Académique Documents
Professionnel Documents
Culture Documents
(K
NERAVNOMERNOI ASIMPTOTICHESKOI
P#M Vo1.27,
No.2,
1963,
Roumanian
(Received
Consider
t)
nonuniform
asymptotic
there
with
are
various
respect
to of
equations,
and the
developed.
Criteria
duced,
of
criteria
Persidskii
ized
of
the
for
which
the
contain
Malkin
of Liapunov
initial
value
leading
of
The criteria
tional)
[91,
considered
stability,
are
for
also
the
special
parameter
shown that
approximation
uniform
[41,
conditions
the
of
ordinary
cases
of
the
results
(see
also
31 and
mentioned
criteria
case
general-
Note
last
intro-
criteria
the
[v, P. 384,
are
are
equivalent
and also
p.721,
behavior
differential
various
of
h,
of
first
is
system
It is shown that
[81, see also
[6, p. 3691).
contain
as a special
case the generalized
Malkin
criteria
the
It
these
in the
a group
of
to nonuniform
a linear
between
1961)
approximation.
conditions
solutions
stability
8 Republic)
12,
first
relationships
[1.21,
criteria
to
in the
types
to
People
January
(relative
stability,
231-243
pp.
REGISH
M.
(Timishoara,
USTOICHIVOSTI)
of
(i.e.
Liapunov.
uncondiof
the
present
paper.
ln this
paper
certain
ideas
of
Krein
[loI
[ill)
are
following
Bellman [121, a theorem of Banach-Steinemployed;
and also,
The paper admits of further
generalization
in
haus [13,141
is employed.
the direction
of the treatment
of the equations
considered,
as Well as
in the direction
of considering
the problem in arbitrary
Banach spaces.
and in the discussion
of the so-called
conditional
stability
(i.e.
dichotomy)
in the
sense
this
occasion
I take
which
tributed
have
influenced
to
the
of
space
a matrix
of
the
completion
all
A will
paper
to thank
the
1. Preliminary
be the
of
paper,
of
of
Massera
A. Khalana
as well
the
and Schaffer
for
his
many Suggestions.
as for
his
aid,
n by R matrices.
be denoted,
which
con-
paper.
[151.
respectively,
344
?he
by
norm of
space
a vector
and R,*
x and
On nonuniform
(5Xi3r
IIsII= II(xih6%n1=
We shall
asymptotic
II A II =
345
stability
(i
II(a&3. i<n I= zz
i=l
consider
uijr
j=l
the
following
differential
equations:
dx
= A (t) 2
v > 0)
dt
dx
- dt = A (0 5 + cp (4
P > 0)
$
where
9(t)
is
= A (t) X
a vector-valued
(1.1)
(l-2)
rt>O)
function
(1.3)
defined
on
in the following
to various
values
in the set R,, and which is subject
conditions;
A(t)
is a function
defined
on [O, + a~) and with values
in
and
such
that
there
exists
a
number
A,
>
0
such
that
11
A(t)
11
<Aa
R,* t
on
LO, + m);
lhe
will
the
solution
x(t)
be denoted
equation
If
tion:
matrix
of
by x(t,
(1.3),
for
I is
the
toI = X(t,
identity
an n by n square
equation
(1.1)
tO,
Analogously,
which
X(t,
x (Ml,&)
X is
no).
X(t,)
matrix
= X,,, will
X(V,)
= x(VI)
If
4t,
is
ro)
to,
If dt,
x0)
to,
is
to h,
a solution
x (b
(1.2),
for
the
which
solution
be denoted
x(t,)
= x0,
X(t)
by X(t,
of
t,,,
the
X0).
of R,*, then,
introducing
the usual
we have the following
relations:
to, I),
= x Wo) x0,
or
matrix.
of
Wo),
(h,h)
nota-
(Wl)
h>,O
equation
(l.l),
then
to,50)= x 0, to)x0
a solution
of
(1.2),
(1.4)
then
(1.5)
1.
Next
have
to
Let
we shall
deal
L be the
values
in
a is
a real
spaces
l
define
a class
frequently
R,,
set
of
the
all
Let
that
Equivalent
functions
measure
zero.
function
on each
us
one
of
introduce
identifies
are
spaces
with
which
we shall
sequel.
measurable
and integrable
number.
(assuming
in
functions
functions
finite
q(t)
interval,
the
following
all
equivalent*
which
coincide
on
and suppose
Banach
that
function
functions):
up to
a set
of
346
M. Rcgish
that
is,
ing
the
the
set
of
inequality
all
p in L (i.e.
functions
between
braces)
measurable
and satisfy-
and
(1.7)
that
is,
curly
II(PInLaP= IIcpIl(p.a)
=
[+fTl
cp(s)lrepsdslyP
[I, + ~1
PE
between
manner
(1.8)
IIcpII
L,oJ
ncpIl(co,
a)
es8SUP,>, IIcp(s)IIeas
(1.9)
Note1.1. If
nab which
as follows:
cp*(s)
possesses
an inverse
isometric
space Lap
(that
on the
preserves
space
lShp.
all
spaces
Q,q =
the
(1.10)
cp(s) e(a--b)s
operator
is,
In particular,
distance
is
an isomorphic
between
elements)
to the
and
map of
the
LOP = LJ.
space
Further,
if n < b, then Lap contains
LbP, and the topology
of the space
LbP is stronger
than the topology
which is induced in LbP by the topology
of
Lap
(in
other
words,
of
Lb P it
sense,
space
2.
Lb P is
employing
stronger
terminology
than the
space
the
of
functions
space
we shall
that
Lap;
the
in
say that
the
L,P) .
the consideration
to
of
on the solutions
of the equations
generalize
the restrictions
which
a series
semi
axis
>
0 by sets
of
measure
of
conditions
to be im-
(1.2)
and (1.3).
These condiare imposed on the motions of
denotes
the
The symbol ess sup,>, ,, y(s)
which are upper bounds of the function
the
norm in the
[15,161,
of
follows
the
a sequence
necessarily
lower
y(s)
zero.
bound
on sets
of
all
which
numbers
differ
from
On nonuniform
systems of
A (h,
first
approximation
a, b).
exist
There
h,
asymptotic
B (a,
b, p).
exist
There
347
stability
(t > s > 0)
w~ep*7,X(t,
s)Ipe-vds<+
ml such
that
00
here
C (a, b, p).
solution
that is
x(t,
to,
of this
The following
section
relations
\
;,
11
p (8) IIPcm ds < 00
is contained
in the following
are valid:
Equivalences
(4
A@, a, 6)wC(a, h 1)
0)
B(&hq)wC(a,hp)
pE(l*+
for $+$=I
{ !7E
001
[I#+
00)
Implications
(7)
c (6
B (a, 6, q),
4 p)
I)
b-s,
A@, U, 6)
p)
for
~Eu.+-l,
qE
[C + =)
given e > 0
lhere exists
IjX(t, s)-X(t,
lhe following
proved:
t,)j\<fjjX(t,
1-a
to)1
for
is an imnediate
sEVO--TT,
consequence
to+T]
0, to),o)
348
M. Regish
Lemma 2.2.
exists
here
real
type
t
vtu
(t,
s)
2.4 (s)
ds
(t
ia01
s
0
defines
a linear
operator
V:LaP *Lb?
(Vu)(t) = Vtu
operator,
that is to say,
there
Furthermore,
a) if
p is in (1,
+ al,
then
1
sup [i
l>O 0
(q = -f--)
e*btllX(t, s)~~qe-qa~ds]4< + 00
P--l
b) if p = 1, then
SUP @IIX(t,
dlle-8<
00
t>a>O
Proof.
Suppose
in a certain
the first
that lt,)
is the sequence of all rational
numbers, arranged
order, and suppose that the linear bounded operators
vk:LaP4Tn
are defined
(k -- 1,2, . I *)
by the formulas
v*u = ebikVtliu
Since,
for
each
II in Lap we have,
(k = 1,2, . * .)
by hypothesis,
that
On
then,
according
there
exists
nonuniform
to the
asymptotic
Banach-Steinhaus
a positive
theorem
number M for
349
stability
1131 it
follows
that
which
IIVkIId M
(k = 1, 2,.
. .)
Consequently
or
(k = 1,2, . . .)
which,
by continuity,
implies
that
bt
IItm,
b)
dMiIUntp,
a)(2.3)
ISX (4 4 u (4 ds II< M IIu IIcpsaj tt>O)OrIIvU
0
Let
us now prove
suppose
columns
that
of
a) Let
the
the
second
s),
the matrix
X( t,
p be
in
(1,
the
s)
theorem.
(j
= 1,
. . . . n) are
the
s).
+ m),
Consider
of
. . . . x,j(t,
xlj(t,
discussion.
part
and let
the
t >
0 be arbitrary,
but fixed
during
functions
ts) = (ij
(i = 1,. . ., n)
tS))l<j<n
where
e-*aszij
(t, s) [i
11X (t, s)
I? emqcu ds]-
$,
8 E IO,4
0
'{j (')=
It
is
easily
seen
ui E
Applying
then
={~
(i,i=l,...,n)
s>t
that
Lap,
inequality
Memb>
'9
IIui II@,a)
(2.3)
<
we then
(i = 1, . . ., n)
obtain
that
x(t,S)Ui(d)dsll=l\S
[$ ($raj(*S)Uij(S))
e,]dsn=
1s
(~;;:
a=1
j=l
~mj(,s)U~j(s)]d~~~~r,~[~
a=1
j=l
zij(t,o)Uij(s)]dSI=
0
j=l
M.Regish
func-
as was desired.
The case p = + m may be treated analogously,
zi j (t, s)
uij (5) =
b)
JIX(f,s))I
et
EIO# tli
s>t
"ij(s)=os
f,, h
(i, i = 1, . . ., a)
a)
YA (T)
= X-lf~,
u)
for
sE.[t0-T,
h-l-T],
to),0
asyaptot
On nonuniform
Consider
351
stability
u (8)
easily
0 <
8 &
ia,
\P
PI
seen that
II11iltl, aJ<
u E Lb,
Applying
s E 1% PI
~0,
0,
It is
ic
the inequality
(2.31,
we obtain
II~0II
that
f
Me-* >
I$
a
=$lX(t~te)~~ll~e-~s~~llX(t,t~)~~lle~~[~-a~
a
and, consequently,
II X (4 to)
t ,, T,
IIB 2,
that
t<T
the notation
h=max[2/min(ea4ed);
we obtain
of x0, we obtain
finally
that
IIX (4 to) IId
*e-Y
t>ta>o
2.2.
Proof
of
obvious.
solution
Theorem
In order
A(h, a, b) =+C(a,
b, 1) is
1) + A(h, a, b), consider the
t
2 (t, ho) = 1 X (4 4 cp(4 ds
0
0 > 0)
(2.4)
352
of
M. Regish
equation
(1.2).
In view
of
the hypotheses.
the
function
a bounded
2.2,
function
from this
for
follows
for
t >,O
the
each
existence
of
p in L,.
implies
C(a,
b,
and B(a,
us prove
us suppose
to
the condition
p)
next
say that
(2.5)
Let
is
that
that
b,
does
a,
b,
not
(2.5)
The equivalence
b).
q) =+ A(h,
holds,
q)
a,
while
us determine
Then there
is
the
which
a sequence
exist
easy
to
case,
is
In order
b).
A(h,
such
{h,)
condi-
a,
b)
to
does
do this,
not,
which
that
inequal-
sequences
that
as nd
{t,),
{s,)
see
then
that
necessarily
(2.6)
impossible.
and (2.7)
that
such
since
X(
. . .)
(2.7)
= + m; indeed,
if
on a compact
set*
hold
is continuous
in each
this
were
argument,
and
subsequences,
and labeling
them.
e
supposed
further
that
lim
may b
ret, =
S)
t,
(n = 1,2,3,
suput,
would
(2.6)
the
valid.
limh,=+m
not
of
analogously.
exist
to Theorem
that
t>s>o
can be proved
q)
6(a,
R(a,
there
A( h,
b,
According
(if
{h,,),
need
{t,),
We now consider
Suppose
a)
6(a,
be)
three
{s,),
it
two cases.
that
b,
4).
there
That
is,
on a set
at least
one limit
sup,sn
exists
such
< + m; in this
a constant
that
element.
every
case,
C > 0 such
bounded
in view
of
the
condition
it
possesses
that
subsequence
of
asymptotic
On nonuniform
e+ [ x
(t,,
s) lf
eqbtn
ds >
353
e+ 11X (t,,
s) If
eqbtn
ds
in
for
stability
all
which
is
an+T
e-qcrr
11X (t,,
sn)
If
eqbfn
ds > $ e-qaThnq+
+-
as
b)
now that
suppose
lim s, = +a, as
SUP,,S~
0~; in
this case we may assume that
we obtain the inequalas in case (a),
- m* and then.
io
In
11x (t,,,
c> e-qar
S)Iv eqbfn
ds >
as n - t a,
which
Finally,
in
(1,
the
e*a
-, -I- CKJ
%a- T
the
al
is
also
impossible.
implication
A(h,
and arbitrary
a,
b) +C(a,
E > 0 follows
in
b -
E,
p)
for arbitrary
a corresponding
manner
from
inequality
t
(b-c); <e(b--r)t
11X
(1
or
s
0
(i
< htf
IIx V, 0, 0) IIe(b--r)f
d ( ~2 IIcp(4 0eonIdbbc)1IIX k.4 IIe ds < hPf
I cplltoo,a) < G
large
constants.
The theorem
is
thus
3.
Application to the problem of stability to the first
approximation. Consider
the equation
dx
dt
= A (t) 2 + @(z,
q,
XER,
(3.1)
354
M. Regish
t>o,
and satisfying
t) is a
P>O)
1141\<
D
(3.2)
the condition
IIW,
(m>,l)
t)II\(fWT
valued
function
f(t)
(3.3)
are specified
more
t),to>o
11
x (t, to)(I< he-(+),
These criteria
second class
to say,
uniform
tion,
constitute
one finds
the extensions
case.
the class
of uniform
criteria.
criteria
[5,8,9,181,
the generalized
of the criteria
For example,
(3.4)
of the first
the condition
class
of Malkin,
that
is
to the non-
in the present
nota-
nX(t, to)I~<hf?t~e-a(f-t~),
m >+
(3.5)
In the
t)
(4
satisfies
a>,h
a>b,
m=+,
m=a
a,
D1,
of first
and that
the following
cases:
fEhlP, 1lfU(,,,)~K,
--
,I
--
,-?
+ ~19
ll~oIl~Dl
(1.1)
> 0,
e@--mb)s(llq, 0)]
1) hm K, 11
P = 1,
a = b,
approximation
a >b
and consider
[I, + c=lv
PE
11
x0 II< [ 2 (m -
/lx0ll<
(7)
A(h,
the inequality
m>t,
11
x0 II< D1,
@I
condition
in the first
mentioned.
--
,-I
f EL07
e- *
IlfIl(P,o,\<~P
to
On nonuniform
(I--L,
where
P--l
every solution
D1 <D,
x(t,
(a), (Pi,
to,
K,,
355
stability
are constants.
p E [1, -+ cc]
x0> of equation
(r) -=+I/z (6
@)=qz
asymptotic
(3.1)
satisfies
A@*
cpct
t>
))20Jj,
the implications
tZ&),
to, "o)(I\<Heaf~-bfl/20(1,
Then
t, >o
where
H = 2h
Proof
Theorem
in case(a)
Ye have that
3.1.
~1 IIdr,
t),t,>o
$0
Setting
t =to+s,
and observing
7=t*+c,
the condition
II3 00+ s,
A(h,
-W,+s&,,(0) +
cp(s) < heat62
a,
b),
to,5) If=
~~~~=ll~oll
gl (4,
we obtain
that
Y
/&f,+@ ,-~(fo+s)f (to $ a) pm (3) dz
s
k=
hefawb)O,
It is readily
h>t
h,
is defined
by the equation
S),O
(3.6)
then,
while,
on the other
hand, a simple
$+btb=
In cases (a),
$5) this
Q(s)'@",
equation
s>o
differentiation
of
(3.6)
leads
to
type.
(3.7)
Solvfng
it,
we
356
M. Regish
obtain
--
- (m - i) ke(-1)s
5 &-@f (to + a) da
0
But
~(0)
1
m-1
8
consequently
1
-cp(s)<
ke-"*q (0) { 1 -
1) km [cp(O)]
(m -
s>o
j &-@j
(3.9)
Case
(a).
Since
s/uo+
In this
k = hecoBb)
case
t0,
and
$+$=i
0) e(a--mb)o
dad IIf llcp,o) IIe(a--mb)a
/Itq,o),
(3.9)
cp(4 B
or,
in the preceding
(a)
1
m-1
for
he(+Woe-b6cp
(0) =2
tkl
with
(3.8)
and (3.9),
~a'oe-Wo+Qcp(o)
notation
f&2&
(3.10)
and Inequality
@) for
in an analogous
Cases (y) and (6). In these cases the proof proceeds
manner, the difference
being that equation
(3.7)
is linear;
however, in
these cases the results
coincide
essentially
with the results
of Krein
[lOI and Kucher[lll.
Theoren
a,
3.2. If equation
(1.2) satisfies
condition
C(a, b, r),
in [l, + ml, then all the conclusions
of Theorem 3.1 are
Further,
in case (p) the following
improvement is possible:
b > 0,
valid.
(p).
Suppose
that
for
a > b,
II >a/b
where
9 E
La,
II\PII@,GKrs
a)
rE[l,
+ ml
W, > 0)
m+--q~o
II*
t>to>o,
fELJ,,
Ilfllo%ot<~~
asymptotic
On nonuniform
Proof.
Since
357
stability
that C(a,
b,
r) =+ A(h,
a,
b),
then
the first
part of the theorem follows automatically.
In order to prove
the last part, let us suppose that (1.2) satisfies
the condition
C(a,
b, r) for some r in 11. + ml, a > b > 0 and that a >, a/b. From C(a,b,r)
it follows that
11x (f, to) nq hCab-bf,
Let us choose
and let
h > l/4,
us prove that
-bt
I = U, 10920)II< hale
provided
t>ro>O
II=0II.
t>,to
that
-- -_1
-
11ebt 1 X (4 8) 9 (4 ds I,, o) }
n
Since
IIz(to, to, ~~)n=nzou<4hIIzoUlc(~~)r,
it follows that, whenever
does indeed
near to, the required inequality
t > to, with t sufficiently
hold. Let us suppose that there exists some number T > t,, for which
II 5 ll f Wea41-1
{ 4wJ,
for
t in [to,
all
T) we have
II 2 k
This supposition
the function
It is readily
will
to,
5)
II <
4hafoe
-bt
(3.11)
II x0 u
seen that
rt,,
From this,
I) that
But,
for
Consider
together
with condition
these values
of t we have
C(a,
b,
II \P lltr, a) < Kr
r),
it
follows,
for
t in
358
M. Regish
that
is
to
say,
From this,
finally
that
in
view
of the
condition
imposed
upon
II *o II, we obtain
that
which
is
impossible,
Thus
Let
the
since
theorem
is
us now prove
values,
then
it
contradicts
the
assumption
proved.
if
that
one always
equation
(1.1)
has a certain
possesses
theorem
of
positive
stability
eigenin the
first
approximation.
Let
us denote
>
PI >e.*
p,
by h, <
dY
dz=
Then the
following
Theorem 3.3.
certain
h > 0,
condition
conclusions
Let
a, <
A, > 0,
the
of
us denote
this
then
for
follows
(i,
j = 1,
equation
(l.lj
(see,
for
<hn
be
the
corresponding
.*.
By N(t)
normal
arbitrary
satisfies
of
hj = eigenvalue
we shall
solutions
just
of (N,j (t),
denote
and by
holds:
(1.1)
by Ni .(t)
(1.11,
Y E R,
- yfl (a
theorem
of
equation
p,,
d(h,
al.
. . . , n)
. . . , Nnj
b < h,,
a,
and a
b) and even
Consequently,
example,
[6,
eigenvalues
the matrix
mentioned
cz <
condition
Ike proof
tions
If
of
theorem
equation
An the eigenvalues
...<
the eigenvalues
below.
a normal
system
pp.323-3321)
(t])
whose columns
the
of
solu-
and let
(j = 1,.
. , nf
consist
of
the
On nonuniform
asymptotic
Ibis
matrix
possesses
the
following
N (t) E K*,
b)
4
d)
If
the elements
e)
If
cl,>
of
one has
where o>O
is,
that
is
of
(see,
the
(s>O)
inverse
matrix
eigenvalues
of
{Ni;(s),
for
of(exp
example,
this
Lemma 3.1.
If
the matrix
of
denoted
N-(s),
by N*-l,
that
is
(i = 1, . . . , n)
irregularity
of
equation
(l.l),
that
by the equation
5 Sp A (U) d)
terminology,
A, = -
(3.12)
(0 > 0)
a=1
the
following
such
are
[191)
Employing
N-(s)
. . . , Ni,(S)}
coefficient
eigenvalue
the
are
. . . >cl,
pi = eigenvalue
then
v, to>,w
= - N- (s) A (s)
ds
properties:
t>o
a)
dN-l(s)
(i=l,..,,n)
t>o
{Nlj(t), - - * Nnj(t)},
359
stability
theorem
there
is
exists
valid:
a number
A)>0
that
11
X
ttt
the,least
to> 0)
eigenvalue
t > 0,
(3.13)
of
kl (A)> 0
the matrix
(3.14)
Similarly
IIN-
to > 0,
ha(P)>0
(3.15)
360
M. Regish
From property
(b),
together
with
(3.14)
and (3.151,
we then deduce
that
IIx @c
to)nG uN 0) 11iv- (to)I]Q he+0 e--A,
where h = h,h,,
Lemma 3.2.
is
to>0
t,
proved.
exists
a number
k = k (e) > 0
such that
] X (t, to) I; < ke(s+W 4 e-h-~)
where o is
the
Let
Proof.
number
by equation
defined
us denote
the
columns
From (b)
we have
zlj
From this,
U--I.),
of
t,
(3.16)
to>0
(3.12).
the
matrix
X( t,
to)
by
(i=l,...,n)
tO))
that
(t, to) = i
keeping
in mind
property
(i,i=
(e),
1,. . .,n)
we obtain
that
a=1
are
CDij, kij = n mar, C,i
6 = q = E > 0, we obtain
finally
where
11X
as
was to
Note
generalized
they also
since
(1,
to) 11(
constants.
ke(6+ac)re-(a,-L)(f-f.)
Consequently,
putting
0, toso)
be shown.
3.1.
The generalized
of Liapunov
criteria
criteria
of Malkin (M, or [6,
follow
from the criteria
expressed
Liapunovs
condition
means
that
m>i+t
[VI
(i
follow
from
the
p. 3691),
and consequently.
in Theorem 3.1.
Indeed,
=i,...,n)
On nonuniform asymptotic
361
stability
where .Wis the number appearing in Theorem 3.1, Ai are the eigenValUeS
of equation (1.1).
and a is the number given by (3.12),
then one may
choose a sufficiently
stnall E $ 0 such that
of Malkins criteria
Now, since,
in view of Lemma 3.2, the conditions
are satisfied,
this is just what was required to be proved. One may
still
note that (as may be easily seen from the examples given by Malkin
([6, p. 3681 or Id)), the criteria
of Malkin are not equivalent
to
Liapunov s criteria.
Note
3.2. The criteria
presented in 193, formulated for the case of
ordinary (i.e.
unconditional)
stability,
are included in Theorem 3.2
for m = 2. a = 2k, b = k, p = + 0~.
BYBLIOGRAPHY
1.
ustoichivosti
integralov
sistem diffePersidskii,
K. P., K teorii
rentsialnykh
uravnenii (On the stability
of systems of differfor Kazan
Univcrs
ity,
ential equations).
Fiz. Math. Obshest.
Vol.
6, No. 3. 1936-1937.
2.
dvizheniia
PO pervomu pribliPersidskii,
K. P., Ob ustoichivosti
zheniiu (On the stability
of motion to the first
approximation).
Maten. Sbornik, Vol. 40. 1933.
3.
Malkin,
bility
4.
5.
po pervomu priblizheniiu
I. G., Ob ustoichivosti
to the first
approximation).
Sbornik
nauchnykh
aviation
institute,
NO. 3. 1935.
Perron,
0..
Zeit.
32.
Liapunov.
General
6.
Yalkin,
7.
6.
A.M., Obschaia
Problem of the
of
Massera,
Vol.
Die StahilitHtsfrage
bei Differentialgleichungen.
5 (Schluss)
Heft. 1930.
I. G. , Teoriia
bility
(On statrudov
Motion).
J. L.,
zadacha
Stability
ustoichiuosti
po
ustoichivosti
of
Motion).
duizheniia
Math.
dvizheniia
(The
Gostekhizdat,
(Theory
of
Kazan.
1950.
the
Sta-
GITTL, 1952.
Contributions
to stability
64, pp. 162-206, 1956.
theory.
Annals of Math.
362
9.
M. Regish
Maizel , A. D. , Ob ustoichivosti
PO pervomu priblizheniiu
(On stability
to the first approximation).
PMM Vol. 14. No. 2, 1950.
10. Krein, M.G., 0 nekotorykh voprosakh sviazannykh s krugom idei Liapunova v teorii
ustoichivosti
(On certain questions related to
Liapunovs ideas concerning the theory of stability).
Uspekhi
Mater.
Nauk,
Vol. 3, No. 3/25, 1948.
11. Kucher, D.L., 0 nekotorykh
sistemy differentsialnykh
boundedness of solutions
Dokl.
Akad.
Nauk
SSSR.
kriteriiakh
ogranichennosti
reshenii
uravnenii (On certain criteria
for
of systems of differential
equations).
Vol. 19. No. 5. 1949.
Lwow,
Trigonometrical
Vol.
series.
Monografye
de la condensation
Matemat
iyczne,
des
Vars-
5. 1935.
equations and
15. Massera, J.L. and Schlffer,
J. J., Linear differential
functional
analysis,
4. Math.
Annalen
139, pp. 287-342, 1940.
equations and
16. Massera, J. L. and SchZiffer, J. J., Linear differential
functional
analysis,
I. Annals
of Math. Vol. 67. NO. 3, 1958.
Theory
17. Bellman, R., Stability
Toronto, London. 1953.
of
Differential
Equations.
New
0..
Zeit.
Die Ordnungszahlen
Vol.
31,
York,
Lucrzrile
der Differentialgleichungssysteme.
1929.
Trans
lated
by
J. B. D.