THE BLACK-SCHOLES MODEL AND EXTENSIONS

EVAN TURNER

Abstract. This paper will derive the Black-Scholes pricing model of a Euro-
pean option by calculating the expected value of the option. We will assume
that the stock price is log-normally distributed and that the universe is risk-
neutral. Then, using Ito’s Lemma, we will justify the use of the risk-neutral
rate in these initial calculations. Finally, we will prove put-call parity in order
to price European put options, and extend the concepts of the Black-Scholes
formula to value an option with pricing barriers.

Contents
1. Introduction 1
2. Derivation 2
3. Ito’s Lemma 6
4. Put-Call Parity 9
5. Barrier Options 10
Acknowledgments 14
References 14

1. Introduction
The Black-Scholes formula developed by Fischer Black and Myron Scholes in
1973 was revolutionary in its impact on the financial industry. Today, many of the
techniques and pricing models used in finance are rooted in the ideas and methods
presented by these two men. This paper will serve as an exposition of the formula
with extensions to more exotic options with barriers and will also highlight two
different methods for solving the options pricing problem.
We will first derive the formula by determining the expected value of the option,
a different method than the one originally employed by Black and Scholes. This
method, although it is somewhat less rigorous, gives the same result, namely that
the price of a European call option is given by
! 2
" ! 2
"
rT + ν 2T + ln SK0 −rT rT − ν 2T + ln SK0
C = S0 N √ − Ke N √ ,
ν T ν T
where S0 is the initial price of the stock, N (x) represents the cumulative distribution
function of a standard normal variable, r is the risk-free interest rate, K is the strike
price of the option, T is the amount of time until the option expires, and ν is the
annual volatility of the stock price.

Date: AUGUST 22, 2010.

1
2 EVAN TURNER

The derivation of this formula requires some non-intuitive assumptions. As a

result, we will define some basic terminology about risk, and then we will invoke
Ito’s Lemma to derive the Black-Scholes equation, named so because it was used
by Black and Scholes in their original derivation. The basic idea here is that, by
hedging away all risk in our portfolio, it becomes perfectly reasonable to assume that
people are risk-neutral. This is a very necessary step, though, for most people are
naturally risk-averse. This section will essentially follow the methods employed by
Black and Scholes and, along with the derivation for barrier options, will highlight
the basic method that they used and a different approach to the problem than that
of expected value.
We will introduce the concept of no-arbitrage, also known as the no-free-lunch
principle, in order to develop the idea of put-call parity. This method of solving
for the European put option price is much simpler than repeating the original
derivation and provides insight into basic ideas in financial mathematics.
Our last task will be to extend the basic principles of the Black-Scholes equation
(not the formula above) to price barrier options, which are options whose validities
are contingent upon hitting some pre-determined stock price. Intuitively, because
they have an extra imposition, barrier options should be worth less than a regu-
lar option. In order to price them, we will use the same technique employed by
Black and Scholes in which they transformed the Black-Scholes equation into the
heat equation. The key difference will be in the boundary conditions, a fact that
emphasizes the versatility of this technique in the pricing of more exotic options.

2. Derivation
We begin with a review of some basic terminology in probability theory.
Definition 2.1. The cumulative distribution function, F , of the random variable
X is defined for all real numbers b, by
F (b) = P{X ≤ b}
We say X admits a probability density function or density f if
# b
P{X ≤ b} = F (b) = f (x) dx
−∞
for some nonnegative function f .
Definition 2.2. X is a normal random variable with parameters µ and σ 2 > 0 if
the density of X is given by
1 −(x−µ)2
f (x) = √ e 2σ2 −∞<x<∞
2πσ
Thus, the cumulative distribution function of a standard normal random variable,
i.e. one with mean 0 and variance 1, is given by
# x
1 −y 2
N (x) = √ e 2 dy
2π −∞
Definition 2.3. If X is a continuous random variable having a probability density
function f (x) then the expected value of X is given by
# ∞
E[X] = xf (x) dx
−∞
 THE BLACK-SCHOLES MODEL AND EXTENSIONS 3

Note that the expected value will always be given by a bold-faced E, while a normal
E merely represents some variable or parameter.
Definition 2.4. The random variable X is log-normally distributed if for some
normally distributed variable Y , X = eY , that is, ln X is normally distributed.

Now that we have made a few basic definitions, we will delve into some ideas that
will be necessary for the derivation.
Definition 2.5. A call option is a contract between two parties in which the holder
of the option has the right (not the obligation) to buy an asset at a certain time in
the future for a specific price, called the strike price.
Definition 2.6. A put option is a contract between two parties in which the holder
of the option has the right (not the obligation) to sell an asset at a certain time in
the future for a specific price, also called the strike price.
A European option is a simply an option that can be exercised only at the expiry
of the option, which is specified in the contract.
Definition 2.7. A universe is a class that contains as its elements all the entities
that one wishes to consider for a given situation.
The concept of a universe allows us to specify and isolate certain conditions that
must hold for a theorem or idea to be true. In the universe in which we will be
dealing, we will assume that the risk-free interest rate, r, is always available. This
allows us to make reasonable simplifications in our argument that will help us reach
our final result.
Construction 2.8. The forward price of a stock is the current price of the stock,
S0 , plus an expected return which will exactly offset the cost of holding the stock
over a period of time t. Thus, as the only cost of holding the stock in our case is
the risk-free interest lost, the forward price is
S0 ert
where r is the risk-free interest rate.
Definition 2.9. A universe is risk-neutral if for all assets A and time periods t,
the value of the asset C(A, 0) at t = 0 is the expected value of the asset at time t
discounted to its present value using the risk-free rate.
C(A, 0) = e−rt E[C(A, t)]
where r is the continuously compounded risk-free interest rate.
Lemma 2.10. Let S0 be the initial value of the stock price, St be the price at time
t, and denote by ν annual volatility in the percent change in the stock price, i.e. the
standard deviation of the percent change in the price over one year. Finally, assume
St is a log-normally distributed random variable, i.e. ln SS0t is normally distributed
with mean µ and variance σ, and let the mean of the log-normal distribution be
located at the forward price of the stock. Then, µ = µ(t), σ = σ(t), and

(2.11) σ = ν2t

ν2
(2.12) µ = (r − )t
2
4 EVAN TURNER

It may not be initially clear why this is an important lemma, but it is actually
quite useful in the derivation of the Black-Scholes formula. It allows us to compute
the drift, µ, of the stock in terms of the risk-free interest rate, annual volatility,
and time and, thus, becomes very useful in simplifying complex expressions and
obtaining a nice result.

Proof. Equation 2.11 follows from induction. It is clear that after one year, ln SS10 has
variance (ν 2 )1. If we assume that after t − 1 years, ln SSt−1
0
has variance (ν 2 )(t − 1),
then after t years
St St−1 St
ln = ln
S0 S0 St−1
St−1 St
= ln + ln
S0 St−1
and has variance (ν 2 )(t − 1) + ν 2 = ν 2 t.

Equation 2.12 results from the following:

F (a) = P{St ≤ a}
= P{S0 ext ≤ a}
a
= P{xt ≤ ln }
S0
# ln Sa
1 0 −(xt −µ)2
= √ e 2σ dx
2σπ −∞
Differentiating with respect to a yields the density function for St , given by

−(ln x −µ)2
S0
f (x) = √ 1 e 2σ
2σπx

By assumption, E[St ] = S0 ert , so

# ∞ −(ln x −µ)2
1 S0
E[St ] = √ xe 2σ dx
0 2σπx
# ∞ −(ln x −µ)2
1 S0
= √ e 2σ dx
2σπ 0
x √
ln S0 −µ dx
Let z = √
σ
, then dz = √
x σ
with x = S0 ez σ+µ
, so that
# ∞ √
S −z 2
E[St ] = √0 e 2 ez σ+µ dz
2π −∞
# ∞ √
S −z 2
= √0 e 2 +z σ+µ dz
2π −∞
# ∞ √
S −(z− σ)2
+µ+ σ
= √0 e 2 2 dz
2π −∞
σ #
S0 eµ+ 2 ∞ −(z−√σ)2
= √ e 2 dz
2π −∞
 THE BLACK-SCHOLES MODEL AND EXTENSIONS 5

√
Letting x = z − σ, we see that
σ #
S0 eµ+ 2 ∞ −x2
(2.13) E[St ] √
= e 2 dx
2π −∞
σ
(2.14) = S0 eµ+ 2
√ ν2 t
except the last integral is equal to 2π. Thus, since σ = ν 2 t, S0 eµ+ 2 = S0 ert
2
and µ = (r − ν2 )t. !
Theorem 2.15. (Black-Scholes) In a risk-neutral universe with an initial stock
price S0 and a log-normally distributed stock price St , as in Lemma 2.6, at time t,
the value C of a European call option at time t = 0 with strike K, and expiration
time T , and r being the continuously compounded risk-free rate is

! " ! "
ν2T ν2T
rT + 2 + ln SK0 −rT rT − 2 + ln SK0
(2.16) C = S0 N √ − Ke N √
ν T ν T
where N is the cumulative distribution function of the standard normal variable.
Proof. We have C(S, T ) = max(ST − K, 0)
By assumption,
C(S, 0) = e−rT E[C(S, T )]
= e−rT E[max(ST − K, 0)]
# ∞ −(ln x −µ)2
−rT 1 S0
= e √ (x − K)e 2ν 2 T dx
K 2πT νx
# ∞ −(ln x −µ)2
# ∞ −(ln x −µ)2
1 S0 1 S0
= e−rT √ e 2ν 2 T dx − e−rT √ Ke 2ν 2 T dx
K 2πT ν K 2πT νx

One can see that the first integral is in fact the same one encountered in Lemma
2.10, hence the first term will simplify to
# ∞
−rT
2
µ+ ν 2T 1 −y2
e S0 e √ e 2 dy
A 2π
with
ln SK0 − µ − ν 2 T
A= √
ν T
To see this, note that this is the same integral as in equation 2.9 except for the
lower limit has been changed to A.
By using equation 2.13 for the value of µ and recognizing that this integral
represents the cumulative distribution function for the standard normal variable,
we see that this is in fact
! ! K 2 "" ! 2 "
ln S0 − rT − ν 2T ln SK0 − rT − ν 2T
S0 1 − N √ = S0 N − √
ν T ν T
! 2
"
ln SK0 + rT + ν 2T
= S0 N √
ν T
which gives the first term of equation 2.16.
6 EVAN TURNER

x
ln −µ
Now we examine the second term. Let z = νS√0 T , then dz = xνdx
√ and
T
# ∞ −(ln x −µ)2
# ∞
1 S 0 1 −z2
−e−rT √ Ke 2ν 2 T dx = −e−rT √ √ e 2 dz
K 2πT νx A+ν T 2π
! ! ""
−rT
√
= −e K 1−N A+ν T
! "
−rT
√
= −Ke N −A−ν T
! "
ν2T
−rT ln SK0 + rT − 2
= −Ke N √
ν T
which is the second term of equation 2.16 and completes the proof.
!

3. Ito’s Lemma
Some readers may have a problem with the above derivation. It certainly leads
to the correct result, but we have made several assumptions that are not necessarily
justified. First of all, it is not clear that a stock price should or will be log-normally
distributed. This presents a real problem for the formula, but it is not something
that we will deal with in this paper. Another problem in our assumptions is that
of risk-neutrality, which is clearly not true. Most people are not risk-neutral, i.e.
for some risky asset A and time period t, they will value the asset at C(A, 0) <
e−rt E[C(A, t)]. In a sense, they must be compensated for the risk that they are
bringing upon themselves. In this section, we will present the ideas behind this
assumption and attempt to justify why it is, in fact, perfectly acceptable to make.
Definition 3.1. A stochastic process, Wt , for t ≥ 0, is a Brownian Motion if
W0 = 0, and for all t and s, with s < t,
Wt − Ws
is continous, has a normal distribution with variance t − s, and the distribution of
Wt − Ws is independent of the behavior Wr for r ≤s.
Definition 3.2. The family X of random variables Xt satisfies the stochastic dif-
ferential equation (SDE),
(3.3) dXt = µ(t, Xt ) dt + σ(t, Xt ) dWt
if for any t,
Xt+h − Xt − hµ(t, Xt ) − σ(t, Xt )(Wt+h − Wt )
is a random variable with mean and variance which are o(h) and Wt is a Brownian
motion.
Definition 3.4. A stochastic process St is said to follow a Geometric Brownian
Motion if it satisfies the stochastic differential equation
(3.5) dSt = µSt dt + σSt dWt
with µ and σ constants and Wt a Brownian motion.
 THE BLACK-SCHOLES MODEL AND EXTENSIONS 7

Definition 3.6. An Ito Process, Xt , is a process that satisfies the stochastic dif-
ferential equation
dXt = µ(t)Xt dt + σ(t)Xt dWt

Theorem 3.7. (Ito’s Lemma) Let Xt be an Ito process satisfying equation 3.3, and
let f (x, t) be a twice-differentiable function; then f (Xt , t) is an Ito process, and
∂f ∂f 1 ∂2f
(3.8) d(f (Xt , t)) = (Xt , t) dt + dXt + dXt2
∂t ∂Xt 2 ∂Xt2
where dXt2 is defined by

(3.9) dt2 = 0

(3.10) dt dWt = 0

(3.11) dWt2 = dt

Remark 3.12. Proving this theorem is beyond our means at this time, but we
would like to say something about equations 3.9-3.11 since they will be essential in
deriving the Black-Scholes equation. Equations 3.9 and 3.10 might seem plausible
based on the fact that dt is infintessimal, i.e. it seems reasonable that dt2 = 0
and dt dWt = 0. However, equation 3.11 requires more justification. In order to
provide a non-rigorous justification, it will be useful to examine random walk on Z.
Imagine that a man takes a step of length 1 or -1 with equal probability at time t
for t > 0, where t is a natural number. Let Wt be the sum of the steps from time
t = 1 to t, then E[Wt ] = 0. Now consider Wt2 . Wt2 = t since each step now √ has
length 1 and there are√t steps. Thus, it seems reasonable that |Wt | = O( t). So
in ∆t steps, Wt = O( ∆t). As ∆t → 0, Wt2 ≈ ∆t, since the time increment and
steps have become arbitrarily small. Equation 3.11 follows. This is certainly not a
proof, but hopefully the reader will accept this for now.

Now let’s get back to risk-neutrality. The standard model for changes in stock prices
is geometric Brownian motion since stock prices are presumed to be log-normally
distributed. Thus, using Ito’s Lemma and equation 3.5, we will now justify the use
of the risk-free interest rate in the derivation.

Theorem 3.13. Given a European call option C(S, t), with expiry T and strike
price K, on a stock with price S that follows a geometric Brownian motion, and
with r being the continuously compounding risk-free interest rate, then
∂C ∂C 1 ∂2C
(3.14) (S, t) + rS + σ 2 S 2 2 (S, t) = rC
∂t ∂S 2 ∂S
This equation is known as the Black-Scholes equation (not to be confused with the
Black-Scholes formula derived earlier, although it was originally used to derive the
formula), and once shown, we will explain why this equation (and proof) justifies
the use of the risk-free rate.
8 EVAN TURNER

Proof. By equation 3.8,

∂C ∂C 1 ∂2C
dC = (S, t) dt + (S, t) dS + 2
(S, t) dS 2
∂t
! ∂S 2 ∂S "
∂C ∂C 1 2 2 ∂2C ∂C
= (S, t) + µS (S, t) + σ S 2
(S, t) dt + σS (S, t) dWt
∂t ∂S 2 ∂S ∂S

since
dS = µS dt + σS dWt
and
dS 2 = µ2 S 2 dt2 + µσS 2 dt dWt + σ 2 S 2 dWt2
= σ 2 S 2 dt
by equations 3.9, 3.10, and 3.11.
Now consider a portfolio consisting of the call option and α stocks. Then the
cost of the portfolio is C+αS. By the same argument as above, we see that
! "
∂C ∂C 1 2 2 ∂2C
d(C + αS) = (S, t) + µS (S, t) + σ S (S, t) + αµS dt
∂t ∂S 2 ∂S 2
! "
∂C
+ σS (S, t) + α dWt
∂S

Now we let α = − ∂C
∂S (S, t) to hedge away all risk in our portfolio.
! "
∂C 1 2 2 ∂2C
d(C + αS) = (S, t) + σ S (S, t) dt
∂t 2 ∂S 2
As one can see, the random component dWt , is now gone. The portfolio has no risk
or randomness. This is an important result, for, since it is risk-free, it
! must grow
"
d
over time at the risk-free rate r. Thus, dt (C + αS) = r(C + αS) = r C − S ∂C
∂S

and
! "
∂C ∂C 1 ∂2C
r C −S = (S, t) + σ 2 S 2 2 (S, t)
∂S ∂t 2 ∂S
Rearranging gives us the Black-Scholes equation:
∂C ∂C 1 ∂2C
(S, t) + rS + σ 2 S 2 2 (S, t) = rC
∂t ∂S 2 ∂S
!
By hedging away all randomness, we make sure that the portfolio has no risk, and
that allows us to use the assumption of risk-neutrality. If we did not do this, then it
seems very natural that the price of an option must take on the perceived risk with
which the investor views the stock. That being said, it is also unlikely that there
would be a unique price, for it is highly improbable that everyone would agree on
the risk level.
 THE BLACK-SCHOLES MODEL AND EXTENSIONS 9

4. Put-Call Parity
Definition 4.1. A portfolio is a collection of assets held by an institution or indi-
vidual.

Definition 4.2. A portfolio is said to be an arbitrage portfolio if today it has zero

value, and in the future it has positive value.

We will be dealing with the assumption of arbitrage-free pricing. Arbitrage

clearly exists in the real world. One example is simply different prices for different
assets in different markets. One might be able to buy something in one market
and sell it for a higher price in a different market and make a profit. However,
instances like these are usually eliminated quickly. People notice when there are
arbitrage opportunities, and they pounce on them, thus adjusting prices so that
these opportunities vanish.
Another instance of arbitrage can occur when dealing with two stocks. Suppose
stock A and B are worth the same at time t = 0, but at time T , A is worth twice
as much as before and B is still at its initial value. Then, one can create a portfolio
that is long A and short B. Clearly, the portfolio has positive value at time T but
zero value at time t = 0.
No-arbitrage means no free lunch, that a person can’t make a riskless profit
when he starts out with some portfolio with no value. Such an assumption makes
things simpler and provides for a certain amount of necessary order in the world of
financial mathematics. Without it, we would not be able to come up with a unique
price for options.

Example 4.3. Suppose 1 dollar exchanges for 5 yen, 1 pound exchanges for 10
yen, and 1 dollar exchanges for 1 pound. Clearly there is an arbitrage opportunity
here, for one could take a dollar, exchange it for a pound, exchange the pound for
10 yen, and exchange the yen for 2 dollars.

Theorem 4.4. If, in an arbitrage-free world, portfolios A and B are such that at
time T , A is worth at least as much as B, then at any time t < T , A will be worth
at least as much as B.

Proof. Let portfolio C consist of being long portfolio A and short portfolio B. Then
at time T, the value of C is A(T ) − B(T ) which is greater than or equal to 0. As
C is an arbitrage-free portfolio, the value of C at any time t < T is A(t) − B(t)
which also must be greater than or equal to 0 since, if it weren’t, there is clearly an
arbitrage opportunity, which is a contradiction. Thus, the value of A at any time
t < T must also be greater than or equal to that of B. !

Theorem 4.5. (Put-Call Parity) Let C(t) be the value of a European call option
on an asset S with strike price K and expiration T . Let P (t) be the value of a
European put option on the same asset S with the same strike price and expiration.
Finally, let S have a final value at expiration of ST , and let B(t, T ) represent the
value of a risk-free bond at time t with final value 1 at expiration time T . If these
assumptions hold and there is no arbitrage, then

(4.6) C(t) + KB(t, T ) = P (t) + St

10 EVAN TURNER

Proof. Consider first a portfolio X that consists of one put option and one share of
S. At time T , portfolio X has value
$
K, if St ≤ K as the option will be worth K − ST and the share ST
Xv =
ST , if ST ≥ K as the option will be worth 0 and the share ST .
Now consider a portfolio Y that consists of one call option and K bonds that
pay 1 at time T with certainty. Then, at time T , portfolio Y has value
$
K, if St ≤ K as the option will be worth 0 and the bonds K
Yv =
ST , if ST ≥ K as the option will be worth ST − K and the bonds K.
We can see that whatever value S takes at time T , portfolios X and Y have the
same value. Thus, from Theorem 4.3, at any time t < T , the portfolios must also
have the same value. It follows then that
C(t) + KB(t, T ) = P (t) + S(t)
!

Now, it is straightforward to obtain the price of a European put option with

strike price K and expiration time T . Recall from equation 2.16 that
! 2
" ! 2
"
rT + ν 2T + ln SK0 −rT rT − ν 2T + ln SK0
C(0) = S0 N √ − Ke N √
ν T ν T
Also note that, if we assume the interest rate r is constant, which we have implicitly
done in this paper, B(0, T ) = e−rT . And so,
P (0) = C(0) − S0 + Ke−rT
! 2
" ! 2
"
rT + ν 2T + ln SK0 −rT −rT rT − ν 2T + ln SK0
= S0 N √ − S0 + Ke − Ke N √
ν T ν T
! ! 2
"" ! ! 2
""
rT + ν 2T + ln SK0 −rT rT − ν 2T + ln SK0
= −S0 1 − N √ + Ke 1−N √
ν T ν T
! 2
" ! 2
"
rT + ν 2T + ln SK0 −rT rT − ν 2T + ln SK0
= −S0 N − √ + Ke N − √
ν T ν T
And we are done.

5. Barrier Options
Now we will examine a specific type of option called a barrier option. Unlike
a normal vanilla option, the barrier option is contingent upon hitting some stock
price, called the barrier, at any time before its expiry. There are two types of barrier
options, knock-out options and knock-in options. A knock-out option becomes
worthless if at any time before the expiry, the stock price reaches the barrier, while
a knock-in option only provides a pay-off once the stock price crosses the barrier.
In this paper, we will address down-and-out call options and, as a consequence,
down-and-in call options. There is a very convenient relationship between the two
that will be apparent once the pay-offs of the two are defined.
 THE BLACK-SCHOLES MODEL AND EXTENSIONS 11

Definition 5.1. The payoff of a down-and-out call option with strike price K and
barrier at B is
max(ST − K, 0)
unless at any time t < T , St passes below B.
One thing that should be pointed out is that it makes no sense for the barrier
to be above the initial stock price. If this were the case, the option is immediately
worthless. Thus, the option is called “down-and-out” because if it goes down past
the barrier, the option “knocks out” and becomes worthless. Even if the final price
of the stock, ST , were well above K, the option’s pay-off is 0 if the stock price ever
hits the barrier. If it does not reach the barrier at any time before the expiry, then
the payoff is identical to a normal European call option.
Definition 5.2. The payoff of a down-and-in call option with strike price K and
barrier at B is
max(ST − K, 0)
if and only if at some time t < T , St passes below B.
St must pass below the barrier for some t < T for the option to pay-off anything.
As with the down-and-out call option, however, hitting the barrier in no way guar-
antees a positive payoff, since ST must still be above K. Otherwise, the payoff is
0.
Because there is an extra condition on the barrier option, intuitively it seems
that it must be worth less than a normal European option. In fact, one can see
that, in a portfolio with exactly one down-and-out call option and one down-and-in
option with the same strike price and barrier and for the same stock, exactly one
of the barrier options will pay max(ST − K, 0), which is the payoff off of a normal
European call option. Thus, there is a very nice relationship between these two
barrier options, namely that
(5.3) CI + CO = Cv
where CI stands for the price of the down-and-in call option, CO for the price of the
down-and-out call option, and Cv for the price of the normal European call option.
We will now derive the price of the down-and-out call option and, afterward, it will
be very simple to determine the price of the down-and-in call option.
Our first goal in pricing barrier options is to reduce the Black-Scholes equation
to the heat equation, along with the extra condition that if the barrier is reached at
any time during the option’s life, the option becomes worthless. The heat equation
is a well-known equation in physics and mathematics, and there are several ways
to solve it. We will use a slight trick that presupposes that a solution exists, but
other methods can be seen in [5].
The barrier option is essentially a normal option with an extra constraint. Thus,
it satisfies the Black-Scholes equation (3.13) with the additional condition that
C(B, t) = 0
where B is the barrier of the option.

∂C ∂C 1 ∂2C
(S, t) + rS + σ 2 S 2 2 (S, t) = rC
∂t ∂S 2 ∂S
C(S, T ) = max(S − K, 0)
12 EVAN TURNER

C(B, t) = 0
Now we will use several change of variables in order to reduce this to the heat
equation. Let S = Bex , then
∂C ∂C ∂S
=
∂x ∂S ∂x
∂C
= S
∂S
and
∂2C ∂ ∂C
2
=
∂x ∂x !
∂x "
∂ ∂C
= S
∂x ∂S
! "
∂S ∂C ∂ ∂C
= +S
∂x ∂S ∂x ∂S
∂C ∂2C
= S +S
∂S ∂S∂x
∂C ∂2C
= S + S2
∂S ∂S
since
∂C ∂C
=S
∂x ∂S
Now we arrive at,
∂C 1 ∂C 1 ∂2C
+ (r − σ 2 ) + σ 2 2 = rC
∂t 2 ∂x 2 ∂x
And the boundary conditions become
C(x, T ) = max(Bex − K, 0) and C(0, t) = 0
Now let t = T − τ / 21 σ 2 . Then,
1 ∂C 1 ∂C 1 ∂2C ∂t 1
− σ2 + (r − σ 2 ) + σ 2 2 = rC since = −1/ σ 2
2 ∂τ 2 ∂x 2 ∂x ∂τ 2
and the boundary conditions become
C(x, 0) = max(Bex − K, 0) and C(0, τ ) = 0
Finally, let C = Beαx+βτ u(x, τ ). Then,
! " ! " ! "
1 2 ∂u 1 2 ∂u 1 2 2 ∂u ∂ 2 u
− σ βu + + (r − σ ) αu + + σ α u + 2α + = ru
2 ∂τ 2 ∂x 2 ∂x ∂x2
We want to eliminate the u term and the ∂u ∂x term so we can set up 2 equations to
find values of α and β such that this will occur.
1 1 1
− σ 2 β + (r − σ 2 )α + σ 2 α2 − r = 0
2 2 2
1 2
r − σ + σ2 α = 0
2
So
r 1 r2 1 r
α=− 2 + and β=− 4 − − 2
σ 2 σ 4 σ
 THE BLACK-SCHOLES MODEL AND EXTENSIONS 13

With these values for α and β, we arrive at the heat equation

∂u ∂2u
=
∂τ ∂x2
for 0 < x < ∞, τ > 0 and with boundary conditions
K −αx
u(x, 0) = U (x) = max(ex−αx − e , 0), x>0
B
and
u(0, τ ) = 0
This problem has now become that of the flow of heat in an infinite bar. However,
there is a slightly new condition in u(0, τ ) = 0. This essentially says that at any
time t, the temperature is 0 at x = 0. This can be modelled by creating a two-bar
semi-infinite problem with one side hot and the other cold, so that the heat flow
exactly cancels out and the temperature is 0 at x = 0.
The heat equation is invariant under reflection so u(x, τ ) and u(−x, τ ) are both
solutions for it. Now, we will solve for all x, instead of just x > 0. Now,
$
U (x), x>0
u(x, 0) =
−U (−x), x < 0
or $
max(ex−αx − KBe
−αx
, 0), x>0
u(x, 0) = αx−x K αx
− max(e − B e , 0), x<0
This guarantees that u(0, τ ) = 0. Rather than integrating this, there is a slight
trick that we can use to find the price of the down-and-out option. (For a more
traditional approach, see [5].) Consider a normal European call option with the
same expiry and exercise price but no barrier. Let its value be Cv (S, T ; K) and
let Uv (x, τ ) be the corresponding solution for the heat equation. When S < K,
Cv (S, T ; K) = 0. Thus, since S = Bex , when x < ln K B then Uv (x, τ ) = 0. As
we assumed that the strike price K was higher than the barrier B, then ln K B > 0.
Thus, if we set U (x) = 0 for x < 0, then we extend U (x) for all x and U (x) is now
equal to Uv (x). We can now write
u(x, 0) = Uv (x) − Uv (−x)
and this holds for all x. Thus,
u(x, τ ) = Uv (x, τ ) − Uv (−x, τ )
since both sides of this equation satisfy all conditions and both solve the heat
equation. By uniqueness of the solution, they must be equivalent.
Cv (S, t; K) = Cv (Bex , t(τ ); K) = Beαx+βτ Uv (x, τ )
shows that
Uv (x, τ ) = e−αx−βτ Cv (Bex , t(τ ); K)/B
and
Uv (−x, τ ) = eαx−βτ Cv (Be−x , t(τ ); K)/B
14 EVAN TURNER

Thus, the value of the down-and-out call option is

CO (S, t) = Beαx+βτ u(x, τ )
= Beαx+βτ (Uv (x, τ ) − Uv (−x, τ ))
= Beαx+βτ (e−αx−βτ Cv (Bex , t(τ ); K)/B − eαx−βτ Cv (Be−x , t(τ ); K)/B)
= Cv (Bex , t(τ ); K) − e2αx Cv (Be−x , t(τ ); K)
! "2α
S
= Cv (S, t; K) − Cv (B 2 /S, t; K)
B
or
! " ! "
ν2T ν2T
rT + 2 + ln SK0 −rT rT − 2 + ln SK0
C O = S0 N √ − Ke N √
ν T ν T
! −2
" ! 2 2 "
S0 −2rσ rT + ν 2T + ln SB0 K
−B N √
B ν T
! −2
" ! ν2T B2
"
S0 1−2rσ rT − 2 + ln S K
− Ke−rT N √ 0

B ν T
And by equation 5.3, it is clear that the price of the down-and-in call option is
! −2
" ! 2 2 "
S0 −2rσ rT + ν 2T + ln SB0 K
CI = B N √
B ν T
! −2
" ! 2 2 "
S0 1−2rσ −rT
rT − ν 2T + ln SB0 K
+ Ke N √
B ν T
Pricing for other types of barrier options is done similarly. For further study, see
[4].
Acknowledgments. It is a pleasure to thank my mentors, Zachary Madden and
Yuan Shao, for their guidance through the paper-writing process. I really enjoyed
working with them and greatly appreciate the work they did in answering any and
all of my questions about math and any other topic that might have come up. I
especially want to thank Roger Lee for providing advice on what direction to take
with my paper, as he was certainly not obligated to do this. Finally, I just want
to thank the REU program in general for allowing me to work on a topic that I
probably wouldn’t have gotten to experience otherwise.

References
[1] M.S. Joshi. The Concepts and Practice of Mathematical Finance. Cambridge University Press.
2003.
[2] Sheldon Ross. A First Course in Probability. Macmillan Publishing Co., Inc. 1976.
[3] Sheldon Ross. An Introduction to Mathematical Finance. Cambridge University Press. 1999.
[4] Sam Howison. Barrier Options. http://people.maths.ox.ac.uk/howison/barriers.pdf.
[5] Mark Gockenbach. Partial Differential Equations: Analytical and Numerical Methods. Society
for Industrial and Applied Mathematics. 2002.
[6] Bernt Oksendal. Stochastic Differential Equations: An Introduction with Applications.
Springer, 5th edition.