STABILITY OF PERIODIC WAVES OF FINITE AMPLITUDE ON THE SURFACE OF A DEEP FLUID

V. E. Zakharov

Zhurnal Prildadnoi Mekhaniki i Tekhnicheskoi Fiziki, Vol. 9, No. 2, pp. 86-94, 1968

ABSTRACT: We study the stability of steady nonlinear waves on the surface The first term in this expression is the kinetic energy and the
of an infinitely deep fluid [1, 2]. In section 1, the equations of hydro- second and third terms are the potential energy in the field of gravity
dynamics for an ideal fluid with a free surface are transformed to and the potential energy due to surface forces. We introduce thequan-
canonical variables: the shape of the surface ~(r, t) and the hydrody- tity 9(r, t) = ~(z, r, t) i z=~" Specifying the quantities ~ and g, fuIIy
namic potential ~(r, t) at the surface are expressed in terms of these defines the fluid finw since the boundary-value problem for Laplace's
variables. By introducing canonical variables, we can consider the equation has a unique solution. Using the equation
problem of the stability of surface waves as part of the more general
o,e o~v on o*
problem of nonlinear waves in media with dispersion [g, 4]. The re-
Ot - - Ot ~ Ot Oz z=n'
sults of the rest of the paper are also easily applicable to the general
case. we obtain

In section 2, using a method similar to van der Pohi's method, we i)T V+]
obtain simplified equations describing nonlinear waves in the small ot 4-,~,~--~V ~ ],',1 + V+l
amplitude approximation. These equations are particularly simple if l I { o o ~2 oqS (1.5)
2 (V(D)+++ 7 \-0-}-) -- ~ - z ( v(D V~)I==~.
we assume that the wave packet is narrow. The equations have an
exact solution which approximates a periodic wave of finiteamplitude. Equations (1.2) and (1.5). together with Laplace's equation, are
equivalent to Eqs. (1.1)-(1.3). We can prove that Eqs. (1.1) and (1.8)
In section 3 we investigate the instability of periodic waves of finite
can be put in the form
amplitude. Instabilities of two types are found. The first type of in-
stability is destructiveinstability, similar to the destructive instability 07 8E 0iF 5E
of waves in a plasma [g, 6]. In this type of instability, a pair of waves ot--5~ ' ot - - - - T ~ - " (1.6)
is simultaneously excited, the sum of the frequencies of which is a
Here E is the energy; the symbols 6E/6~ and 5E/69 denote the
multiple of the frequency of the original wave. The most rapid de-
variational derivative.
structive instability occurs for capillary waves and the slowest for
Consider first the variation of g~. Obviously., the variation of the
gravitational waves. The second type of instability is the negative-
potential energy is zero. We transform the kinetic energy by means
pressure type, which arises because of the dependence of the nonlinear
of Green's formula:
wave velocity on the amplitude; this results in an unbounded increase
in the percentage modulation of the wave. This type of instability
occurs for nonlinear waves through any media in which the sign of ., _ = T ~ "I" - N - +~= - -
the second derivative in the dispersion law with respect to the wave -co s

number (d2~/dk 2) is different from the sign of the frequency shift due t (' ~F 0el)
to the nonlinearity.
8

As announced by A. N. Litvak and V. I. Talanov [7], this type of

Here ds is a differential surface element. The normal derivative
instability was independently observed for nonlinear electromagnetic
0~/0n is linked with ~ by the Green's function for the boundary-value
waves.
problem of Laplace's equation;
1. Canonical variables. We consider the potential flow of an ideal
0 0 (s) I
fluid of infinite depth in a homogeneous gravity field. We choose the On - - a(s, ,h)~F(,h)dsl , (1,7)
coordinates so that the undisturbed surface of the fluid coincides with
the xy-plane. The z-axis is directed away from the surface. In what Here s and s1 are points on the surface. The function G is symmetric;
follows, all vectors are two-dimensional vectors in the xy-plane. i.e., G(s, sD ; G(si, s).
Let ~(r, t) be the shape of the surface of the fluid and let ~(r, z, t) The variation of the kinetic energy has two terms:
be the hydrodynamic potential. The fluid flow is described by Laplace's
equation, 8E* = t f [SW (s) -0@(s)
- T ~ -l-, ~. (s) -g~
0 8 0 ( s ).] d s .
s
0~q) +
Ar =o, (L1)
From (1.7) and the symmetry of the Green's function, we see that
both terms are the same:
with two conditions at the surface,

0~1 .......... 0(I) [ &I) ~ . >O7-

o+ 1/1 + Vq~"dr 9
ot - I ' t + V~l~ 0 ~ - t ~ ' = ~ 7 - - - W l V @ I ~ = ~ , (L2) 8

From this we obtain (1.2) instantaneously.

- ~ - -l- gq = -- "~" (V(I))"-Iz=,~ -- ~" k-~7-z} -[- :zV I t -+- Vq'-' ' Consider now variation in ~ (this simple proof is due to R. M.
Garipov).
and a condition at infinity,
Variation of the potential energy at once gives the terms on the
{b--->-0 aS z-+-- ex-. left-hand side of (2.5). Variation of the kinetic energy gives
Here g is the acceleration due to gravity and a is the coefficient
of surface tension. 8E*:-~-I[(VO)~+('~-z )~]Sq(r)dr@
Equations (1.1)-(1.8) conserve the total energy of the fluid,
0,+,
.[ I* 1-
+,:= ~'+...~~z,- ) [(v,,,)~ +-~),
. ( OtD ,o-I
~ ~z++-i- + 5,++, [(v+,
-co

v +,,i+,++,-+-+S(,+ +- +,,+- +,>+, .+

190
 Here 6~ is the variation in the potential due to a change in the X 8 (k -- kl -- k.2tdk dkl dk~ +
boundary. Since ~ satisfies Laplace's equation, we can apply the Green's
theorem to the second integral: + t !'IV (k, k~, k~, ks)a* (k)a* (kl) a (k~) a (k~) X

8(I3 ds
lil ~ 17 X ~ (k -- kl -- k~ -- k s t d k dk~ dk2 dk~,
(1.ii)

O(l) \ Off) Wher e

(
= i k---37-~ -~- Vq V(D) ~ z=~ 5q (r) dr.
V(k, kl, k 2 ) =
Finally we have
t { {r (k) o~ (kl)~th
--8n'1/'2 [(kk~)+lkltk~[]\ ~0(k2) ) [ ( k k 2 ) - [ I k I ! k " H •
2 [ az } 47 (V~1 Vq)) Oz._z]z=n 8rl (r) d r .

Hence we obtain (1.5). ((o (k) c0 (k2)~V, ( l kx I ~% ,

Thus, Eqs. (1.2) and (1.5) are Hamilton's equations and $ and x\ c0(k~) ) I l k l l k ~ l )
are canonical variables, ~ being a generalized coordinate and ~] a ?0 (k~ ~(k~)y/~ ( Ik I y,q
generalized m o m e n t u m . The energy E of the fluid is the Hamiltonian. +[(klk')+lk~l]k21]\ ~o(k) ) \lk~ltk21) j ; (L.12)
To close Eqs. (1.1) and (1.5) we have to solve the boundary-value
problem for Laplace's equation. We find the solution of this problem
in the form of a series in powers of r/. If we apply a Fourier transforma-
tion to the variables x and y, W (k, k~, k2, k3)

~1(k) = ~ t ~d rt (r) e_i(kr)dr, ~g (k) = - ~t I T (r) e-i(kr)dr,

-- 32~" [(o (k) re (kl) o (k~) ~o (k3)] ~/y (kkl) (k~.ks) +
we obtain the series in a more convenient form.
t
Omitting the details, we i m m e d i a t e l y give the result of the ex-
pansion (up to second-order terms):

(k, z) = e Iklz { ~ (k) -[- ! ~F (k~) "q (kz) I k~ I ~ (k -- k~ -- k2) dkldks - - • [Lw(k~)(o(ka)J (21k}-[ 2 [ k l i - - l k - - k 2 1 - -

t
--~-I[] k --k~,+lk--k2]--,kl) X - - I k - - ka [ - - ] k , - - k2 [ -- [ k l - - ks [) +

• Ia I~ (ka n (k._,)
~ (k~)~ (k -- kl--k~-- k~)]dk,dk~k~}. (i.8) + L (o (k) o) (k~)j (21k2]+2[k3[--
Here 6 denotes the delta function.
If we linearize (1.2) and (1.5) and consider onIy the first term in --]k--k2[--[k--ks] --lkr--k2l--Ikl--k:.~ I)-
(1.8), we obtain the theory of small oscillations for the surface of a
o (kl) to (k~)] 5'
fluid, which describes the propagation of waves with dispersion law ~)( ~ (-k~3)j ( 2 l k x l + 2 ] k : [ - -

(o ( k ) = K g Ik I + a l k p . ~ lk2+kz[-- ]k-- k2] -- lk-- kl I-- ]kx +ka[) --

We now c o m p l e t e the transformation to the c o m p l e x variable a(k)

via the equations - ko)(k,)ro(k3)J (2]kxi+21k2]--

1 I k I'/' --[k2+ ksl -- [kl-- kzl-- Ik-- k~ [-- [k + ksj--

~1(r, t) = 2 ~ S ~ [a (k) et(ur) + a* (k) e-i(kr)] dk,

i ~ ~o'I~ (k)
W~ (k) ~o(k~)l'/: I k2 + ks I '
-- Lco (kl) r9 (k2)j (--' --
~F (r, t) --= 2~ ]/'2 ,) ~ [a (k) eitkr) -- a* (k) e-i(kr)] dk. (1.97
-}- 2 [ k [-}- 2 1k~ 1-- 1 k - - kl [ - - I k - - ks ] - - [ k - b k2 I)--
Then
~-o (kl) co (ka)t G
1 Ik] 5' - - [ ~ ) J (21k'l+2lkal--
n(k)-- V2 mV*(k) [a (k) + a* (-- k)],
i o% (k) -- [k2+ksl--lk--k3l--lk--ksl--!kl+k~i) }. (1.13)
W(k)=-- ]/=g [ k l,/ - [a (k) _ a* ( _ k)] . (1.10)
There are other fourth-order terms in a, proportional to products of
Transformation (1.9) can be considered a canonical transformation the form a*aaa and aaaa and terms conjugate to them. These are ig-
(with c o m p l e x coefficients) to the variables ia*(k) and a(k); Hamilton's nored, since, as will be shown in section 3, their conuibution is small.
equation (1.8) becomes the single equation We note that the functions V and W obey the follovdng equations:
Oa (k) 5E V(k, kl, k s ) = V(k, k~, k 0 = V (ks, kl, k),
Ot __-- i Sa, (k) .
V (-- k, - - k l , -- k2) = V(k, kl, k_o),
Using (1.4), (1.8), and (1.10), we can express the energy in the
form of a series in powers of a(k) and a*(k): W ( k , kl, k~, ks)~---W (kl, k, k2, k 3 ) =
(z.14)
E=I~o(k) a(k)a*(k)dk-bf[V(--k, k,, k 2 ) x = W (k, kl, ks, k 2 ) = W (k2, ks, k, k O,

W (-- k, -- kl, -- k2, ks) = W (k, kl, k2, ks).

X [a* (k) a (k~) a (k~) + a (k) a* (k~) a* (k2i] X
The equation for a(k) has the form
X 8 ( k - - k l - - k~tdk dkl dk~ + Oa (k)
Ot @ in) (k) a (k) =
- -

-r- ~- (k, k~, k2) [a* (k) a* (k~) a* (k~) + a (k) a (k~) a (k~)l X
=--i~{V(--k, kl, k : ) a ( k l ) h ( k ~ ) 6 ( k - - k ~ - - k ~ ) +

191
 + 2V (-- k~, k, k2) a* (k~.) a (k~) 6 (k -- k @ k~) +
+ V (k, kl, k.~) a* (k~) a* (k~) •
X 6 (k @ k l " k~)} dkldk2 --
- - i t W ( k , kl, ke, ka) 5 ( k @ k l - - k 2 - - k a ) X
X a* (k~) a (k~) a (ka) dk~ dk~ dka 9 (3..25)
We seal from (L1.5) that the variables a(k) are the n o r m a l variables
in the problem of s m a l l oscillations.
2. S i m p l i f i e d equations. Equation (1..15) is an a p p r o x i m a t i o n and
is valid for s m a l l n o n l i n e a r i t i e s , roughly speaking, for a / X << 1, where
a is the c h a r a c t e r i s t i c a m p l i t u d e of the wave and X is a c h a r a c t e r i s t i c
w a v e l e n g t h . In this a p p r o x i m a t i o n , we can m a k e a considerable s i m -
p l i f i c a t i o n in Eq. (1..12). To do this we write a(k) as
a (k) - : [A (k, t) 4- t (k, t)] exp [-- io (k) t]. (2.1)
We assume that A(k, t) changes slowly in comparison with f,
where ae << A. We substitute a(k), in the form of (2.1), into the equa-
tion for f and the one for A. In the equation for ac we r e t a i n only terms
which are q u a d r a t i c in k. Assuming A constant as ac varies, we inte-
grate this equation with respect to t i m e . This yields
If w(k) is a m o n o t o n i c function, we note that a sufficient condition
for the e x i s t e n c e of a solution to (2.5) is
(k) > m (kz) + E0 (k -- kl), (2.6)
where k and k I are in the d i r e c t i o n of the same straight line. Indeed,
if (2.6) holds, by adding to k t components perpendicular to k, we
can increase the right-hand side o f ( 2 . 6 ) a n d convert (2.6) into
an equation. On the other hand, if the i n e q u a l i t y _converse to (2.6)
holds, this is a sufficient condition for the absence of solutions to (2.5).
For g r a v i t a t i o n a l waves, with the dispersion law
co(k)=Vglkl ,
an i n e q u a l i t y converse to (2.6) holds. Accordingly, (2.5) cannot have
any solutions and for s m a l l a/,X (2.3) applies. For c a p i l l a r y waves
[k >> (g/a)1/2], with dispersion law ~o(k) = a ~ , (2.6) holds, so that
(2.3), in general, cannot be used; if it is assumed that the w a v e p a c k e t
is sufficiently narrow, i.e., if a(k) is nonzero for t k - ko i << k0, Eq.
(2.5) cannot be satisfied for any co(k), ttence, assuming the w a v e p a c k e t
is narrow, Eq. (2.6) is a p p l i c a b l e for any dispersion laws, in particular,
for c a p i l l a r y waves.
Assuming the wave p a c k e t is narrow, we can m a k e a further sim-
p l i f i c a t i o n in Eq. (2.3). We Introduce the v a r i a b l e z = k -- k0 and
expand co(k) in powers of ~ to terms of second order:

/ ,)~ l'[V(-- k, kx, k~) exp it [o (k) - - re (kl) - - (0 (kQ] x

a ( k ) - - ~o ( k z ) - - ~o (k~) (o (k) = ~0 (k0) + ~ C i- J/2 (~, li ~ 2 4 - ~ 1 ~ i"-),
&o I 8~co c Oeo
• 6(k--k I-k2) A(k~)A(ke)@ c . . . . . oa
. . . i.'.=,.o' ), t: :-: -0~TU...... , s = g' Da~" = ~ ,.=,"

+ 2 V ( - - k : t , k, k2) e x p i t [ ~ 4-~176
co (k) 4- m (kl) - - t0 (k2)
X 6(k [-kl--k~)A*(k~)A(k~)+
+ V (k, k~, k ) _exp it [o (k) 4- o (kl) 4- o (k~)l •
co (k) + ~ (M) 4- ~o ( k )
X 6 (k 4- k l 4- k~) A* (kl) A* (kQ} dkldk~.
X Here ~ x and ~ v are the projections of the vector w- both p a r a l l e l
--
and perpendicular to the vector k0; e is the group v e l o c i t y of t h e w a v e s ;
ka_ is an eigenvector of the tensor DaB. Next we r e p l a c e the approxi-
m a t i o n T(k, kz, kz, k3) by w = T(ko, k0, k0, k0) and introduce the v a r i a b l e
b (k) = :1 (k) exp i [ z c -k i/., ()~ ',i• ~"-'- Lixve)] t ,
Ob
(2 02) =OZ 4- i (zxc 4- '%L !1• ~ + ~/~)'• b=

In the equation for k we r e t a i n only those terms proportional to = - - iw i b* (k) b (k._.) b (k:l) 6 (k -I- lq - - k~ -- ka) dkl [k~dka . (2.7)
A f which contain the most slowly varying exponents. Obviously, all We note that X• is always positive, while xll vanishes for k0 = k~:
the slowly varying exponents are contained in those terms proportional
to A*AA. Gathering a l l these terms together, we obtain 9= I 2 V-~/. % g %

OAOti t ) - - - - i f 7'(k, k~, ke, ka) (3 (k 4- k~ - - k ~ - - k ) •

For k0 < k'~, X Jl is n e g a t i v e , and for k0 > k~, X~ is positive. We apply the
inverse Fourier transformation with respect to ?4:
X exp it [o (k) 4- to (kl) -- (o (k2) - -
b (x, y, t) = ~ I b (• ':v, l) <,xp [ - - i (~z~ 4- yXv) ] d • v 9
-- o (ka)] A * (k~) A (k2) A (ka) dki dk~ dka. (2.3)

Here Here b(x, y, t) is the envelope of the wave packet. We obtain

T (k, kl, k2, ka)= i( O~b O~b )

Ob Ob - - Y .~ 4- s - --_iw[b 7 (2.8)
o~- 4- c -o7 ~"II ~ 9
+kz) V(--k--k, k, kl) V ( - - k ~ - - k a , k2, k~)
__ 4 a (k2 ~o2 (k~ 4 - ka) - - [ o (ke) + o (ka)l ~" To simplify the equation further we introduce the v a r i a b l e g = x -
-- ct (which corresponds to transformation to a system of coordinates
o (ka -- ka) V (-- k, k~, k - - k~) V (-- k~, kr, ka - - kr) m o v i n g with a v e l o c i t y equal to the group v e l o c i t y of the wave);
-- 4 ~ (ka - - kx) - - [ o (ks) - - o (k,)l" " we a s s u m e t h a t the solution depends only on t and z = g c o s a + y s i n a ;
we obtain
__4o (kl--k~)V(--k, ka, k - - ka) V(--.k~, k~, k ~ - - k l )
(0 ~ (k~ - - kl) - - [m (k:t) - - o (k~)] ~ -}- O~g iL 0~
Ot 2 Oz~ - - - - iw [ ~ l~ W,
+ W (k, kl, ke, ks) . (2.4)
L = LII e ~ + s sin2 a . (2.9)
Obviously, the terms o m i t t e d from the H a m i l t o n i a n (1.11.) cannot
contribute to (2.4). Equation (2.3) has the e x a c t solution
In using (2.3), we h a v e to assume that f << A. For this c o n d i t i o n
to be satisfied, i t is necessary that t h e denominators in (2.2) and(2.4) A (k) = b o 5 (k -- ko) exp [-- i t Q (k)], ~ (k) = w [bop. (2.10)
do not vanish. There is a zero denominator if
o (k) = co (kx) -1- o (kz), k = k 1 + k2 (2.5) Here b0 is an arbitrary constant. In terms of the v a r i a b l e s 7? and %
solution (2.10) has the form
h a v e a solution. I1 = a cos ( k z - - o)t), ~F = a sin ( k x - - o)t),
If this system has no solutions, Eq. (2.4) can be used for sufficiently
k'/,
s m a l l a/X, but if it has a solution we h a v e to assume a d d i t i o n a l restric-
a: [bol, r + a(#) .
tions.

192
 C a l c u l a t i o n yields Consider first the case

~q (k) = [ bol~I 4V (-- 2k, k, k)~ ]

[ k - - k o l T ~d0~
->~> o ] b p. (3.6)
9 4(o ~(k)-o)"(2k) --IV(k, k,k,k) =

Then terms proportional to b -~ can be dropped from (3.5). The con-

4o)~"(k) ~ ~o2 (2k) 4 -- t6o) (k) j
dition for the e x i s t e n c e of i n s t a b i l i t y for arbitrarily s m a l l a m p l i t u d e s
In the l i m i t , for s m a l l k this takes the form 9~(k) = l/~(ka)~w(k), is 5 = 0, which is e q u i v a l e n t to the e x i s t e n c e of solutions for the equa-
which coincides with the expression obtained by Stokes in 1847. Thus, tions
solution (2.9) a p p r o x i m a t e s a periodic wave of finite a m p l i t u d e .
When 2o (k) = m (kl) + m (k2), 2k = k l @ k2 . (3.7)

Obviously, these equations have solutions if

:* = \2g) '

r ~ >~176 (3.8)
the frequency d i s p l a c e m e n t b e c o m e s i n f i n i t e ; for l a r g e k , it is n e g a t i v e .
In the l i m i t , as k ~ ~o we h a v e
(a sufficient c o n d i t i o n ) , w h e r e the vectors k~ and k2 are p a r a l l e l to the
t same straight line. I n e q u a l i t y (3.8) is the r e q u i r e m e n t that co(k) be
(k) = - - -iV (/~ay- ~ (/0. c o n v e x upwards. For g r a v i t a t i o n a l waves,

3. S t a b i l i t y of steady waves of f i n i t e a m p l i t u d e . We consider the

d e v e I o p m e n t of sinai1 perturbations against the background of a steady
periodic wave. We seek a(k) in the form
and the i n e q u a l i t y necessarily holds.
Conversely, for c a p i l l a r y waves, the inverse i n e q u a l i t y holds, in-
a(k)=boS(k--ko)e-i~ + at(k, t) e -i~t , o=o(ko)-~O(k0).(3.1) d i c a t i n g that instability of this type is impossible.
Equation (3.7) defines a surface in k-space. Unstable vectors l i e
We assume c~(k) to be s m a l l in the sense that near this surface in a layer of thickness proportional to b 2. The m a x i -
m u m i n c r e m e n t in the i n s t a b i l i t y for g r a v i t a t i o n a l waves is of order
!]a(k) ldk~lbo[.
7~(k~V co (k).
Now we m a k e (1.15) l i n e a r in a(k). To do this, we consider only
terms on the r i g h t - h a n d side which vary slowly with t i m e . We obtain There are higher-order i n s t a b i l i t i e s corresponding to conservation
laws for m:

0a(k) - - 2 i b o e i Y t V ( - - k o , k0, k 0 - - k) et* (ko-- k); no (k) = o (kl) q- (0 (k~), nk -- k 1 i - k~.

[Ot - -
'r : ~0 (k) ~- o (k -- ko) - - m (k0) - - Q (k0). (3.2)
The order of the i n c r e m e n t for such instabilities is y(k) ~ (ka)nw(k).
E l i m i n a t i n g cd~(k0 -- k) from (3.2) , we obtain All these i n s t a b i l i t i e s can be c a l l e d destructive.
0 t i-t 0ct(k) , We turn now to i n s t a b i l i t i e s for which ! kl - kol << k0. To study
these we use Eq. (3.3) directly. The solution
8=0 (k0)
= ~ e-~'rta~-g (-- ko, k, ko - - k) ~ (k) . (3.3) b = b0 exp (-- iwlboi~t)

Equation (3.3) has a solution of the form corresponds to a c o m p l e t e l y finite a m p l i t u d e .

We now seek the solution in the form
ct (k) = ce vt ,
~F - - exp (-- iw[ be [~'t) {be + ~e -i~,* ~~ =- u*ei"'l-i~~

q=*/~i'r~-ylbo[eU~-(--ko, k, k o - - k ) - - * / G ~ . (3.4)

Instability w i l l occur i f the expression under the square-root sign Then for co we have
is positive. In order that t h e r e should be i n s t a b i l i t y for arbitrarily sinai1
b0, the e q u a t i o n ~ = 0 should h a v e a solution. If we n e g l e c t the s m a l l
~o=! g~otb0t~,<0~+li,>,~<~ . (3.9)
term a(k0) in this equation, we arrive at the system of equations (2.5).
As was established in section 2, this system can be solved for c a p i l l a r y We see from (3.9) that instability is possible if w,k < 0, i n s t a b i l i t i e s
waves; thus, instability of this type occurs for c a p i l l a r y waves. Un- only being e x c i t e d for sufficiently sinai1 wave vectors
stable wave vectors are c o n c e n t r a t e d near the surface co(k) = w(k 0 +
+ co(k -- k,) in a layer of thickness proportional to the a m p l i t u d e . The ~w
~~ < ~ - - l b0 [3 .
m a x i m u m i n c r e m e n t in the i n s t a b i l i t y is of order Req ~ (ka) co(k).
This type of i n s t a b i l i t y is impossible for g r a v i t a t i o n a l waves. How- We consider the ease of different wave numbers for the surface
ever, for these waves slower i n s t a b i l i t i e s are possible. We use Eq. (2.3) waves.
and substitute into it A(R) in the form 1. In the region of wave numbers

A (k) = bo6 (k -- ko) e-ia(k~ ~- ct (k~ t) 9 I~o < 1 / g ~ 7 ~ - t (g l ~) ' I~ ,

If we l i n e a r i z e in c~(k, t), we obtain where w > 0, X1 > 0, and ,kit < 0, the d o m a i n of instability in the
o~ (k) plane • Xy is bounded by the i n e q u a l i t i e s 0 < I kl] ~r -- X. • <
Ot -- 2 T (k, k0, k0, k) [ be [-'a (k) + < 41 btaw; i , e . , it lies b e t w e e n the hyperbola .a. y

-~ e-~if~(k~lt T ( k , 2ko -- k, ko, ko) bo~ * (2ko -- k) 9 4lbleto = s jtxac~ - - ~._~x~az

This equation can be reduced to an equation of form (3.2); it has and its asymptotes.
a solution proportional to e qt, where, for q, we have 2. In the d o m a i n
q = ~ / ,2i6 +_ V'ibo ["T ~"(k, 2ko -- k, ko, ko)'-'-- t / ~':~,
]/ VG- I (g / ~)',' <1~o < I / / : 2 (g / ~)',,,
6 = o (k) + o (2ko - - k) - - 20 (ko) -l- 2 [ bo p~[T (k, ko, ko, k) -~-
+ T (2k o - - k , ko, ko, 2 k o - - k) - - T (ko, ko, ko, ko) . (3.5) where ,k_k > 0, kl] > 0, and w > 0, i m t a b i l i t y , in g e n e r a l , is impossible.

193
 3. In the domain of capillary waves turbations increase exponentiatIy as

ko>l/ ]/"2(g/a) %, NA > O , 2. ll > O , w < O , v ~ exp (t VI w~, [ no ) .

the region of instability is the interior of the ellipse Hence we have the case of a negative-pressure type of instability.
We note that (3.9) can be obtained for the increment in the nega-
~"II:~x~"+ ~'A_• " = 41b I~ w .
tive pressure instability if we let k -'~ k 0 in (3.6). Thus, negative-pres-
In (2.9) we make the change of variables sure instability is the limiting case of slow destructive instability of
gravitational waves.
xF=]/'nexp[@Svdz I 9 The author wishes to thank L. V. Ovsyannikov and R. Z. Sagdeev
for fruitful discussions.
Equation (2.9) becomes
REFERENCES
On 3
0-7- § -2i- (n~) = O, 1. H. Lamb, Hydrodynamics [Russian translation], OGIZ-Gostekh-
izdat, 1964.
Ov Ov On k ~" 0 t 02 ]/'E. (3.10) 2. N. N. Moiseev, Surface Waves (introduction) [in Russian],
~ / - + v - O T = - w ~ - z + 2 o~ V~- Oz2
Fizmatgiz, 1960.
These equations are similar to the equations of gasdynamics with 3. S. A. Akhmanov and R. V. Khokhlov, Problems in Nonlinear
an adiabatic relationship between the pressure and the density, Optics [in Russian], Izd-vo AN SSSR, 1964.
4. V. E. Zakharov, "A solvable model for weak turbulence,"
W]%/Z2
P=--T- ' PMTF [Journal of Applied Mechanics and Technical Physics], no. 1,
p. 14, 1965.
and differ from them by an additional term containing the third de- 5. V. N. Oraevskii and R. Z. Sagdeev, "On the stability of steady
rivative with respect to z. If we consider a sufficiently large-scale longitudinal oscillations of a plasma," ZhTF, vol. 32, p. 1921, 1963.
motion with characteristic scale L, then for 6. V. N. Oraevskii, "The stability of nonlinear steady oscillations
of a plasma," Yadernyi sintez, vol. 4, no. 4, p. 263, 1964.
7. k. G. Litvak and V. I. Talanov, "The application of the para-
bolic equation to the calculation of fields in dispersing nonlinear
this term may be neglected. For positive pressure wk> 0, Eq. (3,10) media," Izv. VUZ. Radiofizika, vol. 10, no. 4, p. 539, 1967.
describes sound waves of velocity w~'~00. For negative pressure, the
speed of sound becomes imaginary, which means that the initial per- 22 June 1967 Novosibirsk

194