The non-spherical ground state of Proca stars

C. A. R. Herdeiro,1 E. Radu,1 N. Sanchis-Gual,2 N. M. Santos,1, 3 and E. dos Santos Costa Filho1

1
Departamento de Matemática da Universidade de Aveiro and CIDMA, Campus de Santiago, 3810-183 Aveiro, Portugal
2
Dept. Astronomı́a y Astrofı́sica, U. València, Dr. Moliner 50, 46100, Burjassot (València), Spain
3
CENTRA, Dept. Fı́sica, IST, Universidade de Lisboa, Av. Rovisco Pais 1, 1049-001 Lisboa, Portugal
(Dated: November 2023)
Spherical Proca Stars (PSs) are regarded as the ground state amongst the family of PSs. In
accordance, spherical PSs are thought to have a fundamental branch of stable solutions. In this
Letter, we provide energetic, morphological and dynamical evidence that spherical PSs are actually
excited states. The ground state is shown to be a family of static, non-spherical, in fact prolate,
PSs. The spherical stars in the fundamental branch, albeit stable against spherical perturbations,
turn out to succumb to non-spherical dynamics, undergoing an isometry breaking into prolate PSs.
We also provide evidence for the dynamical formation of prolate PSs, starting from spherical dilute
arXiv:2311.14800v1 [gr-qc] 24 Nov 2023

initial data, via gravitational cooling. Consequently, PSs provide a remarkable example of (possibly
compact) relativistic stars, in General Relativity minimally coupled to a simple, physical, field theory
model, where staticity plus stability implies non-sphericity.

Introduction. In vacuum General Relativity (GR), in the Einstein–Klein-Gordon model

Israel’s theorem [1] establishes that staticity implies
√
Z  
sphericity. This powerful result, generalizable to elec- R
S = d4 x −g + Lmatter , (1)
trovacuum [2], supports that astrophysical isolated black 16πG
holes (BHs), if non-rotating, are spherical.
Relativistic stars, on the other hand, enjoy more free- where Lmatter = LKG = −∂α Φ∂ α Φ∗ − µ2 ΦΦ∗ [15], can
dom. When matter withstands the tyranny of gravity, have a multipolar distribution - see Fig. 3 in [4]. These
richer morphologies are possible for static configurations. multipolar stars may be seen as follows. Turning off
For concreteness, consider boson stars - self-gravitating gravity, the Klein-Gordon equation derived from LKG
massive, complex scalar fields in GR [3]. As static so- on flat spacetime, in spherical coordinates, has solutions,
Φ = e−iωt ℓ,m cℓm Rℓ (r)Yℓm (θ, ϕ), with radial function
P
lutions, they admit a multipolar structure [4]. In the
absence of self-interactions, however, the spherical stars
1  p 
remain the ground state, i.e. the lowest energy, dynam- Rℓ (r) = √ Kℓ+ 12 r µ2 − ω 2 , (2)
ically most robust configuration. Thus, adding the re- r
quirement of stability to staticity still implies sphericity.
where ℓ ∈ N0 ; Kj are modified Bessel functions of the
Boson stars, together with their vector cousin Proca
second kind; cℓm are constants; the harmonic time de-
stars (PSs) [5], are widely studied models of compact
pendence has frequency ω ∈ R+ ; Yℓm are real spheri-
objects. They are self-gravitating solitons [6], relat-
cal harmonics. Solutions (2) are irregular, diverging at
able to fuzzy dark matter [7, 8], often considered as BH
the origin. But they are regularized by turning on grav-
foils [9] and ameanable to fully non-linear numerical dy-
ity within model (1). Their non-linear self-gravitating
namics [10], hence permitting a breadth of applications in
versions, with a single cℓm ̸= 0, yield the multipolar
astrophysics, cosmology, strong gravity and mathemati-
stars [16]. Focusing on the axially symmetric sector
cal physics. Yet, the two models are not mere copies. PSs
(m = 0), we illustrate the domain of existence of the
have distinctive properties, e.g., dynamically robust spin-
monopolar (ℓ = 0), dipolar (ℓ = 1) and quadrupolar
ning stars [11]. Studying their binary dynamics is thus
(ℓ = 2) stars in Fig. 1 (left panel), showing also their mor-
justified, together with the resulting waveforms [12] per-
phology - akin to that of hydrogen orbitals. The ADM
mitting comparisons, and intriguing matchings, to real
mass M increases with ℓ, fixing ω. The ground state,
events [13]. PSs also have distinctive geodesic flows; un-
thus, as in hydrogen, are the spherical states (ℓ = 0).
like scalar ones, spherical PSs along the fundamental
Multipolar PSs - flat spacetime limit. We now
branch support accretion disks similar to those around
consider static PSs in the Einstein—Proca model given
BHs, sourcing a possible degeneracy with BH images [14].
by (1) with Lmatter = LP = − 14 Fαβ F̄ αβ − 12 µ2 Aα Āα [17].
In this Letter we unveil another, unexpected, distinc-
This model has a richer spectrum than its scalar coun-
tive feature of PSs. We establish that PSs have a non-
terpart. As in Maxwell’s theory, it admits both electric
spherical (prolate) ground state, with the spherical stars
and magnetic static solutions. Again we start by turn-
being an excited state and decaying into the former un-
ing off gravity. The Proca equations derived from LP
der non-spherical dynamics. Thus, in a remarkable turn
on flat spacetime in spherical coordinates admit the elec-
of affairs for a strong gravity model of compact stars, for
tric [18], axiallly symmetric solutions described by the
PSs, staticity plus stability implies non-sphericity.
Proca ansatz
Multipolar scalar stars. Static scalar boson stars
 2

FIG. 1. Domain of existence of monopolar (spherical, ℓ = 0), dipolar (ℓ = 1) and quadrupolar (ℓ = 2) scalar boson stars
(left panel) and PSs (right panel) in M vs. ω diagrams. The morphology of surfaces of constant Komar energy density is also
provided. We have selected 12 illustrative solutions in the Proca case, detailed in Table I, for the dynamical studies below.

 
−iωt H1 (r, θ) ∂θ H1 = H2 = 0 to ensure both the metric and the gauge
A=e iH0 (r, θ)dt + dr + H2 (r, θ)dθ ,
r potential are Z2 -even. All solutions are asymptotically
P (3) flat, obeying, at infinity, r = ∞, Fi = Hi = 0, i = 0, 1, 2.
where Hi (r,P θ) = ℓ c i,ℓ hi,ℓ (r)Yℓ0 (θ), for i = 0, 1 and Additionally, regularity on the symmetry axis imposes
H2 (r, θ) = ℓ c2,ℓ h2,ℓ (r)dYℓ0 (θ)/dθ. The radial function θ = 0, π, ∂θ Fi = ∂θ H0 = ∂θ H1 = H2 = 0, together with
h0,ℓ (r) = Rℓ (r) is given by (2) [19]; ci,ℓ , i = 0, 1, 2 are F1 = F2 to ensure the absence of conical singularities, for
constants. Again, these solutions are irregular at the ori- both ℓ = 1, 2 solutions. Finally regularity at the origin,
gin; and again they are regularized by turning on gravity r = 0, requires ∂r Fi = Hi = 0 (i = 0, 1, 2) for ℓ = 1 and
within model (1). Their non-linear self-gravitating ver- ∂r Fi = ∂r H0 = H1 = H2 = 0 for ℓ = 2.
sions, with ci,ℓ ̸= 0 for a single ℓ ∈ N0 , yield multipolar, The field equations were solved numerically employing
axially symmetric PSs. a professional package [20]. The numerical error for the
Construction of axially symmetric multipolar solutions is estimated to be typically less than 10−5 .
PSs. Static, axi–symmetric solutions of (1) possess two Energetics and morphology. The domain of exis-
(commuting) Killing vectors, {ξ, η}. In adapted coordi- tence of the ℓ = 0, 1, 2 PSs is displayed in Fig. 1 (right
nates (t, φ), ξ = ∂t and η = ∂φ . It is further assumed panel). For all ℓ: (i) the bound state condition ω < µ
that the solutions admit a 2–space orthogonal to {ξ, η}, is satisfied; (ii) as ω → µ (Newtonian limit), M van-
in which mutually orthogonal, spherical–like coordinates ishes; (iii) M increases monotonically as ω decreases,
(r, θ) are introduced so that grθ = 0. A line element reaching a maximum Mmax at frequency ω(Mmax ) - this
compatible with these assumptions reads interval shall be dubbed the fundamental branch, and it
contains stable solutions for ground state models; (iv)
ds2 = −e2F0 dt2 + e2F1 dr2 + r2 dθ2 + e2F2 r2 sin2 θdφ2 ,


further decreasing ω, the mass decreases until a back-
(4) bending is reached at ωmin . For ℓ = 0, 1, 2, in units
where t ∈ (−∞, +∞), r ∈ [0, +∞), θ ∈ [0, π], φ ∈ [0, 2π), of µ, {Mmax , ω(Mmax ), ωmin } = {1.06, 0.875, 0.814},
and {F0 , F1 , F2 } are real functions of (r, θ). {0.792, 0.792, 0.698} and {1.15, 0.833, 0.824}.
To solve model (1) with the ansatz (3)-(4) for asymp- Comparing the different ℓ curves in Fig. 1, within
totically flat, axi-symmetric PSs, appropriate boundary each model, unveils a fundamental difference between the
conditions must be imposed. In all cases here, the equato- Proca and scalar cases. For the former, the lowest energy
rial plane, θ = π/2, is a fixed plane of the Z2 north-south configuration for a fixed frequency within the fundamen-
mapping θ → π/2 − θ. The geometry is always Z2 -even, tal branch is not the ℓ = 0, but rather ℓ = 1 PSs. In
whereas the Proca field may be either Z2 -even or Z2 - fact, even ℓ = 2 stars have lower energies than ℓ = 0, ex-
odd, depending on ℓ. The monopolar solutions (ℓ = 0) cept in a vicinity of ω(Mmax ) for ℓ = 0. This is energetic
are well known [5]. The dipolar solutions (ℓ = 1) obey, evidence that prolate PSs are the family’s ground state.
at θ = π/2, ∂θ Fi = H0 = H1 = ∂θ H2 = 0, to ensure the The discovery of the prolate stars also solves a conun-
metric (gauge potential) is Z2 -even (Z2 -odd). Quadrupo- drum of the static PS model. On the one hand, for
lar solutions (ℓ = 2) obey, at θ = π/2, ∂θ Fi = ∂θ H0 = spherical scalar boson stars, the scalar field profile has
 3

no radial nodes for the ground state. Spherical excited

stars are obtained by increasing the number of nodes,
and have a larger energy for the same frequency - see
e.g. Fig. 1 in [21]. This is akin to N s orbitals in hydro-
gen, N ∈ N. For spherical PSs, on the other hand, it
can be proven that the Proca scalar potential [22] has
no nodeless solutions [5]. Consequently, the ground state
of static PSs has been regarded as the 1-node spheri-
cal solutions. But for stationary, rotating PSs there are
nodeless solutions [23, 24], which are actually dynami-
cally robust [11]. Hence the conundrum: what prevents
nodeless static PSs? The answer turns out to be spheric-
ity, since the prolate PSs have no nodes - Fig. 2; they are
the missing link. This is morphological evidence they are
the ground states of the static PSs family.
FIG. 3. Frequency squared Ω2 for spherical perturbations of
monopolar (ℓ = 0) scalar (top) and Proca (bottom) stars.
Dashed (solid) lines are stable (unstable) stars.

Dipolar/Prolate Monopolar/Spherical Quadrupolar

1-(0.970,0.412) 5-(0.970,0.693) 9-(0.965,0.693)
Stable S→P Q→P
2-(0.920,0.626) 6-(0.936,0.924) 10-(0.920,0.955)
Stable S→P ; [Z2 ] S→P (xy) Q→P
3-(0.870,0.735) 7-(0.900,1.04) 11-(0.870,1.11)
Stable S→P ; [Z2 ] S→P (xy) Q→BH
4-(0.820, 0.785) 8-(0.885,1.05) 12-(0.845,1.15)
Stable S→P Q→BH

FIG. 2. Prolate/spherical (main panel/inset) PSs do not TABLE I. Models #-(M, ω), with #= 1 − 12 and their evo-
have/have nodes. Both illustrative solutions have ω/µ = 0.9. lution. [Z2 ] indicates a north-south symmetry is imposed.

Dynamics. Both the scalar boson stars (with no

nodes) and the PSs (with one node) in the fundamental We first analysed if prolate PSs are stable in long term
branch are stable against spherical perturbations. Con- numerical evolutions. All prolate PSs evolved (1-4 in
sidering a perturbation of frequency Ω [5, 25, 26], linear Fig. 1) have exhibited stability throughout the time evo-
stability changes when Ω2 changes sign [27]. In Fig. 3, Ω2 lutions performed (up to t = 2 × 104 , in units of µ) -
is plotted against ω for the set of scalar and PSs between Fig. 4 (left columns). This corroborates that prolate PSs
the maximum and the minimum frequency. It shows that in the fundamental branch are stable.
Ω2 > 0 for the set of stars in the fundamental branch Second, we assessed the stability of spherical stars.
(dashed lines) for both scalar and PSs. Thus, these spher- The evolution of solutions 5-8 shows that at timescales of
ical stars appear to be the ground state for both models. t ∼ 2000 the spherical stars develop a 2-centre morphol-
But prolate stars have lower energy in the Proca family. ogy along the z-axis; one of these centres swiftly dissi-
Thus, non-spherical dynamics has to be analysed. pates leaving an oscillating prolate star, which is kicked
Performing 3D numerical relavitity evolutions with- in the z-direction - Fig. 4 (middle columns). The pro-
out imposing any spatial symmetries (unless otherwise lateness of the kicked star is confirmed by observing the
stated), using the ℓ = 0, 1, 2 PSs as initial data, allows scalar potential acquires a dipolar configuration as in the
tackling both the issue of stability and the end point leftmost column of Fig. 4. This corroborates the spheri-
of unstable stars. These simulations use the einstein cal PSs are decaying into the ℓ = 1 prolate stars.
toolkit [28], together with the formalism and infras- The aformentioned kick is generic in the ℓ = 0 → ℓ = 1
tructure developed in previous works [11, 29–31]. Twelve decay. Both ℓ = 0 and ℓ = 1 PSs have well defined Proca
illustrative PS solutions were selected, highlighted in field parity, even or odd, respectively. Thus, its energy-
Fig. 1 (right panel). Their mass, frequency and dynami- momentum tensor (roughly the Proca field squared) has
cal fate are summarized in Table I. even parity, sourcing a north-south Z2 symmetric metric.
 4

FIG. 4. Snapshots in the x-z plane of the imaginary part of the Proca scalar potential [22] (green panels) and energy density
(blue panels) in PS evolutions. (Two leftmost columns) Solution 3, illustrating the stability of prolate PSs. (Three central
columns) Solution 7, illustrating the isometry breaking of spherical PSs into prolate stars, accompanied by a kick. (Three
rightmost columns) Formation scenario, illustrating the formation of a prolate star from an initial dilute spherical Proca cloud.

The instability of spherical PSs means, however, a dipo- scalar and PSs form via gravitational cooling. Following
lar mode grows around a spherical PS. The resulting su- closely [22] but imposing no symmetries, one observes
perposition has both even and odd parity terms, thus an isometry breaking in the cloud collapse - Fig. 4 (right
sourcing a non-Z2 symmetric geometry. The transition columns) - supporting the formation of a prolate PS.
from a Z2 to a non-Z2 symmetric geometry generates a Remarks. Static PSs include multipolar solutions,
kick of the remnant (proto-)prolate PS [32]. exemplified by the ℓ = 1, 2 PSs, reported here. A much
We have also imposed an artificial Z2 symmetry in larger spectrum of solutions exists, both electric and
some evolutions (e.g. of 6 and 7) effectively evolving a magnetic, and also non-axisymmetric, deserving detailed
single hemisphere, to attempt blocking the ℓ = 0 → ℓ = 1 construction and analysis. Quite unexpectedly, and dif-
decay. A cunning dynamical adaptation ensued. As be- ferently from the scalar model, the energetic, morpholog-
fore, the spherical stars develop a 2-centre morphology ical and dynamical evidence presented herein establishes
along the z-axis; these 2-centres oscillate, merging and that the PSs family has a non-spherical ground state in
separating along the z axis; but eventually the merger is the static sector. The ℓ = 0 PSs, considered hitherto as
followed by a separation along the x-axis. A similar dy- the ground state, succumb to non-spherical dynamics.
namics to that in Fig. 4 (middle panels) follows, but now Spherical PSs attain their maximum energy density
in the xy plane, to comply with the artificial symmetry away from the origin, unlike scalar stars in the ground
imposed, and still generating a prolate PS. state - see Fig. 3 in [34]. For prolate PSs, the energy
Concerning the ℓ = 2 PSs, we find that the evolutions density is attained at the centre, befitting typical ground
of solutions 9 and 10 form ℓ = 1 PSs, roughly resembling states. Still, a deeper understanding of why PSs dynami-
the evolution of the spherical PS in Fig. 4 starting from cally prefer less isometric lumps would be desirable. One
the (2-centre) 3rd panel. The higher energy solutions remark in this direction is that the symmetry loss in the
solutions 11 and 12, by contrast, collapse into BHs. spherical→prolate decay seen here is reminscent of the
Finally, we considered the formation of a prolate PS fate of charged liquid drops in fluid dynamics: beyond
from a dilute spherical distribution of Proca field. Such the Rayleigh limit [35, 36] spherical drops decay into
simulations were previously performed assuming spheri- non-spherical ones, within a multipolar spectrum [37],
cal symmetry [22, 33], providing evidence that spherical as shown by experiment [38] and simulations [39].
 5

Acknowledgements. We would like to thank E. Radu, J. A. Font, S. H. W. Leong, and

Richard Brito for discussions on the perturbation the- A. Torres-Forné, Phys. Rev. D 106, 124011 (2022),
ory of PSs. This work is supported by the Center for arXiv:2208.11717 [gr-qc].
Research and Development in Mathematics and Applica- [13] J. Calderón Bustillo, N. Sanchis-Gual, A. Torres-Forné,
J. A. Font, A. Vajpeyi, R. Smith, C. Herdeiro, E. Radu,
tions (CIDMA) through the Portuguese Foundation for and S. H. W. Leong, Phys. Rev. Lett. 126, 081101 (2021),
Science and Technology (FCT – Fundação para a Ciência arXiv:2009.05376 [gr-qc].
e a Tecnologia), references UIDB/04106/2020 and [14] C. A. R. Herdeiro, A. M. Pombo, E. Radu, P. V. P.
UIDP/04106/2020. The authors acknowledge support Cunha, and N. Sanchis-Gual, JCAP 04, 051 (2021),
from the projects PTDC/FIS-AST/3041/2020, as well arXiv:2102.01703 [gr-qc].
as CERN/FIS-PAR/0024/2021 and 2022.04560.PTDC. [15] Φ denotes a complex scalar field with mass µ and complex
This work has further been supported by the Euro- conjugate Φ∗ . As usual G, is Newton’s constant, g the
metric determinant and R the Ricci scalar.
pean Union’s Horizon 2020 research and innovation [16] These non-linear versions turn on other Yℓm for a given
(RISE) programme H2020-MSCA-RISE-2017 Grant No. ‘seed’ solution, but the original harmonic dominates
FunFiCO-777740 and by the European Horizon Eu- asymptotically. Same holds below for the vector case.
rope staff exchange (SE) programme HORIZON-MSCA- [17] Aα denotes a complex Proca field with mass µ, complex
2021-SE-01 Grant No. NewFunFiCO-101086251. NSG conjugate A∗α and field strenght F = dA.
is supported by the Spanish Ministerio de Universi- [18] In this letter we only consider electric PSs; the magnetic
dades, through a Marı́a Zambrano grant (ZA21-031) case will be considered in a follow up paper.
[19] The expressions for hi,ℓ (r), i = 1, 2 are more involved
with reference UP2021-044, funded within the Euro- but can also be obtained analytically.
pean Union-Next Generation EU. This work is also [20] W. Schonauer and R. Weiss, J. Comput. Appl. Math. 27,
supported by the Spanish Agencia Estatal de Inves- 279 (1989).
tigacion (Grant PID2021-125485NB-C21) funded by [21] M. Brito, C. Herdeiro, E. Radu, N. Sanchis-Gual,
MCIN/AEI/10.13039/501100011033 and ERDF A way and M. Zilhão, Phys. Rev. D 107, 084022 (2023),
of making Europe. NSG thankfully acknowledges the arXiv:2302.08900 [gr-qc].
[22] F. Di Giovanni, N. Sanchis-Gual, C. A. R. Herdeiro,
computer resources at Tirant and the technical support
and J. A. Font, Phys. Rev. D98, 064044 (2018),
provided by UV (FI-2023-2-0002). N. M. S. is supported arXiv:1803.04802 [gr-qc].
by the FCT grant SFRH/BD/143407/2019. E.S.C.F. is [23] C. A. R. Herdeiro and E. Radu, Phys. Rev. Lett. 119,
supported by the FCT grant PRT/BD/153349/2021 un- 261101 (2017), arXiv:1706.06597 [gr-qc].
der the IDPASC Doctoral Program. Computations were [24] C. Herdeiro, I. Perapechka, E. Radu, and Y. Shnir, Phys.
performed at the ARGUS cluster at Aveiro University. Lett. B 797, 134845 (2019), arXiv:1906.05386 [gr-qc].
[25] M. Gleiser and R. Watkins, Nucl. Phys. B319, 733
(1989).
[26] S. H. Hawley and M. W. Choptuik, Phys. Rev. D62,
104024 (2000), arXiv:gr-qc/0007039 [gr-qc].
[1] W. Israel, Phys. Rev. 164, 1776 (1967). [27] In Fig. 3 we have extended the results for perturbations of
[2] W. Israel, Commun. Math. Phys. 8, 245 (1968). PSs away from the maximal mass, the only ones reported
[3] F. E. Schunck and E. W. Mielke, Class. Quant. Grav. 20, in [5].
R301 (2003), arXiv:0801.0307 [astro-ph]. [28] EinsteinToolkit, “Einstein Toolkit: Open software for rel-
[4] C. A. R. Herdeiro, J. Kunz, I. Perapechka, E. Radu, ativistic astrophysics,” http://einsteintoolkit.org/.
and Y. Shnir, Phys. Lett. B 812, 136027 (2021), [29] H. Witek and M. Zilhão, “Canuda,” https://
arXiv:2008.10608 [gr-qc]. bitbucket.org/canuda/.
[5] R. Brito, V. Cardoso, C. A. R. Herdeiro, and E. Radu, [30] H. Witek, M. Zilhao, G. Bozzola, C.-H. Cheng, A. Dima,
Phys. Lett. B752, 291 (2016), arXiv:1508.05395 [gr-qc]. M. Elley, G. Ficarra, T. Ikeda, R. Luna, C. Richards,
[6] D. J. Kaup, Phys. Rev. 172, 1331 (1968). N. Sanchis-Gual, H. Silva, and U. Sperhake, “Canuda: a
[7] A. Suárez, V. H. Robles, and T. Matos, Astrophys. Space public numerical relativity library to probe fundamental
Sci. Proc. 38, 107 (2014), arXiv:1302.0903 [astro-ph.CO]. physics,” (2023).
[8] L. Hui, J. P. Ostriker, S. Tremaine, and E. Witten, [31] M. Zilhão, H. Witek, and V. Cardoso, Class. Quant.
Phys. Rev. D 95, 043541 (2017), arXiv:1610.08297 [astro- Grav. 32, 234003 (2015), arXiv:1505.00797 [gr-qc].
ph.CO]. [32] At the linear level, this kick can be interpreted as a trans-
[9] H. Olivares, Z. Younsi, C. M. Fromm, M. De Lauren- lation of the system’s centre of mass - see Appendix G
tis, O. Porth, Y. Mizuno, H. Falcke, M. Kramer, and in [40].
L. Rezzolla, Mon. Not. Roy. Astron. Soc. 497, 521 (2020), [33] E. Seidel and W.-M. Suen, Phys. Rev. Lett. 72, 2516
arXiv:1809.08682 [gr-qc]. (1994), arXiv:gr-qc/9309015 [gr-qc].
[10] S. L. Liebling and C. Palenzuela, Living Rev. Rel. 15, 6 [34] C. A. R. Herdeiro, A. M. Pombo, and E. Radu, Phys.
(2012), arXiv:1202.5809 [gr-qc]. Lett. B 773, 654 (2017), arXiv:1708.05674 [gr-qc].
[11] N. Sanchis-Gual, F. Di Giovanni, M. Zilhão, C. Herdeiro, [35] L. Rayleigh, Proc. London Math. Soc. 10, 4 (1879).
P. Cerdá-Durán, J. Font, and E. Radu, Physical Review [36] L. Rayleigh, The London, Edinburgh, and Dublin Philo-
Letters 123, 221101 (2019). sophical Magazine and Journal of Science 14, 184 (1882).
[12] N. Sanchis-Gual, J. Calderón Bustillo, C. Herdeiro, [37] O. A. Basaran and L. E. Scriven, Physics of Fluids A:
Fluid Dynamics 1, 795 (1989).
 6

[38] D. Duft, T. Achtzehn, R. Müller, B. A. Huber, and 144501 (2011).

T. Leisner, Nature 421, 128 EP (2003). [40] F. J. Zerilli, Phys. Rev. D 2, 2141 (1970).
[39] J. C. Burton and P. Taborek, Phys. Rev. Lett. 106,