An empirical model of noise sources in subsonic jets, formulated

in a linear resolvent framework

Ugur Karban∗
Middle East Technical University, 06800, Ankara, Turkey

Benjamin Bugeat † and Anurag Agarwal‡

University of Cambridge, Cambridge, CB2 3AP, United Kingdom

Lutz Lesshafft§
CNRS - École Polytechnique, 91128, Palaiseau Cedex, France

Peter Jordan¶
Institut Pprime, CNRS-Université de Poitiers-ENSMA, 86962 Chasseneuil du Poitou, France

We present an analysis aiming to identify the acoustically efficient forcing components in

subsonic jets using an LES database. The forcing terms, which correspond to the source terms
in the classical terminology of acoustic analogies, are defined within the resolvent framework.
We extract from a jet database at Mach number, 𝑀 = 0.4, which contains both state and forcing
data, a low-rank representation of the acoustic field and a low-rank forcing that is correlated
to it. We show that the forcing components that generate 99% of the acoustic energy in the
downstream region correspond to ∼3% of the total forcing energy. In the final paper, we will
further analyse this low-rank forcing to identify the parts that actually generate the downstream
noise We aim also to present a model for these identified sources of subsonic jet noise.

I. Introduction
et noise is one of the most studied subjects in aeroacoustics since the first attempt by Lighthill [1]. Understanding
J the true mechanisms behind jet mixing noise still remains a challenge and motivates many ongoing research activities,
although thanks to 70 years of research, it is now possible to distinguish between certain mechanisms that generate
sound at sideline or at low angles, also at subsonic and supersonic jets. Some review studies that discuss these different
mechanisms can be listed as Tam [2], Jordan and Colonius [3], Brès and Lele [4] and Cavalieri et al. [5].
The nonlinear nature of sound generation by turbulent flows prevents separating the mechanisms that generate
hydrodynamic fluctuations from those that generate waves propagating waves outside the turbulent region, and thus
radiate sound. A common practice that takes place in many approaches to identify this mechanism is to split sound
generation and radiation, yielding the so-called acoustic analogies. There exist a number of analogies in the literature
(Lilley [6], Howe [7], Doak [8], Goldstein [9]; etc.) succeeding that of Lighthill’s [1]. A common property of the
acoustic analogies is that they are exact in the sense that if one has access to the source components with very good
accuracy, they yield the exact response. This was shown in several numerical studies using direct numerical simulation
(DNS) [10–12] or large eddy simulation (LES) [13–15]. The same property holds also for the resolvent analysis, which
is obtained by linearisation of the Navier-Stokes (N-S) equation around the mean flow and reorganisation in input-output
form in the frequency domain [16, 17]. Such an exactitude, however, is not very useful by itself in terms of modeling the
mechanisms behind response generation. Considering noise generation in turbulent flow for instance, a vast amount of
the forcing does not radiate sound, or is not acoustically efficient. Therefore, it is critical to devise source identification
strategies in turbulent flows.
Using a linear operator that takes into account all the linear amplification and scattering mechanisms can be crucial
to simplify the source terms. It is now known that for supersonic jets, the main contribution to jet noise comes from
∗ Post-doctoralresearch fellow, Department of Aerospace Engineering, Universiteler Mh., Dumlupinar Blv. 1/6 D:133
† Post-doctoralresearch fellow, Department of Engineering, Trumpingtion Street
‡ Professor, Department of Engineering, Trumpington Street
§ Research Director, Laboratoire d’Hydrodynamique.
¶ Research Director, Départment Fluides Thermique et Combustion, 11 Boulevard Marie et Pierre Curie.

1
a Mach-wave-type propagation associated to the instability mechanism contained in the shear layer [18–21]. Such
linear models, on the other hand, do not provide predictions of similar accuracy for subsonic jets, although a similar
Mach-wave-type propagation is also observed in subsonic jets at low angles, i.e. towards the downstream region of the
jet [22]. This issue was shown to be associated to not including in these linear models the effect of the jitter mechanism
resulting in coherence decay observed in wavepackets in subsonic jets [23, 24]. A source model to predict subsonic jet
noise at different operating conditions taking into account these nonlinear jitter effect is yet to be developed.
In this study, we focus on the axisymmetric sound generation at low propagation angles which is known to be
dominant part of the overall noise generated by isolated subsonic jets [3, 22]. We present a methodology which combines
a number of tools to provide a low-rank, dynamics-based representation of turbulent jet noise. Resolvent framework is
adopted to achieve the split between the propagator and the source. We consider that although the instability waves
appearing as coherent structures in the jet near field are not sufficient by themselves to explain noise generation in
subsonic jets, the linear mechanisms embedded in the resolvent operator are still relevant and can provide an effective
modeling platform when accompanied by an appropriate source model.
We use the spectral proper orthogonal decomposition (SPOD) method [25–28] to obtain a low-rank description of jet
noise at low angle, and resolvent-based extended SPOD (RESPOD) [29] to extract the associated low-rank higher-order
terms, i.e. forcing. The resolvent framework connects low-rank source terms obtained by the RESPOD method to the
low-rank noise measured at the acoustic field via the resolvent operator. Up to this point, the analysis is systematic, i.e.,
can be applied to any flow with any particular response in focus. An implementation on wall-bounded flows was shown
in Karban et al. [29].
In the final paper, with further inspection of these low-rank sources in terms of the acoustic matching criterion
[5, 30, 31], we aim to identify the parts that actively generate the sound at low angles. Following this identification
process, we aim to present an empirical model for this low-rank active source, based on tests performed using the
resolvent framework. The empirical model will take into account the effects of frequency, the jet Mach number and the
flight stream.
The structure of the remaining parts is as follows: the mathematical framework for resolvent analysis is revisited in
§II. The details about the numerical database and the tool to perform resolvent analysis are given in §III. The process to
identify the low-rank forcing components that generate substantial amount of the downstream jet noise is explained in
§IV. Finally, some concluding remarks are provided in §V.

II. Governing equations in resolvent form

The compressible Navier-Stokes (N-S) equations are given in a compact form as,

𝜕𝑡 q = N (q), (1)

where q = [𝜈 𝑢 𝑥 𝑢𝑟 𝑢 𝜃 𝑝] ⊤ is the state vector with 𝜈 and 𝑝 being the specific volume and the pressure, respectively,
and u = [𝑢 𝑥 𝑢𝑟 𝑢 𝜃 ] ⊤ being the velocity vector given in cylindrical coordinates in the order of streamwise, radial and
azimuthal directions, and N denotes the nonlinear N-S operator. All variables are properly non-dimensionalised by
ambient speed of sound, 𝑐 ∞ , density, 𝜌∞ , and the nozzle diameter, 𝐷. We consider a discretisatized system in space, for
which linearisation around the mean, q, yields

𝜕𝑡 q′ − Aq′ = f, (2)

where A = 𝜕𝑞 N | q is the linear operator obtained from the Jacobian of N and f denotes all the remaining nonlinear
terms, i.e., the forcing vector. In resolvent framework, Eq. (2) is Fourier transformed and rearranged to obtain

q̂ = Rf̂, (3)

where the hat indicates Fourier transformed quantity and R = (𝑖𝜔I − A) −1 denotes the resolvent operator. The resolvent
operator can be modified to limit the response to certain measurements, which can be defined as any linear transformation
of the state vector, as;

ŷ = Cq̂, (4)
ŷ = R̃f̂, (5)

2
where C denotes the measurement matrix and R̃ ≜ CR is the modified resolvent operator. Throughout this paper, we
will focus on the pressure in the acoustic field as our measured quantity, therefore

C = [0 0 0 0 𝛼], (6)

where 𝛼 = 0 in the near-field with non-negligible hydrodynamic fluctuations, and 1 in the acoustic field.
It is also possible to limit forcing by introducing a control matrix, B on f̂ in Eq. (3), yielding

ŷ 𝐵 = R̃Bf̂. (7)

Note that depending on the measured quantity, ŷ, and the control matrix, B, the condition,

ŷ 𝐵 = ŷ (8)

may or may not be satisfied. We will test different B matrices to identify irrelevant forcing components whose absence
do not cause violation of Eq. (8) at least to a certain extent.

III. Numerical database and resolvent analysis tool

The numerical database used in this study consists of large eddy simulation (LES) of a subsonic, isothermal, ideally
expanded jet at jet Mach number, 𝑀 𝑗 ≜ 𝑈 𝑗 /𝑐 𝑗 = 0.4 using the unstructured flow solver ‘Charles’ [32]. The database
was validated against experimental data in Brès et al. [33], where detailed information about the database can also be
found. In the final paper, three other LES datasets will be used to investigate Mach number and flight stream effects on
noise generation and forcing modeling.
Resolvent-based prediction of the response using the forcing data is achieved via a custom resolvent analysis code
[34]. The code uses finite-volume method to solve linearised N-S equations decomposed into azimuthal Fourier modes.
Note that since the mean flow in jets is axisymmetric, Eq. (3) can be written for a given azimuthal mode, 𝑚, as,

q̂ (𝑚) = Rf̂ (𝑚) . (9)

As mentioned earlier, the scope of the study is limited to noise generation in the first azimuthal mode, 𝑚 = 0. The
response given in Eq. (9) is computed using the inverted system,

Lq̂ (𝑚) = f̂ (𝑚) , (10)

where L is the sparse linear operator satisfying L−1 = R. The resolvent code solves the linear system of equations given
in Eq. (10) via LU decomposition, using PETSc library [35]. Further details can be found in Bugeat et al. [34].

IV. Identification of the acoustically efficient forcing components

The final goal of this study is to provide a model for the forcing responsible for noise generation in subsonic jets. To
achieve this, in the following, we first present a methodology to identify these acoustically efficient forcing components,
using the LES database, which contains both state and forcing data. As mentioned earlier, the scope of the study is
limited to the first azimuthal mode, 𝑚 = 0. We also limit the study to mechanisms associated to noise generation at
low-propagation angles, which we refer to as downstream noise. As it was discussed in the literature [22] and will
also be demonstrated in the following, sidestream and downstream noise are generated by different mechanisms, and
therefore, separate analyses are required for modeling them. In what follows, we will present results for the frequency
range, 𝑆𝑡 = [0.4, 1.0].

A. Masking the forcing vector

We start forcing identification by applying spatial and componentwise masking in the forcing data to observe the
effect of the masked regions/components on noise generation. Given that the system is discretised, both masking
correspond to zeroing certain parts in the forcing vector using a control matrix, B. Figure 1 depicts comparison of the
effects of different masking in space and in variables on the PSD of the acoustic pressure at 𝑆𝑡 = 0.6. One can see that
masking the forcing in 𝑟 > 2𝐷 or 𝑟 > 4𝐷 yields nearly identical results in the entire flow domain. Masking in 𝑟 > 1.5𝐷
also yields nearly identical noise fields in the downstream region, 𝑥 > 6𝐷, while a slight discrepancy is observed in the

3
Fig. 1 PSD of pressure predicted using resolvent analysis with masking applied in space (top) and in variables
(bottom) in comparison to the LES data at 𝑟 = 5𝐷 and 𝑆𝑡 = 0.6.

region, 𝑥/𝐷 = [3, 6]. Given the convergence of the noise field, we focus on the forcing within 𝑟 < 2𝐷 for the rest of the
analysis.
Componentwise masking of the forcing shows that the variables, 𝑓 𝜈 and 𝑓 𝑝 , which are the forcing corresponding to
the specific volume and the pressure, respectively, have negligible effect on noise generation. Masking the variable, 𝑓𝑢𝑟 ,
which is the forcing corresponding to the radial velocity, causes significant reduction in the sidestream noise while not
affecting the noise generation in the downtream region. Opposite to that, masking the variable, 𝑓𝑢𝑥 , which is the forcing
corresponding to the streamwise velocity, causes significant reduction in the downstream noise while having limited
effect on the sidestream noise generation. These results indicate that the forcing term, 𝑓𝑢𝑥 , contained within 𝑟 < 2𝐷
is the only responsible for downstream noise propagation. In the following, we will continue the analysis to identify
acoustically efficient forcing using only this component.

B. Applying RESPOD method to obtain low-rank forcing

In this part, we will obtain a low-rank representation of the forcing that is associated to the most-energetic response
structures in the acoustic field. To achieve this, we will first compute the SPOD modes of the acoustic field, and then
use the RESPOD approach to extract the associated forcing modes, which is discussed in detail in Karban et al. [29].
In Fig. 2, the SPOD gains of the pressure in the downstream acoustic field, defined as 𝑥/𝐷, 𝑟/𝐷 ∈ [6, 30] × [5, 6],
and the forcing term, 𝑓𝑢𝑥 , in the near field, defined as 𝑥/𝐷, 𝑟/𝐷 ∈ [0, 30] × [0, 2], are shown for 𝑆𝑡 = 0.6. It is seen
that the leading SPOD gain corresponds to more than 75% of the total acoustic energy, and the sum of the first five
SPOD gains correspond to 99% of the total acoustic energy, which indicates a low-rank organisation in the acoustic
field. For the forcing in the near field, on the other hand, we see that the leading SPOD mode contains less than 6% of
the total energy in 𝑓𝑢𝑥 , and one needs over 100 modes to reach 90% of the total forcing energy, indicating an extremely
high-rank structure.
Using the RESPOD method, one can extract from this high-rank forcing data, a low-rank portion which is correlated
to the low-rank pressure structures observed in the acoustic field. In Fig. 3, we show the leading SPOD mode of pressure
in the acoustic field, and the associated RESPOD mode of forcing in the near field, together with the energy distribution

4
Fig. 2 SPOD gains of the pressure in the acoustic field (top) and streamwise forcing, 𝑓𝑢𝑥 in the near field at
𝑆𝑡 = 0.6.

5
Fig. 3 Optimal SPOD mode of acoustic pressure (top) and the associated RESPOD mode of forcing (center)
together with the energy distribution in the first twenty RESPOD modes of forcing (bottom) at 𝑆𝑡 = 0.6.

6
Fig. 4 PSD of acoustic pressure generated using rank-5 and rank-1 forcing, respectively, obtained by RESPOD,
in comparison to acoustic field obtained from LES data (corresponding to full-rank forcing in the ideal case) at
different frequencies ranging from 𝑆𝑡 = 0.4 to 1 (from top to bottom).

of the RESPOD modes of the forcing. The leading SPOD mode is well converged while the associated RESPOD mode
of forcing contains many wiggles as an indicator of being under-converged. It is seen that the first RESPOD mode
contains less than 0.8% of the total forcing energy while it is associated to the leading SPOD mode of the acoustic
field, corresponding to 75% of the total noise in the downstream region. This difference may explain the different
convergence observed in the response and forcing modes, as it may require extremely long simulation time to converge
the forcing structures that contain less than 1% of the total forcing energy. Also note that the energy in the RESPOD
modes of forcing does not decay in higher modes. Although this might also be due to under-convergence, a decay in the
RESPOD modes are not guaranteed even in the fully converged scenario, since in RESPOD, forcing modes are ordered
with respect to the energy of the response they generate. One should note that this under-convergence observed in the
forcing modes does not pose a problem in the following analysis. The SPOD modes of the response and the RESPOD
modes of the forcing are computed using realisations of response and forcing that exactly correspond to the same time
instant. In the ideal case of an error-free database, Eq. (3) is satisfied for each pair of response-forcing realisations.
Therefore, no matter how under-converged the forcing data is, the structures generating the converged acoustic field is,
by construction, ensured to be contained in the forcing mode seen in Fig. 3.
The first RESPOD mode contains all the structures that are correlated to the leading SPOD mode [? ]. This indicates
that the remaining forcing that amounts to 99% of the total forcing energy can generate only 25% of the total acoustic
energy. In Fig. 4, we show a comparison of the total acoustic field to the acoustic fields obtained using rank-5 and
rank-1 forcing extracted using RESPOD method, respectively, for a number of frequencies, 𝑆𝑡 ∈ [0.4, 1]. We see that
the rank-5 forcing recovers nearly the entire acoustic field in the downstream region. The rank-1 forcing also recovers
a significant portion of the downstream acoustic field while deviation becomes more apparent as the probe moves

7
upstream. It should be noted that although the acoustic field predicted by the rank-1 forcing should correspond to 75%
of the total acoustic energy, the actual prediction amounts to less than this ratio. This is due to the errors contained in
the LES database, causing a loss in the correlation information between the response and the forcing. Despite all the
limitations of the existing database, we see that it is still possible to define a rank-1 forcing which can generate most of
the downstream noise.
Note that this rank-1 forcing, even in the ideal case of error-free database, can contain correlated-but-silent
components, i.e., components that are correlated to the acoustic field but does not take active role in noise generation. A
detailed discussion about these correlated silent components can be found in Karban et al. [29]. In the final paper, we
will further decompose the rank-1 forcing obtained by RESPOD to extract the active components that actually generate
the leading SPOD mode of the acoustic pressure seen in Fig. 3.

V. Conclusion
A process to obtain low-rank forcing associated to jet noise in the downstream region is presented. We demonstrated
that a substantial amount of the downstream acoustic energy is associated to a forcing mode that contains less than 1%
of the total forcing energy. The presented approach is generic in the sense that it can be applied to any flow. The final
aim of this study is to identify the components in this low-rank forcing that actually drives the downstream noise, and to
provide a model for these components. The analysis will be extended to cover Mach number and flight stream effects as
well.

Acknowledgments
This work has received funding from the Clean Sky 2 Joint Undertaking under the European Union’s Horizon 2020
research and innovation programme under grant agreement No 785303. U.K. has received funding from TUBITAK
2236 Co-funded Brain Circulation Scheme 2 (Project No: 121C061).

References
[1] Lighthill, M. J., “On sound generated aerodynamically I. General theory,” Proceedings of the Royal Society of London. Series A.
Mathematical and Physical Sciences, Vol. 211, No. 1107, 1952, pp. 564–587. https://doi.org/10.1098/rspa.1952.0060.

[2] Tam, C. K. W., “Jet Noise: Since 1952,” Theoretical and Computational Fluid Dynamics, Vol. 10, 1998, pp. 393–405.

[3] Jordan, P., and Colonius, T., “Wave Packets and Turbulent Jet Noise,” Annual Review of Fluid Mechanics, Vol. 45, No. 1, 2013,
pp. 173–195. https://doi.org/10.1146/annurev-fluid-011212-140756.

[4] Brès, G. A., and Lele, S. K., “Modelling of jet noise: a perspective from large-eddy simulations,” Philosophical Transactions
of the Royal Society A: Mathematical, Physical and Engineering Sciences, Vol. 377, No. 2159, 2019, p. 20190081.
https://doi.org/10.1098/rsta.2019.0081, URL https://royalsocietypublishing.org/doi/abs/10.1098/rsta.2019.0081.

[5] Cavalieri, A. V. G., Jordan, P., and Lesshafft, L., “Wave-Packet Models for Jet Dynamics and Sound Radiation,” Applied
Mechanics Reviews, Vol. 71, No. 2, 2019. https://doi.org/10.1115/1.4042736, 020802.

[6] Lilley, G. M., “On the noise from jets,” AGARD CP-131, 1974, pp. 13–1.

[7] Howe, M. S., “Contributions to the theory of aerodynamic sound, with application to excess jet noise and the theory of the
flute,” Journal of Fluid Mechanics, Vol. 71, No. 4, 1975, p. 625–673. https://doi.org/10.1017/S0022112075002777.

[8] Doak, P. E., “Fluctuating total enthalpy as a generalized acoustic field,” Acoustical Physics, Vol. 41, 1995, pp. 677–685.

[9] Goldstein, M. E., “A generalized acoustic analogy,” Journal of Fluid Mechanics, Vol. 488, 2003, p. 315–333. https:
//doi.org/10.1017/S0022112003004890.

[10] Colonius, T., Lele, S. K., and Moin, P., “Sound generation in a mixing layer,” Journal of Fluid Mechanics, Vol. 330, 1997, p.
375–409. https://doi.org/10.1017/S0022112096003928.

[11] Freund, J. B., “Noise sources in a low-Reynolds-number turbulent jet at Mach 0.9,” Journal of Fluid Mechanics, Vol. 438, 2001,
p. 277–305. https://doi.org/10.1017/S0022112001004414.

[12] Sandberg, R. D., and Sandham, N. D., “Direct numerical simulation of turbulent flow past a trailing edge and the associated
noise generation,” Journal of Fluid Mechanics, Vol. 596, 2008, p. 353–385. https://doi.org/10.1017/S0022112007009561.

8
[13] Karabasov, S. A., Afsar, M. Z., Hynes, T. P., Dowling, A. P., McMullan, W. A., Pokora, C. D., Page, G. J., and McGuirk,
J. J., “Jet Noise: Acoustic Analogy informed by Large Eddy Simulation,” AIAA Journal, Vol. 48, No. 7, 2010, pp. 1312–1325.
https://doi.org/10.2514/1.44689, URL https://doi.org/10.2514/1.44689.

[14] Bassetti, A., and Nichols, J. W., Analysis of LES for source modeling in jet noise, 2014. https://doi.org/10.2514/6.2014-2905,
URL https://arc.aiaa.org/doi/abs/10.2514/6.2014-2905.

[15] Leib, S. J., Ingraham, D., and Bridges, J. E., Evaluating Source Terms of the Generalized Acoustic Analogy using the Jet Engine
Noise REduction (JENRE) Code, 2017. https://doi.org/10.2514/6.2017-0459, URL https://arc.aiaa.org/doi/abs/10.2514/6.2017-
0459.

[16] McKeon, B. J., and Sharma, A. S., “A critical-layer framework for turbulent pipe flow,” Journal of Fluid Mechanics, Vol. 658,
2010, pp. 336–382. https://doi.org/10.1017/S002211201000176X.

[17] Hwang, Y., and Cossu, C., “Amplification of coherent streaks in the turbulent Couette flow: an input–output analysis at low
Reynolds number,” Journal of Fluid Mechanics, Vol. 643, 2010, p. 333–348. https://doi.org/10.1017/S0022112009992151.

[18] Mclaughlin, D. K., Morrison, G. L., and Troutt, T. R., “Experiments on the instability waves in a supersonic jet and their
acoustic radiation,” Journal of Fluid Mechanics, Vol. 69, No. 1, 1975, p. 73–95. https://doi.org/10.1017/S0022112075001322.

[19] Papamoschou, D., “Mach Wave Elimination in Supersonic Jets,” AIAA Journal, Vol. 35, No. 10, 1997, pp. 1604–1611.
https://doi.org/10.2514/2.19, URL https://doi.org/10.2514/2.19.

[20] Tam, C. K. W., Chen, P., and Seiner, J. M., “Relationship between the instability waves and noise of high-speed jets,” AIAA
Journal, Vol. 30, No. 7, 1992, pp. 1747–1752. https://doi.org/10.2514/3.11132, URL https://doi.org/10.2514/3.11132.

[21] Tam, C. K. W., “Supersonic Jet Noise,” Annual Review of Fluid Mechanics, Vol. 27, No. 1, 1995, pp. 17–43. https:
//doi.org/10.1146/annurev.fl.27.010195.000313, URL https://doi.org/10.1146/annurev.fl.27.010195.000313.

[22] Tam, C. K. W., Viswanathan, K., Ahuja, K. K., and Panda, J., “The sources of jet noise: experimental evidence,” Journal of
Fluid Mechanics, Vol. 615, 2008, p. 253–292. https://doi.org/10.1017/S0022112008003704.

[23] Cavalieri, A. V., Jordan, P., Agarwal, A., and Gervais, Y., “Jittering wave-packet models for subsonic jet noise,” Journal of
Sound and Vibration, Vol. 330, No. 18, 2011, pp. 4474–4492. https://doi.org/https://doi.org/10.1016/j.jsv.2011.04.007, URL
https://www.sciencedirect.com/science/article/pii/S0022460X11002641.

[24] Cavalieri, A. V. G., and Agarwal, A., “Coherence decay and its impact on sound radiation by wavepackets,” Journal of Fluid
Mechanics, Vol. 748, 2014, p. 399–415. https://doi.org/10.1017/jfm.2014.186.

[25] Lumley, J. L., “Toward a turbulent constitutive relation,” Journal of Fluid Mechanics, Vol. 41, No. 2, 1970, p. 413–434.
https://doi.org/10.1017/S0022112070000678.

[26] Picard, C., and Delville, J., “Pressure velocity coupling in a subsonic round jet,” International Journal of Heat and
Fluid Flow, Vol. 21, No. 3, 2000, pp. 359 – 364. https://doi.org/https://doi.org/10.1016/S0142-727X(00)00021-7, URL
http://www.sciencedirect.com/science/article/pii/S0142727X00000217.

[27] Towne, A., Schmidt, O. T., and Colonius, T., “Spectral proper orthogonal decomposition and its relationship to dynamic mode
decomposition and resolvent analysis,” Journal of Fluid Mechanics, Vol. 847, 2018, p. 821–867. https://doi.org/10.1017/jfm.
2018.283.

[28] Schmidt, O. T., Towne, A., Rigas, G., Colonius, T., and Brès, G. A., “Spectral analysis of jet turbulence,” Journal of Fluid
Mechanics, Vol. 855, 2018, p. 953–982. https://doi.org/10.1017/jfm.2018.675.

[29] Karban, U., Martini, E., Cavalieri, A. V. G., Lesshafft, L., and Jordan, P., “Self-similar mechanisms in wall turbulence studied
using of resolvent analysis,” , 2021.

[30] Ffowcs Williams, J. E., “The noise from turbulence convected at high speed,” Philosophical Transactions of the Royal Society
of London. Series A, Mathematical and Physical Sciences, Vol. 255, No. 1061, 1963, pp. 469–503. https://doi.org/10.1098/rsta.
1963.0010, URL https://royalsocietypublishing.org/doi/abs/10.1098/rsta.1963.0010.

[31] Crighton, D., “Basic principles of aerodynamic noise generation,” Progress in Aerospace Sciences, Vol. 16, No. 1, 1975, pp.
31–96. https://doi.org/https://doi.org/10.1016/0376-0421(75)90010-X, URL https://www.sciencedirect.com/science/article/pii/
037604217590010X.

9
[32] Brès, G., Ham, F., Nichols, J., and Lele, S., “Unstructured large-eddy simulations of supersonic jets,” AIAA Journal, Vol. 55,
No. 4, 2017, pp. 1164–1184. https://doi.org/10.2514/1.J055084.

[33] Brès, G. A., Jordan, P., Jaunet, V., Le Rallic, M., Cavalieri, A. V. G., Towne, A., Lele, S. K., Colonius, T., and Schmidt, O. T.,
“Importance of the nozzle-exit boundary-layer state in subsonic turbulent jets,” Journal of Fluid Mechanics, Vol. 851, 2018, p.
83–124. https://doi.org/10.1017/jfm.2018.476.

[34] Bugeat, B., Chassaing, J.-C., Robinet, J.-C., and Sagaut, P., “3D global optimal forcing and response of the supersonic boundary
layer,” Journal of Computational Physics, Vol. 398, 2019, p. 108888.

[35] Balay, S., Gropp, W. D., McInnes, L. C., and Smith, B. F., “Efficient Management of Parallelism in Object Oriented Numerical
Software Libraries,” Modern Software Tools in Scientific Computing, Birkhäuser Press, 1997, pp. 163–202.

10