Stochastic Processes and their Applications 119 (2009) 2970–2991

www.elsevier.com/locate/spa

Lévy driven moving averages and semimartingales

Andreas Basse ⇤ , Jan Pedersen
Department of Mathematical Sciences, University of Aarhus, Ny Munkegade, DK-8000 Aarhus C, Denmark

Received 19 June 2008; received in revised form 19 February 2009; accepted 23 March 2009
Available online 28 March 2009

Abstract

The aim of the present paper is to study the semimartingale property of continuous time moving averages
driven by Lévy processes. We provide necessary and sufficient conditions on the kernel for the moving
average to be a semimartingale in the natural filtration of the Lévy process, and when this is the case we
also provide a useful representation. Assuming that the driving Lévy process is of unbounded variation, we
show that the moving average is a semimartingale if and only if the kernel is absolutely continuous with a
density satisfying an integrability condition.
c 2009 Elsevier B.V. All rights reserved.

MSC: 60G48; 60H05; 60G51; 60G17

Keywords: Semimartingales; Moving averages; Lévy processes; Bounded variation; Absolutely continuity; Stable
processes; Fractional processes

1. Introduction
The present paper is concerned with the semimartingale property of moving averages (also
known as stochastic convolutions) which are driven by Lévy processes. More precisely, let
(X t )t 0 be a moving average of the form
Z t
Xt = (t s)dZ s , t 0, (1.1)
0
where (Z t )t 0 is a Lévy process and : R+ ! R is a deterministic function for which the integral
exists. We are interested in the question whether (X t )t 0 is an (FtZ )t 0 -semimartingale, where
(FtZ )t 0 denotes the natural filtration of (Z t )t 0 . In addition, two-sided moving averages (see
(1.6)) are studied as well.
⇤ Corresponding author. Tel.: +45 89423534; fax: +45 86131769.
E-mail addresses: basse@imf.au.dk (A. Basse), jan@imf.au.dk (J. Pedersen).

0304-4149/$ - see front matter c 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.spa.2009.03.007
 A. Basse, J. Pedersen / Stochastic Processes and their Applications 119 (2009) 2970–2991 2971

According to [1, page 533], a stationary process is a moving average if and only if its spectral
measure is absolutely continuous. Key examples of moving averages are the Ornstein–Uhlenbeck
process, the fractional Brownian motion, and their generalizations, the Ornstein–Uhlenbeck
type process (see [2]) and the linear fractional stable motion (see [3]). Moving averages occur
naturally in many different contexts, e.g. in stochastic Volterra equations (see [4]), in stochastic
delay equations (see [5]), and in turbulence (see [6]). Moreover, to capture the long-range
dependence of log-returns in financial markets it is natural to consider the fractional Brownian
motion instead of the Brownian motion in the Black–Scholes model (see [7, Part III]), and to
capture also heavy tails one is often led to more general moving averages.
It is often important that the process of interest is a semimartingale, and in particular the
following two properties are crucial: Firstly, if (X t )t 0 models an asset price which is locally
bounded and satisfies the No Free Lunch with Vanishing Risk condition then (X t )t 0 has
to be an (FtZ )t 0 -semimartingale (see [8, Theorem 7.2]). Secondly, it is possible to define a
Rt
“reasonable” stochastic integral 0 Hs dX s for all locally bounded (FtZ )t 0 -predictable processes
(Ht )t 0 if and only if (X t )t 0 is an (FtZ )t 0 -semimartingale due to the Bichteler–Dellacherie
Theorem (see [9, Theorem 7.6]). In view of the numerous applications of moving averages it is
thus natural to study the semimartingale property of these processes.
Let (Z t )t 0 denote a general semimartingale, : R+ ! R be absolutely continuous with a
bounded density and let (X t )t 0 be given by (1.1). Then by a stochastic Fubini result it follows
that (X t )t 0 is an (FtZ )t 0 -semimartingale, see e.g. [4, Theorem 3.3] or [5, Theorem 5.2]. In
the case where (Z t )t2R is a two-sided Wiener process, 2 L 2 (R+ , ) ( denotes the Lebesgue
measure) and (X t )t 0 is given by
Z t
Xt = (t s)dZ s , t 0, (1.2)
1

Knight [10, Theorem 6.5] shows that (X t )t 0 is an (FtZ ,1 )t 0 -semimartingale if and only if
is absolutely continuous with a square integrable density (FtZ ,1 := (Z s : 1 < s  t)).
Related results can be found in [11–13]. Moreover, results characterizing when (X t )t 0 is an
(FtX,1 )t 0 -semimartingale are given in [14,15].
The above presented results only provide sufficient conditions on or are only concerned
with the Brownian case. In the present paper we study the case where (Z t )t 0 is a Lévy process
and we provide necessary and sufficient conditions on for (X t )t 0 , given by (1.1), to be an
(FtZ )t 0 -semimartingale. Assume that (Z t )t 0 is of unbounded variation and has characteristic
triplet ( , 2 , ⌫). Our main result is the following:
(X t )t 0 is an (FtZ )t 0 -semimartingale if and only if is absolutely continuous on R+ with a
density 0 satisfying
Z tZ ⇣ ⌘
2
x 0 (s) ^ x 0 (s) ⌫(dx)ds < 1, 8t > 0, if 2 = 0, (1.3)
0 [ 1,1]
Z t
0 2 2
(s) ds < 1, 8t > 0, if > 0. (1.4)
0

In the case where (Z t )t 0 is a symmetric ↵-stable Lévy process, (1.3) corresponds to

0 2 L ↵ ([0, t], ) for all t > 0 when ↵ 2 (1, 2) and to 0 log+ ( 0 ) 2 L 1 ([0, t], ) for
all t > 0 when ↵ = 1.
2972 A. Basse, J. Pedersen / Stochastic Processes and their Applications 119 (2009) 2970–2991

Assume that (Z t )t 0 is of unbounded variation. If (X t )t 0 is an (FtZ )t 0 -semimartingale it

can be decomposed as
Z Z !
t u
0
X t = (0)Z t + (u s)dZ s du, t 0. (1.5)
0 0

As a corollary of (1.5) it follows that (X t )t 0 is càdlàg and of bounded variation if and only if
it is absolutely continuous, which is also equivalent to is absolutely continuous on R+ with a
density satisfying (1.3)–(1.4) and (0) = 0.
Finally we study two-sided moving averages, i.e. where (X t )t 0 is given by
Z t
Xt = ( (t s) ( s))dZ s , t 0, (1.6)
1
(Z t )t2R is a two-sided Lévy process and , : R ! R are deterministic functions for which the
integral exists. Note that in this case (X t )t 0 has stationary increments, and when = 0 it is
a stationary process. Several examples, including fractional Lévy processes and hence also the
linear fractional stable motion, are given in Section 5.
The conditions on from the one-sided case translate into necessary conditions in the
two-sided case. That is, if (Z t )t2R is of unbounded variation and (X t )t 0 is an (FtZ ,1 )t 0 -
semimartingale then is absolutely continuous on R+ with a density satisfying (1.3)–(1.4).
Moreover, [10, Theorem 6.5] is extended from the Gaussian case to the ↵-stable case with
↵ 2 (1, 2].
The paper is organized as follows. In Section 2 we collect some preliminary results. The
main results are presented in Section 3. All proofs are given in Section 4. The two-sided case is
considered in Section 5.

2. Preliminaries

Throughout the paper (⌦ , F, P) denotes a complete probability space. Let (Z t )t 0 denote a

Lévy process with characteristic triplet ( , 2 , ⌫), that is for t 0, E[ei✓ Z t ] = et(✓) for all
✓ 2 R, where
Z
2 2
(✓) = i ✓ ✓ /2 + ei✓ s 1 i✓ s1{|s|1} ⌫(ds), ✓ 2 R. (2.1)

For a general treatment of Lévy processes we refer to [16,17] or [18]. Let f : R ! R denote
a measurable function. Following [19, page 460] we say that f is Z -integrable R if there exists a
sequence of simple functions ( f n )n 1 such that f n ! f -a.s. and limn A f n (s)dZ s exists in
probability for all AR 2 B([0, t]) and all t > 0 (recall that Rdenotes the Lebesgue measure). In
t t
this case we define 0 f (s)dZ s as the limit in probability of 0 f n (s)dZ s . By [19, Theorem 2.7],
f is Z -integrable if and only if the following three conditions are satisfied for all t > 0:
Z t
f (s)2 2 ds < 1, (2.2)
0
Z tZ ⇣ ⌘
|x f (s)|2 ^ 1 ⌫(dx)ds < 1, (2.3)
0
Z t ✓ Z ◆
f (s) + x(1{|x f (s)|1} 1{|x|1} ) ⌫(dx) ds < 1. (2.4)
0
 A. Basse, J. Pedersen / Stochastic Processes and their Applications 119 (2009) 2970–2991 2973
Rt 2, ⌫ )
In this case 0f (s)dZ s is infinitely divisible with characteristic triplet ( f, f f given by
Z t Z !
f = f (s) + x(1{|x f (s)|1} 1{|x|1} ) ⌫(dx) ds, (2.5)
0
Z t
2
f = f (s)2 2
ds, (2.6)
0
⌫ f (A) = (⌫ ⇥ )((x, s) 2 R ⇥ [0, t] : x f (s) 2 A \ {0}), A 2 B(R). (2.7)
If
R t f is locally square integrable it is easily shown that (2.2)–(2.4) are satisfied and hence
0 f (s)dZ s is well-defined for all t 0. Note also that (2.4) is satisfied if (Z t )t 0 is symmetric.
Recall that (Z t )t 0 is a symmetric ↵-stable Lévy process with ↵ 2 (0, 2] if = 2 = 0 and ⌫
has density s 7! c |s| 1 ↵ for some c > 0 when ↵ 2 (0, 2), and ⌫ = 0 and = 0 when ↵ = 2.
In this case (2.2)–(2.4) reduce to f 2 L ↵ ([0, t], ) for all t > 0.
A function f : R+ ! R is said to be of bounded variation if on each finite interval [0, t] the
total variation of f is finite, that is
n
X
Vt ( f ) := sup | f (ti ) f (ti 1 )| < 1, (2.8)
i=1

where the sup is taken over all partitions 0 = t0 < · · · < tnR = t, n 1 of [0, t]. Note that a
Lévy process (Z t )t 0 is of bounded variation if and only if [ 1,1] |s| ⌫(ds) < 1 and 2 = 0
(see e.g. [16, Theorem 21.9]). Let I denote an interval and f : I ! R. Then f is said to be
absolutely continuous if there exists a locally integrable function h such that
Z t
f (t) f (u) = h(s)ds, u, t 2 I, u  t, (2.9)
u
and in this case h is called the density of f . If f : I ! R and g: R+ ! R+ are two measurable
functions, then f is said to have locally g-moment if
Z t
g(| f (s)|)ds < 1, u, t 2 I, u  t. (2.10)
u
If (2.10) is satisfied with g(x) = x ↵ for some ↵ > 0 then f is said to have locally ↵-moment.
An increasing family of -algebras (Ft )t 0 is called a filtration if it satisfies the usual
conditions of right-continuity and completeness. For each process (Yt )t 0 we let (FtY )t 0 denote
its natural filtration, i.e. (FtY )t 0 is the least filtration for which (Yt )t 0 is (FtY )t 0 -adapted. Let
(Ft )t 0 denote a filtration. We say that (X t )t 0 is an (Ft )t 0 -semimartingale if it admits the
following representation
X t = X 0 + Mt + A t , t 0, (2.11)
where (Mt )t 0 is a càdlàg local (Ft )t 0 -martingale starting at 0 and (At )t 0 is (Ft )t 0 -
adapted, càdlàg, of bounded variation and starting at 0, and X 0 is F0 -measurable. (Recall that
càdlàg means right-continuous with left-hand limits.)
We need the following standard notation: For functions f, g: R ! (0, 1) we write f (x) ⇡
g(x) as x ! 1 if f /g is bounded above and below on some interval (K , 1), where K > 0.
Furthermore we write f (x) = o(g(x)) as x ! 1 if f (x)/g(x) ! 0 as x ! 1. A similar
notation is used as x ! 0.
2974 A. Basse, J. Pedersen / Stochastic Processes and their Applications 119 (2009) 2970–2991

Assume that ⌫ has positive mass on [ 1, 1]. Similar to [20] we let ⇠ : [0, 1) ! [0, 1) be
given by
Z ⇣ ⌘
⇠(x) = |sx|2 ^ |sx| ⌫(ds), x 0. (2.12)
[ 1,1]

Note that ⇠ is 0 at 0, continuous and increasing and satisfies:

R
(i) ⇠(x)/x ! [ 1,1] |s| ⌫(ds) 2 (0, 1] as x ! 1,
R
(ii) If [ 1,1] |s|↵ ⌫(ds) < 1 for ↵ 2 (1, 2] then ⇠(x) = o(x ↵ ) as x ! 1.
To show (i)–(ii) let
Z Z
H (x) = x |s| ⌫(ds) and K (x) = x 2 s 2 ⌫(ds), (2.13)
x 1 |s|1 |s|<x 1

and note that ⇠(x) = H (x) + K (x) for x > 1. We have

Z Z
1
|s| ⌫(ds)  ⇠(x)x  |s| ⌫(ds), x > 1, (2.14)
x 1 |s|1 [ 1,1]

where the first inequality follows from H  ⇠ and the second from (2.12) since |xs|2 ^
|xs|
R  |xs|. Hence by (2.14) and monotone convergence (i) follows. To show (ii) assume
[ 1,1] |s| ⌫(ds) < 1 for some ↵ 2 (1, 2]. For all ✏ > 0 we have
↵

Z
lim sup H (x)x ↵  |s|↵ ⌫(ds), (2.15)
x!1 [ ✏,✏]

and
Z
↵
K (x)x  |s|↵ ⌫(ds), (2.16)
|s|<x 1

which shows ⇠(x)x ↵ ! 0 as x ! 1 and completes the proof of (ii).

Assume that ⌫ is absolutely continuous in a neighborhood of zero with a density f satisfying
f (x) ⇡ |x| ↵ 1 as x ! 0 for some ↵ 2 (0, 2) (this is satisfied in the ↵-stable case). An easy
calculation shows:
(1) ⇠(x) ⇡ x ↵ as x ! 1 if ↵ 2 (1, 2),
(2) ⇠(x)
R ⇡ x log(x) as x ! 1 if ↵ = 1,
(3) [ 1,1] |s| ⌫(ds) < 1 if ↵ 2 (0, 1).

3. Main results

Let (Z t )t 0 denote a nondeterministic Lévy process with characteristic triplet ( , 2 , ⌫) and

: R+ ! R be a measurable function which is Z -integrable (see (2.2)–(2.4)). Throughout this
section we let (X t )t 0 be the moving average
Z t
Xt = (t s)dZ s , t 0. (3.1)
0
Theorem 3.1 below is the main result of the paper. It provides a complete characterization of
when (X t )t 0 is an (FtZ )t 0 -semimartingale. Recall the definition of the function ⇠ in (2.12).
 A. Basse, J. Pedersen / Stochastic Processes and their Applications 119 (2009) 2970–2991 2975

Theorem 3.1. Assume that (Z t )t 0 is of unbounded variation. Then (X t )t 0 is an (FtZ )t 0 -

semimartingale if and only if is absolutely continuous on R+ with a density 0 which is locally
square integrable when 2 > 0 and has locally ⇠ -moment when 2 = 0 (that is, 0 satisfies
(1.3)–(1.4)).
Assume that (Z t )t 0 is of bounded variation. Then (X t )t 0 is an (FtZ )t 0 -semimartingale if
and only if it is of bounded variation which is also equivalent to is of bounded variation.
R
In particular, if 2 = 0, [ 1,1] |x|↵ ⌫(dx) < 1 for some ↵ 2 (1, 2] and is absolutely
continuous on R+ with a density having locally ↵-moment, then it follows by (ii) and the
above theorem that (X t )t 0 is an (FtZ )t 0 -semimartingale. In the case where (X t )t 0 is a
semimartingale the next proposition provides a useful representation of this process.

Proposition 3.2. Assume that (Z t )t 0 is of unbounded variation and (X t )t 0 is an (FtZ )t 0-

semimartingale. Then
Z Z !
t u
0
X t = (0)Z t + (u s)dZ s du, t 0, (3.2)
0 0

0
Ru 0 (u
where denotes the density of and ( 0 s)dZ s )u 0 is chosen measurable.
Hence we obtain the following corollary.

Corollary 3.3. Assume that (Z t )t 0 is of unbounded variation. Then the following four
statements are equivalent:
(a) (X t )t 0 is càdlàg and of bounded variation,
(b) (X t )t 0 is absolutely continuous,
(c) (X t )t 0 is an (FtZ )t 0 -semimartingale and (0) = 0,
(d) is absolutely continuous with a density satisfying (1.3)–(1.4) and (0) = 0.
In the symmetric ↵-stable case with ↵ 2 (1, 2) the equivalence between (b) and (d) follows
by [21, Theorem 6.1]. Braverman Rt and Samorodnitsky [22] study, among other things, processes
(Yt )t 0 on the form Yt = 0 f (t, s)dZ s , where (Z t )t 0 is a symmetric Lévy process and f
is a deterministic function. Their Theorem 5.1 provides necessary and sufficient conditions on
f (t, s) for (X t )t 0 to be absolutely continuous. In [23,24] necessary and sufficient conditions on
are obtained for (X t )t 0 to have locally bounded or continuous sample paths.
The next corollary follows by Theorem 3.1 and the estimates on ⇠ given in (1)–(3).

Corollary 3.4. Assume that 2 = 0 and ⌫ is absolutely continuous in a neighborhood of zero

with a density f satisfying f (x) ⇡ |x| ↵ 1 as x ! 0 for some ↵ 2 (0, 2) (this is satisfied in the
↵-stable case with ↵ 2 (0, 2)). Then (X t )t 0 is an (FtZ )t 0 -semimartingale if and only if
(i) is absolutely continuous with a density having locally ↵-moment when ↵ 2 (1, 2),
(ii) is absolutely continuous with a density having locally x log+ (x)-moment when ↵ = 1,
(iii) is of bounded variation when ↵ 2 (0, 1).
Here log+ denotes the positive part of log, i.e. log+ (x) = log(x) for x 1 and 0 otherwise.
In the following let (X t )t 0 be the Riemann–Liouville fractional integral given by
Z t
Xt = (t s)⌧ dZ s , t 0, (3.3)
0
2976 A. Basse, J. Pedersen / Stochastic Processes and their Applications 119 (2009) 2970–2991

where ⌧ is such that the integral exists. If (Z t )t 0 is a Wiener process and ⌧ > 1/2, (X t )t 0
is called a Lévy fractional Brownian motion (see [25, page 424]). Assume that (Z t )t 0 has no
Brownian component (i.e. 2 = 0). Using (2.2)–(2.4) it follows that for (X t )t 0 to be well-
defined one of the following (I)–(III) must be satisfied:
(I) ⌧ > 1/2, R
(II) ⌧ = 1/2 and [ 1,1] x 2 |log |x|| ⌫(dx) < 1,
R 1/⌧
(III) ⌧ < 1/2 and [ 1,1] |x| ⌫(dx) < 1.
Condition (I) is also sufficient for (X t )t 0 to be well-defined and when (Z t )t 0 is symmetric,
the conditions (I)–(III) are both necessary and sufficient for (X t )t 0 to be well-defined. When
⌧ = 0, (X t )t 0 = (Z t )t 0 ; thus let us assume ⌧ 6= 0. As a consequence of Theorem 3.1 we have
the following.

Corollary 3.5. Let (X t )t 0 be given by (3.3) and assume that (Z t )t 0 has no Brownian
component. Then (X t )t 0 is an (FtZ )t 0 -semimartingale if and only if one of the following
(1)–(3) is satisfied:
(1) ⌧ > 1/2, R
(2) ⌧ = 1/2 and [ 1,1] x 2 |log |x|| ⌫(dx) < 1,
R
(3) ⌧ 2 (0, 1/2) and [ 1,1] |x|1/(1 ⌧ ) ⌫(dx) < 1.
Note that 1/(1 ⌧ ) 2 (1, 2) when ⌧ 2 (0, 1/2). Let us in particular consider
Z t
1/↵
Xt = (t s) H dZ s , t 0, (3.4)
0
where (Z t )t 0 is a symmetric ↵-stable Lévy process with ↵ 2 (0, 2] and H > 0 (note
that (X t )t 0 is well-defined). To avoid trivialities assume H 6= 1/↵. As a consequence of
Corollary 3.5 (↵ 2 (0, 2)) and Theorem 3.1 (↵ = 2) it follows that (X t )t 0 is an (FtZ )t 0 -
semimartingale if and only if H > 1 when ↵ 2 [1, 2] or H > 1/↵ when ↵ 2 (0, 1).

4. Proofs

Throughout this section (X t )t 0 is given by (3.1). We extend to a function from R into R

by setting (s) = 0 for s 2 ( 1, 0). For any function f : R ! R, let t f denote the function
s 7! t ( f (1/t + s) f (s)) for all t > 0. We start by the following extension of [26, Theorem 24].

Lemma 4.1. Let I be either R+ or R, f : I ! R be locally integrable and g: R+ ! R+ be an

increasing convex function satisfying g(x)/x ! 1 as x ! 1 and let (rk )k 1 be a sequence
satisfying rk ! 1. Then f is absolutely continuous with a density having locally g-moment if
and only if (g( rk f ))k 1 is bounded in L 1 ([a, b], ) for all a, b 2 I with a < b. In this case
{g(| t f |) : t > ✏} is bounded in L 1 ([a, b], ) for all a, b 2 I with a < b and all ✏ > 0.
If (Z t )t 0 is of unbounded variation the above lemma can be applied with ⇠ playing the role
of g (⇠ is given by (2.12)), since in this case ⇠ satisfies all the conditions imposed on g except ⇠
is not convex. But h, defined by h(x) = x 2 1{x1} + (2x 1)1{x>1} for all x 0, is convex and
if we let
Z
g(x) = h(|xs|) ⌫(ds), x 0, (4.1)
[ 1,1]
 A. Basse, J. Pedersen / Stochastic Processes and their Applications 119 (2009) 2970–2991 2977

then g satisfies all the conditions in the lemma and g/2  ⇠  g. Thus, if f : I ! R is locally
integrable then f is absolutely continuous with a density having locally ⇠ -moment if and only if
(⇠( rk f ))k 1 is bounded in L 1 ([a, b], ) for all a, b 2 I with a < b.
Proof. Note that g is continuous and x 7! g(|x|) is a convex function from R into R, since g is
increasing and convex. Let a, b 2 I satisfying a < b be given and assume that (g( rk f ))k 1
is bounded in L 1 ([a, b], ). Since g(x)/x ! 1 as x ! 1, { rk f : k 1} is uniformly
integrable and hence weakly sequentially compact in L 1 ([a, b], ) (see e.g. [27, Chapter IV.8,
Corollary 11]). Choose a subsequence (n k )k 1 of (rk )k 1 and an h 2 L 1 ([a, b], ) such that
1
n k f ! h in the weak L ([a, b], )-topology. For all c, d 2 [a, b] with c < d we have
Z d Z d
nk f d ! hd , as k ! 1. (4.2)
c c
Moreover,
Z d Z Z !
d+1/n k d
nk f d = nk fd fd
c c+1/n k c
Z d+1/n k Z c+1/n k
= nk fd nk f d ! f (d) f (c), as k ! 1,(4.3)
d c
for ⇥ -a.a. c < d. Thus, we conclude that f is absolutely continuous with density h. Since
nk f ! h in the weak L 1 ([a, b], )-topology we may choose a sequence (n )n 1 of convex
combinations of ( n k f )k 1 such that n ! h in L 1 ([a, b], ), see [28, Theorem 3.13]. By
convexity and continuity of g we have
Z b Z b Z b
g(|h|)d  lim inf g(|n |)d  sup g( n k f )d < 1, (4.4)
a n!1 a k 1 a

which shows that h has g-moment on [a, b]. This completes the proof of the if -part.
Assume conversely that f is absolutely continuous with a density, h, having locally g-
moment. For all t > ✏, we have by Jensen’s inequality that
Z b Z s+1/t ! Z b Z 1/t !
g t h(u)du ds  t g(|h(u + s)|)du ds
a s a 0
Z 1/t Z b Z b+1/✏
=t g(|h(u + s)|)dsdu  g(|h(s)|)ds < 1, (4.5)
0 a a

which shows that {g(| t f |) : t > ✏} is bounded in L 1 ([a, b], ) and completes the proof. ⇤
In what follows, we are going to use two Lévy–Itô decompositions of (Z t )t 0 (see e.g. [16,
Theorem 19.2]).
(a) Decompose (Z t )t 0 as Z t = Z t1 + Z t2 , where (Z t1 )t 0 and (Z t2 )t 0 are two independent
Lévy processes with characteristic triplets (0, 2 , ⌫ ) respectively ( , 0, ⌫ ), where ⌫ =
1 2 1
⌫|[ 1,1] and ⌫2 = ⌫|[ 1,1]c . (Z t1 )t 0 and (Z t2 )t 0 are (FtZ )t 0 -adapted. Moreover, when is
locally bounded we let
Z t Z t
1 1 2
Xt = (t s)dZ s , and X t = (t s)dZ s2 , t 0. (4.6)
0 0
2978 A. Basse, J. Pedersen / Stochastic Processes and their Applications 119 (2009) 2970–2991

(b) Decompose (Z t )t 0 as Z t = Wt + Yt , where (Wt )t 0 is a Wiener process with variance

parameter 2 and (Yt )t 0 is a Lévy process with characteristic triplet ( , 0, ⌫). (Wt )t 0 and
(Yt )t 0 are independent and (FtZ )t 0 -adapted. Moreover, let
Z t Z t
X tW = (t s)dWs , and X tY = (t s)dYs , t 0. (4.7)
0 0

If 2 = 0 and (X t )t 0 is càdlàg it follows by [29, Theorem 4] and a symmetrization argument

that by modification on a set of Lebesgue measure 0, we may and do choose càdlàg.
The following lemma is closely related to [10, Theorem 6.5].

Lemma 4.2. We have the following:

(i) (X t )t 0 is an (FtZ )t 0 -semimartingale if is absolutely continuous on R+ with a locally
square integrable density.
(ii) Assume that (Z t )t 0 is a Wiener process. Then is absolutely continuous on R+ with a
locally square integrable density if (X t )t 0 is an (FtZ )t 0 -semimartingale.
Proof. (i): Decompose (Z t )t 0 and (X t )t 0 as in (a) above. Since both and (Z t2 )t 0 are
càdlàg and of bounded variation, (X t2 )t 0 is càdlàg and of bounded variation as well. Hence,
it is enough to show that (X t1 )t 0 is an (FtZ )t 0 -semimartingale. Since
Z t
X t1 = ( (t s) (0))dZ s1 + (0)Z t1 , t 0, (4.8)
0
we may and do assume (0) = 0. Then, is absolutely continuous on R with locally square
integrable density and hence for all T > 0, k t k L 2 ([ T,T ], )  K for some constant K > 0
2
and all t > 1/T by Lemma 4.1 with g(x) = x 2 . By letting c = E[ Z 11 ] we have (recall that
is zero on ( 1, 0))
E[(X t1 X u1 )2 ] = c k (t ·) (u ·)k2L 2 ([0,t], )
2 2
 cK (t u) , 8 0  u  t  T, (4.9)
which by the Kolmogorov–C̆entsov Theorem (see [30, Chapter 2, Theorem 2.8]) shows that
(X t1 )t 0 has a continuous modification (also to be denoted (X t1 )t 0 ). Moreover, for all 0 = t0 <
· · · < tn = T we have
" #
Xn Xn
p
E X t1i X t1i 1  X t1i X t1i 1 2  cK T, (4.10)
L (P)
i=1 i=1

which shows that (X t1 )t 0 is of integrable variation and hence an (FtZ )t 0 -semimartingale.

To show (ii) assume that (Z t )t 0 is a standard Wiener process and (X t )t 0 is an (FtZ )t 0 -
semimartingale. Since (X t )t 0 is a Gaussian process, Stricker [31, Proposition 4+5] entails that
(X t )t 0 is an (FtZ )t 0 -quasimartingale on each compact interval [0, N ]. For 0  u  t we have
h i Z u
E E[X t X u |FuZ ] = E (t s) (u s) dZ s
0
r Z u
2
= (t s) (u s) dZ s
⇡ 0 L 2 (P)
 A. Basse, J. Pedersen / Stochastic Processes and their Applications 119 (2009) 2970–2991 2979
r Z !1/2
2 u
2
= (t s) (u s) ds
⇡ 0
r Z !1/2
2 u
2
= (t u + s) (s) ds , (4.11)
⇡ 0

where the second equality follows by Gaussianity, which implies that

Z !1/2
nN
X h i Nn N /2
2
E E[X i/n X (i 1)/n |F(iZ 1)/n ] p (1/n + s) (s) ds .
i=1 ⇡2 0
(4.12)
Since (X t )t 0 is an (FtZ )t 0 -quasimartingale on [0, N ], the left-hand side of (4.12) is bounded
in n (see [32, Chapter VI, Definition 38]), showing that ( n )n 1 is bounded in L 2 ([0, N /2], ).
By Lemma 4.1 with g(x) = x 2 this shows that is absolutely continuous on R+ with a locally
square integrable density. ⇤
1
Lemma 4.3. If (X t )t 0 is an (FtZ )t 0 -semimartingale then (X t1 )t 0 is an (FtZ )t 0-
semimartingale.
Proof. Assume that (X t )t 0 is an (FtZ )t 0 -semimartingale, fix T > 0 and let

A := { Z t2 = 0 8t 2 [0, T ]}. (4.13)

Note that P(A) > 0 and (Z t1 )t 0 is P-independent of A. Let Q A denote the probability measure
given by Q A (B) := P(B \ A)/P(A).(X t )t 0 is an (FtZ )t 0 -semimartingale under Q A , since
Q A is absolutely continuous with respect to P. Moreover, since (Z t )t 0 and (Z t1 )t 0 are Q A -
1
indistinguishable it follows that (X t1 )t 0 is an (FtZ )t 0 -semimartingale under Q A and since A
1
is independent of (Z t )t 0 this is also true under P. ⇤
In the next lemma we study the jump structure of (X t )t 0.

Lemma 4.4. Assume that 2 = 0 and (X t )t 0 is càdlàg. Then ( X t 1{ Z t 6=0} )t 0 and

( (0) Z t )t 0 are indistinguishable.
Before proving the lemma we note the following:

Remark 4.5. (a) Let (X t )t 0 and (Yt )t 0 denote two independent càdlàg processes such that
P( X t = 0) = P( Yt = 0) = 1 for all t 0. Then as a consequence of Tonelli’s Theorem
we have P( X t Yt = 0, 8t 0) = 1. Rt
(b) If ⌫ is concentrated on [ 1, 1] then the mapping t 7! 0 (t s)dZ s is continuous from R+
into L 1 (P). This follows by approximating with continuous functions.
Rt
Proof of Lemma 4.4. Since X t = 0 ( (t s) 1)dZ s + Z t we may and do assume (0) 6= 0.
Recall that is chosen càdlàg; moreover, (0) = (0).
First we show the lemma in the case where ⌫ is a finite measure. Let ⌧n denote the time of the
nth jump of (Z t )t 0 ((⌧n+1 ⌧n )n 1 is thus an i.i.d. sequence of exponential distributions) and
let ( n )n 1 ✓ [0, 1) denote the jump times of . Note that the event
B := {9 ( j, k) 6= ( j 0 , k 0 ) : ⌧ j + k = ⌧ j0 + k 0 }, (4.14)
2980 A. Basse, J. Pedersen / Stochastic Processes and their Applications 119 (2009) 2970–2991

has probability zero. Since (Z t )t 0 only has finitely many jumps on each compact interval we
may regard (X t )t 0 as a pathwise Lebesgue–Stieltjes integral and hence it follows that
!
X
( X t )t 0 = Zt k ( k) . (4.15)
k 1 t 0
P
Let us show that on Bc
the series k 1 Z t k ( k ) has at most one term which differs from
zero for all t 0. Indeed, to see this assume that Z t k ( k ) and Z t k 0 ( k 0 ) both
differ from zero, where k 6= k 0 . Then there exist n, n 0 1 such that ⌧n = t k and ⌧n 0 = t k0
which implies ⌧n + k = ⌧n 0 + k 0 , and hence we have a contradiction. In particular, if Z t 6= 0
then Z t (0) 6= 0 and thus X t = Z t (0) = (0) Z t .
Now let (Z t )t 0 be a general Lévy process for which 2 = 0. For each n 1, decompose
(Z t )t 0 as Z t = Ytn + Utn , where (Ytn )t 0 and (Utn )t 0 are two independent Lévy processes with
characteristic triplets (0, 0, ⌫|[ 1/n,1/n] ) respectively (0, 0, ⌫|[ 1/n,1/n]c ). Moreover, set
Z t Z t
n n
X tY = (t s)dYsn and X tU = (t s)dUsn . (4.16)
0 0

Since (Utn )t 0 has piecewise constant sample paths the second integral is a pathwise
n n
Lebesgue–Stieltjes integral. Hence (X tU )t 0 is càdlàg and it follows that (X tY )t 0 is càdlàg as
well. Set
\ n
C := { X tY Utn = 0, 8t 0}, (4.17)
n 1
\ n
D := { X tU 1{ Utn 6=0} = (0) Utn , 8t 0}. (4.18)
n 1
n
From Remark 4.5(b) it follows that P( X tY = 0) = 1 for all t 0 which together with
Remark 4.5(a) shows that C has probability one. Moreover, from the first part of the proof it
follows that D has probability one. When Z t 6= 0, choose n 1 such that | Z t | > 1/n. Thus,
n n
Utn 6= 0, and hence X tY = 0 on C, which shows X t = X tU = (0) Utn = (0) Z t
on C \ D and completes the proof. ⇤

Lemma 4.6. Assume that 2 = = 0, ⌫ is concentrated on [ 1, 1] and (X t )t 0 is a special

(FtZ )t 0 -semimartingale. Then ( (0)Z t )t 0 is the martingale component of (X t )t 0 .

Proof. Let X t = Mt + At denote the canonical decomposition of (X t )t 0 . Since (Z t )t 0 is a

Lévy process, it is quasi-left-continuous (see [33, Chapter II, Corollary 4.18]) and thus there
exists a sequence of totally inaccessible stopping times (⌧n )n 1 which exhausts the jumps of
(Z t )t 0 . On the other hand, since (At )t 0 is predictable there exists a sequence of predictable
times ( n )n 1 which exhausts the jumps of (At )t 0 . From the martingale representation theorem
for Lévy processes (see [33, Chapter III, Theorem 4.34]) it follows that (Mt )t 0 is a purely
discontinuous martingale which jumps only when (Z t )t 0 does. Furthermore, since
P(9 n, k 1 : ⌧n = k < 1) = 0, (4.19)
Lemma 4.4 shows
(0) Z ⌧n = X ⌧n = M ⌧n + A ⌧n = M⌧n , P-a.s. on {⌧n < 1} 8 n 1. (4.20)
 A. Basse, J. Pedersen / Stochastic Processes and their Applications 119 (2009) 2970–2991 2981

Hence ( Mt )t 0 and ( (0) Z t )t 0 are indistinguishable which implies that (Mt )t 0 and
( (0)Z t )t 0 are indistinguishable since they both are purely discontinuous martingales (see [33,
Chapter I, Corollary 4.19]). This completes the proof. ⇤
The following lemma is concerned with the bounded variation case and it relies on an inequality
by Marcus and Rosiński [20].

Lemma 4.7. Assume that = 2 = 0, ⌫ is concentrated on [ 1, 1] and (Z t )t 0 is of unbounded

variation. Then (X t )t 0 is càdlàg and of bounded variation if and only if is absolutely
continuous on R+ with a density having locally ⇠ -moment and (0) = 0.
Recall the definition of t and of Vt ( f ) in (2.8).
Proof. Let N 1 be given. We start by showing the following (i) and (ii) under the assumptions
stated in the lemma:
(i) If (X t )t 0 is of bounded variation then E[V N (X )] < 1 for all N 1.
(ii) For all N 1,
8 ! !1/2 9
N < Z N /2 Z N /2 =
sup ⇠(| 2n (s)|)ds ^ ⇠(| 2n (s)|)ds
8 n 1: N /2 N /2 ;
⇢Z N
 E[VD
N (X )]  3N sup ⇠(| 2n (s)|)ds + 1 , (4.21)
n 1 N

where for each f : R+ ! R we let

2n N
X
VD
N ( f ) = sup f (i/2n ) f ((i 1)/2n ) . (4.22)
n 1 i=1

To show (i) assume that (X t )t 0 is of bounded variation. By Rosiński [29, Theorem 4], (· s)
is of bounded variation for -a.a. s 2 R+ ; in particular there exists an s 2 R+ such that (· s)
is of bounded variation. Hence is of bounded variation. Let T := [0, N ] \ Q, X : ⌦ ! RT
denote the canonical random element induced by (X t )t2T and let µ be given by

µ(A) = ( ⇥ ⌫) (s, x) 2 [0, t0 ] ⇥ R : x (· s) 2 A \ {0} , A 2 B(RT ). (4.23)

1
For all t1 , . . . , tn 2 T , (X t1 , . . . , X tn ) is infinitely divisible with Lévy measure µ pt1 ,...,t n
, where
T T
pt1 ,...,tn ( f ) = ( f (t1 ), . . . , f (tn )) for all f 2 R . For f 2 R let q( f ) denote the total variation
of f on T . Then q: RT ! [0, 1] is clearly a lower-semicontinuous pseudonorm on RT (see [34,
page 998]). Since ⌫ has compact support and is of bounded variation there exists an r0 > 0
such that µ( f 2 RT : q( f ) > r0 ) = 0 and hence by Lemma 2.2 in [34], E[e✏q(X ) ] < 1 for
some ✏ > 0. In particular (X t )t 0 is of integrable variation on [0, N ].
(ii) From [20, Corollary 1.1] we have
1/2 ⇥ ⇤ 1/2
1/4 min(ai,n , ai,n )  E 2n (X i/2n X (i 1)/2n )  3 max(ai,n , ai,n ), (4.24)
where
Z (i 1)/2n
ai,n := ⇠(| 2n (s)|)ds. (4.25)
1/2n
2982 A. Basse, J. Pedersen / Stochastic Processes and their Applications 119 (2009) 2970–2991

By monotone convergence we have

2n N
1 X ⇥ ⇤
E[VD
N (X )] = sup n E 2n (X i/2n X (i 1)/2n ) , (4.26)
n 1 2 i=1

and hence
N ⇥ ⇤
sup inf E 2n (X i/2n X (i 1)/2n )  E[VD
0,N (X )]
2 n 1 2n N /2<i2n N
⇥ ⇤
 N sup sup E 2n (X i/2n X (i 1)/2n ) , (4.27)
n 1 1i2n N

which by (4.24) shows (4.21).

Assume that (X t )t 0 is càdlàg and of bounded variation and hence by (i) of integrable
variation. From (ii) it follows that (⇠( 2n ))n 1 is bounded in L 1 ([ a, a], ) for all a > 0.
Conversely, if (⇠( 2n ))n 1 is bounded in L 1 ([ a, a], ) for all a > 0, (ii) shows that
E[VD D
N (X )] < 1; in particular V N (X ) < 1 P-a.s. Since in addition (X t )t 0 is right-continuous
in probability by Remark 4.5(b) it has a càdlàg modification (also to be denoted (X t )t 0 ), which
is of bounded variation since V N (X ) = VD N (X ) < 1 P-a.s.
Finally, the discussion just below Lemma 4.1 completes the proof, since (Z t )t 0 is of
unbounded variation. ⇤
We have the following consequence of the Bichteler–Dellacherie Theorem.

Lemma 4.8. Let (Yt )t 0 , (Ut )t 0 , (Ỹt )t 0 and (Ũt )t 0 denote four processes such that (Yt )t 0 is
D
(Ft )t 0 -adapted, (Ỹt )t 0 is (FtŨ )t 0 -adapted and (Y· , U· ) = (Ỹ· , Ũ· ). If (Yt )t 0 is an (FtU )t 0 -
U

semimartingale then (Ỹt )t 0 has a modification which is an (FtŨ )t 0 -semimartingale.

D
Proof. Since (Yt )t 0 , by assumption, is càdlàg and (Yt )t 0 = (Ỹt )t 0 we may choose a
càdlàg modification of (Ỹt )t 0 (also to be denoted (Ỹt )t 0 ). By the Bichteler–Dellacherie
Theorem (see [32, Theorem 80]) we must show that for all t > 0 the set of random variables
given by
( )
Xn
Ũ
H̃ti 1 (Ỹti Ỹti 1 ) : n 1, 0  t0 < · · · < tn  t, H̃ti 2 Fti , H̃ti  1 (4.28)
i=1

is bounded in L 0 (P). Since each H̃s 2 FsŨ satisfying H̃s  1 is given by

H̃s = lim Fn ((Ũu )us+1/n ) P-a.s., (4.29)

n!1

for some Fn : R[0,s+1/n] ! [ 1, 1] which is B(R)[0,s+1/n] -measurable, our assumptions imply

that for each random variable in the above set there exist Hti 2 FtUi satisfying Hti  1 for
i = 0, . . . , n 1 such that
n
X n
X
D
H̃ti 1 (Ỹti Ỹti 1 ) = Hti 1 (Yti Yti 1 ). (4.30)
i=1 i=1

Thus since (Yt )t 0 is an (FtU )t 0 -semimartingale, another application of the Bichteler–

Dellacherie Theorem shows that the set given in (4.28) is bounded in L 0 (P). ⇤
 A. Basse, J. Pedersen / Stochastic Processes and their Applications 119 (2009) 2970–2991 2983

We are now ready to prove Theorem 3.1.

Proof of Theorem 3.1. We prove the result in the following three steps (1)–(3). Recall the
Lévy–Itô decompositions (a) and (b).
(1) Let 2 > 0. Rt
Assume that (X t )t 0 is an (FtZ )t 0 -semimartingale. Let Z̃ t = Yt Wt and X̃ t = 0 (t
D
s) d Z̃ s . We have FtZ = FtW _FtY = Ft W _FtY = FtZ̃ and since (X · , Z · ) = ( X̃ · , Z̃ · ), Lemma 4.8
shows that ( X̃ t )t 0 is an (FtZ )t 0 -semimartingale. Therefore (X tW )t 0 := ((X t X̃ t )/2)t 0 is
an (FtZ )t 0 -semimartingale and thus an (FtW )t 0 -semimartingale, and by Lemma 4.2(ii) we
conclude that is absolutely continuous on R+ with a locally square integrable density.
On the other hand, if is absolutely continuous with a locally square integrable density it
follows by Lemma 4.2(i) that (X t )t 0 is an (FtZ )t 0 -semimartingale.
(2) Let 2 = 0 and (Z t )t 0 be of unbounded variation.
Assume that (X t )t 0 is an (FtZ )t 0 -semimartingale. By Lemma 4.3 it follows that (X t1 )t 0 is
1
an (FtZ )t 0 -semimartingale. Let T = Q \ [0, t], q( f ) = sups2T | f (s)| for all f 2 RT and µ
be given by (4.23) with ⌫ replaced by ⌫1 . Since ⌫1 has compact support and is locally bounded
(recall that is chosen càdlàg) there exists an r0 > 0 such that µ( f 2 RT : q( f ) r0 ) = 0 and
hence, according to Rosiński [34, Lemma 2.2], E[sups2[0,t] X s1 ] < 1. This shows that (X t1 )t 0
1 1
is a special (FtZ )t 0 -semimartingale. Let X t1 = Mt + At denote the canonical (FtZ )t 0 -
decomposition of (X t1 )t 0 . Then Lemma 4.6 yields (Mt )t 0 = ( (0)Z t1 )t 0 and hence (At )t 0 ,
given by
Z t
At = (t s)dZ s1 , t 0, (4.31)
0

where (t) = (t) (0) for t 0, is of bounded variation. Thus, by Lemma 4.7 we conclude
that , and hence also , is absolutely continuous on R+ with a density having locally ⇠ -moment.
Assume conversely that is absolutely continuous with a density having locally ⇠ -moment.
Since and (Z t2 )t 0 are càdlàg and of bounded variation it follows that (X t2 )t 0 is càdlàg and of
bounded variation as well. Let (At )t 0 be given by (4.31). By Lemma 4.7 it follows that (At )t 0
is càdlàg and of bounded and hence (X t1 )t 0 = ( (0)Z t1 + At )t 0 is an (FtZ )t 0 -semimartingale
and we have shown that (X t )t 0 is an (FtZ )t 0 -semimartingale.
(3) Let (Z t )t 0 be of bounded variation.
Assume that (X t )t 0 is an (FtZ )t 0 -semimartingale. By arguing as in (2) it follows that
(At )t 0 given by (4.31) is of bounded variation. Hence [29, Theorem 4] and a symmetrization
argument shows that , and hence also , is of bounded variation.
Assume conversely that is of bounded variation. Since (Z t )t 0 is càdlàg and of bounded
variation it follows that (X t )t 0 is càdlàg and of bounded variation and hence an (FtZ )t 0 -
semimartingale. ⇤

To show Proposition 3.2 we need the following Fubini type result.

Lemma 4.9. Let T > 0, µ denote a finite measure on R+ and let f : R2+ ! R be a measurable
function such that either (i) or (ii) is satisfied, where
(i) 2 = 0, ⇠(| f (t, ·)|) 2 L 1 ([0, T ], ) for all t 0 and ⇠(| f |) 2 L 1 (R+ ⇥ [0, T ], µ ⇥ ).
(ii) 2 > 0, f (t, ·) 2 L 2 ([0, T ], ) for all t 0, and f 2 L 2 (R+ ⇥ [0, T ], µ ⇥ ).
2984 A. Basse, J. Pedersen / Stochastic Processes and their Applications 119 (2009) 2970–2991
RT
Then ( 0 f (t, s)dZ s )t 0 can be chosen measurable and in this case
Z Z ! Z Z !
T T
f (t, s)dZ s µ(dt) = f (t, s) µ(dt) dZ s P-a.s. (4.32)
0 0

Proof. Assume that (i) is satisfied. To show (4.32) we may and do assume that (Z t )t 0 has
characteristic triplet (0, 0, ⌫) where ⌫ is concentrated on [ 1, 1]. Let g be given by (4.1). Since
g is 0 at 0, symmetric, increasing, convex, limx!1 g(x) = 1 and g(2x)  4g(x) for all x 0,
g is a Young function satisfying the 2 -condition (see [35, page 5+22]). Let L g ([0, T ], ) denote
the Orlicz space of measurable functions with finite g-moment on [0, T ] equipped with the norm
⇢ Z T
1
khkg = inf c > 0 : g(c h(s))ds  1 . (4.33)
0

According to Chapter 3.3, Theorem 10, and Chapter 3.5, Theorem 1, in [35], L g ([0, T ], ) is
a separable Banach space. Let f t := f (t, ·) for all t 0. Since ⇠(| f t |) 2 L 1 ([0, T ], ) for
RT
all t 0, it is easy to check that f t satisfies (2.2)–(2.4) and hence Yt := 0 f t (s)dZ s is well-
defined for all t 0. We show that (Yt )t 0 has a measurable modification. Since L g ([0, T ], ) is
separable and t 7! k f t hkg is measurable for all h 2 L g ([0, T ], ) it follows that t 7! f t is a
measurable mapping from R+ into L g ([0, T ], ). Furthermore, since L g ([0, T ], ) is separable
there exists (h nk )n,k 1 ✓ L g ([0, T ], ) and disjoint measurable sets (Ank )k 1 for all n 1 such
that with
X
f tn (s) = h nk (s)1 Ank (t), (4.34)
k 1
P RT
we have f t f tn g
 2 n for all t 0. Set Ytn = k 1 0 h nk (s)dZ s 1 Ank (t) for all t 0 and
n 1. Then (Ytn )t 0 is a measurable process and by [20, Theorem 2.1] it follows that

Ytn Yt L 1 (P)
 3 f tn ft g
3⇥2 n
, 8t 0, 8n 1. (4.35)

For all t 0 and ! 2 ⌦ let Ỹt (!) = limn Ytn (!) when the limit exists in R and zero otherwise.
Then (Ỹt )t 0 is measurable and for all t 2 R, Ỹt = Yt P-a.s. by (4.35). Thus we have constructed
a measurable modification of (Yt )t 0 .
Let us show that both sides of (4.32) are well-defined. Since g/2  ⇠  g and ⇠(ax) 
(a + 1)2 ⇠(x) for all x, a > 0, it follows by Jensen’s inequality that
Z T Z ! Z Z
2(µ(R) + 1)2 T
⇠ | f (t, s)| µ(dt) ds  ⇠(| f (t, s)|) µ(dt)ds < 1. (4.36)
0 µ(R) 0

Thus, the right-hand side of (4.32) is well-defined. The left-hand side is well-defined as well
since
Z Z T Z Z T ! Z T !1/2
E f (t, s)dZ s µ(dt)  3 ⇠(| f t (s)|)ds _ ⇠(| f t (s)|)ds µ(dt)
0 0 0
< 1, (4.37)
 A. Basse, J. Pedersen / Stochastic Processes and their Applications 119 (2009) 2970–2991 2985

where the first inequality follows by [20, Corollary 1.1]. Furthermore, (4.32) is obviously true
for simple f on the form
n
X
f (t, s) = ↵i 1(si 1 ,si ] (t)1(ti 1 ,ti ] (s). (4.38)
i=1
If f is a given function satisfying (i) we can choose a sequence of simple ( f n )n 1 converging to
f and satisfying | f n |  | f |. We have
Z Z T ! Z T Z !
f n (u, s)dZ u µ(ds) = f n (u, s) µ(ds) dZ u , (4.39)
0 0

and by estimates as above it follows that we can go to the limit in L 1 (P) in (4.39), which shows
(4.32).
The case (ii) follows by a similar argument. In this case we have to work in L 2 ([0, T ], )
instead of L g ([0, T ], ). ⇤
Proposition 3.2 is an immediate consequence of Theorem 3.1 and Lemma 4.9, since
Z t s Z t
0
(t s) = (0) + (u)du = (0) + 1{su} 0 (u s)du, s 2 [0, t]. (4.40)
0 0

5. The two-sided case

Let (X t )t 0 be given by
Z t
Xt = ( (t s) ( s)) dZ s , t 0, (5.1)
1
where (Z t )t2R is a (two-sided) nondeterministic Lévy process with characteristic triplet
( , 2 , ⌫) and , : R ! R are measurable functions for which the integral exists (still in
the sense of [19, page 460]). Also assume that and are 0 on ( 1, 0) and let (FtZ ,1 )t 0
denote the least filtration for which (Z s : 1 < s  t) ✓ FtZ ,1 for all t 0. From [19,
Theorem 2.8] it follows that (X t )t 0 is well-defined if and only if
Z t Z 0
X t1 = (t s)dZ s , and X t2 = ( (t s) ( s))dZ s , (5.2)
0 1
are well-defined. Similar to Lemma 4.8 we have the following.

Lemma 5.1. Let (Yt )t 0 , (Ut )t2R , (Ỹt )t 0 and (Ũt )t2R denote four processes such that (Yt )t 0
U,1 D
is (Ft )t 0 -adapted, (Ỹt )t 0 is (FtŨ ,1 )t 0 -adapted and (Y· , U· ) = (Ỹ· , Ũ· ). If (Yt )t 0
is an (FtU,1 )t 0 -semimartingale then (Ỹt )t 0 has a modification which is an (FtŨ ,1 )t 0 -
semimartingale.

Lemma 5.2. Assume that (Z t )t2R is symmetric. Then (X t )t 0 is an (FtZ ,1 )t 0 -semimartingale

if and only if (X t1 )t 0 is an (FtZ )t 0 -semimartingale and (X t2 )t 0 is càdlàg and of bounded
variation.
Proof. The if -part is trivial. To show the only if -part assume that (X t )t 0 is an (FtZ ,1 )t 0 -
semimartingale. Let X̃ t = X t1 X t2 and let Z̃ t = Z t for t 0 and Z̃ t = Z t when t < 0.
2986 A. Basse, J. Pedersen / Stochastic Processes and their Applications 119 (2009) 2970–2991

D
Since (Z t )t2R is symmetric (X · , Z · ) = ( X̃ · , Z̃ · ) and from Lemma 5.1 it follows that ( X̃ t )t 0
is an (FtZ̃ ,1 )t 0 -semimartingale and hence an (FtZ ,1 )t 0 -semimartingale since (FtZ̃ ,1 )t 0 =
(FtZ ,1 )t 0 . Thus, (X t1 )t 0 = ((X t + X̃ t )/2)t 0 is an (FtZ ,1 )t 0 -semimartingale and hence
an (FtZ )t 0 -semimartingale. Moreover, (X t2 )t 0 is an (FtZ ,1 )t 0 -semimartingale and hence
càdlàg and of bounded variation since X t2 is F0Z ,1 -measurable for all t 0. ⇤
We have the following consequence of Lemma 5.2 and Theorem 3.1.

Proposition 5.3. Let (X t )t 0 be given by (5.1) and assume that it is an (FtZ ,1 )t 0 -

semimartingale.
If (Z t )t2R is of unbounded variation then is absolutely continuous on R+ with a density 0
satisfying (1.3)–(1.4).
If (Z t )t2R is of bounded variation then (X t )t 0 is of bounded variation and is of bounded
variation as well.
Proof. Let Z̃ t = Z t Z t0 where (Z t0 )t2R is an independent copy of (Z t )t2R and let (X t0 )t 0 be
given by
Z t
0
Xt = ( (t s) ( s)) dZ s0 , t 0. (5.3)
1
0
By Lemma 5.1, (X t0 )t 0 is an (FtZ ,1 )t 0 -semimartingale, which by independence of filtrations
0
shows that ( X̃ t )t 0 := (X t X t0 )t 0 is a semimartingale in the (FtZ ,1 _ FtZ ,1 )t 0 -filtration and
hence in the (FtZ̃ ,1 )t 0 -filtration. Since ( Z̃ t )t2R is symmetric Lemma 5.2 shows that ( X̃ t1 )t 0
is an (FtZ̃ )t 0 -semimartingale and since ( Z̃ t )t 0 has characteristic triplet (0, 2 2 , ⌫˜ ) where
⌫˜ (A) = ⌫(A) + ⌫( A), the proposition follows by Theorem 3.1. ⇤
Let (X t )t 0 denote a fractional Lévy motion, that is
Z t
Xt = ((t s)⌧ ( s)⌧+ )dZ s , t 0, (5.4)
1
where ⌧ is such that the integral exists and x+ := x _ 0 for all x 2 R. In the following let us
assume that (Z t )t2R has no Brownian component. Recall the definition of X t2 in (5.2). From [19,
Theorem 2.8] it follows that it is necessary (and sufficient when (Z t )t 0 is symmetric) that
Z Z !
1
2
x((t + s)⌧ s ⌧ ) ^ 1 ⌫(dx)ds < 1 (5.5)
0

for X t2 to be well-defined. A simple calculation shows that (5.5) is satisfied if and only if
Z
⌧ < 1/2 and |x|1/(1 ⌧ ) ⌫(dx) < 1. (5.6)
[ 1,1]c

Thus it is necessary that (5.6) and (I)–(III) are satisfied for (X t )t 0 to be well-defined, and when
(Z t )t2R is symmetric these conditions are also sufficient.
R Marquardt [36] studies processes
R of the
form (5.4) under the assumptions that 2 = 0, [ 1,1]c |x|2 ⌫(dx) < 1, = [ 1,1]c x ⌫(dx)
and 0 < ⌧ < 1/2. See also [37] for a study of the well-balanced case.
To avoid trivialities assume ⌧ 6= 0. As an application of Proposition 5.3 and Corollary 3.5 we
have the following.
 A. Basse, J. Pedersen / Stochastic Processes and their Applications 119 (2009) 2970–2991 2987

Corollary 5.4. Assume that (Z t )t2R has no Brownian component and let (X t )t 0 be
given
R by (5.4). If (X t )t 0 is an (FtZ ,1 )t 0 -semimartingale then ⌧ 2 (0, 1/2) and
1/(1 ⌧ )
[ 1,1] |x| ⌫(dx) < 1.
In particular let (X t )t 0 denote a linear fractional stable motion with indexes ↵ 2 (0, 2] and
H 2 (0, 1), that is
Z t ⇣ ⌘
H 1/↵
Xt = (t s) H 1/↵ ( s)+ dZ s , t 0, (5.7)
1

where (Z t )t2R is a symmetric ↵-stable Lévy process (see [3, Definition 7.4.1]). For ↵ = 2,
(X t )t 0 is a fractional Brownian motion (fBm) with Hurst parameter H (up to a scaling constant).
From Corollary 5.4 it follows that (X t )t 0 is an (FtZ ,1 )t 0 -semimartingale if and only if
H = 1/↵.
Let (X t )t 0 be given by (5.1) and assume that (Z t )t2R is a symmetric ↵-stable Lévy process
with ↵ 2 (1, 2]. If (X t )t 0 is an (FtZ ,1 )t 0 -semimartingale it follows by Proposition 5.3 and
result (1) at the end of Section 2 that is absolutely continuous on R+ with a density having
locally ↵-moment. The next result shows that this condition is actually necessary and sufficient
for (X t )t 0 to be an (FtZ ,1 )t 0 -semimartingale if we delete “locally”. Thus, extending [10,
Theorem 6.5] from ↵ = 2 to ↵ 2 (1, 2] we have the following.

Proposition 5.5. Let (X t )t 0 be given by (5.1) and assume that (Z t )t2R is a symmetric ↵-stable
Lévy process with ↵ 2 (1, 2]. Then (X t )t 0 is an (FtZ ,1 )t 0 -semimartingale if and only if is
absolutely continuous on R+ with a density in L ↵ (R+ , ).
Let B denote a Banach space (not necessarily separable) and assume that there exists a countable
subset D of the unit ball of B 0 (the topological dual space of B) such that
kxk = sup |F(x)| , 8x 2 B. (5.8)
F2D
Pn
Following [38, page 133], a B-valued random element X is called ↵-stable if i=1 ai Fi (X ) is a
real-valued ↵-stable random variable for all n 1, F1 , . . . , Fn 2 D and a1 , . . . , an 2 R.
Let T denote an interval in R+ and let B denote the subspace of RT containing all functions
which are càdlàg and of bounded variation. Then B is a Banach space in the total variation norm
(but not separable) and since the unit ball of B 0 consists of F of the form
n
X
F( f ) = ai ( f (ti ) f (ti 1 )), f 2 B, (5.9)
i=1
n
where (ai )i=1 n
✓ [ 1, 1] and (ti )i=0 is an increasing sequence in T , it follows that B satisfies
(5.8).
Proof of Proposition 5.5. For ↵ = 2 the result follows by [11, Theorem 3.1]; thus let us assume
↵ 2 (1, 2).
Assume that (X t )t 0 is an (FtZ ,1 )t 0 -semimartingale. According to Lemma 5.2 (X t2 )t 0 is
càdlàg and of bounded variation. Consider (X t2 )t 0 as an ↵-stable random element with values
in the Banach space consisting of functions which are càdlàg and of bounded variation equipped
with the total variation norm. Hence from [38, Proposition 5.6] it follows that (X t2 )t 0 is of
2988 A. Basse, J. Pedersen / Stochastic Processes and their Applications 119 (2009) 2970–2991

integrable variation on each compact interval. Moreover, by [20, Corollary 1.1] we have
h i 1 p
2
E n(X i/n X (i2 1)/n ) ai,n ^ ai,n , i, n 1, (5.10)
4
where
Z 1 Z
ai,n := ⇠ (| n (s)|)ds, and ⇠ (x) := (|xs|2 ^ |xs|) ⌫(ds).
˜ ˜ (5.11)
(i 1)/n

Since i 7! ai,n is decreasing it follows that

n
X h i 1 p
E[V1 (X 2 )] sup E 2
X i/n X (i2 1)/n sup an,n ^ an,n . (5.12)
n 1 i=1 n 1 4

By (5.12) we conclude that (an,n )n 1 is bounded and hence (⇠˜ (| n |))n 1 is bounded in
L 1 ([1, 1), ). A straightforward calculation shows ⇠˜ (x) = c1 x ↵ for all x 0 for some constant
c1 > 0, which implies that ( n )n 1 is bounded in L ↵ ([1, 1), ). Since ↵ > 1, a sequence in
L ↵ ([1, 1), ) is bounded if and only if it is weakly sequentially compact (see [27, Chapter IV.8,
Corollary 4]). Thus, by arguing as in Lemma 4.1 it follows that is absolutely continuous with
a density in L ↵ ([1, 1), ). Furthermore, since (X t1 )t 0 is an (FtZ )t 0 -semimartingale it follows
by Corollary 3.4 that is absolutely continuous on R+ with a density locally in L ↵ (R+ , ). This
shows the only if -part.
Assume conversely that is absolutely continuous on R+ with a density in L ↵ (R+ , ). By
Corollary 3.4 (X t1 )t 0 is an (FtZ )t 0 -semimartingale. Thus it is enough to show that (X t2 )t 0
is càdlàg and of bounded variation. Since is absolutely continuous on R+ with a density in
L ↵ (R+ , ) it follows by arguing as in Lemma 4.1 that k (t ·) (u ·)k L ↵ (( 1,0), ) 
c(t u) for some c > 0 and all 0  u  t. For all p 2 [1, ↵) and all u, t 0 we have

X t2 X u2 = K p,↵ k (t ·) (u ·)k L ↵ (( 1,0), )  K p,↵ c |t u| , (5.13)

L p (P)
for some constant K p,↵ > 0 only depending on p and ↵. By letting p 2 (1, ↵), (5.13) and
the Kolmogorov–C̆entsov Theorem show that (X t2 )t 0 has a continuous modification. Moreover,
by letting p = 1 (5.13) shows that this modification is of integrable variation on each compact
interval. This completes the proof. ⇤
Motivated by Lemma 5.2 we study in the following proposition infinitely
R divisible processes
(X t )t 0 of bounded variation, where (X t )t 0 is on the form X t = R f (t, s)dZ s . Assume that
(X t )t 0 is càdlàg and of bounded variation. Rosiński [29, Theorem 4] shows that t 7! f (t, s) is
of bounded variation for -a.a. s 2 R. Extending this we show that the total variation
R of f (·, s)
must satisfy an integrability condition which is equivalent to the existence of R Vt ( f (·, s))dZ s
for all t > 0 when (Z t )t2R is symmetric and has no Brownian component.

Proposition
R 5.6. Let f : R+ ⇥ R ! R denote a measurable function such that X t =
R f (t, s)dZ s is well-defined for all t 0. If (X t )t 0 is càdlàg and of bounded variation then
ZZ ⇣ ⌘
1 ^ |x Vt ( f (·, s))|2 ⌫(dx)ds < 1, 8t > 0. (5.14)

Let (✏i )i 1 denote a Rademacher sequence, i.e. (✏i )i 1 is an i.i.d. sequence P

such that
1
P(✏1 = 1) = P(✏1 = 1) = 1/2. It is well-known that if (↵i )i 1 ✓ R then i=1 ✏i ↵i
 A. Basse, J. Pedersen / Stochastic Processes and their Applications 119 (2009) 2970–2991 2989
P1 2
converges P-a.s. if and only if i=1 ↵i < 1. Let B denote a Banach space satisfying (5.8).
Following [38, page 99], a B-valued random element XPis called a vector-valued Rademacher
1
series if there exists a sequence P1(xi )i 1 in B suchPthat i=1 F 2 (xi ) < 1 for all F 2 D and
1
(F1 (X ), . . . , Fn (X )) equals ( i=1 ✏i F1 (xi ), . . . , i=1 ✏i Fn (xi )) in distribution for all n 1
and all F1 , . . . , Fn 2 D.

Proof of Proposition 5.6. By a symmetrization argument we may and do assume that 2 =0

and (Z t )t2R is symmetric. Define
1
X
Yt = ✏ j C j f (t, U j ), t 0, (5.15)
j=1

where (✏ j ) j 1 is a Rademacher sequence, (⌧ j ) j 1 are the partial sums of i.i.d. standard

exponential random variables and (U j ) j 1 are i.i.d. standard normal random variables with
density ⇢, and (✏ j ) j 1 , (⌧ j ) j 1 and (U j ) j 1 are independent. Let ⌫ : R+ ! R+ denote
the right-continuous inverse of the mapping x 7! ⌫((x, 1)), that is, ⌫ (s) = inf{x > 0 :
⌫((x, 1))  s}, and let C j := ⌫ (⌧ j ⇢(U j )) for all j 1. By [29, Proposition 2], the series
(5.15) converges P-a.s. and (Yt )t 0 has the same finite dimensional distributions as (X t )t 0 .
Thus, (Yt )t 0 has a càdlàg modification of locally bounded variation. Hence we may and do
assume that (X t )t 0 is given by (5.15). Moreover, we may define (✏ j ) j 1 on a probability
space (⌦ 0 , F 0 , P 0 ), (⌧ j ) j 1 and (U j ) j 1 on a probability space (⌦ 00 , F 00 , P 00 ) and (X t )t 0 on
the product space. Let T = [0, t] denote a compact interval in R+ and let B denote the subspace
of RT consisting of functions which are càdlàg and of bounded variation. Inspired by Marcus
and Rosiński [23] let us fix !00 2 ⌦ 00 and consider X = (X t )t2T as a B-valued Rademacher
2
series under P 0 . From [38, Theorem 4.8] it follows that E 0 [e↵ k X k ] < 1 for all ↵ > 0, which
in particular shows that (X t )t2T is of P 0 -integrable variation. By Khinchine’s inequality there
exists a constant c > 0 such that E 0 [|X t X u |] c kX t X u k L 2 (P 0 ) for all u, t 0. Together
with the triangle inequality in l 2 this shows that
" # !1/2
n
X n
X 1
X
E 0
X ti X ti 1 c C 2j ( f (ti , U j ) f (ti 1 , U j ))
2

i=1 i=1 j=1

!2 !1/2
n
1 X
X
c C j ( f (ti , U j ) f (ti 1 , U j ))
j=1 i=1
!2 !1/2
1
X n
X
=c Cj f (ti , U j ) f (ti 1, U j ) . (5.16)
j=1 i=1

Thus, by monotone convergence we conclude

!1/2
1
X
0 2
E [Vt (X )] c C j Vt ( f (·, U j )) , (5.17)
j=1

and 2
P1 in particular (C j Vt ( f (·, U j ))) j 1 2 l . Thus, we have shown that the series
j=1 ✏ j C j Vt ( f (·, U j )) converges P-a.s. and from Theorem 2.4 and Proposition 2.7 in [39]
2990 A. Basse, J. Pedersen / Stochastic Processes and their Applications 119 (2009) 2970–2991

it follows that
Z 1Z
1 ^ H (u, v)2 ⇢(v)dvdu < 1, (5.18)
0
where H (u, v) = ⌫ (u⇢(v)) Vt ( f (·, v)). Furthermore, (5.18) equals
ZZ
1
1 ^ (⌫ (u) Vt ( f (·, v)))2 du ⇢(v)dv
⇢(v)
ZZ ⇣ ⌘
= 1 ^ (u Vt ( f (·, v)))2 ⌫(du)dv, (5.19)

which shows (5.14). ⇤

References

[1] J.L. Doob, Stochastic Processes, Wiley Classics Library, John Wiley & Sons Inc., New York, 1990, Reprint of the
1953 original, A Wiley-Interscience Publication.
[2] A. Rocha-Arteaga, K.-i. Sato, Topics in Infinitely Divisible Distributions and Lévy Processes, in: Aportaciones
Matemáticas: Investigación [Mathematical Contributions: Research], vol. 17, Sociedad Matemática Mexicana,
México, 2003.
[3] G. Samorodnitsky, M.S. Taqqu, Stable non-Gaussian random processes, in: Stochastic Modeling, Chapman & Hall,
New York, 1994, Stochastic models with infinite variance.
[4] P. Protter, Volterra equations driven by semimartingales, Ann. Probab. 13 (2) (1985) 519–530.
[5] M. Reiß, M. Riedle, O. van Gaans, On Émery’s inequality and a variation-of-constants formula, Stochastic Anal.
Appl. 25 (2) (2007) 353–379.
[6] O.E. Barndorff-Nielsen, J. Schmiegel, Ambit processes: With applications to turbulence and tumour growth,
in: Stochastic Analysis and Applications, in: Abel Symp., vol. 2, Springer, Berlin, 2007, pp. 93–124.
[7] F. Biagini, B. Øksendal, A. Sulem, N. Wallner, An introduction to white-noise theory and Malliavin calculus for
fractional Brownian motion, Proc. Roy. Soc. Lond. Ser. A Math. Phys. Eng. Sci. 460 (2041) (2004) 347–372,
Stochastic analysis with applications to mathematical finance.
[8] F. Delbaen, W. Schachermayer, A general version of the fundamental theorem of asset pricing, Math. Ann. 300 (3)
(1994) 463–520.
[9] K. Bichteler, Stochastic integration and L p -theory of semimartingales, Ann. Probab. 9 (1) (1981) 49–89.
[10] F.B. Knight, Foundations of the Prediction Process, in: Oxford Studies in Probability, vol. 1, The Clarendon Press,
Oxford University Press, Oxford Science Publications, New York, 1992.
[11] A. Cherny, When is a moving average a semimartingale? MaPhySto – Research Report2001–28, 2001. Available
from http://www.maphysto.dk/cgi-bin/gp.cgi?publ=318.
[12] P. Cheridito, Gaussian moving averages, semimartingales and option pricing, Stochastic Process. Appl. 109 (1)
(2004) 47–68.
[13] A. Basse, Spectral representation of Gaussian semimartingales. Thiele Centre – Research Report 2008–03, 2008b.
Available from http://www.imf.au.dk/publs?id=672.
[14] T. Jeulin, M. Yor, Moyennes mobiles et semimartingales, Sémin. Probab. XXVII (1557) (1993) 53–77.
[15] A. Basse, Gaussian moving averages and semimartingales, Electron. J. Probab. 13 (39) (2008) 1140–1165.
[16] K. Sato, Lévy Processes and Infinitely Divisible Distributions, in: Cambridge Studies in Advanced Mathematics,
vol. 68, Cambridge University Press, Cambridge, 1999, Translated from the 1990 Japanese original, Revised by the
author.
[17] J. Bertoin, Lévy Processes, in: Cambridge Tracts in Mathematics, vol. 121, Cambridge University Press,
Cambridge, 1996.
[18] P.E. Protter, Stochastic Integration and Differential Equations, Second ed., in: Applications of Mathematics (New
York), vol. 21, Springer-Verlag, Berlin, 2004, Stochastic Modelling and Applied Probability.
[19] B.S. Rajput, J. Rosiński, Spectral representations of infinitely divisible processes, Probab. Theory Related Fields
82 (3) (1989) 451–487.
[20] M.B. Marcus, J. Rosiński, L 1 -norms of infinitely divisible random vectors and certain stochastic integrals, Electron.
Comm. Probab. 6 (2001) 15–29 (electronic).
 A. Basse, J. Pedersen / Stochastic Processes and their Applications 119 (2009) 2970–2991 2991

[21] J. Rosiński, On stochastic integral representation of stable processes with sample paths in Banach spaces, J.
Multivariate Anal. 20 (2) (1986) 277–302.
[22] M. Braverman, G. Samorodnitsky, Symmetric infinitely divisible processes with sample paths in Orlicz spaces and
absolute continuity of infinitely divisible processes, Stochastic Process. Appl. 78 (1) (1998) 1–26.
[23] M.B. Marcus, J. Rosiński, Sufficient conditions for boundedness of moving average processes, in: Stochastic
Inequalities and Applications, in: Progr. Probab., vol. 56, Birkhäuser, Basel, 2003, pp. 113–128.
[24] S. Kwapień, M.B. Marcus, J. Rosiński, Two results on continuity and boundedness of stochastic convolutions, Ann.
Inst. H. Poincaré Probab. Statist. 42 (5) (2006) 553–566.
[25] B.B. Mandelbrot, J.W. Van Ness, Fractional Brownian motions, fractional noises and applications, SIAM Rev. 10
(1968) 422–437.
[26] G.H. Hardy, J.E. Littlewood, Some properties of fractional integrals. I, Math. Z. 27 (1) (1928) 565–606.
[27] N. Dunford, J.T. Schwartz, Linear Operators: Part I: General Theory, in: Pure and Applied Mathematics, vol. 7,
Interscience Publishers, Inc., New York, 1957.
[28] W. Rudin, Functional Analysis, Second ed., in: International Series in Pure and Applied Mathematics, McGraw-Hill
Inc., New York, 1991.
[29] J. Rosiński, On path properties of certain infinitely divisible processes, Stochastic Process. Appl. 33 (1) (1989)
73–87.
[30] I. Karatzas, S.E. Shreve, Brownian Motion and Stochastic Calculus, Second ed., in: Graduate Texts in Mathematics,
vol. 113, Springer-Verlag, New York, 1991.
[31] C. Stricker, Semimartingales gaussiennes—application au problème de l’innovation, Z. Wahrsch. Verw. Gebiete 64
(3) (1983) 303–312.
[32] C. Dellacherie, P.-A. Meyer, Probabilities and Potential B: Theory of Martingales, in: North-Holland Mathematics
Studies, vol. 72, North-Holland Publishing Co., Amsterdam, 1982.
[33] J. Jacod, A.N. Shiryaev, Limit Theorems for Stochastic Processes, Second ed., in: Grundlehren der Mathematischen
Wissenschaften [Fundamental Principles of Mathematical Sciences], vol. 288, Springer-Verlag, Berlin, 2003.
[34] J. Rosiński, G. Samorodnitsky, Distributions of subadditive functionals of sample paths of infinitely divisible
processes, Ann. Probab. 21 (2) (1993) 996–1014.
[35] M.M. Rao, Z.D. Ren, Theory of Orlicz Spaces, in: Monographs and Textbooks in Pure and Applied Mathematics,
vol. 146, Marcel Dekker Inc., New York, 1991.
[36] T. Marquardt, Fractional Lévy processes with an application to long memory moving average processes, Bernoulli
12 (6) (2006) 1099–1126.
[37] A. Benassi, S. Cohen, J. Istas, On roughness indices for fractional fields, Bernoulli 10 (2) (2004) 357–373.
[38] M. Ledoux, M. Talagrand, Probability in Banach Spaces, in: Ergebnisse der Mathematik und ihrer Grenzgebiete (3)
[Results in Mathematics and Related Areas (3)], vol. 23, Springer-Verlag, Berlin, 1991, Isoperimetry and processes.
[39] J. Rosiński, On series representations of infinitely divisible random vectors, Ann. Probab. 18 (1) (1990) 405–430.