This draft was prepared using the LaTeX style file belonging to the Journal of Fluid Mechanics 1

Wavepackets in turbulent flows over airfoils

Leandra I. Abreu1,2 †, Alvaro Tanarro3 André V. G. Cavalieri2 ,
Philipp Schlatter3 , Ricardo Vinuesa3 Dan S. Henningson3 and
Ardeshir Hanifi3
1
São Paulo State University (UNESP), Campus of São João da Boa Vista, 13876-750, SP,
Brazil
2
Divisão de Engenharia Aeronáutica, Instituto Tecnológico de Aeronáutica, 12228-900, São
José dos Campos, SP, Brazil
3
Linné FLOW Centre, KTH Mechanics, Stockholm, SE-100-44, Sweden

(Received xx; revised xx; accepted xx)

Motivated by the recent analysis by Sano et al. (2019) of spanwise-coherent structures

in the turbulent flow around airfoils and their connection to trailing-edge noise, we carry
out a thorough characterisation of such structures in three simulation databases. We
analyse two different numerical simulations of incompressible flow in turbulent regime,
both at chord Reynolds number of 4 · 105 : a large-eddy simulation for a NACA 0012
profile at zero angle of attack, and a direct numerical simulation for a NACA 4412 airfoil
with an angle of attack of 5 degrees. Snapshots of the flow field were analysed using
Spectral Proper Orthogonal Decomposition (SPOD), in order to extract the dominant
coherent structures of the flow. Focus is given to the aforementioned spanwise-coherent
fluctuations, which two-dimensional disturbances in the computational domain due to
the use of periodic boundary conditions. The leading SPOD modes show that the
most energetic coherent structures are wavepackets, extending over the whole turbulent
boundary layers around the airfoils with significant amplitudes near the trailing-edge.
Higher amplitudes are observed in the region of stronger adverse pressure gradient at
the suction side of the NACA 4412 airfoil. To understand how such structures in the
turbulent field can be modelled, the linear response of the flow using the singular value
decomposition of the linearised resolvent operator was performed, using the mean field
as a base flow and considering a locally parallel approximation. Such analysis shows
that the leading SPOD modes can be associated to optimal, linearised flow responses,
particularly for stations far from the trailing edge; the latter introduces a discontinuity
in boundary conditions, and the locally parallel approximation becomes questionable.
We then focus on evaluating the dependence of such wavepackets on the domain size, to
ensure that these structures are not an artifact of the use of periodic boundary conditions
in small computational boxes. To do so, we performed an incompressible LES of a zero-
pressure gradient turbulent boundary layer (ZPGTBL), for three different spanwise sizes:
Lz = 32δ ∗ , Lz = 64δ ∗ and Lz = 128δ ∗ , where δ ∗ is a reference displacement thickness
in a region of developed turbulent flow, with Reynolds number matching the values in
the airfoil simulations. The signature of such wavepackets is seen in non-premultiplied
spanwise wavenumber spectra, which reaches, for the three domain sizes, a plateau for
spanwise wavelengths going to infinity (or wavenumbers going to zero); this plateau is
representative of the spanwise-coherent structures seen in the airfoil simulations. Similar
SPOD and resolvent analyses were carried out for the zero spanwise wavenumber of
the ZPGTBL, and the same coherent wavepackets were observed for the three domains,
with very similar amplitudes. Such wavepackets were also accurately modelled using the
† Email address for correspondence: leandra.abreu@unesp.br
2
optimal resolvent response. These results confirm that the spanwise-elongated structures
are not domain-size dependent for the studied simulations, and are thus a feature of
turbulent boundary layers.

Key words: SPOD, resolvent analysis, wall-bounded turbulence, wavepackets.

1. Introduction
Coherent structures play an important role in turbulent flows, for both drag and noise
generation. The modelling of these structures using a linearisation of the Navier–Stokes
system, for flows such as jets, boundary layers and wakes, has been attempted recently by
Del Alamo & Jimenez (2006), McKeon & Sharma (2010), Cavalieri et al. (2013), Abreu
et al. (2019), to cite a few examples. Usually such linearised models lead to a definition of
a set of modes that together describe coherent turbulent structures. Some of these may
play an important role for aeroacoustics (Cavalieri et al. 2013; Jeun et al. 2016; Sano
et al. 2019). More specifically, when one considers sound radiation by airfoils, wings
and blades at low angles of attack, the dominant aeroacoustic mechanism is referred
to as trailing-edge noise. In that case turbulent fluctuations are scattered into acoustic
waves by the edge, in a mechanism with high acoustic efficiency at low Mach numbers
(Ffowcs Williams & Hall 1970). It is thus natural to use the cited linearised models to
study how coherent turbulent structures may be associated with trailing-edge noise.
To educe coherent structures in experimental or numerical data, a useful, data-driven
approach, is Proper Orthogonal Decomposition (POD) (Lumley 2007). POD is a quanti-
tative method often applied to instantaneous fields in order to analyse coherent structures
in turbulent flows. POD provides a basis for the modal decomposition as an ensemble of
functions, called POD modes, which are extracted from a set of data. In terms of energy,
the leading POD modes capture the most energetic structures. Thus, if the dynamics of
the flow have a few predominant flow structures, the data can often be well represented
using just some of the first modes. The most energetic modes will then represent the
dominant flow structures. This method has been successfully used in various types of
flows, such as mixing layers (Wei & Freund 2006), jets (Freund & Colonius 2009), channel
flows (Alfonsi et al. 2001), cylinder wakes (Noack et al. 2003) and boundary layers (Aubry
et al. 1988); the last case is the closest to that of the present study.
POD in frequency domain, also labeled as spectral POD, or SPOD (Picard & Delville
2000), has been used to obtain coherent structures in turbulent airfoil flows (Sano
et al. 2019), since SPOD is better suited for identifying physically meaningful coherent
structures (Towne et al. 2018). Recent works have explored the connection of SPOD
modes with the flow responses to stochastic forcing; such responses are obtained using the
linearised resolvent operator (McKeon & Sharma 2010). These linearised responses can
often be related to results of hydrodynamic stability theory, with modes corresponding
to instability waves or to non-modal mechanisms such as lift-up (Jovanovic & Bamieh
2005), but in recent years resolvent analysis has been used to study coherent structures in
turbulent wall bounded flows (see the review of McKeon (2017) and references therein).
An important result is that if the forcing is white noise, a direct correspondence between
SPOD and resolvent modes is expected; furthermore, whenever the resolvent operator has
an optimal gain much larger than suboptimals, the leading flow response, predicted by
resolvent analysis, often also dominates the flow statistics and appears as the first SPOD
 3
mode (Beneddine et al. 2016; Towne et al. 2018). This way, a combined analysis of the
flow, with SPOD serving for the signal processing of snapshots of the field and resolvent
analysis as a theoretical framework, enables a reduced-order model of the dynamically-
relevant flow features. For this reason, SPOD will be used in this work to study turbulent
structures, coupled with resolvent analysis as a modelling framework.
A recent study by Sano et al. (2019) used flow-acoustic correlation and SPOD to
analyse the compressible turbulent flow over a NACA 0012 airfoil. The results showed
spanwise coherent disturbances in the boundary layer, which are two-dimensional in
the simulation due to spanwise periodicity, and they are strongly correlated to the
acoustic field; leading SPOD modes showed wavy structures throughout the airfoil
surface. Those spanwise-elongated structures are not often studied in wall turbulence
due to the widespread use of pre-multiplied spectra (Hutchins & Marusic 2007), but
nonetheless they are relevant for aerocoustics (Amiet 1976; Nogueira et al. 2017). Thus,
two-dimensional perturbations associated to spanwise wavenumber equal to zero is the
main focus in the current paper.
In the present study, we analyse two different numerical simulations in turbulent
regime, both for incompressible flow at chord Reynolds number of Rec = 400, 000: a
large-eddy simulation (LES) for a NACA 0012 airfoil at zero angle of attack (Tanarro
et al. 2018), and a direct numerical simulation (DNS) for a NACA 4412 profile with an
angle of attack of 5◦ (Hosseini et al. 2016). Both databases are studied using SPOD as
a signal post-processing approach. The mean flow will also be used as a basis for the
computation of the linearised flow response to harmonic forcing, represented by singular
modes of the resolvent operator at specific frequencies. This allows to identify the optimal
harmonic forcing, and verify whether it can be related to observed relevant structures
obtained by SPOD.
In order to demonstrate the domain-size independence and the physical sources of these
wavepackets, we performed an incompressible LES of a zero-pressure gradient turbulent
boundary layer (ZPGTBL) for three different spanwise sizes with a Reynolds number
similar to the airfoil simulations. We repeated the analysis using SPOD in the ZPGTBL
databases, using the correspondent analysed frequency in the airfoils.
Results of the present study show the presence of wavepackets on wall-bounded turbu-
lent flows, which can be modeled using locally parallel resolvent analysis, i.e., a simple
reduced-order model. This knowledge can be used to better understand mechanisms of
trailing-edge noise, which can lead to wing modifications so as to reduce sound radiation.
Understanding and modelling such wavepackets can point to novel approaches in order
to design more silent aircraft.
The paper is organized as follows: in § 2 we present information about the numerical
simulations for both airfoils and the flat-plate boundary layer used in the present analysis.
In § 3 we present the techniques used to analyse the numerical databases: the SPOD
method, the resolvent formulation and the relation between them. The modeling of
wavepackets found in the airfoils surfaces is shown in § 4. Finally, the domain-size effect
on the spanwise-averaged disturbances is demonstrated in § 5 for the flat-plate boundary
layer simulation.

2. Databases
2.1. NACA 0012
The NACA 0012 airfoil database analysed in this work is described in Tanarro et al.
(2018). In this study, a well-resolved large-eddy simulation (LES) was performed to
4
analyse the incompressible turbulent flow around a NACA 0012 airfoil at a chord
Reynolds number of Rec = 400, 000, with zero angle of attack. An unsteady volume force
was used to force transition to turbulence, at 10% of the chord on both top and bottom
sides of the wing. The incompressible spectral-element Navier–Stokes solver Nek5000
(Fischer et al. 2008) was used to carry out the simulation.
The spatial discretisation is performed by means of Lagrange interpolants of poly-
nomial order N = 11 and the spatial resolution in the boundary layer is expressed in
wall units: ∆x+ = 18, ∆y + = (0.64, 11) and ∆z + = 9. Note that x, y and z denote
streamwise, wall-normal and spanwise directions, respectively. The scaling in wall units
is in terms of the friction velocity uτ and kinematic viscosity ν. The domain considered
here is a C-mesh with streamwise length Lx = 6c, wall-normal length Ly = 4c and
periodic spanwise length Lz = 0.1c, where c is the wing chord. A preliminary RANS
simulation was used to obtain the velocity distributions used as boundary conditions.
The domain of the NACA 0012 is discretised using a total of 220,000 spectral elements,
which amounts to approximately 380 million grid points.
For this database we extracted 515 snapshots from the simulation, with a non-
dimensional timestep between them of ∆tc = 0.01, normalized by the freestream velocity
U∞ and the wing chord c.

2.2. NACA 4412

The NACA 4412 airfoil database analysed in this work is described in Hosseini et al.
(2016). The incompressible direct numerical simulation (DNS) was performed for a
turbulent flow over a NACA 4412 airfoil at a chord Reynolds number of Rec = 400, 000,
with an angle of attack AoA = 5◦ . An unsteady volume force was also used to force
transition to turbulence at 10% of the chord on both top and bottom sides of the wing.
The simulation was also carried out using Nek5000, this time without a subgrid scale
model. Figure 1 shows a visualization of turbulent structures in a flow snapshot.
The spatial discretisation is performed by means of Lagrange interpolants of polyno-
mial order N = 11 and the mesh is structured, and is designed to meet the following
criteria around the wing section: ∆x+ < 10, ∆y + < 0.5 and ∆z + < 5. The domain
considered here is a C-mesh with streamwise length Lx = 6c, wall-normal length Ly = 4c
and periodic spanwise length Lz = 0.1c, where c is the aerodynamic chord. The mesh
was optimized to properly resolve all relevant scales in the flow, and comprises around
3.2 billion grid points. A preliminary RANS simulation was used to obtain the velocity
distribution used as a boundary condition.
For this database we extracted 1124 snapshots from the simulation, with a non-
dimensional timestep between them of ∆tc = 0.008.

2.3. Zero-pressure gradient turbulent boundary layer

It would be too computationally expensive to run other LES or DNS with different
domain sizes for the present airfoil configurations. In order to examine the role of
computational domain size in a simpler, canonical setting, we thus performed three
incompressible LES of a zero pressure gradient turbulent boundary layer (ZPGTBL),
using the pseudo-spectral solver, SIMSON, as in Schlatter et al. (2014). The horizontal
and vertical lengths for all the three simulations are Lx = 1000δ0∗ and Ly = 90δ0∗ ,
respectively, and simulations are carried out for three spanwise lengths, Lz = 96δ0∗ ,
Lz = 192δ0∗ and Lz = 384δ0∗ , where δ0∗ is the displacement thickness at the inflow
boundary, a location at which the displacement-thickness Reynolds number is set to
450 for all cases. The number of points used to define the mesh for the horizontal and
 5

(a) (b)

Figure 1. (a) Schematic three-dimensional layout of the set up for direct numerical simulation.
The chord length is denoted as c. The domain extends to 5c downstream, and 1c upstream, top
and bottom from the leading edge. (b) Instantaneous visualization of the DNS results showing
coherent vortices identified by means of the λ2 criterion. Figures extracted from Hosseini et al.
(2016).

vertical directions are nx = 256 and ny = 241, respectively, and in the spanwise direction
they are nz = 48, nz = 96 and nz = 192, respectively in increasing order of each spanwise
length.
Towards the end of the domain the Reynolds number based on local displacement
thickness Reδ∗ is 1350. This Reynolds number corresponds to Reδ∗ near the trailing
edge region, x/c = 0.75, of the NACA 0012 airfoil. All the analysis throughout this
paper concerning ZPGTBL databases will be pursuit considering a normalization by the
local displacement thickness δ ∗ . At this boundary-layer position, the domain sizes become
Lz = 32δ ∗ , 64δ ∗ and 128δ ∗ . The corresponding spanwise length for the airfoil calculations,
considering a position x/c = 0.75 of the NACA 0012 simulation, is Lz ≈ 28δ ∗ . An
unsteady volume force was used to force transition to turbulence at the beginning of the
domain, following Schlatter & Örlü (2012).
For these databases we extracted 10000 snapshots from the simulations, with a non-
dimensional timestep between them of ∆tδ∗ = 0.3, normalized by the freestream velocity
U∞ and local δ ∗ , which correspond to the value of ∆tδ∗ at x/c = 0.75 from the NACA
0012 airfoil.
Figure 2 (a) shows the instantaneous field of the streamwise velocity component u,
for the ZPGTBL simulation with Lz = 64δ ∗ . Figure 2 (b) shows the friction coefficient
for the three ZPGTBL simulations, which is essentially identical in the three cases as
expected, where the tripping location is at x/δ ∗ ≈ 18, the red dashed lines delimit the
fully turbulent region, and the sponge zone is 300 6 x/δ ∗ 6 400.

3. Methods
To standardize the parameters used throughout the paper, we consider the standard
Reynolds decomposition q = q̄ + q0 , where q̄ is time average, q0 is the fluctuation, and
q = [u; v; p] a vector with horizontal and vertical velocity components and pressure,
respectively.
6
10-3
7
*
32
6 *
64
5 *
124

4
3
2
1
50 100 150 200 250 300

(a) (b)

Figure 2. ZPGTBL simulation results: (a) instantaneous field of the streamwise velocity
component u, for the ZPGTBL simulation with Lz = 64δ ∗ ; (b) friction coefficient for the three
ZPGTBL simulations

3.1. Spectral Proper Orthogonal Decomposition

POD consists in finding, within an ensemble of realizations of the flow field, orthogonal
basis functions that maximize the mean square energy (Berkooz et al. 1993). In this study,
POD in the frequency domain, or SPOD (Picard & Delville 2000; Towne et al. 2018),
was employed to analyze all the numerical databases described in § 2.
Prior to the SPOD, we perform a spanwise Fourier transform of the flow fluctuations.
Since the simulations described here use periodic boundary conditions in the spanwise
direction z, the span can be considered as a homogeneous direction and Fourier modes
in z-axis become the optimal orthogonal basis functions in this direction (Berkooz et al.
1993); due to the periodicity in z the decomposition becomes thus a Fourier series,
with discrete spanwise wavenumbers kz = 2πn/Lz , with n = 0, 1, 2..., where Lz is the
domain size in z. Focus is given to the two-dimensional mode, which is most relevant for
low-frequency trailing-edge noise (Amiet 1976; Nogueira et al. 2017; Sano et al. 2019).
Accordingly, in what follows it is implicit that a spanwise average has been taken for all
flow quantities, which no longer have a z-dependence. The truncation of computational
boxes is well studied for turbulent channel flow, and it is known that accurate statistics
can be obtained with relatively small domains, with structures larger than the domain
size appearing as constant along a given direction (Lozano-Durán & Jiménez 2014). We
thus interpret the structures studied here as spanwise-elongated disturbances, whose
wavelength in span is much larger than the boundary-layer thickness.
A further decomposition can be applied to the fields prior to computation of SPOD
modes. Since the NACA 0012 airfoil at zero angle of attack is symmetric about the
streamwise x-axis, the flow along upper and lower surfaces of the airfoil should have the
same statistics; accordingly, fluctuations can be split into parts that are even (symmet-
ric) or odd (antisymmetric) with respect to the airfoil chord. Since acoustic radiation
by trailing-edges was seen to be antisymmetric (Sano et al. 2019), focus is given to
streamwise velocity fluctuations and pressure disturbances which are odd with respect to
the chord; the wall-normal velocity must in turn be symmetric to satisfy the continuity
equation. For the cambered NACA 4412 airfoil, with angle of attack of 5 degrees, such
decomposition is not possible, and thus we apply SPOD to the whole field.
SPOD is applied to the fluctuations of velocity components u0 and v0 and pressure
disturbances p0 , and the norm used here is the turbulent kinetic energy (TKE), sim-
ilarly to Sano et al. (2019). First, a Fourier transform is performed to the velocity
and pressure fields in time to obtain the field for a specific frequency f , qb0 (x, f ) =
u0 (x, f ); v
[b b 0 (x, f ); p
b 0 (x, f )], where hats denote Fourier-transformed quantities; then we
 7
apply SPOD to the transformed field, which amounts to solving the integral equation:
Z
C(x, x̃, f )W(x̃)Ψ (x̃, f )dx̃ = λ(f )Ψ (x, f ), (3.1)

where Ψ represents the basis functions, or SPOD modes; λ is the corresponding SPOD
eigenvalues; W is a weighting function, with weights equal to zero for the pressure field,
since the norm used here is TKE; and C is the two-point cross-spectral density tensor,
given by:
H
C(x, x̃, f ) = E(qb0 (x, f )qb0 (x̃, f )), (3.2)
where E(·) is the expected value operator, and the superscript H denotes a conjugate
transpose of the vector. C is Hermitian, and thus its eigenvalues are real and eigenfunc-
tions are orthogonal.
The realizations of the flow, which in the frequency domain amount to short-time
Fourier transforms taken at segments of data, can be expanded as:
Nb
X
qb0 (x, f ) = ak (f )Ψ k (x, f ), (3.3)
k=1

with ak (f ) = hqb0 (x, f ), Ψ k (x, f )i, and Nb is the total number of blocks required by the
Welch’s method. For lower computational cost, the snapshot method (Sirovich 1987) is
applied, as described by Towne et al. (2018).
For the NACA 0012 airfoil the short-time FFTs required for SPOD have been taken
considering blocks of 32 snapshots with 50% overlap, leading to a total number of blocks
of Nb = 31; and for the NACA 4412, were considered blocks of 64 snapshots with 50%
overlap, leading to a total number of blocks of Nb = 34. SPOD was also evaluated for
both airfoils using blocks containing 48 snapshots with 50% and 75% of overlap, leading
to very similar results.
For the three ZPGTBL databases the short-time FFTs required for SPOD have been
taken considering blocks of 512 snapshots with 75% overlap, leading to a total number
of blocks of Nb = 59. SPOD was also evaluated for the three simulations using blocks
containing 400 snapshots with 50% and 75% of overlap, leading to very similar results.

3.1.1. Convergence analysis

In order to further verify the reliability of the computed SPOD modes, for both airfoils,
we carry out a convergence analysis, by dividing the total dataset into two equal parts
each corresponding to 75% of the original dataset, and performing the SPOD on each
part, so the correlation coefficient is given by:

hΨk , Ψi,k i
µi,k = , (3.4)
||Ψk || · ||Ψi,k ||
where i = (1; 2) indicates each subset and k = 1, ..., Nb each SPOD mode. This kind
of analysis was also performed by Lesshafft et al. (2019), Sano et al. (2019) and Abreu
et al. (2019). For well-converged statistics, the SPOD modes for each subset are nearly
identical, and thus the normalized projection of eq. (3.4) leads to µi,k close to 1.
In the present database the first SPOD modes were found to be well converged in all
analysed cases studied in this paper, with µi,k > 0.95 for k = 1, 2 and 3. We conclude
that the first three modes are well converged and can be meaningfully analysed. In this
paper we focus our analysis on the first mode for brevity.
8
3.2. Resolvent analysis
The present resolvent analysis follows the procedure outlined by Towne et al. (2018).
Resolvent analysis provides two orthonormal bases, one for forcings and the other one
for the associated flow responses, and each pair of forcing and response modes is related
by a gain. This provides a hierarchy of forcing-responsed pairs, ordered by the gain. If
the optimal forcing has much larger gain than suboptimal ones, the corresponding flow
response is expected to dominate the flow statistics in the turbulence field (Beneddine
et al. 2016; Towne et al. 2018; Cavalieri et al. 2019).
Resolvent analysis requires the linearisation of the Navier–Stokes equations (N-S)
isolating the non-linear terms at the right-hand side. We will for simplicity consider
the locally-parallel problem, where the mean flow q̄ = [ū; v̄; p̄] has its divergence in x
neglected and thus varies only in y. The base-flow ū(y) becomes thus independent of x-
axis and time t, and the linearised equations are homogeneous along these directions,
which allows the parallel-flow ansatz: q0 (x, y, t) = qb0 (y)ei(ωt−kx x) , where hats now
denote quantities that are Fourier transformed in both x and t, ω = 2πf is the
angular frequency and kx is the streamwise wavenumber. Notice that as discussed
in the previous section, we will focus on two-dimensional disturbances representative
of spanwise-elongated structures (Sano et al. 2019), and thus the ansatz above has
no z-dependence. Substitution of the above equations in the linearised Navier–Stokes
and continuity equations, and considering that the Reynolds number, frequency ω and
streamwise wavenumber kx are given, leads to a forced linear problem. The problem can
be written in state-space formulation, in operator notation:

(
Lqb0 = Bb f
(3.5)
yb0 = C qb0

where L = (iωI − A) is the linear operator applied to qb0 = [ub0 ; vb0 ; pb0 ]; b
f represents the
nonlinear terms from N-S; yb0 = [ub0 ; vb0 ]; C is the linear operator used to select certain
flow variables/regions of interest, in this case u0 and v 0 ; and B is the linear operator
that enforces known properties of nonlinearity in certain flows; in the present case, B
guarantees that no force will be applied in the continuity equation. Thus, the linear
operators B and C impose restrictions in forcing terms, and in some quantities of interest
in the output, respectively.
The problem is closed with homogeneous Dirichlet boundary conditions for the velocity
fluctuations. At the wall we enforce u0 (y = 0) = v 0 (y = 0) = 0. In the far field, we have
u0 (y → ∞) = v 0 (y → ∞) = 0. The resolvent problem requires to write eq. (3.5) in
input-output form, as:

yb0 = C(iωI − A)−1 Bb

f = Rb
f, (3.6)
where R is the resolvent operator. Now we can obtain a relationship between input and
output by applying singular-value decomposition (SVD) of the resolvent operator R,
given by:

R = UΣVH , (3.7)
which decomposes R into two orthonormal bases: the output basis, or response modes
U, and the input basis, or forcing modes V. The matrix of gains Σ is diagonal with real
positive values, with gains in decreasing order σ1 > σ2 > ... > σn .
The analyses here were discretised using a Chebyshev pseudo-spectral method (Tre-
 9
fethen 2000). A total of 301 Chebyshev polynomials have been used in the discretisation,
and we have verified that increasing the number of polynomials does not modify the
results.

3.3. SPOD vs. resolvent analysis

The mathematical relationship between SPOD and resolvent analysis can be expressed
as the relation between the realisations qb0 and the resolvent operator R for a problem
with harmonic forcing b
f as:

qb0 = Rb
f. (3.8)
The analysis of stochastic fields requires a formulation in terms of two-point statistics.
This can be obtained by multiplying eq. (3.8) by its Hermitian and taking the expected
value E(·). This leads to:

Ψ ΛΨ H = RE(b f H )RH ,
fb (3.9)
where the eigenvalue decomposition of the CSD E(b bH ) was applied to the left-hand
qq
side, leading to SPOD modes. If the forcing is white noise in space, or E(bf H ) = I, eq.
fb
(3.9) becomes:
Ψ ΛΨ H = UΣ 2 UH , (3.10)
meaning that the SPOD modes are equal to the response modes Ψ = U, and SPOD
eigenvalues equal to the square of resolvent gains Λ = Σ 2 .
The expressions above consider an Euclidean inner product, which is appropriate for
matrices. The non-uniform grids used in this work require the use of integration weights
for the discretisation of the inner product. Resolvent analysis and SPOD should be
modified so as to account for integration weights; appropriate expressions are presented
by Towne et al. (2019) and Lesshafft et al. (2019) and not repeated here for brevity.

4. Wavepackets in airfoils
4.1. SPOD results
In this section we will analyse both airfoil databases using SPOD to highlight the
dominant two-dimensional structures in the flow. Figures 3 (a) and (b) show the first ten
SPOD eigenvalues λ as a function of the Strouhal number St for both NACA 0012 and
4412 airfoils, respectively. The Strouhal number is a normalized frequency, defined as
St = f c/U∞ , where f is the frequency and U∞ is the freestream velocity. Results show
for both airfoils a peak, more pronounced in the first eigenvalue λ1 , for St ≈ 7. We can
also notice that the larger difference between first and second eigenvalues correspond to
the same Strouhal number of the peak, St ≈ 7; such difference indicates a dominance of
the leading mode in the flow fluctuations.
The recent study by our group (Sano et al. 2019) using a compressible LES of a NACA
0012 airfoil, showed a peak of trailing-edge noise at about St = 7, which is the same peak
frequency found here for the velocity fields of both analysed airfoils. Therefore, we focus
our analysis throughout this paper on St ≈ 7 for both airfoils.
Figures 4 (a) and (b) show the first SPOD mode of the NACA 0012 airfoil, corre-
sponding to St ≈ 7, for the horizontal u and vertical v velocity components of the flow,
respectively. Figures 5 (a) and (b) also show the first SPOD mode, but for the NACA
4412 airfoil, for the same quantities at the same Strouhal√number. All SPOD results are
shown multiplied by the square-root of the eigenvalue, λ1 , as in Sinha et al. (2014),
10
1 1

2 10-6 2
10-7 3 3

4 4

5 10-7 5

6 6
10-8 7 7

8 10-8 8

9 9

10 10
0 10 20 30 40 50 0 10 20 30 40 50

(a) (b)

Figure 3. First ten SPOD eigenvalues as a function of the frequency St for the (left) NACA
0012 and (right) NACA 4412 airfoils.

such that these correspond to fluctuation amplitudes in the flow. The real part of the
modes is shown; the imaginary part forms a similar pattern, and both correspond to a
wave with phase speed towards following the positive x axis.
All the results here show a coherent two-dimensional wave in all extension of the profile
after the boundary-layer tripping, with more pronounced amplitudes at the trailing-edge
region, and higher amplitudes for the NACA 4412 airfoil. Such wavepackets shown in
all quantities of the first SPOD mode appear, even though the boundary layer is at
an already developed turbulent state; this shows that once careful post-processing is
performed, the turbulent boundary layer exhibits spanwise-coherent structures forming
a hydrodynamic wave. The trailing edge is a region of adverse pressure gradient in the
boundary layer, which is known to enhance boundary-layer instability for laminar flows
(Schmid & Henningson 2012), and the present results suggest that a similar mechanism
may be at work in the turbulent boundary layers near the trailing edge.
The SPOD results for the streamwise velocity component u in all cases (see Figures 4
(a) and 5 (a)) show near-wall fluctuations with phase opposition to disturbances towards
the boundary-layer edge, which suggests a behaviour reminiscent of Tollmien-Schlichting
waves, although this is a region of turbulent boundary layer. A similar behaviour was
found by Sano et al. (2019) for the NACA 0012 airfoil and by Kaplan et al. (2019) for a
turbulent boundary layer inside a jet nozzle. The present results show that wavepackets
are not an artifact of a particular database.
In order to further highlight the structures in the boundary-layer region, the wake
contribution is neglected in SPOD by imposing zero at the integral weights W through
the wake region, where x/c > 1. We refer to these as boundary-layer weighted SPOD
modes. The results in that case are shown in Figure 6 (a) and (b) for the NACA 0012
and in Figure 7 (a) and (b) for the NACA 4412 airfoil. The SPOD results now emphasise
the wavepackets in the airfoils’ surfaces, inside the boundary layer. Since the wake region
typically has a Kelvin–Helmholtz unstable mode which is much more energetic then
the ones from turbulent boundary layer, which has just stable modes, boundary-layer
weighted SPOD modes highlight more clearly structures within the boundary layer;
inspection of the figures shows that these are the same of standard SPOD modes, but
convergence-related noise is lower when the wake is discarded. As the wake dynamics is
more thoroughly studied in the literature (see for instance Araya et al. (2017)), we will
thus focus for the remainder of our analysis on boundary-layer-weighted modes.
 11

(a) Streamwise velocity component u

(b) Wall-normal velocity component v

Figure 4. First SPOD mode at St ≈ 7 for the NACA 0012 airfoil.

4.2. Comparison between SPOD and resolvent modes

In this section we will compare the SPOD results obtained in Section 4.1 for both
airfoils with the most amplified responses from resolvent analysis. To do so, we will make
a locally parallel assumption for the resolvent analysis, and compare the first resolvent
mode with the first SPOD mode for some specific stations x/c of the airfoil by taking a
“slice” from the boundary-layer weighted SPOD results obtained for the entire field in
Figures 6 and 7.

4.2.1. The base flow

The base flow ū(y) was extracted from the simulations considering time and spanwise
average. For the NACA 0012 airfoil we analysed three different stations: x/c = 0.6, 0.7
and 0.9. For the NACA 4412 airfoil we also analysed three different stations: x/c = 0.6
and 0.9 for the suction side (top) and 0.5 on the pressure side (bottom), since this airfoil
has no symmetry with respect to y = 0. These choices allow consideration of boundary-
layer profiles with various degrees of adverse pressure gradient, which is stronger close
to the trailing edge, particularly for the NACA 4412 profile, and close to zero for the
pressure side of the same airfoil. For all stations considered here, in both airfoils, the
boundary-layer profiles are turbulent. The mean velocity profiles at some x/c positions
for both airfoils are show in Figure 8 scaled in outer units (wing chord c and freestream
velocity U∞ ). Figure 9 shows the mean velocity profiles for the same stations in inner
units, denoted with a + superscript; quantities are made non-dimensional using the
friction velocity uτ and the kinematic viscosity ν. The mean velocity profiles in Figure
9 show the expected shape of wall-bounded turbulent flows when plotted in inner scale
as a function of wall distance y + . Adverse pressure gradients lead to departures of the
12

(a) Streamwise velocity component u

(b) Wall-normal velocity component v

Figure 5. First SPOD mode at St ≈ 7 for the NACA 4412 airfoil.

typical zero-pressure-gradient profiles, as evident for the NACA 4412 results (see also
Vinuesa et al. (2016)).

4.2.2. Resolvent parameters

Resolvent analysis was performed for St = 7 for comparison with the SPOD results of
§ 4.1. The streamwise wavenumber kx is also a parameter of resolvent analysis, and is
not known a priori, since different wavelengths may appear in the forcing. To study the
effect of kx in the resolvent analysis, we first varied kx and studied the leading gains for
the three analysed stations of each airfoil. These results are show in Figure 10 for the
respective stations analysed in the NACA 0012 airfoil, and in Figure 11 for the NACA
4412.
We can notice a sharp peak in all results, which corresponds to disturbances outside
the boundary layer, where the phase velocity is equal to the free stream velocity. Analysis
of forcing and response modes for these wavenumbers shows that they correspond to free-
stream modes, similar to the observations of Dergham et al. (2013) for a backward-facing
step. We can also notice smooth peaks in all stations at the first gain σ1 . These peaks
correspond to disturbances inside the boundary layer, where the phase velocity is lower
than the free stream velocity. Moreover, for the smooth peaks there is a clear dominance
of the optimal forcing, which leads to a response with a gain that is at least one order
of magnitude higher than the suboptimals. To perform the resolvent analysis we thus
used the kx corresponding to the larger smooth peaks in each analysed section for both
airfoils.
 13

(a) Streamwise velocity component u

(b) Wall-normal velocity component v

Figure 6. First boundary-layer weigthed SPOD mode at St ≈ 7 for the NACA 0012 airfoil.

4.2.3. Comparison between SPOD and resolvent modes

Results of the comparison between first boundary-layer weighted SPOD mode and
optimal response from resolvent analysis are show in Figures 12 and 13 for the NACA
0012 airfoil for horizontal and vertical velocity components, respectively; and in Figures
14 and 15 for the NACA 4412 airfoil also for horizontal and vertical velocity components,
respectively. As commented before, the boundary-layer weighted SPOD modes were taken
to minimise statistical noise.
Since we deal with a turbulent boundary layer, an exact match is not expected, but the
comparisons presented here show that resolvent analysis provides a good reduced-order
model for the two-dimensional disturbances considered here, which play an important
role in the radiation of trailing-edge noise (Sano et al. 2019). There is close agreement
with streamwise velocity fluctuations of the leading SPOD mode, which has an amplitude
distribution similar to that of Tollmien-Schlichting waves, with a phase jump at a position
with vanishing amplitude, also visible in the SPOD fields in Figures 4-7. The relatively
low amplitude of the two-dimensional disturbances within the boundary layer tends to
ease the application of linearised models.
For both airfoils, for positions closer to the trailing edge (x/c = 0.9) the agreement
deteriorates, but the general features of the SPOD mode are captured by the optimal
response from resolvent analysis. The worse agreement at downstream positions may be
due to the proximity of the trailing edge, which leads to a stronger inhomogeneity in x,
making the locally-parallel assumption invalid.
Resolvent analysis also shows consistent trends when the two airfoils are compared.
The comparison of Figures 8 and 9 shows that resolvent gains are larger for boundary-
layer stations with adverse pressure gradient. This is observed in the SPOD modes of
14

(a) Streamwise velocity component u

(b) Wall-normal velocity component v

Figure 7. First boundary-layer weigthed SPOD mode at St ≈ 7 for the NACA 4412 airfoil.
0.2 0.2 0.2

0.15 0.15 0.15

x/c=0.4 x/c=0.4 x/c=0.4
x/c=0.5 x/c=0.5 x/c=0.5
x/c=0.6 x/c=0.6 x/c=0.6
0.1 0.1 0.1
x/c=0.7 x/c=0.7 x/c=0.7
x/c=0.8 x/c=0.8 x/c=0.8
x/c=0.9 x/c=0.9 x/c=0.9
0.05 0.05 0.05

0 0 0
0 0.5 1 1.5 0 0.5 1 1.5 0 0.5 1 1.5

(a) (b) (c)

Figure 8. Mean velocity profiles at different positions x/c for (left) NACA 0012 and NACA
4412 on (centre) suction and (right) pressure sides, scaled in outer units.
30 50 25

25 40 20
20
30 15
15
20 10
10

5 linear law
10 linear law
5 linear law
log law log law log law
0 0 0
100 102 104 100 102 104 100 102 104

(a) (b) (c)

Figure 9. Mean velocity profiles at different positions x/c for (left) NACA 0012 and NACA
4412 on (centre) suction and (right) pressure sides, scaled in inner units. In all figures, the
blue dashed line represents the linear law, ū+ = y + , and the red dashed line is the log law,
ū+ = (1/κ) ln(y + ) + B, where κ = 0.41 and B = 5.2.
 15
1 1 1
10 10 10

100 100 100

10-1 10-1 10-1

40 45 50 55 60 40 45 50 55 60 40 45 50 55 60

(a) x/c = 0.6 (b) x/c = 0.7 (c) x/c = 0.9

Figure 10. First three resolvent gains as a function of the streamwise wavenumber at St = 7
for the NACA 0012 airfoil.

101 101 101

100 100 100

10-1 10-1 10-1

30 40 50 60 30 40 50 60 30 40 50 60

(a) x/c = 0.5 (bottom) (b) x/c = 0.6 (top) (c) x/c = 0.9 (top)

Figure 11. First three resolvent gains as a function of the streamwise wavenumber at St = 7
for the NACA 4412 airfoil.
0.1 0.1 0.1

0.08 0.08 0.08

0.06 0.06 0.06

0.04 0.04 0.04

0.02 0.02 0.02

0 0 0
10-10 10-5 10-10 10-5 10-10 10-5

(a) x/c = 0.6 (b) x/c = 0.7 (c) x/c = 0.9

Figure 12. First SPOD mode (red dashed line) compared with the optimal response from
resolvent analysis (black solid line) of the streamwise velocity component u, for the NACA 0012
airfoil, at stations x/c = 0.6, 0.7 and 0.9.

Figures 4-7, which display larger amplitudes at the suction side of the NACA 4412 airfoil,
particularly close to the trailing edge. The NACA 0012 airfoil fluctuation amplitudes are
less significant, and the lowest levels are seen at the pressure side of the NACA 4412
profile. These trends are also observed in the resolvent gains.

5. Domain-size effect on wavepackets

We have found two-dimensional coherent structures on the airfoil surfaces using SPOD,
which were modelled using resolvent analysis. However, the use of spanwise-periodic
boundary conditions may be a concern, as it may lead to artificially coherent structures.
This has been observed in turbulent Couette flow by Tsukahara et al. (2006) and more
recently by Lee & Moser (2018); in this flow, computational boxes with short streamwise
lengths led to artificially streamwise coherent structures that remained straight through-
out the domain, and the use of larger boxes creates a meandering behaviour. We here
16
0.1 0.1 0.1

0.08 0.08 0.08

0.06 0.06 0.06

0.04 0.04 0.04

0.02 0.02 0.02

0 0 0
10-10 10-5 10-10 10-5 10-10 10-5

(a) x/c = 0.6 (b) x/c = 0.7 (c) x/c = 0.9

Figure 13. First SPOD mode (red dashed line) compared with the optimal response from
resolvent analysis (black solid line) of the wall-normal velocity component v, for the NACA
0012 airfoil, at stations x/c = 0.6, 0.7 and 0.9.
0.1 0.1 0.1

0.08 0.08 0.08

0.06 0.06 0.06

0.04 0.04 0.04

0.02 0.02 0.02

0 0 0
10-10 10-5 10-10 10-5 10-10 10-5

(a) x/c = 0.5 (bottom) (b) x/c = 0.6 (top) (c) x/c = 0.9 (top)

Figure 14. First SPOD mode (red dashed line) compared with the optimal response from
resolvent analysis (black solid line) of the streamwise velocity component u, for the NACA 4412
airfoil, at stations x/c = 0.5 in the pressure side (bottom), x/c = 0.6 and 0.9 in the suction side
(top).
0.1 0.1 0.1

0.08 0.08 0.08

0.06 0.06 0.06

0.04 0.04 0.04

0.02 0.02 0.02

0 0 0
10-10 10-5 10-10 10-5 10-10 10-5

(a) x/c = 0.5 (bottom); w1 (b) x/c = 0.6 (top); w1 (c) x/c = 0.9 (top); w0

Figure 15. First SPOD mode (red dashed line) compared with the optimal response from
resolvent analysis (black solid line) of the wall-normal velocity component v, for the NACA
4412 airfoil, at stations x/c = 0.5 in the pressure side (bottom), x/c = 0.6 and 0.9 in the suction
side (top).

investigate this issue using simulations of the flow over a flat plate, forming a ZPGTBL,
with computational domains of different spanwise lengths.

5.1. Turbulence statistics and spectral analysis

Using standard Reynolds decomposition u = ū + u0 , the mean horizontal velocity
+
profiles ū+ and the variance profiles of the horizontal velocity fluctuations (ū0 )2 in inner
∗
scaling are shown in Figures 16 (a) and (b) respectively, for the position x/δ = 250. The
mean velocity profiles for all the three simulations collapse as expected, and show the
typical shape of wall-bounded turbulent flows when plotted as a function of wall distance
 17
15
25

20
10
15
L z = 32 *
10 L z = 64 * 5
L z = 128 *
5
linear law
log law
0 0
100 101 102 103 100 101 102 103

(a) (b)

Figure 16. (a) The mean flow and (b) horizontal velocity fluctuations at x/δ ∗ = 250 scaled in
inner units. In both figures, the blue dashed line represents the linear law, ū+ = y + ; and the
red dashed line is the log law, ū+ = (1/κ) ln(y + ) + B, where κ = 0.41 and B = 5.2.

y + , following linear and log laws. Variance profiles also collapse for the three simulations,
and have the expected pattern characteristic of wall-bounded turbulent flows, with a
near-wall peak in the buffer layer at y + ≈ 15 (Eitel-Amor et al. 2014).
The inner-scaled premultiplied power-spectral density of horizontal velocity fluctua-
tions kz+ Euu+
as a function of spanwise wavelength λ+ +
z = 2π/kz is shown in Figure 17
∗
(a) at x/δ = 250 for all the three simulations. Similarly to the mean value and the
variance, the pre-multiplied spectra for the three simulations collapse, showing further
indication of independence of domain size. Results in Figure 17 (a) show a peak in
the spanwise wavelength corresponding to λ+ z ≈ 100, which is representative of the
signature of the near-wall cycle. This has been documented in many studies across a
range of Reynolds numbers and flow types, due to the near-wall cycle of streaks and
quasi-streamwise vortices (Jiménez 2013). The same inner-scaled power-spectral density
+
of horizontal velocity fluctuations, without premultiplication Euu , is shown in Figure 17
∗
(b) at the same position, x/δ = 250, for all the three simulations. The results in Figure
17 (b) show that the power-spectral density contours do not close when premultiplication
is not considered, even with the increasing spanwise length Lz , indicating that there is
non-zero power spectral density for very large values of λ+ z . In order to also show the
wavenumber kz+ = 0, which corresponds to λ+ z → ∞ and consequently does not appear in
the plot with the horizontal axis in log scale, the spanwise power-spectral density, without
premultiplication, is shown in Figure 17 (c). Results in Figure 17 (c) show clearly that
there are structures in kz+ = 0, which correspond to two-dimensional perturbations in
the computational domain. This suggests that this corresponds to the wavepackets we
found around the airfoils for kz+ = 0. The wavenumber kz+ → 0 corresponds to spanwise-
coherent velocity fluctuations, which have not been studied in detail in the literature
and are the main focus of this study. This limit is thus independent of domain size for
the computations shown here. As discussed by Lozano-Durán & Jiménez (2014), two-
dimensional fluctuations in the simulation, for kz+ = 0, represent disturbances whose
spanwise length scale is larger than the size of the computational box; they should be
thus understood as fluctuations with λz  δ, which are responsible for the plateau of
power-spectral density shown in Figure 17 (b).
18
L z = 32 *
2 L z = 64 * 102
10 102
L z = 128 *

101 101 101

100 100 100

101 102 103 101 102 103 0 0.005 0.01 0.015 0.02

(a) (b) (c)

Figure 17. Spanwise power-spectral density of the streamwise velocity fluctuation u0 at

x/δ ∗ = 250 for the three ZPGTBL simulations: (a) with premultiplication kz+ Euu+
and x-axis as
λ+
z shown in log scale; (b) without premultiplication E +
uu and λ+
z shown in log scale; (c) without
+
premultiplication Euu and kz+ shown in linear scale.

5.2. SPOD results

As discussed previously, in order to compare the results obtained with the ZPGTBL
simulations with those from the airfoils, we normalized the parameters using δ ∗ taken
at 75% of the NACA 0012 chord. Thus, for the NACA 0012 airfoil case at x/c = 0.75
we have Reδ∗ ≈ 1400, which was considered as the region of interest in the ZPGTBL
simulation; moreover, the frequency used to perform the SPOD in the airfoils, St ≈ 7,
corresponds to Stδ∗ ≈ 0.025. We will then perform the SPOD in the ZPGTBL databases
considering Stδ∗ ≈ 0.025, with displacement thickness taken in the middle of the region
of interest, defined below.
Similarly to the analysis of the airfoil databases, we have taken a spanwise average of
flow fluctuations before performing the SPOD, to focus our analysis on spanwise-coherent
structures. SPOD for the three ZPGTBL databases was performed in a fully turbulent
region delimited by 170 6 x/δ ∗ 6 270. The focus on the region of interest ensures that
only fluctuations in the developed turbulent region are considered, with minimal influence
from the upstream tripping. Furthermore, the size of the region ensures that disturbances
are able to travel across the entire region within the period of each block.
Figure 18 shows the SPOD eigenvalues for the three ZPGTBL simulations, which have
nearly no change among the three domain sizes. We thus conclude that the coherent
structures in all cases have nearly the same amplitude. Figure 19 shows the first SPOD
mode for horizontal and vertical velocities components, for all the three ZPGTBL
simulations. We can observe the presence of two-dimensional coherent waves for all the
cases, even with the increase of the spanwise length of the domain Lz . The SPOD results
for the ZPGTBL databases in Figure 19 display nearly the same mode shape for the three
considered domain sizes, and the structures are similar to those around the airfoils, shown
in Figures 6 and 7. The present SPOD results confirm that the wavepackets found in the
present study, corresponding to spanwise-coherent disturbances in turbulent boundary
layers, have a physical meaning and are not just an artifact from the simulations.

5.3. SPOD and resolvent comparison

Figure 20 (a) and (b) shows the first SPOD mode compared with the optimal response
from resolvent analysis for the horizontal and vertical velocity components, respectively,
for the three analysed ZPGTBL simulations at x/δ ∗ = 230. The frequency used to
compute the resolvent analysis here was Stδ∗ = 0.025, and the horizontal wavenumber
kx = 0.205, chosen with the same criterion as in the airfoil case (see Section 4.2.2). There
is a close agreement between the SPOD modes and the optimal response from resolvent
 19
10 -1
Lz=32 *
Lz=64 *
-2 Lz=128 *
10

10 -3

10 -4
0 5 10 15 20 25 30

Figure 18. SPOD eigenvalues for the three ZPGTBL simulations.

(a) Lz = 32δ ∗ ; u (b) Lz = 64δ ∗ ; u (c) Lz = 128δ ∗ ; u

(d) Lz = 32δ ∗ ; v (e) Lz = 64δ ∗ ; v (f) Lz = 128δ ∗ ; v

Figure 19. First SPOD mode of horizontal and vertical velocities components for the three
ZPGTBL simulations at Stδ∗ = 0.025.

30 30
resolvent
*
25 L z=32 25
*
L z=64
20 20
*
L z=128
15 15

10 10

5 5

0 0
-8 -6 -4 -8 -6 -4
10 10 10 10 10 10

(a) horizontal velocity component (b) wall-normal velocity component

Figure 20. First SPOD mode (red lines) compared with the optimal response from resolvent
analysis (black lines) for the three analysed ZPGTBL simulations at station x/δ ∗ = 230 and
Stδ∗ = 0.025.

analysis. For the present simulation, the absence of a trailing edge ensures that the mean
flow is only slowly developing, and the use of the locally-parallel resolvent analysis is
more clearly justified. The observed agreement further confirms that spanwise-coherent
structures in turbulent boundary layers can be modelled as the most amplified flow
response, obtained using the linearised Navier–Stokes operator.
20
6. Conclusions
In the current work we have used SPOD to extract the most energetic spanwise
coherent structures from the turbulent flow around two different airfoils: NACA 0012
with zero angle of attack and NACA 4412 with an angle of attack AoA = 5◦ , both for
chord Reynolds number Rec = 400, 000. By doing so, we achieved a proper separation
of spatially and temporally coherent structures, which are either hidden in stochastic
turbulent fluctuations or spread over a wide frequency range. We found in the leading
SPOD mode, which is statistically converged in our database, a two-dimensional coherent
wave, or wavepacket, that propagates all the way from the transition region to the trailing
edge. The study by Sano et al. (2019), shows that such wavepackets are a dominant non-
compact source in trailing edge noise. Thus, although not much studied in the literature
of wall-bounded turbulence, such spanwise-coherent structures are likely relevant for
aeroacoustics (Amiet 1976; Nogueira et al. 2017).
The observed waveapckets in the leading SPOD modes were modelled using locally
parallel resolvent analysis, where turbulent mean profiles from specific stations x/c in
both airfoils were used as base flows. Results of comparisons for all the cases presented
here for both airfoils show a fair agreement between the optimal flow response and the
leading SPOD mode. Moreover, the optimal response has a gain which is at least an order
of magnitude higher than the suboptimals, which explains its presence in the data series
for the turbulent boundary layers. The results highlight resolvent analysis as a pertinent
reduced-order model for the relevant fluctuations, leading to predictions of the structure
of advecting waves reaching the edge.
In order to show the domain size independence and ensure the physical relevance of
such wavepackets found around the airfoils, we run three simulations of a ZPGTBL
for three different spanwise domain lengths: Lz = 32δ ∗ , Lz = 64δ ∗ and Lz = 128δ ∗ .
A spectral analysis of the flow, considering a power-spectral density of the horizontal
velocity fluctuations without the usual premultiplication adopted in wall turbulence,
showed the presence of two-dimensional structures with kz+ = 0, which thus have no
signature in standard pre-multiplied spectra. SPOD results also showed the presence of
such wavepackets for all three computational domains, with nearly the same amplitude.
Thus, our results indicate that such wavepackets have physical meaning and are not
just an artifact from the simulation. They should be understood as spanwise-coherent
disturbances, the characteristic length of which is much larger than the boundary-layer
thickness, which appear as two-dimensional disturbances in numerical simulations; in this
sense they are analogous to very long streaky structures, which appear as infinitely long
in the streamwise direction for finite domains (Lozano-Durán & Jiménez 2014).
In summary, the present work reveals that among the dominant features in the
turbulent boundary layers around the airfoil, such as streaky structures with the usual
spanwise spacing of 100 viscous units, there are coherent, low-amplitude two-dimensional
hydrodynamic wavepackets, with peak at St ≈ 7, i.e. the same frequency found to be
correlated to the acoustic radiation in Sano et al. (2019). The structures studied in this
work are not the standard streaky structures of turbulent boundary layers; these are
instead low-energy, coherent spanwise-elongated waves reminiscent of those identified
by Hussain & Reynolds (1972) in their early work on forced oscillations in turbulent
channel flow. In this paper we have shown the presence of such wavepackets in various
wall-bounded turbulent flows, and these are appropriately modelled as the optimal flow
response from resolvent analysis. Further numerical and experimental work aiming at the
characterisation of such structures is promising.
 21
Acknowledgments
The authors acknowledge the financial support received from Conselho Nacional de
Desenvolvimento Cientı́fico e Tecnológico, CNPq, under grant No. 310523/2017-6, and by
CAPES through the PROEX program. Financial support provided by the Knut and Alice
Wallenberg Foundation is also gratefully acknowledged. We also acknowledge funding
from CISB. The simulations were performed on resources provided by the Swedish
National Infrastructure for Computing (SNIC) at the PDC Center for High Performance
Computing at KTH, Stockholm.

REFERENCES
Abreu, L.I., Cavalieri, A.V.G., Schlatter, P., Vinuesa, R. & Henningson, D. 2019
Reduced-order models to analyse coherent structures in turbulent pipe flow. TSFP11
paper .
Alfonsi, G., Primavera, L., Passoni, G. & Restano, C. 2001 Proper orthogonal
decomposition of turbulent channel flow. In Computational Fluid Dynamics 2000 , 1st
edn., , vol. 1, pp. 473–478. Berlin, Heidelberg: Springer Berlin Heidelberg.
Amiet, R. K. 1976 Noise due to turbulent flow past a trailing edge. Journal of Sound and
Vibration 47 (3), 387–393.
Araya, Daniel B, Colonius, Tim & Dabiri, John O 2017 Transition to bluff-body dynamics
in the wake of vertical-axis wind turbines. Journal of Fluid Mechanics 813, 346–381.
Aubry, N., Holmes, P., Lumley, J. L. & Stone, E. 1988 The dynamics of coherent structures
in the wall region of a turbulent boundary layer. Journal of Fluid Mechanics 192, 115–173.
Beneddine, S., Sipp, D., Arnault, A., Dandois, J. & Lesshafft, L. 2016 Conditions for
validity of mean flow stability analysis. Journal of Fluid Mechanics 798, 485–504.
Berkooz, G., Holmes, P. & Lumley, J. L. 1993 The proper orthogonal decomposition in the
analysis of turbulent flows. Annual review of fluid mechanics 25 (1), 539–575.
Cavalieri, A.V.G., Jordan, P. & Lesshafft, L. 2019 Wave-packet models for jet dynamics
and sound radiation. Applied Mechanics Reviews 71 (2), 020802.
Cavalieri, A. V. G., Rodrı́guez, D., Jordan, P., Colonius, T. & Gervais, Y. 2013
Wavepackets in the velocity field of turbulent jets. Journal of Fluid Mechanics 730, 559–
592.
Del Alamo, J.C. & Jimenez, J. 2006 Linear energy amplification in turbulent channels.
Journal of Fluid Mechanics 559, 205–213.
Dergham, G, Sipp, D & Robinet, J-Ch 2013 Stochastic dynamics and model reduction of
amplifier flows: the backward facing step flow. Journal of Fluid Mechanics 719, 406–430.
Eitel-Amor, Georg, Örlü, Ramis & Schlatter, Philipp 2014 Simulation and validation
of a spatially evolving turbulent boundary layer up to Reθ = 8300. International Journal
of Heat and Fluid Flow 47, 57–69.
Ffowcs Williams, J. E. & Hall, L. H. 1970 Aerodynamic sound generation by turbulent flow
in the vicinity of a scattering half plane. Journal of Fluid Mechanics 40 (04), 657–670.
Fischer, P.F., Lottes, J.W. & Kerkemeier, S.G. 2008 Nek5000: Open source spectral
element CFD solver. available at: http://nek5000.mcs.anl.gov.
Freund, J. B. & Colonius, T. 2009 Turbulence and sound-field pod analysis of a turbulent
jet. International journal of aeroacoustics 8 (4), 337–354.
Hosseini, S.M., Vinuesa, R., Schlatter, P., Hanifi, A. & Henningson, D.S. 2016 Direct
numerical simulation of the flow around a wing section at moderate Reynolds number.
International Journal of Heat and Fluid Flow 61, 117–128.
Hussain, AKMF & Reynolds, WC 1972 The mechanics of an organized wave in turbulent
shear flow. part 2. experimental results. Journal of Fluid Mechanics 54 (2), 241–261.
Hutchins, N. & Marusic, I. 2007 Evidence of very long meandering features in the logarithmic
region of turbulent boundary layers. Journal of Fluid Mechanics 579, 1–28.
Jeun, Jinah, Nichols, Joseph W & Jovanović, Mihailo R 2016 Input-output analysis of
high-speed axisymmetric isothermal jet noise. Physics of Fluids 28 (4), 047101.
Jiménez, J. 2013 Near-wall turbulence. Physics of Fluids 25 (10), 101302.
22
Jovanovic, M.R. & Bamieh, B. 2005 Componentwise energy amplification in channel flows.
Journal of Fluid Mechanics 534, 145–183.
Kaplan, O., Jordan, P., Cavalieri, A.V.G. & Brés, G.A. 2019 The nozzle dynamics driving
wavepackets in turbulent jets. AIAA paper p. 2019.
Lee, Myoungkyu & Moser, Robert D 2018 Extreme-scale motions in turbulent plane
Couette flows. Journal of Fluid Mechanics 842, 128–145.
Lesshafft, L., Semeraro, O., Jaunet, V., Cavalieri, A.V.G. & Jordan, P. 2019
Resolvent-based modelling of coherent wavepackets in a turbulent jet. arXiv preprint
arXiv:1810.09340 .
Lozano-Durán, Adrián & Jiménez, Javier 2014 Effect of the computational domain on
direct simulations of turbulent channels up to Reτ = 4200. Physics of Fluids 26 (1),
011702.
Lumley, J. L. 2007 Stochastic tools in turbulence. Courier Corporation.
McKeon, B.J. 2017 The engine behind (wall) turbulence: perspectives on scale interactions.
Journal of Fluid Mechanics 817.
McKeon, B. J. & Sharma, A. S. 2010 A critical-layer framework for turbulent pipe flow.
Journal of Fluid Mechanics 658, 336–382.
Noack, B. R., Afanasiev, K., Morzynski, M., Tadmor, G. & Thiele, F. 2003 A hierarchy
of low-dimensional models for the transient and post-transient cylinder wake. Journal of
Fluid Mechanics 497, 335–363.
Nogueira, P. A. S., Cavalieri, A. V. G. & Jordan, P. 2017 A model problem for sound
radiation by an installed jet. Journal of Sound and Vibration 391, 95–115.
Picard, C. & Delville, J. 2000 Pressure velocity coupling in a subsonic round jet.
International Journal of Heat and Fluid Flow 21 (3), 359–364.
Sano, A, Abreu, LI, Cavalieri, AVG & Wolf, WR 2019 Trailing-edge noise from the
scattering of spanwise-coherent structures. Physical Review Fluids .
Schlatter, P., Li, Q., Örlü, R., Hussain, F. & Henningson, D. S. 2014 On the near-
wall vortical structures at moderate Reynolds numbers. European Journal of Mechanics-
B/Fluids 48, 75–93.
Schlatter, Philipp & Örlü, Ramis 2012 Turbulent boundary layers at moderate reynolds
numbers: inflow length and tripping effects. Journal of Fluid Mechanics 710, 5–34.
Schmid, P.J. & Henningson, D.S. 2012 Stability and transition in shear flows, , vol. 142.
Springer Science & Business Media.
Sinha, A., Rodrı́guez, D., Brès, G. A. & Colonius, T. 2014 Wavepacket models for
supersonic jet noise. Journal of Fluid Mechanics 742, 71–95.
Sirovich, Lawrence 1987 Turbulence and the dynamics of coherent structures. i. coherent
structures. Quarterly of applied mathematics 45 (3), 561–571.
Tanarro, A, Vinuesa, R & Schlatter, P 2018 History effects for cambered and symmetric
wing profiles. arXiv preprint arXiv:1812.03754 .
Towne, Aaron, Lozano-Durán, Adrian & Yang, Xiang 2019 Resolvent-based estimation
of space-time flow statistics. arXiv preprint arXiv:1901.07478 .
Towne, A., Schmidt, O.T. & Colonius, T. 2018 Spectral proper orthogonal decomposition
and its relationship to dynamic mode decomposition and resolvent analysis. Journal of
Fluid Mechanics 847, 821867.
Trefethen, L. N. 2000 Spectral methods in MATLAB , , vol. 10. Siam.
Tsukahara, Takahiro, Kawamura, Hiroshi & Shingai, Kenji 2006 Dns of turbulent couette
flow with emphasis on the large-scale structure in the core region. Journal of Turbulence
(7), N19.
Vinuesa, Ricardo, Bobke, Alexandra, Örlü, Ramis & Schlatter, Philipp 2016 On
determining characteristic length scales in pressure-gradient turbulent boundary layers.
Physics of fluids 28 (5), 055101.
Wei, M. & Freund, J. B. 2006 A noise-controlled free shear flow. Journal of Fluid Mechanics
546, 123–152.