PHYSICAL REVIEW FLUIDS 10, 063201 (2025)

Forced synchronization of globally stable and unstable flames

Manikandan Balasubramaniyan ,1 Linghan Chen,2 Wen Ao,2 Peijin Liu,2

Yu Guan,3 and Larry K. B. Li1,*
1
Department of Mechanical and Aerospace Engineering,
The Hong Kong University of Science and Technology, Clear Water Bay, Hong Kong
2
National Key Laboratory of Solid Rocket Propulsion, Northwestern Polytechnical University, Xi’an, China
3
Department of Aeronautical and Aviation Engineering, The Hong Kong Polytechnic University,
Kowloon, Hong Kong

(Received 2 April 2024; accepted 20 May 2025; published 3 June 2025)

We experimentally investigate the forced synchronization of a turbulent, lean-premixed,

bluff-body-stabilized flame undergoing self-excited oscillations due to global hydro-
dynamic instability. We acoustically force the flame at different amplitudes (α) and
frequencies ( f f ) around its natural global frequency ( fn ), while measuring its heat release
rate (HRR) response via time-resolved CH* chemiluminescence imaging. As α increases
at a fixed f f , the flame initially oscillates quasiperiodically at both fn and f f , but syn-
chronizes with the forcing above a critical α, consistent with the behavior of a canonical
self-excited oscillator. The minimum α required for synchronization increases with the
detuning (| f f − fn |), but not symmetrically around f f / fn = 1, resulting in a skewed Arnold
tongue. The HRR amplitude grows at small detuning due to resonant amplification but
decays at large detuning due to asynchronous quenching. Comparing the globally unstable
flame with an equivalent globally stable flame, we find that both exhibit a range of coupled
states, including desynchronization, phase synchronization, and generalized synchroniza-
tion. Crucially, the phase locking value varies spatially throughout the flame body, with the
shear layers synchronizing more readily than the wake and recirculation zones. At higher
detuning levels (i.e., when the forcing frequency is below 0.9 or above 1.1 times the natural
frequency, f f / fn < 0.9 or f f / fn > 1.1), the HRR amplitude of the globally unstable flame
is observed to be lower than that of the globally stable flame. This finding suggests that
global hydrodynamic instability in a flame may serve as a passive mechanism to weaken
self-excited thermoacoustic oscillations in combustion systems such as gas turbines and
rocket engines.

DOI: 10.1103/PhysRevFluids.10.063201

I. INTRODUCTION
Gas turbines are widely used in industrial and aeronautical applications due to their high
operational efficiency and competitive capital cost [1]. However, their NOx emissions can contribute
to air pollution and global warming [2]. Operating gas turbines in lean premixed mode can reduce
NOx emissions, but it also increases the risk of thermoacoustic instability [3,4]. This self-excited
oscillatory phenomenon arises from a constructive phase relationship between the unsteady heat
release rate (HRR) of a flame and the acoustic modes of the surrounding combustor [5–8]. Under
certain conditions, thermoacoustic instability can lead to high-amplitude flow oscillations at the
resonant acoustic modes of the combustor, accelerating component wear and degrading system

*
Contact author: [email protected]

2469-990X/2025/10(6)/063201(23) 063201-1 ©2025 American Physical Society

 MANIKANDAN BALASUBRAMANIYAN et al.

performance [9,10]. These challenges underscore the need for effective methods to control ther-
moacoustic oscillations.
Over the past several decades, various techniques have been developed to suppress thermoa-
coustic oscillations in combustion devices [1,9]. These methods can be broadly categorized into
passive and active control strategies [6]. In most industrial combustors, passive control is preferred
over active control because it does not rely on actuators, sensors, or controllers—all of which are
prone to failure in the harsh environment of combustion systems [11,12]. Examples of passive
control include modifying the fuel injector or combustor geometry to disrupt the coupling be-
tween HRR and pressure [13,14], and introducing damping devices such as acoustic liners and
Helmholtz resonators [15]. However, implementing passive control effectively requires a good
understanding of the root cause of the instability—knowledge that is not always available due
to the nonlinear and sensitive nature of the thermoacoustic feedback loop [10]. Consequently,
passive control can be expensive and time-consuming to implement, often requiring ad hoc system
modifications late in the development cycle, which is typically when thermoacoustic instability fully
manifests [9,10].
Thermoacoustic instability can be driven by several physical mechanisms, including equivalence-
ratio coupling [16–20], intrinsic modes [21–24], entropy waves [25–28], and hydrodynamic
instabilities [29–35]. In the present study, we experimentally investigate how the presence of a
specific type of hydrodynamic instability—namely, global hydrodynamic instability [36–40]—can
reduce the HRR response of a flame to incident flow perturbations. This HRR desensitization may
serve as the basis for a passive approach to mitigating thermoacoustic oscillations in combustion
systems.

A. Hydrodynamic instabilities in combustors

In the combustors of gas turbines, a variety of flow features can coexist, including mixing
layers, swirling jets and wakes, and precessing vortex cores [3,32,35]. These features are often
hydrodynamically unstable and can amplify incident flow perturbations [41], such as the acoustic
velocity perturbations associated with the pressure oscillations of a thermoacoustically unstable
combustor [42]. In addition to amplifying external perturbations, these flow features can also support
self-excited oscillations of their own [36]. To explore this further, it is necessary to first review the
concepts of absolute and convective instability in hydrodynamics.
Locally convectively unstable flows are highly sensitive to incident perturbations and can
amplify external disturbances as they propagate downstream in space [41]. In contrast, locally
absolutely unstable flows are relatively insensitive to incoming disturbances, but can still be
disrupted if the amplitude of the perturbations is sufficiently large [43–45]. If a flow contains
an extended region of local absolute instability, it can become globally unstable and oscillate in
a self-excited limit cycle at a natural frequency determined by the flow configuration [36,46].
Such a self-excited flow is said to exhibit global hydrodynamic instability, or simply global
instability [36].
In a typical combustor, convective instabilities in the reactant flow can cause vortical waves
to grow and convect along the flame surface [5]. This wrinkles the flame, producing large-scale
coherent structures and spatially organized pockets of high HRR [29,47]. These structures are
generated periodically, leading the flame to release heat in a periodic fashion as well. If the HRR is
sufficiently in phase with the pressure field, thermal energy can be transferred from the flame to the
acoustic modes of the combustor, leading to thermoacoustic instability [9,48].
If the reactant flow is globally unstable, its self-excited hydrodynamic oscillations can cause
the flame to release heat at its natural frequency, while remaining relatively insensitive to external
perturbations at other frequencies. This insensitivity to off-resonant forcing can potentially be
leveraged to weaken thermoacoustic oscillations, provided that the natural frequency of the globally
unstable flame is tuned away from that of the combustor’s acoustic mode [37,38]. However, if the
natural hydrodynamic frequency is close to the natural thermoacoustic frequency, synchronization

063201-2
 FORCED SYNCHRONIZATION OF GLOBALLY STABLE AND …

between the flow and acoustic fields may occur, resulting in a wide range of complex dynamics
[31,49–52].

B. Synchronization in combustion systems

Synchronization is a universal process in nonlinear science, involving the adjustment of rhythms
between two (or more) oscillatory systems due to their coupled interactions [53,54]. In self-excited
combustion systems, when a globally unstable flame interacts with thermoacoustic oscillations,
the natural frequencies of the two systems may shift toward each other in a process known as
mutual synchronization [53,54]. Although mutual synchronization has been shown to play a role
in the generation of thermoacoustic instability [55], the natural frequency of the thermoacoustic
mode is typically quite robust—resistant to significant deviation from the natural frequency of
the combustor’s acoustic mode—because the energy gain and loss in each limit cycle are small
compared to the total energy already present in the oscillations [42]. This suggests that flame-
acoustic interactions can often be approximated as a unidirectional forcing scenario, in which the
globally unstable flame is perturbed by external forcing at a fixed temporal frequency—a process
referred to as forced synchronization [53,54]. In the combustion community, forced synchronization
has been most commonly studied in the context of active control, particularly open-loop control of
self-excited thermoacoustic systems [56–62].
Previous studies have also examined forced synchronization in various globally unstable flows,
including low-density jets [45,63], jet diffusion flames [37,40], cross-flowing jets [64], and bluff-
body wakes [33,65]. Some key findings from these studies include (i) a transition from an unforced
periodic limit cycle to quasiperiodicity on a T 2 torus via a Neimark-Sacker bifurcation as the forcing
amplitude increases; (ii) two distinct routes from T 2 quasiperiodicity to 1:1 synchronization—
one via an inverse Neimark-Sacker bifurcation, and the other via a saddle-node bifurcation; (iii)
asynchronous quenching at large detuning, leading to a reduction in overall oscillation amplitude;
(iv) resonant amplification at small detuning, resulting in increased oscillation amplitude; (v)
the emergence of a ∨-shaped 1:1 synchronous region, known as an Arnold tongue; and (vi) the
ability to phenomenologically model these synchronization dynamics using a low-order canonical
oscillator subjected to external time-periodic forcing. In summary, the synchronization phenomena
that emerge when a globally unstable flow interacts with external forcing are complex and not yet
fully characterized—particularly in the context of combustion systems. This research gap forms the
motivation for the present study.

C. Contributions of the present study

In this experimental study, we address two research questions:
(1) How do globally unstable and stable flames differ in their response to external forcing?
Globally unstable and stable flows are known to respond differently to external disturbances
(Sec. I A), but how and to what extent combustion can influence this response under various forcing
conditions are still not fully understood. Our aim is to determine whether HRR suppression or HRR
amplification occurs in such flames, revealing whether these instability modes aid or impede the
passive weakening of thermoacoustic instability.
(2) How does the forced response vary at different flame locations (e.g., shear layer, recirculation
zone, and wake)? For efficient implementation of active control, these critical flame regions may be
targeted to achieve a larger reduction in HRR amplitude using a lower actuator power. To do this,
it is not enough to deploy just a temporal analysis but is necessary to incorporate a spatiotemporal
analysis as well, revealing the critical flame regions that contribute significantly to either HRR sup-
pression or HRR amplification. In addition, we map out a variety of forced synchronization states.
Although some of these states have been identified before in forced thermoacoustically self-excited
combustors [57–59], they have received comparatively little attention in forced hydrodynamically
self-excited flames.

063201-3
 MANIKANDAN BALASUBRAMANIYAN et al.

FIG. 1. (a) Illustration of the bluff-body stabilized turbulent premixed combustor and its fuel/air supply
system. The labels “P1” and “P2” denote the two pressure transducers, and “HSC” denotes the high-speed
camera. Panel (b) shows a side view of the combustor, superimposed over an instantaneous CH* chemilumi-
nescence image of a globally unstable flame. This figure is adapted from our previous work [66].

II. EXPERIMENTAL SETUP

Experiments are performed in a bluff-body stabilized turbulent lean-premixed combustor (see
Fig. 1), which is identical to that of our recent study on global hydrodynamic instabilities in turbulent
reacting flows [66]. The flame is produced with methane (CH4 , 99.9%) as the fuel and air as the
oxidizer. The methane is supplied from a compressed gas cylinder, while the air is supplied from a
centralized compressor. Particle and oil filters (HEPA standard) are installed in both the fuel and air
lines to prevent any dust or oil droplets from entering the combustor. The fuel and air flow rates are
individually controlled via two Parker Series II mass flow controllers, which have full-scale ranges
of 250 and 1670 l/min (±0.1% full scale) for fuel and air, respectively.
The overall setup is 2140 mm long and consists of three main parts: a settling chamber, a flow
conditioning duct, and a combustor. The settling chamber is made of two conical sections sand-
wiching a cylindrical section, and its total length is 800 mm. A mixing grid is installed at the
inlet of the settling chamber to enhance mixing of the fuel and air. Downstream from the settling
chamber is a flow conditioning duct, which consists of three parts with a total length of 950 mm. The
cross-sectional area of the flow conditioning duct is 60 × 60 mm2 . Two loudspeakers (FaitalPRO
6FE100) are mounted on the side of the flow conditioning duct (300 mm downstream from the
settling chamber exit) for acoustic perturbations to be applied to the flame. Downstream from the
flow conditioning duct is the combustor whose length is 390 mm. We intentionally use such a short
combustor to avoid resonant coupling between the longitudinal acoustic modes and the hydrody-
namic mode of the flame, enabling us to investigate the latter in isolation, without contamination
from potential thermoacoustic feedback. The cross-sectional area of the combustor is identical to
that of the flow conditioning duct. A circular cylinder of diameter D = 10 mm is installed in the
combustor to stabilize the flame [see the magenta circle in Fig. 1(b)].
A globally unstable flame is examined at a Reynolds number of Re = 1360 ± 6% and an
equivalence ratio of φ = 0.47 ± 3%. A globally stable flame is examined at the same Re but a
slightly higher φ of 0.51 ± 3%. Here Re ≡ Ub D/ν, where Ub is the bulk flow velocity, D is the
diameter of the cylindrical bluff body, and ν is the kinematic viscosity of the air-CH4 reactant
mixture. We examine the synchronization dynamics across a wide range of forcing conditions:
0.4  f f / fn  1.6 and 0  α  0.25, where f f is the forcing frequency, fn = 48 Hz is the natural

063201-4
 FORCED SYNCHRONIZATION OF GLOBALLY STABLE AND …

frequency of the global hydrodynamic mode, and α is the forcing amplitude (defined below). For
external forcing, we use a function generator (Tektronix AFG3022C) to produce a sinusoidal wave-
form, which is then amplified (OSD XMP300) and fed into the combustor via the two loudspeakers.
The forcing amplitude is defined as α ≡ uf /Ub , where uf is the acoustic velocity perturbation
computed from the two-microphone method. To estimate uf , we measure the acoustic pressure at
two separate locations using pressure transducers (PCB103B02), which are wall mounted 350 mm
(P1) and 200 mm (P2) upstream of the combustor exit plane, as per Fig. 1(a).
The flame dynamics are characterized via the HRR fluctuations, which are measured via the CH*
chemiluminescence signal captured by a high-speed digital camera (Photron FASTCAM SA-5). In
this study, the globally unstable flame is produced near the lean blow-off limit, where the CH*
chemiluminescence signal is relatively weak. To achieve a reliable signal-to-noise ratio, we use an
image intensifier (Lambert HiCATT 18) coupled to a bandpass optical filter centered at 432 nm,
with 90% transmission between 422 and 442 nm, to capture the CH* chemiluminescence emission.
The camera resolution is fixed at 1024 × 512 pixels2 during the entire experimental campaign,
yielding a spatial resolution of around 0.17 mm per pixel; the yellow dotted box shown in Fig. 1(b)
indicates the field of view. As noted earlier, the maximum forcing frequency is f f / fn = 1.6, which is
equivalent to around f f = 77 Hz, so the image sampling frequency is set to 2000 frames per second,
which is more than 25 times the maximum f f (or more than 40 times fn ). For each experimental
run, we record 6000–8000 flame images, corresponding to around 140–190 natural cycles of the
self-excited flame oscillations.

III. RESULTS AND DISCUSSION

A. Characterization of globally stable and unstable flames

Figure 2(a) shows the normalized RMS amplitude of the upper flame edge oscillations ζrms,U /D
as a function of φ. The amplitudes of the upper and lower flame branches are quite similar; therefore,
we show only the amplitude of the upper flame edge oscillations. In this study, the upper flame
edge oscillation ζU is tracked using an edge detection algorithm. The fluctuation of the upper edge
position is calculated by subtracting the time-averaged mean value from the instantaneous upper
edge position. We applied 5%–15% Otsu thresholding and have included error bars. The values of

ζrms,U /D are extracted at the axial location x/D = 8, where the flame oscillation is at a maximum.

These ζrms,U /D values are determined when increasing (hollow marker) as well as decreasing (solid
marker) the φ values. However, the results are quite similar, indicating the absence of hysteresis and
bistability. When φ > 0.575, the flame amplitude is exceedingly low, indicating steady behavior,
which is also evident from the power spectral density (PSD) of ζU [Fig. 2(b)]. Regardless of the
variation direction of φ, the PSD shows only weak broadband components, indicating that the flame
is in a “steady regime.” When φ < 0.575, the flame amplitude increases as φ decreases, leading to
two distinct types of behavior: a varicose regime where symmetric flame oscillations are dominant,
and a sinuous regime where asymmetric flame oscillations are dominant. Here, the varicose mode
is globally stable, with local convective instability in the shear layer producing Kelvin-Helmholtz
vortices [36]. By contrast, the sinuous mode is globally unstable, with a large region of absolute
instability producing an asymmetric Bénard–Von Kármán (BVK) vortex street [67]. In this study,
we consider φ = 0.47 for investigating the forced synchronization of the globally unstable flame;
similarly, we consider φ = 0.51 for the globally stable flame.
Figure 2(c) shows the time trace of the amplitude of global HRR fluctuations Q  = Q /Q. Here,

Q represents the fluctuating component of the flame global HRR, whereas Q represents the time-
averaged flame global HRR. The frequency spectrum here is obtained from the HRR signal that
shows a single dominant peak at around 48 Hz, representing the natural frequency fn of the flame
[Fig. 2(d)]. To confirm the presence of a globally unstable mode, we calculate the Strouhal number
(St = fn D/Ub ), which is found to be 0.24, indicating the presence of BVK vortices, a classic form of
global hydrodynamic instability [39,44,68]. For further confirmation, we conducted this experiment

063201-5
 MANIKANDAN BALASUBRAMANIYAN et al.

FIG. 2. Transition to global instability in a bluff-body stabilized flame: (a) the normalized RMS amplitude

of the upper flame edge oscillations (ζrms,U /D) extracted at x/D = 8 as a function of the equivalence ratio

(φ); (b) the PSD of ζU along the forward path (bottom panel) and the backward path (top panel); Subfigures
(c)–(f) show the characteristics of the globally unstable flame, such as (c) time traces of the normalized
global HRR (Q  = Q /Q), (d) the PSD, (e) the correlation coefficient (rU,L ) between the upper and lower flame
branches, and (f) instantaneous CH* chemiluminescence images of the flame in the sinuous regime. Data for the
globally unstable flame are collected at Re = 1360 and φ = 0.47. Similarly, subfigures (g)–(j) are analogous
to subfigures (c)–(f) but for the globally stable flame at Re = 1360 and φ = 0.51. The solid red and dotted blue
lines represent the instantaneous flame edge and its time-averaged position, respectively. The magenta circle
represents the circular bluff body. Some of the images are adapted from our previous study [66].

at different flow rates and found that fn changes linearly with the flow rate and St remains constant
at around 0.19–0.26; this range of St values indicates the presence of a globally unstable mode. To
quantify the flame asymmetry, we extract the edge positions of the upper and lower flame branches
and compute the correlation coefficient (rU,L ) defined as follows [39]:

ζU (t )ζL (t )
rU,L =  , (1)
ζU (t )2 ζL (t )2 

where ζU and ζL are the upper and lower flame-edge oscillations, respectively. In general, rU,L is
between −1 and 1; here, rU,L > 0 indicates a positive symmetric correlation and rU,L < 0 indicates
a negative antisymmetric correlation. Similarly, rU,L = 0 indicates no correlation. In this study,
Fig. 2(e) shows the negative correlation (rU,L < 0) that confirms the presence of an asymmetric
mode, indicating BVK vortices. Figure 2(f) shows an asymmetric flame produced by the construc-
tive interaction between the two diametrically opposite shear layers, which are consistent with BVK
vortices.
In Figs. 2(g)–2(j), we show results that correspond to the globally stable flame. Unlike the glob-
ally unstable flame, the PSD spectrum of the globally stable flame shows broadband components
around 0–100 Hz, indicating the flame oscillates at multiple frequencies. This broad spectrum
could result from interactions between various flow structures and the flame, leading to a wider
range of oscillation frequencies. The correlation coefficient (rU,L ) between the upper and lower
flame branches shows marginally positive values, which suggests the presence of symmetric flame

063201-6
 FORCED SYNCHRONIZATION OF GLOBALLY STABLE AND …

FIG. 3. (a), (b) Instantaneous and (c), (d) time-averaged velocity vectors superimposed over in-plane
vorticity contours of unforced globally unstable (left) and stable (right) flames. Subplots (e) and (f) show
the normalized time-averaged axial velocity of globally unstable and stable flames at x/D = 1, 3, 5, and 8.

oscillations. More details of the characterization of globally unstable and stable flames can be
found in [66].
The instantaneous velocity fields for the globally unstable and stable flames are presented in
Figs. 3(a) and 3(b), respectively, while the corresponding time-averaged velocity fields are shown
in Figs. 3(c) and 3(d). In these figures, velocity vectors are superimposed over in-plane vorticity
contours, which clearly illustrate the sinuous and varicose modes of oscillations exhibited by the
flames. Furthermore, we have provided the normalized time-averaged axial velocity (U /Ulip , where
Ulip represents the freestream axial velocity above the bluff body) at different axial stations (x/D =
1, 3, 5, and 8). As shown, the flow velocity is consistently lower along the centerline and gradually
increases toward the freestream velocity. This trend is observed at all axial stations. Additionally,
the negative velocity in the centerline region indicates the presence of a recirculation zone. The
freestream velocity is reduced near the combustor walls due to boundary layer effects.

B. Forced response of globally stable and unstable flames

As mentioned earlier, one of the key objectives of the present study is to investigate the forced
response of globally unstable and stable flames. We aim to find out how their HRR amplitude
 
changes under various forcing conditions. Figure 4 shows the forced response η = Qrms /Q0,rms of

the globally unstable and stable flames. Here, Q0,rms is the global HRR obtained at the unforced
condition. The forcing amplitude was kept constant at α = 0.01, 0.05, and 0.08, and the forcing
frequency was varied within 0.75  f f / fn  1.25. Here, the hollow marker represents desynchro-
nized conditions, whereas the solid marker represents lock-in conditions. Similarly, the light- and
dark-green shaded regions represent amplitude reduction and amplification, respectively. At low
forcing amplitude α = 0.01 [Fig. 4(a)], the globally unstable and stable flames are unlocked with

063201-7
 MANIKANDAN BALASUBRAMANIYAN et al.

 
FIG. 4. Forcing response η = Qrms /Q0,rms of the globally unstable (φ = 0.47) and stable (φ = 0.51) flames
at various forcing conditions (0.75  f f / fn  1.25): (a) α = 0.01, (b) α = 0.05, and (c) α = 0.08. In this
study, fn of both the globally unstable and the stable flame is 48 Hz. The shaded region of the light and
dark green represents the amplitude reduction and amplification, respectively. The hollow and solid markers
represent the unlocked and locked conditions, respectively.

the forcing, and the corresponding η values are close to 1, indicating the HRR response of both
the globally unstable and stable flames are not affected much by weak forcing. When increasing
the forcing amplitude to α = 0.05 [Fig. 4(b)], the globally unstable and stable flames are locked
into the forcing at several frequency ratios ( f f / fn ); here, η increases close to f f / fn = 1, whereas
they decrease at higher detuning levels (i.e., f f  fn or f f  fn ). This increasing and decreasing
trend of η is consistent for both the globally unstable and stable flames; however, the amplitude
suppression was more pronounced for the globally unstable flame (i.e., η is lower for the globally
unstable flame than for the globally stable flame). When further increasing the forcing amplitude
to α = 0.08 [Fig. 4(c)], η of the globally unstable and stable flames exhibits amplitude reduction
when f f / fn < 0.9 or > 1.1 and amplification at around 0.9 < f f / fn < 1.1. In summary, the HRR
amplitude is suppressed for both the globally unstable and stable flames. However, the amplitude
reduction is notably stronger for the globally unstable flame. The stronger amplitude suppression
in the globally unstable flame is primarily due to its higher spatial sensitivity to external forcing,
which will be discussed in the next section.

C. Synchronization dynamics of globally unstable flames

1. Amplitude suppression
Figure 5(a) shows the time trace of normalized amplitude of the flame global HRR (Q  = Q /Q).

Here, Q represents the fluctuating component of the global HRR, whereas Q represents the
time-averaged global HRR. The amplitude of Q  monotonically decreases as the forcing amplitude
increases towards lock-in (α  0.075); this indicates the gradual suppression of the globally unsta-
ble mode due to the forcing. However, the value of Q  slightly increases beyond lock-in (α > 0.075).
Figure 5(b) shows the frequency spectrum for the four α values including the unforced condition.
Here, fn indicates the natural frequency of the globally unstable mode, which is indicated by a red
vertical line; similarly, f f indicates the frequency of the forced mode, which is indicated by a blue
vertical line. At low forcing amplitude (α = 0.038), the frequency spectrum shows a dominant peak
at fn , and while increasing the forcing amplitude to α = 0.063, the dominant peak shifts from fn
to f f ; however, the fn mode still coexists. At low forcing amplitudes along with fn and f f , some
additional peaks emerged around fn and f f , which are known as the sidebands, and they are caused
by the nonlinear interactions due to fn and f f [40]. When further increasing the forcing amplitude
(α  0.075), the fn locks into f f , leaving no sign of the natural mode in the frequency spectrum.
This is referred to as the lock-in (also known as synchronization). The onset of lock-in here occurs
at α = 0.075.

063201-8
 FORCED SYNCHRONIZATION OF GLOBALLY STABLE AND …

FIG. 5. Amplitude suppression of the globally unstable flame ( f f / fn = 0.81): (a) the time trace of
normalized amplitude of the flame global HRR Q  = Q /Q; (b) the PSD; (c) the instantaneous CH* chemilumi-
nescence of the flame image; (d) the normalized amplitude of HRR response  I = I/Imax ; (e) the phase-locking
value κ; and (f) the probability of recurrence P(τ ) of the forcing and HRR signal as the function of time
delay τ . The data are shown for five different forcing amplitudes including the unforced case (from bottom
to top): α = 0, α = 0.038, α = 0.063, α = 0.075, and α = 0.100. The onset of lock-in occurs at α = 0.075.
The parameters used for computing P(τ ) are the embedding dimension=10, the time delay τ = 1 ms, and the
recurrence of threshold ε = 25% of the maximum attractor dimensions.

063201-9
 MANIKANDAN BALASUBRAMANIYAN et al.

The instantaneous CH* chemiluminescence flame images for the aforementioned forcing con-
ditions are shown in Fig. 5(c). Here, images clearly show two modes of flame oscillations:
(i) sinuous mode and (ii) varicose mode. At unforced conditions, the flame exhibits a sinusoidal
mode of oscillations where the top and bottom flame branches are asymmetrically oscillating.
This sinusoidal mode continues even after introducing the forcing (however, up to a lower forcing
amplitude, α = 0.038). When the forcing amplitude rises (α  0.038), the flame oscillation changes
from sinuous to varicose mode. Here, the varicose mode (i.e., symmetric oscillation) represents the
nature of forcing; therefore, the change of flame oscillations from the sinuous to varicose mode can
be considered as an indicator of the onset of lock-in.
Figure 5(d) shows the distribution of normalized HRR amplitude ( I = I/Imax ) for various forcing
conditions. Here, I represents the time-averaged intensity of an individual pixel of the flame,
whereas Imax represents the highest time-averaged intensity for all the pixels of the unforced
and forced flames. Here,  I is between 0 (black) and 1 (yellow) across the entire flame region.
At the unforced condition, the maximum flame response (yellow region) is observed around the
recirculation zone and wakes due to the presence of a sinuous mode; however, when the forcing is
introduced, the amplitude of the HRR response changes over the entire flame region. At low forcing
amplitude (α  0.063), the HRR amplitude of the shear layer region is higher than the other flame
regions, whereas increasing the forcing amplitude (α  0.075) lowers the HRR amplitude of the
overall flame regions such as the shear layers, recirculation zones, and wakes.
In this study, we determine the degree of synchronization using the phase-locking value κ. We
calculate κ as [69]
 N 
 
 
κ = N −1  ei θt , (2)
 
t=1

where θt represents the instantaneous relative phase at the tth time instant and N represents the
total number of samples. A quantitative approach to characterizing the synchronization behavior is
relatively rare in combustion instability problems. Mondal et al. [70] investigated the emergence of
thermoacoustic instability in a turbulent combustor and calculated κ to estimate the temporal varia-
tion of global synchrony during the transition from combustion noise to thermoacoustic instability.
Recently, Jegal et al. [71] calculated the phase-locking values to identify mutual synchronization
between two flame modes in a coupled combustor system.
Figure 5(e) shows the distribution of phase-locking value κ, calculated from the phase difference
between the HRR fluctuations of each pixel to the forcing signal. Here, the value of κ is between
0 and 1 (i.e., κ = 0 represents the state of unlocked, whereas the κ = 1 represents the state of
lock-in). The gray region here indicates the absence of flame, indicating the net HRR value is
zero; therefore, the κ values are undefined. At low forcing conditions (α  0.063), the κ values
at the flame shear layer regions are higher (κ > 0.5, blue color) than the other flame regions
such as the recirculation zone and wakes (κ < 0.5, green color). This indicates the flame shear
layers are relatively more responsive to the forcing, locking into the forcing even at low forcing
amplitude, whereas the recirculation zone and the wake regions remain unlocked. We computed
the amplitude-weighted spatial average of the phase-locking value κ̂ of the entire flame region that
corresponds to the low forcing condition (α  0.038). We found that the value of κ̂ is lower than
0.5, indicating the majority of the flame is not under lock-in. When increasing the forcing amplitude
(α  0.063), the phase-locking value of most flame regions increases (i.e., the flame regions become
more blue), indicating the majority of the flame regions are locked into the forcing. We computed
κ̂ values of around 0.82–0.88 (>0.5), indicating the majority of the flame regions are under the
lock-in state.
To determine the various dynamical states of the flame, we compute the probability of recurrence
P(τ ) based on the global HRR and the forcing signal. Recurrence is a fundamental property of
dynamical systems, and the trajectory of a dynamical system returns to the neighborhood of a
given point in the phase space after a time lag τ . There are many different techniques in nonlinear

063201-10
 FORCED SYNCHRONIZATION OF GLOBALLY STABLE AND …

dynamics that exploit the concept of recurrence; however, in this study, we compute the probability
of recurrence [P(τ )] to determine various dynamical states of synchronization. Here, the P(τ ) is
calculated based on an equation provided by Romano and co-workers [72,73].
1 −τ
N
1
P(τ ) = (ε − xi − xi+τ ), (3)
N1 − τ i=1

where N1 is the total number of reconstructed vectors, is the Heaviside step function [i.e., (X ) =
0 if X < 0 and (X ) = 1 if X  0], ε is a predefined recurrence threshold, and x is the state of the
system. More details about calculating the probability of recurrence can be found in [72,74].
Figure 5(f) shows the P(τ ) as the function of time delay τ . At low forcing condition (α  0.063),
we see that there is no correspondence between the peaks of P(τ ) of both the forcing signal
and the global HRR signal, indicating these two signals are not in synchrony. When increasing
the forcing amplitude (α = 0.075), we see that the position corresponds to where the forcing
and global HRR signals overlap with each other; however, the heights of these peaks are yet
to be matched, indicating the system undergoes phase synchronization (PS) [75]. When further
increasing the forcing amplitude (α > 0.075), the position as well as the height of the peaks for
the forcing and global HRR signals overlap, indicating generalized synchronization (GS) [76].
We compute the correlation coefficient of the probability of recurrence , which can be used as
an indicator of the degree of synchronization. Here, the values span between 0 and 1 (where
= 0 represents no synchronization, whereas = 1 represents the strongest synchronization). At
low forcing amplitude (α = 0.038), = 0.14 (<0.5), indicating weak synchronization. However,
when increasing the forcing amplitude (α  0.063), the degree of synchronization monotonically
increases, reaching a maximum of = 0.91 at the highest forcing amplitude.

2. Amplitude resonance
Figure 6 presents results analogous to Fig. 5, but under amplitude resonance conditions: 0  α 
0.026 at f f / fn = 1.04. Figure 6(a) shows the time trace of the normalized global HRR, Q  = Q /Q.
The amplitude of Q  increases linearly with α and reaches a maximum at the highest forcing


amplitude (α = 0.026), indicating amplitude resonance. The corresponding frequency spectra in

Fig. 6(b) show a single dominant peak at the natural frequency fn in the unforced case. At low
forcing amplitudes (α  0.015), two peaks appear near fn (red line) and f f (blue line), suggesting
the coexistence of modes, i.e., lock-in has not yet occurred. When α increases beyond 0.019,
fn locks into f f , and only the forcing frequency remains in the spectrum—clear evidence of
synchronization. Phase-averaged CH* chemiluminescence images for various forcing amplitudes
are shown in Fig. 6(c). A sinuous flame structure dominates both in the unforced case and at low
forcing amplitudes (α  0.015), consistent with a globally unstable (but unsynchronized) state. For
α  0.019, the flame mode transitions to a varicose structure, characteristic of the forcing mode,
confirming lock-in.
Figure 6(d) displays the spatial distribution of normalized HRR amplitude,  I = I/Imax . In the
unforced case, peak HRR amplitudes are observed in the recirculation zone and wake regions
(yellow), typical of sinuous mode dynamics. At low forcing amplitudes, the shear layer dominates
the HRR response. When α  0.019, the HRR amplitude increases across other flame regions,
consistent with broader spatial lock-in. Figure 6(e) shows the spatial distribution of the phase-
locking value κ for different forcing amplitudes. The results are qualitatively similar to those in
Fig. 5(d). At low forcing (α  0.015), only the shear layer exhibits phase locking, while other
regions remain desynchronized. For α  0.019, most of the flame synchronizes with the forcing.
Finally, Fig. 6(f) compares the probability density function P(τ ) for the forcing and HRR signals.
At low forcing levels, peak positions of P(τ ) differ, indicating no synchronization. At α = 0.019, the
peaks align in position but not height, indicating PS. At α = 0.026, both position and height match,
signaling GS.

063201-11
 MANIKANDAN BALASUBRAMANIYAN et al.

FIG. 6. Amplitude resonance of the globally unstable flame ( f f / fn = 1.04): (a) the time trace of the nor-
malized amplitude of the global HRR Q  = Q /Q; (b) the PSD; (c) the instantaneous CH* chemiluminescence
of the flame images; (d) the normalized amplitude of the HRR response  I = I/Imax ; (e) the phase-locking value
κ; and (f) the probability of recurrence P(τ ) of the forcing and the HRR signal as the function of time delay
τ . The data are shown for the five different forcing amplitudes including the unforced case (from bottom to
top): α = 0, α = 0.008, α = 0.015, α = 0.019, and α = 0.026. The onset of lock-in occurs at α = 0.019.
The parameters used for computing P(τ ) are the embedding dimension=10, the time delay τ = 1 ms, and the
recurrence of threshold ε = 25% of the maximum attractor dimensions.

063201-12
 FORCED SYNCHRONIZATION OF GLOBALLY STABLE AND …

FIG. 7. Amplitude suppression of the globally stable flame ( f f = 40 Hz): (a) the time trace of the
normalized amplitude of the flame global HRR Q  = Q /Q; (b) the PSD; (c) the instantaneous CH* chemi-
luminescence of the flame images; (d) the normalized amplitude of the flame HRR response  I = I/Imax ;
(e) the phase-locking value κ; and (f) the probability of recurrence P(τ ) of the forcing and the flame HRR
signal as the function of time delay τ . The data are shown for the four different forcing amplitudes including
the unforced case (from bottom to top): α = 0, α = 0.01, α = 0.05, and α = 0.08. The onset of lock-in occurs
at α = 0.05. The parameters used for computing P(τ ) are embedding dimension = 10, the time delay τ = 1
ms, and the recurrence of threshold ε = 25% of the maximum attractor dimensions.

D. Synchronization dynamics of globally stable flames

1. Amplitude suppression
Figure 7 illustrates the amplitude suppression behavior of a globally stable flame under various
forcing conditions (0  α  0.08 at f f = 48 Hz). As previously noted, the globally stable flame
does not exhibit a single dominant frequency but oscillates across a broad range (0–100 Hz).
To investigate suppression, forcing was applied near the flame’s dominant centerline frequency.

063201-13
 MANIKANDAN BALASUBRAMANIYAN et al.

Figure 7(a) shows the time trace of the normalized global HRR, Q  = Q /Q. The amplitude of Q 
decreases monotonically with increasing α, reaching a minimum at α = 0.08. In the corresponding
frequency spectra [Fig. 7(b)], the unforced flame shows multiple peaks between 0 and 100 Hz.
At low forcing (α = 0.01), multiple peaks persist, though the forcing frequency ( f f , marked by
a blue line) begins to dominate. Once the forcing amplitude exceeds a critical threshold (α 
0.05), f f becomes the dominant frequency, with other peaks significantly suppressed—indicating
lock-in.
Instantaneous CH* chemiluminescence images [Fig. 7(c)] show that the varicose flame mode
is present in both unforced and forced cases, unlike the globally unstable flame where a mode
transition occurs upon lock-in. Figure 7(d) presents the spatial distribution of normalized HRR
amplitude  I = I/Imax across forcing conditions. In the unforced case, the maximum amplitude is
concentrated along the shear layers, likely due to the inherent varicose mode. With increasing
forcing—especially beyond α  0.05—the HRR amplitude diminishes across all flame regions
(shear layers, recirculation zones, and wakes), indicating suppression of oscillations. Figure 7(e)
shows the spatial distribution of phase-locking values κ. At low forcing (α = 0.01), the amplitude-
weighted average κ̂ is below 0.5, suggesting weak or no lock-in. For higher forcing amplitudes
(α  0.05), κ̂ exceeds 0.5, indicating the flame has entered a phase-locked state. Finally, Fig. 7(f)
shows the probability of recurrence P(τ ) for the flame HRR and forcing signal. At low forcing
levels, the peaks of P(τ ) for the flame and forcing are misaligned, confirming the absence of
lock-in. When α  0.05, both the position and the height of the peaks align, indicating the
onset of GS.

2. Amplitude resonance
Figure 8 presents results analogous to Fig. 7, but for the amplitude resonance condition
(0  α  0.08 at f f = 40 Hz). As discussed earlier, the globally stable flame exhibits multiple
oscillation frequencies (0–100 Hz), rather than a single dominant frequency. Thus, the flame is
forced near its dominant centerline frequency. Figure 8(a) shows the time trace of the normalized
global HRR, Q  = Q /Q. Here, the amplitude of Q
 increases monotonically with forcing amplitude,
reaching a maximum at α = 0.08. The corresponding frequency spectra in Fig. 8(b) show that the
unforced flame has multiple peaks across 0–100 Hz. At low forcing amplitude (α = 0.01), the
forcing frequency ( f f ) begins to emerge as dominant (marked by a blue vertical line), though other
peaks remain. Once α  0.05, f f dominates and other peaks disappear, indicating the onset of
synchronization. Instantaneous CH* chemiluminescence images [Fig. 8(c)] show that the varicose
mode of flame oscillation persists under both unforced and forced conditions, unlike in the globally
unstable flame case, where a mode transition is observed.
The spatial distribution of normalized HRR amplitude,  I = I/Imax , is shown in Fig. 8(d). In the
unforced case, the highest amplitudes (yellow regions) are concentrated near the shear layers, likely
due to varicose structures. When forcing is applied—especially for α  0.05—the HRR amplitude
increases significantly throughout the flame, including in the shear layers, recirculation zones, and
wake regions. Figure 8(e) shows the phase-locking distribution κ for different forcing conditions.
Values of κ range from 0 (no phase locking) to 1 (complete phase locking). At low forcing amplitude
(α = 0.01), the amplitude-weighted average phase-locking value, κ̂, is below 0.5, indicating the
flame is not locked in. For α  0.05, κ̂ exceeds 0.5, signifying phase lock-in has been achieved.
The probability of recurrence P(τ ) is shown in Fig. 8(f). At low forcing amplitudes, the HRR and
forcing signals show uncorrelated peaks, confirming the absence of synchronization. However, when
α  0.05, both the positions and heights of the peaks align, indicating the system has entered a
state of GS.
In summary, both globally unstable and globally stable flames exhibit amplitude suppression and
resonance depending on the forcing condition—whether f f is far from or near the natural frequency
fn . Notably, the globally unstable flame shows stronger amplitude suppression and amplification
compared to the stable flame. Amplitude suppression is more pronounced at higher detuning (i.e.,

063201-14
 FORCED SYNCHRONIZATION OF GLOBALLY STABLE AND …

FIG. 8. Amplitude resonance of the globally stable flame ( f f = 48 Hz): (a) the time trace of the normalized
amplitude of the flame global HRR Q  = Q /Q; (b) the PSD; (c) the instantaneous CH* chemiluminescence of
the flame images; (d) the normalized amplitude of the flame HRR response  I = I/Imax ; (e) the phase-locking
value κ; and (f) the probability of recurrence P(τ ) of the forcing and HRR signal as the function of time delay τ .
The data are shown for the four different forcing amplitudes including the unforced case (from bottom to top):
α = 0, α = 0.01, α = 0.05, and α = 0.08. The onset of lock-in occurs at α = 0.05. The parameters used for
computing P(τ ) are the embedding dimension=10, the time delay τ = 1 ms, and the recurrence of threshold
ε = 25% of the maximum attractor dimensions.

when f f is far from fn ). Additionally, both flame types exhibit a range of dynamic behaviors as they
progress toward forced synchronization.

E. Lock-in map of globally unstable flames

As mentioned above, the amplitude suppression achieved in the globally unstable flame is higher
than in the globally stable flame. However, the amplitude suppression may or may not be uniform

063201-15
 MANIKANDAN BALASUBRAMANIYAN et al.

FIG. 9. (a) lock-in map shown as contours of the normalized global HRR (η = Q /Q0 , where Q0 represents
the fluctuation of HRR amplitude of the unforced case) as the function of forcing frequency ratio ( f f / fn )
and forcing amplitude (α). The black line with solid circles represents the minimum α required for lock-in.
Subfigures within the lock-in map depict the universal phenomena of phase locking and phase drifting obtained
from the instantaneous phase lag ( ψQ ,α ) between the global HRR and forcing signal. Subfigures (b), (c), (d),
and (e) show the normalized global HRR (η) for the forcing frequency ratio of f f / fn = 0.81, f f / fn = 0.94,
f f / fn = 1.04, and f f / fn = 1.17, respectively. The hollow marker represents the unlocked cases whereas the
solid marker represents the lock-in cases.

throughout the entire flame region because the forcing response of shear layers, recirculation zone,
and wakes could be different, leading to a different level of amplitude suppression. Therefore, we
apply spatiotemporal analysis to the globally unstable flame subjected to various forcing conditions.
Figure 9(a) shows the lock-in map for various forcing conditions: 0.75  f f / fn  1.25 and
0  α  0.24. Here, Q0 represents the global HRR of unforced flame. In Fig. 9(a), the black line
with circles represents the lock-in boundary between the locked and unlocked states. For example,
α values lower than the black line correspond to the unlocked states, whereas the α values higher
than the black line correspond to the lock-in states. The results show that when f f is close to fn ,
the flame locks into forcing at lower α. However, when f f is away from fn , the flame locks into
forcing at higher α. In other words, the α required for lock-in increases linearly with increasing
detuning (i.e., the gap between f f and fn or | f f − fn |). The increasing trend of α exhibits a
V-shaped lock-in curve around the flame’s natural frequency ( f f / fn = 1), which is referred to
as Arnold tongue in dynamical system theory and has already been reported in several forced
synchronization studies [37,40,45,58]. Although the lock-in curve rises on both sides of f f / fn = 1,
the curve is not symmetric with respect to f f / fn = 1 (i.e., the lock-in curve is steeper on f f / fn < 1
than on f f / fn > 1). This indicates, when f f < fn the flame is more resistant to lock-in than
when f f > fn .
The bluish region (η < 1) indicates the asynchronous quenching where the overall flame re-
sponse amplitude decreases below the unforced case. In subfigures within the lock-in map, we
show the instantaneous phase lag ψQ ,α corresponding to asynchronous quenching ( f f / fn = 0.81)
and resonant amplification ( f f / fn = 1.04). The results show that zero phase lag arises at lock-in
conditions, whereas it is unboundedly drifting for unlocked conditions. These results are consistent
whether f f is far from or nearby fn .
Figures 9(b)–9(e) show the normalized global HRR amplitude (η) corresponding to four forcing
frequency ratios: f f / fn = 0.81, 0.94, 1.04, and 1.17. The hollow markers correspond to the un-
locked forcing conditions, whereas the solid markers correspond to the lock-in conditions. Here,
we choose these frequency ratios as they are to the left and right of the Arnold tongue ( f f < fn

063201-16
 FORCED SYNCHRONIZATION OF GLOBALLY STABLE AND …

FIG. 10. Partial lock-in of the globally unstable flame at low forcing amplitude ( f f / fn = 0.81 and α =
0.038): (a) the phase-locking value κ between the forcing and local HRR signal; (b) the normalized amplitude
of HRR response  I = I/Imax ; (c) the dominant frequency spectrum f ∗ ; (d) the time trace of the normalized
amplitude of local HRR  q = q /q; (e) the PSD; (f) the instantaneous phase lag ψq ,α between the forcing and
local HRR q signal; and (g) the probability of recurrence P(τ ) of the forcing and HRR signal as the function of
time delay τ . The data are shown for three locations: recirculation zone (orange), shear layer (light magenta),
and wake region (dark magenta). The q is extracted from 10 × 10 pixels2 . The parameters used for computing
P(τ ) are the embedding dimension = 10, the time delay τ = 1 ms, and the recurrence of threshold ε = 25%
of the maximum attractor dimensions.

and f f > fn ), so that we can quantitatively characterize the flame response (i.e., in terms of η) at
various forcing conditions. The results depict that, increasing the α value leads to the reduction of
η and reaches a lower value nearby lock-in; however, further increasing of α beyond the lock-in
value leads to the rise of η. This decreasing (i.e., up to nearby lock-in) and increasing (i.e., beyond
lock-in) trend of η is consistent for the asynchronous quenching case, whereas the η values are
continuously rising before and after lock-in is achieved during resonant amplification; however, at
higher α (i.e., well beyond the lock-in value), the η value drops again. From the overall lock-in map,
it is worth mentioning that the maximum amplitude reduction is achieved at higher detuning levels
(i.e., f f  fn or f f  fn ).

F. Forcing response at local regions

1. Partial lock-in of globally unstable flames
As reported above, the forcing response of the globally unstable flame significantly varies with
different flame regions (i.e., shear layers, recirculation zone, and wakes). To examine this, we
employ spatiotemporal analysis for the forced flame that attains local and global synchronization
while increasing the forcing amplitudes. In this study, we choose the flame that undergoes amplitude
suppression (shown in Fig. 5) that exhibits local and global synchronization at low and high
forcing amplitudes. Figure 10 shows the flame response at low forcing amplitude f f / fn = 0.81 and
α = 0.038. Here, the orange, light magenta, and dark magenta indicate the recirculation zone, shear
layers, and wakes, respectively. We extract the local HRR q out of the box size of 10 × 10 pixels2
and use it for further analysis. We also perform a sensitivity analysis for different box sizes (i.e.,
5 × 5 and 15 × 15 pixels2 ), and found that the results are relatively similar. Figure 10(a) shows
the distribution of κ values, which are calculated from the flame HRR and the forcing signal.

063201-17
 MANIKANDAN BALASUBRAMANIYAN et al.

Here, the κ values span between 0 and 1 (κ = 0 represents no lock-in between the flame and
forcing; whereas, κ = 1 represents lock-in). Figure 10(b) shows the distribution of normalized
HRR amplitude  I = I/Imax ; here, the 
I spans between 0 and 1 across the flame region, indicating
no response when  I = 0 (black region) whereas, the maximum response at  I = 1. Figure 10(c)
shows the distribution of the dominant frequency f ∗ whose values span over 40–80 Hz, where
the yellow and green colors represent the frequency of natural mode fn and forcing mode f f ,
respectively.
Figure 10(d) shows the time trace of the normalized amplitude of the flame local HRR  q =
 
q /q signal for 1 s. The time trace of q corresponds to the shear layer region slightly higher than
the recirculation and wake region, which is evident from the frequency spectrum [Fig. 10(e)]. The
frequency spectrum also reveals that the shear layer locks into the forcing, indicating a single peak
appears at around f f with no sign of fn . On the other hand, the frequency spectrum for the wake
and recirculation zone shows a peak at around fn , indicating that the flame oscillations remain at
the natural mode fn . We calculate the instantaneous phase lag ψq ,α between the HRR and forcing
signal, and found that the zero phase lag between the HRR and forcing signal corresponds to the
flame shear layer region. This indicates that the shear layers undergo lock-in, whereas the phase lag
corresponding to the wake and recirculation zone drifts unboundedly, indicating these flame regions
are not under lock-in. Finally, we compute the probability of recurrence P(τ ) of the forcing and
local HRR signals as the function of time delay τ . As expected, the position and peak of the HRR
and forcing signals overlap, indicating GS for the shear layer region (where = 0.87). However,
the correlation between the flame HRR and forcing signals is very low (i.e., almost zero) for the
wake and recirculation zone regions, indicating that these regions are not under the lock-in state.

2. Global lock-in of globally unstable flames

Figure 11 illustrates the forcing response of the globally unstable flame at the high forcing
amplitude ( f f / fn = 0.81 and α = 0.075): recirculation zone (orange), shear layer (light magenta),
and wake region (dark magenta). Figure 11(a) shows the distribution of κ, which is calculated
from the flame HRR and the forcing signal; here, the κ values span between 0 and 1, where
“κ = 0” represents no phase synchronization; whereas, “κ = 1” represents phase synchronization.
Figure 11(b) shows the distribution of the normalized flame HRR amplitude  I = I/Imax ; here, the

I spans between 0 and 1 across the flame region, indicating no response when  I = 0, whereas
maximum flame response at  I = 1. Figure 11(c) shows the distribution of dominant frequency
f ∗ whose values span 40–80 Hz, where the yellow and green color represents the fn and f f ,
respectively.
Figure 11(d) shows the time trace of the normalized amplitude of the flame local HRR  q = q /q

signal for 1 s. The time trace of  q from the shear layer region is significantly higher than the
wake and recirculation zone, which is more evident from the frequency spectrum [Fig. 11(e)]. The
Frequency spectrum shows a single dominant peak at around f f for all the flame regions (shear
layer, wake, and recirculation zone), indicating that the entire flame locks into the forcing. We
calculate the instantaneous phase lag ψq ,α between the HRR and forcing signal, and found that
the zero phase lag corresponds to all the flame regions. This further confirms that the entire flame
undergoes lock-in. Finally, we compute the probability of recurrence P(τ ) of the forcing and the
flame local HRR signal as a function of time delay τ . The results from the shear layer region
indicate the position and peak of the HRR and forcing signals are well overlapped, indicating
the shear layers undergo GS, whereas the wake and recirculation zone of the flame regions
undergo PS.

IV. CONCLUSIONS
In this study, we experimentally investigated the forced synchronization of globally unstable
and stable flames generated in a bluff-body-stabilized turbulent premixed combustor. By analyzing

063201-18
 FORCED SYNCHRONIZATION OF GLOBALLY STABLE AND …

FIG. 11. Global lock-in of the globally unstable flame at high forcing amplitude ( f f / fn = 0.81 and α =
0.075): (a) the phase-locking value κ between the forcing and HRR signal; (b) the normalized amplitude of the
flame HRR response  I = I/Imax ; (c) the dominant frequency f ∗ ; (d) the time trace of the normalized amplitude
of the flame local HRR  q = q /q; (e) the PSD; (f) the instantaneous phase lag ψq ,α between the HRR and
forcing signal; and (g) the probability of recurrence P(τ ) of the forcing and HRR signals as the function of
time delay τ . The data are shown for three locations: recirculation zone (orange), shear layer (light magenta),
and wake region (dark magenta). In this study, the q is extracted from 10 × 10 pixels2 region. The parameters
used for computing P(τ ) are the embedding dimension = 10, the time delay τ = 1 ms, and the recurrence of
threshold ε = 25% of the maximum attractor dimensions.

time-resolved CH* chemiluminescence images, we extracted the HRR response under various
forcing conditions. When the forcing amplitude (α) exceeds a critical threshold, the natural
frequency of the flame ( fn ) aligns with the forcing frequency ( f f ), resulting in a state of perfect
lock-in or synchronization. The key findings are summarized as follows:
(1) Both globally stable and unstable flames exhibit asynchronous quenching (amplitude sup-
pression) and resonant amplification of the HRR under different forcing conditions. Specifically,
when f f is far from fn , asynchronous quenching is predominantly observed, whereas resonant
amplification occurs when f f is close to fn . Additionally, the HRR amplitude suppression in globally
unstable flames is stronger than that in globally stable flames. At higher detuning levels (i.e., when
f f is farther from fn ), the suppression in both flame types becomes more pronounced.
(2) From the HRR response, we observe an asymmetric V-shaped lock-in boundary centered
around the natural frequency of the global hydrodynamic mode ( f f / fn = 1). This asymmetry
indicates that the flame is less resistant to lock-in for f f / fn > 1 than for f f / fn < 1.
(3) Based on the probability of recurrence [P(τ )] computed from the HRR and forcing signals,
we identify three distinct regimes of synchronization: desynchronization, PS, and GS.
(4) By computing the spatially distributed phase-locking value κ, we gain insight into how
the degree of synchronization varies across different flame regions, such as the shear layers,
recirculation zones, and wakes. The results show that the shear layers tend to lock into the forcing
at lower α, whereas the recirculation zones and wakes require higher α for lock-in. This suggests
that the shear layers are more receptive to time-periodic forcing compared to the recirculation zones
and wakes.
In summary, we have demonstrated that both globally stable and unstable flames can synchronize
with external acoustic forcing at different detuning levels (i.e., when the forcing frequency deviates
from the natural frequency of the flames). Additionally, we observed that various flame regions,

063201-19
 MANIKANDAN BALASUBRAMANIYAN et al.

such as the recirculation zone, shear layers, and wakes, exhibit different rates of HRR amplitude
suppression when they synchronize with the acoustic forcing. Thus, we conclude that amplitude
suppression in both globally stable and unstable flames primarily arises from dissipative coupling
at higher detuning levels. In globally unstable flames, this suppression is further enhanced by their
spatial sensitivity to the applied forcing. For instance, the presence of large recirculation zones and
wakes in globally unstable flames contributes to a greater reduction in the HRR amplitude. This
suggests that globally unstable flames may be more useful for developing passive and active control
strategies to mitigate thermoacoustic instability. Furthermore, identifying the critical regions of a
flame is crucial for developing effective and efficient control strategies for combustion systems.

ACKNOWLEDGMENT
This work was supported by the Research Grants Council of Hong Kong (Project No. 16200220).
The authors have no conflicts of interest to disclose.

DATA AVAILABILITY
The data supporting the findings of this study are available from the authors upon reasonable
request.

[1] T. C. Lieuwen and V. Yang, Combustion Instabilities in Gas Turbine Engines: Operational Experience,
Fundamental Mechanisms, and Modeling (AIAA, Reston, VA, 2005).
[2] T. C. Lieuwen and V. Yang, Gas Turbine Emissions (Cambridge University Press, Cambridge, UK, 2013).
[3] Y. Huang and V. Yang, Dynamics and stability of lean-premixed swirl-stabilized combustion, Prog.
Energy Combust. Sci. 35, 293 (2009).
[4] J. O’Connor, S. Hemchandra, and T. C. Lieuwen, Combustion instabilities in lean premixed systems, in
Lean Combustion (Elsevier, New York, 2016), pp. 231–259.
[5] A. P. Dowling, Nonlinear self-excited oscillations of a ducted flame, J. Fluid Mech. 346, 271 (1997).
[6] S. Candel, Combustion dynamics and control: Progress and challenges, Proc. Combust. Inst. 29, 1 (2002).
[7] N. Noiray, D. Durox, T. Schuller, and S. Candel, A unified framework for nonlinear combustion instability
analysis based on the flame describing function, J. Fluid Mech. 615, 139 (2008).
[8] T. Schuller, T. Poinsot, and S. Candel, Dynamics and control of premixed combustion systems based on
flame transfer and describing functions, J. Fluid Mech. 894, P1 (2020).
[9] T. Poinsot, Prediction and control of combustion instabilities in real engines, Proc. Combust. Inst. 36, 1
(2017).
[10] M. P. Juniper and R. I. Sujith, Sensitivity and nonlinearity of thermoacoustic oscillations, Annu. Rev.
Fluid Mech. 50, 661 (2018).
[11] A. P. Dowling and A. S. Morgans, Feedback control of combustion oscillations, Annu. Rev. Fluid Mech.
37, 151 (2005).
[12] D. Zhao, Z. Lu, H. Zhao, X. Li, B. Wang, and P. Liu, A review of active control approaches in stabilizing
combustion systems in aerospace industry, Prog. Aerosp. Sci. 97, 35 (2018).
[13] E. Gutmark, K. Schadow, M. Nina, and G. Pita, Suppression of combustion instability by geometrical
design of the bluff-body stabilizer, J. Propul. Power 11, 456 (1995).
[14] N. Noiray, D. Durox, T. Schuller, and S. Candel, Passive control of combustion instabilities involving
premixed flames anchored on perforated plates, Proc. Combust. Inst. 31, 1283 (2007).
[15] D. Zhao and X. Y. Li, A review of acoustic dampers applied to combustion chambers in aerospace industry,
Prog. Aerosp. Sci. 74, 114 (2015).
[16] T. Lieuwen, H. Torres, C. Johnson, and B. Zinn, A mechanism of combustion instability in lean premixed
gas turbine combustors, J. Eng. Gas Turbines Power 123, 182 (2001).

063201-20
 FORCED SYNCHRONIZATION OF GLOBALLY STABLE AND …

[17] Shreekrishna, S. Hemchandra, and T. Lieuwen, Premixed flame response to equivalence ratio perturba-
tions, Combust. Theory Model. 14, 681 (2010).
[18] R. Bluemner, C. O. Paschereit, and K. Oberleithner, Generation and transport of equivalence ratio
fluctuations in an acoustically forced swirl burner, Combust. Flame 209, 99 (2019).
[19] V. Kather, F. Lückoff, C. O. Paschereit, and K. Oberleithner, Interaction of equivalence ratio fluctuations
and flow fluctuations in acoustically forced swirl flames, Int. J. Spray Combust. Dyn. 13, 72 (2021).
[20] M. Vogel, M. Bachfischer, J. Kaufmann, and T. Sattelmayer, Experimental investigation of equivalence
ratio fluctuations in a lean premixed kerosene combustor, Exp. Fluids 62, 93 (2021).
[21] M. Hoeijmakers, V. Kornilov, I. L. Arteaga, P. de Goey, and H. Nijmeijer, Intrinsic instability of flame–
acoustic coupling, Combust. Flame 161, 2860 (2014).
[22] T. Emmert, S. Bomberg, and W. Polifke, Intrinsic thermoacoustic instability of premixed flames, Combust.
Flame 162, 75 (2015).
[23] P. Buschmann, G. Mensah, and J. Moeck, Intrinsic thermoacoustic modes in an annular combustion
chamber, Combust. Flame 214, 251 (2020).
[24] L. Xu, J. Zheng, G. Wang, Z. Feng, X. Tian, L. Li, and F. Qi, Investigation on the intrinsic thermoacoustic
instability of a lean-premixed swirl combustor with an acoustic liner, Proc. Combust. Inst. 38, 6095
(2021).
[25] F. Marble and S. Candel, Acoustic disturbance from gas non-uniformities convected through a nozzle,
J. Sound Vib. 55, 225 (1977).
[26] A. S. Morgans, C. S. Goh, and J. A. Dahan, The dissipation and shear dispersion of entropy waves in
combustor thermoacoustics, J. Fluid Mech. 733, R2 (2013).
[27] E. Motheau, F. Nicoud, and T. Poinsot, Mixed acoustic–entropy combustion instabilities in gas turbines,
J. Fluid Mech. 749, 542 (2014).
[28] L. Li and D. Zhao, Coupling between entropy and unsteady heat release in a thermoacoustic system with
a mean flow, J. Sound Vib. 382, 73 (2016).
[29] K. Schadow and E. Gutmark, Combustion instability related to vortex shedding in dump combustors and
their passive control, Prog. Energy Combust. Sci. 18, 117 (1992).
[30] K. Matveev and F. Culick, A model for combustion instability involving vortex shedding, Combust. Sci.
Technol. 175, 1059 (2003).
[31] S. R. Chakravarthy, O. J. Shreenivasan, B. Boehm, A. Dreizler, and J. Janicka, Experimental character-
ization of onset of acoustic instability in a nonpremixed half-dump combustor, J. Acoust. Soc. Am. 122,
120 (2007).
[32] T. C. Lieuwen, Unsteady Combustor Physics (Cambridge University Press, Cambridge, UK, 2012).
[33] B. Emerson and T. C. Lieuwen, Dynamics of harmonically excited, reacting bluff body wakes near the
global hydrodynamic stability boundary, J. Fluid Mech. 779, 716 (2015).
[34] K. Oberleithner, S. Schimek, and C. Paschereit, Shear flow instabilities in swirl-stabilized combustors and
their impact on the amplitude dependent flame response: A linear stability analysis, Combust. Flame 162,
86 (2015).
[35] J. O. O’Connor, Understanding the role of flow dynamics in thermoacoustic combustion instability, Proc.
Combust. Inst. 39, 4583 (2023).
[36] P. Huerre and P. A. Monkewitz, Local and global instabilities in spatially developing flows, Annu. Rev.
Fluid Mech. 22, 473 (1990).
[37] M. P. Juniper, L. K. B. Li, and J. W. Nichols, Forcing of self-excited round jet diffusion flames, Proc.
Combust. Inst. 32, 1191 (2009).
[38] M. P. Juniper, Absolute and convective instability in gas turbine fuel injectors, in Proceedings of ASME
Turbo Expo (ASME International Gas Turbine Institute, Copenhagen, Denmark, 2012), GT2012-68253.
[39] B. Emerson, J. O’Connor, M. P. Juniper, and T. C. Lieuwen, Density ratio effects on reacting bluff-body
flow field characteristics, J. Fluid Mech. 706, 219 (2012).
[40] L. K. B. Li and M. P. Juniper, Lock-in and quasiperiodicity in hydrodynamically self-excited flames:
Experiments and modelling, Proc. Combust. Inst. 34, 947 (2013).
[41] C.-M. Ho and P. Huerre, Perturbed free shear layers, Annu. Rev. Fluid Mech. 16, 365 (1984).

063201-21
 MANIKANDAN BALASUBRAMANIYAN et al.

[42] F. Culick, Unsteady Motions in Combustion Chambers for Propulsion Systems, North Atlantic Treaty
Organisation, 2006, AGARDograph AG-AVT-039.
[43] M. Provansal, C. Mathis, and L. Boyer, Bénard-von kármán instability: transient and forced regimes,
J. Fluid Mech. 182, 1 (1987).
[44] P. A. Monkewitz, The absolute and convective nature of instability in two-dimensional wakes at low
reynolds numbers, Phys. Fluids. 31, 999 (1988).
[45] L. K. B. Li and M. P. Juniper, Lock-in and quasiperiodicity in a forced hydrodynamically self-excited jet,
J. Fluid Mech. 726, 624 (2013).
[46] J. M. Chomaz, P. Huerre, and L. G. Redekopp, Bifurcations to local and global modes in spatially
developing flows, Phys. Rev. Lett. 60, 25 (1988).
[47] K. Schadow, E. Gutmark, T. Parr, D. Parr, K. Wilson, and J. Crump, Large-scale coherent structures as
drivers of combustion instability, Combust. Sci. Technol. 64, 167 (1989).
[48] T. Poinsot, A. C. Trouve, D. P. Veynante, S. M. Candel, and E. J. Esposito, Vortex-driven acoustically
coupled combustion instabilities, J. Fluid Mech. 177, 265 (1987).
[49] Y. Guan, L. K. B. Li, B. Ahn, and K. T. Kim, Chaos, synchronization, and desynchronization in
a liquid-fueled diffusion-flame combustor with an intrinsic hydrodynamic mode, Chaos 29, 053124
(2019).
[50] R. I. Sujith and V. R. Unni, Dynamical systems and complex systems theory to study unsteady combustion,
Proc. Combust. Inst. 38, 3445 (2021).
[51] A. Karmarkar, S. Gupta, I. Boxx, S. Hemchandra, and J. O’Connor, Impact of precessing vortex core
dynamics on the thermoacoustic instabilities in a swirl-stabilized combustor, J. Fluid Mech. 946, A36
(2022).
[52] A. B. Britto and S. Mariappan, Self-excited vortex-acoustic lock-in in a bluff body combustor, J. Fluid
Mech. 975, A16 (2023).
[53] A. Pikovsky, M. Rosenblum, J. Kurths, and J. Kurths, Synchronization: A Universal Concept in Nonlinear
Sciences (Cambridge University Press, Cambridge, UK, 2003), Vol. 12.
[54] A. Balanov, N. Janson, D. Postnov, and O. Sosnovtseva, Synchronization: From Simple to Complex
(Springer Science & Business Media, Berlin, 2008).
[55] R. I. Sujith and S. A. Pawar, Thermoacoustic Instability, Series in Synergetics (Springer, New York,
2021).
[56] B. Bellows, A. Hreiz, and T. C. Lieuwen, Nonlinear interactions between forced and self-excited acoustic
oscillations in premixed combustor, J. Propul. Power 24, 628 (2008).
[57] S. Balusamy, L. K. B. Li, Z. Han, M. P. Juniper, and S. Hochgreb, Nonlinear dynamics of a self-excited
thermoacoustic system subjected to acoustic forcing, Proc. Combust. Inst. 35, 3229 (2015).
[58] Y. Guan, V. Gupta, K. Kashinath, and L. K. B. Li, Open-loop control of periodic thermoacoustic
oscillations: Experiments and low-order modelling in a synchronization framework, Proc. Combust. Inst.
37, 5315 (2019).
[59] S. Mondal, S. A. Pawar, and R. I. Sujith, Forced synchronization and asynchronous quenching of periodic
oscillations in a thermoacoustic system, J. Fluid Mech. 864, 73 (2019).
[60] Y. Guan, W. He, M. Murugesan, Q. Li, P. Liu, and L. K. B. Li, Control of self-excited thermoa-
coustic oscillations using transient forcing, hysteresis and mode switching, Combust. Flame 202, 262
(2019).
[61] A. Sahay, A. Roy, S. A. Pawar, and R. I. Sujith, Dynamics of coupled thermoacoustic oscillators under
asymmetric forcing, Phys. Rev. Appl. 15, 044011 (2021).
[62] M. L. Passarelli, A. Kazbekov, V. Salazar, K. Venkatesan, and A. M. Steinberg, Experimental study of
forced synchronization and cross-coupling in a liquid-fuelled gas turbine combustor at elevated pressure,
Proc. Combust. Inst. 39, 4751 (2023).
[63] L. K. B. Li and M. P. Juniper, Phase trapping and slipping in a forced hydrodynamically self-excited jet,
J. Fluid Mech. 735, 705 (2013).
[64] A. R. Karagozian, Transverse jets and their control, Prog. Energy Combust. Sci. 36, 531 (2010).
[65] S. A. Pawar, R. I. Sujith, B. Emerson, and T. C. Lieuwen, Characterization of forced response of density
stratified reacting wake, Chaos 28, 023108 (2018).

063201-22
 FORCED SYNCHRONIZATION OF GLOBALLY STABLE AND …

[66] M. Balasubramaniyan, A. Kushwaha, Y. Guan, J. Feng, P. Liu, V. Gupta, and L. K. B. Li, Global
hydrodynamic instability and blowoff dynamics of a bluff-body stabilized lean-premixed flame, Phys.
Fluids 33, 034103 (2021).
[67] M.-H. Yu and P. A. Monkewitz, The effect of nonuniform density on the absolute instability of two-
dimensional inertial jets and wakes, Phys. Fluid 2, 1175 (1990).
[68] K. R. Anderson, J. Hertzberg, and S. Mahalingam, Classification of absolute and convective instabilities
in premixed bluff body stabilized flames, Combust. Sci. Technol. 112, 257 (1996).
[69] S. Aydore, D. Pantazis, and R. M. Leahy, A note on the phase locking value and its properties, NeuroImage
74, 231 (2013).
[70] S. Mondal, S. Pawar, and R. I. Sujith, Synchronous behaviour of two interacting oscillatory systems
undergoing quasiperiodic route to chaos, Chaos 27, 103119 (2017).
[71] H. Jegal, K. Moon, J. Gu, L. K. B. Li, and K. T. Kim, Mutual synchronization of two lean-premixed gas
turbine combustors: Phase locking and amplitude death, Combust. Flame 206, 424 (2019).
[72] M. C. Romano, M. Thiel, J. Kurths, I. Kiss, and J. Hudson, Detection of synchronization for non-phase-
coherent and non-stationary data, Europhys. Lett. 71, 466 (2005).
[73] M. C. Romano, M. Thiel, J. Kurths, and C. Grebogi, Estimation of the direction of the coupling by
conditional probabilities of recurrence, Phys. Rev. E 76, 036211 (2007).
[74] S. A. Pawar, A. Seshadri, V. R. Unni, and R. Sujith, Thermoacoustic instability as mutual synchronization
between the acoustic field of the confinement and turbulent reactive flow, J. Fluid Mech. 827, 664 (2017).
[75] M. G. Rosenblum, A. S. Pikovsky, and J. Kurths, Phase synchronization of chaotic oscillators, Phys. Rev.
Lett. 76, 1804 (1996).
[76] N. F. Rulkov, M. M. Sushchik, L. S. Tsimring, and H. D. I. Abarbanel, Generalized synchronization of
chaos in directionally coupled chaotic systems, Phys. Rev. E 51, 980 (1995).

063201-23