The transition from two-dimensional

to three-dimensional waves in falling

liquid films: Wave patterns and
transverse redistribution of local flow
rates
Cite as: Phys. Fluids 27, 114106 (2015); https://doi.org/10.1063/1.4935958
Submitted: 08 May 2015 . Accepted: 28 October 2015 . Published Online: 20 November 2015

S. M. Kharlamov , V. V. Guzanov, A. V. Bobylev, S. V. Alekseenko, and D. M. Markovich

ARTICLES YOU MAY BE INTERESTED IN

Three-dimensional solitary waves on falling liquid film at low Reynolds numbers

Physics of Fluids 17, 121704 (2005); https://doi.org/10.1063/1.2158428

Three-dimensional instabilities of film flows

Physics of Fluids 7, 55 (1995); https://doi.org/10.1063/1.868782

Three-dimensional localized coherent structures of surface turbulence. I.

Scenarios of two-dimensional–three-dimensional transition
Physics of Fluids 19, 114103 (2007); https://doi.org/10.1063/1.2793148

Phys. Fluids 27, 114106 (2015); https://doi.org/10.1063/1.4935958 27, 114106

© 2015 AIP Publishing LLC.

 PHYSICS OF FLUIDS 27, 114106 (2015)

The transition from two-dimensional to three-dimensional

waves in falling liquid films: Wave patterns and transverse
redistribution of local flow rates
S. M. Kharlamov,1,2,a) V. V. Guzanov,1,2 A. V. Bobylev,1,2 S. V. Alekseenko,1,2
and D. M. Markovich1,2
1
Kutateladze Institute of Thermophysics, Siberian Branch of the Russian Academy of Science,
Lavrentiev Ave. 1, Novosibirsk 630090, Russia
2
Novosibirsk State University, Pirogova St. 2, Novosibirsk 630090, Russia
(Received 8 May 2015; accepted 28 October 2015; published online 20 November 2015)

This article presents the results of experimental investigations of the process of tran-
sition from two-dimensional (2D) to three-dimensional (3D) waves in liquid films
falling down a vertical plate. The method of laser induced fluorescence was used to
obtain instant shapes of three dimensional waves and to investigate the regularities
of formation of 3D wave patterns arising due to transverse instability of 2D waves.
The obtained results were compared to the results from the published literature on
the modeling of 3D wave regimes of film flow. Although many details of 3D wave
patterns correspond well, there are a few significant distinctions between our exper-
iments and modeling. In particular, during 2D-3D wave transition, we observed a
strong transverse redistribution of liquid leading to the formation of rivulets on the
surface of isothermal liquid film, which is a phenomenon not described previously.
Possible discrepancies between modeling and experiments, including applicability of
boundary layer models and downstream periodic boundary conditions, are discussed.
The authors hope that the results presented in the article are of interest not only for
modeling of film flows but also for practical applications because at large distances
from the film inlet due to 2D-3D wave transition the local flow rates can differ several
times at the transverse distances of about 1 cm, which is an effect that cannot be
neglected. C 2015 AIP Publishing LLC. [http://dx.doi.org/10.1063/1.4935958]

I. INTRODUCTION
Gravity-driven film flow of a viscous liquid down a flat plane is a well known example of con-
vectively unstable free surface flow with instabilities of different kinds leading to the appearance and
interaction of surface waves with a great variety of characteristics. At moderate Reynolds numbers,
lying between a few tens and a few hundreds, typical downstream evolution of the waves on the
surface of a liquid film falling down a vertical plate looks as follows. Two-dimensional (2D) waves
arise and rapidly develop to the saturated state at the domain close to the film inlet. Saturated 2D
waves travel in a quasi-stationary manner at the distances between a few and a few tens wavelengths.
At this stage, a distinct difference can be observed between the cases of noise-driven wave evolution
and evolution of the waves forced by time-periodic perturbations applied at the inlet. Coarsening of a
wave pattern with downstream distance is observed in the former case,1 whereas periodically forced
2D waves can travel at the distances up to a few tens wavelengths conserving their periodicity.2 But,
in any case, transverse distortions of 2D waves arise and grow with distance, which inevitably leads
to the formation of three-dimensional (3D) waves. An example of described wave evolution is shown
in Fig. 1.

a) Electronic mail: kharlamov@itp.nsc.ru.

1070-6631/2015/27(11)/114106/25/$30.00 27, 114106-1 © 2015 AIP Publishing LLC

114106-2 Kharlamov et al. Phys. Fluids 27, 114106 (2015)

FIG. 1. Shadow image of waves on vertically falling film of water at Re = 25. The case of noise-driven evolution. Distance x
is counted downstream the film inlet.

Regimes with developed 3D waves are assumed to be the final stage of wave evolution on verti-
cally falling films at moderate Reynolds numbers. At this stage, free surface of liquid film is covered by
numerous 3D waves randomly interacting with each other. It is supposed1–6 that in spite of the strong
interaction, the interacting waves can be considered actually as stable three-dimensional coherent
114106-3 Kharlamov et al. Phys. Fluids 27, 114106 (2015)

structures or localized 3D waves with predetermined properties and that the description of developed
regimes of falling film flow can be drastically simplified at moderate Reynolds numbers by using the
concept of coherent structures.
Over the past decade, there have been significant advancements in the description of localized
3D waves on falling liquid films. Scenarios of transition from 2D to 3D waves, steady-state travel-
ing solutions in the form of solitary horseshoe waves and some aspects of interaction between such
waves were described on the base of the Kapitza-Shkadov model in Refs. 4, 5, 7, and 8. Furthermore,
theoretically obtained and experimentally measured8–10 characteristics of the solitary 3D waves were
shown to be in good agreement. Inspired by this agreement between theory and experiment, the au-
thors7,8 proposed to apply their modeling approach to describe fully developed 3D wave regimes on
falling films though recognizing that the applied model have certain deficiencies discussed, among
others, in Refs. 3, 6, and 11. More accurate modeling approaches6,11 and direct numerical simula-
tion12,13 for the cases allowing comparison with available experimental results on 3D waves were
implemented only with the imposition of periodic boundary conditions in both lateral and transverse
directions which makes it difficult to define the domain of applicability of the obtained results to the
real film flows. Concerning the transition from 2D to 3D waves in vertically falling films, the results
of modeling6,11–13 were compared mainly with the experimental findings14 obtained by the method
of shadowgraphy, which is insufficient to reconstruct the wave shape as a whole. As a consequence,
comparison between modeling and experiment leads to ambiguous conclusions. Thus, forming 3D
waves observed in Ref. 14 appear to be horseshoe waves in accordance with modeling,6,7,11 whereas
the same waves look like downstream elongated 3D waves with filled trailing edges in accordance
with computations.12 Elongated 3D waves were also observed in films flowing at moderate Reynolds
numbers in experiments15,16 and were referred as streak-like waves. The concept of both horseshoe
and streak-like waves is shown in Fig. 2.
It should be noted that quantitative characteristics of stationary horseshoe waves in vertically
falling films were obtained on the base of different modeling equations in Refs. 5, 17, and 18 and also
were measured in the experiments7–10,19 using special methods of the waves excitation, whereas exper-
imentally observed streak-like waves were described mainly in a qualitative manner. Consequently,
the horseshoe waves, also referred to as Λ-solitons,5,7,8 are considered to be the main candidate for
the role of three-dimensional coherent structures determining the principal traits of the developed
falling film flows at moderate Reynolds numbers. At the same time, there is a lack of experimental
information on the regularities of 3D waves formation in the course of 2D waves decaying through
a mechanism of transverse instability. In particular, it is not known whether or not horseshoe waves
can arise in such a process and if not, how experimentally observed 3D waves actually do arise. It
was shown recently20 for the case of inclined film flow that formation of the wave patterns similar to
those observed in vertically falling films is associated with long-wave mode of transverse instability
and occurs only for sufficiently large angles of inclination, whereas at small angles of inclination the

FIG. 2. Horseshoe wave (a) and streak-like wave (b). The horseshoe wave has curved crest with elongated trough behind
it with film thickness in the trough smaller than thickness of undisturbed film. Streak-like wave is elongated in the flow
direction x and has filled trailing edge.
114106-4 Kharlamov et al. Phys. Fluids 27, 114106 (2015)

short-wave instability and principally different wave patterns are realized. But although the authors
of Ref. 20 used synthetic Schlieren method21 allowing reconstruction of 3D form of waves, they
investigated primarily the initial stage of 3D wave patterns formation and did not consider the form
of 3D waves in detail. In our recent short communication,22 we have shown in a rather brief manner
that 3D waves forming during 2D–3D transition have no horseshoe form and that such transition
is accompanied by strong redistributions of fluid in transverse direction. In this work, we present
detailed experimental results on 2D–3D wave transition in vertically falling films with a focus on the
traits of this phenomenon that have previously escaped notice.
The outline of the present work is as follows: In Sec. II, we describe the experimental setup and
measurement technique. In Sec. III, we present the results on initial stage of 2D–3D transition for
the case of the periodically forced 2D waves. These results are supplementary to those presented in
Refs. 7 and 14 and mostly concern the features of 2D–3D transition not considered previously. In
Sec. IV, we give brief insight into wave patterns arising at the large distances downstream film inlet.
Finally, in Sec. V, we give concluding remarks.

II. EXPERIMENTAL METHOD

A. Experimental setup
The sketch of experimental setup is shown in Fig. 3. Experiments were carried out on a vertical
glass plate with dimensions 20 cm in transversal and 25 cm in vertical directions. Liquid film was
formed by a slot distributor with adjustable slot width. The method of laser induced fluorescence
(LIF) was used to measure instantaneous local film thickness on the area of 12 × 12 cm with spatial
resolution 0.125 mm.

B. Method LIF
The LIF method is based on the reconstruction of local film thickness in accordance with local
light intensity emitted by a small amount of fluorescent dye dissolved in working liquid and is quite

FIG. 3. Experimental setup. (1) CCD camera; (2) laser; (3) scattering fluorocarbon film; (4) expanding lens; (5) glass plate;
(6) slit distributor; (7) liquid film. White arrows show flow direction.
114106-5 Kharlamov et al. Phys. Fluids 27, 114106 (2015)

similar to the method of “fluorescent imaging” originally introduced to study wave dynamics on in-
clined films.23 The main difference between the LIF and fluorescent imaging methods is that in the
former coherent laser light is used for fluorescence excitation and additional measures should be taken
to suppress speckles in illuminated area. As a simple remedy from the speckles, we placed scattering
film made of porous fluorocarbon between the laser and expanding lens (Fig. 3). This technique allows
us to obtain speckleless LIF images of liquid films with appropriate spatial resolution (at least in the
range between 60 and 250 µm per pixel).
In our experiments working liquids were doped with small additions, about 30–40 mg/l, of fluo-
rescent dye Rhodamine 6G. Prepared solutions of the dye absorb green and emit orange light. Pulsed
Nd:YAG laser with a wavelength of 532 nm, pulse energy 50 mJ, and duration 10 ns was used as a light
source for fluorescence excitation. Low noise CCD camera (Kodak MegaPlus ES 1.0) with low-pass
optical filter (>550 nm) was synchronized with the laser and used for the registration of fluorescent
images of falling films. Duration of the laser pulse determines the time of the image exposure and
correspondingly the sampling time of our system. To minimize optical distortions arising due to wavy
free surface of the film, we arranged the light source and the camera at the dry side of the glass plate,
i.e., excitation and registration were carried out through glass. Our preparatory tests performed with
flat layers of different fluids, mostly water-alcohol and water-glycerol solutions (WGSs) of different
concentrations, showed that for the LIF system layout depicted in Fig. 3, the relation between local
intensity of fluorescent image J(x, z) and local film thickness h(x, z) can be described by the following
expression:
J(x, z) = C(x, z) · (1 − exp(−α · h(x, z))) × (1 + K · exp(−α · h(x, z))) × (1 + K) + D(x, z), (1)
where α is absorption coefficient for exciting light, K is light reflection index from free boundary of
liquid layer, D(x, z) is dark level of the CCD camera, and C(x, z) is correcting matrix introduced to
compensate the non-uniformity of lighting across the measurement area and difference in sensitivity
of camera pixels. Eq. (1) can also be obtained analytically assuming that incident and re-emitted light
propagate normal to the liquid layer and neglecting absorption of re-emitted light. Both α and D(x, z)
were measured before and after each experimental run. To measure absorption, we used optical cu-
vettes of different path length. When α and D(x, z) are known, C(x, z) can be obtained by measuring
intensity of fluorescent image J(x, z) for liquid layer with fixed thickness. For the case of flat layers,
the error in determination of h in accordance with the described procedure was 2%–3% in the range
0.05 mm ≤ h ≤ 1 mm, which exceeds the range of film thicknesses realized in our experiments.
Reconstruction of the correcting matrix C(x, z) during experiments was performed in situ at
low liquid flow rates under assumption that at low Reynolds numbers, when flowing film is waveless
in the measurement area, its thickness is equal to the Nusselt thickness hN = (3qν/g)1/3, where q is
volumetric flow rate per unit of film width, ν is kinematic viscosity, and g is gravity acceleration.
Taking into account the errors in the measurements of liquid flow rate (1%) and viscosity (1%), the
total error in the determination of film thickness was estimated to be no more than 5%.
Additional measurement error can be attributed to the fact that Eq. (1) is valid for the case of flat
film, whereas we study film flows with undulated free boundary. For the used layout of measuring sys-
tem, this error arises due to deviations in intensity of reflected light because of its focusing/defocusing
under curved free boundary as well as due to the dependence of reflection index on the inclination
angle of free boundary. Approximate estimate using geometrical optics approach shows that in all
cases when this error has significance to our experiments, it leads to the appearance of bright spots
on the LIF images and consequently to the appearance of sharp spikes in the field of film thickness.
When the spikes are absent this error can be neglected.

C. Investigated regimes
In this paper, we investigate transition to 3D waves only for the case of initially regular 2D waves
with additional restriction that regularity retains for the whole measurement area. This restriction was
caused by limitation of our measurement system, which allows making consecutive measurements
in time intervals not less than 1/15 s. For this reason, we used stroboscopic technique which makes it
114106-6 Kharlamov et al. Phys. Fluids 27, 114106 (2015)

TABLE I. Frequency bands for which regular wave patterns retain in the
whole measurement area.

Water WGS

Re Range of F (Hz) Re Range of F (Hz)

19 10–17 14 11–12
25 11–17 25 15–17
31 13–19 36 16–17
40 13–22 47 16–17
50 15–23 ... ...
65 15 ... ...

possible to increase temporal sampling rate while studying periodic processes. For each forcing fre-
quency F, we set a frequency of measurement repetition F1 in such a way that maximal displacement
of waves between successive measurements was less than half wavelength of capillary ripples which
determine the maximal frequency in the wave spectrum. The results of successive measurements at
each point of measurement area were interpreted as obtained for the field of film thicknesses h(x, z, t)
with a temporal sampling rate satisfying the sampling theorem.
The regular 2D waves with prescribed frequency F were forced by harmonic modulation of liquid
flow rate. Distilled water with density ρ = 998 kg/m3, kinematic viscosity ν = 0.994 × 10−6 m2/s, sur-
face tension σ = 0.073 kg/s2, and WGS with ρ = 1110 kg/m3, ν = 3.9 × 10−6 m2/s and σ = 0.073 kg/s2
were used as working fluids. In the process of transition from 2D to 3D waves, the regularity of wave
patterns was retained in narrow bands of frequencies shown in Table I.
Uniformity of flow rate across the plate and the effects of side walls was checked by the form of
the crest line of 2D waves. In the case when flow rate is distributed uniformly across the plate, the crest
line of the waves at the flow inlet is straight and perpendicular to the flow direction. With downstream
propagation of the waves, the sidewall effect begins to declare itself similarly to the case of inclined
film flow.24,25 The waves begin decelerate and distort near the sidewalls and this distortion grows out
with downstream distance towards the centerline of the plate. In our experiments, the growth rate
of such distortions was larger in water films and in the worst case they affected the flow no further
than 1.5 cm from the sidewalls at the bottom of the plate. The distance between the sidewalls and
lateral boundaries of the measurement area was 4 cm, so we suppose that the sidewall effect can be
disregarded.
In this article, we study both the natural and forced scenarios of 2D-3D transition. In the first case,
the transition occurs due to transverse instability of 2D waves and in the second case, we used the
regular combs of needles that touched the surface of a liquid near the film inlet to destruct 2D waves
in the manner described in Refs. 7 and 14. It is supposed that forced destruction makes it possible to
skip the initial stage of 2D-3D transition and get an insight into the wave patterns arising at the large
distances downstream.
In addition, in Sec. IV, we give brief insight on wave patterns arising at the large distances (up
to 130 cm) downstream the film inlet without forced destruction of 2D waves. To obtain this infor-
mation, we have carried out additional experiments on the vertical plate of large dimensions. Since
the wave patterns are always irregular far downstream the film inlet, even in the case of initially
regular 2D waves, we have used for their registration another type of LIF system equipped with a
high-speed camera. For the sake of consistency, more detailed description of this experimental setup
and measuring system is given in Sec. IV.

D. Dimensionless parameters and results representation

There exist several sets of dimensionless parameters which are usually used to characterize film
flows. Useful comments on the choice of different independent dimensionless groups for different
regimes of film flow can be found among others in Refs. 1, 2, 6, 26, and 27. In this article, we use
114106-7 Kharlamov et al. Phys. Fluids 27, 114106 (2015)

FIG. 4. Formation of regular 2D waves in the upper region of water film flow. Re = 40, F = 13 Hz. 1—Nusselt thickness hN.
2—Substrate thickness hs.

Nusselt thickness hN as a measure of spatial scale, Reynolds number Re = q/ν as a measure of liquid
flow rate. It is a common practice to characterize physical properties of fluids in films by dimension-
less Kapitza number Ka = σ ρ−1 ν −4/3 g−1/3. In our experiments Ka = 3660 for water and Ka = 494
for WGS. In addition, to compare our results with theoretical findings,4 it is useful to introduce the
Reynolds number of substrate film Res. The concept of substrate film or residual layer (see Fig. 4) is
helpful in the case of low frequency waves when evolution of each wave can be considered separately
as the evolution of single solitary wave on thin liquid layer (substrate) with thickness hs sufficiently
lower than hN.
Experimentally measured values of hs were used to obtain the dependency between Res and Re
(see Fig. 5).

FIG. 5. Dependency between Res and Re. 1—Water film. 2—WGS film. Error in determination of Res is about 15%.
114106-8 Kharlamov et al. Phys. Fluids 27, 114106 (2015)

For the case of solitary 2D wave propagated on residual layer characterized by sufficiently large
Reynolds number Demekhin et al.4 obtained asymptotic values for maximal growth rate αm of trans-
verse modulations, transverse wave length of most growing mode λ m , and transverse wave length of
neutral mode λ 0 which in accordance with Eq. (20) of Ref. 4 can be rewritten in dimensional form as
( g ) 1/3
α m ∼ 0.0214Res −1/9Ka−1/3 2 ,
ν
( 2 ) 1/3
ν
λ m ∼ 19.66Res 1/9Ka1/3 , (2)
g
( 2 ) 1/3
ν
λ 0 ∼ 8.53Res 1/9Ka1/3 .
g
In our experiments, all measured values of Res are sufficiently large to apply asymptotic values of
Eq.(2) for the water films. For the case of WGS films, this solutions are applicable at Re > 30. We
will use these asymptotic solutions to compare with the results described in Sec. III.
To reduce the quantity of uniform graphics, the obtained results are illustrated on examples of
water films with short comments on the difference or likeness between the cases of water and WGS
flow. The data for graphical representation are filtered by a low-pass digital filter which can diminish
the amplitudes of capillary precursor up to 10% in the worst cases.

III. INITIAL STAGE OF 2D–3D TRANSITION

A. Natural 2D-3D transition
We observed natural formation of 3D wave patterns at Re > 30 in water film flow and at Re > 12
in the case of WGS film. For smaller Reynolds numbers, the crest line of 2D waves remained straight
in the whole area of observation with no evident signs of transition to 3D regimes. An example of 3D
wave patterns formation is shown in Fig. 6.
As can be seen, in the process of 2D-3D transition, the growing transverse modulation of 2D
waves results in the formation of running forward leading humps with horseshoe curved leading edges

FIG. 6. Natural formation of 3D structures on the surface of water film. (a) LIF image of the waves. (b) and (c) Wave patterns
for the regions 1 and 2 depicted in (a). Re = 40, F = 19 Hz.
114106-9 Kharlamov et al. Phys. Fluids 27, 114106 (2015)

FIG. 7. Main features of forming 3D wave patterns. A—amplitude of developing transverse modulations of 2D waves. λx is
streamwise wavelength and λz is wavelength of transverse modulations.

and trailing humps with flatter leading edges. This terminology for describing the forming 3D struc-
tures was introduced by Dietze et al.13 The main features of forming wave patterns are sketched in
Fig. 7.
In our experiments, we observed only synchronous scenario28 of 3D wave patterns evolution,
when transverse modulations of successive 2D waves are in phase. One more interesting trait is that
as leading humps develop the wakes are formed behind them with film thickness in the wakes greater
than in the residual layer behind the trailing humps. Such wakes are clearly seen in Fig. 6(c). Despite
their small amplitudes not exceeding 5–10 µm, these wakes can promote formation of synchronous
3D patterns by disturbing following 2D waves and evoke formation of following trailing humps at
the same places. Note that formation of synchronous 3D patterns on falling liquid films can also be
found, for example, in Fig. 4 of Ref. 29 and in Fig. 6 of Ref. 30. In addition, Adomeit and Renz15
noted when describing 3D waves in developed 3D wavy film flows that “due to wake effects the waves
tend to line up.”

1. Evolution of the transverse modulations amplitude

Evolution of the transverse modulations amplitude A with downstream distance is shown in Fig. 8
for the case of water film flow. At the initial stage, the leading humps gradually run away from the

Re=65, F=15 Hz

Re=50, F=15~22 Hz

Re=40, F=13~16 Hz

Re=40, F=19 Hz

Re=40, F=22 Hz

FIG. 8. Downstream evolution of the transverse modulations amplitude for water films. 1—Re = 40, F = 22 Hz, 2—Re = 40,
F = 19 Hz, 3—Re = 40, F = 13—16 Hz, 4—Re 50, all F except 23 Hz, and 5—Re = 65, F = 15 Hz.
114106-10 Kharlamov et al. Phys. Fluids 27, 114106 (2015)

trailing humps. The amplitude A grows exponentially A ∼ exp(αx) up to the values A/λx ≈ 0.1–0.15,
where λx is the wave length of 2D waves. Farther downstream growth of the amplitude gradually
diminishes and for some cases after reaching a maximum value the amplitude starts decreasing since
the leading humps are slowing down and trailing humps start to overtake them. For the case of WGS,
no saturation of A was observed in the measurement area.
At Re = 40, growth rate α increases with the decrease of excitation frequency and becomes
frequency-independent in the range 13–16 Hz. At Re = 50, α is frequency-independent for almost
all investigated frequencies except 23 Hz, in the range 15–22 Hz. Measured frequency-independent
value of α = 0.21 ± 0.01 cm−1. Frequency independence of the measured growth rate in the region
of low frequencies as well as the above-mentioned fact that in our experiments with water films Res
is always sufficiently large to apply asymptotic solutions,4 make it appropriate to compare α and
maximal growth rate α m . For water film flow, Eq. (2) gives α m ≈ 0.23 cm−1. In the case of WGS,
flow for all investigated regimes α = 0.18 ± 0.01 cm−1, whereas α m ≈ 0.19 cm−1. Thus, the measured
values of α agree well with theoretically predicted4 values of maximal growth rate for both working
liquids.

2. Transverse redistribution of liquid

Development of 3D structures downstream is accompanied by gradual decrease in the amplitude
of trailing humps in the regions adjacent to leading humps (Fig. 6(c)), which may indicate that some
transverse reflux of the liquid occurs. Time averaged fields of the film thickness,

−1
i=N
⟨h(x, z)⟩t = (1/N) h(x, z,t + i∆t), (3)
i=0

where N∆t = 1/F and N is the number of samplings per period, indicates that the transverse redistri-
bution of liquid actually takes place. As can be seen in Fig. 9, such representation reveals formation
of “rivulets” on the film surface, with ⟨h⟩t decreasing downstream in x-y sections through the regions
adjacent to the leading humps and increasing in the sections through central parts of the leading
humps.
The downstream evolution of wave form along the lines I, II, and III indicated in Fig. 9 is shown
in Fig. 10. Note that the values of film thicknesses averaged over spatial period of the waves ⟨h⟩ λ
agree well with ⟨h(x, z)⟩t in different parts of the film flow.
Whereas at Re = 40 water gradually refluxes from the regions adjacent to leading humps and
accumulates mainly in the regions crossed by them, at increased flow rates sufficient accumulation
of water is also observed in the regions crossed by trailing humps as shown in Fig. 11.
For the regimes studied, time averaged value of film thickness can vary along z-direction in the
lower part of the plate up to 30%–40% over distances comparable to transverse size of rivulets Lr (see
Fig. 11(e)). It seems that Lr is independent of F and Re and in all cases it was equal to 10.5 ± 0.5 mm
for the water films and 13 ± 1 mm for the WGS films. Transverse redistribution of liquid must be
accompanied by transverse redistribution of its time averaged local flow rates ⟨q(x, z)⟩. Rough esti-
mate under the assumption ⟨q(x, z)⟩t ∼ ⟨h(x, z)3⟩t shows that averaged local flow rates under the ridges
of rivulets can be several times greater than flow rates at their peripheries. In the case of low Re, when
2D waves stay undistorted in the measurement area, no formation of rivulets is observed and the fields
of ⟨h(x, z)⟩t are flat.

B. Forced scenarios of 2D-3D transition

In the case of forced destruction of 2D waves by a regular comb of needles, 3D patterns consisting
of leading and trailing humps arise downstream the comb in such an arrangement that the centers of
leading humps have the same transverse positions as the needles have and the fields of time averaged
film thickness behind the comb always show the presence of rivulets downstream the needles. In the
experiments, the needles touched the surface of liquid 1 cm downstream the film inlet. Characteristic
114106-11 Kharlamov et al. Phys. Fluids 27, 114106 (2015)

x=15
x=11
x=20

FIG. 9. Time averaged thickness of water film. Re = 40, F = 19 Hz. (a) ⟨h(x, z)⟩t for the whole measurement area shown
in Fig. 6(a), (b) transverse y-z sections for (a) at the downstream distances 1–11 cm, 2–15 cm, 3–20 cm. (c) downstream
x-y sections for (a) along the lines indicated in (b): I (along the ridge, i.e., through the centers of leading humps), II (along
the trough), and III (through the centers of trailing humps). Open symbols correspond to the film thicknesses averaged over
spatial period of the waves along the same lines (see Fig. 10).

level of film thickness distortions inserted by the needles is shown in Fig. 12. As can be seen, the
needles imposed on the film surface “positive” perturbations in terms of Ref. 14.
For the Reynolds numbers at which natural formation of 3D patterns was observed, we investi-
gated the effect of the distance between the needles on formation of 3D waves. At these experiments,
the distance between neighboring needles Ln was varied in the range from 0.5 up to 5.5 cm with
the step 0.5 cm. For the case of water film, no influence of the needles on formation of 3D patterns
was observed at Ln ≤ 1.0 cm, which coincides with observations.7,8,14 An example how Ln affects
formation of 3D patterns at Ln ≥ 1.5 cm is shown in Fig. 13.
As can be seen, with growth of Ln additional leading humps appear between the ones induced
by needles in such a manner that transverse distance λz between leading humps varies in the range
1.1 < λz < 2.1 cm for the water films regardless of the distance between the needles. For the case
of WGS, additional leading humps appear in the same manner as for the case of water and λz varies
in the range between 1.5 and 2.5 cm. In accordance with Eq. (2), transverse wave lengths of most
114106-12 Kharlamov et al. Phys. Fluids 27, 114106 (2015)

FIG. 10. Downstream evolution of the wave form along the lines I, II, and III for the case shown in Fig. 9. Corresponding
values of ⟨h⟩λ in millimeters for the upper and lower regions of the measurement area are indicated by bold numerals at the
left and right sides of each inset.

growing and neutral modes for water are, respectively, λ m ≈ 1.9 cm and λ 0 ≈ 0.8 cm and for WGS
λ m ≈ 2.3 cm and λ 0 ≈ 1.0 cm. Thus, despite the transverse scales of initial perturbation of 2D waves,
all experimentally observed transverse scales of forming 3D patterns for both working liquids fall in
the narrow band between values slightly greater than λ m and values approximately 1.5 times greater
than λ 0.
One more interesting trait of forming 3D patterns, which is seen in Fig. 13(b), is forming of
V-shaped indentation on trailing hump, which was discussed recently13 with open question whether
such shape is related to the imposition of constant λz in the simulations or its absence in the previous
experiments is connected with some difference between simulated and experimentally implemented
flow conditions. As can be seen, such indentation is forming in our experiments at Ln = 2 cm,
which coincides with λz in the simulations,14 but, in the experiments, its formation shows oscillatory
pattern and V-shaped form of trailing humps almost completely faded away a few wavelengths
downstream after its formation. Note also that with growth of Ln such V-shaped indentations develop
in additional leading humps. Apparently, these traits of the indentations evolution complicate their
observation in experiments. Detailed form of the film surface in the region of the indentation framed
in Fig. 13(b) is shown in Fig. 14. In addition, the field of curvatures K and the field of pressure
under film surface P = −σK for this region are also shown in Fig. 14. The curvature K is defined
as2
(1 + h2z )h x x − 2h x hz h x z + (1 + h2x )hz z
K= , (4)
(1 + h2x + h2z )3/2
where h = h(x, z) and values with subscript represent partial derivatives. To obtain smooth fields
shown in Figs. 14(b) and 14(c) additional low-pass filtering of the experimental data was performed.
Though additional filtering reduces maximal values of K and P approximately two times, from
Fig. 14(c) can be seen that V-shape of indentation is associated with saw-tooth transverse profile of
pressure in front of trailing hump.
Though qualitative characteristics of 3D wave patterns evolving due to forced destruction of 2D
waves are to some extent dependent on forcing conditions, there exist common traits of the patterns
evolution, which will be described in short below.
At low Reynolds numbers, when no natural transition to 3D waves is observed, trailing humps
either gradually overtake leading humps and in this case transverse liquid fluxes are directed towards
114106-13 Kharlamov et al. Phys. Fluids 27, 114106 (2015)

FIG. 11. Natural formation of 3D structures on the surface of water film at Re = 50 and F = 17 Hz. (a) LIF image of the waves.
(b) and (c) Wave patterns for the regions 1 and 2 depicted in (a). (d) ⟨h(x, z)⟩t for the region 1.5 ≤ z ≤ 6.5 cm. (e) transverse
y-z sections for (d) at the downstream distances from the film inlet 1–11.5 cm, 2–16 cm, 3–21.5 cm.

leveling liquid distribution over plate as shown in Fig. 15 or they move with the same celerity as
leading humps and in this case non-uniformity of liquid distribution remains unchanged downstream
as shown in Fig. 16.
At larger Reynolds numbers, two scenarios of the patterns evolution were observed. For low
frequencies of 2D waves excitation, the leading humps initially run away from the trailing humps.
Then their amplitudes and velocities decrease so that the trailing humps which move all the time with
approximately constant celerity begin to overtake the leading humps. As the trailing humps approach
the leading ones, the latter begin to restore their amplitudes and the pattern structure almost repeats
itself through several wave lengths. An example of such oscillatory scenario obtained previously in
modeling6,13 and observed in experiments22 is shown in Fig. 17(a). For high excitation frequencies, the
site between trailing humps, which was left by running downstream leading hump, is filled by another
leading hump that arrives from above so that the pattern structure also almost repeats itself through
several wave lengths as shown in Fig. 17(b). We will refer to such scenario as “passing through”
one.
114106-14 Kharlamov et al. Phys. Fluids 27, 114106 (2015)

FIG. 12. Film thickness distribution 2 cm downstream the needles. Water film flow at Re = 50. The arrows show positions of
the needles.

To characterize the three-dimensionality of periodic wave patterns on liquid films their energies
of deformation, Ex and Ez are often used. These values are defined as quadratic sums of spatial Fourier
coefficients for film thickness obtained over the period of the wave pattern in streamwise and trans-
verse directions.6,11,13,31 For experimental data processing, it is useful to redefine the deformation
energies in spatial domain as follows:
Nz  N x 1/2
1   
2
E x (x) = h m, j − ⟨h⟩ j  ,

Nx Nz j=1 m=1 

Nx
1 
⟨h⟩ j = h m, j ,
Nx m=1
(5)
N x  N
 z  1/2
1 
Ez (x) = (hn,l − ⟨h⟩l )  ,
2
Nx Nz l=1  n=1 

Nz
1 
⟨h⟩l = hn,l ,
Nz n=1

where Nx and Nz are the numbers of experimental points of film thickness over streamwise and trans-
verse periods of 3D structure, respectively, and coordinate x is defined as a midpoint between the
coordinates of two successive trailing humps. In accordance with Parseval’s theorem, this definition
of the deformation energies is equivalent to their definition in Fourier space and allows to compare our
results with those obtained by summation of spatial harmonics. The energies of deformation for the
cases in Fig. 17 are shown in Fig. 18. As can be seen, both scenarios demonstrate oscillatory dynamics
in terms of Ex and Ez with levels of oscillation about 30%–40%. For comparison, the energies of
deformation for the case of natural formation of 3D waves (see Fig. 11) are also shown in Fig. 18. As
can be seen, in the bottom part of the flow, where the amplitude of transverse modulations for natural
3D patterns reaches the maximum, Ex and Ez for forced and natural scenarios agree well, which may
be indicative that forced destruction of 2D waves does not affect the main regularities of 3D waves
formation.
In the case of WGS film for all flow rates only “passing through” scenario was observed.
114106-15 Kharlamov et al. Phys. Fluids 27, 114106 (2015)

FIG. 13. Influence of the distance Ln between neighboring needles on 3D wave patterns. Water film flow at Re = 50 and
F = 17 Hz. The very left and the very right leading humps on each inset are induced by neighboring needles. Ln is shown
above the insets. For the framed region with V-shaped indentation, the detailed structure of film thickness is shown in Fig. 14.

IV. THREE DIMENSIONAL WAVE PATTERNS AT LARGE DISTANCES DOWNSTREAM

THE FILM INLET
In this section, preliminary results on 3D wave patterns at large distances downstream the film
inlet for the case of water flow will be shortly presented. The experiments were performed on 4 mm
thick glass plate with dimensions 45 cm in transversal and 140 cm in vertical directions at Reynolds
numbers within the range 4 < Re < 140.
The wave structure over the whole area of the plate was monitored using shadowgraphy, for which
the plate was illuminated from the side covered by the flowing film. The shadow image was formed
on a white screen behind the plate. It should be noted that different elements of the wave structure
can be well recognized by their shadows at different distances between the plate and the screen. At
short distances, the structure of a capillary precursor in the front of the large waves is clearly seen
and the shadow images are similar to those presented in Refs. 14, 29, and 30. At greater distances,
114106-16 Kharlamov et al. Phys. Fluids 27, 114106 (2015)

FIG. 14. Detailed structure of fields in the region of V-shaped indentation (framed in Fig. 13(b)). (a) Film thickness.
(b) Curvature of the film surface. (c) Pressure under film surface.

the shadow images of the elements of waves with much smaller surface curvature become visible.
To obtain the images presented in Fig. 1 and Fig. 19, the screen was positioned at a distance of 7 cm
behind the plate. In this case, the structure of capillary precursor is indistinguishable but the humps
of 3D waves and the wakes behind them are clearly seen.
Shadow images shown in Figs. 1 and 19 were obtained for pure water. When water is tinted
with fluorescent dye the wakes behind forming leading humps are clearly seen on the film surface by
unaided eye.
To reconstruct 3D form of liquid film, we used a high-speed LIF system32 composed of a contin-
uous green laser and high-speed camera. The LIF images were recorded for 13 × 13 cm2 areas with a
spatial resolution of 0.132 mm per pixel at various distances from the liquid inlet with the frame rate
of 1000 fps. This technique allows investigating 3D wave regimes arising both in the case of forced
regular 2D wave evolution and in the case of natural wave evolution. The time averaged fields of film
thickness ⟨h(x, z)⟩t were obtained for the records with the length 1–2 s. In all cases, the maximum
fluctuations of averaged values with change of averaging time did not exceed 1.5% when this time
was greater than 0.5 s.
Transition to 3D wave regimes was observed for Re > Re1, where 5 < Re1 < 10. At Re < Re1
the transverse dimensions of the waves were much greater than their longitudinal dimensions at all
distances from the film inlet (see Fig. 20). In this case, no rivulets on the fields of ⟨h(x, z)⟩t were
observed and the distribution of liquid over the plate surface remained uniform at all downstream
distances.
114106-17 Kharlamov et al. Phys. Fluids 27, 114106 (2015)

FIG. 15. Evolution of 3D wave patterns at Re = 30, Ln = 2 cm, and F = 13 Hz. (a) LIF image of the waves. (b) ⟨h(x, z)⟩t for
the measurement region. (c) Transverse y-z sections for (b) in upper (1), middle (2), and bottom (3) parts of the measurement
region.

At Re > Re1, a transition to 3D wave regimes occurs, so that 2D waves decay into 3D waves with
characteristic longitudinal and transverse dimensions about 1–2 cm. For such flow regimes, formation
of rivulets can be observed on the time-averaged fields of film thickness. For Reynolds numbers lying
in the range Re1 < Re < Re2, where 40 < Re2 < 60, a significant difference between the natural wave
evolution and regular wave evolution is observed over the entire flow length as shown in Fig. 19. For
such Reynolds numbers, the rivulets formation has the most pronounced character. At Re > Re2, a
clear difference between the two scenarios of wave evolution is observed only at small, up to about
ten wavelengths, distances from the film inlet and the development of rivulets is non-monotonic, but
a detailed description of such regimes lays beyond the scope of this article.
In the case of Re < Re2, most sufficient difference between natural and forced scenarios of wave
evolution is observed at the distances up to 50–60 cm downstream the liquid inlet. For the forced
waves, 2D–3D wave transition occurs in the same manner as described in Sec. III for the case of film
flowing down the plate of small dimensions. Evolution of transverse modulations of 2D waves leads
to the formation of leading humps behind which the wakes with increased film thickness arise. Such
wakes are clearly seen as light vertical streaks in Fig. 19. In the region where leading humps start to
appear, the formation of rivulets can be observed on the time-averaged fields of film thickness. For
the case of forced 2D waves presented in Fig. 19(b), the leading humps with streak-like wakes as well
as rivulets appear at a distance of 20–25 cm downstream the film inlet and at the distances 30–40 cm
the rivulets are clearly seen as shown in Figs. 21(d) and 21(e). In the case of natural wave evolution
(Fig. 19(a)) leading humps with streak-like wakes behind them arise sufficiently farther downstream
than in the case of forced 2D waves, and in the region of film flow where crest lines of natural waves
are already strongly distorted but leading humps are still absent, no transverse redistribution of liquid
occurs and time-averaged fields of film thickness are flat as shown in Figs. 21(a) and 21(b). For this
case, leading humps with streak-like wakes as well as rivulets on the time-averaged field of film
thickness begin to appear at the distances between 40 and 50 cm.
114106-18 Kharlamov et al. Phys. Fluids 27, 114106 (2015)

FIG. 16. Evolution of 3D wave patterns at Re = 19, Ln = 2 cm, and F = 11.5 Hz. (a) LIF image of the waves. (b) ⟨h(x, z)⟩t for
the measurement region. (c) Transverse y-z sections for (b) in upper (1), middle (2), and bottom (3) parts of the measurement
region.

Despite the considerable differences of the wave patterns in the upstream part of the flow, the
characteristic dimensions and shapes of 3D waves at large distances from the film inlet are almost
identical for both scenarios of wave evolution. In the case of natural wave evolution, 3D waves in the
lower part of the flow also line up in vertical chains, although not as regular as those observed for the

FIG. 17. Scenarios of 3D wave patterns evolution. Re = 50 and Ln = 2 cm. (a) F = 17 Hz, oscillatory pattern. (b) F = 20 Hz,
passing through pattern.
114106-19 Kharlamov et al. Phys. Fluids 27, 114106 (2015)

FIG. 18. Energies of deformations for water film flow at Re = 50. 1—For the case Ln = 2 cm and F = 17 Hz (see Fig. 17(a)).
2—For the case Ln = 2 cm and F = 19 Hz (Fig. 17(b)). 3—For the case of natural evolution and F = 17 Hz (Fig. 11).

case of forced waves (see Fig. 19). As a result, at large distances from the film inlet the rivulets are
observed for both scenarios of wave evolution as shown in Fig. 22.
At the same time, while the transition to 3D waves of forced 2D waves leads at large distances
to the formation of rivulets with almost completely suppressed wave motions in the regions between

FIG. 19. Shadow images of waves on water film flowing down the large plate at Re = 14. (a) The case of natural evolution.
(b) The case of forced 2D waves with F = 6 Hz. Transition from pattern (a) to pattern (b) and back occurs in a few seconds
after switching on or switching off forcing of 2D waves.
114106-20 Kharlamov et al. Phys. Fluids 27, 114106 (2015)

FIG. 20. A wave pattern at the distances between 120 and 130 cm downstream the film inlet at Re = 5 for the case of natural
wave evolution.

FIG. 21. Characteristics of water film flow at Re = 14 at the distances between 30 and 40 cm downstream the film inlet. Left
panels—the case of natural wave evolution. (a) ⟨h(x, z)⟩t, (b) y-z section for (a) going through points 1 and 2, (c) time traces
of film thickness for points 1 and 2. Right panels—the case of forced wave evolution with F = 6 Hz. (d) ⟨h(x, z)⟩t, (e) y-z
section for (d) going through points 1 and 2, (f) time traces of film thickness for points 1 and 2.
114106-21 Kharlamov et al. Phys. Fluids 27, 114106 (2015)

FIG. 22. Characteristics of water film flow at Re = 14 at the distances between 120 and 130 cm downstream the film inlet.
Left panels—the case of natural wave evolution. (a) ⟨h(x, z)⟩t, (b) y-z section for (a) going through points 1 and 2, (c) time
traces of film thickness for points 1 (on the periphery of a rivulet) and 2 (on the ridge of a rivulet). Right panels—the case of
forced wave evolution with F = 6 Hz. (d) ⟨h(x, z)⟩t, (e) y-z section for (d) going through points 1 and 2, (f) time traces of film
thickness for points 1 and 2.

FIG. 23. A wave pattern at the distances between 120 and 130 cm downstream the film inlet at Re = 14 for the case of natural
wave evolution.
114106-22 Kharlamov et al. Phys. Fluids 27, 114106 (2015)

FIG. 24. Time averaged local flow rates in the region of a full-flowing rivulet for the case shown in Fig. 19. The downstream
distance from the film inlet is 120 cm. (1) The case of forced 2D waves. (2) The case of natural wave evolution after switching
off the forcing.

them (see Fig. 22(f)); in the case of natural wave evolution, the wave motion in the regions between
rivulets exhibits intermittency between the suppressed wave motion and the passage of wave trains
(see Fig. 22(c)).
In all cases, 3D waves observed at large distances downstream the film inlet have characteristic
transverse dimensions of about 1–2 cm and resemble the streak-like waves more than the horseshoe
waves, as can be seen in Fig. 23.
Formation of rivulets appears to be not the final stage of 3D wave evolution at least in the case
of forced regular waves. For all investigated regimes, we observed that a part of neighboring rivulets
gradually merge and form more full-flowing rivulets with more strong redistribution of local flow
rates. Such events of rivulet merging are clearly seen in Fig. 19(b). As an example, the results of
direct measurements of time averaged local flow rates in the region of full-flowing rivulet by a water
sampler with transverse sampling length of 4 mm are shown in Fig. 24. This example shows how
strong the transverse nonuniformity of local flow rates at the distances comparable with rivulet width
can be. But the investigation of downstream evolution and merging of rivulets is yet to be done.

V. CONCLUSIONS
We have investigated experimentally the process of transition from 2D to 3D waves in liquid films
falling down a vertical plate with a primary goal to examine to which extent the modern modeling
approaches adequately describe real film flows. In the first place, we were interested to what extent
the shape of forming 3D waves is alike to the shape of Λ-solitons, which is supposed to be the main
candidates for the role of typical 3D waves in falling films,5,7,8 and in which measure the results of
modeling of 3D waves in falling liquid films with the imposition of periodic boundary conditions are
valid for description of experimentally observed wave evolution. To answer these questions, we used
the method of laser induced fluorescence to obtain instant shapes of three dimensional waves and to
investigate the regularities of formation of 3D wave patterns arising due to transverse instability of
2D waves.
Our results show that at the initial stage of 2D–3D wave transition, when the amplitudes of
transverse modulations of 2D waves grow exponentially, the measured values of growth rate agree
well with the theoretically predicted4 values of maximal growth rate and that in the case of forced
destruction of 2D waves by the regular combs of needles, the transverse scales of forming 3D wave
patterns fall in a relatively narrow band inside the theoretically predicted4 range of instability despite
the transverse scales of initial perturbation of 2D waves.
114106-23 Kharlamov et al. Phys. Fluids 27, 114106 (2015)

At the same time, forming 3D waves always resemble streak-like waves15 but not Λ-solitons.5,7,8
Though both types of the waves have approximately the same transversal dimensions and though the
form of capillary precursor in front of leading edge is horseshoe-shaped for both of them, these types
of the waves differ significantly. Λ-solitons have curved crest with elongated trough behind them, as
shown in Fig. 2. As discussed by Demekhin et al.,8 such shape of the wave leads to the following
consequences. Excessive liquid from excited solitons drains back along their ridges, which results
in the formation of a checkerboard pattern of 3D waves. As our experiments show, behind forming
3D waves there arise wakes in the form of vertical strips with increased film thickness, which may
indicate that excessive liquid from the waves drains with the wakes. These wakes disturb the following
2D wave and as a result synchronous 3D wave pattern arises and at large distances downstream the
film inlet 3D waves line up in vertical chains (see Fig. 19(b)). Previously such arrangement of 3D
waves was described for the film falling down inner wall of vertical tube.15 Formation of the wakes
on the initial stage of 2D–3D wave transition was not observed in other works, apparently because
of small curvature of the film surface in the region of the wakes. As discussed in Sec. IV, special
measures should be taken to distinguish the wakes behind forming leading humps.
Though the shapes of the observed 3D waves agree well with the results of modeling using peri-
odic boundary conditions,12,13 including many tiny details, there exist a few principal distinctions
between experiment and modeling. As shown in our experiments, in the region of film flow where
leading humps with wakes behind them are formed, on the time averaged fields of film thickness
the formation of rivulets is observed accompanied by the transverse redistribution of time averaged
liquid flow rates with much greater flow rates in the central parts of the rivulets than in their periph-
eries. Concurrently, a weaker but quite distinguishable liquid inflow to the central parts of trailing
humps also occurred (see, e.g., Fig. 11). Although the resulting liquid redistribution is strong, time
averaged local flow rates in transverse direction ⟨qz(x, z)⟩t are apparently small in comparison with
local streamwise flow rates ⟨q(x, z)⟩t. Thus, for the cases presented in Figs. 9 and 11, rough estimate
under the assumption ⟨q(x, z)⟩t ∼ ⟨h(x, z)3⟩t shows that local transverse flow rates are approximately
an order of magnitude smaller than streamwise flow rates.
Apparently, the initial stage of transverse liquid redistribution is related to transverse liquid
drainage from the residual layer to the curved forward parts of decaying 2D waves in such a way that,
in terms of streamwise evolution of time averaged film thickness, stronger spanwise flow rates are
associated with more curved forward regions, i.e., with the central parts of forming leading humps,
which is clearly seen in Figs. 9 and 11. Although some alternative explanations are possible, in our
opinion, the transverse liquid redistribution can be associated with one of two plausible mechanisms.
The first of them can be related with spanwise flux of liquid in the region of the capillary pre-
cursor of the humps, if the net balance of the liquid flowing along capillary ridges and troughs of
the precursor results in the inflow of the liquid towards the humps. It should be noted that directly
opposite process of spanwise drainage of liquid from leading hump through the capillary precursor
was obtained recently in modeling with downstream periodic boundary conditions.13 As stated in
Ref. 13, due to this drainage the amplitudes and celerity of running forward leading humps decrease,
which leads to the oscillations of 3D wave patterns and in the process of such drainage the spanwise
flow rates can be comparable to the streamwise flow rates. Although our results on the initial stage of
2D–3D transition show a directly opposite tendency of liquid accumulation under the running forward
and decreasing in amplitude leading humps (see, e.g., Fig. 10), the “capillary precursor induced”
mechanism of liquid inflow to the humps still can hold if the structure of capillary precursor for the
developing humps is sufficiently different from that obtained in Ref. 13 and induces influx of liquid
to the humps instead of its drainage.
But if “capillary precursor induced” mechanism of the liquid inflow to the humps does not work,
there exists a possibility that pressure under the humps is lower than the value induced by the curvature
of film surface to such an extent that inflow into the humps still takes place. If the latter is true, it may
imply that boundary layer models in which pressure in the liquid films is defined only through the
free boundary curvature of the film are not applicable to simulations of 3D wave regimes in falling
films. It should be noted that little deviations of pressure from the values induced by the free boundary
curvature of liquid film in front of large 2D waves were obtained, e.g., in direct numerical simula-
tions.33 Although these deviations are assumed to be negligible for the applicability of boundary layer
114106-24 Kharlamov et al. Phys. Fluids 27, 114106 (2015)

models,6 in the case of 3D waves, if such deviations exist in front of the protruded forward parts of
3D waves, when uniformity along the z-axis is disrupted, they can induce transverse liquid fluxes in
the circumferential residual layer. Whereas for vertical plate other forces in transverse direction are
absent, small but persistent deviations of pressure, can lead to strong effects. Finally, the applicability
of boundary layer models is accepted after their validation by experiments or exact analysis.
Additional considerations should be given to the imposition of periodic boundary conditions
in downstream direction in modeling of 3D wave patterns. Even if such conditions look justified
in modeling of saturated states, their implementation in the case of evolving waves can lead to the
discrepancies between theory and experiment as discussed, e.g., in Ref. 6. The results presented in this
article show that in all cases downstream evolution of 3D wave patterns occurs in such a manner that
streamwise local flow rates are changed on the scales sufficiently large in comparison with stream-
wise wavelength. Whereas periodic boundary conditions are usually applied to the computational
domains with the dimensions of one or a few wavelengths, our results show that in the experiments
such conditions do not hold because of inequality of inflow and outflow local flow rates at the distances
of a few wavelengths as can be seen, e.g., from Figs. 9 and 11. It is a subtle matter whether or not
this discrepancy between modeling and experimental conditions will result in a sufficient difference
between the calculated and experimentally observed wave patterns but this discrepancy should be
taken into account.
At this stage, we cannot obtain direct experimental evidence as to the real mechanism of the
phenomenon of transverse liquid redistribution, which is to be further investigated. In any case, the
results obtained are of interest not only for modeling of film flows but also for practical applications
because, as was shown above, at large distances from the film inlet due to 2D-3D wave transition the
local flow rates can differ several times at the transverse distances of about 1 cm, which is an effect
that cannot be neglected.

ACKNOWLEDGMENTS
This work was supported by the Russian Science Foundation Grant No. 14-22-00174.
1 H.-C. Chang and E. A. Demekhin, Complex Wave Dynamics on Thin Films (Elsevier, New York, 2002).
2 S. V. Alekseenko, V. E. Nakoryakov, and B. G. Pokusaev, Wave Flow of Liquid Films (Begell House, New York, 1994).
3 S. Saprykin, E. A. Demekhin, and S. Kalliadasis, “Two-dimensional wave dynamics in thin films. II. Formation of lattices

of interacting solitary pulses,” Phys. Fluids 17, 117106 (2005).

4 E. A. Demekhin, E. N. Kalaidin, S. Kalliadasis, and S. Yu. Vlaskin, “Three-dimensional localized coherent structures of

surface turbulence. I. Scenarios of two-dimensional-three-dimensional transition,” Phys. Fluids 19, 114103 (2007).
5 E. A. Demekhin, E. N. Kalaidin, S. Kalliadasis, and S. Yu. Vlaskin, “Three-dimensional localized coherent structures of

surface turbulence. II. Λ solitons,” Phys. Fluids 19, 114104 (2007).

6 S. Kalliadasis, C. Ruyer-Quil, B. Scheid, and M. G. Velarde, Falling Liquid Films (Springer, London, 2012).
7 E. A. Demekhin, E. N. Kalaidin, and A. S. Selin, “Three-dimensional localized coherent structures of surface turbulence.

III. Experiment and model validation,” Phys. Fluids 22, 092103 (2010).
8 E. A. Demekhin, E. N. Kalaidin, S. Kalliadasis, and S. Yu. Vlaskin, “Three-dimensional localized coherent structures of

surface turbulence: Model validation with experiments and further computations,” Phys. Rev. E 82, 036322 (2010).
9 S. V. Alekseenko, V. A. Antipin, V. V. Guzanov, S. M. Kharlamov, and D. M. Markovich, “Three-dimensional solitary waves

on falling liquid film atlow Reynolds numbers,” Phys. Fluids 17, 121704 (2005).
10 S. V. Alekseenko, V. A. Antipin, V. V. Guzanov, D. M. Markovich, and S. M. Kharlamov, “Stationary solitary three-

dimensional waves on a vertically flowing fluid film,” Dokl. Phys. 50, 598 (2005).
11 B. Scheid, C. Ruyer-Quill, and P. Manneville, “Wave patterns in film flows: Modelling and three-dimensional waves,” J.

Fluid Mech. 562, 183 (2006).

12 G. F. Dietze and R. Kneer, “Flow separation in falling liquid films,” Front. Heat Mass Transfer 2, 033001 (2011).
13 G. F. Dietze, W. Rohlfs, K. Nährich, R. Kneer, and B. Scheid, “Three-dimensional flow structures in laminar falling liquid

films,” J. Fluid Mech. 743, 75 (2014).

14 C. D. Park and T. Nosoko, “Three-dimensional wave dynamics on a falling film and associated mass transfer,” AIChE J.

49, 2715 (2003).

15 P. Adomeit and U. Renz, “Hydrodynamics of three-dimensional waves in laminar falling films,” Int. J. Multiphase Flow 26,

1183 (2000).
16 U. Gross, Th. Storch, Ch. Philipp, and A. Doeg, “Wave frequency of falling liquid films and the effect on reflux condensation

in vertical tubes,” Int. J. Multiphase Flow 35, 398 (2009).

17 V. I. Petviashvili and O. Yu. Tsvelodub, “Horseshoe-shaped solitons on an inclined viscous liquid film,” Dokl. Acad. Nauk.

SSSR 238, 1321 (1978).

18 S. Saprykin, E. Demekhin, and S. Kalliadasis, “Two-dimensional wavedynamics in thin films. I. Stationary solitary pulses,”

Phys. Fluids 17, 117105 (2005).

114106-25 Kharlamov et al. Phys. Fluids 27, 114106 (2015)

19 S. V. Alekseenko, V. V. Guzanov, D. M. Markovich, and S. M. Kharlamov, “Characteristics of solitary three-dimensional

waves on vertically falling liquid films,” Tech. Phys. Lett. 36, 1024 (2010).
20 N. Kofman, S. Mergui, and C. Ruyer-Quil, “Three-dimensional instabilities of quasi-solitary waves in a falling liquid film,”

J. Fluid Mech. 757, 854 (2014).

21 F. Moisy, M. Rabaud, and K. Salsac, “A synthetic Schlieren method forthe measurement of the to pography of a liquid

interface,” Exp. Fluids 46, 1021 (2009).

22 S. V. Alekseenko, V. V. Guzanov, D. M. Markovich, and S. M. Kharlamov, “Specific features of a transition from the regular

two-dimensional to three-dimensional waves on falling liquid films,” Tech. Phys. Lett. 38, 739 (2012).
23 J. Liu, J. D. Paul, and J. P. Gollub, “Measurements of the primary instabilities of film flow,” J. Fluid Mech. 250, 69 (1993).
24 V. Leontidis, J. Vatteville, M. Vlachogiannis, N. Andritsos, and V. Bontozoglou, “Nominally two-dimensional waves in

inclined film flow in channels of finite width,” Phys. Fluids 22, 112106 (2010).
25 A. Georgantaki, J. Vatteville, M. Vlachogiannis, and V. Bontozoglou, “Measurements of liquid film flow as a function of

fluid properties and channel width: Evidence for surface-tension-induced long-range transverse coherence,” Phys. Rev. E
84, 026325 (2011).
26 C. E. Meza and V. Balakotaiah, “Modeling and experimental studies of large amplitude waves on vertically falling films,”

Chem. Eng. Sci. 63, 4704 (2008).

27 Yu. Ya. Trifonov, “Wavy film flow down a vertical plate: Comparisons between the results of integral approaches and full

scale computations,” J. Eng. Thermophys. 17, 30 (2008).

28 J. Liu, J. B. Schneider, and J. P. Gollub, “Three-dimensional instabilities of film flows,” Phys. Fluids 7, 55 (1995).
29 P. N. Yoshimura, T. Nosoko, and T. Nagata, “Enhancement of mass transfer into falling laminar liquid film by two-

dimensional surface waves—Some experimental observations and modeling,” Chem. Eng. Sci. 51, 1231 (1996).
30 T. Nosoko, P. N. Yoshimura, T. Nagata, and K. Oyakawa, “Characteristics of two-dimensional waves on a falling liquid

films,” Chem. Eng. Sci. 51, 725 (1996).

31 B. Scheid, S. Kalliadasis, C. Ruyer-Quil, and P. Colinet, “Interaction of three-dimensional hydrodynamic and thermocap-

illary instabilities in film flows,” Phys. Rev. E 78, 066311 (2008).

32 S. Alekseenko, A. Cherdantsev, M. Cherdantsev, S. Isaenkov, S. Kharlamov, and D. Markovich, “Application of a high-speed

laser-induced fluorescence technique for studying the three-dimensional structure of annular gas–liquid flow,” Exp. Fluids
53, 77 (2012).
33 D. Gao, N. B. Morley, and V. Dhir, “Numerical simulation of wavy falling film flow using VOF method,” J. Comput. Phys.

192, 624 (2003).