Professional Documents
Culture Documents
DECEMBER 1997
I. INTRODUCTION
1063-651X/97/56~6!/7053~14!/$10.00
7053
56
7054
56
where the integration is taken over all the points in the two
objects. We assume that the potential V(x1 2x2 2u) depends
only on the distance r5 @ (x1 2x2 2u) 2 1h 2 # 1/2 between the
points x1 and x2 1u on the two objects, and decays sufficiently rapidly with r.
The assumption that the charges are randomly distributed
on the rods ~plates!, with average charge density
r 5
r1
2
r 2 , leads to a Gaussian distribution of the quenched
variations of density d r i (x)5 r i (x)2
r , with the second moment
d r i ~ x! d r j ~ x8 ! 5g d i j d ~ x2x8 ! ,
~2!
r 1 1
r 2 is the number density of charges, and the
where g[
bar denotes averaging over the ensemble of all possible realizations of quenched disorder on the rods, with the probability distribution
FIG. 1. Schematic drawing of ~a! two randomly charged parallel
rods, one of length L and the other infinite, separated by a vertical
distance h and displaced by a distance u. ~b! A similar drawing of
two plates, one of area A and the other infinite.
W @ d r # 5N 21 e 2S @ d r # ,
where
S@dr#[
N[
EE
~1!
d r 2 ~ x!
2g
D @ d r # e 2S @ d r #
~4!
~5!
is the normalization factor ~* D @ d r # denotes functional integration over the density variations!.
In order to avoid complications associated with the
change of overlap between the two rods of lengths L and L 8
~or two plates of areas A and A 8 ! upon parallel displacement,
we assume that L!L 8 , and consider only relative displacements for which the two rods overlap ~a similar assumption,
A!A 8 , is made for the case of two plates!. The energy E(u)
associated with a particular relative shift of the two objects
differs, in general, from the average interaction energy
E,
E 5
r 2M
E ~ u! 5
dx
and
II. MODEL
~3!
dx V ~ x! ,
~6!
d @ E2 d E ~ u!# u @ L 2 ~ u!# ,
(
extr
~7!
where the sum goes over the positions u of all the energy
extrema, each of which is defined by the relation
56
7055
E 6[
E xx 6E y y
.
2
L 1 5E 1 1 A
~9!
~10!
d @ E2 d E ~ u!# u @ L 2 ~ u!# .
(
extr
du d $ ] @ d E ~ u !# / ] u % u ] 2 @ d E ~ u !# / ] u 2 u ...,
~12!
~8!
5 E du ( d ~ u2u ext!
(
ext
ext
u ~ L 2 @ u# ! 5
~11!
dL d L2
] 2 @ d E ~ u !#
.
]u2
~13!
Substituting Eqs. ~13! and ~12! into Eq. ~11!, and changing
the order of averaging over d r 2 and integration over u
~which gives the length L 8 of the infinite rod!, we obtain
the following representation for the energy distribution function:
N min~ E ! 5L 8
dL L
1
N
D @ d r 2 # e 2S @ d r 2 #
3 d $ ] @ d E ~ u !# / ] u % d @ E2 d E ~ u !# d $ L
2 ] 2 @ d E ~ u !# / ] u 2 % .
~14!
d @ E2 d E ~ u !# 5
d
d L2
J E
J E
dk 1
exp $ ik 1 @ E2 d E ~ u !# % ,
2p
H
H F
] @ d E ~ u !#
5
]u
dk 2
] @ d E ~ u !#
exp 2ik 2
, ~15!
2p
]u
] 2 @ d E ~ u !#
5
]u2
dk 3
] 2 @ d E ~ u !#
exp ik 3 L2
2p
]u2
GJ
7056
d r *2 ~ x ! [2ig
dx 8 K ~ x2x 8 2u ! d r 1 ~ x 8 !
~16!
and
K ~ x ! 5k 1 V ~ x ! 1k 2 ] V ~ x ! / ] x1k 3 ] 2 V ~ x ! / ] x 2 .
~17!
Substituting Eq. ~16! into the definition of S @ d r 2 # , we observe that the product d r 1 (x) d r 1 (x 8 ) in the integral over the
coordinates x and x 8 can be replaced by its average over the
length of the rod. In the thermodynamic limit ~the length
of the shorter rod L is much larger than the average spacing
between the charges on the rod, 1/g! we can replace the latter
average by the quenched average ~2!,
d r 1 ~ x ! d r 1 ~ x 8 ! d r 1 ~ x ! d r 1 ~ x 8 ! 5g d ~ x2x 8 ! ,
~18!
)E
j51
3
dL L
dk j
2p
3exp ik 1 E1ik 3 L2
g 2L
2
(jl D jl k j k l
,
~19!
m 0~ h !
0
D5
2m 1 ~ h !
0
m 1~ h !
0
2m 1 ~ h !
0
.
m 2~ h !
~20!
m 1 ~ h ! [2
dr V 2 ~ r ! ,
m 2 ~ h ! [2
E F
`
E F] ] G
`
dr
V~ r ! 2
,
r
~21!
] 2V~ r ! 2
dr
.
]r2
N min~ E ! 5L 8
c
dL P~ L ! P ~ E u L ! .
~22!
Here
c is the average concentration of potential wells ~number of potential minima per unit length of the rod of length
L 8 ! for parallel rods separated by a distance h
c5
S D
1 m 2~ h !
2 p m 1~ h !
1/2
e 2k
Ar 2 1h 2
Ar 2 1h 2
~24!
where k 21 is the Debye screening length ~we set the constant coefficient in front of this expression to unity!. The
qualitative picture discussed in the preceding paragraph
remains valid in the range g 21 ,h, k 21 . A straightforward calculation shows that in the limit of strong
screening ( k h@1), m 1 ; k 1/2h 25/2 exp(22kh) and m 2
; k 3/2h 27/2 exp(22kh), and therefore
c ;( k /h) 1/2. The conclusion that the concentration of energy minima at distances
exceeding the screening length is larger in the screened than
in the unscreened Coulomb case, is quite unexpected. However, as will be shown in the following, the energy of these
potential wells is exponentially reduced by screening effects.
The function P~L! which appears in Eq. ~22!, is the probability of finding an energy minimum with a given stiffness
~inverse squared radius of curvature! L
P~ L ! 5
L
L2
exp
2
.
g 2 Lm 2 ~ h !
2g 2 Lm 2 ~ h !
~25!
This function has a maximum at L max5g@Lm2(h)#1/2 and decays to zero for both large and small radii of curvature. The
function P(E u L) which also appears in Eq. ~22! is the conditional probability to find a minimum with energy E, given
that its stiffness is L,
P~ EuL !5
1
~ 2 ps 2E ! 1/2
exp 2
@ E2 d E min~ L !# 2
2 s 2E
. ~26!
56
~23!
d E min~ L ! 52
m 1~ h !
L,0,
m 2~ h !
~27!
s 2E 5g 2 L @ m 0 ~ h ! 2m 21 ~ h ! /m 2 ~ h !#
~28!
56
7057
d E min52
S D
gm 1 ~ h !
pL
4
m 2~ h !
1/2
~29!
L8
g AhL
F SDG
p min
E h
g L
1/2
where
p min~ y ! [
S D
F S D
&
y2
exp
2
16p 2
2p
3 5 exp 2
S D G
y
y2
1y erf
2y , ~30!
25p
5 Ap
FIG. 3. Plot of the number of minima per unit energy per unit
length, N min(E)/L8, as a function of the energy d E, for different
vertical separations between the rods. The broken line corresponds
to h5L, and the solid line to h50.1L ~the density of charges g is
unity!.
2`
dL 8 L u L 8 u
D @ d r 2 # e 2S @ d r 2 #
3 d $ ] @ d E ~ u !# / ] u % d $ L2 ] 2 @ d E ~ u !# / ] u 2 %
3 d $ ] @ d E ~ u1w !# / ] u % d $ L 8 2 ] 2 @ d E ~ u1w !# / ] u 2 % .
~31!
The functional integral over d r 2 (x) can be calculated as before, with the result
c ~ L u w ! 5g 24 L 22 @ 12 h 21 ~ w !# 21/2@ 12 h 22 ~ w !# 21/2
3
1
~ 2 p ! m 1m 2
2
2`
dL 8 L u L 8 u
L 2 1 ~ L 8 ! 2 22 h 2 ~ w ! LL 8
3exp 2
2g 2 Lm 2 @ 12 h 22 ~ w !#
~32!
where we defined
1
N
h 1~ w ! [
h 2~ w ! [
1
m1
1
m2
2`
dr
] V ~ r ! ] V ~ r1w !
.
]r2
]r2
2
2`
] V ~ r ! ] V ~ r1w !
,
]r
]r
dr
~33!
7058
56
;h3/2L 1/5, i.e., the width of the minimum is a weakly increasing function of its stiffness.
The average width of an energy well is defined as
w ~ L max! .
2 P~ L ! q ~ L u w ! ,
c ~ L u w ! 5c
where
q~ Luw !5
12 h 22 ~ w !
12 h 21 ~ w !
D F
1/2
exp 2
L2
g Lm 2 @ h 22
2 ~ w ! 21 #
2
~34!
Am 1 m 3
m2
exp 2
L2
,
w g Lm 3
2 2
w!
S D
m2
m3
E F
`
dr
] 3V~ r ! 2
.
]r3
L
g ALm 3
~38!
21 @ 12 h 22 ~ w !# 3/2
2c
@ 11 h 22 ~ w !#@ 12 h 21 ~ w !# 1/2
~39!
~36!
1/2
~35!
where
m 3 [2
m 2~ h !
m 3~ h !
1/2
F G
~37!
We now consider what happens with a given energy minimum when we change the distance h between plates by a
small value d h. There are two types of changes: the minimum is displaced by the vector d u5( ] u/ ] h) d h, and its
depth and curvature change. Differentiating the equation for
the extrema, ] @ d E(u) # / ] u50, with respect to h, and taking
into account both the implicit and the explicit dependence of
the energy on h, we find
]u
] 2 @ d E ~ u !# / ] u ] h
52 2
]h
] @ d E ~ u !# / ] u 2
Q
52
L
where
] 2 @ d E ~ u !#
Q[
.
]u]h
~40!
56
P~ QuL !5
1
~ 2 ps 2Q ! 1/2
exp 2
where s 2Q is defined by
s 2Q 5g 2 L 2
E F ]] ] G
2
dr
Q2
2 s 2Q
~41!
1
V~ r ! 2
] m 1~ h !
2
h r
4m 1 ~ h !
]h
GJ
2
.
~42!
S DU
]u
]h
s 2Q
5 2.
L
L
]L
L ] lnm 2 ~ h !
.
5
]h L 2
]h
~44!
In order to obtain the dependence of the stiffness of the average potential well on the distance h between rods, we integrate this equation and obtain
~ h ! 5L
~ h !@ m ~ h ! /m ~ h !# 1/2.
L
0
2
2 0
~43!
~45!
7059
d E.6
SD
L
h
1/2
Q 2h
h
56g
SD
L
h
1/2
~46!
t . AgL!gL,
~47!
D. Scaling estimate
7060
L 2 ~ u! L 1 ~ u! u @ L 2 ~ u !# 5
dL 2
L2
C ~ E,E 1 ,E 2 ,E xy !
dL 1 L 2 L 1
3d @ L 2 2L 2 ~ u!#
3 d @ L 1 2L 1 ~ u!# ,
~49!
Substituting Eqs. ~49! and ~48! into Eq. ~11!, and changing
the order of averaging over d r 2 and integration over u
~which gives the area A 8 of the infinite plate!, we obtain the
following representation for the energy distribution function
E E
EEE
N min~ E ! 5A 8
3
dL 2
L2
dL 1 L 2 L 1
E pE
dk
2
3exp ikE1i
3 d ~ L 1 2E 1 2 A
E 22 1E 2xy ! .
~50!
22kq 1 m 81 ~ h !# ,
m 08 ~ h ! [2 p
3 d @ E2 d E ~ u!#
m 28 ~ h ! [
] @ d E ~ u !#
]u
J E
] @ d E ~ u!#
d
5
]u
d @ E j 2E j ~ u!# 5
dL 2
L2
dL 1 L 2 L 1
V~ r ! 2
,
r
~55!
dr r
V~ r ! 1 ]V~ r ! 2
2
.
r2
r ]r
~56!
22 k h
e
8~ kh !4
dL 2
dL 1 P~ L 2 ,L 1 ! P ~ E u L 2 1L 1 ! ,
~57!
dk
exp$ ik @ E2 d E ~ u!# % ,
2p
~ 112 k h !
2Ei~ 2 k h ! ,
~ 2kh !2
3 ~ 316 k h12 k 2 h 2 24 k 3 h 3 ! ,
~52!
~53!
dp
] @ d E ~ u!#
,
2 exp 2ip
]u
~2p!
~51!
dr r
m 82 ~ h ! 5 p k 4 Ei~ 2 k h ! 1
3 ~ L 1 2L 2 ! C ~ E,L 2 ,L 1 ! .
d @ E2 d E ~ u!# 5
)j d @ E j 2E j ~ u!#
E F] ] G
E F] ]
m 81 ~ h ! 5 p k 2 e 22 k h
pA8
2
m 18 ~ h ! [ p
m 08 ~ h ! 52 p Ei~ 2 k h ! ,
[C ~ E,L 2 ,L 1 !
N min~ E ! 5
~54!
dr rV 2 ~ r ! ,
C ~ E,E 1 ,E 2 ,E xy !
g 2A
2
D @ d r 2 # e 2S @ d r 2 # d
(j q j E j 2
dE 1 dE 2 dE xy C ~ E,E 1 ,E 2 ,E xy !
dq j
2p
)j
3 @ p2 m 81 ~ h ! 1k 2 m 80 ~ h ! 1 ~ 2q 21 1q 22 1q 2xy ! m 82 ~ h !
3 d ~ L 2 2E 1 1 A
1
N
dp
2
~ p !2
E 22 1E 2xy !
56
dq j
exp$ iq j @ E j 2E j ~ u!# % .
2p
c5
m 28 ~ h !
2 p ) m 81 ~ h !
~58!
56
SD
3
P~ L 2 ,L 1 ! 5
p
1/2
L 2 L 1 u L 1 2L 2 u
16@ g 2 Am 28 ~ h !# 5/2
3exp 2
3L 22 13L 21 22L 2 L 1
16g 2 Am 82 ~ h !
. ~59!
1
~ 2 ps 2E ! 1/2
exp
2 @ E2 d E ~ L 2 1L 1 !# 2
2 s 2E
~60!
d E ~ L 2 1L 1 ! 52
m 18 ~ h ! L 2 1L 1
,0,
2
m 82 ~ h !
~61!
s 2E 5g 2 A @ m 80 ~ h ! 2m 81 2 ~ h ! /2m 82 ~ h ! # .
~62!
7061
FIG. 5. Plot of the average energy of the minima, d E min ~in units
of gA 1/2!, as a function of the separation between the plates h ~in
arbitrary units of length, for different values of the inverse screening length k: 1 ~solid line!, 0.2 ~circles!, and 5 ~crosses!.
d E min528gm 18 ~ h !
A
3 p m 82 ~ h !
1/2
~63!
7062
two plates are randomly frozen ~and uncorrelated! and proceed to analyze the response of plate 2 to the displacement of
plate 1 ~see Fig. 6!.
As we have showed in Sec. IV A, the presence of frozen
random charge distributions on the plates leads to the appearance of a large number of energy minima which correspond
to different relative parallel displacements of the plates.
Since, for macroscopic plates, the typical energy associated
with these minima scales as A 1/2, for sufficiently large plates
it can be much larger that the thermal energy k B T and the
plates will get stuck in one of the minimal energy configurations. The electrostatic repulsion between the plates
scales as A and, therefore, the normal force needed to keep
the plates at a fixed separation from each other will be dominated by the average repulsion, and will not be significantly
affected by the relative displacement.
The presence of these minima will lead to the appearance
of macroscopic restoring forces for relative displacement of
the plates @10#. We showed earlier that the depth of an energy minimum has a typical value d E min @Eq. ~63!#. Since
such energy wells are densely distributed with respect to the
relative displacement of the plates, the system will occupy
one of these characteristic minima. Plate 2 will move together with plate 1 ~stick phase! until the force on it exceeds
some critical value f crit at which point it will recoil back
~slip!. This value can be estimated from the characteristics of
the typical energy well:
f crit.
FS D G
]E~ u !
]u
2 1/2
5g AAm 81 ~ h ! ,
E uE
2p
rdr
Au
~64!
E ~ w ,h,u ! 52g
56
~66!
56
w * 52
SD S
9
7
1/2
h
u
21
R &h
1/2
~67!
w *5
2 arcsin~ u/R !
p
when u<R
when u.R,
~68!
7063
FIG. 8. Schematic picture of the relative orientation of the mirror imaged charge distributions on the two discs. One of the discs
~shaded! is fixed at the origin, and the relative orientation of the
other disc, corresponding to the minimal energy configuration for
different displacements, is shown. The orientations of the minima
are w * 50 in the region to the left of the dotted curve and w *
5 p to the right of the dashed curve. The region between the two
curves corresponds to 0, w * , p .
7064
In the above analysis we assumed perfect correlation between the frozen charge distributions on the two plates. If the
charges on the two discs are only partially correlated, the
attractive interaction will be reduced by a factor a proportional to the fraction of the correlated charges, and the incoherent contribution to the energy will remain unchanged. The
ratio of coherent to incoherent contributions will be reduced
by a factor a, and therefore a strong energy minimum corresponding to alignment of the coherent charge distributions
will now appear only when the discs approach separations of
order a R. Thus perfect recognition of the correlated parts of
the charge distributions on the two discs will still take place
when they are brought into contact.
The discussion in this section may appear to be based on
a somewhat artificial model, since the preparation of such
mirror imaged randomly frozen charge distribution may pose
formidable difficulties. Note, however, that since the theory
makes no assumptions about the dimensions of the charged
objects, one can construct such a system by assembling a
random array of negative and positive electrodes and then
constructing its mirror image. Another example to which the
general methodology described in this section can be applied
is that of an array of parallel magnets with their north and
south poles randomly alternating along the normal to the
array. A mirror image of this random array can be prepared in a thin ferromagnetic film which consists of disordered magnetic domains ~with directions pointing randomly
in and out of the surface of the film!, if this film is brought
near the Curie temperature and placed in contact with the
random magnetic array. A mirror image of the array will
form in the film, and can subsequently be frozen by decreasing the temperature. The resulting magnetic interactions between the film and the array will vary with their separation
and relative orientation in a manner closely resembling our
electrostatic model.
Finally, one can try to speculate on the relevance of our
model to recognition in biological systems. Although many
different types of attractive and repulsive interactions ~electrostatic, hydrophobic, hydrogen bonding, van der Waals,
etc.! are present in these systems @11#, the resulting potential
energy surfaces can be modeled by introducing random ~or
partially correlated! distributions of effective charges. The
justification for this statement is that the random combination of attractive and repulsive long range forces in our
theory can give rise to effective interaction potentials of arbitrary complexity and, therefore, the theory can serve as a
generic model of interactions between biological systems in
which perfect lock-and-key arrangement of the two objects at contact is preceded by prealignment at a distance,
when the separation of the approaching objects becomes of
the order of their dimensions ~or on the order of the Debye
screening length, in the presence of salt!. This prealignment
can provide a mechanism by which metastable traps at
contact can be avoided by sensing the energy landscape at
distances where the energy barriers are small enough to allow the system to find the true energy minimum ~for example, when the depth of the coherent minimum is larger
than k B T and the depths of the incoherent ones are smaller
than k B T, at separations of the order of the size of the objects!.
56
VI. DISCUSSION
56
7065
7066
56
represented by its average over the quenched ensemble ~see, for example, spin glasses @2#!. We proceed to check whether such
problems arise in our model.
Consider the variation of the number of energy minima per unit energy, N min(E), for different realizations of the disordered
charge distribution on the rods. A possible way to check whether intermittent behavior arises in our theory is to estimate higher
order moments of the function N min(E). Repeating the steps which led to Eq. ~19!, we obtain
) FE
F(
n
N nmin~ E ! 5
q51
L8
3exp i
du q
)E
j51
3
dL q L q
k q1 E1i
(q k q3 L q 2
dk q j
2p
g 2L
2
jl,qq 8
8
D qq
jl k q j k q 8 l ,
~A1!
~A2!
m 0 ~ h ! h 0 ~ u q 2u q 8 !
0
D8 5
2m 1 ~ h ! h 1 ~ u q 2u q 8 !
0
m 1 ~ h ! h 1 ~ u q 2u q 8 !
0
h 0~ u ! [
1
m0
2`
2m 1 ~ h ! h 1 ~ u q 2u q 8 !
0
.
m 2 ~ h ! h 2 ~ u q 2u q 8 !
~A3!
dr V ~ r ! V ~ r1u ! ,
g 2 L ] lnN min~ E !
3 12
2L 8
]E
2`
GJ
2
dr V ~ r !
~A6!
In all physically relevant situations, some degree of screening is present, and the integral over the potential converges.
Since ] lnNmin(E)/]E.Ah/g 2 L @see Eq. ~30!#, the variance
of lnNmin(E) is of order h/L 8 , and we conclude that intermittency effects come into play only when the separation between the rods exceeds the length of the longer rod, i.e.,
when the average number of energy minima becomes of the
order of or smaller than unity. It can be shown that the corresponding correction in the case of two randomly charged
plates in the presence of screening is of order ( k 2 A) 21 .
N nmin~ E ! 5 @ N min~ E !# n 11
3
] lnN min~ E !
]E
g 2 L n ~ n21 !
L8
2
2`
GJ
2
dr V ~ r !
~A5!