Skip to main content

Second-order integro-differential parabolic variational inequalities arising from the valuation of American option

Abstract

This article deals with the existence and uniqueness of solution to the R d 1 -valued parabolic variational inequalities with integro-differential terms which arise from the valuation of American option. The authors use the penalty method to construct a sequence of approximation parabolic problem and hence obtain the existence and uniqueness of solution to the approximation problem by using fixed point theory. Then the solution of parabolic variational inequalities is obtained by showing that the solution of this penalty problem converges to the variational inequalities. The uniqueness of the solution is also proven.

MSC:35B40, 35K35.

1 Introduction

In this paper, we are concerned with the existence and uniqueness of solution to an R d 1 -valued parabolic variational inequality with the integro-differential terms

{ ( u k ) t L k u F k ( x , t , u , u ) 0 , in  Q T , u k ( x , t ) u 0 k ( x ) , in  Q T , ( ( u k ) t L k u F k ( x , t , u , u ) ) ( u k u 0 k ( x ) ) = 0 , in  Q T , u k ( x , 0 ) = u 0 k ( x ) , on  Ω , u k ( x , t ) = g k ( x , t ) , on  Ω × ( 0 , T ) ,
(1)

where Ω R d is an open set with a smooth boundary Ω, Q T =Ω×[0,T] is a parabolic domain for some T>0. L=( L 1 , L 2 ,, L d 1 ), and L k = L k (x,t) is a divergence form second-order elliptic operator satisfying

L k u= r = 1 d 1 i , j = 1 d x j ( a k r i j ( x , t ) x i u r ) r = 1 d 1 i = 1 d b k r i (x,t) x i u r r = 1 d 1 c k r (x,t) u r .

Moreover, the integro-differential operator F k is defined by

F k (x,t,u,Δu)= Ω f k ( x , t , y , u ( x , t ) , u ( x , t ) ) dy

which is a continuous integral operator as defined in [1].

The concrete motivations of studying (1) can be easily found in the literature. If d 1 =1, F k (x,t,u,Δu)=0, and the operator L k degenerates to

Lu= t u σ 2 2 2 x 2 u+ ( σ 2 2 r ) x u+ru,

(1) becomes the linear variational inequality based on the famous Black-Scholes equation (see [2]). Such a variational inequality arises in many applications of American option pricing (see [3, 4]). To deal with this problem, some scholars often introduce a free and moving boundary problem. By adding a certain penalty term to the Black-Scholes equation, the solution of this variational inequality is extended to a fixed domain. Furthermore, this penalty term forces the solution to stay above the payoff function at expiry. Throughout the last decade, a number of papers addressing penalty schemes for American options have been published (see, for instance, [59] and references therein).

However, several Black-Scholes models proposed in recent years, such as the model found in [8, 9], have allowed the risk assets driven by levy processes. In this model, the risk asset S as in the Black-Scholes model, follows the stochastic process (assume d 1 =1)

d S t =μ S t dt+σ S t d z t ,

from which we obtain the following PIDE:

t u= σ 2 2 [ x x u x u]+ [ u ( x + y , t ) u ( x , t ) ( e y 1 ) x u ] v(dx),
(2)

where z t is a kind of levy process ( for details, see [10, 11]). The authors in [9] generalize (2) and prove the existence and uniqueness of a classical solution to a more general problem in the parabolic domain Q T =Ω×[0,T], where Ω is an open, unbounded subset of R d , with a smooth boundary Ω. A related work in the context of quantum mechanics has been studied in [12, 13].

Therefore, the authors of this paper intend to study a more complex variational inequality involved in American option based on the more complicated PIDE type Black-Scholes equation than (2). And we will consider R d 1 -valued parabolic variational inequality (1) using the penalty method. The rest of this article is as follows. In Section 2, we state the main result. Section 3 discusses the penalty problem which will be used to prove our main result. In Section 4, we show the proof of our main result.

2 Main result

In this section, we present some notations and lemmas which are important to be used to prove our main result. Define L p (Ω)= L p (Ω; R d 1 ) as the space of all R d 1 -value functions u=( u 1 , u 2 ,, u d 1 ) satisfying

u L p ( Ω ) p = k = 1 d 1 u k L p ( R d ) p <.

Let 1p and kN. By W p k , we mean that the -value functions space satisfies

W p k (Ω)= { u L p ( Ω ; R ) | D α u L p ( Ω ; R ) , 1 α k }

which is Banach spaces with the norm

v W p k ( Ω ) = 0 | α | k D α v L p ( Ω ) .

Further, we use the following Sobolev spaces which are modified from -value version to R d 1 -value version

W p k (Ω)= { u L p ( Ω ; R d 1 ) | D α u L p ( Ω ; R d 1 ) , 1 α k }

with the norm

v W p k ( Ω ) = 0 | α | k D α v L p ( Ω ) .

As mentioned in [14], we shall use C 2 k + δ , k + δ / 2 ( Q ¯ T ) to denote the Holder space which satisfies

C 2 k + δ , k + δ / 2 ( Q ¯ T )= { u | D α t ρ C δ , δ / 2 ( Q ¯ T ) , | α | + 2 ρ 2 k } .

For 0<T<, we also use some following common notation from PDEs without mentioning:

H k ( Ω ) = W 2 k ( Ω ) , H k ( Ω ) = W 2 k ( Ω ) , H k ( Ω ) = W 2 k ( Ω ) , H k ( Ω ) = W 2 k ( Ω ) , L p ( 0 , T ; H k ( Ω ) ) = { v ( , t ) | v ( , t ) H k ( Ω )  in  ( 0 , T ) ; v ( , t ) H k ( Ω ) L p ( 0 , T ) } , L p ( 0 , T ; H k ( Ω ) ) = { v ( , t ) | v ( , t ) H k ( Ω )  in  ( 0 , T ) ; v ( , t ) H k ( Ω ) L p ( 0 , T ) } , u L p ( 0 , T ; H k ( Ω ) ) p = 0 T u ( , t ) H k ( Ω ) p d t , u L p ( 0 , T ; H k ( Ω ) ) p = 0 T u ( , t ) H k ( Ω ) p d t ,

where 0<T<.

Moreover, we use the following lemmas to show the existence and uniqueness of the solution for the penalty approximation of our main problem.

Definition 2.1 Assume that X is a real Banach space, the space C([0,T];X) consists of all continuous functions u:[0,T]X with

u C ( [ 0 , T ] ; X ) = max 0 t T u X <.

Then C([0,T];X) is a Banach space endowed with the norm u C ( [ 0 , T ] ; X ) .

Definition 2.2 A mapping A:XX is called to be compact if and only if the sequence { A [ u ] } k = 1 is precompact for each bounded sequence { u } k = 1 , that is, there exists a subsequence { u k j } j = 1 such that { A [ u k j ] } j = 1 converges in X.

Lemma 2.3 (Schaefer’s fixed point theorem)

Suppose that A:XX is a continuous and compact mapping. Further, assume that the set

{ u X : u = λ A [ u ] ,  for some  0 λ 1 }

is bounded. Then A has a fixed point.

The following two lemmas from the linear theory of parabolic partial differential equations can be found in [15].

Lemma 2.4 (Energy estimates)

Consider the problem

{ v k t L k v = f k ( x , t ) , in  Q T , v k ( x , 0 ) = 0 , on  Ω , v k ( x , t ) = 0 , on  Ω × ( 0 , T )
(3)

with f k L 2 (0,T; L 2 (Ω)). Then there exists a unique solution v L 2 (0,T; H 0 1 ( B R ))C([0,T]; L 2 ( B R )) to problem (9) that satisfies

max 0 < t < T v ( t ) L 2 ( Ω ) + v ( t ) L 2 ( 0 , T ; H 0 1 ( Ω ) ) + v t L 2 ( 0 , T ; H 0 1 ( Ω ) ) C f L 2 ( 0 , T ; L 2 ( Ω ) ) ,

where C is a positive constant depending only on Ω, T and the operator L.

Lemma 2.5 (Improved regularity)

There exists a unique weak solution

u L 2 ( 0 , T ; H 0 1 ( B R ) ) C ( [ 0 , T ] ; L 2 ( B R ) )

to problem (3), with u t L 2 (0,T; H 1 (Ω)). Moreover,

u L 2 ( 0 , T ; H 2 ( Ω ) ) L ( 0 , T ; H 0 1 ( Ω ) ) , u t L 2 ( 0 , T ; H 1 ( Ω ) ) .

We also have the estimate

ess sup 0 < t < T u ( t ) H 0 1 ( Ω ) + u ( t ) L 2 ( 0 , T ; H 2 ( Ω ) ) + u t L 2 ( 0 , T ; L 2 ( Ω ) ) C f L 2 ( 0 , T ; L 2 ( Ω ) ) ,
(4)

where C is a positive constant depending only on Ω, T and the operator L.

Throughout this section, we impose the following assumptions:

(A1) The coefficients a k r i j (x,t), b k r i (x,t), c k r (x,t) belong to the Holder space C δ , δ / 2 ( Q ¯ T ).

(A2) The operator t L k is (uniformly) parabolic, that is, there exists a constant θ>0 such that

θ k = 1 d 1 | v k | 2 k , r = 1 d 1 i , j = 1 d a k r i j (x,t) i v r j v k for all (x,t) Q T v R d .

(A3) There exists a positive constant B such that for all v H 0 1 (Ω) and (x,t) Q T , we have

B k = 1 d 1 i v k v k k , r = 1 d 1 b k r i (x,t) i v r v k , c k r (x,t)>0,k,r=1,2,, d 1 .

(A4) u 0 k (x) and g k (x,t) belong to the Holder spaces C 2 + δ ( R d ) and C 2 + δ , 1 + δ / 2 ( Q ¯ T ), respectively. Moreover, for all (x,t) Q T , we have that u 0 k (x)< g k (x,t), k=1,2,, d 1 .

(A5) The following two consistency conditions

g k (x,0)= u 0 k (x)and g k (x,0) L k (x,0) u 0 k (x)=0

hold for all xΩ, k=1,2,, d 1 .

(A6) f k (x,t,y,z,p) is nonnegative and belongs to C 1 ( Q T ×Ω× R d 1 × R d 1 ), k=1,2,, d 1 , and there exists 0<ε<1 satisfying

| f k ( , , , z 1 , p 1 ) f k ( , , , z 2 , p 2 ) | ε| z 1 z 2 |+ε| p 1 p 2 |.

(A7) For some C 0 >0, f=( f 1 , f 2 ,, f d 1 ) satisfies the estimate

| f k ( x , t , y , z , p ) | C 0 ( f 0 , k + r = 1 d 1 | z r | + r = 1 d 1 | p r | )

for all (x,t,y,z,p) Q T ×Ω× R d 1 × R d 1 , where C 0 is a positive constant independent of f, f 0 , k L 2 (Ω).

(A8) If u H loc 2 ( R d ), then F(x,t,u,u) L 2 (0,T; L loc 2 (R)). If w n w in L 2 (0,T; H 0 1 (Ω)), then F k (x,t, w n ,Δ w n ) F k (x,t,w,Δw) in L 2 (0,T; L 2 (Ω)).

Using these assumptions, we will elaborate our main result.

Theorem 2.6 Under hypotheses (A1)-(A8), there exists a unique solution u L 2 (0,T; H 2 (Ω))C([0,T]; L 2 (Ω)) to variational inequality (1) satisfying

u L 2 ( 0 , T ; H 0 2 ( Ω ) ) N ( u 0 L 2 ( Ω ) + g L 2 ( 0 , T ; L 2 ( Ω ) ) + F ( , , 0 , 0 ) L 2 ( 0 , T ; L 2 ( Ω ) ) ) .

The proof of Theorem 2.6 will be given in Section 4.

3 The penalty problem

In order to prove the existence and uniqueness of the solution, we consider the following penalty approximation of problem (1)

{ ( u ε k ) t L k u ε F k ( x , t , u ε , u ε ) + β ε ( u ε k u 0 k ( x ) ) = 0 , in  Q T , u ε k ( x , 0 ) = u 0 k ( x ) , on  Ω , u ε k ( x , t ) = g k ( x , t ) , on  Ω × ( 0 , T ) ,
(5)

where β ε , k () is the penalty function satisfying

0 < ε 1 , β ε , k ( x ) C 2 ( R ) , β ε , k ( x ) 0 , β ε , k ( x ) 0 , β ε , k ( x ) 0 , lim ε 0 β ε , k ( x ) = { 0 , x > 0 , , x < 0 .
(6)

Next, we introduce a change of variables

v(x,t)= u ε (x,t)φ(x,t)

to transform problem (5) into the zero boundary condition of the form

{ ( v k ) t L k v = F k ( x , t , v + φ , ( v + φ ) ) β ε ( v k + φ k u 0 k ( x ) ) , in  Q T , v k ( x , 0 ) = 0 , on  Ω , v k ( x , t ) = φ k ( x , t ) = 0 , on  Ω × ( 0 , T ) .
(7)

Here φ=( φ 1 , φ 2 ,, φ d 1 ) C 2 + δ , 1 + δ / 2 ( Q ¯ T ) is the unique solution to the problem

{ ( u k ) t L k u = 0 , in  Q T , u k ( x , 0 ) = u 0 k ( x ) , on  Ω , u k ( x , t ) = g k ( x , t ) , on  Ω × ( 0 , T )
(8)

satisfying

φ L 2 ( 0 , T ; H 2 ( Ω ) ) N ( u 0 L 2 ( Ω ) + g L 2 ( 0 , T ; L 2 ( Ω ) ) ) .

For further details, see Theorem 10.4.1 in [15].

Definition 3.1 v is said to be a weak solution to problem (7) if

v L 2 ( 0 , T ; H 0 1 ( Ω ) ) , v t L 2 ( 0 , T ; H 1 ( Ω ) )

and

Ω ( v k t g + r = 1 d 1 i , j = 1 d a k r i j ( x , t ) i v r j g ) d x + Ω ( r = 1 d 1 i = 1 d b k r i ( x , t ) i v r g + r = 1 d 1 c k r ( x , t ) v r g ) d x = Ω F k ( x , t , v + φ , ( v + φ ) ) g d x Ω β ε ( v k + φ k u 0 k ) g d x
(9)

for all g H 0 1 (Ω).

Lemma 3.2 If v L 2 (0,T; H 0 1 (Ω)) and v t L 2 (0,T; H 1 (Ω)), then

vC ( [ 0 , T ] ; L 2 ( Ω ) ) .

The mapping t v ( t ) L 2 ( Ω ) 2 is absolutely continuous with

d d t v k ( t ) L 2 ( Ω ) 2 =2 Ω v k t v k dt, d d t v ( t ) L 2 ( Ω ) 2 =2 k = 1 d 1 Ω v k t v k dt

for any 0tT. The proof is quite standard and can be found in [15].

Lemma 3.3 If v is a weak solution to problem (7), then there exists a positive constant C such that

v ( t ) L 2 ( 0 , T ; H 0 1 ( Ω ) ) C,

where C is independent of v.

Proof Choosing v H 0 1 (Ω) as the test function in (9), we obtain

Ω ( v k t v k + r = 1 d 1 i , j = 1 d a k r i j ( x , t ) i v r j v k ) d x + Ω ( r = 1 d 1 i , j = 1 d b k r i ( x , t ) i v r v k + r = 1 d 1 c k r ( x , t ) v r v k ) d x = Ω F k ( x , t , v + φ , ( v + φ ) ) v d x Ω β ε ( v k + φ k u 0 k ) v k d x .
(10)

From (1), we easily have that

v k + φ k u 0 k for all (x,t) Q T .

This and (6) lead to

| Ω β ε ( v k + φ k u 0 k ) v k d x | | β ε ( 0 ) | Ω | v k | dx.
(11)

By (A7), so that we get

| F k ( x , t , v + φ , ( v + φ ) ) | | Ω f k ( x , t , y , v + φ , ( v + φ ) ) d y | C 0 Ω ( f 0 , k + k = 1 d 1 | v k | + k = 1 d 1 | v k | + k = 1 d 1 | φ k | + k = 1 d 1 | φ k | ) d x ,
(12)

where C 0 is a positive constant which depends only on the region Ω. It follows by (A2) and (A3) that

θ k = 1 d 1 | v k | 2 k , r = 1 d 1 i , j = 1 d a k r i j i v r j v k ,
(13)
Ω k , r = 1 d 1 i = 1 d b k r i ( x , t ) i v r v k d x B k = 1 d 1 i = 1 d Ω i v k v k d x = B 2 k = 1 d 1 i = 1 d Ω i ( v k ) 2 d x = 0
(14)

and

Ω c k r (x,t) v 2 dx0.
(15)

Now, we use Theorem 10.4.1 in [15] and substitute (11)-(15) into (10) to arrive at

1 2 d d t v k L 2 ( Ω ) 2 + θ k = 1 d 1 Ω | v k | 2 d x C 0 Ω ( f 0 , k + k = 1 d 1 | v k | + k = 1 d 1 | v k | + k = 1 d 1 | φ k | + k = 1 d 1 | φ k | ) k = 1 d 1 | v k | d x + | β ε ( 0 ) | Ω k = 1 d 1 | v k | d x .
(16)

It follows, by using the Cauchy inequality with ω>0, that

1 2 d d t v k L 2 ( Ω ) 2 + θ k = 1 d 1 Ω | v k | 2 d x C 0 ( ω Ω f 0 , k 2 d x + 1 4 ω Ω ( k = 1 d 1 | v k | ) 2 d x ) + C 0 Ω ( k = 1 d 1 | v k | ) 2 d x + C 0 ( ω Ω ( k = 1 d 1 | v k | ) 2 d x + 1 4 ω Ω ( k = 1 d 1 | v k | ) 2 d x ) + C 0 ( ω Ω ( k = 1 d 1 | φ k | ) 2 d x + 1 4 ω Ω ( k = 1 d 1 | v k | ) 2 d x ) + C 0 ( ω Ω ( k = 1 d 1 | φ k | ) 2 d x + 1 4 ω Ω ( k = 1 d 1 | v k | ) 2 d x ) + | β ε ( 0 ) | ( ω Ω d x + 1 4 ω Ω ( k = 1 d 1 | v k | ) 2 d x ) .

Next, we choose 0<ω1 to arrive at

1 2 d d t v k L 2 ( Ω ) 2 + C 1 v H 0 1 ( Ω ) 2 C 2 v L 2 ( Ω ) 2 + C ˜ (t),

where C 1 and C 2 are positive constants and

C ˜ (t)= C 0 ω Ω ( f 0 , k 2 + ( k = 1 d 1 | φ k | ) 2 + ( k = 1 d 1 | φ k | ) 2 ) dx+ω | β ε ( 0 ) | Ω 1dx.

Note that φ k C 2 + δ , 1 + δ / 2 ( Q ¯ T ). Therefore we choose a positive constant C 3 = sup 0 < t < T C ˜ (t) to obtain

1 2 d d t v ( t ) L 2 ( Ω ) 2 + C 1 v H 0 1 ( Ω ) 2 C 2 v L 2 ( Ω ) 2 + C 3 .
(17)

On the one hand, letting η(t)= v ( t ) L 2 ( Ω ) 2 , (17) gives

η (t)2 C 2 η(t)+2 C 3 .

Using the differential form of Gronwall inequality with η(0)= v ( 0 ) L 2 ( Ω ) 2 =0, we get

v ( t ) L 2 ( Ω ) 2 2 e 2 C 2 t C 3 t

and

max 0 < t < T v ( t ) L 2 ( Ω ) 2 C 4 ,
(18)

where C 4 is a positive constant independent of v.

On the other hand, an integration of (17) from 0 to T with v ( 0 ) L 2 ( Ω ) 2 =0 yields

C 1 0 T v H 0 1 ( Ω ) 2 dt C 2 0 T v L 2 ( Ω ) 2 dt+ C 3 T+ lim t T v ( t ) L 2 ( Ω ) 2 .

Using inequality (18), we obtain

v ( t ) L 2 ( 0 , T ; H 0 1 ( Ω ) ) C 2 C 4 C 1 T+ C 3 C 1 T+ C 4 C 1 .

Thus, the proof is ended by letting C= C 2 C 4 C 1 T+ C 3 C 1 T+ C 4 C 1 >0. □

Lemma 3.4 The solution to problem (7) satisfies

u k (x,t) u 0 k (x).

Proof Using (A8), there exists a positive constant M satisfying

| F k ( x , t , u , u ) | M.

Here we plan to finish the proof by using contradiction. Assume that k = 1 d 1 { u k (x,t)< u 0 k (x)} is not empty, that is, at least there exists one k satisfying u k < u 0 k . Let β ε (0)=M, from (16), one gets

u t Lu= F k (x,t,u,u) β ε ( u k u 0 k ( x ) ) > F k (x,t,u,u) β ε (0)0.

Using the standard maximum principle and (A4), one gets

u k (x,t) u 0 k (x).

This is obviously contradictory. Therefore, we conclude that u k (x,t) u 0 k (x) for any (x,t) Q T . □

Lemma 3.5 If (A1)-(A8) are satisfied, there exists a unique weak solution v L 2 (0,T; H 0 2 (Ω))C([0,T]; L 2 (Ω)) to problem (7) satisfying

v L 2 ( 0 , T ; H 0 2 ( Ω ) ) N F ( , , 0 , 0 ) L 2 ( 0 , T ; L 2 ( Ω ) ) +N | β ε ( 0 ) | T,
(19)

where N is a positive constant independent of v.

Proof Given w L 2 (0,T; H 0 1 (Ω)), set

f w =( f 1 , w , f 2 , w ,, f d 1 , w ), f k , w (x,t)= F k ( x , t , w + φ , ( w + φ ) ) β ε ( w k + φ k u 0 k ) .

The use of (A8) leads to

f w (x,t) L 2 ( 0 , T ; L 2 ( Ω ) ) .
(20)

By Lemma 2.4, there exists a unique solution v L 2 (0,T; H 0 2 (Ω))C([0,T]; L 2 (Ω)) to

{ ( v k ) t L k v = f k , w ( x , t ) in  Q T , v k ( x , 0 ) = 0 , on  Ω , v k ( x , t ) = 0 , on  Ω × ( 0 , T ) .
(21)

Define the mapping

M: L 2 ( 0 , T ; H 0 2 ( Ω ) ) L 2 ( 0 , T ; H 0 2 ( Ω ) ) ,wM(w)=v,

where v is derived from w via (21).

Here we plan to prove the existence and uniqueness by Schaefer’s fixed point theorem. So that we need to present the continuity and compactness of the mapping M. In this proof we only prove the continuity of the mapping M. The compactness can be obtained by following similar arguments, so we omit it here.

Let { w n } n L 2 (0,T; H 0 1 (Ω)) be a sequence such that

w k win  L 2 ( 0 , T ; H 0 1 ( Ω ) ) .
(22)

By the improved regularity (4), we obtain

v n L 2 ( 0 , T ; H 2 ( Ω ) ) f w k L 2 ( 0 , T ; L 2 ( Ω ) ) for  v n =M[ w n ],n=1,2,.

By using (A8) and Lemma 3.4 with β ε (x)0, β ε (x)0, we have

f w k L 2 ( 0 , T ; L 2 ( Ω ) ) F k ( x , t , w + φ , w + φ ) L 2 ( 0 , T ; L 2 ( Ω ) ) + | β ε ( 0 ) | d 1 T F k ( x , t , w + φ , w + φ ) F k ( x , t , 0 , 0 ) L 2 ( 0 , T ; L 2 ( Ω ) ) + F k ( x , t , 0 , 0 ) L 2 ( 0 , T ; L 2 ( Ω ) ) + | β ε ( 0 ) | d 1 T α w L 2 ( 0 , T ; L 2 ( Ω ) ) + α w L 2 ( 0 , T ; L 2 ( Ω ) ) + F k ( x , t , 0 , 0 ) L 2 ( 0 , T ; L 2 ( Ω ) ) + | β ε ( 0 ) | d 1 T α w L 2 ( 0 , T ; H 2 ( Ω ) ) + F k ( x , t , 0 , 0 ) L 2 ( 0 , T ; L 2 ( Ω ) ) + | β ε ( 0 ) | d 1 T .

The two inequalities above lead to

v n L 2 ( 0 , T ; H 2 ( Ω ) ) α w n L 2 ( 0 , T ; H 2 ( Ω ) ) + F k ( x , t , 0 , 0 ) L 2 ( 0 , T ; L 2 ( Ω ) ) + | β ε ( 0 ) | d 1 T .
(23)

Next, we pay our attention to the sequence { f w n L 2 ( 0 , T ; L 2 ( Ω ) ) } n . Since w n w in L 2 (0,T; H 0 1 (Ω)), from (A8) we have

f w n ( x , t ) L 2 ( 0 , T ; L 2 ( Ω ) ) f w ( x , t ) L 2 ( 0 , T ; L 2 ( Ω ) ) .

This and (20) lead to the fact that the sequence { f w n L 2 ( 0 , T ; L 2 ( Ω ) ) } n is bounded, that is,

f w n L 2 ( 0 , T ; L 2 ( Ω ) ) C 6 .
(24)

Combing (23) with (24), the sequence { v n } n is bounded uniformly in L 2 (0,T; H 2 (Ω)). In a similar way, { v n t } n is uniformly bounded in L 2 (0,T; H 1 (Ω)). By using Rellich’s theorem (see [16]), there exist a subsequence { v j } j L 2 (0,T; H 0 1 (Ω)) and a function v L 2 (0,T; H 0 1 (Ω)) which satisfies

v n j vin  L 2 ( 0 , T ; H 0 1 ( Ω ) )  as j

such that

Ω ( v n j k t ϕ L k v n j ϕ ) dx= Ω f k , w n j (x,t)ϕdxfor each ϕ H 0 1 (Ω).
(25)

We combine (22) with (25) to arrive at

Ω ( v k t ϕ L k v ϕ ) dx= Ω f k , w (x,t)ϕdx.

Therefore,

v=A[w],A[ w n ]A[w]in  L 2 ( 0 , T ; H 0 1 ( Ω ) ) .

Further, by Lemma 3.3, we have that {w L 2 (0,T; H 0 1 (Ω)):w=A[w]} is bounded. Hence the existence and uniqueness of this theorem are proven by using Lemma 2.3 with λ=1.

Finally, we pay our attention to the estimate (19). Letting n in (23), we obtain

v L 2 ( 0 , T ; H 2 ( Ω ) ) α v L 2 ( 0 , T ; H 2 ( Ω ) ) + F k ( x , t , 0 , 0 ) L 2 ( 0 , T ; L 2 ( Ω ) ) + | β ε ( 0 ) | d 1 T.

It follows by 0<α<1 that

v L 2 ( 0 , T ; H 2 ( Ω ) ) 1 1 α F k ( x , t , 0 , 0 ) L 2 ( 0 , T ; L 2 ( Ω ) ) + 1 1 α | β ε ( 0 ) | d 1 T.

Therefore, the proof is complete. □

4 The proof of the main result

In this section, we prove that the solution to problem (5) converges to that of problem (1) when ε0. From Lemma 3.3 and Lemma 3.5, we conclude that u ε in (7) exists a unique solution satisfying

u ε L 2 ( 0 , T ; H 0 1 ( Ω ) ) C , u ε L 2 ( 0 , T ; H 0 2 ( Ω ) ) N F ( , , 0 , 0 ) L 2 ( 0 , T ; L 2 ( Ω ) ) + N | β ε ( 0 ) | d 1 T .

Thus, there exists a subsequence of { u ε } still denoted by itself for convenience, and uL(0,T; H 2 (Ω)) such that

u ε k weak u k in  L ( 0 , T ; H 2 ( Ω ) ) , k = 1 , 2 , , d 1 , u ε k uniformly u k in  C ( Ω T ) , k = 1 , 2 , , d 1 .

Letting ε0 in (5) and using the maximum principle, we arrive at

{ ( u k ) t L k u F k ( x , t , u , u ) 0 , u k ( x , 0 ) = u 0 k ( x ) , u k ( x , t ) = g k ( x ) .
(26)

Comparing (1) with (26), we only need to prove that

[ ( u k ) t L k u F k ( x , t , u , u ) ] ( u k ( x , t ) u 0 k ( x ) ) =0

holds. Note that u k (x,0) u 0 k (x) derived from Lemma 3.3. Thus, we can end the proof by showing

( u k ) t L k u F k (x,t,u,u)=0when  u k (x,t) u 0 k (x).
(27)

In fact, there exists a positive constant such that for all ( x 0 , t 0 ){ u k (x,t)> u 0 k (x)},

u ε k ( x 0 , t 0 )>α+ u 0 k ( x 0 )

holds when ε is sufficiently small. Further, from (15) we have that

0 β ε ( u ε k u 0 k ) β ε (α)0.
(28)

Therefore, we conclude that

( u k ) t L k u F k (x,t,u,u)=0in L ( 0 , T ; H 2 ( Ω ) ) .

Now, we prove the uniqueness by contradiction. Assume that u 1 =( u 1 1 , u 1 2 ,, u 1 d 1 ) and u 2 =( u 2 1 , u 2 2 ,, u 2 d 1 ) are the solutions of (1) and u 1 u 2 . That is, there exists at least one k satisfying u 1 k u 2 k . For simplicity, we assume that { u 1 k > u 2 k } is not empty (if not, we assume { u 1 k < u 2 k } is not empty), so that we have

u 1 k > u 2 k u 0 k (x)in  { u 1 k > u 2 k } .

This and (27) lead to

( u 1 k ) t L k u 1 F k ( x , t , u 1 , Δ u 1 ) = 0 in  { u 1 k > u 2 k } , ( u 2 k ) t L k u 2 F k ( x , t , u 2 , Δ u 2 ) 0 in  { u 1 k > u 2 k } , ( u 1 k u 2 k ) t L k ( u 1 u 2 ) ( F k ( x , t , u 1 , Δ u 1 ) F k ( x , t , u 2 , Δ u 2 ) ) 0 .

By the maximum principle and (A8), we have that

u 1 k u 2 k 0in  { u 1 k > u 2 k } .

This is obviously contradictory. Therefore, we conclude that problem (1) has a unique solution. Moreover the estimate can be easily obtained by Lemma 3.5 and (28) with ε0.

References

  1. 1.

    Blanchet A: On the regularity of the free boundary in the parabolic obstacle problem application to American options. Nonlinear Anal. 2006, 65: 1362-1378. 10.1016/j.na.2005.10.009

    MathSciNet  Article  MATH  Google Scholar 

  2. 2.

    Cox J, Ross S: The valuation of options for alternative stochastic processes. J. Financ. Econom. 1976, 3: 145-166. 10.1016/0304-405X(76)90023-4

    Article  Google Scholar 

  3. 3.

    Jeunesse M, Jourdain B: Regularity of the American Put option in the Black-Scholes model with general discrete dividends. Stoch. Process. Appl. 2012, 122: 3101-3125. 10.1016/j.spa.2012.05.009

    MathSciNet  Article  MATH  Google Scholar 

  4. 4.

    Kohler M, Krzyzak A: Pricing of American options in discrete time using least squares estimates with complexity penalties. J. Stat. Plan. Inference 2012, 142: 2289-2307. 10.1016/j.jspi.2012.02.031

    MathSciNet  Article  MATH  Google Scholar 

  5. 5.

    Forsyth PA, Vetzal KR: Quadratic convergence for valuing American options using a penalty method. SIAM J. Sci. Comput. 2002, 23: 2095-2122. 10.1137/S1064827500382324

    MathSciNet  Article  MATH  Google Scholar 

  6. 6.

    Khaliq AQM, Voss DA, Kazmi SH: A linearly implicit predictor-corrector scheme for pricing American options using a penalty method approach. J. Bank. Finance 2006, 30: 489-502. 10.1016/j.jbankfin.2005.04.017

    Article  Google Scholar 

  7. 7.

    Fasshauer G, Khaliq AQM, Voss DA: Using meshfree approximation for multi asset American options. J. Chin. Inst. Eng. 2004, 27: 563-571. Mesh free methods 10.1080/02533839.2004.9670904

    Article  Google Scholar 

  8. 8.

    Marcozzi M: On the approximation of optimal stopping problems with application to financial mathematics. SIAM J. Sci. Comput. 2001, 22: 1865-1884.

    MathSciNet  Article  MATH  Google Scholar 

  9. 9.

    Halluin Y, Forsyth PA, Labah G: A penalty method for American options with jump-diffusion processes. Numer. Math. 2004, 97: 321-352. 10.1007/s00211-003-0511-8

    MathSciNet  Article  MATH  Google Scholar 

  10. 10.

    Mariani MC, SenGupta I: Solutions to an integro-differential parabolic problem arising in the pricing of financial options in a Lévy market. Nonlinear Anal., Real World Appl. 2012, 12: 3103-3113.

    MathSciNet  Article  Google Scholar 

  11. 11.

    Florescu I, Mariani MC: Solutions to an integro-differential parabolic problem arising in the pricing of financial options in a Lévy market. Electron. J. Differ. Equ. 2010, 62: 1-10.

    MathSciNet  Google Scholar 

  12. 12.

    SenGupta I: Spectral analysis for a three-dimensional super radiance problem. J. Math. Anal. Appl. 2011, 375: 762-776. 10.1016/j.jmaa.2010.10.003

    MathSciNet  Article  MATH  Google Scholar 

  13. 13.

    SenGupta I: Differential operator related to the generalized super radiance integral equation. J. Math. Anal. Appl. 2010, 369: 101-111. 10.1016/j.jmaa.2010.02.034

    MathSciNet  Article  MATH  Google Scholar 

  14. 14.

    Adams RA: Sobolev Spaces. Academic Press, New York; 1975.

    Google Scholar 

  15. 15.

    Evans LC Grad. Stud. Math. 19. In Partial Differential Equations. 2nd edition. Am. Math. Soc., Providence; 2010.

    Google Scholar 

  16. 16.

    Folland GB: Introduction to Partial Differential Equations. 2nd edition. Princeton University Press, Princeton; 1995.

    Google Scholar 

Download references

Acknowledgements

This work was supported by the National Nature Science Foundation of China (Grant No. 71171164) and the Doctorate Foundation of Northwestern Polytechnical University (Grant No. CX201235). The authors are sincerely grateful to the referees and the associate editor handling the paper for their valuable comments.

Author information

Affiliations

Authors

Corresponding author

Correspondence to Yudong Sun.

Additional information

Competing interests

The authors declare that they have no competing interests.

Authors’ contributions

YS and YS carried out the proof of the main part of this article, MW corrected the manuscript and participated in its design and coordination. All authors read and approved the final manuscript.

Rights and permissions

Open Access This article is distributed under the terms of the Creative Commons Attribution 2.0 International License (https://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Reprints and Permissions

About this article

Cite this article

Sun, Y., Shi, Y. & Wu, M. Second-order integro-differential parabolic variational inequalities arising from the valuation of American option. J Inequal Appl 2014, 8 (2014). https://doi.org/10.1186/1029-242X-2014-8

Download citation

Keywords

  • American option
  • integro-differential
  • variational inequalities
  • existence
  • uniqueness