (24)
vortex. Re @xj @xj Re @xj @xj
054106-11 Sengupta et al. Phys. Fluids 30, 054106 (2018)

Thus the growth/decay rate of ⌦d is determined by the In Figs. 7(a) and 7(b), D⌦d /Dt is plotted in the domain for
vortex stretching terms given on the RHS of the above equa- the case of linear instability as indicated above, i.e., positive ⌦d
tion inside the curly brackets. These will not be present for 2D will increase where D⌦d /Dt is positive, as shown in Fig. 7(a).
disturbance fields. The term that is identified as a strictly dissi- In Fig. 7(b), the case of instability is indicated for ⌦d < 0 shown
pation term in Eq. (14) for ⌦1 can now contribute to growth of in the flooded region marked in the various frames for which
⌦d , as shown here by the last term on RHS. We also note that, D⌦d /Dt < 0. Growth rate contours at the indicated times show
unlike ⌦1 (which is strictly positive definite), ⌦d can be either variation in instability that changes by orders of magnitude.
positive or negative, as this is given as (2! !m ·! ! d ). It is noted Only the growth rate regions are marked by flooded zone for
that this ⌦d term also appears in Eq. (7) for E d . However, both the cases in the frames which increases with time. The
positive ⌦d does not make the RHS of Eq. (7) necessarily region of instabilities is qualitatively different for the positive
negative. Thus, positive ⌦d may not indicate instability via and negative values of ⌦d . For the positive values, disturbances
DME.49,58 originate outside the shear layer for t = 20. However by t = 28,
By contrast, instability for ⌦d can arise for ⌦d being either strong growth rates are also seen on the wall near x = 5. These
positive or negative. The growth rate for ⌦d will be different disturbances create vertical eruptions and disrupt the outer vor-
for different signs of the quantity, i.e., for positive ⌦d , an insta- tical structures. This sequence of events are also noted for t = 30
bility is indicated when D⌦d /Dt > 0 and for negative ⌦d , its and 50. By the later time, large growth rates are seen to be
amplitude will grow, when D⌦d /Dt < 0. These conditions are confined inside the shear layer, while the outer disturbances
investigated next with results of the vortex-induced instability for larger x stations have smaller growth rates. In Fig. 7(b)
caused by a convecting vortex in the free stream. at t = 20, one notices larger growth rates near the wall,

FIG. 7. (a) The linear growth rate of ⌦d as given in Eq. (15) for the vortex-induced instability problem. On the frames shown are the cases with positive ⌦d at
the indicated time instants. (b) The linear growth rate of ⌦d as given in Eq. (15) for the vortex-induced instability problem. On the frames shown are the cases
with negative ⌦d at the indicated time instants.
054106-12 Sengupta et al. Phys. Fluids 30, 054106 (2018)

FIG. 8. (a) The nonlinear growth rate of ⌦d for the vortex-induced instability problem. On the frames shown are the cases with positive ⌦d at the indicated time
instants. (b) The nonlinear growth rate of ⌦d for the vortex-induced instability problem. On the frames shown are the cases with negative ⌦d at the indicated
time instants.

i.e., a predominance of wall mode as compared to the free notices the wall-normal streaky structures piercing through the
stream mode noted for ⌦d > 0. By t = 28, the wall modes are shear layer. For negative ⌦d , results shown in Fig. 8(b) show
shown to eject vortical structures in the wall-normal direction, instability all along the shear layer at t = 20, along with a small
which is also seen at t = 30. At a much later time (t = 50), free stream mode attached to the leading edge. By t = 28, this
one notices the presence of free stream mode near the lead- free stream mode grows, while one notices wall-normal vorti-
ing edge of the plate showing a coherent structure. However, cal streaks near x = 5 which is localized. With passage of time,
in the downstream stations for x > 10, one notices stronger at t = 30 and 50, one notices further growth of the coherent
growth near the wall and an upper deck with lower growth free stream mode originating from the leading edge, while the
rate. streaky structures are seen to be present downstream of the
In Figs. 8(a) and 8(b), we show the nonlinear growth rate free stream mode, originating from the wall.
for ⌦d , once again showing the positive and negative values Comparing the linear and nonlinear growth rates of dis-
of ⌦d , respectively. As noted for the linear growth rates in turbance enstrophy in Figs. 7 and 8, one notices a moderation
Figs. 7(a) and 7(b), we also see the same trends for the non- of the growth rates for the nonlinear cases, as compared to the
linear growth rates. For positive ⌦d , at t = 20, once again one corresponding linear cases. Here, the vortex-induced instabil-
notices the dominant free stream mode with higher growth ity caused by the convecting vortex for c = 0.3 is used to explain
rates near the leading edge of the plate. However, at t = 28 and the developed theory of instability based on enstrophy.
30, one notices the appearances of stronger instabilities at the
wall which causes eruption in the wall-normal direction dis-
VII. SUMMARY AND CONCLUSIONS
rupting the coherent free stream vortical structure. By t = 50,
the free stream mode is not noted and the disturbance field con- A theory of instability based on disturbance enstrophy of
sists of strictly wall modes up to x = 10, and beyond that one incompressible fluid flows is developed here. The developed
054106-13 Sengupta et al. Phys. Fluids 30, 054106 (2018)

theory is tested with the help of a numerical simulation of a we would focus upon the receptivity of the zero pressure gra-
bypass transition caused by convecting vortex in free stream, dient boundary layer for other parameters, such as circulation
over a zero pressure gradient boundary layer. This is also the ( ), convection speed (c), and the height (H) of the free stream
topic of vortex-induced instability investigated experimentally convecting vortex. This will also be attempted along with the
in the work of Lim et al.26 and computationally in the work of corresponding cases for 3D simulations. The preliminary 3D
Sengupta et al.49 These studies were in turn motivated by the results in Appendix B are included to show the qualitative
observations in the studies of Monin and Yaglom30 and Tay- similarity with the 2D case studied at early times. In the 3D
lor 60 on the effect of free-stream turbulence. The motivation simulations, one would like to note the evolution of all the
for the present work is to show how unstable vortical structures three components of vorticity and the physical mechanism(s)
are created during this bypass transition. of growth.
Although total enstrophy is a positive quantity and is asso- Finally, we note that the presented theories are for incom-
ciated with viscous dissipation for incompressible fluid flows, pressible flows, for which the NSE is an accepted govern-
here we develop an evolution equation for the disturbance ing equation for the fluid flow. Corresponding development
component of enstrophy. This disturbance component can be for compressible flows is beyond the scope of the present
either positive or negative and is shown to drive the distur- investigation, although the disturbance energy equation has
bance mechanical energy (DME) distribution. The latter has been described for compressible flow6 involving enthalpy as
been developed49 from the NSE without any approximations the measure. We also note that there are efforts,15 where
to explain bypass transition observed experimentally.26 The researchers have tried to look beyond NSE in describing Bur-
present disturbance enstrophy (⌦d )-based theory is comple- nett and super-Burnett equations. Such studies require search
mentary to the DME-based theory in explaining creation of for instability mechanisms related to Burnett hydrodynamics.
rotationality for general inhomogeneous flows. The detailed
derivation given here leads to Eq. (24), derived from the enstro- ACKNOWLEDGMENTS
phy transport equation,55 which has its origin in the VTE. It is
seen that the DME equation originates from the inviscid part The authors acknowledge the Department of Science and
of the NSE for the incompressible flow, while the equation for Technology, New Delhi (India) for providing the necessary
⌦d contains both the inviscid and the viscous operators of the HPC facility at IIT Kanpur, India used for the simulations
NSE. Thus, one would see finer structures for the disturbance reported. The first author is supported by the Cambridge India
quantities obtained by Eq. (24), as compared to those obtained Ramanujan Scholarship of DST and Cambridge Trust for her
from Eq. (7). We reason that these two nonlinear theories have graduate studies at University of Cambridge, UK.
a similar relationship, as that is present between Rayleigh’s
and Orr-Sommerfeld equations for linearized disturbance APPENDIX A: NUMERICAL METHODS AND GRID
growth. USED FOR VORTICITY-VELOCITY FORMULATION
From Eq. (7), we note that the DME is driven by ⌦d , at any 1. Numerical methods and grids
instant of time. For this reason, one can reproduce linear spatial
theory results based on the OSE by application of Eq. (7).52 A staggered variable arrangement17,33 is adopted here for
By contrast, the governing equation for ⌦d given by Eq. (24) is the 2D simulation. This arrangement reduces errors that arise
an evolution equation providing temporal growth of rotation- due to aliasing and are smaller in a staggered grid, as compared
ality in the flow. It is apparent that enstrophy-based approach to a non-staggered grid using this 2D formulation.19 In the
can provide temporal disturbance growth rate at any point. transformed (⇠, ⌘)-plane, the staggered grid is shown in Fig. 9,
This, however, does not imply that DME-based approach is with contravariant components of velocity marked at the mid-
only suitable for spatial instability and the ⌦d -based approach points of control surface over which the component is normal.
is for temporal growth only. Both the theories based on DME In Fig. 9, the vorticity is placed at the nodes.
and ⌦d are ideally suited for the most general spatio-temporal Time advancement of ! is done by the RK 4 -scheme,
instability. The dominance of the ⌦d equation over the equa- while convective derivatives ( @! @!
@⇠ , @⌘ ) are discretized using
tion for DME is apparent with DNS data showing all finer the OUCS3 compact scheme,50 and second derivatives are
spatial and temporal scales. discretized using the second-order central difference scheme.
The instability theory developed here helps explain lin- To evaluate convective derivatives in the staggered grid, it is
ear/nonlinear growth for high Reynolds number incompress- required to interpolate u and 3 velocities at the locations of
ible flows. Apart from explaining creation of rotationality for the vorticity, which is carried out by an optimized compact
general inhomogeneous flows and the cascade of eddies by mid-point interpolation scheme.32 After solving Eq. (8) for
enstrophy transport equation, we have also described the linear the vorticity, we solve the Poisson equation for the u and 3
and nonlinear growth of disturbances during vortex-induced components of velocity.
instability. The Poisson equation is solved by the Bi-CGSTAB
When one compares the linear and nonlinear growth rates method,63 with a prescribed tolerance limit of ✏ = 10 8 for
of ⌦d in Figs. 7 and 8, one notices qualitative similarity convergence. Same numerical methods are used to obtain
between these two cases. A moderation of growth rates is noted the equilibrium and the disturbance flow, as excited by the
for the nonlinear cases, as compared to the corresponding lin- free stream vortex. Once the steady equilibrium flow is
ear cases. While the vortex-induced instability caused by the established, the free stream vortex is convected to excite the
convecting vortex is shown for c = 0.3 only, in future studies, flow.
054106-14 Sengupta et al. Phys. Fluids 30, 054106 (2018)

!
FIG. 9. Staggered grid system used in ( V , !)-
formulation for 2D problems. The velocity components
are at the mid-point of the elemental surface over which
it is normal.

Brinckman and Walker 5 and Obabko and Cassel34 have spectral plane to achieve better dispersion relation preserva-
used methods which are O( x t, y t) accurate. In the tion (DRP) properties.1 Using Fourier-Laplace transform, one
computations by Sengupta et al.,49 the authors have used can express f̂j = s F̂(k) eikxj dk and fj = s F(k) eikxj dk. Then
OUCS3 compact schemes for spatial discretization of con- the resolution properties and the performance of an interpola-
vection terms, which has more than seven times higher spec- tion scheme can be characterized by a transfer function given
tral resolution, as compared to second-order accurate explicit by TF = F̂(k)/F(k), which is the ratio of the Fourier ampli-
schemes. Also, the dispersion error of the method49 is low tude of the interpolated function to that of the original function.
for the space-time discretization schemes. The formulation From Eq. (A1), this transfer function is obtained as
and used staggered compact scheme here further improves " #
aI cos(kh/2) + bI cos(3kh/2)
accuracy by reducing aliasing and dispersion errors.45,53 It has TF(kh) = (A2)
been firmly established that in computing space-time depen- 1 + 2↵1 cos(kh)
dent problems, dispersion error is the largest source of error, as with uniform grid spacing h. Then, to minimize the error in
compared to spatial and temporal discretization errors viewed the spectral plane, one can define an objective function given
in isolation. Thus, the role of error dynamics for scientific as ⌅
computing cannot be overemphasized.45,53,59
Iinterp = 1 TF(kh) d(kh) (A3)
2. Interpolation scheme 0
which is nothing but the phase error integrated over the range
We have noted that in solving the NSE using velocity- of kh from zero to (which is less than the Nyquist limit).
vorticity formulation, the velocity components are shifted by The TF given above is shown plotted as a function of
half-grid spacing from the nodes. Thus in solving the vorticity kh in Fig. 10(a) for the indicated values of ↵ I . The objective
transport equation and the Poisson equation, the variables have function I interp is shown plotted in Fig. 10(b), as a function of
to be interpolated by a scheme given as ↵ I for the indicated values of . It is noted clearly that this
aI bI objective function is minimum at ↵ I = 0.42 for = ⇡. This
↵I f̂j 1 + f̂j + ↵I f̂j+1 = (f 1 + fj+ 1 ) + (f 3 + fj+ 3 ), (A1)
2 j 2 2 2 j 2 2 optimized fourth order scheme has been used here. However,
where fˆj s are the interpolated values at the jth location obtained for non-periodic problems, the interpolation scheme given by
from the known f j±n/2 values at the (j ± n/2)th locations. For Eq. (A1) has to be supplemented by the following boundary
4th order accuracy, one should have aI = 18 (9 + 10↵I ) and stencil given as
1
bI = 18 (6↵I 1), with ↵ I as a free parameter.32 Further- f̂1 = (f 1 + f 3 ), (A4)
3 2 2 2
more, a choice of ↵I = 10 yields a 6th order accuracy for
the interpolation scheme. However, here ↵ I = 0.42 is obtained 1
f̂N = (f 1 + fN+ 1 ). (A5)
by optimizing the integrated phase error of the scheme in the 2 N 2 2
054106-15 Sengupta et al. Phys. Fluids 30, 054106 (2018)

Similarly, due to the grid clustering at the wall, the trans-


formation function in the wall-normal direction is given as
" #
tanh[ y (1 ⌘)]
y(⌘) = ymax 1 , (A7)
tanh y
where 0  ⌘  1. Here, x and y are parameters that control
the grid clustering in the streamwise and wall-normal direc-
tions, respectively. For the cases reported in the present work,
x = y = 2 is used. This transformation of the problem from
the physical plane to computational plane requires the scale
1 1
factors h1 and h2 as h1 = (x⇠2 + y2⇠ ) 2 and h2 = (x⌘2 + y⌘2 ) 2 ,
@x @y
respectively. Since here, x⌘ = @⌘ = 0 and y⇠ = @⇠ = 0, the
scale factors are simply given as h1 = x ⇠ and h2 = y⌘ . The
above grid transformation in the streamwise direction given
by Eq. (A6) ensures the continuity of both h1 and dh
d⇠ at x = x s .
1

The grid used has 1001 points in the streamwise direction and
301 points in the wall-normal direction.

APPENDIX B: SOLUTION OF 3D VORTEX-INDUCED


INSTABILITY PROBLEM
1. Governing equations and numerical methods
for 3D receptivity problem
Schematic diagram used for 3D receptivity computations
for a zero pressure gradient boundary layer to 2D free stream
convecting vortex excitation is shown in Fig. 11(a). The lead-
ing edge of the plate is retained inside the computational

FIG. 10. (a) Transfer function of the staggered interpolation scheme given
by Eq. (A2) plotted as a function of kh for indicated values of ↵I . (b) Inte-
grated phase error plotted as a function of ↵I for the stencil with indicated
value of .

3. Grid generation
In order to resolve the points near the leading edge, a
non-uniform stretched grid is used in the streamwise direction
that clusters points at the leading edge. The tangent-hyperbolic
function is used for grid clustering, which has been shown to
cause lower aliasing error.45 Similar clustering of grid points
has been performed in the wall-normal direction, as well, to
accurately resolve the boundary layer and events very close to
the wall for the receptivity problem, as one expects high gradi-
ents of the flow variables near the wall. The physical problem
defined in the physical plane is transformed to a uniform com-
putational (⇠, ⌘)-coordinate system. The specific form of grid
transformation function in the streamwise direction is given
as, for 0  ⇠  ⇠ 1 and x in  x  x s ,
" #
tanh[ x (1 ⇠)]
x(⇠) = xin + (xs xin ) 1
tanh x
and, for ⇠ 1  ⇠  1 [and x s  x(⇠)  x out ],
"✓ ◆ ✓ ⇠ ⇠ ◆#
x 1
x(⇠) = xs + (xs xin ) , (A6)
tanh x ⇠1
1
where ⇠1 = 1+A1 and
! ! FIG. 11. (a) Schematic diagram of the 3D receptivity problem for a 2D
xout xs tanh x
A1 = . ZPG boundary layer, (b) staggered grid arrangement of velocity and vorticity
xs xin x components.
054106-16 Sengupta et al. Phys. Fluids 30, 054106 (2018)

domain. The origin of the reference co-ordinate system is At the spanwise boundaries: Periodic conditions on all
located at the mid-point of the leading edge of the plate. The the six variables (three components of velocity and three
computational domain is given as x in  x  x out along the components of vorticity) are used.
streamwise direction with x in < 0, 0  y  ymax along the wall- At the wall: No-slip boundary conditions on u, 3, and 4
normal direction, and zmax /2  z  zmax /2 along the spanwise components are used. The boundary conditions on the vorticity
direction. components on the plate surface are given as
The simulations are performed in the transformed (⇠, ⌘,
⇣)-plane such that x = x(⇠), y = y(⌘) and z = z(⇣). The rotational ⌦⌘ = 0, (B10)
!! ! !
variant of the ( V , ⌦)-formulation of the incompressible NSE @w @vw @vw @u
is used here3 and given by ⌦⇠ = and ⌦⇣ = . (B11)
@y @z @x @y
!
@⌦⇠ 1 @H⇣ 1 @H⌘
+ = 0, (B1) The sharp leading edge of the flat plate is assumed as
@t h2 @⌘ h3 @⇣
! the locus of stagnation points for this flow and hence, at the
@⌦⌘ 1 @H⇠ 1 @H⇣ bottom plane ahead of the leading edge, as shown in Fig. 11(a),
+ = 0, (B2)
@t h3 @⇣ h1 @⇠ symmetry conditions are used on all the six variables given by
!
@⌦⇣ 1 @H⌘ 1 @H⇠ @u @w @⌦⌘
+ = 0, (B3) =v= = ⌦⇠ = = ⌦⇣ = 0. (B12)
@t h1 @⇠ h2 @⌘ @⌘ @⌘ @⌘
! !
where ⌦ = (⌦⇠ , ⌦⌘ , ⌦⇣ ) and V = (u, v, w). The scale factors At the outflow: Convective Sommerfeld boundary condi-
@x @y @z
h1 , h2 , and h3 are given as h1 = @⇠ , h2 = @⌘ , and h3 = @⇣ . In tions are applied on the variables u, ⌦⌘ , and ⌦⇣ as
Eqs. (B1)–(B3), the terms H ⇠ , H ⌘ , and H ⇣ are given as
! @u @u
1 1 @⌦⇣ 1 @⌦⌘ + Uc = 0, (B13)
H⇠ = (w⌦⌘ v⌦⇣ ) + , (B4) @t @x
ReL h2 @⌘ h3 @⇣
!
1 1 @⌦⇠ 1 @⌦⇣
H⌘ = (u⌦⇣ w⌦⇠ ) + , (B5)
ReL h3 @⇣ h1 @⇠
!
1 1 @⌦⌘ 1 @⌦⇠
H⇣ = (v⌦⇠ u⌦⌘ ) + . (B6)
ReL h1 @⇠ h2 @⌘
!!
For the ( V , ⌦)-formulation, the corresponding velocity vec-
tors are obtained from the velocity Poisson equations
! !
r2 V = r ⇥ ⌦, and in the transformed plane these are given
as !
@⌦⌘ @⌦⇣
r2⇠⌘⇣ u = h1 h2 h3 h1 , (B7)
@⇣ @⌘
!
2 @⌦⇣ @⌦⇠
r⇠⌘⇣ v = h2 h3 h1 h2 , (B8)
@⇠ @⇣
!
@⌦⇠ @⌦⌘
r2⇠⌘⇣ w = h3 h1 h2 h3 , (B9)
@⌘ @⇠
where the operator r2⇠⌘⇣ is given as
! !
@ h2 h3 @ @ h3 h1 @
h1 h2 h3 r2⇠⌘⇣ = +
@⇠ h1 @⇠ @⌘ h2 @⌘
!
@ h1 h2 @
+ .
@⇣ h3 @⇣
In deriving these equations, the free stream velocity U 1
and L are used as the velocity and length scales. The Reynolds
number based on L is ReL = U 1 L/⌫ = 105 , for all the simu-
lations reported here. While solving the receptivity problem,
the Poisson equations for u, 3, and w components of velocity
given by Eqs. (B7)–(B9) are solved.
2. Boundary conditions
!
At inflow and the far-field: The components of velocity V
!
and vorticity ⌦ are obtained by using the Biot-Savart law for FIG. 12. Spanwise vorticity at the indicated times for the 3D receptivity case
the induced velocity as explained for the 2D case. of a convecting vortex for c = 0.3 and = 2 shown in the plane z = zmax2 .
054106-17 Sengupta et al. Phys. Fluids 30, 054106 (2018)

@⌦⌘ @⌦⌘ initial conditions, the 3D solver is run for approximately 1000
+ Uc = 0, (B14)
@t @x iterations (when all the unsteady terms fall below machine
@⌦⇣ @⌦⇣ zero) so that the flow field adjusts itself to the 3D domain and
+ Uc = 0. (B15) boundary conditions. Subsequently, 2D free stream excitation
@t @x
The boundary condition on ⌦⇠ at the outflow boundary is by the convecting vortex is initiated.
derived from the solenoidality condition of vorticity as
!
@⌦⇠ @⌦⌘ @⌦⇣ 4. Numerical methods and results
= + . (B16)
@x @y @z
For the 3D computations, we stagger the variables as indi-
The boundary conditions for the 3- and 4-components of cated in Fig. 11(b), with the components of vorticity defined
velocity at the outflow boundary are derived from the definition at the center of each sides of the cell, while the velocity com-
of the vorticity components ⌦⌘ and ⌦⇣ as ponents are placed at the center of the faces of the elementary
@v @u cube in the transformed plane. Optimized staggered compact
= + ⌦⇣ , (B17) scheme (OSCS) is used for the purpose of both interpolation
@x @y
and evaluation of first derivatives, as defined for the 2D com-
@w @u putations. The second or mixed derivative terms are evaluated
= ⌦⌘ . (B18)
@x @y by repeated application of the OSCS scheme for the evaluation
of first derivatives. The ORK 3 scheme3 is used to integrate the
VTEs [Eqs. (B1)–(B3)] with a time step of t = 8 ⇥ 10 5 . To
3. Initial condition
suppress numerical spanwise spurious oscillations, a periodic
Once the 2D equilibrium flow is established, the 3D equi- sixth-order filter with the filter coefficient ↵ f = 0.49 is used in
librium flow is obtained by specifying u, 3, and ⌦⇣ variables at the spanwise direction. For the purpose of de-aliasing, a 2D
all the discrete spanwise stations, while prescribing other vari- filter in (⇠, ⌘)-plane with a filter coefficient of ↵ 2D = 0.18 is
ables, i.e., 4, ⌦⇠ , and ⌦⌘ , as zero at all locations. With these used every twelfth time step. Equations (B1)–(B3) are solved

zmax
FIG. 13. (a) The vorticity components at t = 26, for the 3D receptivity case of a convecting vortex for c = 0.3 and = 2 shown in the plane z = 2 (contour).
(b) The vorticity components at t = 28, for the 3D receptivity case of a convecting vortex for c = 0.3 and = 2 shown in the plane z = zmax
2 .
054106-18 Sengupta et al. Phys. Fluids 30, 054106 (2018)

in the computational domain by domain decomposition tech- turbulence of fluid flows,” NACA Technical Memorandum Wash. No 1291,
nique and using MPI parallelization framework. We have used 1951).
19 Huang, H. and Li, M., “Finite-difference approximation for the velocity-
a grid with 1001 ⇥ 301 ⇥ 129 points in ⇠, ⌘, and ⇣ directions, vorticity formulation on staggered and non-staggered grids,” Comput.
respectively. The parallel code was run with 150 processors. Fluids 26(1), 59–82 (1997).
The run time is of the order of 26 h 30 min to advance the 20 Hunt, J. C. R. and Durbin, P., “Perturbed vortical layers and shear

computational time by unity. sheltering,” Fluid Dyn. Res. 24(6), 375–404 (1999).
21 Kendall, J. M., “Boundary layer receptivity to free stream turbulence,”
In Fig. 12, we show the computed results for the spanwise AIAA Paper No. 90-1504, 1990.
vorticity plotted in the plane z = zmax
2 at the indicated times,
22 Kerr, R. M., “Dissipation and enstrophy statistics in turbulence: Are the

when the 3D receptivity case is considered with the convect- simulations and mathematics converging?,” J. Fluid Mech. 700, 1–4 (2012).
23 Klebanoff, P. S., Tidstrom, K. D., and Sargent, L. M., “The three-
ing counter-clockwise vortex moving at a speed of c = 0.3
dimensional nature of boundary-layer instability,” J. Fluid Mech. 12, 1–34
and the circulation of = 2. One can clearly see the primary (1962).
and secondary vortices formed, as noted in the case of 2D 24 Landahl, M. T. and Mollo-Christensen, E., Turbulence and Random Pro-

computations also. cesses in Fluid Mechanics (Cambridge University Press, New York, USA,
In Figs. 13(a) and 13(b), we show the vorticity compo- 1992).
25 Leib, S. J. and Wundrow, D. W., and Goldstein, M. E., “Generation and
nents, ⌦⇠ , ⌦⌘ , and ⌦⇣ at t = 26 and t = 28, respectively. growth of boundary layer disturbances due to free-stream turbulence,”
As the free stream vortical excitation is two-dimensional, AIAA Paper No. 99-0408, 1999.
26 Lim, T. T., Sengupta, T. K., and Chattopadhyay, M., “A visual study of
the response field also clearly reveals the flow field to be
vortex-induced subcritical instability on a flat plate laminar boundary layer,”
two-dimensional during these primary and secondary stages Exp. Fluids 37, 47–55 (2004).
of vortex-induced instability. Similar such behavior was 27 Loiseau, J.-C., Robinet, J.-C., Cherubini, S., and Leriche, E., “Investigation

recorded in the experimental results shown in Ref. 26. Such of the roughness-induced transition: Global stability analyses and direct
two-dimensionality will yield to three-dimensionality with numerical simulations,” J. Fluid Mech. 760, 175–211 (2014).
28 Mathieu, J. and Scott, J., An Introduction to Turbulent Flows (Cambridge
progress in time. University Press, UK, 2000).
29 Matsubara, M. and Alfredsson, P. H., “Disturbance growth in boundary
1 Bhaumik, S. and Sengupta, T. K., “On the divergence-free condition layers subjected to free-stream turbulence,” J. Fluid Mech. 430, 149–168
of velocity in two-dimensional velocity-vorticity formulation of incom- (2001).
pressible Navier-Stokes equation,” in 20th AIAA Computational Fluid 30 Monin, A. S. and Yaglom, A. M., Statistical Fluid Mechanics: Mechanics
Dynamics Conference, 27–30 June, 2011, Honululu, Hawaii, USA (AIAA, of Turbulence (The MIT Press, Cambridge, MA, 1971).
2011). 31 Morkovin, M. V., “Panoramic view of changes in vorticity distribution in
2 Bhaumik, S. and Sengupta, T. K., “Precursor of transition to turbulence: transition, instabilities and turbulence,” in Transition to Turbulence, edited
Spatiotemporal wave front,” Phys. Rev. E 89(4), 043018 (2014). by Reda, D. C., Reed, H. L., and Kobyashi, R. (ASME FED Publication,
3 Bhaumik, S. and Sengupta, T. K., “A new velocity-vorticity formulation 1991), Vol. 114, pp. 1–12.
for direct numerical simulation of 3D transitional and turbulent flows,” 32 Nagarajan, S., Lele, S. K., and Ferziger, J. H., “A robust high-order com-
J. Comput. Phys. 284, 230–260 (2015). pact method for large eddy simulation,” J. Comput. Phys. 19, 392–419
4 Breuer, K. S. and Kuraishi, T., “Bypass transition in two and three (2003).
dimensional boundary layers,” AIAA Paper No. 93-3050, 1993. 33 Napolitano, M. and Pascazio, G., “A numerical method for the vorticity-
5 Brinckman, K. W. and Walker, J. D. A., “Instability in a viscous flow driven velocity Navier-Stokes equations in two and three dimensions,” Comput.
by streamwise vortices,” J. Fluid Mech. 432, 127–166 (2001). Fluids 19, 489–495 (1991).
6 Chu, B. T., “On energy transfer to small disturbances in fluid flow (Part-I),” 34 Obabko, A. V. and Cassel, K. W., “Navier-Stokes solutions of unsteady
Acta Mech. 1, 215–234 (1965). separation induced by a vortex,” J. Fluid Mech. 465, 99–130 (2002).
7 Daube, O., “Resolution of the 2D Navier-Stokes equations in velocity- 35 Peridier, V. J., Smith, F. T., and Walker, J. D. A., “Vortex-induced boundary-
vorticity form by means of an influence matrix technique,” J. Comput. Phys. layer separation. Part 1. The unsteady limit problem. Re ! 1,” J. Fluid
103, 402–414 (1992). Mech. 232, 99–131 (1991).
8 Davies, S. J. and White, C. M., “An experimental study of the flow of water 36 Peridier, V. J., Smith, F. T., and Walker, J. D. A., “Vortex-induced boundary-
in pipes of rectangular section,” Proc. R. Soc. A 119, 92 (1928). layer separation. Part 2. Unsteady interacting boundary-layer theory,”
9 Degani, A. T., Walker, J. D. A., and Smith, F. T., “Unsteady separation past J. Fluid Mech. 232, 133–165 (1991).
moving surfaces,” J. Fluid Mech. 375, 1–38 (1998). 37 Perraud, J., Arnal, D., Seraudie, A. and Tran, D., “Laminar-turbulent tran-
10 Doering, C. R. and Gibbon, J. D., Applied Analysis of the Navier-Stokes sition on aerodynamic surfaces with imperfections,” Paper presented at the
Equations (Cambridge University Press, UK, 1995). RTO AVT Specialists’ Meeting on “Enhancement of NATO Military Flight
11 Doligalski, T. L., Smith, C. R., and Walker, J. D. A., “Vortex interaction Vehicle Performance by Management of Interacting Boundary Layer Tran-
with wall,” Annu. Rev. Fluid Mech. 26, 573–616 (1994). sition and Separation,” RTO-MP-AVT-111 (STO Publisher RTO, Prague,
12 Donzis, D. A. and Yeung, P. K., “Resolution effects and scaling in numerical Czech Republic, 2004), pp. 14-1–14-13.
simulations of passive scalar mixing in turbulence,” Phys. D 239, 1278–1287 38 Raudkivi, A. J. and Callander, R. A., Advanced Fluid Mechanics: An
(2010). Introduction (Edward Arnold, USA, 1975).
13 Donzis, D. A., Yeung, P. K., and Sreenivasan, K. R., “Energy dissipation 39 Robinson, S. K., “Coherent motions in the turbulent boundary layer,” Annu.
rate and enstrophy in isotropic turbulence: Resolution effects and scaling in Rev. Fluid Mech. 23, 601–639 (1991).
direct numerical simulations,” Phys. Fluids 20, 045108 (2008). 40 Schlichting, H., “Zur entstehung der turbulenz bei der plattenströmung,”
14 Drazin, P. G. and Reid, W. H., Hydrodynamic Stability (Cambridge Nach. Gesell. d. Wiss. z. Gött., MPK, 181–208 (1933).
University Press, UK, 1981). 41 Schmid, P. J., “Linear stability theory and bypass transition in shear flows,”
15 Garcia-Colin, L. S., Velasco, R. M., and Uribe, F. J., “Beyond the Navier- Phys. Plasmas 7, 1788 (2000).
Stokes equations: Burnett hydrodynamics,” Phys. Rep. 465(4), 149–189 42 Schmid, P. J. and Henningson, D. S., Stability and Transition in Shear Flow
(2008). (Springer-Verlag, New York, USA, 2001).
16 Gatski, T. B., Grosch, C. E., and Rose, M. E., “A numerical study of 43 Schubauer, G. B. and Skramstad, H. K., “Laminar boundary layer oscil-
the 2-dimensional Navier-Stokes equations in vorticity-velocity variables,” lations and the stability of laminar flow,” J. Aeronaut. Sci. 14, 69–78
J. Comput. Phys. 48, 1–22 (1982). (1947).
17 Guj, G. and Stella, F., “A vorticity-velocity method for the numerical of 3D 44 Sengupta, T. K., Instabilities of Flows and Transition to Turbulence (CRC
incompressible flows,” J. Comput. Phys. 106, 286–298 (1993). Press, Taylor & Francis Group, Florida, USA, 2012).
18 Heisenberg, W., “Über stabilität und turbulenz von Flüssigkeitsströmen,” 45 Sengupta, T. K., High Accuracy Computing Methods: Fluid Flows and Wave
Ann. der Phys. 379, 577–627 (1924) (translated as “On stability and Phenomenon (Cambridge University Press, New York, USA, 2013).
054106-19 Sengupta et al. Phys. Fluids 30, 054106 (2018)

46 Sengupta, T. K. and Dipankar, A., “Subcritical instability on the 57 Smith, C. R., Walker, J. D. A., Haidari, A. H., and Soburn, U., “On the
attachment-line of an infinite swept wing,” J. Fluid Mech. 529, 147–171 dynamics of near-wall turbulence,” Philos. Trans. R. Soc., A 336, 131–175
(2005). (1991).
47 Sengupta, T. K. and Bhaumik, S., “Onset of turbulence from the receptivity 58 Sommerfeld, A., Partial Differential Equation in Physics (Academic Press,

stage of fluid flows,” Phys. Rev. Lett. 107, 154501 (2011). New York, USA, 1949).
48 Sengupta, T. K., Chattopadhyay, M., Wang, Z.-Y., and Yeo, K. S., “By- 59 Suman, V. K., Sengupta, T. K., Durga Prasad, C. J., Mohan, K. S., and

pass mechanism of transition to turbulence,” J. Fluids Struct. 16, 15–29 Sanwalia, D., “Spectral analysis of finite difference schemes for convection
(2002). diffusion equation,” Comput. Fluids 150, 95–114 (2017).
49 Sengupta, T. K., De, S., and Sarkar, S., “Vortex-induced instability of an 60 Taylor, G. I., “Statistical theory of turbulence. V. Effects of turbulence on

incompressible wall-bounded shear layer,” J. Fluid Mech. 493, 277–286 boundary layer,” Proc. R. Soc. A 156(888), 307–317 (1936).
(2003). 61 Tennekes, H. and Lumley, J. L., A First Course in Turbulence (MIT Press,
50 Sengupta, T. K., Ganeriwal, G., and De, S., “Analysis of central and upwind Cambridge, MA, 1971).
compact schemes,” J. Comput. Phys. 192(2), 677–694 (2003). 62 Tollmien, W., “Über die enstehung der turbulenz. I,” English translation
51 Sengupta, T. K., Rao, A. K., and Venkatasubbaiah, K., “Spatiotemporal NACA TM 609, 1931.
growing wave fronts in spatially stable boundary layers,” Phys. Rev. Lett. 63 Van der Vorst, H. A., “Bi-CGSTAB: A fast and smoothly converging variant

96, 224504 (2006). of Bi-CG for the solution of non-symmetric linear systems,” SIAM J. Sci.
52 Sengupta, T. K., Rao, A. K., and Venkatasubbaiah, K., “Spatio-temporal Stat. Comput. 12, 631–644 (1992).
growth of disturbances in a boundary layer and energy based receptivity 64 Wu, X. H., Wu, J. Z., and Wu, J. M., “Effective vorticity-velocity formula-

analysis,” Phys. Fluids 18, 094101 (2006). tions for 3D incompressible viscous flows,” J. Comput. Phys. 122, 68–82
53 Sengupta, T. K., Dipankar, A., and Sagaut, P., “Error dynamics: Beyond von (1995).
Neumann analysis,” J. Comput. Phys. 226(2), 1211–1218 (2007). 65 Wu, X., Jacobs, R. G., Hunt, J. C. R., and Durbin, P., “Simulation of boundary
54 Sengupta, T. K., Bhole, A., and Sreejith, N. A., “Direct numerical simulation layer transition induced periodically passing wakes,” J. Fluid Mech. 399,
of 2D transonic flows around airfoils,” Comput. Fluids 88, 19–37 (2013). 109–153 (1999).
55 Sengupta, T. K., Singh, H., Bhaumik, S., and Chowdhury, R. R., “Diffusion 66 Wu, X., Moin, P., and Hickey, J.-P., “Boundary layer bypass transition,”

in inhomogeneous flows: Unique equilibrium state in an internal flow,” Phys. Fluids 26, 091104 (2014).
Comput. Fluids 88, 440–451 (2013). 67 Yeung, P. K., Donzis, D. A., and Sreenivasan, K. R., “Dissipation, enstrophy
56 Shapiro, A. H., The Dynamics and Thermodynamics of Compressible Fluid and pressure statistics in turbulence simulations at high Reynolds numbers,”
Flow (John Wiley & Sons, USA, 1977), Vol. 1. J. Fluid Mech. 700, 5–15 (2012).

You might also like