Physics 347 Problem Set 2 due: Oct.

8 page 1
(Note: Renumbering of problems compared with hard copy handed out in class.)
6. PKT Problem 3.1 (p. 115)
7. PKT Problem 3.2 (p.115)
8. I will describe a game to you. Your job is to use a probability analysis to
calculate the probability of winning using one of two strategies:
You are a contestant on a game show. In this show there are three closed doors A,
B, and C. Behind one of them is a prize; behind the other two there is nothing. If
you pick the right door, you get the prize. You don’t know which door has the prize
behind it but the host does know. The host asks you to pick a door. You choose
door A. The host now opens one of the other doors, say, door B and shows you that
there is nothing behind it. The host then asks you “Would you like to change your
guess to C or would you like to stick with door A?”
(a) What should you do? What is your probability of winning if you switch? What
is your probability of winning if you stick? Whichever answer you give, EXPLAIN
your reasoning carefully.
Note: You will be tempted to answer that your probability of winning is the same
(1/3) with both strategies. This answer is wrong.
(b) What is your probability of winning if there are N(>3) doors?
9. This is a grab-bag of smaller problems:
(a) Testing the Stirling formula. Calculate 10! exactly and compare with the
Stirling approximation (without 1/n corrections). What is the absolute error? What
is the fractional error? Show that the fractional error is well approximated by the
1/12n correction, i.e., show that the error NOT accounted for by 1/12n is of order
1/n2.
(b) Consider a continuous probability distribution with the form of a one-sided
# x2
% " 2
Gaussian: p( x) = $Ce 2a , x > 0 . What is the normalization factor C? Calculate
%
& 0, x < 0
<x>, <x2>, and σ2 for this distribution.
(c) Certain molecular motors walk on cellular filaments. The walking has a fixed
step
! length a but it is not unidirectional. It has at each step a probability p of going
forward one step, a probability q of going backwards one step, and a probability r of
falling off the “track” (p+q+r=1). Calculate the mean distance <X>=a<M> traveled
along the track before falling off.
( p " q)
Hints: 1. The result turns out to be X = a .
r
2. In one way to do this problem, the result of problem 10 (a) is useful.

!
 page 2
10. In class, I analysed the random-walk/coin-toss problem for an unbiased coin
p+ = p" = 1/2 . In this problem, I want you to do the same analysis for a biased
coin p+ = p , p" = q , with p + q = 1. Consider a walk of N steps. Proceed as in
class:
! (a) Show by direct calculation that M = N ( p " q) ,, and " 2 # 4 Npq .
N!
! (b) !Show that the!discrete probability distribution is PN,M = p N+ q N"
N + !N " !
(c) Show that for N>>1 and ! M " M ~ N the!discrete probability distribution is
given approximately (but to and excellent approximation for N>>1) by the
2
! $
( M$ M )
1
(discrete) Gaussian!distribution, PN,M " e 8Npq with M as above.
2 #Npq
Hints: This is the hard part of the problem. You will need to show that
N " Np % " Nq %
e N ( ) with FN ( M ) = N + ln$
F M
PN,M = + N
' !( $ln ' . Then,
2 "N + N # ! # N + & # N (&
carefully find the maximum of FN(M) and expand around it. The maximum is NOT
generally at M=0.
! (d) Now, reinterpret the!problem as a random walk with step-length a and time-
step τ. Without taking any limits, show that the continuum probability distribution
( X#vt ) 2
1 #
can be written as P ( X,t ) = e 4Dt , where
4 "Dt
a ( p " q) 2 pqa 2
v= is a drift velocity and D = is a diffusion coefficient.
# "
This time-dependent
! probability distribution represents a particle starting at the
origin at time t=0 and subject to a combination of drift and diffusion.
(e) Consider a lattice model as done in class for the case without drift. Show that
! with the biased walk the ! discrete equation becomes
Pn ( t + " ) # Pn (t) = p( Pn#1 ( t ) # Pn (t )) + q(Pn+1 ( t ) # Pn ( t )) .
Now, take an appropriate continuous limit and show that
"P "P "2P
= #v +D , which is sometimes called the Smoluchowski equation (for a
! "t "X "X 2
linear potental).
(f) Show that the diffusive drift distribution derived in (d) satisfies the
Smoluchowski equation.
!