You are on page 1of 8

SERIES  ARTICLE

Snippets of Physics
20. Random Walk Through Random Walks – II

T Padmanabhan

W e c o n tin u e o u r e x p lo ra tio n o f ra n d o m w a lk s
w ith so m e m o r e c u r io u s r e su lts. W e d isc u ss th e
d im e n sio n d e p e n d e n c e o f so m e o f th e fe a tu r e s o f
th e ra n d o m w a lk , d e sc r ib e a n u n e x p e c te d c o n -
n e c tio n b e tw e e n ra n d o m w a lk s a n d e le c tric a l n e t-
w o r k s a n d ¯ n a lly d isc u ss so m e re m a r k a b le fe a -
T Padmanabhan works at
tu re s o f r a n d o m w a lk w ith g e o m e tr ic a lly d e c re a s- IUCAA, Pune and is
in g ste p -le n g th . interested in all areas
of theoretical physics,
In th e la st in sta lm en t, w e lo o k ed a t sev era l elem en ta ry especially those which
fea tu res o f ra n d o m w a lk a n d , in p a rticu la r, o b ta in ed a have something to do with
g en era l fo rm u la fo r th e p ro b a b ility P N (x ), fo r th e p a rti- gravity.

cle to b e fo u n d a t p o sitio n x a fter N step s. T h is resu lt,


in th e ca se o f ra n d o m w a lk in a cu b ic la ttice ca n b e
w ritten a s th e in teg ra l
Z¼ Ã D
!N
dD k 1 X
P N (x ) = [co s(k ¢ x )] co s k j : (1 )
¡¼ (2 ¼ )D D
j= 1

T h is resu lt { fo r a n a rb itra ry d im en sio n D { m ig h t d e-


ciev e y o u in to b eliev in g th a t th e b eh av io u r o f ra n d o m
w a lk in , say, D = 1 ;2 ;3 is a ll essen tia lly th e sa m e. In
fa ct th ey a re n o t, a s ca n b e illu stra ted b y stu d y in g th e
p h en o m en o n k n ow n a s recu rren ce.
R ecu rren ce refers to th e p ro b a b ility fo r th e ra n d o m w a lk -
in g p a rtcle to co m e b a ck to th e o rig in , w h ere it sta rted
fro m , in th e co u rse o f its p era m b u la tio n , w h en w e w a it
fo rev er. L et u n d en o te th e p ro b a b ility th a t a p a rticle
retu rn s to th e o rig in o n th e n th step a n d let R b e th e Keywords
Random walk, decreasing steps,
ex p ected n u m b er o f tim es it retu rn s to th e o rig in . recurrence, Watson integrals,
grid of resistors.

RESONANCE  August 2009 799


SERIES  ARTICLE

C lea rly,
X1
R = un : (2 )
n= 0

W e ca n n ow d istin g u ish b etw een tw o d i® eren t scen a rio s. If th e series in (2 )


d iv erg es, th en th e m ea n n u m b er o f retu rn s to th e o rig in is in ¯ n ite a n d w e say
th a t th e ra n d o m w a lk is recu rren t. If th e series is co n v erg en t, lea d in g to a ¯ n ite
R , th en w e say th a t th e ra n d o m w a lk is tra n sien t.
T h is id ea ca n b e rein fo rced b y th e fo llow in g a ltern a tiv e in terp reta tio n o f R . S u p -
p o se u is th e p ro b a b ility fo r th e p a rticle to retu rn to th e o rig in . T h en th e n o rm a l-
ized p ro b a b ility fo r it to retu rn ex a ctly k tim es is u k (1 ¡ u ). T h e m ea n n u m b er
o f retu rn s to th e o rig in is, th erefo re,
X1
R = k u k ¡ 1 (1 ¡ u ) = (1 ¡ u )¡1 : (3 )
k= 1

O b v io u sly, if R = 1 , th en u = 1 , sh ow in g th a t th e p a rticle w ill d e¯ n itely retu rn


to th e o rig in . B u t if R < 1 , th en u < 1 a n d w e ca n 't b e certa in th a t th e p a rticle
w ill ev er co m e b a ck h o m e.
L et u s co m p u te u n a n d R fo r ra n d o m w a lk s in D = 1 ;2 ;3 d im en sio n s w ith th e
la ttice sp a cin g set to u n ity fo r sim p licity. F ro m (1 ), settin g x = 0 , w e h av e:
Z¼ Ã D
!n
dD k 1 X
u n (x ) = D
co s k j : (4 )
¡ ¼ (2 ¼ ) D j= 1

D o in g th e su m in (2 ) w e g et
Z¼ Ã !¡ 1
X1 dD k 1 X
D
R = un = 1¡ co s k j : (5 )
n= 0 ¡¼ (2 ¼ )D D j= 1

W e w a n t to k n ow w h eth er th is in teg ra l is ¯ n ite o r d iv erg en t. C lea rly, th e d iv er-


g en ce, if a n y, ca n o n ly a rise d u e to its b eh av io u r n ea r th e o rig in in k -sp a ce. U sin g

800 RESONANCE  August 2009


SERIES  ARTICLE

th e T ay lo r series ex p a n sio n o f th e co sin e fu n ctio n , w e see th a t, n ea r th e o rig in w e


h av e th e b eh av io u r:
Z Z
d k 1 d k 2 :::d k D ¡ 2 2
¢
2 ¡1 2D k D ¡1 d k
R ¼ 2D k 1 + k 2 + ¢¢¢ + k D / : (6 )
k ¼0 (2 ¼ )D (2 ¼ )D k ¼ 0 k 2

T h e d im en sio n d ep en d en ce is n ow o b v io u s. In D = 1 ;2 th e in teg ra l is d iv erg en t


a n d R = 1 ; so w e co n clu d e th a t th e ra n d o m w a lk in D = 1 ;2 is recu rren t a n d
th e p a rticle w ill d e¯ n itely retu rn to th e o rig in if it w a lk s fo rev er. B u t in D = 3 ,
th e R is ¯ n ite a n d th e w a lk is n o n -recu rren t. T h ere is a ¯ n ite p ro b a b iltity th a t
th e p a rticle w ill co m e b a ck to th e o rig in b u t th ere is a lso a ¯ n ite p ro b a b ility th a t
it w ill n o t. A d ru n k en m a n w ill d e¯ n itely co m e h o m e (g iv en en o u g h tim e) b u t a
d ru n k en b ird o n ° ig h t m ay o r m ay n o t!
T h e m ea n n u m b er o f recu rren ces in D = 3 is g iv en b y th e W a tso n in teg ra l
Z¼ Z¼ Z¼
3
R = 3
dk1 dk2 d k 3 [1 ¡ (co s k 1 + co s k 2 + co s k 3 )]¡1 ; (7 )
(2 ¼ ) ¡ ¼ ¡¼ ¡¼

w h ich is n o to rio u sly d i± cu lt to eva lu a te a n a ly tica lly [1 ]. S in ce th e a n sw er h a p -


p en s to b e
p µ ¶ µ ¶ µ ¶ µ ¶
6 1 5 7 11
R = 3
¡ ¡ ¡ ¡ ; (8 )
32¼ 24 24 24 24

y o u a n y w ay n eed to lo o k it u p in a ta b le so o n e m ig h t a s w ell d o th e in teg ra l


n u m erica lly w h ich is triv ia l in M athem atica. (O f co u rse, if y o u lik e th e ch a llen g e
o f a d e¯ n ite in teg ra l, try it o u t. I d o n o t k n ow o f a sim p le w ay o f d o in g it; n eith er
d o th e ex p erts in ra n d o m w a lk I h av e ta lk ed to !) T h e resu lt is R ¼ 1 :5 1 6 4 g iv in g
th e retu rn p ro b a b ility u ¼ 0 :3 4 0 5 .
In th e ca se o f D = 1 ;2 it is a lso ea sy to o b ta in u n ex p licitly b y co m b in a to rics.
In 1 -d im en sio n , th e p a rticle ca n retu rn to th e o rig in o n ly if it h a s ta k en a n ev en
n u m b er o f step s, h a lf to th e rig h t a n d h a lf to th e left. T h e p ro b a b ility fo r th is is
clea rly

2n 1
u 2n = C n : (9 )
22n

RESONANCE  August 2009 801


SERIES  ARTICLE

Fp o r su ± cien tly la rg e n , w e capn u se S terlin g 's a p p rox im a tio n fo r fa cto ria ls (n ! ¼


2 ¼ n e ¡ n n n ) to g et u 2 n ¼ 1 = ¼ n . T h e series in (2 ) in v o lv es th e a sy m p to tic su m
w h ich is d iv erg en t:
X X 1
m = u 2n ¼ p = 1 : (1 0 )
n n
¼n

O b v io u sly, th e 1 -d im en sio n a l ra n d o m w a lk is recu rren t.


In terestin g ly en o u g h th e resu lt fo r D = 2 is ju st th e sq u a re o f th e resu lt fo r
D = 1 . T h e in teg ra l in (4 ) b eco m es fo r D = 2 :
Z¼ Z¼
1 1
u n (x ) = dk1 d k 2 (co s k 1 + co s k 2 )n : (1 1 )
(2 ¼ )D 2 n ¡¼ ¡¼

If y o u n ow ch a n g e va ria b les o f in teg ra tio n to (k 1 + k 2 ) a n d (k 1 ¡ k 2 ) it is ea sy to


sh ow th a t th is in teg ra l b eco m es th e p ro d u ct o f tw o in teg ra ls g iv in g
· ¸2
1 2n
u 2n = Cn ; (1 2 )
22n

w h ich is th e sq u a re o f th e resu lt fo r D = 1 . N ow th e series in (2 ) w ill b e d o m in a ted


a sy m p to tica lly b y
X 1
m ¼ = 1 ; (1 3 )
n
¼n

m a k in g th e D = 2 ra n d o m w a lk recu rren t a g a in . Y o u m ig h t g u ess a t th is sta g e,


th a t in 3 -D , th e a sy m p to tic series w ill in v o lv e su m ov er n ¡3 = 2 (a n d h en ce w ill
co n v erg e) m a k in g th e 3 -D ra n d o m w a lk n o n -recu rren t. T h is is p a rtia lly tru e a n d
th e 3 -d im en sio n a l series is b o u n d ed fro m a b ov e b y th e su m ov er n ¡3 = 2 . B u t th e
3 -d im en sio n a l ca se is n ot th e p ro d u ct o f th ree 1 -d im en sio n a l ca ses.
W e n ow tu rn o u r a tten tio n to a n o th er cu rio u s resu lt. S u m m in g P N (x ) ov er a ll
N o n e ca n co n stru ct th e q u a n tity P (x ) w h ich is th e p ro b a b ility o f rea ch in g x .
U sin g (1 ) a n d d o in g th e g eo m etric su m , w e ¯ n d in D = 2 , th is q u a n tity to b e:
Z¼ Z¼ µ ¶¡1
d k 1d k 2 1
P (x ) = 2
[co s(k ¢ x )] 1 ¡ (co s k 1 + co s k 2 ) : (1 4 )
¡¼ ¡¼ (2 ¼ ) 2

802 RESONANCE  August 2009


SERIES  ARTICLE

C o n sid er n o w th e ex p ressio n
Z¼ Z¼
1 dk1dk2 [1 ¡ co s(k ¢ x )]
R = (P (x ) ¡ P (0 )) = : (1 5 )
2 ¡¼ ¡¼ 8¼ 2 [1 ¡ 12 (co s k 1 + co s k 2 ]

In cred ib ly en o u g h th is p rov id es th e so lu tio n to a co m p letely d i® eren t p ro b lem ! It


g iv es th e e® ectiv e resista n ce b etw een a la ttice p o in t x a n d th e o rig in in a n in ¯ n ite
g rid o f 1 -o h m resisto rs co n n ected b etw een th e la ttice sites. L et u s see h ow th is
co m es a b o u t b y a n a ly sin g th e g rid o f resisto rs.
L et a n o d e x in th e in ¯ n ite p la n a r sq u a re la ttice b e d en o ted b y tw o in teg ers
(m ;n ) a n d let a cu rren t I m ;n b e in jected a t th a t n o d e. T h e ° ow o f cu rren t w ill
in d u ce a v o lta g e a t ea ch n o d e a n d , co m b in in g K irch o ® 's a n d O h m 's law s fo r th e
1 -o h m resisto rs, w e ca n w rite th e rela tio n :
Im ;n = (V m ;n ¡ Vm + 1 ;n ) + (V m ;n ¡ Vm ¡1 ;n ) + (V m ;n ¡ Vm ;n + 1 ) + (V m ;n ¡ Vm ;n ¡1 )

= 4V m ;n ¡ Vm + 1 ;n ¡ Vm ¡1 ;n ¡ Vm ;n + 1 ¡ Vm ;n ¡1 ; (1 6 )

w h ere V m ;n is th e p o ten tia l a t th e n o d e (m ;n ) d u e to th e cu rren t. T h is eq u a tio n


ca n a g a in b e so lv ed b y in tro d u cin g th e F o u rier tra n sfo rm o n th e d iscrete la ttice.
If w e w rite
Z¼Z¼
1
I m ;n = 2
d k 1 d k 2 I (k 1 ;k 2 )e i(m k 1 + n k 2 ) ; (1 7 )
4 ¼ ¡¼ ¡ ¼
Z¼Z¼
1
V m ;n = 2
d k 1 d k 2 V (k 1 ;k 2 )e i(m k 1 + n k 2 ) ; (1 8 )
4 ¼ ¡¼ ¡ ¼

th en o n e ca n o b ta in fro m (1 6 ) th e resu lt in th e F o u rier sp a ce:


I (k 1 ;k 2 ) = 2 V (k 1 ;k 2 ) [2 ¡ co s(k 1 ) ¡ co s(k 2 )] : (1 9 )

S u p p o se w e n ow in ject a cu rren t o f 1 a m p at (0 ,0 ) a n d ¡ 1 a m p a t (N ;M ). T h en
I m ;n = ± m ;n ¡ ± m ¡ M ;n ¡ N , lea d in g to
I (k 1 ;k 2 ) = 1 ¡ e ¡i(M k1+ N k2)
; (2 0 )

so th a t (1 9 ) g iv es th e v o lta g e to b e
1 1 ¡ e ¡i(M k 1 + N k 2 )
V (k 1 ;k 2 ) = £ : (2 1 )
2 2 ¡ co s(k 1 ) ¡ co s(k 2 )

RESONANCE  August 2009 803


SERIES  ARTICLE

T h e eq u iva len t resista n ce b etw een n o d es (0 ,0 ) a n d (M ,N ) w ith a ° ow o f u n it


cu rren t is ju st th e v o lta g e d i® eren ce b etw een th e n o d es:
R M ;N = V 0 ;0 ¡ V M ;N
Z¼ Z¼
1 £ i(M k 1 + N k 2 )
¤
= d k 1 d k 2 V (k 1 ;k 2 ) 1 ¡ e
4 ¼ 2 ¡¼ ¡¼
Z¼ Z¼
1 1 (1 ¡ e ¡ i(M k 1 + N k 2 ) )(1 ¡ e i(M k 1 + N k2)
)
= d k 1 d k 2
4 ¼ 2 ¡ ¼ ¡¼ 2 2 ¡ co s(k 1 ) ¡ co s(k 2 )
Z¼ Z¼
1 1 ¡ co s(M k 1 + N k 2 )
= 2
dk1 dk2 ; (2 2 )
4 ¼ ¡ ¼ ¡¼ 2 ¡ co s(k 1 ) ¡ co s(k 2 )

w h ich is ex a ctly th e sa m e a s th e in teg ra l in (1 5 )!


T h e in ¯ n ite g rid o f sq u a re la ttice resisto rs is a cla ssic p ro b lem a n d th e e® ectiv e
resista n ce b etw een tw o a d ja cen t n o d es is a `trick q u estio n ' th a t is a fav o u rite o f
ex a m in ers. T h e a n sw er (0 .5 o h m ) ca n b e fo u n d b y triv ia l su p erp o sitio n b u t su ch
trick s a re u seless to ¯ n d th e e® ectiv e resista n ce b etw een a rb itra ry n o d es. In fa ct,
th e e® ectiv e resista n ce b etw een tw o d ia g o n a l n o d es o f th e b a sic sq u a re { th e (0 ,0
a n d (1 ,1 ), say { is g iv en b y th e in teg ra l
Z¼ Z¼
1 1 ¡ co s(k 1 + k 2 )
R 1 ;1 = 2
dk1 dk2 : (2 3 )
4 ¼ ¡¼ ¡¼ 2 ¡ co s(k 1 ) ¡ co s(k 2 )

T h is is d o a b le, b u t n o t ex a ctly ea sy, a n d th e a n sw er is 2 = ¼ . (N ex t tim e so m eo n e


lectu res y o u a b o u t th e p ow er o f clev er a rg u m en ts, a sk h er to u se th em to g et th is
a n sw er, w h ich h a s a ¼ in it!)
B u t w h y d o es th is w o rk ? W h a t h a s ra n d o m w a lk o n a la ttice to d o w ith resisto r
n etw o rk s? T h ere a re d i® eren t lev els o f so p h istica tio n a t w h ich o n e ca n a n sw er th is
q u estio n a n d a n en tire b o o k [2 ] d ea lin g w ith th is su b ject ex ists. T h e m a th em a tica l
rea so n h a s to d o w ith th e fa ct th a t b o th th e ra n d o m w a lk p ro b a b ility to v isit a
n o d e a n d th e v o lta g e o n a n o d e (w h ich d o es n o t h a v e a n y cu rren t in jected o r
rem ov ed ) a re h a rm o n ic fu n ctio n s. T h ese a re fu n ctio n s w h o se va lu e a t a n y g iv en
n o d e is g iv en b y th e av era g e o f th e va lu e o f th e fu n ctio n o n th e a d ja cen t la ttice
sites. T h is is o b v io u s in th e ca se o f ra n d o m w a lk b eca u se a p a rticle w h ich rea ch es
th e n o d e (m ;n ) m u st h av e h o p p ed to th a t n o d e w ith eq u a l p ro b a b ility fro m o n e
o f th e n eig h b o u rin g n o d es (m § 1 ;n § 1 ). In th e ca se o f a resisto r n etw o rk , th e
sa m e resu lt is o b ta in ed fro m (1 6 ) w h en I m n = 0 . If y o u n ow in ject th e v o lta g es 1

804 RESONANCE  August 2009


SERIES  ARTICLE

a n d 0 a t tw o sp eci¯ c n o d es A a n d B , th en th e v o lta g e a t a n y o th er n o d e X ca n
b e in terp reted a s th e p ro b a b ility th a t a ra n d o m w a lk er sta rtin g a t X w ill g et to
A b efo re B . O n e ca n th en u se th is in terp reta tio n to m a k e a fo rm a l co n n ectio n
b etw een v o lta g e d istrib u tio n in a n electric n etw o rk a n d a ra n d o m w a lk p ro b lem .
T h e in terested rea d er ca n ¯ n d m o re in th e b o o k [2 ] referred to a b ov e.
H av in g d o n e a ll th ese in th e la ttice, w e n ow g o b a ck to th e ra n d o m w a lk in
fu ll sp a ce fo r w h ich w e h a d o b ta in ed th e resu lt in th e la st in sta llm en t, w h ich ,
sp ecia lised to o n e d im en sio n , is g iv en b y :
Z1 N
d k ik x Y
P N (x ) = e p n (k ) : (2 4 )
¡1 (2 ¼ ) n= 1

W e w a n t to co n sid er a situ a tio n in w h ich th e step s a re ra n d o m a n d u n co rrela ted


b u t th eir len g th s a re d ecrea sin g m o n o to n ica lly. (T h is is w h a t w ill h a p p en if th e
d ru n ka rd g ets tired a s h e w a lk s!) In p a rticu la r, w e w ill a ssu m e th a t ea ch step
len g th is a fra ctio n ¸ o f th e p rev io u s o n e w ith ¸ < 1 a n d th e ¯ rst step is o f u n it
len g th . It is clea r th a t P N (x ) is n ow g iv en b y
Z1 N
d k ik x Y
P N (x ) = e co s(k ¸ n ) : (2 5 )
¡1 (2 ¼ ) n= 1

W e ca n n ow stu d y th e lim it o f N ! 1 a n d a sk h ow th e p ro b a b ility P (x ) ´ P 1 (x )


is d istrib u ted . T h is fu ctio n sh ow s in cred ib ly d iv erse p ro p erties d ep en d in g o n th e
va lu e o f ¸ a n d is still n o t co m p letely a n a ly sed . I w ill co n ¯ n e m y self to a sim p le
situ a tio n , referrin g y o u to th e litera tu re if y o u a re in terested [3 ].
L et u s ¯ rst co n sid er th e ca se w h en ¸ = 1 = 2 . In th is ca se th e releva n t in ¯ n ite
p ro d u ct is g iv en b y
1
Y k sin k
co s n
= : (2 6 )
n= 1
2 k

(T h is is a cu te resu lt w h ich y o u m ig h t w a n t to p rov e fo r y o u rself; A ll y o u n eed


to d o is to w rite co s(k = 2 n ) = (1 = 2 )[sin (k = 2 n ¡ 1 )= sin (k = 2 n )], ta k e a p ro d u ct o f N
term s ca n cellin g o u t th e sin es a n d th en ta k e th e lim it N ! 1 .) S in ce th e fo u rier
tra n sfo rm o f (sin k = k ) is ju st a u n ifo rm d istrib u tio n , w e g et th e ta n ta lisin g resu lt
th a t P (x ) is ju st a u n ifo rm d istrib u tio n in th e in terva l (¡ 1 ;1 ) a n d zero elsew h ere!

RESONANCE  August 2009 805


SERIES  ARTICLE

T h is trick w h ich h a s b een u sed to g et (2 6 ) a lso w o rk s fo r ¸ = 2 ¡1 = 2 ;2 ¡ 1 = 4 :::; etc.


F o r ex a m p le, w h en ¸ = 2 ¡1 = 2 th e in ¯ n ite p ro d u ct is
1 p
Y sin k sin 2 k
n =2
co s(k = 2 ) = p : (2 7 )
n= 1
k 2 k

T h e F o u rier tra n sfo rm o f th is in v o lv es a co n v o lu tio n o f tw o recta n g u la r d istrib u -


tio n s a n d is ea sily seen to b e a tria n g u la r p ro b a b ility d istrib u tio n . (I w ill let y o u
ex p lo re th e g en era l ca se o f ¸ = 2 ¡1 = k .) It tu rn s o u t th a t th e in ¯ n ite p ro d u ct o f
co s(k ¸ n ) is ex tra o rd in a rily sen sitiv e to ch a n g es in ¸ . F o r a lm o st ev ery ¸ in th e
in terva l (0 :5 ;1 ) th is p ro d u ct is sq u a re in teg ra b le. B u t o n ce in a w h ile, it is n o t
so . F u rth er, if ¸ < (1 = 2 ), th e su p p o rt fo r P (x ) h a p p en s to b e th e Cp a n to r set.
T h e rea lly b iza rre b eh av io u r o ccu rs w h en ¸ is th e g o ld en ra tio g = ( 5 ¡ 1 )= 2 .
C lea rly, th ere a re en o u g h su rp rises in sto re in th e stu d y o f ra n d o m w a lk s.

Suggested Reading

[1] This integral was first evaluated by Watson in terms of elliptic integrals and a “simpler” result was obtained by
Glasser and Zucker: G N Watson, Three triple integrals, Quarterly J. Math., Vol.10, p.266, 1939;
M L Glasser and I J Zucker, Extended Watson integrals for the cubic latice, Proc. Natl. Acad. Sci., USA,Vol.74,
p.1800, 1977.
[2] There is a large literature on this subject, most of which can be found from the references in the book: P G Doyle
and J Laurie Snell, Random Walks and Electric Networks, Mathematical Association of America, Oberlin, OH,
1984. The book is available on the web as a pdf file at: http://math.dartmouth.edu/~doyle docs/walks/walks.pdf.
[3] This interesting topic does not seem to have been explored extensively. A good discussion is available in the paper:
P L Krapivsky and S Redner, Am. J. Phys., Vol.72(5), p.591, 2004. Also see, K E Morrison, Random Walks with
Decreasing Steps, available at: http://www.calpoly.edu/~kmorriso/Research/RandomWalks.pdf

Address for Correspondence


T Padmanabhan, IUCAA, Post Bag 4, Pune University Campus ,Ganeshkhind, Pune 411 007, India.
Email: paddy@iucaa.ernet.in, nabhan@iucaa.ernet.in

806 RESONANCE  August 2009