Vous êtes sur la page 1sur 7

Nonl. Analysis and Dierential Equations, Vol. 1, 2013, no.

2, 75 - 81
HIKARI Ltd, www.m-hikari.com
On Nonlinear Black-Scholes Equations
R. Agliardi
Department of Mathematics
University of Bologna
40126 Bologna, Italy
P. Popivanov and A. Slavova
Institute of Mathematics
Bulgarian Academy of Sciences
Soa 1113, Bulgaria
Copyright c 2013 R. Agliardi et al. This is an open access article distributed under the
Creative Commons Attribution License, which permits unrestricted use, distribution, and
reproduction in any medium, provided the original work is properly cited.
Abstract
This paper revisits some solution methods for Black-Scholes equation
and some of its nonlinear versions arising in option pricing theory.
Keywords: nonlinear PDEs; Black-Scholes; Mellin transform
1 Introduction
Modern nance started with Black-Scholes linear partial dierential equation
[2], which was obtained under several model restrictions. When some as-
sumptions (e.g. no transaction costs, agents are price takers and there are no
feedback eects from the trading activity, the market is perfectly liquid, etc)
are relaxed, then the linear Black-Scholes equation needs to be replaced by a
nonlinear one. Several models have been proposed to address the case of the
price impact eects from large traders (see [8], [9] and [11], for example). In
[7] many models and methods from this stream of literature are collected. The
most comprehensive equation providing the price of a European option in such
a framework is of the form:
V
t
+
1
2
(s,
V
s
,

2
V
s
2
)
2
s
2
2
V
s
2
+ rs
V
s
rV = 0
76 R. Agliardi, P. Popivanov and A. Slavova
where V is the price of the European option under study, s is the price of the
underlying stock, and r is the risk-free interest rate.
In [1] we proved a comparison principle for these nonlinear PDEs and investi-
gated some properties of the viscosity solutions. The aim of this paper is to
focus on the analytical solution method. The approach we follow to get ex-
act solutions for the nonlinear Black-Scholes equation diers from [3] and [4].
Before giving the main result in Section 3, in Section 2 we revisit the Mellin
transform approach, which is an alternative solution method that is seldom
used in Finance (see [10]).
2 Cauchy problem for the linear Black-Scholes
equation via the Mellin transformation
Consider the Cauchy problem V (s, T) = f(s) for the backward degenerate
parabolic equation:
V
t
+
1
2

2
s
2

2
V
s
2
+ rs
V
s
rV = 0, (1)
where 0 t T, s > 0, = const > 0, r = const; V = V (s, t).
In the original paper [2], the Cauchy problem for (1) - with nal payo f(s) =
max(s k, 0) was transformed into a Cauchy problem for the heat equation
Y
z
= Y
xx
by changing to variables:
V (s, t) = e
r(tT)
Y (x, z) with x = (
2r

2
1)(ln
s
K
+ (r

2
2
)(T t)) and
z =
2

2
(r

2
2
)
2
(T t). In this section we follow an alternative method based
on the Mellin transform and consider a generic terminal datum f(s). As we
know the Mellin transform
u(z) = (Mu)(z) =


0
t
z1
u(t)dt, z C
1
. (2)
In many cases u(z) is analytic in a strip of the type {z : < Re z < }. Under
several conditions an inverse formula for (2) is valid, namely
(M
1
h)(t) =
1
2i

t
z
h(z)dz (3)
and

= {z C : Re z = Re = c}. Details about Mellin transformation, its


properties, Mellin pseudodierential operators can be found in [5],[6],[12]. For
example, M(t
d
dt
u) = z(Mu)(z), k N : M(t
k d
k
u
dt
k
) = (1)
k
(z)
k
(Mu)(z),
where (z)
k
= z(z + 1) . . . (z + k 1), i.e. M(t
2 d
2
dt
2
u) = z(z + 1)(Mu)(z).
Put

V (z, t) =

0

z1
V (, t)d. Then for the Cauchy problem (1) with V (s, T) =
f(s) one has:
On nonlinear Black-Scholes equations 77

V
t
= (r(z + 1)
1
2

2
(z
2
+ z))

V ,

V |
t=T
=

f(z) = (Mf)(z); (4)
(Mf)(z) =

z1
f()d.
The change = T t, 0 t T in the ODE (4) leads to:

= (z + 1)(r
1
2

2
z)

V ,

V |
=0
=

f(z). (5)
Consequently,

V (z, ) =

f(z)e
(z+1)(
1
2

2
zr)
.
Remark 1 The fundamental solution

E(z, ) of the ODO

+ (z + 1)(r
1/2
2
z) with z C being a parameter is:

E(z, ) = ()Z(z, ), where
dZ
d
=
(z + 1)(
1
2

2
z r), Z|
=0
= 1, i.e.

E(z, ) = ()e
(z+1)(
1
2

2
zr)

+ (z + 1)(r
1
2

2
z)

E = ().
According to the inverse formula (3) we have:
V (s, t) = M
1
zs

V =
(T t)
2i

c+i
ci
s
z

f(z)e
(Tt)(z+1)(r+
1
2

2
z)
= (6)
=
(T t)
2i

c+i
ci
s
z

f(z)

E(z, T t)dz.
Put f(x) g(x) =

0
= f()g(
x

)
d

. Then it is well known [12] that


M(f g)(z) = (Mf)(z).(Mg)(z) =

f(z) g(z) (7)
f(x) (g(x) = M
1
(

f g).
V (s, t) = (T t)M
1
zs
(

f

E) = (T t)f(s) E(s, t). (8)
Therefore, we are looking for a function E(s, t) and such that M
sz
E =

E(z, t) = (T t)e
(Tt)(z+1)(

2
2
zr)
, i.e.
E(s, t) = (T t)M
1
zs
(e
(Tt)(z+1)(

2
2
zr)
= (9)
=
(T t)
2i

c+i
ci
s
z
e
(Tt)(z+1)(

2
2
zr)
dz, s > 0.
78 R. Agliardi, P. Popivanov and A. Slavova
By using the fact that the underintegral function is analytic we apply Cauchy
integral theorem. Thus, 0 t T,
I =
1
2i

c+i
ci
e
zln s+(Tt)[

2
2
z
2
+(

2
2
r)zr]
dz,
z = c + iy, y (, +). In other words,
I =
1
2i

c+i
ci
e
az
2
+bz+p
dz,
where, dz = idy, 0 < a = (T t)

2
2
, b = (T t)(

2
2
r) ln s, p = r(T t).
Evidently,
az
2
+ bz + p = (

az +
b
2

a
)
2
+ p
b
2
4a
, z = c + iy.
Let d =
b
2a
and consider the straight line in C
1
: l
3
: z = d + iy. It is
obvious that (

az +
b
2

a
)
2
= ay
2
on l
3
. We apply Cauchy integral formula
to the analytic (entire) function e
az
2
+bz+p
in the rectangle l
1
: {z : z = c +iy},
l
2
: {z : z = x + iN, d x c}, l
3
, l
4
: {z : z = x iN, d x c} and then
let N . It is easy to understand that
I =
1
2

e
ay
2
dy.e
p
b
2
4a
=
1
2

a
e
p
b
2
4a
= (10)
1
2

2
T t
.
1

e
(Tt)r
[(Tt)(

2
2
r)ln s]
2
2
2
Tt)
.
The conclusion is that V (s, t) = (T t)f(s) E(s, t) and E(s, t) is given by
(9). If in the integral for the convolution f(s) E(s, t) we change variables
= e

, then we get the formula


V (s, t) =
e
r(Tt)

2(Tt)

f(e

)e

(ln s+(Tt)(

2
2
r))
2
2
2
(Tt)
d
for the solution for the Cauchy problem (1).
3 Solutions into explicit form for several non-
linear variants of the Black-Scholes equation
In this section we construct explicit solutions for the nonlinear modications
of Black-Scholes which are relevant in Finance. We consider the case of a zero
risk-free interest rate for simplicity of exposition. Consider the PDE
u
t
+ F(s
2
u
ss
,
1
(s)u
ss
) = 0, (11)
On nonlinear Black-Scholes equations 79
where
1
(s) = s
k+1
, k N
0
, F(, ) C
2
(R
2
). Changing to variables:

z = logs + at, a = const. = 0, i.e. e


z
= se
at
v = us
k1
, s > 0, t [0, T]
and look for a solution v depending on the new single variable z only: v(z) =
u(t, s)s
k1
. In other words, u(t, s) = s
1k
v(log s + at) is a solution of (11)
having that special form. Evidently, u
t
= av

(z)s
1k
, u
s
= v

(z)s
k
+ (1
k)s
k
v(z), u
ss
= s
k1
(k(k 1)v + (1 2k)v

+ v

). Substituting in (11) one


gets
av

s
1k
+F(s
k+1
(k(k 1)v +(12k)v

+v

), k(k 1)v +(12k)v

+v

) = 0
(12)
i.e.
v

+
1
a
s
k1
F(s
k+1
(k(k 1)v +(12k)v

+v

), k(k 1)v +(12k)v

+v

) = 0
(13)
This is the main assumption in the section:
F(, ) = F(, ), R
1
, > 0. (14)
Putting = s
1k
, = k(k 1)v +(1 2k)v

+v

we rewrite (13) in the form


v

+
1
a
F(k(k 1)v + (1 2k)v

+ v

, k(k 1)v + (1 2k)v

+ v

) = 0. (15)
In the case k = 0 we have
v

+
1
a
F(v

+ v

, v

+ v

) = 0 (16 0)
The standard substitution w = v

leads to a rst order autonomous ODE:


w + G(w + w

) = 0, where G() =
1
a
F(, ). Similarly we get
(k = 1) w + G(w

w) = 0 (16 1)
Assume now that k 2. Making the change v

(z) = p(v) v

=
dp
dv
p and
with v as a new independent variable we reduce the order of the ODE to 1,
namely
p + G(k(k 1)v + (1 2k)p + p
dp
dv
) = 0 (16 k)
This ODE with k 2 is a nonautonomous rst order ODE with unknown
p = p(v).
Examples: a) Let F(, ) = b, b = const > 0. Then (14) is satised and
with = s
2
u
ss
we obtain the classical linear Black-Scholes equation. Therefore
(16-1) is written as w +
b
a
(w

w) = 0, w = v

(z), z = log s + at etc.


80 R. Agliardi, P. Popivanov and A. Slavova
b) F(, ) =

(1)
2
, || < 1 (see [9]). Then evidently (14) holds.
c) F(, ) = (), () 0, R
1
or || d, d = const. > 0. Then
G() =
1
a
F(, ) =

a
() and the equation (16-k), k 2 takes the form:
p +
k(k 1)v + (1 2k)p + p
dp
dv
a
(k(k 1)v + (1 2k)p + p
dp
dv
) = 0;
p = p(v), 0.
Finally, we recall a comparison principle for the nonlinear Black-Scholes equa-
tion
u
t
+

F(s, u
ss
) = 0,
where F C
2
(R
2
). We refer to [1] for a proof in a more general setting.
Let p = (

t, s)
T
= {(t, s) : 0 < t < T, a < s < b}. The parabolic part of the
boundary will be denoted by
T
= I II III, where I = {0 < t < T, s = a},
II = {a < s < b, t = T}, III = {0 < t < T, s = b}; IV = {t = 0, a < s < b},
a < 0 < b.
Theorem 1 (Comparison principle for classical solutions in

T
). Suppose,
that

F
U
(s, U) 0 in

T
R and u, v are two classical solutions of u
t
+

F(s, u
ss
) = 0 such that u|

T
v|

T
. Then u v in
T
.
References
[1] R. Agliardi, P.Popivanov, A.Slavova, Nonhypoellipticity and comparison
principle for partial dierential equations of Black-Scholes type, Nonlinear
Analysis, Series B: Real World Applications, 12 (2011), 14291436
[2] F. Black, M. Scholes, The pricing of options and corporate liabilities, J.
Polit. Econ., 81 (1973), 637-659
[3] L. Bordag, A. Chmakova, Explicit solutions for a nonlinear model of -
nancial derivatives, Intern. J. Theoret. Appl. Finance, 10 (2007), 1-21
[4] L. Bordag, R. Frey, Nonlinear option pricing models for illiquid markets:
scaling properties and explicit solutions, L. Bordag, R. Frey, in: Nonlinear
Models in Mathematical Finance. New Research Trends in Option Pricing,
Nova Science Publisher, 2009, pp. 103130
On nonlinear Black-Scholes equations 81
[5] S. Colombo, J. Lavoine, Transformations de Laplace et de Mellin, for-
mulaires, mode dutilisation, Memorial des Sc.Mathem., 169 (1972),
Gauthier-Villars
[6] L. Debnath, D. Bhatta, Integral transforms and their applications, Chap-
man and Hall/CRC, 2007
[7] M. Ehrhardt, Nonlinear Models in Mathematical Finance, in: New Re-
search Trends in Option Pricing, Nova Science Publisher, Hauppauge,
New York, 2009
[8] R. Frey, P. Patie, Risk Management for Derivatives in Illiquid Markets:
A Simulation Study, Springer, Berlin, 2002
[9] R. Frey, A. Stremme, Market volatility and feedback eects from dynamic
hedging, Math. Finance, 4 (1997), 351-374
[10] L. Jodar, P. Sevilla-Peris, J.C. Cortes, R. Sala, A new direct method for
solving the Black-Scholes equation, Appl. Math. Letters, 18 (2005), 29-32
[11] M. Kratka, No mystery behind the smile, Risk, 9 (1998), 67-71
[12] B.W. Schulze, Pseudodierential Operators on Manifolds with Singulari-
ties, North Holland, Studies in Math. and its Applications, vol. 24, 1991
Received: January, 2013