Ann.
Acad.
Rom.
Sci.
Ser.
Math.
Appl.
ISSN
2066-6594
Vol.
18,
No.
3/2026
LOCAL
ERGODIC
PROPERTIES
FOR
POWERS
OF
OPERATORS
IN
BANACH
SPACES
∗
Laurian
Suciu
†
Dedicated
to
the
memory
of
Professor
Mihail
Megan
for
his
scientific
contributions
and
enduring
enthusiasm
for
mathematics
DOI
10.56082/annalsarscimath.2026.3.129
Abstract
In
this
paper
we
deal
with
the
local
convergence
of
a
sequence
{
T
n
}
in
the
convex
hull
of
the
powers
of
an
operator
T
on
a
Banach
space
X
.
We
refer
to
the
case
when
the
sequence
{
T
n
(
T
−
I
)
}
is
bounded,
and
to
the
subspaces
of
X
on
which
these
two
sequences
converge
to
zero,
respectively.
Also,
we
study
the
behavior
of
f
(
T
n
x
)
or
f
(
T
n
(
T
−
I
)
x
)
for
x
∈
X
and
f
in
the
dual
space
X
∗
of
X
,
in
particular
when
X
is
a
Banach
algebra
and
f
a
character
of
X
.
Some
consequences
concerning
the
Ces`
aro
means
of
higher
order
are
also
derived.
Keywords:
Ces`
aro
mean,
ergodicity,
ascent
of
an
operator.
MSC:
47A35,
47A10,
47A16.
1
Introduction
and
preliminaries
Let
X
,
Y
be
Banach
spaces
and
B
(
X
,
Y
)
be
the
Banach
space
of
all
bounded
linear
operators
T
:
X
→
Y
.
As
usual
denote
B
(
X
)
=
B
(
X
,
X
),
which
is
a
Banach
algebra
with
the
identity
operator
I
∈
B
(
X
).
The
kernel,
the
range
and
the
spectrum
of
T
∈
B
(
X
)
will
be
denoted
by
N
(
T
)
,
R
(
T
)
and
σ
(
T
),
∗
Accepted
for
publication
on
June
17,
2026
†
laurians2002@yahoo.com
,
Lucian
Blaga
University
of
Sibiu,
Department
of
Mathe-
matics
and
Informatics,
Sibiu,
Romania
129
Local
ergodic
properties
for
powers
of
operators
130
respectively.
For
a
subspace
M
⊂
X
,
M
stands
for
the
norm
closure
of
M
into
X
.
Also,
X
∗
and
T
∗
∈
B
(
X
∗
)
denote
the
dual
space
of
X
and
the
adjoint
operator
of
T
,
respectively.
The
ascent
of
T
,
denoted
asc(
T
)
is
the
smallest
nonnegative
integer
n
with
N
(
T
n
)
=
N
(
T
n
+1
).
For
such
an
operator
T
let
co
(
T
)
be
the
(infinite)
convex
hull
of
the
powers
T
n
(
n
≥
0)
of
T
.
More
exactly,
an
element
T
∈
co
(
T
)
has
the
form
T
=
j
≥
0
a
j
T
j
,
a
j
≥
0
,
j
≥
0
a
j
=
1
,
(1)
such
that
if
a
j
>
0
for
infinitely
many
values
of
j
,
then
the
series
is
norm
convergent
in
B
(
X
).
In
this
paper
we
study
some
local
ergodic
properties
for
the
sequences
{
T
n
}
n
≥
0
⊂
co
(
T
)
for
an
operator
T
∈
B
(
X
).
So,
we
continue
a
study
in
this
sense
begun
in
[
1
,
14
,
16
,
17
].
These
investigations
are
in
connection
with
other
ergodic
results
obtained
by
[
6
,
7
,
9
–
13
]
and
many
other
authors.
The
results
of
the
paper
are
applied
to
the
Ces`
aro
means
of
higher
order
denoted
M
(
p
)
n
(
T
),
for
T
∈
B
(
X
)
and
integers
p,
n
≥
0.
We
put
M
(
p
)
0
(
T
)
=
I
,
M
(0)
n
(
T
)
=
T
n
,
and
for
n,
p
≥
1
are
defined
by
M
(
p
)
n
(
T
)
=
p
(
n
+
1)
...
(
n
+
p
)
n
j
=0
(
j
+
p
−
1)!
j
!
M
(
p
−
1)
j
(
T
)
(2)
=
p
(
n
+
1)
...
(
n
+
p
)
n
j
=0
(
n
−
j
+
p
−
1)!
(
n
−
j
)!
T
j
=
p
n
+
p
n
j
=0
(1
−
j
n
+
1
)(1
−
j
n
+
2
)
...
(1
−
j
n
+
p
−
1
)
T
j
.
It
is
easy
to
see
that
these
satisfy
for
p,
n
≥
1
the
following
relations:
M
(
p
)
n
(
T
)(
T
−
I
)
=
p
n
+
1
(
M
(
p
−
1)
n
+1
(
T
)
−
I
)
,
(3)
TM
(
p
)
n
(
T
)
=
n
+
p
+
1
n
+
1
M
(
p
)
n
+1
(
T
)
−
p
n
+
1
I,
(4)
n
+
p
+
1
n
+
1
M
(
p
)
n
+1
(
T
)
−
M
(
p
)
n
(
T
)
=
p
n
+
1
M
(
p
−
1)
n
+1
(
T
)
,
(5)
M
(
p
)
n
(
T
)
≤
max
0
≤
j
≤
n
M
(
p
−
1)
j
(
T
)
≤
max
0
≤
j
≤
n
T
j
.
(6)
L.
Suciu
131
Remark
by
the
spectral
mapping
theorem
that,
if
1
∈
σ
(
T
),
then
1
∈
σ
(
M
(
p
)
n
(
T
))
for
all
n,
p
,
hence
M
(
p
)
n
(
T
)
≥
1
in
this
case.
The
operator
T
is
p
-Ces`
aro
ergodic
,
or
p
-Ces`
aro
bounded
if
the
sequence
{
M
(
p
)
n
(
T
)
}
is
strongly
convergent,
or
it
is
bounded,
in
B
(
X
),
respectively.
When
T
is
p
-Ces`
aro
ergodic
the
limit
operator
is
the
bounded
projection
P
∈
B
(
X
)
onto
R
(
P
)
=
N
(
T
−
I
)
with
N
(
P
)
=
R
(
T
−
I
)
(see,
e.g.,
[
8
,
Corollary
VIII.
5.2,
p.
662],
[
12
,
Theorem
1.3,
p.
73]).
In
this
paper
we
study
some
local
ergodic
properties
for
the
sequences
{
T
n
}
⊂
co
(
T
)
which
are
not
necessary
bounded,
but
they
satisfy
local
bound-
edness
conditions.
So,
in
Section
2
we
refer
to
the
case
when
{
T
n
}
is
bounded
on
R
(
T
−
I
),
or
only
on
R
((
T
−
I
)
p
)
for
p
>
1,
in
order
to
investigate
the
subspaces
on
which
{
T
n
}
and
{
T
n
(
T
−
I
)
}
converge
to
zero.
In
the
same
sense,
we
assume
some
boundedness
and
regularity
conditions
in
vectors
from
X
,
to
obtain
the
convergence
of
the
corresponding
operator
sequences
in
those
vectors.
Our
investigations
are
based
on
the
intertwining
of
T
with
an
operator
on
another
space,
which
has
good
ergodic
properties.
In
Section
3
we
explore
the
similar
boundedness
and
regularity
conditions
for
sequences
{
T
n
}
⊂
co
(
T
),
when
T
acts
on
a
Banach
algebra
X
.
We
assume
such
condition
with
respect
to
characters
of
X
(the
linear
and
multiplicative
functionals
on
X
).
These
conditions
permit
to
obtain
the
convergence
to
zero
of
the
corresponding
sequences,
with
respect
to
the
respective
functionals
in
X
∗
and
the
vectors
from
X
.
Other
consequences
regarding
the
means
M
(
p
)
n
(
T
)
are
derived
from
these
results.
2
Ergodic
properties
for
sequences
in
the
convex
hull
of
an
operator
In
what
follows
we
investigate
ergodic
properties
from
the
context
of
Ces`
aro
means
for
more
general
operator
means.
Namely,
we
refer
to
sequences
from
the
convex
hull
of
the
powers
of
an
operator
on
a
Banach
space
X
.
We
begin
with
the
following
Theorem
1.
Let
T
∈
B
(
X
)
and
{
T
n
}
⊂
co
(
T
)
be
such
that
T
n
|
R
(
T
−
I
)
=
O
(1)
as
n
→
∞
.
(7)
Then
there
exists
a
Banach
space
X
and
the
operators
Q
∈
B
(
X
,
X
)
and
V
∈
B
(
X
)
satisfying
QT
=
V
Q
,
R
(
Q
)
=
X
and
N
(
Q
)
=
{
x
∈
X
:
T
n
(
T
−
I
)
x
→
0
,
n
→
∞}
.
(8)
Local
ergodic
properties
for
powers
of
operators
132
Moreover,
for
every
integer
m
≥
1
we
have
X
0
:=
{
x
∈
X
:
T
n
x
→
0
,
n
→
∞}
⊂
N
(
Q
)
∩
R
((
T
−
I
)
m
)
,
(9)
and
the
following
conditions
are
equivalent:
(i)
R
((
T
−
I
)
m
)
⊂
N
(
Q
)
;
(ii)
(
V
−
I
)
m
=
0
;
(iii)
X
0
=
R
((
T
−
I
)
m
+1
)
.
Proof.
The
subspace
X
0
of
X
from
(
9
)
is
invariant
for
T
and
T
n
because
TT
n
=
T
n
T
and
{
T
n
}
satisfies
(
7
).
Let
S
=
T
|
X
0
,
S
n
=
T
n
|
X
0
for
n
∈
N
.
As
usually,
X
∗
0
denotes
the
dual
space
of
X
0
.
Then
for
every
x
∈
X
0
and
f
∈
X
∗
0
we
have
(by
(
9
))
(
S
∗
n
f
)
x
=
f
(
S
n
x
)
→
0
as
n
→
∞
.
But
S
∗
n
are
convex
combinations
of
powers
of
S
∗
(by
the
choice
{
T
n
}
⊂
co
(
T
)).
So,
if
f
∈
N
(
S
∗
−
I
)
then
S
∗
n
f
=
f
,
and
by
the
above
convergence
one
obtains
f
(
x
)
=
0
for
x
∈
X
0
,
that
is
f
=
0.
Thus
N
(
S
∗
−
I
)
=
{
0
}
which
implies
N
((
S
∗
−
I
)
m
)
=
{
0
}
for
every
integer
m
≥
1.
Hence
we
have
X
0
=
R
((
S
−
I
)
m
)
⊂
R
((
T
−
I
)
m
)
,
for
m
≥
1
.
(10)
Since
sup
n
∈
N
T
n
(
T
−
I
)
x
<
∞
for
x
∈
X
one
can
define
the
(continu-
ous)
seminorm
γ
on
X
by
γ
(
x
)
=
lim
sup
n
→∞
T
n
(
T
−
I
)
x
,
x
∈
X
.
So
N
(
γ
)
=
{
x
∈
X
:
T
n
(
T
−
I
)
x
→
0
,
n
→
∞}
is
a
closed
subspace
of
X
,
which
is
invariant
for
T
and
T
n
,
because
T
n
T
=
TT
n
(
n
∈
N
).
Let
now
X
be
the
completion
of
the
normed
quotient
space
X
/
N
(
γ
)
with
the
norm
x
=
γ
(
x
)
,
x
=
x
+
N
(
γ
)
∈
X
.
Let
Q
be
the
quotient
mapping
of
X
into
X
,
that
is
Qx
=
x
for
x
∈
X
.
Then
R
(
Q
)
=
X
,
N
(
Q
)
=
N
(
γ
),
as
in
(
8
),
and
clearly
X
0
⊂
N
(
γ
)
which
together
with
(
10
)
give
the
inclusion
(
9
).
On
the
other
hand
we
have
by
the
above
relations
that
QTx
=
Tx
=
γ
(
Tx
)
≤
T
γ
(
x
)
≤
T

Qx
L.
Suciu
133
for
all
x
∈
X
.
So,
there
exists
an
operator
V
∈
B
(
X
)
such
that
QT
=
V
Q
.
This
implies
Q
(
T
−
I
)
m
=
(
V
−
I
)
m
Q
for
every
integer
m
≥
1.
There-
fore
R
((
T
−
I
)
m
)
⊂
N
(
Q
)
if
and
only
if
(
V
−
I
)
m
=
0,
which
means
the
equivalence
of
(i)
with
(ii).
But
the
inclusion
from
(i)
implies
that
R
((
T
−
I
)
m
+1
)
⊂
X
0
,
and
later
by
(
10
)
this
implies
that
X
0
=
R
((
T
−
I
)
m
+1
),
that
is
(iii).
Hence
the
conditions
(i)
and
(iii)
are
equivalent,
what
completes
the
proof.
Corollary
1.
Let
T,
T
n
and
X
0
be
as
in
Theorem
1
.
If
T
n
(
T
−
I
)
m
x
=
o
(1)
as
n
→
∞
for
an
integer
m
≥
1
and
each
x
∈
X
,
then
X
0
=
R
((
T
−
I
)
m
)
and
asc
(
T
∗
−
I
)
≤
m
.
Corollary
2.
Let
T
and
Q
be
as
in
Theorem
1
,
and
let
S
=
T
|
N
(
Q
)
.
Then
asc
(
S
∗
−
I
)
≤
1
.
Proof.
Preserving
the
notation
from
the
proof
of
Theorem
1
and
denoting
S
n
=
T
n
|
N
(
Q
)
for
n
∈
N
,
we
have
for
f
∈
N
(
Q
)
∗
and
x
∈
N
(
Q
),
(
S
∗
n
(
S
∗
−
I
)
f
)
x
=
f
(
S
n
(
S
−
I
)
x
)
→
0
,
as
n
→
∞
.
Now,
if
f
∈
N
((
S
∗
−
I
)
2
)
then
(
S
∗
−
I
)
f
∈
N
(
S
∗
−
I
),
and
so
for
x
∈
N
(
Q
)
we
have
((
S
∗
−
I
)
f
)
x
=
(
S
∗
(
S
∗
−
I
)
f
)
x
=
(
S
∗
n
(
S
∗
−
I
)
f
)
x,
taking
into
account
that
S
∗
n
∈
co
(
S
∗
)
because
T
n
∈
co
(
T
).
Thus
by
the
above
convergence
we
obtain
(
S
∗
−
I
)
f
=
0,
that
is
f
∈
N
(
S
∗
−
I
).
Having
in
view
the
choice
of
f
we
infer
that
N
((
S
∗
−
I
)
2
)
=
N
(
S
∗
−
I
),
this
meaning
that
asc(
S
∗
−
I
)
≤
1.
Remark
1.
Concerning
Theorem
1
it
is
possible
to
have
X
0
=
R
((
T
−
I
)
2
)
=
R
(
T
−
I
)
,
in
general.
For
example,
let
T
∈
B
(
X
)
with
(
T
−
I
)
2
=
0,
T
=
I
.
Then
M
n
(
T
)(
T
−
I
)
=
nM
(2)
n
−
1
(
T
)(
T
−
I
)
2
+
T
−
I
=
T
−
I
=
0
,
n
≥
1
and
X
0
=
R
(
T
−
I
)
2
=
{
0
}
when
T
n
=
M
n
(
T
).
This
sequence
just
strongly
converges
on
R
(
T
−
I
),
but
it
is
unbounded
on
X
because
M
n
(
T
)
=
n
(
n
−
1)
M
(3)
n
−
2
(
T
)(
T
−
I
)
2
+
n
(
T
−
I
)
+
I
=
n
(
T
−
I
)
+
I,
n
≥
2
.
Local
ergodic
properties
for
powers
of
operators
134
Since
M
n
(
T
)(
T
−
I
)
2
=
0
we
have
R
(
T
−
I
)
⊂
N
(
Q
),
the
inclusion
being
strict
in
this
case
(as
T
−
I
=
0).
Also,
we
remark
that
R
(
T
−
I
)
∩
N
(
T
−
I
)
=
R
(
T
−
I
)
=
{
0
}
,
a
condition
which
is
not
compatible
with
ergodic
theory.
Corollary
3.
Let
T
∈
B
(
X
)
and
{
T
n
}
⊂
co
(
T
)
be
a
bounded
sequence.
Then
there
exists
a
Banach
space
X
and
the
operators
Q
0
∈
B
(
X
,
X
)
,
V
0
∈
B
(
X
)
such
that
Q
0
T
=
V
0
Q
0
,
R
(
Q
0
)
=
X
and
N
(
Q
0
)
=
{
x
∈
X
:
T
n
x
→
0
,
n
→
∞}
⊂
R
((
T
−
I
)
m
)
for
every
integer
m
≥
1
.
Moreover,
N
(
Q
0
)
=
R
(
T
−
I
)
if
and
only
if
V
0
=
I
,
and
in
this
case
asc
(
T
∗
−
I
)
≤
1
.
Proof.
Since
{
T
n
}
is
bounded
one
can
also
define
the
seminorm
γ
0
on
X
by
γ
0
(
x
)
=
lim
sup
n
→∞
T
n
x
,
x
∈
X
.
Proceeding
as
in
the
proof
of
Theorem
1
we
get
a
Banach
space
X
and
the
operators
Q
0
and
V
0
satisfying
Q
0
T
=
V
0
Q
0
,
R
(
Q
0
)
=
X
and
N
(
Q
0
)
=
{
x
∈
X
:
T
n
x
→
0
,
n
→
∞}
⊂
R
((
T
−
I
)
m
)
,
m
≥
1
.
Since
Q
0
(
T
−
I
)
=
(
V
−
I
)
Q
0
it
follows
that
R
(
T
−
I
)
⊂
N
(
Q
0
)
if
and
only
if
V
0
=
I
.
In
this
case
from
the
above
inclusion
we
infer
N
(
Q
0
)
=
R
((
T
−
I
)
m
)
for
m
≥
1.
This
gives
R
(
T
−
I
)
=
R
((
T
−
I
)
2
)
that
is
asc(
T
∗
−
I
)
≤
1.
Remark
2.
Notice
that
only
the
boundedness
condition
of
a
sequence
{
T
n
}
⊂
co
(
T
)
cannot
give
information
concerning
the
convergence
of
{
T
n
}
,
in
any
way.
More
precisely,
under
this
condition
we
cannot
describe
the
whole
set
of
(strongly)
convergence
of
{
T
n
}
,
neither
to
provide
the
equality
in
the
inclusion
(
A
)
X
0
⊕
N
(
T
−
I
)
⊂
{
x
∈
X
:
{
T
n
x
}
converges
in
X}
.
When
the
sequence
{
T
n
}
is
ergodic
that
is
X
0
=
R
(
T
−
I
),
the
equality
in
(A)
holds,
which
means
V
0
=
I
by
Corollary
3
,
that
is
an
extreme
case.
We
want
to
see
if
the
equality
occurs
by
other
conditions
on
{
T
n
}
and
V
0
,
for
instance
V
0
to
be
an
isometry.
This
idea
leads
to
another
ergodic
concept,
as
follows.
L.
Suciu
135
We
say
that
a
sequence
{
T
n
}
⊂
co
(
T
)
is
(
uniformly
)
regular
if
it
is
bounded
and
there
exists
an
integer
n
0
≥
0
such
that
{
TT
n
−
T
n
+
n
0
}
strongly
(respectively,
uniformly)
converges
to
zero.
In
this
case,
{
T
n
}
is
called
(uni-
formly)
regular
of
order
n
0
.
Clearly,
the
ergodicity
means
regularity
of
order
n
0
=
0.
But
{
T
n
}
can
be
regular
of
an
order
n
0
≥
1
without
being
ergodic.
Ces`
aro
averages
of
order
p
form
a
regular
sequence
of
order
1
(see
relation
(
4
)
below)
but
these
means
are
not
necessary
ergodic
sequences,
in
general.
Other
examples
can
be
provided
as
follows
Example
1.
Let
T
∈
B
(
X
)
be
Ces`
aro
bounded
such
that
{
1
n
T
n
}
does
not
strongly
converge
to
zero.
(Such
operators
are
provided
in
[
6
],
for
instance;
or
on
C
2
can
be
considered
the
Assani’s
example
T
=
−
1
2
0
−
1
).
Let
{
T
n
}
∈
co
(
T
)
given
by
T
n
=
1
2
(
M
n
(
T
)
+
M
n
+1
(
T
))
,
n
∈
N
.
Then
{
T
n
}
is
bounded
and
by
the
relation
(
4
)
we
have
TT
n
−
T
n
+1
=
1
2
(
TM
n
(
T
)
−
M
n
+1
(
T
)
+
TM
n
+1
(
T
)
−
M
n
+2
(
T
))
→
0
,
hence
{
T
n
}
is
regular
of
order
1
.
Clearly,
by
recurrence
we
can
find
such
a
sequence
for
any
order
n
0
>
1
.
On
the
other
hand
one
has
T
n
(
T
−
I
)
=
1
2
T
n
+1
n
+
1
(
T
+
I
)
−
T
n
+2
(
n
+
1)(
n
+
2)
−
2
n
+
3
(
n
+
1)(
n
+
2)
I
.
As
{
M
n
(
T
)
}
is
bounded,
the
second
term
in
the
right
side
uniformly
con-
verges
to
zero.
Thus,
assuming
that
R
(
T
+
I
)
=
X
,
under
the
above
assumption
on
T
it
follows
from
this
relation
that
{
T
n
(
T
−
I
)
}
does
not
strongly
converge
to
zero,
that
is
{
T
n
}
is
not
ergodic.
As
another
case,
we
can
consider
T
an
invertible
isometry
on
X
such
that
1
,
−
1
/
∈
σ
(
T
)
and
T
n
=
1
2
(
T
n
+
T
n
+1
)
,
n
∈
N
.
Here
TT
n
−
T
n
+1
=
0
therefore
{
T
n
}
is
regular
of
order
1,
but
T
n
(
T
−
I
)
=
1
2
T
n
(
T
2
−
I
)
does
not
converge
strongly
to
zero,
hence
it
is
not
ergodic.
Notice
that
in
[
17
]
the
regularity
condition
was
used
to
derive
certain
er-
godic
properties
for
the
Ces`
aro
means
of
higher
order,
as
well
as
to
obtain
two
extensions
of
the
Esterle-Katznelson-Tzafriri
theorem
for
some
sequences
in
the
convex
hull
of
an
operator.
Now
we
present
results
on
local
convergence
for
some
sequences
of
oper-
ators
which
satisfy
local
regularity
conditions.
Local
ergodic
properties
for
powers
of
operators
136
Theorem
2.
Let
T
∈
B
(
X
)
with
σ
(
T
)
⊂
D
∪
{
1
}
and
{
T
n
}
⊂
co
(
T
)
such
that
T
n
x
0
=
O
(1)
and
(
TT
n
−
T
n
+
n
0
)
x
0
=
o
(1)
as
n
→
∞
,
for
some
x
0
∈
X
and
n
0
≥
1
.
Then
T
n
(
T
−
I
)
x
0
=
o
(1)
as
n
→
∞
.
Proof.
Since
the
sequence
{
T
n
x
0
}
is
bounded,
we
can
define
the
seminorm
Γ
on
B
(
X
)
by
Γ(
S
)
=
lim
sup
n
→∞
ST
n
x
0
,
S
∈
B
(
X
)
.
(11)
In
order
to
apply
[
2
,
Theorem
5]
to
the
operator
T
in
the
Banach
algebra
B
(
X
)
relative
to
Γ,
we
firstly
note
that
Γ
is
continuous
on
B
(
X
),
and
it
is
needed
to
verify
that
Γ(
T
m
)
=
O
(1)
as
m
→
∞
.
We
write
TT
n
=
(
TT
n
−
T
n
+
n
0
)
+
T
n
+
n
0
,
and
so
for
every
m
≥
1
we
have
Γ(
T
m
)
=
lim
sup
n
→∞
T
m
T
n
x
0
≤
T
m
−
1
lim
n
→∞
(
TT
n
−
T
n
+
n
0
)
x
0
+
lim
sup
n
→∞
T
m
−
1
T
n
+
n
0
x
0
=
lim
sup
n
→∞
T
m
−
1
T
n
x
0
=
Γ(
T
m
−
1
)
.
Hence
Γ(
T
m
)
≤
Γ(
T
m
−
1
)
≤
Γ(
T
)
≤
Γ(
I
)
for
m
≥
1.
Since
σ
(
T
)
⊂
D
∪
{
1
}
,
by
the
above
quoted
theorem
we
obtain
Γ(
T
m
+1
−
T
m
)
→
0
as
m
→
∞
.
On
the
other
hand,
we
observe
that
T
2
T
n
=
T
(
TT
n
−
T
n
+
n
0
)
+
(
TT
n
+
n
0
−
T
n
+2
n
0
)
+
T
n
+2
n
0
,
and
by
induction,
for
every
m
≥
1
we
get
the
relation
T
m
T
n
=
m
−
1
j
=0
T
m
−
j
−
1
(
TT
n
+
jn
0
−
T
n
+(
j
+1)
n
0
)
+
T
n
+
mn
0
.
(12)
Having
in
view
this
expression
of
T
m
T
n
,
we
get
the
inequality
(
T
m
+1
−
T
m
)
T
n
x
0
≥
T
n
+
mn
0
(
T
−
I
)
x
0
−
m
−
1
j
=0
T
m
−
j
−
1
(
T
−
I
)

(
TT
n
+
jn
0
−
T
n
+(
j
+1)
x
0
)
x
0
.
(13)
Finally,
from
(
11
),
(
12
)
and
(
13
)
we
infer
the
following
relations
for
m
≥
1,
lim
sup
n
→∞
T
n
(
T
−
I
)
x
0
=
lim
sup
n
→∞
T
n
+
mn
0
(
T
−
I
)
x
0
L.
Suciu
137
≤
Γ(
T
m
+1
−
T
m
)
+
m
−
1
j
=0
T
m
−
j
−
1
(
T
−
I
)
lim
n
→∞
(
TT
n
+
jn
0
−
T
n
+(
j
+1)
n
0
)
x
0
=
Γ(
T
m
+1
−
T
m
)
.
Clearly,
here
we
also
used
that
(
TT
n
−
T
n
+
n
0
)
x
0
→
0
as
n
→
∞
.
Since
Γ(
T
m
+1
−
T
m
)
→
0
as
m
→
∞
,
we
get
lim
n
→∞
T
n
(
T
−
I
)
x
0
=
0,
which
ends
the
proof.
From
this
theorem
we
derive
a
version
concerning
the
local
convergence
for
Ces`
aro
means,
of
the
well-known
result
of
Hille
[
11
],
this
later
being
a
strengthening
of
Gelfand’s
theorem
involving
the
doubly
power
bounded
operators
with
the
spectrum
reduced
to
{
1
}
.
For
an
invertible
operator
T
in
B
(
X
)
we
denote
M
(
p
)
−
n
(
T
)
=
M
(
p
)
n
(
T
−
1
)
for
n,
p
∈
N
.
Corollary
4.
Let
T
∈
B
(
X
)
with
0
/
∈
σ
(
T
)
⊂
D
∪
{
1
}
,
and
let
x
0
∈
X
such
that
M
(
p
)
n
(
T
)
x
0
=
O
(1)
as
|
n
|
→
∞
,
for
some
integer
p
≥
1
.
Then
M
(
p
−
1)
n
(
T
)
x
0
=
o
(
|
n
|
)
and
T
n
x
0
=
o
(
|
n
|
p
)
,
as
|
n
|
→
∞
.
Moreover,
if
σ
(
T
)
=
{
1
}
and
M
(
p
)
n
(
T
)
=
O
(1)
as
|
n
|
→
∞
then
T
=
I
.
Proof.
We
apply
Theorem
2
for
the
operators
T
n
=
M
(
p
)
n
(
T
)
and
S
n
=
M
(
p
)
n
(
T
−
1
)
(
n
∈
N
)
respectively,
because
by
the
relation
(
4
)
and
the
bound-
edness
of
{
T
n
}
and
{
S
n
}
,
we
have
(
TT
n
−
T
n
+1
)
x
0
→
0
and
(
SS
n
−
S
n
+1
)
x
0
→
0,
as
n
→
∞
.
So,
we
obtain
M
(
p
)
n
(
T
)(
T
−
I
)
x
0
→
0
as
|
n
|
→
∞
,
which
means
by
(
3
)
that
1
n
M
(
p
−
1)
n
(
T
)
x
0
→
0,
and
finally
1
n
p
T
n
x
0
→
0,
as
|
n
|
→
∞
.
This
gives
the
first
assertion.
Now,
if
the
sequence
{
M
(
p
)
n
(
T
)
}
n
∈
Z
is
bounded,
then
M
(
p
)
n
(
T
)
=
o
(
n
)
as
|
n
|
→
∞
,
and
by
(
3
)
it
follows
that
M
n
(
T
)
=
o
(
n
p
)
as
|
n
|
→
∞
.
By
Drissi-Zem´
anek
theorem
in
[
7
],
this
means
(
T
−
I
)
p
=
0.
But
the
bounded-
ness
of
{
M
(
p
)
n
(
T
)
}
ensures
asc(
T
−
I
)
≤
1,
that
is
N
(
T
−
I
)
=
N
((
T
−
I
)
p
)
=
X
.
Hence
T
=
I
,
which
ends
the
proof.
Let
us
recall
(see
[
9
])
that
a
positive
operator
T
on
a
Banach
lattice
for
which
the
Ces`
aro
means
of
an
(arbitrary)
order
p
≥
2
are
bounded
is,
in
fact,
Ces`
aro
bounded.
Thus,
in
this
case
Theorem
2
shows
that
1
n
T
n
→
0
strongly,
if
σ
(
T
)
⊂
D
∪
{
1
}
and
M
(
p
)
n
(
T
)
=
O
(1)
as
n
→
∞
,
for
an
integer
p
≥
1.
But,
in
this
context
we
have
the
following
more
general
result,
which
gives
also
another
version
of
Gelfand’s
theorem.
This
theorem
can
be
related
to
similar
results
of
de
Pagter-Bernau-Huijsmans
in
[
5
]
for
power
bounded
operators.
Local
ergodic
properties
for
powers
of
operators
138
Theorem
3.
Let
X
be
a
Banach
lattice
and
T
∈
B
(
X
)
be
a
positive
operator
with
σ
(
T
)
=
{
1
}
.
Let
{
T
n
}
⊂
co
(
T
)
and
x
0
∈
X
be
as
in
Theorem
2
.
Then
T
n
x
0
=
o
(1)
as
n
→
∞
.
Moreover,
if
T
n
=
O
(1)
and
TT
n
−
T
n
+
n
0
=
o
(1)
as
n
→
∞
,
for
some
n
0
≥
1
,
then
T
=
I
.
Proof.
Let
r
∈
(0
,
1)
fixed.
Then
by
a
result
of
Meyer-Nieberg
[
15
]
there
exists
an
integer
N
≥
0
such
that
T
N
=
(1
−
r
)
N
I
.
Let
{
T
n
}
and
x
0
∈
X
be
as
in
Theorem
2
.
From
the
proof
of
this
theorem
we
have
Γ(
T
Nm
)
≤
Γ(
I
)
for
m
≥
1,
and
since
σ
(
T
N
)
=
{
1
}
,
by
[
2
,
Theorem
5]
we
obtain
Γ(
T
Nm
(
T
N
−
I
))
→
0
as
m
→
∞
.
But
I
−
T
N
=
(1
−
δ
)
I
where
δ
=
(1
−
r
)
N
∈
(0
,
1),
therefore
Γ(
T
Nm
)
→
0
as
m
→
∞
.
Then
using
the
expression
of
T
Nm
T
n
given
by
(
12
)
as
well
as
the
convergence
condition
from
hypothesis,
we
get
for
m
≥
1
(as
in
the
proof
of
Theorem
2
)
lim
sup
n
→∞
T
n
x
0
=
lim
sup
n
→∞
T
n
+
Nmn
0
x
0
≤
Γ(
T
Nm
)
,
hence
lim
n
→∞
T
n
x
0
=
0.
This
gives
the
first
assertion
of
theorem.
Suppose
now
that
T
n
=
O
(1)
and
TT
n
−
T
n
+
n
0
=
o
(1),
as
n
→
∞
.
Consider
the
(continuous)
seminorm
Γ
0
on
B
(
X
)
given
by
Γ
0
(
S
)
=
lim
sup
n
→∞
ST
n
,
S
∈
B
(
X
)
.
Then
we
have
(as
above)
for
every
m
∈
N
,
Γ
0
(
T
Nm
)
=
lim
sup
n
→∞
T
Nm
T
n
≤
lim
sup
n
→∞
T
n
+
Nmn
0
+
Nm
−
1
j
=0
T
Nm
−
j
−
1
lim
n
→∞
TT
n
+
jn
0
−
T
n
+(
j
+1)
n
0
≤
sup
n
∈
N
T
n
<
∞
.
This
assures
as
above
(using
[
2
,
Theorem
5])
that
Γ
0
(
T
Nm
)
→
0
as
m
→
∞
.
Latterly,
using
the
relation
(
12
)
and
the
condition
of
convergence
from
our
assumption,
we
obtain
T
n
→
0,
n
→
∞
.
But
this
ensures
that
T
=
I
.
Indeed,
by
contrary,
we
firstly
obtain
by
Theorem
1
that
X
=
R
(
T
−
I
)
and
N
(
T
−
I
)
=
{
0
}
.
On
the
other
hand,
as
N
≥
1
(by
the
assumption
T
=
I
)
from
the
relation
I
−
T
N
=
(1
−
δ
)
I
follows
that
R
(
T
−
I
)
=
X
.
Thus,
T
−
I
is
invertible
in
B
(
X
)
that
is
1
/
∈
σ
(
T
),
which
contradicts
the
hypothesis.
We
conclude
T
=
I
,
and
this
ends
the
proof.
L.
Suciu
139
Corollary
5.
Let
T
be
a
positive
operator
on
a
Banach
lattice
X
with
σ
(
T
)
=
{
1
}
such
that
M
n
(
T
)
x
0
=
O
(1)
as
n
→
∞
for
some
x
0
∈
X
.
Then
M
n
(
T
)
x
0
=
o
(1)
as
n
→
∞
.
In
addition,
if
T
is
Ces`
aro
bounded
then
T
=
I
.
3
Ergodic
conditions
involving
characters
of
Banach
algebras
The
above
results
show
that
some
ergodic
properties
cannot
occur
for
the
operators
T
=
I
,
when
T
−
I
is
not
nilpotent
of
order
1.
So,
from
the
ergodic
point
of
view
it
is
important
to
know
when
T
−
I
is
nilpotent
of
an
order
k
≥
2,
or
more
general,
to
have
information
about
the
behavior
of
the
sequence
(
T
−
I
)
k
x
for
some
(any)
x
∈
X
.
Clearly,
a
nilpotent
operator
of
an
order
k
≥
2
cannot
be
p
-Ces`
aro
bounded
unless
it
is
identity,
but
such
operators
can
satisfy
certain
conditions
of
pointwise
boundedness
of
the
Ces`
aro
means,
even
of
their
powers.
Recall
that
some
results
in
this
context
are
given
by
Atzmon
[
3
,
4
]
in
connection
to
the
invariant
subspace
problem.
In
another
sense,
Esterle
and
Zarrabi
in
[
10
,
Theorem
1.1]
studied
the
behavior
of
f
((
T
−
I
)
k
x
)
for
x
∈
X
and
f
∈
X
∗
.
They
proved
that
if
|
f
(
T
n
x
)
|
=
o
(
n
k
+1
)
as
n
→
∞
then
either
lim
sup
n
→∞
n
|
f
((
T
−
I
)
n
x
)
|
1
/n
>
0,
or
f
((
T
−
I
)
n
x
)
=
0
for
n
≥
k
+
1.
Notice
that
some
theorems
for
Ces`
aro
means
give
conditions
which
assure
the
hypothesis
of
this
result.
Other
local
properties
concerning
the
sequences
in
the
convex
hull
of
the
powers
of
an
operator,
will
be
given
below.
Theorem
4.
Let
X
be
a
Banach
algebra
with
unit
element
e
,
and
let
T
∈
B
(
X
)
with
σ
(
T
)
⊂
D
∪
{
1
}
and
{
T
n
}
⊂
co
(
T
)
.
If
|
f
(
T
n
x
)
|
=
O
(1)
and
|
f
((
TT
n
−
T
n
+
n
0
)
x
)
|
=
o
(1)
as
n
→
∞
for
some
x
∈
X
,
f
a
character
of
X
and
n
0
≥
1
,
then
|
f
(
T
n
(
T
−
I
)
x
)
|
=
o
(1)
as
n
→
∞
.
Proof.
Based
on
the
boundedness
condition
from
hypothesis,
we
can
define
the
seminorm
γ
on
X
by
γ
(
y
)
=
lim
sup
n
→∞
|
f
(
yT
n
x
)
|
,
y
∈
X
.
We
prove
that
{
γ
(
T
m
e
)
}
is
bounded.
Indeed,
as
in
the
proof
of
Theorem
Local
ergodic
properties
for
powers
of
operators
140
2
we
have
for
m
≥
1,
γ
(
T
m
e
)
=
lim
sup
n
→∞
|
f
(
T
m
T
n
x
)
|
≤
sup
0
≤
k
≤
m
−
1
|
f
(
T
k
e
)
|
m
−
1
j
=0
lim
n
→∞
|
f
((
TT
n
+
jn
0
−
T
n
+(
j
+1)
n
0
)
x
)
|
+
lim
sup
n
→∞
|
f
(
T
n
+
mn
0
x
)
|
=
γ
(
e
)
.
Here
we
used
the
assumption
that
f
is
linear
and
multiplicative
on
X
,
and
the
assumption
that
f
((
TT
n
−
T
n
+
n
0
)
x
)
→
0
as
n
→
∞
.
Also,
as
f
is
continuous
on
X
it
follows
γ
(
y
)
≤
γ
(
e
)
y
,
for
y
∈
X
,
hence
γ
is
continuous
on
X
.
So,
by
[
2
,
Theorem
5]
we
infer
γ
((
T
m
+1
−
T
m
)
e
)
→
0
as
m
→
∞
.
This
leads
(using
a
similar
argument
as
in
the
proof
of
Theorem
2
)
to
the
conclusion
that
f
((
T
n
(
T
−
I
)
x
)
→
0
as
n
→
∞
,
which
ends
the
proof.
The
conclusion
of
this
theorem
is
also
true
for
f
∈
X
∗
(without
being
multiplicative),
under
the
stronger
condition
(
TT
n
−
T
n
+
n
0
)
x
→
0
as
n
→
∞
,
for
some
n
0
≥
1
and
x
∈
X
.
Corollary
6.
Let
X
be
a
Banach
algebra
with
unit
e
,
and
T
∈
B
(
X
)
with
σ
(
T
)
⊂
D
∪{
1
}
such
that
|
f
(
M
(
p
)
n
(
T
)
x
)
|
=
O
(1)
as
n
→
∞
,
for
some
x
∈
X
,
f
a
character
of
X
and
p
≥
1
.
Then
|
f
(
T
n
x
)
|
=
o
(
n
p
)
as
n
→
∞
.
Moreover,
one
of
the
following
statements
hold
:
(
i
)
f
(
T
n
x
)
=
0
for
any
n
≥
0
,
or
(
ii
)
lim
sup
n
→∞
n
|
f
((
T
−
I
)
n
x
)
|
1
/n
>
0
,
or
(
iii
)
f
((
T
−
I
)
n
x
)
=
0
for
n
≥
p
and
lim
sup
n
→∞
n
|
f
(
T
n
x
)
|
1
/n
>
0
.
Proof.
We
apply
Theorem
4
to
T
and
T
n
=
M
(
p
)
n
(
T
),
n
∈
N
.
So,
we
obtain
f
(
M
(
p
)
n
(
T
)(
T
−
I
)
x
)
→
0
as
n
→
∞
which
implies
(
f
being
a
character)
|
f
(
M
(
p
)
n
(
T
)(
T
−
I
)
p
x
)
|
=
|
f
((
T
−
I
)
p
−
1
e
)
f
(
M
(
p
)
n
(
T
)(
T
−
I
)
x
)
|
→
0
.
But,
by
(
3
)
this
yields
that
1
n
p
f
(
T
n
x
)
→
0.
Now
by
[
10
,
Theorem
1.1]
we
obtain:
either
lim
sup
n
→∞
n
|
f
((
T
−
I
)
n
x
)
|
1
/n
>
0
that
is
the
statement
(
ii
),
or
f
((
T
−
I
)
n
x
)
=
0
for
n
≥
p
.
L.
Suciu
141
In
the
second
case,
we
can
also
apply
the
same
theorem
for
the
operator
T
−
I
instead
of
T
.
We
conclude
that
either
lim
sup
n
→∞
n
|
f
(
T
n
x
)
|
1
/n
>
0,
or
f
(
T
n
x
)
=
0
for
n
≥
1.
The
first
alternative
just
means
the
statement
(
iii
),
while
the
last
equality
together
with
the
relation
f
((
T
−
I
)
p
x
)
=
0
of
above
give
f
(
x
)
=
0.
Hence
f
(
T
j
x
)
=
0,
for
any
j
≥
0
which
is
the
statement
(
i
).
This
ends
the
proof.
Notice
that
Theorem
4
is
a
version
of
Theorem
2
in
the
context
of
Banach
algebras.
Some
conditions
in
the
hypothesis
of
Theorem
4
can
be
modified
for
Banach
spaces,
in
order
to
obtain
a
weaker
conclusion,
as
follows.
Theorem
5.
Let
T
∈
B
(
X
)
with
σ
(
T
)
∩
T
⊂
{
1
}
.
Let
{
T
n
}
⊂
co
(
T
)
and
f
∈
X
∗
such
that
|
f
(
T
n
y
)
|
=
O
(1)
as
n
→
∞
for
any
y
∈
R
(
T
−
I
)
,
and
|
f
((
TT
n
−
T
n
+
n
0
)
x
)
|
=
o
(1)
as
n
→
∞
,
for
some
n
0
≥
1
and
any
x
∈
X
.
Then
either
|
f
(
T
n
(
T
−
I
)
x
)
|
=
o
(1)
as
n
→
∞
for
any
x
∈
X
,
or
σ
p
(
T
∗
)
=
{
1
}
and
|
f
(
T
n
(
T
−
I
)
2
x
)
|
=
o
(1)
as
n
→
∞
,
for
any
x
∈
X
.
Moreover,
the
first
conclusion
certainly
occurs
when
asc
(
T
∗
−
I
)
≤
1
.
Proof.
By
hypothesis
the
sequence
{
f
(
T
n
(
T
−
I
)
x
)
}
is
bounded,
for
x
∈
X
,
so
one
can
define
the
seminorm
γ
on
X
by
γ
(
x
)
=
lim
sup
n
→∞
|
f
(
T
n
(
T
−
I
)
x
)
|
,
x
∈
X
.
Since
f
((
TT
n
−
T
n
+
n
0
)
y
)
→
0
for
any
y
∈
X
,
we
obtain
γ
(
Tx
)
=
lim
sup
n
→∞
|
f
((
TT
n
−
T
n
+
n
0
)(
T
−
I
)
x
)
+
f
(
T
n
+
n
0
(
T
−
I
)
x
)
|
=
lim
sup
n
→∞
|
f
(
T
n
+
n
0
(
T
−
I
)
x
)
|
=
γ
(
x
)
.
Hence
γ
induces
an
isometry
V
on
a
Banach
space
(as
in
the
proof
of
The-
orem
1
).
Since
σ
(
V
)
⊂
σ
(
T
),
and
σ
(
T
)
∪
T
⊂
{
1
}
,
it
follows
σ
(
V
)
⊂
{
1
}
.
If
1
/
∈
σ
(
V
)
then
necessary
X
=
N
(
γ
),
so
f
(
T
n
(
T
−
I
)
x
)
→
0
for
any
x
∈
X
.
In
the
case
that
σ
(
V
)
=
{
1
}
one
has
V
=
I
.
This
gives
R
(
T
−
I
)
⊂
N
(
γ
)
that
is
f
(
T
n
(
T
−
I
)
2
x
)
→
0
for
any
x
∈
X
.
In
this
case
we
have
also
σ
p
(
T
∗
)
=
{
1
}
,
by
the
spectral
condition
from
hypothesis.
Now
denote
S
n
=
T
n
|
R
(
T
−
I
)
,
f
0
=
f
|
R
(
T
−
I
)
.
Then
the
previous
conclu-
sion
means
(
f
0
◦
S
n
)(
T
−
I
)
2
x
→
0
for
x
∈
X
.
Since
the
sequence
{
f
0
◦
S
n
}
is
bounded,
it
follows
that
(
f
0
◦
S
n
)
y
→
0
for
any
y
∈
R
((
T
−
I
)
2
)
=
R
(
T
−
I
),
this
equality
being
assured
under
the
assumption
asc(
T
∗
−
I
)
≤
1.
We
con-
clude
in
this
case
that
f
(
T
n
(
T
−
I
)
x
)
→
0
for
any
x
∈
X
,
which
ends
the
proof.
Local
ergodic
properties
for
powers
of
operators
142
Corollary
7.
Let
T
∈
B
(
X
)
with
σ
(
T
)
∩
T
⊂
{
1
}
and
f
∈
X
∗
satisfying
(for
some
p
≥
1
)
one
of
the
conditions
:
(
a
)
|
f
(
M
(
p
)
n
(
T
)
x
)
|
=
O
(1)
as
n
→
∞
for
x
∈
R
(
T
−
I
)
and
asc
(
T
∗
−
I
)
≤
1
,
or
(
b
)
|
f
(
M
(
p
)
n
(
T
)
x
)
|
=
O
(1)
as
n
→
∞
for
any
x
∈
X
,
and
σ
(
T
)
⊂
D
∪{
1
}
.
Then
|
f
(
T
n
x
)
|
=
o
(
n
p
)
as
n
→
∞
,
for
any
x
∈
X
.
Moreover,
for
every
x
∈
X
one
of
the
statements
(
i
)
,
(
ii
)
or
(
iii
)
of
Corollary
6
hold.
Proof.
If
condition
(
a
)
is
satisfied,
an
argument
similar
to
the
one
in
the
proof
of
Theorem
5
can
be
used,
with
the
seminorm
γ
given
by
γ
(
x
)
=
lim
sup
n
→∞
1
n
|
f
(
M
(
p
−
1)
n
(
T
)
x
)
|
,
x
∈
X
.
So,
by
(
4
)
it
is
immediate
that
γ
(
Tx
)
=
γ
(
x
)
for
x
∈
X
,
and
as
in
the
previous
proof
we
obtain
one
of
the
assertions:
(i)
either
1
n
f
(
M
(
p
−
1)
n
(
T
)
x
)
→
0
for
any
x
∈
X
,
or
(ii)
σ
p
(
T
∗
)
=
{
1
}
and
1
n
f
(
M
(
p
−
1)
n
(
T
)(
T
−
I
)
x
)
→
0
for
any
x
∈
X
.
The
last
convergence
in
(ii)
means
f
(
M
(
p
)
n
(
T
)(
T
−
I
)
2
x
)
→
0
for
x
∈
X
,
which
together
with
the
conditions
of
(
a
)
imply
that
f
(
M
(
p
)
n
(
T
)(
T
−
I
)
x
)
→
0
for
x
∈
X
.
In
both
cases
we
infer
by
(
3
)
that
1
n
p
f
(
T
n
x
)
→
0,
for
x
∈
X
.
Now
we
suppose
the
condition
(
b
)
satisfied.
This
means
|
(
T
n
f
)
x
|
=
O
(1)
as
n
→
∞
for
any
x
∈
X
,
that
is
T
n
f
=
O
(1)
as
n
→
∞
,
where
T
n
=
M
(
p
)
n
(
T
∗
).
Since
we
have
also
(
TT
n
−
T
n
+1
)
f
=
p
n
+
1
(
T
n
+1
−
I
)
f
→
0
,
by
Theorem
2
we
obtain
T
n
(
T
∗
−
I
)
f
→
0.
This
implies
f
(
M
(
p
)
n
(
T
)(
T
−
I
)
x
)
→
0
for
every
x
∈
X
,
which
leads
to
1
n
p
f
(
T
n
x
)
→
0
for
all
x
∈
X
.
Having
in
view
the
conclusion
obtained
in
both
cases
(
a
)
and
(
b
),
the
alternatives
(
i
),
(
ii
)
and
(
iii
)
mentioned
in
Corollary
7
follows
as
in
the
proof
of
Corollary
6
.
Acknowledgements.
We
are
grateful
to
the
reviewer
for
the
careful
eval-
uation
and
detailed
reading
of
our
work.
References
[1]
A.
Aleman
and
L.
Suciu,
On
ergodic
operator
means
in
Banach
spaces,
Integr.
Equations
Oper.
Theory
85
(2016),
259-287.
L.
Suciu
143
[2]
G.R.
Allan,
Power-bounded
elements
and
radical
Banach
algebras,
Ba-
nach
Center
Publ.
38,
Institute
of
Mathematics,
Polish
Academy
of
Sciences,
Warszawa,
1997,
9-16.
[3]
A.
Atzmon,
Operators
which
are
annihilated
by
analytic
functions
and
invariant
subspaces,
Acta
Math.
144
(1980),
27-63.
[4]
A.
Atzmon,
On
the
existence
of
hyperinvariant
subspaces,
J.
Oper.
Theory
11
(1984),
3-40.
[5]
S.J.
Bernau
and
C.B.
Huijsmans,
On
the
positivity
of
the
unit
element
in
a
normed
lattice
ordered
algebra,
Stud.
Math.
97
(1990),
143-149.
[6]
Y.
Derriennic,
On
the
mean
ergodic
theorem
for
Ces`
aro
bounded
oper-
ators,
Colloq.
Math.
84/85
(2000),
443-455.
[7]
D.
Drissi
and
J.
Zem´
anek,
Gelfand-Hille
theorems
for
Ces`
aro
means,
Quaestiones
Math.
23
(2000),
375-381.
[8]
N.
Dunford
and
J.T.
Schwartz,
Linear
Operators,
Part
I:
General
The-
ory
,
Interscience
Publishers,
New
York,
1958.
[9]
R.
Emilion,
Mean
bounded
operators
and
mean
ergodic
theorems,
J.
Funct.
Anal.
61
(1985),
1-14.
[10]
J.
Esterle
and
M.
Zarrabi,
Local
properties
of
powers
of
operators,
Arch.
Math.
(Basel)
65
(1995),
53-60.
[11]
E.
Hille,
Remarks
on
ergodic
theorems,
Trans.
Amer.
Math.
Soc.
57
(1945),
246-269.
[12]
U.
Krengel,
Ergodic
Theorems
,
Walter
de
Gruyter
Studies
in
Mathe-
matics
6,
Berlin–New
York,
1985.
[13]
M.
Lin,
On
the
uniform
ergodic
theorem,
Proc.
Amer.
Math.
Soc.
43
(1974),
337-340.
[14]
M.
Lin,
D.
Shoikhet
and
L.
Suciu,
Remarks
on
uniform
ergodic
theo-
rems,
Acta
Sci.
Math.
81
(2015),
251-283.
[15]
P.
Meyer-Nieberg,
Banach
Lattices
,
Springer,
Berlin–Heidelberg–New
York,
1991.
[16]
L.
Suciu
and
J.
Zem´
anek,
Growth
conditions
on
Ces`
aro
means
of
higher
order,
Acta
Sci.
Math.
79
(2013),
545-581.
Local
ergodic
properties
for
powers
of
operators
144
[17]
L.
Suciu,
Ergodic
behaviors
of
the
regular
operator
means,
Banach
J.
Math.
Anal.
11
(2017),
239-265.