Professional Documents
Culture Documents
Theoretical Molecular Biophysics
Theoretical Molecular Biophysics
Theoretical Molecular Biophysics
biomedical engineering
Theoretical
Molecular Biophysics
123
Professor Dr. Philipp Scherer
Professor Dr. Sighart F. Fischer
Technical University München, Physics Department T38
¨
85748 Munchen, Germany
E-mail: philipp.scherer@ph.tum.de, sighart.fischer@ph.tum.de
Biophysics deals with biological systems, such as proteins, which fulfill a vari-
ety of functions in establishing living systems. While the biologist uses mostly
a phenomenological description, the physicist tries to find the general con-
cepts to classify the materials and dynamics which underly specific processes.
The phenomena span a wide range, from elementary processes, which can
be induced by light excitation of a molecule, to communication of living sys-
tems. Thus, different methods are appropriate to describe these phenomena.
From the point of view of the physicist, this may be Continuum Mechanics to
deal with membranes, Hydrodynamics to deal with transport through vessels,
Bioinformatics to describe evolution, Electrostatics to deal with aspects of
binding, Statistical Mechanics to account for temperature and to learn about
the role of the entropy, and last but not least Quantum Mechanics to under-
stand the electronic structure of the molecular systems involved. As can be
seen from the title, Molecular Biophysics, this book will focus on systems
for which sufficient information on the molecular level is available. Compared
to crystallized standard materials studied in solid-state physics, the biological
systems are characterized by very big unit cells containing proteins with thou-
sands of atoms. In addition, there is always a certain amount of disorder, so
that the systems can be classified as complex. Surprisingly, the functions like
a photocycle or the folding of a protein are highly reproducible, indicating a
paradox situation in relation to the concept of maximum entropy production.
It may seem that a proper selection in view of the large diversity of phe-
nomena is difficult, but exactly this is also the challenge taken up within this
book. We try to provide basic concepts, applicable to biological systems or
soft matter in general. These include entropic forces, phase separation, coop-
erativity and transport in complex systems, like molecular motors. We also
provide a detailed description for the understanding of elementary processes
like electron, proton, and energy transfer and show how nature is making use
of them for instance in photosynthesis. Prerequisites for the reader are a basic
understanding in the fields of Mechanics, Electrostatics, Quantum Mechanics,
and Statistics. This means the book is for graduate students, who want to
VI Preface
4 Debye–Hückel Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.1 Electrostatic Shielding by Mobile Charges . . . . . . . . . . . . . . . . . . 45
4.2 1–1 Electrolytes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.3 Charged Sphere . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.4 Charged Cylinder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.5 Charged Membrane (Goüy–Chapman Double Layer) . . . . . . . . . 52
4.6 Stern Modification of the Double Layer . . . . . . . . . . . . . . . . . . . . . 57
Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
5 Protonation Equilibria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
5.1 Protonation Equilibria in Solution . . . . . . . . . . . . . . . . . . . . . . . . . 61
5.2 Protonation Equilibria in Proteins . . . . . . . . . . . . . . . . . . . . . . . . . 64
5.2.1 Apparent pKa Values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.2.2 Protonation Enthalpy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
5.2.3 Protonation Enthalpy Relative to the Uncharged State . 68
5.2.4 Statistical Mechanics of Protonation . . . . . . . . . . . . . . . . . 69
5.3 Abnormal Titration Curves of Coupled Residues . . . . . . . . . . . . 70
Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
6 Formal Kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
6.1 Elementary Chemical Reactions . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
6.2 Reaction Variable and Reaction Rate . . . . . . . . . . . . . . . . . . . . . . 75
6.3 Reaction Order . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
6.3.1 Zero-Order Reactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
6.3.2 First-Order Reactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
6.3.3 Second-Order Reactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
6.4 Dynamical Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
6.5 Competing Reactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
6.6 Consecutive Reactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
6.7 Enzymatic Catalysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
6.8 Reactions in Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
6.8.1 Diffusion-Controlled Limit . . . . . . . . . . . . . . . . . . . . . . . . . . 83
6.8.2 Reaction-Controlled Limit . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Part IX Appendix
Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 365
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 369
Part I
In this chapter, we study simple statistical models for the entropic forces which
are due to the large number of conformations characteristic for biopolymers
like DNA or proteins (Fig. 1.1).
R
O
H H
Ψ Φ H
H + Cα N C
N Cα
H H C ω
H
O R
Since the different units are independent, the joint probability distribution
factorizes
M
P (b1 , . . . , bM ) = P (bi ). (1.4)
i=1
N
R= bi (1.5)
i=1
and hence
∞
1 1 2 sin kbi
δ(|bi | − b)e ikbi 3
d bi = 2π dbi δ(bi − b)b2i
4πb2 0 4πb 2 kbi
sin kb
= , (1.13)
kb
and finally we have
M
1 3 −ikR sin kb
P (R) = d ke . (1.14)
(2π)3 kb
The function M
sin kb
(1.15)
kb
has a very sharp maximum at kb = 0. For large M , it can be approximated
quite accurately by a Gaussian:
M
sin kb
≈ e−M(1/6)k
2 2
b
(1.16)
kb
which gives
3/2
1 −ikR −(M/6)k2 b2 3
e−3R
2
/(2b2 M)
P (R) ≈ 3
d ke e = .
(2π)3 2πb2 M
(1.17)
3R2
F = −T S = −kB T ln P (R) = kB T + const. (1.18)
2b2 M
The quadratic dependence on L is very similar to a Hookean spring. For a
potential energy
6 1 Random Walk Models for the Conformation
ks 2
V = x , (1.19)
2
the probability distribution of the coordinate is
ks
e−ks x /2kB T
2
P (x) = (1.20)
2πkB T
which gives a free energy of
ks x2
F = −kB T ln P = const + . (1.21)
2
By comparison, the apparent spring constant is
3kB T
ks = . (1.22)
M b2
We consider now a chain with one fixed end and an external force acting in
x-direction at the other end [1] (Fig. 1.3).
The projection of the ith segment onto the x-axis has a length of (Fig. 1.4)
b
κ
b
φ
θ
x
The chain members are divided into n groups according to their bond
projections bj . The number of units in each group is denoted by Mi so that
n
Mi = M (1.24)
i=1
sin(θ)dθdφ
P (θ, φ)dθdφ = . (1.26)
4π
Since we are only interested in the probability of the li , we integrate over φ
sin(θ)dθ
P (θ)dθ = (1.27)
2
and transform variables to have
1 1
P (l)dl = P (−b cos θ)d(−b cos θ) = d(−b cos θ) = dl. (1.28)
2b 2b
The canonical partition function is
M ! Mi z Mi
Z(L, M, T ) = zi = M! i
. (1.29)
{Mi }
Mi li =L j Mj ! i M
{Mi }
i i!
0.5
0
L/Mb
–0.5
–1
–20 –10 0 10 0
κb/kBT
In this new ensemble, the summation over L simplifies the partition function:
z Mi (zeκli /kB T )Mi
Δ= eκ Mi li /kB T
M! = M!
Mi ! Mi !
{Mi } {Mi }
M
= zeκli /kB T = ξ(κ, T )M . (1.33)
F = −kB T ln Z
L − M lβ M lα − L
= −kB T ln zα − kB T ln zβ − kB T (M ln M − M )
lα − lβ lα − lβ
L − M lβ L − M lβ L − M lβ
= −kB T − ln +
lα − lβ lα − lβ lα − lβ
M lα − L M lα − L M lα − L
− ln + . (1.41)
lα − lβ lα − lβ lα − lβ
lα lβ
zb /za = 5
2
κ(1a–1b )/kBT
zb /za = 1
0
1/6 1/2 1/1.2
zb /za = 0.2
–2
–4
lα − l β δ zβ
κ = ln + ln . (1.44)
kB T 1−δ zα
The mean extension for zero force is obtained by solving κ(L) = 0:
z α lα + z β lβ
L0 = M , (1.45)
zα + zβ
L̄0 − M lβ zα
δ0 = = . (1.46)
M (lα − lβ ) zα + zβ
Taylor series expansion around L0 gives the linearized force–extension rela-
tion:
∂F kB T (zα + zβ )2
κ= = (L − L̄0 ) + · · ·
∂L M (lα − lβ ) 2 zα zβ
kB T (zα + zβ )2
≈ (δ − δ0 )
lα − lβ zα zβ
kB T 1
= (δ − δ0 ). (1.47)
lα − lβ δ0 (1 − δ0 )
1.2 Two-Component Model 11
∗∗∗αα∗∗∗ +1 +1 +2
∗∗∗αβ∗∗∗ +1 +1 +2
∗∗∗βα∗∗∗ +1 +1 +2
∗∗∗ββ∗∗∗ +1 +2 −1 +2
∗∗∗αα∗∗∗ +1 −1 +2 +2
∗∗∗αβ∗∗∗ +1 +1 +2
∗∗∗βα∗∗∗ +1 +1 +2
∗∗∗ββ∗∗∗ +1 +1 +2
α α 2 0 1 0 0 −2 0
β β 0 2 0 0 1 0 −2
α β 1 1 0 1 0 −1 −1
α α 2 0 2 0 0 0 0
β β 0 2 0 0 2 0 0
α β 1 1 0 2 0 0 0
Fig. 1.10. Determination of the constants. c1 and c2 are zero for periodic boundary
conditions and in the range −2, . . . , 0 for open boundaries
1
Nββ = Mβ − Nαβ , (1.51)
2
w = wαα + wββ − 2wαβ (1.52)
to have the interaction energy
1 1
W = wαα Mα − Nαβ + wββ Mβ − Nαβ + wαβ Nαβ
2 2
w
= wαα Mα + wββ (M − Mα ) − Nαβ . (1.53)
2
The canonical partition function is
g(Mα , Nαβ )e−W (Nαβ )/kB T
M
Z(Mα , T ) = zαMα zβ β
Nαβ
Mα
(M−Mα )
= zα e−wαα /kB T zβ e−wββ /kB T
× g(Mα , Nαβ )eNαβ w/2kB T . (1.54)
Nαβ
αα ββ α β
αα β α ββ
α ββ αα β
α β αα ββ
ββ αα β α
ββ α β αα
β αα ββ α
β α ββ αα
Fig. 1.11. Degeneracy factor g. The possible eight configurations are shown for
M = 4, Mα = 3, Mβ = 3, and Nαβ = 3
ααα ααα
ααα α αα
ααα αα α
ααα ααα
α αα α αα
ααα α α α
ααα α αα
αα α αα α
ααα αα α
ααα ααα
Fig. 1.12. Distribution of three segments over three blocks. This is equivalent to
the arrangement of three segments and two border lines
to distribute the (Nαβ − 1)/2 walls over the total of Mα − 1 objects which is
given by (Fig. 1.12)
(Mα − 1)! Mα !
Nαβ −1 N +1
≈ Nαβ N
. (1.55)
( 2 )!(Mα − αβ2 )! ( 2 )!(Mα − 2αβ )!
The same consideration for the β-segments gives another factor of
(M − Mα )!
N Nαβ
. (1.56)
( 2αβ )!(M − Mα − 2 )!
∂
w ∂
0= ln g(Mα , Nαβ )ewNαβ /2kB T = + ln g(Mα , Nαβ ).
∂Nαβ 2kB T ∂Nαβ
(1.58)
Stirling’s approximation gives
∗ ∗ ∗
w 1 Nαβ 1 Nαβ Nαβ
0= + ln Mα − + ln M − Mα − − ln
2kB T 2 2 2 2 2
(1.59)
or
N∗ N∗
w (Mα − 2αβ )(M − Mα − 2αβ )
0= + ln N∗
. (1.60)
kB T ( αβ )2 2
Taking the exponential gives
∗ ∗ 2
Nαβ Nαβ Nαβ
Mα − M − Mα − = e−w/kB T . (1.61)
2 2 2
∗
For Nαβ , we obtain approximately
∗ w
Nαβ = 2M δ(1 − δ) + (δ − δ 2 )2 + ··· . (1.69)
kB T
Let us now apply the maximum term method, which approximates the
logarithm of a sum by the logarithm of the maximum summand:
F = −kB T ln Z(Mα , T )
≈ −kB T Mα ln zα − kB T (M − Mα ) ln zβ + Mα wαα + (M − Mα )wββ
∗
∗
wNαβ
−kB T ln g(Mα , Nαβ )− . (1.70)
2
The force–length relation is now obtained from
L−Mlβ
∂F ∂F ∂ lα −lβ 1 ∂F
κ= = =
∂L ∂Mα ∂L lα − lβ ∂Mα
1 ∂
= −kB T ln zα + kB T ln zβ + wαα − wββ − kB T ln g
lα − lβ ∂Mα
∗ ∗
1 ∂N αβ ∂ wN αβ
+ ∗ −kB T ln g − . (1.71)
lα − lβ ∂Mα ∂Nαβ 2
∗
The last part vanishes due to the definition of Nαβ . Now using Stirling’s
formula, we find
N∗
∂ Mα M − Mα − 2αβ δ 1−δ−γ
ln g = ln + ln ∗ = ln + ln
∂Mα M − Mα Mα −
N αβ 1−δ δ−γ
2
(1.72)
and substituting γ, we have finally
(lα − lβ ) zβ e−wββ /kB T (1 − δ)
κ = ln −w /k T
− ln
kB T zα e αα B δ
2
w w
+(2δ − 1) + δ(3δ − 2δ 2 − 1) . (1.73)
kB T kB T
Linearization now gives
zα e−wαα /kB T
δ0 = ,
zα e−wαα /kB T + zβ e−wββ /kB T
kB T 1
κ= (δ − δ0 )
lα − lβ δ0 (1 − δ0 )
2
w w
+ (2δ0 − 1) + (3δ02 − δ0 − 2δ03 )
kB T kB T
2
w w
+ 2 + (6δ0 − 6δ02 − 1) (δ − δ0 ). (1.74)
kB T kB T
16 1 Random Walk Models for the Conformation
10
–5
–10
0 0.2 0.4 0.6 0.8 1
δ
Fig. 1.13. Force–length relation for the interacting two-component model. Dashed
curves: exact results for w/kB T = 0, ±2, ±5. Solid curves: series expansion (1.73)
which gives a good approximation for |w/kB T | < 2
For negative w, a small force may lead to much larger changes in length
than with no interaction. This explains, for example, how in proteins huge
channels may open although the acting forces are quite small. In the case
of Myoglobin, this is how the penetration of oxygen in the protein becomes
possible (Fig. 1.13).
Problems
1.1. Gaussian Polymer Model
The simplest description of a polymer is the Gaussian polymer model which
considers a polymer to be a series of particles joined by Hookean springs
n
n−1
(a) The vector connecting monomers n−1 and n obeys a Gaussian distribution
with average zero and variance
(rn − rn−1 )2 = b2 .
N
f
V = (rn − rn−1 )2 − κ(xN − x0 )
n=1
2
1.2 Two-Component Model 17
P ∼ e−V /kB T .
ri = R1 R2 · · · Ri−1 r1 θ
with ⎛ ⎞ φ
0 ri+1
r1 = ⎝ 0 ⎠
b ri
⎛ ⎞⎛ ⎞
cos φi sin φi 0 cos θi 0 − sin θi
Ri = Rz (φi )Ry (θi ) = ⎝ − sin φi cos φi 0 ⎠ ⎝ 0 1 0 ⎠.
0 0 1 sin θi 0 cos θi
2
N
Calculate the mean square of the end-to-end distance i=1 ri for the
following cases:
(a) The averaging · · · includes averaging over all angles φi and θi .
(b) The averaging · · · includes averaging over all angles φi while the angles
θi are held fixed at a common value θ.
(c) The averaging · · · includes averaging over all angles φi while the angles
θi are held fixed at either θ2i+1 = θa or θ2i = θb depending on whether
the number of the link is odd or even.
2
The rotation matrices act in the laboratory fixed system (xyz). Trans-
formation into the coordinate system of the segment (x y z ) changes
the order of the matrices. For instance, r2 = R(y, θ1 )R(z, φ1 )r1 =
R(z, φ1 )R(y , θ1 )R−1 (z, φ1 )R(z, φ1 )r1 = R(z, φ1 )R(y , θ1 )r1 (this is sometimes
discussed in terms of active and passive rotations).
18 1 Random Walk Models for the Conformation
(d) How large must N be, so that it is a good approximation to keep only
terms which are proportional to N ?
(e) What happens in the second case if θ is chosen as very small (wormlike
chain)?
Hint. Show first that after averaging over the φi , the only terms of the matrix
which have to be taken into account are the elements (Ri )33 . The appearing
summations can be expressed as geometric series.
(a) Make use of the asymptotic expansion3 of the logarithm of the Γ -function
√ 1 1
ln(Γ (z + 1)) = (ln z − 1)z + ln( 2π) + ln z + + O(z −3 ),
2 12z
N ! = Γ (N + 1)
to calculate the leading terms of the force–extension relation which is
obtained from
∂
κ= (−kB T ln Ω(L, M, T )).
∂L
Discuss the error of Stirling’s approximation for M = 1,000 and lβ /lα = 2.
(b) Now switch to an ensemble with constant force κ. The corresponding
partition function is
Z(κ, M, T ) = eκL/kB T Ω(L, M, T ).
L
3
Several asymptotic expansions can be found in [5].
2
Flory–Huggins Theory for Biopolymer
Solutions
In the early 1940s, Flory [6] and Huggins [7], both working independently,
developed a theory based upon a simple lattice model that could be used to
understand the nonideal nature of polymer solutions. We consider a lattice
model where the lattice sites are chosen to be the size of a solvent molecule
and where all lattice sites are occupied by one molecule [1].
ΔSm = kB (N ln N − N − Nα ln Nα + Nα − Nβ ln Nβ + Nβ )
= kB (N ln N − Nα ln Nα − Nβ ln Nβ )
Nα Nβ
= −kB Nα ln − kB Nβ ln . (2.2)
N N
Inserting the volume fractions
Nα Nβ
φα = , φβ = , (2.3)
Nα + Nβ Nα + Nβ
the mixing entropy can be written in the well-known form
20 2 Flory–Huggins Theory for Biopolymer Solutions
solvent
solute
0 χ =2
–0.2
χ =2
–0.6 χ =0
Fig. 2.2. Free energy change ΔF/N kB T of a binary mixture with interaction
F = −kB T ln Z = −Nα kB T ln zα − Nβ kB T ln zβ
c c Nα Nβ cw
= +Nα wαα + Nβ wββ − + N kB T (φα ln φα + φβ ln φβ ).
2 2 2N
(2.12)
Nα Nβ cw cw
ΔEm = − = −N φα φβ = N kB T χφα φβ . (2.16)
2N 2
The last equation defines the Flory interaction parameter (Fig. 2.2):
cw
χ=− . (2.17)
2kB T
For χ > 2, the free energy has two minima and two stable phases exist. This
is seen from solving
22 2 Flory–Huggins Theory for Biopolymer Solutions
∂ΔF ∂
0= = N kB T (χφα (1 − φα ) + φα ln φα + (1 − φα ) ln(1 − φα ))
∂φα ∂φα
φα
= N kB T χ(1 − 2φα ) + ln . (2.18)
1 − φα
This equation has as one solution φα = 1/2. This solution becomes instable
for χ > 2 as can be seen from the sign change of the second derivative
∂ 2 ΔF 1 − 2χφα + 2χφ2α
= N k B T = N kB T (4 − 2χ). (2.19)
∂φ2α φ2α − φα
Nα M Nβ
φα = , φβ = (2.20)
Nα + M Nβ Nα + M Nβ
N = Nα + M Nβ . (2.21)
It consists of the mixing entropy and a contribution due to the different con-
formations of the polymers (disorder entropy). The latter can be eliminated
by subtracting the entropy for Nα = 0:
ln Ω(Nα , Nβ )
ΔSm = ΔS − ΔSd = kB . (2.23)
ln Ω(0, Nβ )
solvent
solute
a r= 2 r =3 r=4 r=5
3 4
1 2 1 3 2 1 1 2
b r=5
3 4
2 1
Fig. 2.4. Available positions. The second segment can be placed onto four positions,
the third segment onto three positions. For the following segments, it is assumed
that they can be also placed onto three positions (a). Configurations like (b) are
neglected
lattice, taking into account only the available volume but not the configuration
of all the other polymers. Under that conditions, Ω factorizes
1
Nβ
Ω= νi , (2.24)
Nβ ! i=1
where νi counts the number of possibilities to put the ith polymer onto the lat-
tice. It will be calculated by adding one segment after the other and counting
the number of possible ways:
M
νi+1 = ni+1
s . (2.25)
s=1
The first segment of the (i + 1)th polymer molecule can be placed onto
ni+1
1 = N − iM (2.26)
ΔSd = ΔS(Nα = 0)
c−1
= −kB Nβ ln Nβ + kB Nβ + kB Nβ (M − 1) ln
M Nβ
+kB (M Nβ ln M Nβ − M Nβ ). (2.34)
2.2 Polymeric Solution 25
0.6
0.5
ΔSm /kBT
0.4
0.3
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1
φβ
ΔSm = ΔS − ΔSd
= kB (N ln N − N + Nα − Nα ln Nα − M Nβ ln M Nβ + M Nβ )
+kB Nβ (M − 1)(ln M Nβ − ln N )
= kB (N ln N − Nα ln Nα − M Nβ ln M Nβ + M Nβ ln M Nβ
−M Nβ ln N − Nβ ln M Nβ + Nβ ln N )
Nα M Nβ
= −kB Nα ln + Nβ ln = −kB Nα ln φα − kB Nβ ln φβ
N N
φβ
= −N kB φα ln φα + ln φβ . (2.35)
M
In comparison with the expression for a solution of molecules without internal
flexibility, we obtain an extra contribution to the entropy of
M Nβ Nβ
− Nβ kB ln + Nβ kB ln = −Nβ kB ln M. (2.36)
N N
Next, we calculate the change of energy due to mixing, ΔEm . wαα is the
interaction energy between nearest-neighbor solvent molecules, wββ between
nearest-neighbor polymer units (not chemically bonded), and wαβ between
one solvent molecule and one polymer unit. The probability that any site
is occupied by a solvent molecule is φα and by a polymer unit is φβ . We
introduce an effective coordination number c which takes into account that a
solvent molecule has c neighbors, whereas a polymer segment interacts only
with c-2 other molecules. Then
Nα Nβ
Nαα = cφα , Nββ = M cφβ , (2.37)
2 2
Nαβ = cφα Nβ . (2.38)
26 2 Flory–Huggins Theory for Biopolymer Solutions
In the pure polymer φβ = 1 and Nββ = M cNβ /2, whereas in the pure solvent
Nαα = cNα/2 . The energy change is
Nα Nβ
ΔEm = cwαα φα + wββ M cφβ + wαβ M cφα Nβ
2 2
Nα Nβ
−wαα c − wββ M c
2 2
Nα Nβ
= −cwαα φβ − cM wββ φα + wαβ M cφα Nβ
2 2
wαα + wββ
= wαβ − cN φα φβ
2
w
= − cN φα φβ = N kB T χφα φβ , (2.39)
2
with the Flory interaction parameter
wc
χ=− . (2.40)
2kB T
For the change in free energy, we find (Figs. 2.6–2.8)
ΔFm ΔEm ΔSm φβ
= − = φα ln φα + ln φβ + χφα φβ . (2.41)
N kB T N kB T N kB M
0.2
0.1
0
ΔFm / NkBT
– 0.1
– 0.2
– 0.3
– 0.4
0 0.2 0.4 0.6 0.8 1
Φβ
– 0.1
ΔFm / NkBT
– 0.2
– 0.3
– 0.4
– 0.5
0 0.2 0.4 0.6 0.8 1
Φβ
0.2
0.1
ΔFm / NkBT
– 0.1
– 0.2
0 0.2 0.4 0.6 0.8 1
Φβ
if the free energy of the two-phase system is lower, that is, the following
equation can be fulfilled:
or
φβ − φβ φβ − φβ
N = N , (N − N ) = N . (2.45)
φβ − φβ φβ − φβ
28 2 Flory–Huggins Theory for Biopolymer Solutions
stable
φβ’
φβ
φβ’’
metastable
φβ’ φβ φβ’’
φβ’’
φβ
φβ’ instable
φβ − φβ
φβ − φβ
Δfm (φβ ) > Δf m (φ ) + Δfm (φβ ), (2.49)
φβ − φβ β
φβ − φβ
as
Δf (φβ − h ) − Δf (φβ ) Δf (φβ + h ) − Δf (φβ )
+ <0 (2.51)
h h
but that means the curvature has to be negative or locally
∂ 2 Δf (φβ )
< 0. (2.52)
∂φ2β
∂2 1 1
0= Δf (φb ) = + − 2χ. (2.53)
∂ 2 φb 1 − φb M φb
In principle, the critical χ value can be found from solving this quadratic
equation for φb and looking for real roots in the range 0 ≤ φβ ≤ 1. Here,
however, a simpler strategy can be applied. At the boundaries of the interval
[0, 1] of possible φb -values, the second derivative is positive
∂2 1
Δf (φb ) → >0 for φb → 0, (2.54)
∂ 2 φb M φb
∂2 1
Δf (φb ) → > 0 for φb → 1. (2.55)
2
∂ φb 1 − φb
Hence, we look for a minimum of the second derivative, that is, we solve
∂3 1 1
0= Δf (φb ) = − . (2.56)
∂ 3 φb (1 − φb )2 M φ2b
∂2 ∂3
2
Δf (φb ) = 3 Δf (φb ) = 0. (2.58)
∂ φb ∂ φb
Inserting φbc into the second derivative gives
1 2
0=1+ + √ − 2χ, (2.59)
M M
which determines the critical value of χ
√
(1 + M )2 1 1 1
χc = = +√ + . (2.60)
2M 2 M 2M
this gives for the vapor pressure (Figs. 2.10 and 2.11)
pα
= e(μα −μα )/kB T = (1 − φβ )eχφβ +(1−1/M)φβ
0 2
(2.66)
p0α
and since the exponential is a monotonous function, another condition for the
critical point is
∂ pα ∂ 2 pα
= = 0. (2.67)
∂φβ p0α ∂ 2 φβ p0α
1.5 χ= 3
P/P0
1
χ= 2
χ= 1
0.5
χ= 0
0
0 0.2 0.4 0.6 0.8 1
φβ
1.0001
1.0000
0.9999
p/p0
0.9998
0.9997
0.9996
0.9995
0 0.02 0.04 0.06 0.08 0.1
Φβ
Fig. 2.11. Vapor pressure of a polymer solution with interaction for M = 1,000, χ =
0.5, 0.532, 0.55
T 0 χ0
χ= (2.68)
T
and discuss the free energy change as a function of φβ and T :
φβ
ΔF = N kB T (1 − φβ ) ln(1 − φβ ) + ln φβ + N kB T0 χ0 φβ (1 − φβ ).
M
(2.69)
This defines the spinodal curve which separates the instable from the meta-
stable region (Fig. 2.12).
Which is the minimum free energy of a two-phase system? The free energy
takes the form
with
M Nβj
N j = Nαj + M Nβj , φjβ = . (2.73)
Nj
32 2 Flory–Huggins Theory for Biopolymer Solutions
0.8
0.6
φβ
0.4
0
0 0.5 1 1.5 2
T / (T0χ0)
The minimum free energy can be found from the condition that exchange of
solvent or solute molecules between the two phases does not change the free
energy, that is, the chemical potentials in the two phases are the same:
∂ΔF 1 ∂ΔF 2 ∂ΔF 1 ∂ΔF 2
0 = dΔF = − dNα + − dNβ , (2.74)
∂Nα1 ∂Nα2 ∂Nβ1 ∂Nβ2
and hence the slope of the free energy has to be the same for both phases.
Inserting into the first equation then gives
Δf 1 − Δf 2 = (φ1β − φ2β )Δf , (2.79)
which shows that the concentrations of the two phases can be found by the
well-known “common tangent” construction (Fig. 2.13).
These so-called binodal points give the border to the stable one-phase
region. Between spinodal and binodal the system is metastable. It is stable in
relation to small fluctuations since the curvature of the free energy is positive.
It is, however, instable against larger-scale fluctuations (Fig. 2.14).
2.3 Phase Transitions 33
2
φβ
1
φβ
stable one−phase
meta meta
stable stable
stable
two−phase
spinodal binodal
φβ
Problems
2.1. Osmotic Pressure of a Polymer Solution
Polymer solution Pure Solvent
μ1(P,T) = μ01(P’,T)
∂μ0α
μ0α (P , T ) − μ0α (P, T ) = −Π .
∂P
Taylor series expansion gives the virial expansion of the osmotic pressure
as a series of powers of φβ . Truncate after the quadratic term and discuss
qualitatively the dependence on the interaction parameter χ (and hence also
on temperature).
Since an explicit treatment of all solvent atoms and ions is not possible in
many cases, the effect of the solvent on the protein has to be approximated
by implicit models which replace the large number of dynamic solvent modes
by a continuous medium [8–10].
e−U(X,Y )
P (X, Y ) = , (3.1)
dXdY e−U (X,Y )
where X stands for the coordinates of the protein (including the protonation
state) and Y stands for the coordinates of the solvent. The potential energy
can be formally split into three terms:
The mean value of a physical quantity which depends only on the protein
coordinates Q(X) is
Q = dXdY, Q(X)P (X, Y ) = dXQ(X)P(X), (3.3)
e−W (X)/kB T
P (X) = (3.5)
dXe−W (X)/kB T
with
e−W (X)/kB T = e−Uprot (X)/kB T dY e−(Usolv (Y )+Uprot,solv (X,Y ))/kB T
ΔW accounts implicitly but exactly for the solvents effect on the protein.
and thermodynamic integration gives the change of free energy due to Coulom-
bic interactions:
⎛ ⎞
1 qi qj
ΔFelec = λdλ ⎝ + fij qj qi ⎠
0 4πεrij ij
i=j
1 qi qj 1
= + fij qj qi . (3.10)
2 4πεrij 2 ij
i=j
The first part is a property of the protein and hence included in Uprot . The
second part is the mean force potential:
1
ΔWelec = fij qi qj . (3.11)
2 ij
3.3 Born Model 39
Ions
N N N N N
O O O O O
dipolar residues
Asparagine Glutamine
Threonine Serine N
N O
OH OH O
N N
N
O O N
Methionine O
Cysteine O
SH S
N N
O O
polarizable residues
Tyrosine
H
Phenylalanine Tryptophan
O
N
N N N
O O O
Fig. 3.1. Charged, polar, and polarizable groups in proteins. Biological macro-
molecules contain chemical compounds with certain electrostatic properties. These
are often modeled using localized electric multipoles (partial charges, dipoles, etc.)
and polarizabilities
q
ε1
r
ε2
a
Θ
−q +q R
−s s
ε1
ε2
The general solution can be written as the sum of a special solution and a
harmonic function. The special solution is given by the multipole expansion:
q 1 s
n
∞
q
= Pn (cos θ). (3.18)
4πε1 |r − s+ | 4πε1 r n=0 r
Since the potential has to be finite at large distances, outside it has the form
∞
φ2 = Bn r−(n+1) Pn (cos θ) (3.19)
n=0
which for ε2
ε1 is approximately
q2 1 1 1 s
2n q2 1 1 1 1
− =− − . (3.28)
2 4πR ε2 ε1 R 2 4πR ε1 ε2 1 − s2 /R2
φR = φR R
+ + φ− (3.29)
−q 2 1 1 1 1
+ qφR (s) + qφR (s) + (−q)φR R
+ (−s) + (−q)φ− (−s). (3.30)
4πε1 (2s) 2 + 2 − 2 2
By comparison we find
q2 1 q2 1 1 1 1
f++ = qφR (s) = − − , (3.31)
2 2 +
2 4πR ε1 ε2 1 − s2 /R2
(−q)2 1 q2 1 1 1 1
f−− = (−q)φR − (−s) = − − , (3.32)
2 2 2 4πR ε1 ε2 1 − s2 /R2
(−q)q 1
f−+ = (−q)φR + (−s)
2 2
∞
(−q) qsn −1−2n (ε1 − ε2 )(n + 1)
= R (−s)n
2 n=1 4πε1 nε1 + (n + 1)ε2
∞ s
2n
−q 2 1 1 1
≈ − (−)n
2 4πR ε2 ε1 n=0 R
2
(−q ) 1 1 1 1
=− − , (3.33)
2 4πR ε1 ε2 s /R2 + 1
2
q(−q) 1 (−q)q
f+− = qφR − (s) = f−+ , (3.34)
2 2 2
3.5 Generalized Born Models 43
1 1 1 1 1 1 1 1
Welec ≈ −q 2
− − (−q )
2
−
4πR ε1 ε2 1 − s /R
2 2 4πR ε1 ε2 s /R2 + 1
2
2
(2qs) 1 1 1
=− − . (3.36)
8πR ε1 ε2 1 − s4 /R4
(2) p2 (ε1 − ε2 )
Welec = . (3.38)
4πε1 R3 ε1 + 2ε2
where the effective Born radius ai accounts for the effect of neighboring solute
atoms. Several methods have been developed to calculate appropriate values
of the ai [13–16].
1
This has been associated with several names (Bell, Onsager, and Kirkwood).
44 3 Implicit Continuum Solvent Models
and the sum of the individual Born energies plus the change in Coulombic
energy
⎛ ⎞
1 1 1 ⎝ qi2 qi qj ⎠ √
Welec → − + for rij
ai aj (3.42)
8π ε2 ε1 i
a i rij
i=j
Mobile charges like Na+ or Cl− ions are very important for the functioning
of Biomolecules. In a stationary state, their concentration depends on the
local electrostatic field which is produced by the charge distribution of the
Biomolecule, which in turn depends for instance on the protonation state
and on the conformation. In this chapter, we present continuum models
to describe the interaction between a Biomolecule and surrounding mobile
charges [17–20].
Debye and Hückel [21] used Boltzmann’s theorem to determine the mobile
charge density. Without the presence of fixed charges, the system is neutral:
0 = 0mob = Zi eNi0 (4.3)
i
φ0 = 0. (4.4)
The fixed charges produce a change of the potential. The electrostatic energy
of an ion of sort i is
46 4 Debye–Hückel Theory
Wi = Zi eφ(r) (4.5)
and the density of such ions is given by a Boltzmann distribution:
If the solution is very dilute, we can expect that the ion–ion interaction is
much smaller than thermal energy:
Zi eφ kB T (4.10)
e e2 e
Δφ(r) + N 0 (e−eφ(r)/kB T − eeφ(r)/kB T ) = − fix (r) (4.14)
kB T εkB T εkB T
4.3 Charged Sphere 47
c1 e−κR Q
φ1 (R) = φ2 (R) → c3 = − , (4.21)
R 4πε1 R
−κR
∂ ∂ Q e e−κR
ε1 φ1 (R) = ε2 φ2 (R) → − = c 1 ε 2 − − κ , (4.22)
∂r ∂r 4πR2 R2 R
R a
ε1
ε2
κ
10
1 κ = 0.1
potential Φ κ = 1.0
0.1
κ = 2.0
0.01
0 2 4
r
Fig. 4.2. Charged sphere in an electrolyte. The potential φ(r) is shown for a = 0.1,
R = 1.0
and
Q Q
c3 = − + . (4.24)
4πε1 R 4πε2 R(1 + κR)
Together, we find the potential inside the sphere
Q 1 1 Q
φ1 (r) = − + (4.25)
4πε1 r R 4πε2 R(1 + κR)
Q e−κ(r−R)
φ2 (r) = . (4.26)
4πε2 (1 + κR) r
0.6 κ = 2.0
0.5
charge density
κ = 0.5
0.4
0.3 κ = 1.0
0.2
0.1
0
1 1.5 2 2.5 3 3.5 4
r
Fig. 4.3. Charge density around the charged sphere for a = 0.1, R = 1.0
Let the charge be concentrated on the surface of the inner sphere. Then we
have
Q 1 1 Q
φ1 (a) = − + . (4.29)
4πε1 a R 4πε2 R(1 + κR)
Q2 κ
ΔGsol = W = − . (4.33)
8πε (1 + κR)
r
a
z
d2 1 d
φ(x) + φ(x) − φ(x) = 0. (4.37)
dx2 x dx
Solutions of this equation are the modified Bessel functions of order zero
denoted as I0 (x) and K0 (x). For large values of x,
and hence the potential in the outer region has the form
Inside the cylinder surface, the electric field is given by Gauss’ theorem:
dφ(r) Ne
E(r) = − = , (4.41)
dr 2πε1 rl
and hence the potential inside is
Ne
φ(r) = C2 − ln r. (4.42)
2πε1 l
4.4 Charged Cylinder 51
φ(r) = − γ + ln + ln r (4.51)
2πlε2 2
and inside is
Ne 2 Ne Ne
φ(r) = ln −γ + ln a − ln r
2πlε2 κa 2πε1 l 2πε1 l
Ne γ 1 κ 1 1 1
= − − ln + − ln a − ln r . (4.52)
2πl ε2 ε2 2 ε1 ε2 ε1
Outside the potential consists of the potential of the charged line (ln r) and
an additional contribution from the screening of the ions (Fig. 4.5).
52 4 Debye–Hückel Theory
1
2 4 6 8 10
r
Fig. 4.5. Potential of a charged cylinder with unit radius a = 1 for κ = 0.05. Solid
curve: K0 (κr)/(κK1 (κ)). Dashed curve: approximation by − ln(κ/2) − ln r − γ
+ − − −
+ − +
+ −
+ −
+ − +
+
+ − −
+ − −
+ −
+
+ −
− +
+ +
+ − −
+ − −
d2
f (x) − f (x) = 0 (4.57)
dx2
with the solution
f (x) = f0 e±x (4.58)
or going back to the potential
kB T
φ(x) = f0 e±κx = φ0 e±κx . (4.59)
e
The membrane potential is related to the surface charge density. Let us assume
that on the left side (x < 0), the medium has a dielectric constant of ε1 and
on the right side ε. Since in one dimension the field in a dielectric medium
does not decay, we introduce a shielding constant κ1 on the left side and take
the limit κ1 → 0 to remove contributions not related to the membrane charge.
The potential then is given by
φ0 e−κx x > 0,
φ(x) = (4.60)
φ0 eκ1 x x < 0.
a φ(x)
φo
x
b
ρ(x)
Fig. 4.7. Electrolytic double layer. (a) The potential is continuous and decreases
exponentially in the electrolyte. (b) The charge density has a sharp peak on the
membrane which is compensated by the exponentially decreasing charge in the
electrolyte
1
co-ions
Φ
0
total
–1
counter-ions
–2
–3
0 1 2 3 4 5
r
Fig. 4.8. Charge density of counter- and co-ions. For an exponentially decaying
potential Φ(r) (full curve), the density of counter- and co-ions (dashed curves) and
the total charge density (dash-dotted curve) are shown
we find
d f (x)
f (x) = ±2 sinh . (4.72)
dx 2
Separation of variables then gives
df
= ±dx (4.73)
2 sinh(f /2)
For f (0) > 0, the argument of the logarithm becomes negative. Hence, we
replace in (4.74) C by C + iπ to have (Fig. 4.9)
x C iπ x C
f (x) = 2 ln tanh + + = −2 ln tanh + , (4.76)
2 2 2 2 2
1
We consider only solutions for the region x > 0 in the following.
56 4 Debye–Hückel Theory
a b
1 5
0.8 4
0.6 3
f(x)
0.4 2
0.2 1
0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
x x
a b
15
1
10
f’’(x)
0.5
5
0 0
0 1 2 3 4 5 0 1 2 3 4 5
x x
and especially
C 1
f (0) = tanh − (4.82)
2 tanh(C/2)
and we have to solve the equation
1 eσ0
t− =− = B, (4.83)
t kB T κε
which yields2 √
B B2 + 4
t= + , (4.84)
2 2
√
B B2 + 4
C = 2arctanh + . (4.85)
2 2
φ(x)
φM
φ0
d x
+ −
+ − − + −
+ −
+
+ − − +
+
+ −
+
+ − − − − +
2
The second root leads to imaginary values.
58 4 Debye–Hückel Theory
Within the Stern layer of thickness d, there are no charges and the potential
drops linearly from the membrane potential φM to a value φ0 :
φM − φ0
φ(x) = φM − x, 0 < x < d. (4.86)
d
In the diffusive Goüy–Chapman layer, the potential decays exponentially
Assuming the same dielectric constant for both layers, we have to fulfill the
boundary condition
dφ φM − φ0
(d) = − = −κφ0 (4.88)
dx d
and hence
φM
φ0 = . (4.89)
1 + κd
The total ion charge in the diffusive layer now is
∞ ∞
qdiff = (x)dx = −Aεκ2 φ0 e−κ(x−a) dx = −Aεκφ0
d d
εκ
= −A φM (4.90)
1 + κd
and the capacity is
−qdiff Aε
C= = , (4.91)
φM d + κ−1
which are just the capacities of the two layers in series
1 Ad AλDebye 1 1
= + = + . (4.92)
C ε ε CStern Cdiff
Problems
4.1. Membrane Potential
Consider a dielectric membrane in an electrolyte with an applied voltage V .
V
I II III
κ κ
εw εm εw
0 L x
4.6 Stern Modification of the Double Layer 59
d2
φ(x) = κ2 (φ(x) − φ(0) )
dx2
with boundary conditions
ΔV = φ(L) − φ(0).
Calculate the charge density mob (x) and the integrated charge on both sides
of the membrane. What is the capacity of the membrane?
Z 2 e2 κ
μ − μideal = kB T ln a − kB T ln c = Δ(μα − μ0α )Gsolv = − .
8πε (1 + κR)
O O
Aspartic Acid
C C
O− Glutamic Acid
OH
NH2 NH+3
C C Arginine
NH NH
N NH+ Histidine
H2O H2O
H2O
H2O H2O H2O
H2O H2O
+
H3O H2O
H2O H2O
H2O H2O
H2O
O −
OH
R C
H2O H2O
OH H2O
H2O
H2O
H2O H2O H2O
H2O
and the free enthalpy difference between protonated and deprotonated form
of the molecule is
z1
G1 − G0 = −kB T ln + pΔV. (5.3)
z0
In the following, the volume change will be neglected.1 If we now put such
N molecules into the solution, the number of protonated molecules can fluc-
tuate by exchanging protons with the solvent. Removal of one proton from
the solvent costs a free enthalpy of
ΔG = −μH+ , (5.4)
1
At atmospheric pressure, the mechanic work term pΔV is very small. We pre-
fer to discuss the free enthalpy G in the following since experimentally usually
temperature and pressure are constant.
2
We omit the index H+ in the following.
5.1 Protonation Equilibria in Solution 63
0.8
protonation degree
0.6
0.4
0.2
0
–4 –2 0 2 4
μ − ΔGp
∂ z1 eμ/kB T
M= ln Ξ = N (5.8)
∂(μ/kB T ) z0 + z1 eμ/kB T
and the protonation degree, that is, the fraction of protonated molecules is
(Fig. 5.3)
M 1 1
= = . (5.9)
N 1 + (z0 /z1 )e−μ/kB T 1 + e(G1 −G0 −μ)/kB T
In physical chemistry [27, 28], the following quantities are usually introduced:
• The activity of a species ai is defined by
μi = μ0i + kB T ln ai (5.10)
in analogy to μi = μ0i + kB T ln pi /p0i for the ideal gas. For very dilute
solutions, it can be approximated by
ci
μi = kB T ln Ni − kB T ln zi = μ0i + kB T ln , (5.11)
c0
where (0) indicates the standard state (usually p0 = 1 atm, c0 = 1 mol/l).
• The concentration of protons is measured by the pH value3
μH+ c(H+ )
pH = − log10 aH+ = − ≈ − log10 . (5.12)
kB T ln(10) c0
• The standard reaction enthalpy of an acid–base equilibrium
AH A− + H+ , (5.13)
3
The standard enthalpy of formation for a proton is zero per definition.
64 5 Protonation Equilibria
where
0= νi μi = ΔG0r + kB T νi ln ai , (5.14)
a(A− )a(H+ )
Ka = e−ΔGr /kB T =
0
, (5.15)
a(AH)
is measured by the pKa value:
1 ΔG0r a(A− )a(H+ )
pKa = − log10 (Ka ) = = − log10
ln(10) kB T a(AH)
⎛ − ⎞
c(A ) c(H+ )
≈ − log10 ⎝ c0c(AH)c0 ⎠ , (5.16)
c0
4
But you have to be aware that all concentrations have to be taken in units of the
standard concentration c0 . The argument of a logarithm should be dimensionless.
5.2 Protonation Equilibria in Proteins 65
H2O
H2O
H2O
ΔGsol – +
RMH RM + H
ΔG1 ΔG0
PMH PM−+ H+
ΔGprot
In the following, the volume change will be neglected. The pKa value of a
group in the protein
1 1
pKaprot = ΔGprot = (ΔGsolv + ΔΔG)
kB T ln(10) kB T ln(10)
ΔΔG
= pKamodel + (5.23)
kB T ln(10)
Npstates = 2N
can be very big for real proteins.5 Proteins are very flexible and have a
large number of configurations. These will be denoted by the symbol γ which
summarizes all the orientational angles between neighboring residues of the
protein. The apparent pKa values will in general depend on this configura-
tion vector, since for instance distances between the residues depend on the
configuration.
The enthalpy change by protonating the ith residue is denoted as
5
We do not take into account that some residues can be in more than two
protonation states.
5.2 Protonation Equilibria in Proteins 67
s si = 0 s
i−1 i+1
s si = 1 s
i−1 i+1
Fig. 5.6. Protonation of one residue. The protonation state of the ith residue
changes together with its charge. Electrostatic interactions are divided into an
intrinsic part and the interactions with all other residues
6
That is, the charge distribution of the protein backbone and the nontitrable
residues.
68 5 Protonation Equilibria
AH A–+ H+ BH+ B + H+
s 1 0 1 0
0
s 1 1 0 0
0
q = s−s 0 −1 +1 0
qc = 1−2s0 −1 −1 1 1
Fig. 5.7. Protonation states. Protonation state si , protonation of the neutral state
s0i , charge qi , and charge of the non-neutral state qic are correlated
Note that Wij = Wji and therefore the last sum can be simplified
1 & '
Wi,j (si − s0i )(sj − s0j ) + s0i sj + si s0j − 2s0i s0j
2 i
j=i
1
= Wi,j (qi qj + 2(si − s0i )s0j ). (5.35)
2 i
j=i
5.2 Protonation Equilibria in Proteins 69
N
N
1,0 0,0
ΔW = qi (Ei,j − Ei,j + s0j Wi,j ). (5.36)
i=1 j=i,j=1
is given by
Z(Q) = e−ΔG/kB T
{s, qi =Q} γ
= e −[ (si −s0i )ΔGi,intr + 12 Wi,j qi qj ]/kB T
. (5.42)
{s, qi =Q} γ
7
The Coulombic interaction vanishes if one of the residues is in the neutral state.
70 5 Protonation Equilibria
becomes the product of one factor which relates to the internal degrees of
freedom which are usually assumed to be configuration-independent8 and a
second factor which depends only on the configurational degrees of freedom:
e−[ (si −si )(ΔEi,bg +ΔEi,Born )+ 2
0 1
Wi,j qi qj ]/kB T
Zconf (s) = . (5.46)
γ
0.8
protonation degree
0.6
0.4
0.2
0
–10 0 10
μ
Fig. 5.8. Abnormal titration curves. Two interacting residues with neutral proto-
nated states (AH), ΔG1,intr = ΔG2,intr = 1.0, W = −5, 0, 5, 10
8
Protonation-dependent degrees of freedom can be important in certain cases [30].
5.3 Abnormal Titration Curves of Coupled Residues 71
1 + e−(−ΔG2,intr +μ)/kB T
s1 = ,
Ξ
1 + e−(−ΔG1,intr +μ)/kB T
s2 = . (5.49)
Ξ
Problems
5.1. Abnormal Titration Curves
Consider a simple example of a model protein with only two titrable sites
of the same type. Determine the relative free enthalpies of the four possible
states.
(B,B)
+
+2H
B B
(B,BH+)
B +H+
BH+
(BH+,B)
BH+ B +H+
(BH+,BH+)
BH+ BH+
From the grand canonical partition function (the number of protons is not
fixed), calculate the protonation degree for both sites and discuss them as a
function of the interaction energy W .
Part III
Reaction Kinetics
6
Formal Kinetics
H2 + J2 2HJ. (6.1)
A + hν A* A* + B A + B* bonds are
absorption energy transfer conserved
reorganization
D*A D+ A– D+A– B D+A B– of bond lengths
charge separation charge transfer and angles
A−H + B A− + B−H+
proton transfer
bimolecular
H− Cl H Cl H− Br
+ +
bonds are
Br−Br Br Br Br− Cl
broken or
activated
complex formed
CH2
CH –3 CH CH2 unimolecular
CH2 CH2
isomerization
as
ci − ci,0
x= (6.4)
νi
and the reaction rate as its time derivative
dx 1 dci
r= = . (6.5)
dt νi dt
with the rate constant k. For such a system, the exponent2 of the ith term is
called the order of the reaction with respect to this substance and the sum of
all the exponents is called the overall reaction order.
2
For more complicated reactions, the exponents need not be integers. For simple
reactions, they are given by the stoichiometric coefficients ni = |νi | of the educts
(the products for the back reaction).
6.3 Reaction Order 77
A+B → · · · (6.13)
1 b(a − x)
ln = k2 t. (6.17)
a − b a(b − x)
If two molecules of the same type react with each other, we have instead
dcA
− = −k2 c2A (6.18)
dt
which gives an algebraic decay
1
cA (t) = 1 , (6.19)
k2 t + a
k1
→
A B. (6.22)
←
k−1
The reaction variable of the back reaction will be denoted by y. The concen-
trations are
cA (t) = a − x + y, (6.23)
cB (t) = b + x − y, (6.24)
z =x−y (6.27)
6.6 Consecutive Reactions 79
k1 k2
A → B → C. (6.35)
cA = a − x, (6.36)
3
With negligible back reactions
80 6 Formal Kinetics
cB = b + x − y, (6.37)
cC = c + y, (6.38)
and their time derivatives are
dcA dx
=− = −k1 cA = −k1 (a − x), (6.39)
dt dt
dcB dx dy
= − = k1 cA − k2 cB = k1 a − k2 b + (k2 − k1 )x − k2 y, (6.40)
dt dt dt
dcC dy
= = k2 cB = k2 (b + x − y). (6.41)
dt dt
The first equation gives an exponential decay:
cA = ae−k1 t . (6.42)
Integration of
dcB
+ k2 cB = k1 ae−k1 t (6.43)
dt
gives the concentration of the intermediate state:
k1 a −k1 t k1 a
cB = e + b− e−k2 t . (6.44)
k2 − k1 k2 − k1
k1
E + S ES E + P. (6.47)
k−1
6.7 Enzymatic Catalysis 81
we find
c0E
cES (t) = (1 − e−(k1 +k−1 +k2 )t ),
1 + Kc0M
S
c0E c0S −(k1 +k−1 +k2 )t
cE (t) = c0
1 + e . (6.55)
1 + KSM KM
82 6 Formal Kinetics
rmax=k2cE,tot
1
reaction rate r
cSk2cE,tot/KM
0.5
0
0 0.5 1
cS
The stationary state is stable, since any deviation will decrease exponentially.
The overall rate of the enzyme-catalyzed reaction is given by the rate of
product formation:
dcP dcS
r= =− = k2 cES (6.56)
dt dt
and with the total concentration of enzyme
we have
c E cS (cE,tot − cES )cS
cES = = (6.58)
KM KM
and hence
cE,tot cS
cES = . (6.59)
cS + K M
The overall reaction rate is given by the Michaelis–Menten equation (Fig. 6.3)
k2 cE,tot cS
r= , (6.60)
K M + cS
r cS
= , rmax = k2 cE,tot . (6.61)
rmax cS + K M
B
A
solvent cage
A B
a b c
reaction rate
0.1
0.01
0 5 0 5 0 5
time
k2 c{AB} ≈ k1 cA cB (6.64)
and hence for the overall reaction rate
dcC
= k2 c{AB} ≈ k1 cA cB . (6.65)
dt
The reaction rate is determined by the formation of the reactive complex.
If, on the other hand k2 k±1 , an equilibrium between reactants and reactive
complex will be established
c{AB} k1
A + B { A B}, =K= . (6.66)
cA cB k−1
Now, the overall reaction rate
dcC
= k2 c{AB} = k2 KcA cB (6.67)
dt
is determined by the reaction rate k2 and the constant of the diffusion
equilibrium.
Problems
6.1. pH Dependence of Enzyme Activity
Consider an enzymatic reaction where the substrate can be in two protonation
states
S− + H+ HS
and the enzyme reacts only with the deprotonated form
E + S− ES− → E + P.
c2M = Kc2M ,
k1
A+B X → C + D.
Zi2 e2 κ e2
μi = μ0i + kB T ln ci − , κ2 = Ni Zi2 .
8π kB T
Calculate the reaction rate
dcC
r= .
dt
7
Kinetic Theory: Fokker–Planck Equation
The average of the stochastic force has to be zero because the equation of
motion for the average velocity should be
88 7 Kinetic Theory: Fokker–Planck Equation
dv(t)
= −γv(t). (7.4)
dt
We assume that many collisions occur during τ and therefore forces at different
times are not correlated:
mv 2 kB T mC 2γkB T
= = →C= . (7.9)
2 2 2 2γ m
The mean square displacement of a particle starting at x0 with velocity v0 is
( 2 ) t t
t
(x(t) − x(0))2 = dt1 v(t1 ) = v(t1 )v(t2 )dt1 dt2
0 0 0
t t
2 −γ(t1 +t2 ) kB T −γ|t1 −t2 |
= v0 e + e dt (7.10)
0 0 m
and since t 2
t
−γ(t1 +t2 ) 1 − e−γt
e dt1 dt2 = (7.11)
0 0 γ
7.2 Probability Distribution 89
and
t t t t1
−γ|t1 −t2 |
e dt1 dt2 = 2 dt1 e−γ(t1 −t2 ) dt2
0 0 0 0
2 2
= t − 2 (1 − e−γt ), (7.12)
γ γ
we obtain
kB T (1 − e−γt )
2
2kB T 2kB T
(x(t) − x(0)) = v0 −
2 2
+ t− (1 − e−γt ).
m γ2 mγ mγ 2
(7.13)
If we had started with an initial velocity distribution for the stationary state
then the first term in (7.12) would vanish. For very large times, the leading
term is1
(x(t) − x(0))2 = 2Dt (7.15)
with the diffusion coefficient
kB T
D= . (7.16)
mγ
To derive an expression for the differential ∂W (v, t)/∂t, we need the transition
probability P (v, t + τ |v , t) for small τ . Introducing Δ = v − v , we expand the
integrand in a Taylor series
we find n
∞
(−1)n ∂
W (v, t + τ ) = Mn (v, t, τ )W (v, t). (7.21)
n=0
n! ∂v
Expanding the moments into a Taylor series
1 1
Mn (v, t, τ ) = Mn (v, t, 0) + D(n) (v, t)τ + · · · , (7.22)
n! n!
we have finally2
∞
n
∂
W (v, t + τ ) − W (v, t) = − D(n) (v, t)W (v, t)τ + · · · , (7.23)
1
∂v
If this expansion stops after the second term,4 the general form of the one-
dimensional Fokker–Planck equation results
∂W (v, t) ∂ (1) ∂ 2 (2)
= − D (v, t) + 2 D (v, t) W (v, t) (7.25)
∂t ∂v ∂x
7.3 Diffusion
Consider a particle performing a random walk in one dimension due to
collisions. We use the stochastic differential equation5
dx
= v0 + f (t), (7.26)
dt
where the velocity has a drift component v0 and a fluctuating part f (t) with
gives
D(1) = v0 . (7.30)
The second moment is
M2 (x0 , t, τ )
τ τ τ
= v02 τ 2 + v0 τ f (t )dt + f (t1 )f (t2 )dt1 dt2 . (7.31)
0 0 0
The second term vanishes and the only linear term in τ comes from the double
integral
τ τ τ τ −t1
f (t1 )f (t2 )dt1 dt2 = dt1 qδ(t )dt = qτ. (7.32)
0 0 0 −t1
Hence
q
D(2) = (7.33)
2
and the corresponding Fokker–Planck equation is the diffusion equation
We can easily find the solution for a sharp initial distribution W (x, 0) =
δ(x − x0 ) by taking the Fourier transform:
∞
* (k, t) =
W dx W (x, t) e−ikx . (7.35)
−∞
* (k, t)
∂W * (k, t)
= (−Dk 2 + iv0 k)W (7.36)
∂t
92 7 Kinetic Theory: Fokker–Planck Equation
0.5
0.4
0.3
0.2
0.1
0
–4
–2 1
0 2
2
x 4 3 t
6
8 4
10
12 5
Fig. 7.1. Solution (7.38) of the diffusion equation for sharp initial conditions
which is solved by
+ ,
* (k, t) = W
W *0 exp (−Dk 2 + iv0 k)t + ikx0 . (7.37)
6
with the proper normalization factor
7.4 Fokker–Planck Equation for Brownian Motion 93
a b
1
2
0.9
1 0.8
W(x,t)
0.7
CA(t)
0
0.6
–1 0.5
0.4
–2
0.3
–2 0 2 0 2 4 6 8 10
x t
Fig. 7.2. Solution from the mirror principle. (a) The probability density distribution
(7.42) at t = 0.25, 0.5, 1, 2, 5 and (b) the decay of the total concentration (7.43) are
shown for 4D = 1
W (x = 0, t) = 0. (7.40)
7
Here and in the following, we use F (t) = 0.
94 7 Kinetic Theory: Fokker–Planck Equation
gives
D(1) (v, t) = −γv. (7.46)
The second moment follows from
(v(τ ) − v0 )2
τ τ
= (v0 γτ )2 + (1 + γ(t1 + t2 − 2τ · · · ))F (t1 )F (t2 )dt1 dt2 . (7.47)
0 0
∂y ∂2y
= γy + De2γt 2 . (7.58)
∂t ∂ρ
To solve this equation, we introduce new variables again
y = χeγt , (7.59)
which results in
∂χ ∂2χ
= De2γt 2 . (7.60)
∂t ∂ρ
Now, we introduce a new time scale
1 2γt
θ= (e − 1), (7.61)
2γ
with
dvdx e−(mv
2
/2+U (x))/kB T
Z= . (7.78)
where Wn is the probability to find the integer value n and wm→n is the
transition probability. For continuous x, the summation has to be replaced by
an integration
∂W (x, t)
= (wx →x W (x , t) − wx→x W (x, t)) dx . (7.86)
∂t
The Fokker–Planck equation is a special form of the master equation with
∂ ∂2
wx →x = − D(1) (x) + 2 D(2) (x) δ(x − x ). (7.87)
∂x ∂x
So far, we have discussed only Markov processes where the change of proba-
bility at time t only depends on the probability at time t. If memory effects
are included, the generalized Master equation results.
Problems
7.1. Smoluchowski Equation
Consider a one-dimensional random walk. At times tn = nΔt, a particle at
position xj = jΔx jumps either to the left side j − 1 with probability wj− or
7.8 Master Equation 99
to the right side j + 1 with probability wj+ = 1 − wj− . The probability to find
a particle at site j at the time n + 1 is then given by
+ −
Pn+1,j = wj−1 Pn,j−1 + wj+1 Pn,j+1 .
Show that in the limit of small Δx, Δt, the probability distribution P (t, x)
obeys a Smoluchowski equation
w−j +
wj
x
j−1 j j+1
kA km
kB km
dPN (t)
= − [(kA + kB )(M − N ) + 2km N ] PN (t)
dt
+ (kA + kB )(M − N + 1)PN −1 (t) + 2km (N + 1)PN +1 (t). (7.88)
2
σ2 = N 2 − N
and the mean diffusion current
dNB dNA
JAB = − ,
dt dt
where NA,B is the number of ions in the upper of lower half-space.
8
Kramers’ Theory
We assume that the perturbation due to the reaction is small and make the
following ansatz (Fig. 8.2)
U(x)
A B C
U*
UA
x0 0 reaction coordinate x
Wstat
y
x0 0
Fig. 8.2. Ansatz function. The stationary solution of a harmonic well is multiplied
with the function y(x, v) which switches from 1 to 0 at the saddle point
2. On the right side of the barrier (x > 0), all particles A are converted into
C and there will be no particles of the A species here
Let us now insert the ansatz function into the Klein–Kramers equation. For
the reversible part, we have
H H
Lrev exp − y(x, v) = exp − Lrev y(x, v) (8.9)
kB T kB T
8.2 Kramers’ Calculation of the Reaction Rate 103
With
∂ ∂ ∂ ∂ ∂
x = η, v = ξ + λη, = −λ , = , (8.14)
∂x ∂η ∂ξ ∂v ∂ξ
(8.12) is transformed to
∂ ∂ γkB T ∂ 2
0 = (ξ + λη) y + ((λ − γ)ξ + (λ2 − λγ − ω 2 )η) y + y. (8.15)
∂η ∂ξ m ∂ξ 2
Now choose
γ γ2
λ= + + ω ∗2 (8.16)
2 4
to have the simplified equation
∂ ∂ γkB T ∂ 2
0 = (ξ + λη) y + ((λ − γ)ξ) y + y (8.17)
∂η ∂ξ m ∂ξ 2
which obviously has solutions which depend only on ξ and obey
∂ γkB T ∂ 2
ξ Φ(ξ) = − Φ(ξ). (8.18)
∂ξ m(λ − γ) ∂ξ 2
m 2 # % $
mωa v + U (x) m(λ − γ)
= C2 exp − 2 1 + erf (v − λx)
2πkB T kB T 2γkB T
(8.21)
and the flux over the barrier is approximately
S = dv vW (0, v)
# % $
mωa −U (x∗ )/kB T −mv 2 /2kB T m(λ − γ)
= C2 e dv ve 1 + erf v
2πkB T 2γkB T
mωa −U (0)/kB T 2kB T γ
= C2 e 1− . (8.22)
2πkB T m λ
In the A-well, we have approximately
# mωa2 (x−xa )2
$
mv 2
mωa 2 +
W (x, v) ≈ 2C2 exp − 2
. (8.23)
2πkB T kB T
The total concentration is approximately
[A] = dx dvWA (x, v) = 2C2 . (8.24)
Hence, we find
ωa −U(0)/kB T γ
e
S = [A] 1− . (8.25)
2π λ
The square root can be written as
! γ 1 γ2
! !− + ∗2
!1 − γ ! 2 4 +ω
" 1 =" 1
γ2 γ2
γ
2 + 4 + ω ∗2 γ
2 + 4 +ω
∗2
! 1 2
!
! − γ + γ 2 + ω ∗2
! 2 4
=!" 2
2
− γ4 + γ4 + ω ∗2
1
2
− γ2 + γ4 + ω ∗2
= (8.26)
ω∗
8.2 Kramers’ Calculation of the Reaction Rate 105
The bracket is Kramers’ correction to the escape rate. In the limit of high
friction, series expansion in γ −1 gives
ωa ω ∗ −U(0)/kB T
k = γ −1 e . (8.28)
2π
In the limit of low friction, the result is
ωa −U(0)/kB T
k= e . (8.29)
2π
9
Dispersive Kinetics
k
D*
D+A−
k −1
kda
kcr
DA
Fig. 9.1. Electron transfer in a fluctuating medium. The rates are time dependent
master equation:
d W (+) −α β W (+)
= , (9.2)
dt W (−) α −β W (−)
W (+) = C1 + C2 e−(α+β)t ,
α
W (−) = C1 − C2 e−(α+β)t . (9.3)
β
Obviously, the equilibrium values are
β α
Weq (+) = , Weq (−) = (9.4)
α+β α+β
and the correlation function is (with Q± = ±1)
Combination of the two systems of equations (9.1) and (9.2) gives the equation
of motion
d
W = AW (9.6)
dt
for the four-component state vector
⎛ ⎞
W (D∗ , +)
⎜ W (D∗ , −) ⎟
W=⎜ ⎝ W (D+ A− , +) ⎠
⎟ (9.7)
W (D+ A− , −)
Generally, the solution of this equation can be expressed by using the left and
right eigenvectors and the eigenvalues λ of the rate matrix which obey
ARν = λν Rν , (9.9)
Lν A = λν Lν . (9.10)
9.1 Dichotomous Model 109
α
α D+A– (−)
D*(+) D*(–)
+−
k D A (+)
β
In the opposite limit, we expand the square root for small k −1 to find
α+β k 1& '
λ3,4 = − − ± k + (α − β) + 2αβk −1 + · · · , (9.26)
2 2 2
αβ
λ3 = −β + + ··· , (9.27)
k
⎛ ⎞
0
⎜ 1 ⎟ & '
R3 = ⎜ ⎟
⎝ 0 ⎠ , L3 = α k 0 0 , (9.28)
−1
λ4 = −k − α + · · · , (9.29)
⎛ ⎞
k
⎜ −α ⎟ & '
R4 = ⎜ ⎟
⎝ −k ⎠ , L4 = 1 0 0 0 (9.30)
α
and the time evolution is approximately
⎛ ⎞
β
e−kt
α+β
⎜ ⎟
⎜ α
e−βt − βk e−kt ⎟
W(t) = ⎜ ⎟.
α+β
⎜ ⎟ (9.31)
⎝
β
(1 − e−kt )
α+β
⎠
α
α+β 1 − e−βt + βk e−kt
a 10 0 b
10 –1
occupation of the initial state
10 –2
10 –3
c 10 0 d
10 –1
10 –2
10 –3
0 5 10 0 5 10
time
Fig. 9.3. Nonexponential decay. Numerical solutions of (9.12) are shown for α = 0.1,
β = 0.9, (a) k = 0.2, (b) k = 2, (c) k = 5, and (d) k = 10. Dotted curves show the
two components of the initial state and solid curves show the total occupation of
the initial state
The fluctuations of the coordinate X(t) are described as random jumps [34,
35]. The time intervals between the jumps (waiting time) and the coordinate
changes are random variables with independent distribution functions:
X
X(t)
0 t1 t2 t3 t4 t
∞
P (X, t) = P (X, 0) ψ(t )dt
t
t ∞
+ dt dX ψ(t − t )f (X, X )P (X , t ). (9.34)
0 −∞
Two limiting cases are well known from the theory of collisions. The correlated
process with
f (X, X ) = f (X − X ) (9.35)
and
we have
∂P ∂P ∂2P
P (X, t) + Δt + · · · = P (X, t) + (q − p)ΔX + ΔX 2 + · · · (9.39)
∂t ∂X ∂X 2
The leading terms constitute a diffusion equation
∂P ΔX ∂P ΔX 2 ∂ 2 P
P = (q − p) + (9.40)
∂t Δt ∂X Δt ∂X 2
with drift velocity (q − p)(ΔX/Δt) and diffusion constant ΔX 2 /Δt.
The uncorrelated process, on the other hand, with
Coupling of motion along the coordinate X with the reactions gives the
following system of equations [36, 37]:
∂ & '
P (X, t) = −k(X) + L1 − τ1−1 P (X, t) + k−1 (X)C(X, t),
∂t
∂ & '
C(X, t) = −k−1 (X) + L2 − τ2−1 C(X, t) + k(X)P (X, t), (9.55)
∂t
where P (X, t)ΔX and C(X, t)ΔX are the probabilities of finding the system
in the electronic state D∗ and D+ A− , respectively; L1,2 are operators describ-
−1
ing the motion in the two states and the rates τ1,2 account for depopulation
via additional channels. For the uncorrelated Markovian process (9.54), the
rate equations take the form
∂ P (X, t)
∂t C(X, t)
k(X) + τ1−1 + τ −1 −k−1 (X) P (X, t)
=−
−k(X) k−1 (X) + τ2−1 + τ −1 C(X, t)
φ1 (X) P (t)
+τ −1 , (9.56)
φ2 (X) C(t)
116 9 Dispersive Kinetics
R(X, t) = exp(−A(X)t)R(X, 0)
t
−1
+τ exp(A(X)(t − t))B(X)R(t )dt (9.62)
0
R(t) = exp(−At)R(0)
t
+ τ −1 exp(A(t − t))BR(t )dt , (9.63)
0
∞ t
−st
e dt exp(A(t − t))BR(t )dt
0 0
−1
= (s + A) BR̃(s). (9.66)
is solved by
4 5−1
R̃(s) = 1 − τ −1 (s + A)−1 B (s + A)−1 R(0). (9.68)
We assume that initially, the system is in the initial state D∗ and the motion
is equilibrated:
φ1 (X)
R(X, 0) = . (9.69)
0
For simplicity, we treat here only the case of τ12 → ∞. Then we have
k + τ −1 −k−1
A= , (9.70)
−k k−1 + τ −1
−1 1 s + τ −1 + k−1 k−1
(s + A) =
(s + τ )(s + τ −1 + k + k−1 )
−1 k s + τ −1 + k
(9.71)
and
f (X)1,2 = φ1,2 (X)f (X)dX, (9.73)
we find
as well as
(s + A)−1 B
−1 − (s+τ −1 )2 αk1 (s+τ −1 )2 αk−1 2
1 1 1
= s+τ (9.75)
(s+τ −1 )2 αk1 s+τ −1 − (s+τ −1 )2 αk−1 2
1 1 1
(s + τ −1 ) + τ −1 s−1 αk−1 2 s + K2
P (s) = −1 −1
= 2
s + τ s (1 + αk1 + αk−1 2 )
2 s + s(K1 + K2 )
(9.78)
with
k−1 (X)
K2 = τ −1 1 + s−1 dXφ2 (X) (9.79)
1 + s+τ1 −1 (k(X) + k−1 (X))
k−1 (X)
≈ dXφ2 (X) , (9.80)
1 + τ (k(X) + k−1 (X))
k(X)
K1 = dXφ1 (X) . (9.81)
1 + τ (k(X) + k−1 (X))
K2 + K1 e−(K1 +K2 )t K1
P (t) = , C(t) = (1 − e−(K1 +K2 )t ). (9.87)
K1 + K2 K1 + K2
Let us now consider the special case that the back reaction is negligible and
k(X) = kΘ(X). Here, we have
s(s + τ −1 ) − sαk1
P̃ (s) = , (9.88)
s(s2 + τ −1 (s + αk1 ))
9.3 Power Time Law Kinetics 119
k(X)
φ(X)
−1
τ
X
k
∞
k(X) k
αk1 = dXφ1 (X) k(X)
= φ1 (X) k
dX
1+ s+τ −1 0 1+ s+τ −1
∞ 0
s + τ −1
= bk , b= φ1 (X)dX, a = 1−b = φ1 (X)dX, (9.89)
k + s + τ −1 0 −∞
−1
(s + τ −1 − bk k+s+τ
s+τ
−1 ) s + τ −1 + k(1 − b)
P̃ (s) = = . (9.90)
(s2 + τ −1 (s + bk s+τ −1
k+s+τ −1 )) s2 + s(τ −1 + k) + bkτ −1
Inverse Laplace transformation gives a biexponential behavior
with
μ± = τ −1 ± k 2 + τ −2 + 2kτ −1 (1 − 2b). (9.92)
If the fluctuations are slow τ −1 k, then
k 2 + τ −2 + 2kτ −1 (1 − 2b) = k + (1 − 2b)τ −1 + · · · , (9.93)
k + μ+ k + μ−
= k + ··· , = bτ −1 + · · · (9.95)
2 2
We assume that the rate k is negligible for X < 0 and very large for
X > 0. Consequently, only jumps X < 0 → X > 0 are considered. Then the
probability obeys the equation
P (X, t)|X<0
∞ 0 t
= P (X, 0) ψ(t )dt + dX dt ψ(t − t )f (X)P (X , t )
t −∞ 0
t 0
= φeq (X)Ψ0 (t) + φeq (X) dt ψ(t − t ) dX P (X , t ) (9.96)
0 −∞
∞
1 − ψ̃(s)
Ψ0 (t) = ψ(t )dt , Ψ˜0 (s) = (9.97)
t s
and the total occupation of inactive configurations is
0 t
P< (t) = dXφeq (X) Ψ0 (t) + dt ψ(t − t )P< (t )
−∞ 0
t
= a Ψ0 (t) + dt ψ(t − t )P< (t ) . (9.98)
0
with 0
a= dXφeq (X) (9.100)
−∞0
and the decay of the initial state is given by
aΨ̃0 (s) 1
P̃< (s) = = . (9.101)
1 − aψ̃(s) s + 1−a
aΨ ˜(s) 0
aΨ̃0 (s)
P̃< (s) ≈ → 0, s → 0, (9.105)
1−a
a
P< (t) → Ψ0 (t), t → ∞. (9.106)
1−a
To describe a power time law at long times
P< (t) → t−β , t → ∞, (9.107)
P̃< (s) → sβ−1 , s → 0, (9.108)
the waiting time distribution has to be chosen as
1
Ψ0 (t) ∼ , t → ∞,
(zt)β
which implies
Ψ̃0 (s) ∼ z −β sβ−1 , (9.109)
−1
where z is the characteristic time for reaching the asymptotics. Finally, we
find
1 1
P̃< (s) ∼ 1−a β 1−β = . (9.110)
s+ a z s
s(1 + (z̃/s)β )
In the time domain, this corresponds to the Mittag–Leffler function3
∞
(−1)l (z̃t)βl
P< (t) = = Eβ (−(z̃t)β ) (9.111)
Γ (βl + 1)
l=0
Problems
9.1. Dichotomous Model for Dispersive Kinetics
k(X)
P(D*,X)
α
P(D*,−) P(D*,+)
β
X
k− k+
α
P(D+A−,−) P(D+A−,+)
β
3
which has also been discussed for nonexponential relaxation in inelastic solids and
dipole relaxation processes corresponding to Cole–Cole spectra.
122 9 Dispersive Kinetics
Determine the eigenvalues of the rate matrix M . Calculate the left and right
eigenvectors approximately for the two limiting cases:
(a) Fast fluctuations k± α, β. Show that the initial state decays with an
average rate.
(b) Slow fluctuations k±
α, β. Show that the decay is nonexponential.
Part IV
Transport Processes
10
Nonequilibrium Thermodynamics
energy
JE
JA
A+B C JC
JB
JS
entropy
Fig. 10.1. Open nonequilibrium system. Energy and particles are exchanged with
the environment. Chemical reactions change the concentrations and consume or
release energy. Entropy is produced by irreversible processes like diffusion and heat
transport
∂ r
k = −div(k vk ) + mk νkj rj
∂t j=1
r
= −div(mk Jk ) + mk νkj rj (10.5)
j=1
we have
∂
r
= −div k v k + mk νkj rj . (10.7)
∂t j=1 k
Due to conservation of mass for each reaction, the last term vanishes
mk νkj = 0 (10.8)
k
we have
∂
= −div(v). (10.10)
∂t
10.2 Energy Conservation 127
has contributions from the local entropy production dSi and from exchange
with the surroundings dSe :
dSi ≥ 0, (10.21)
Together, we have
∂(s)
= −div(Js ) + σ, (10.24)
∂t
σ ≥ 0. (10.25)
From (10.17) the time derivative of the entropy per volume is
dSe
dSi
dSe
Fig. 10.2. Entropy change. The local entropy production dSi is always positive,
whereas entropy exchange with the environment dSe can have both signs
10.3 Entropy Production 129
of the fluxes
Jq , Jk , rj (10.34)
2
The definition of the heat flux is not unique in the case of simultaneous diffusion
transport. Our definition is in analogy to the relation T dS = dH −
and heat
V dp − k μk dNk for an isobaric system. With this convention, the diffusion flux
does not depend on the temperature gradient explicitly.
130 10 Nonequilibrium Thermodynamics
The entropy production is also zero and entropy has its maximum value. If
the deviation from equilibrium is small, the fluxes3 can be approximated as
linear functions of the forces:
Jα = Lα,β Kβ , (10.38)
β
T T+ΔT
Jq
forces. If the forces K1 , . . . , Km are kept fixed and a stationary state is reached,
then the fluxes corresponding to the remaining forces vanish:
Jα = 0, α = m + 1, . . . , n (10.40)
A simple example is the coupling of diffusion and heat transport between two
reservoirs with fixed temperatures T and T + ΔT (Fig. 10.3).
In a stationary state, the diffusion current must vanish Jd = 0 since any dif-
fusion would lead to time-dependent concentrations. The entropy production
is
σ = Jq Kq + Jd Kd . (10.42)
If the force
1 ΔT
Kq = grad ≈− 2 (10.43)
T T
is fixed by keeping the temperature distribution fixed, then in the stationary
state there is only transport of heat but not of mass. The entropy production
is
hence in the stationary state. Obviously, the chemical potential adjusts such
that
1 Lqd ΔT
Kd = − Δμ = , (10.47)
T Ldd T 2
which leads to a concentration gradient6 :
Δc Lqd ΔT
≈− . (10.48)
c Ldd kB T 2
Problems
6
This is connected to the Seebeck effect.
11
Simple Transport Processes
Jq
T1 T2
JS
where the interaction mobilities Lk,k vanish for very small ion concen-
trations, whereas the direct mobilities Lk,k are only weakly concentration
dependent [39, 40]. Inserting the forces, we find
kB 1
Jk = − Lk,k gradck − grad(Φ) Lk,k Zk e, (11.13)
ck T
k k
which for small ion concentrations can be written more simply by introducing
the diffusion constant as
ck Zk eDk
Jk = −Dk grad ck − grad Φ, (11.14)
kB T
2
The concentration (particles per volume) is ck = k /mk = xk /mk .
11.2 Diffusion in an External Electric Field 135
− +
− grad(c)
− grad(Φ)
In equilibrium, we have
μk = const (11.24)
and hence
kB T ln ck + Zk eΦ = const (11.25)
or
ck = c0k e−Zk eΦ/kB T . (11.26)
This is sometimes described as an equilibrium of the currents due to concen-
tration gradient on the one side and to the electrical field on the other side
(Fig. 11.2).
The Nernst–Planck equation has to be solved together with the Poisson–
Boltzmann equation for the potential
div(εgrad Φ) = − Zk eck . (11.27)
Problems
11.1. Synthesis of ATP from ADP
In mitochondrial membranes, ATP is synthesized according to
where (in) and (out) refer to inside and outside of the mitochondrial mem-
brane.
Express the equilibrium constant K as a function of the potential difference
c(H+ +
in )/c(Hout ). (11.31)
12
Ion Transport Through a Membrane
μ = μ0 + kB T ln c, (12.1)
μ − μ0
c = exp . (12.2)
kB T
Assuming linear variation of the chemical potential over the thickness a of the
membrane, we have
x
μ(x) = μinside + Δμ. (12.3)
a
The diffusion current is
∂μ Δμ
J = −C = −C (12.4)
∂x a
and the constant is determined from the diffusion equation in the center of
the membrane:
∂c ∂ ∂ kB T ∂c ∂ ∂
0= = − (J) = − −C = D(x) c
∂t ∂x ∂x c ∂x ∂x ∂x
D
→C = c (12.5)
kB T
with
μinside + (Δμ/2) μoutside + μinside √
c = exp = exp = cinside coutside . (12.6)
kB T 2kB T
140 12 Ion Transport Through a Membrane
cinside coutside
J
μoutside μinside
μ0 μ0
DcΔμ Δμ
J =− =− . (12.7)
akB T R
I
inside
g VN
V
outside
μoutside μinside
R R
Jm
C
μ0 μ0
The transport through biological membranes occurs often via the formation
of pore–ligand complexes. The number of these complexes depends on the
chemical potential of the ligands and needs time to build up. It may therefore
be more realistic to associate also a capacity Cm with the membrane, which
we define in analogy to the electronic capacity via the change of the potential
(Fig. 12.4)
dμm
Jm = Cm . (12.14)
dt
The change of the potential is
dμm kB T dc
= (12.15)
dt c dt
and the current is from the continuity equation
Jm dc
− divJ = = . (12.16)
a dt
Hence, the capacity is
ac
Cm = . (12.17)
kB T
membrane
outside JNa+ inside
JK+
JCl−
charge
density
potential Φm
0 d x
Let us first consider only one type of ions and constant temperature. From
(11.25), we have
kB T
Φ+ ln ck = const. (12.18)
Zk e
Usually, the potential is defined as zero on the outer side and
kB T ck,outside
Vk = Φinside − Φoutside = ln (12.19)
Zk e ck,inside
d
Jk eyk dx = −D (ck eyk )dx = −D ck,in eZk eΦm /kB T − ck,out .
0 0 dx
(12.22)
and we find
e2 Φm ck,in eZk eΦm /kB T − ck,out
Dk Zk2 = 0. (12.27)
kB T d eZk eΦm /kB T − 1
With ZNa+ = ZK+ = 1, ZCl− = −1 and
eΦm
ym = , b m = ey m , (12.28)
kB T
we have
cNa,in bm − cNa,out cK+ ,in bm − cK+ ,out
DNa+ + DK+ (12.29)
bm − 1 bm − 1
−1
cCl,in bm − cCl,out
+DCl− = 0. (12.30)
b−1
m −1
and we find
DNa+ cNa,out + DK+ cK,out + DCl− cCl,in
bm = , (12.33)
DNa+ cNa,in + DK+ cK,in + DCl− cCl,out
and finally
kB T D + cNa,out + DK+ cK,out + DCl− cCl,in
Φm = ln Na . (12.34)
e DNa+ cNa,in + DK+ cK,in + DCl− cCl,out
This formula should be compared with the Nernst equation:
kB T ck,outside
Vk = Φinside − Φoutside = ln . (12.35)
Zk e ck,inside
The ionic contributions appear weighted with their mobilities. The Nernst
equation is obtained if the membrane is permeable for only one ionic species.
Ik = gk (V − Vk ), (12.38)
Vinside
Im INa IK ILeak
Iext
Cm
Voutside
α(V )
closed open. (12.39)
β(V )
The variable p which denotes the number of open gates obeys a first-order
kinetics:
dp
= α(1 − p) − βp. (12.40)
dt
Hodgkin and Huxley introduced a gating particle n describing the potas-
sium conductance (Fig. 12.7) and two gating particles for sodium (m being
activating and h being inactivating). They used the specific equations
gK = gk p4n , (12.41)
K+
membrane membrane
n−gates
40
potential (mV)
0
– 40
– 80
1.0
probability
0.5
0.0
0 100 200 300 400 500
time (ms)
Fig. 12.8. Hodgkin–Huxley model. Equation (12.43) is solved numerically [43]. Top:
excitation by two short rectangular pulses Iext (t) (solid ) and membrane potential
V (t) (dashed ). Bottom: fraction of open gates pm (full ), pn (dashed ), and ph (dash-
dotted )
146 12 Ion Transport Through a Membrane
40
potential (mV)
0
– 40
–80
1.0
probability
0.5
0.0
0 100 200 300 400 500
time (ms)
13.1 Derivation
We assume for the diffusion coefficients the simplest case that the different
species diffuse independently (but possibly with different diffusion constants)
148 13 Reaction–Diffusion Systems
and that the diffusion fluxes are parallel to the corresponding concentration
gradients. We write the diffusion–reaction equation in matrix form
⎛ ⎞ ⎛ ⎞ ⎛ ⎞ ⎛ ⎞
c1 D1 c1 F1 ({c})
∂ ⎜ . ⎟ ⎜ .. ⎟ ⎜ .. ⎟ ⎜ .. ⎟
⎝ .. ⎠ = ⎝ . ⎠Δ⎝ . ⎠ + ⎝ . ⎠ (13.5)
∂t
cN DN cN FN ({c})
or briefly
∂
c = DΔc + F(c). (13.6)
∂t
13.2 Linearization
Since a solution of the nonlinear equations is generally possible only numeri-
cally, we investigate small deviations from an equilibrium solution c0 1 with
∂
c0 = 0, (13.7)
∂t
Δc0 = 0. (13.8)
Obviously, the equilibrium corresponds to a root
F(c0 ) = 0. (13.9)
c = c0 + ξ (13.10)
Plane waves. Plane waves are solutions of the linearized problem. Using the
Ansatz
ξ = ξ0 ei(ωt−kx) (13.14)
gives
iωξ = −k 2 Dξ + M0 ξ = Mk ξ. (13.15)
For a stable plane wave solution, λ = iω is an eigenvalue of
Mk = M0 − k 2 D (13.16)
with
(λ) ≤ 0. (13.17)
If there are purely imaginary eigenvalues for some k, they correspond to stable
solutions which are spatially inhomogeneous and lead to formation of certain
patterns.
u3
u̇ = u − − v + I(t), (13.18)
3
v̇ = (u + a − bv). (13.19)
The standard parameter values are
R
C
V
tunnel L
diode
E
2
v
–1
–3 –2 –1 0 1 2
u
This defines a region around ueq = ±1 where the eigenvalues have a nonzero
imaginary part (Fig. 13.3).
For standard parameters, we distinguish the regions
1 √
I : |ueq | > 1 + b + 2 M = 1.277,
II : 0.706 < |ueq | < 1.277,
1
√
III : |ueq | < 1 + b − 2 = 0.706. (13.32)
Within region II, the real part of both eigenvalues is (Fig. 13.4)
1
λ = (1 − u2eq − b) (13.33)
2
and since √ √
1 + b − 2 < u2eq < 1 + b + 2 , (13.34)
we have
√ √
1 − u2eq − b − 2 + 2b < 0 < 1 − u2eq − b + 2 + 2b, (13.35)
0.3
I II III II I
0.2
0.1
Im(λ)
–0.1
–0.2
–2 –1 0 1 2
ueq
III
II II
0
Re(λ) I I
–1
–2
–3
–2 –1 0 1 2
ueq
√ √
λ + b − < 0 < λ + b + , (13.36)
√
|λ + b| < . (13.37)
For standard parameters,
k
A B, (14.2)
k
1
An overview over the development of rate theory during the past century is given
in [45].
156 14 Equilibrium Reactions
reactants products
Ea
E’a
ΔH0
reaction coordinate
T = const, (14.5)
A= μk νk = 0. (14.6)
k
μk = μ0k + kB T ln ck , (14.7)
we have
μ0k νk + kB T νk ln ck = 0, (14.8)
k k
which gives the van’t Hoff relation for the equilibrium constant
0
μ νk ΔG0
ln(Kc ) = νk ln ck = − k k = − . (14.9)
kB T kB T
k
The standard reaction free energy can be divided into an entropic and an
energetic part:
ΔG0 −ΔH 0 ΔS 0
− = + . (14.10)
kB T kB T k
Since volume changes are not important at atmospheric pressure, the free
reaction enthalpy gives the activation energy difference
Ea − Ea = ΔH 0 . (14.11)
A catalyst can only change the activation energies but never the difference
ΔH 0 .
14.2 Statistical Interpretation of the Equilibrium Constant 157
A B. (14.16)
The partition functions for the two species are (Fig. 14.2)
zA = e−n (A)/kB T , zB = e−n (B)/kB T . (14.17)
n=0,1,... n=0,1,...
In equilibrium,
μB − μA = 0, (14.18)
ε3(B)
ε2(B)
ε3(A)
ε1(B)
ε2(A)
ε0(B)
ε1(A)
Δε0
ε0(A)
A B
This is nothing but a thermal distribution over all energy states of the system.
15
Calculation of Reaction Rates
The activated behavior of the reaction rate can be understood from a sim-
ple model of two colliding atoms (Fig. 15.1). In this chapter, we discuss the
connection to transition state theory, which takes into account the internal
degrees of freedom of larger molecules and explains not only the activation
energy, but also the prefactor of the Arrhenius law [27, 28, 46].
vr = v A − vB . (15.1)
The two particles collide if the distance between their centers is smaller than
the sum of the radii:
d = rA + rB . (15.2)
During a time interval dt, the number of possible collision partners is
cB πd2 vr dt (15.3)
vr
d = ra +rb
vr dt
where E0 is the minimum energy necessary for a reaction to occur (Fig. 15.2).
The distribution of relative kinetic energy can be determined as follows.
From the one-dimensional Maxwell distribution
m
e−mv /2kB T dv,
2
f (v)dv = (15.12)
2πkB T
15.1 Collision Theory 161
σ(E)
Eo E
we find the distribution of relative kinetic energy for one particle as1
ma −Ea /kB T dEa 1
f (Ea )dEa = 2 e √ =√ e−Ea /kB T dEa .
2πkB T 2mEa πkB T Ea
(15.13)
The joint distribution for the two particles is
f (Ea , Eb )dEa dEb
1 1
= √ e−Ea /kB T dEa √ e−Eb /kB T dEb (15.14)
πkB T Ea πkB T Eb
and the distribution of the total kinetic energy is given by
E
f (E) = f (Ea , E − Ea )dEa
0
E
1 1
= e−Ea /kB T −(E−Ea )/kB T dEa
πkB T 0 Ea (E − Ea )
E
1 −E/kB T dEa
= e . (15.15)
πkB T 0 Ea (E − Ea )
This integral can be evaluated by substitution
Ea = E sin2 θ, dEa = 2E sin θ cos θ dθ,
E π/2 π/2
dEa 2E sin θ cos θdθ
= √ = 2dθ = π (15.16)
0 Ea (E − Ea ) 0 E sin2 θE cos2 θ 0
and we find
1 −E/kB T
f (E) = e . (15.17)
kB T
Now
∞ ∞
f (E)σ(E)dE = f (E)πd2 dE = πd2 e−E0 /kB T (15.18)
0 E0
1
There is a factor of 2 since v 2 = (−v)2 .
162 15 Calculation of Reaction Rates
For nonspherical molecules, the reaction rate depends on the relative orienta-
tion. Therefore, a so-called steric factor p is introduced:
%
8kB T −E0 /kB T
r = ca cb pπd2 e . (15.20)
πμ
According to transition state theory [47, 48], the reactants form an equilib-
rium amount of an activated complex which decomposes to yield the products
(Figs. 15.3 and 15.4).
The reaction path [49] leads over a saddle point called the transition state.
According to TST, the reaction of two substances A and B is written
K k1
A+B [A−B]‡ → C+D . (15.22)
reactants activated complex products
activated complex
reactants products
reaction coordinate
Fig. 15.3. Transition state theory. The reactants are in chemical equilibrium with
the activated complex which can decompose into the products
2
The quantity πd2 /4 is equivalent to the collision cross section used in connection
with nuclear reactions.
15.2 Transition State Theory 163
A+BC
RAB
[A−B−C]
R*AB
AB+C
R*BC RBC
Fig. 15.4. Potential energy contour diagram. The potential energy for the reaction
A+BC → [A−B−C]‡ → AB+C is shown schematically as a function of the distances
RAB and RBC . In general, three coordinates are needed to give the relative positions
of the nuclei, but if atom A approaches the molecule BC, the direction of minimum
potential energy is along the line of centers for the reaction. Arrows indicate the
reaction coordinate. The activated complex corresponds to a saddle point of the
potential energy surface
kB T z ‡ −ΔE0 /kB T
K= e , (15.25)
ω zA zB
If not every activated complex decays into products, the expression for the
rate constant has to be modified by a transmission coefficient κ:
kB T z ‡ −ΔE0 /kB T
k2 = κ e . (15.27)
2π zA zB
In most cases, κ is close to 1. Important exceptions are the formation of
a diatomic molecule, since the excess energy can only be eliminated by
third-body collisions or radiation and reactions that involve changes from
one type of electronic state to another – for instance in certain cis–trans
isomerizations.
(2πmkB T )3/2
zT = . (15.28)
h3
For the activated complex, we consider the rotation around the center of mass.
The moment of inertia is
mA mB
I = (rA + rB )2 (15.29)
mA + mB
and the partition function is
8π 2 IkB T
zR = . (15.30)
h2
The rate expression from TST is
kB T
k2 = Kc (15.39)
2π
4
The activated complex is treated as a normal molecule with the exception of the
special mode.
166 15 Calculation of Reaction Rates
d ln k2 1 ∂ ΔG0‡
= − . (15.41)
dT T ∂T kB T
Now for ideal gases or solutions, the chemical potential has the form
z
μ = kB T ln c − kB T ln (15.42)
V
hence zi
ΔG0‡ = −kB T νi ln (15.43)
V
and
∂ ΔG0‡ ∂ Ui
=− νi ln zi = νi = ΔU 0‡ . (15.44)
∂T kB T ∂T kB T 2
Comparison with the Arrhenius law
d Ea Ea
ln A − = (15.45)
dT kB T kB T 2
ln(k)
V0
E tunneling
activated
a 1/kBT
Fig. 15.5. Tunneling through a rectangular barrier. Left: the tunneling probability
depends on the height and width of the barrier. Right: at low temperatures, tunneling
dominates (dashed ). At higher temperatures, activated behavior is observed (dash-
dotted )
activated complex
ln(k)
Eo(H)
Eo(D)
reactants
Ea(H)
Ea(D)
Eo(H)
Eo(D)
1/kBT
The potential energy surface has a saddle point in the transition state region.
The surface S ∗ , passing through that saddle point along the direction of steep-
est ascent, defines the border separating the reactants from the products in
quite a natural way. We introduce the so-called intrinsic reaction coordinate
(IRC) qr as the path of steepest descent from the cole connecting reactants
and products. We set qr = 0 for the points on the surface S ∗ . The instanta-
neous rate r(t) is given by the flux of systems that cross the surface S ∗ at time
t and are on the product side at t → ∞. This definition allows for multiple
crossings of the surface S ∗ . In general, however, it will depend on how exactly
the surface S ∗ is chosen. Only if the fluxes become stationary, this dependence
vanishes.
Classically, the flux is the product of the number of systems passing through
the surface S ∗ and their velocity vr normal to that surface. As we defined the
reaction coordinate qr to be normal to the surface S ∗ , the velocity is
dqr
vr = (15.50)
dt
and the classical flux is
pr
F = dn−1 q dn p vr W (q, p, t) = dn q dn p δ(qr ) W (q, p, t).
mr
(15.51)
However, we need the projector P (t) on that states that at some time t in the
future are on the product side. If H is the system Hamiltonian, then
and that part of the density matrix ρ(t) that will end in the product side in
the far future is
lim P (t − t)ρ(t)P (t − t). (15.57)
t →∞
The time dependence of the density matrix is given by the van Neumann
equation
dρ(t)
i = [H, ρ(t)]. (15.59)
dt
In thermal equilibrium,
This expression can be modified such that the limit operation does not appear
explicitly any more. To that end, we note that the trace vanishes for t =06
and
t = ∞
−1
r = Q tr e −βH iHt /
e θ(qr ) e −iHt /
F
t =0
∞
Ultimately, this expression for the reaction rate has been used a lot in
numerical computations. It is quite general and may easily be extended to
nonadiabatic reactions.
Problems
15.1. Transition State Theory
v
[AB]
A+B
δx
For a one-dimensional model, assume that the transition state is an exit point
from the parabola at some position x = x‡ with energy ΔE = mω 2 x‡2 /2 and
evaluate the TST rate constant
ΔE
x x
16
Marcus Theory of Electron Transfer
In 1992, Rudolph Marcus received the Nobel Prize for chemistry. His theory
is currently the dominant theory of electron transfer in chemistry. Originally,
Marcus explained outer-sphere electron transfer, introducing reorganization
energy and electronic coupling to relate the thermodynamic transition state
to nonequilibrium fluctuations of the medium polarization [50–52].
D + A → D+ + A − . (16.1)
If the charge is transferred from one molecule to the other, this is a charge
resonance process
D− + A → D + A− (16.2)
or
D + A+ → D+ + A. (16.3)
In the following, we want to treat all these kinds of processes together. There-
fore from now on, the symbols D and A implicitly have the meaning DZD and
AZA and the general reaction scheme
D + A → D+ + A − . (16.5)
1. The two species approach each other by diffusive motion and form the
so-called encounter complex:
kdif
‡
D + A [D − A] . (16.6)
k−dif
ket
‡ ‡
[D − A] [D A− ] .
+
(16.7)
k−et
The actual ET is much faster than the motion of the nuclei. The nuclear
conformation of the activated complex and the solvent polarization do not
change. According to the Franck–Condon principle, the two states [D − A]‡
and [D+ A− ]‡ have to be isoenergetic.
3. After the transfer, the polarization returns to a new equilibrium and the
products separate
ksep
‡
[D A− ] → D+ + A− .
+
(16.8)
Gr (Q)
Gp(Q)
ER
ΔG
ΔG
Q =0 Qs Q1 Q
ER −ΔG
activationless
activated inverted
ln(k)
If the polarization change induced by the ET is not too large (no dielectric
saturation), the potential curve for the final state will have the same curvature
but is shifted to a new equilibrium value Q1 :
a a
p + (Q − Q1 ) = Gr + ΔG + (Q − Q1 ) .
Gp (Q) = G(0) 2 (0) 2
(16.19)
2 2
In the new equilibrium, the final state is stabilized by the reorganization
energy1
a
ER = −(Gp (Q1 ) − Gp (0)) = Q21 . (16.20)
2
The transition between the two states is only possible if due to some thermal
fluctuations the crossing point is reached which is at
a 2
2 Q1+ ΔG ER + ΔG
Qs = = √ . (16.21)
aQ1 2aER
The corresponding activation energy is
a 2 (ΔG + ER )2
ΔG‡ = Qs = (16.22)
2 4ER
hence the rate expression becomes
ωN (ΔG + ER )2
ket = κet exp − . (16.23)
2π 4ER kB T
If we plot the logarithm of the rate as a function of the driving force ΔG,
we obtain the famous Marcus parabola which has a maximum at ΔG = −ER
(Fig. 16.2).
1
Sometimes the reorganization energy is defined as the negative quantity Gp (Q1 )−
Gp (0).
16.3 Free Energy Contribution of the Nonequilibrium Polarization 177
D = ε0 E + P = εE. (16.25)
Here P is the polarization, which is nothing else but the influenced dipole
density. It consists of two parts [53]:
1. The optical or electronical polarization Pop . It is due to the induced dipole
moments due to the response of the molecular electrons to the applied field.
At optical frequencies, the electrons still move fast enough to follow the
electric field adiabatically.
2. The inertial polarization Pin which is due to the reorganization of the
solvent molecules (reorientation of the dipoles, changes of molecular geom-
etry). This part of the polarization corresponds to the static dielectric
constant. The nuclear motion of the solvent molecules cannot follow the
rapid oscillations at optical frequencies.
The total frequency-dependent polarization is
P(ω) = Pop (ω) + Pin (ω) = D(ω) − ε0 E(ω) = (ε(ω) − ε0 )E(ω). (16.26)
In general, rot D
= 0 and the calculation of the field for given charge distribu-
tion is difficult. Therefore, we use a model of spherical ions behaving as ideal
conductors with the charge distributed over the surface. Then in our case, the
displacement is normal to the surface and rot D = 0 and the displacement is
the same as it would be generated by the charge distribution in vacuum.
The corresponding electric field in vacuum would be ε−1 0 D and the field in
178 16 Marcus Theory of Electron Transfer
(2)
P
Peq
(1)
DI/II D*I/II D
Now, we have to calculate the free energies of the activated complexes. Con-
sider a thermal fluctuation of the inertial polarization. We will do this in two
steps (Fig. 16.3).
In the first step, it is assumed that the polarization is always in equilib-
rium with the charge distribution. However, this charge density will be chosen
such that the nonequilibrium polarization P‡in of the activated complex is
generated. The electric fields are D∗ and E∗ :
1 ‡ 1 1
Pin = − D∗ . (16.37)
εop εop εst
In the second step, the electric field of the charge distribution will be reduced
to the equilibrium value while keeping the nonequilibrium polarization fixed.2
Then this fixed polarization acts as an additional displacement:
1 1
E= (D − P‡in ) − Pop , (16.39)
ε0 ε0
where the optical polarization depends on the electric field according to (16.28)
1 ‡ εop
E = (D − Pin ) − − 1 E. (16.40)
ε0 ε0
2 ‡
We tacitly assume that this is possible. In general, the polarization Pin will also
modify the charge distribution on the reactants. If, however, the distance between
the two ions is large in comparison with the radii, these changes can be neglected.
Marcus calls this the point-charge approximation.
180 16 Marcus Theory of Electron Transfer
and the total free energy change due to the polarization fluctuation is
G‡el − Geq 1
el = ΔGel + ΔGel
2
1 P‡2 − P2
=− 1 d3 r in 2 in
εop − εst
1 2εop
1 1
− 1 d3 r 2 P‡in (Pin,eq − P‡in )
εop − εst
1 εop
2
1 3 1 P‡in − Pin,eq
= 1 d r . (16.44)
εop − εst
1 2 εop
δG‡el = 0, (16.46)
we find that
δP = 0, G‡el = Geq
el . (16.47)
16.4 Activation Energy 181
Marcus asks now, for which polarizations P‡in of the two states I and II become
isoenergetic, i.e.
ΔU ‡ = UII‡ − UI‡ = 0. (16.48)
We want to rephrase the last condition in terms of the free energy
G = U − T S + pV ≈ U − T S. (16.49)
We conclude that
ΔG‡ = G‡II − G‡I = −T ΔS ‡ . (16.50)
Here (Fig. 16.4)
2
1 1 δPI
G‡I = 3
d r + Geq eq ∞
el,I (R) − Gel,I (∞) + GI ,
1
εop − 1
εst
2 εop
2
1 1 δPII
G‡II = d3 r + Geq eq ∞
el,II (R) − Gel,II (∞) + GII . (16.51)
1
εop − 1
εst
2 εop
The quantity
ΔG∞ = G∞ ∞
II − GI = ΔGvac + ΔGsolv,∞ (16.52)
is the difference of free energies for an ideal solution of A and B on the
one side and A+ and B− on the other side at infinite distance R → ∞.
Usually, it is determined experimentally. It contains all the information on
the internal structure of the ions. The entropy difference is the difference of
internal entropies of the reactants since the contribution from the polarization
drops out due to P‡in,I = P‡in,II . Usually, this entropy difference is small. The
electrostatic energies Geqel,I/II are composed of the polarization contribution
(the solvation free energy) and the mutual interaction of the reactants which
will be considered later. The reaction free energy is
It can be easily seen that the condition ΔG‡ = 0 will be fulfilled for infinitely
many polarizations P‡in . We will therefore look for that one which minimizes
simultaneously the free energy of the transition state G‡I,II . We introduce a
Lagrange parameter m to impose the isoenergeticity condition:
6 7
δ G‡I (P‡in ) + m(G‡II (P‡in ) − G‡I (P‡in )) = 0. (16.54)
The variation is with respect to P‡in or, equivalently with respect to δPI for
fixed change of the inertial polarization:
ΔPin = δPI − δPII = (P‡in − PI,in ) − (P‡in − PII,in ) = PII,in − PI,in . (16.55)
182
R=∞ R=∞ R
∞ eq eq
ΔGsolv,I GI (R) – GI (∞)
A B eq ¹ ¹
AI AI B,I B GI (R) GI(R,P )
Gvac A B,I B
Gvac A G∞ G∞
∞
∞ eq
Geq
16 Marcus Theory of Electron Transfer
#
GI/II
GIeq
δ PI#
δ PII GIIeq
#
Pin,I Pin Pin,II Pin
ΔPin
m 1−m
We have to solve
6 7
δ (1 − m)G‡I (PI,in + δPI ) + mG‡II (PII,in − ΔPin + δPI,in ) = 0. (16.56)
Due to the quadratic dependence of the free energies, we find the condition
− T ΔS = ΔG‡ (P‡in )
= G‡II (Pin,II + (m − 1)ΔPin ) − G‡I (Pin,I + mΔPin )
2
1 1 ΔPin
= ΔGeq + ((m − 1)2 − m2 ) 1 d3
r
εop − εst
1 2 εop
= ΔGeq + (1 − 2m)λ, (16.59)
where
ΔGeq + T ΔS ‡ + λ
m= . (16.61)
2λ
The free energy of activation is given by
2
1 1 ΔPin
ΔGa = G‡I − Geq
I = m
2
d3 r
1
εop − 1
εst
2 εop
(ΔGeq + T ΔS ‡ + λ)2
= m2 λ = . (16.62)
4λ
Finally, we insert this into the TST rate expression to find
ωN ωN (ΔGeq + T ΔS ‡ + λ)2
ket = κet e−ΔGa /kB T = κet exp − . (16.63)
2π 2π 4λkB T
aA aB
For a charge shift reaction of the type QA,I = QB,II = −e, QB,I = QA,II = 0,
e2 e2
Geq
I = , Geq
II = , (16.82)
8πεst aA 8πεst aB
16.6 The Energy Gap as the Reaction Coordinate 187
ΔG
+
D
IP(D*)
D*
IP(D)
hω A
EA(A)
−
D A
Fig. 16.7. Photoinduced charge separation. At large distance, the energy difference
ΔG∞ is given by EA(A) − IP(D∗ ) = EA(A) − (IP(D) − ω)
e2 e2
Geq
I = , Geq
II = . (16.83)
8πεst aB 8πεst aA
The activation energy now does not contain a Coulombic term
(ΔG∞ + T ΔS ‡ + λ)2
ΔGa = . (16.84)
4λ
ΔU
λ λ
I
Geq
ΔGeq
II
Geq
0 2λ δU
The fluctuations of the energy gap can for instance be investigated with
molecular dynamics methods. They can be modeled by diffusive motion in
a harmonic potential U (Q) = Q2 /4λ:
2
∂ ∂ 1 ∂ ∂U
W (Q, t) = −DE W + W − κδ(Q − λ)W, (16.93)
∂t ∂Q2 kB T ∂Q ∂Q
where the diffusion constant is
2kB T λ
DE = (16.94)
τL
and τL is the longitudinal relaxation time of the medium.
16.8 The Transmission Coefficient for Nonadiabatic Electron Transfer 189
gr2n −gr2
FC(0, n) = e (16.96)
n!
and the sum over all transitions gives the rate expression
∞ 2n
ω −g2 g (ΔG + ER + nωv )2
k= κel e exp − , (16.97)
2π n=0
n! 4ER kB T
k= √ . (16.98)
4πλkB T
3
A more detailed discussion follows later.
190 16 Marcus Theory of Electron Transfer
n=0
n’ = 0 n’ = 0
Fig. 16.9. Accepting mode. In the inverted region, vibrations can accept the excess
energy
1
electron transfer rate k
0.1
0.01
0 0.5 1 1.5 2
ΔG
Problems
kAA
A− + A → A + A−
k
A+D → A− + D+
where kAA and kDD are the rate constants of the self-exchange reactions,
Keq is the equilibrium constant of the cross reaction and f is a factor,
which is usually close to unity.
Part VI
Elementary Photophysics
17
Molecular States
which has to be solved separately for each set of nuclear coordinates. Using
now the Born–Oppenheimer product ansatz, we have
Hψs (r, Q)χs (Q) = TN ψs (r, Q)χs (Q) + Es (Q)ψs (r, Q)χs (Q)
= ψs (r, Q)(TN + Es (Q))χs (Q)
2 ∂ 2 ψs (r, Q) ∂χs (Q) ∂ψs
− χs (Q) + . (17.6)
j
2mj ∂Q2j ∂Qj ∂Qj
is diagonalized
Djj urj = ωr2 urj (17.10)
j
The vibrations form a very dense manifold of states for biological molecules.
This is often schematically represented in a Jablonsky diagram3 (Figs. 17.1–
17.4)
E .
. .
. .
.
. S2
. T2
.
.
S1 .
.
T1
.
.
.
S0
3
which usually also shows electronic states of higher multiplicity.
4
Sometimes called channels.
5
Without a magnetic field, the molecular wave functions can be chosen real-valued.
198 17 Molecular States
O CH3 CH3
H CH3
2 3 H
CH3 1
I II 4
N N H
H
H H H
CH3 N N
IV III CH3
H V
H
CH2 H
O O
CH2 O
phytyl O O CH3
300
0.1 /cm–1
number of normal modes
250
200
150
100
50
0
0 1000 2000 3000 4000
normal mode frequency (cm–1)
16
10
DOS (1/cm–1)
12
10
8
10
4
10
0
10 0 1000 2000 3000 4000
frequency (cm–1)
Fig. 17.4. Density of vibrational states. The density of vibrational states including
overtones and combination tones was evaluated for bacteriopheophytine-b with a
simple counting algorithm
17.2 Nonadiabatic Interaction 199
nad
Vs,s = dr dQ χs (Q)ψs (r, Q) Vnad ψs (r, Q)χs (Q)
2
∂2
=− dQ χs (Q) χs (Q) dr ψs (r, Q) ψs (r, Q)
j
2mj ∂Q2j
2
∂ ∂
− dQ χs (Q) χs (Q) dr ψs (r, Q) ψs (r, Q) .
2mj ∂Qj ∂Qj
(17.16)
√ √
with
FCrs s (nr nr ) = dqr χs r,n (r) (qr ) χs,r,n(r) (qr ). (17.22)
6
This is known as the Condon approximation.
18
Optical Transitions
The dipole moment function Meg (Q) is expanded as a series with respect to
the nuclear coordinates around the equilibrium configuration
∂Meg
Meg (Q) = Meg (Qeq ) + (Q − Qeq ) + · · · (18.5)
∂Q
202 18 Optical Transitions
If its equilibrium value does not vanish for symmetry reasons, the dipole
moment can be approximated by neglecting all higher-order terms (this is
known as Condon approximation)
2π
= |E0 μeg |2 FCF(ω0 ). (18.8)
where we introduced the Hamiltonians for the nuclear motion in the two states:
with
Hg(e) χg(e) = ωg(e) χg(e) (18.11)
1
and ω00 is the pure electronic, so-called “0–0” transition energy. The energy
of the transition
|ψg χg → |ψe χe (18.12)
is
ωeg = ω00 + ωe − ωg . (18.13)
The interaction between molecule and radiation field is within the Condon
approximation2
H = |ψe (−E0 μeg )ψg | + h.c. (18.14)
After replacing the δ-function by a Fourier integral
∞
1
δ(ω) = eiωt dt, (18.15)
2π −∞
1
Including the zero-point energy difference.
2
We consider here only the nondiagonal term.
18.2 Time Correlation Function Formalism 203
(18.7) becomes
|E0 μeg |2
k= dt P (χg )|χg |χe |2 ei(ωe −ωg +ω00 −ω0 )t , (18.16)
2 χg χe
which can be written as a thermal average over the vibrations of the ground
state
|E0 μeg |2
e−βHg
−iωg t
k= dt χ g e |χ e eiωe t χe |χg ei(ω00 −ω0 )t
2 χg χe
Q g
2 8 9
|E0 μeg |
= dt e−i(ω0 −ω00 )t e−itHg / eitHe / (18.17)
2 g
|E0 μeg |2
= dt e−i(ω0 −ω00 )t F (t) (18.18)
2
with the correlation function
8 9
F (t) = e−itHg / eitHe / (18.19)
g
which will be evaluated for some simple models in the following chapter. From
comparison with (18.8), we find that the Franck–Condon-weighted density of
states is given by
1
FCF(ω) = dte−i(ω0 −ω0 0) F (t). (18.20)
2π
Problems
18.1. Absorption Spectrum
Within the Born–Oppenheimer approximation, the rate for optical transi-
tions of a molecule from its ground state to an electronically excited state is
proportional to
α(ω) = Pi |i|μ|f |2 δ(+ωf − ωi − ω).
i,f
Show that this golden rule expression can be formulated as the Fourier integral
of the dipole moment correlation function
1
dt eiωt μ(0)μ(t)
2π
and that within the Condon approximation, this reduces to the correlation
function of the nuclear motion.
19
The Displaced Harmonic Oscillator Model
We would like now to discuss a more specific model for the transition between
the vibrational manifolds. We apply the harmonic approximation for the
nuclear motion which is described by Hamiltonians
Hg,e = ωr(g,e) br(g,e)+ br(g,e) (19.1)
r
with
Q = tr(e−Hg /kB T ) (19.5)
206 19 The Displaced Harmonic Oscillator Model
hω00
−ωr b†r br /kB T −itωr b†r br itωr (b†r +gr )(br +gr )
Fr (t) = Q−1
r tr e e e
8 † †
9
= e−itωr br br eitωr (br +gr )(br +gr ) . (19.7)
a b
0.80
0.60
f (hΔω)
0.40
0.20
0.00
–400 0 400 – 400 0 400
Δhω (cm–1)
gr2j −gr2
F C(0, j) = e . (19.13)
j!
ωr2 t2
eiωr t = 1 + iωr t − + ··· (19.15)
2
and find
# $
1 ∞
ωr2 t2
−iΔω
f (Δω) = dt e exp − gr2 iωr t
+ 1) gr2 (2n̄r
2π −∞ r r
2
% 2 2
1 2π (Δω − g ωr )
= 2 2 exp − 2 2 r r .
2π g ω
r r r (2n̄ r + 1) 2 ( g ω
r r r (2n̄ r + 1))
(19.16)
208 19 The Displaced Harmonic Oscillator Model
f( hΔω)
Δ
Er hΔω
In the high-temperature limit (all modes can be treated classically), the width
becomes (Fig. 19.3)
Δ2 → 2gr2 ωr kB T = 2Er kB T (19.20)
and
1 (Δω − Er )2
f (Δω) → √ exp − . (19.21)
4πEr kB T 4Er kB T
20
Spectral Diffusion
20.1 Dephasing
We study a semiclassical model of a two-state system. Due to interaction with
the environment, the energies of the two states are fluctuating quantities. The
system is described by a time-dependent Hamiltonian
E1 (t) 0
H0 = (20.1)
0 E2 (t)
and the time evolution of the unperturbed system
t
1
ψ(t) = exp H0 dt ψ(0) = U0 (t)ψ(0) (20.2)
i 0
can be described with the help of a propagator
t (1/i) t E (t)dt
1 e 0 1
U0 (t) = exp H0 dt = t . (20.3)
i 0 e(1/i) 0 E2 (t)dt
and find
d d
(H0 + H )ψ = i ψ = H0 U0 (t)φ + iU0 (t) φ. (20.6)
dt dt
Hence
d
φ = U0 (−t)H U0 (t)φ.
i (20.7)
dt
The product of the operators gives
t
0 μeiωt−(i/) 0 (E2 −E1 )dt
U0 (−t)H U0 (t) = t (20.8)
μe−iωt+(i/) 0 (E2 −E1 )dt 0
where the time-dependent dipole moment μ(t) obeys the differential equation
d 1
μ(t) = (E2 (t) − E1 (t))μ(t)
dt i
= −iω21 (t)μ(t) = −i(ω21 + δω(t))μ(t). (20.10)
P (t) 2π|μ0 |2
→ F̂ (ω − ω21 ). (20.18)
t 2
We assume that in the simplest case, the correlation of the frequency fluctu-
ations decays exponentially
t2
e−t/τc − 1 + t/τc = + ··· (20.27)
2τc2
F (t) = e−Δ
2 2
t /2
(20.28)
and the transition probability is
|μ0 |2 t t i(t −t )(ω−ω21 ) −Δ2 (t −t )2 /2
P (t) = 2 dt dt e e . (20.29)
0 0
20.2 Gaussian Fluctuations 213
100
10 –1
F(t)
10 –2
10 –3
–10 –5 0 5 10
time t
Fig. 20.1. Time correlation function of the Kubo model. F (t) from (20.26) is shown
for Δ = 1 and several values of τc . For large correlation times, the Gaussian
exp(−Δ2 t2 /2) is approximated whereas for short correlation times the correla-
tion function becomes approximately exponential and the width increases with
τ −1 . Correspondingly, the Fourier transform becomes sharp in this case (motional
narrowing)
hence
P (+, t + τ | + t0 )
= P (+, t + τ |+, t)P (+, t| + t0 ) + P (+, t + τ |−, t)P (−, t| + t0 )
= (1 − ατ )P (+, t|+, t0 ) + βτ P (−, t|+, t0 ) (20.38)
with
P (+, t|+, 0) P (+, t|−, 0)
P(t) = (20.42)
P (−, t|+, 0) P (−, t|−, 0)
and the rate matrix
−α β
A= . (20.43)
α −β
Let us define the quantity
8 t 9
Q(X, t|X , 0) = e−i 0 ω(t )dt , (20.44)
X,X
where the average is taken under the conditions that the system is in state X
at time t and in state X at time 0. Then we find
8 t 9
Q(X, t + τ |X , 0) = e−i 0 ω(t )dt eiω(t)τ
X,X
= Q(X, t + τ |+, t)Q(+, t|X , 0)
+Q(X, t + τ |−, t)Q(−, t|X , 0). (20.45)
and
ω+
Ω= . (20.50)
ω−
Equation (20.48) is formally solved by
For a steady-state system, we have to average over the initial state and to sum
over the final states to obtain the dephasing function. This can be expressed
as
β β
F (t) = (1, 1)Q(t) α+β
α = (1, 1) exp {(−iΩ + A)t} α+β
α . (20.52)
α+β α+β
s + (α + β) + i βωα+β
2 +αω1
= . (20.54)
(s + iω1 )(s + iω2 ) + (α + β)s + i(αω2 + βω1 )
Since the dephasing function generally obeys the symmetry F (−t) = F (t)∗ ,
the lineshape is obtained from the real part
∞ ∞ ∞
iωt iωt −iωt ∗
2π F̂ (ω) = e F (t)dt = e F (t)dt + e F (t)dt
−∞ 0 0
= 2 F (−iω)
αβ (ω1 − ω2 )2
=2
2 .
α+β
(ω − ω1 )2 (ω − ω2 )2 + (α + β)2 ω − αω2 +βω1
α+β
(20.55)
60
a b c
50
40
30
20
10
0
0 1 2 0 1 2 0 1 2
frequency ω
Fig. 20.2. Motional narrowing. The lineshape (20.55) is evaluated for ω1 = 1.0,
ω2 = 1.5 and α = β =(a) 0.02, (b) 0.18, and (c) 1.0
In the limit of slow fluctuations (ωc → 0), two sharp resonances appear at
ω = ω1,2 with relative weights given by the equilibrium probabilities
With increasing ωc , the two resonances become broader and finally merge
into one line. For further increasing ωc the resonance, which is now centered
at ω, becomes very narrow (this is known as the motional narrowing effect)
(Fig. 20.2).
Problems
20.1. Motional Narrowing
Discuss the poles of the lineshape function and the motional narrowing effect
(20.55) for the symmetrical case α = β = (2τc )−1 and ω1,2 = ω ± Δω/2.
21
Crossing of Two Electronic States
We will assume that the electronic wave functions are chosen real-valued (no
magnetic field). From the adiabatic energies
dr ψsad (Tel + V )ψsad ad
= δs,s Es (Q) (21.4)
2 nad,(2) 2 ∂ 2 ad
− Vs,s (Q) = − drψsad (r, Q) ψ (r, Q) , (21.7)
2m 2m ∂Q2 s
we construct the Hamiltonian for the nuclear wave function
2 ∂ 2 2 nad ∂ 2 nad,(2)
H(Q) = − + E ad
(Q) − V (Q) − V (Q) (21.8)
2m ∂Q2 2m ∂Q 2m s,s
with the matrices
- ad . / nad nad
0
E1 (Q) V11 (Q) V12 (Q)
E ad (Q) = , V nad
(Q) = ,
E2ad (Q) V21nad nad
(Q) V22 (Q)
⎡ ⎤
nad,(2) nad,(2)
V (Q) V (Q)
V nad,(2) (Q) = ⎣ ⎦.
11 12
(21.9)
nad,(2) nad,(2)
V21 (Q) V22 (Q)
The nuclear part of the wave function is determined by the eigenvalue
equation:
χ1 (Q) χ1 (Q)
H(Q) =E . (21.10)
χ2 (Q) χ2 (Q)
Now for real-valued functions, we have
nad ∂ ad
Vss (Q) = dr ψsad (r, Q) ψ
∂Q s
∂ ∂ ad
= (r, Q) −
dr ψsad (r, Q)ψsad dr ψs ψsad
∂Q ∂Q
& '
= − Vsnad s (Q) (21.11)
nad ∗
= −(Vs ,s ) .
1 nad
For complex wave functions instead Vs,s
21.1 Adiabatic and Diabatic States 221
2
for instance by a series expansion
222 21 Crossing of Two Electronic States
E1ad
E1dia
2Δ
Q
Edia
2
Ead
2
where they assumed a constant velocity in the vicinity of the crossing point.
They investigated the time-dependent model Hamiltonian
1 ∂ΔE
E1 (Q(t)) V − 2 ∂t t V
H(t) = = 1 ∂ΔE (21.34)
V E2 (Q(t)) V 2 ∂t t
P
high velocity
low velocity
Fig. 21.2. Velocity dependence of the transition. At low velocities, the electronic
wave function follows the nuclear motion adiabatically corresponding to a transition
between the diabatic states. At high velocities, the probability for this transition
(21.42) becomes slow. Full : diabatic states and dashed : adiabatic states
Let us consider the limit of small V and calculate the transition probability
in lowest order. From the initial condition a1 (−∞) = 1, a2 (−∞) = 0, we get
t
∂ΔE t2
(E2 (t ) − E1 (t ))dt =
, (21.39)
0 ∂t 2
∞ %
1 1 ∂ΔE t2 V 2π
a2 (∞) ≈ V exp − dt = (21.40)
i −∞ i ∂t 2 i −i(∂ΔE/∂t)
and the transition probability is
2πV 2
P12 = |a2 (∞)|2 = . (21.41)
|∂ΔE/∂t|
Landau and Zener calculated the transition probability for arbitrary coupling
strength (Fig. 21.2)
2πV 2
P12 = 1 − exp −
LZ
. (21.42)
|∂ΔE/∂t|
mω 2 2 mω 2
E1 (Q) = Q , E2 (Q) = ΔG + (Q − Q1 )2 , (21.43)
2 2
the energy gap is with
2λ
Q1 = , (21.44)
mω 2
mω 2 2 √
E2 − E1 = ΔG + (Q1 − 2Q1 Q) = ΔG + λ − ω 2mλQ. (21.45)
2
21.4 Crossing in More Dimensions 225
1 (1−P12)P12
(1−P12)P12
2
1−P12
P12
Fig. 21.3. Multiple crossing. The probability of crossing during one period of the
oscillation is 2P12 (1 − P12 ) if the two crossings are independent
The probability of curve crossing during one period of the oscillation is given
by 2P12 (1 − P12 ) ≈ 2P12 (see Fig. 21.3).
Together, this gives a transmission coefficient
2πV 2 2πV 2 2π 1
κel = 2 √ 1 =
√ (21.48)
ω 4πλkT
ω 2λ 2kT
π
and a rate of
2πV 2 1
k= √ e−ΔGa /kT . (21.49)
4πλkT
to obtain
∂ ad ∂ ad
dr dr ψs (r, Q) ψsad ad
(r)ψs (r ) ψs (r , Q)
∂Qi ∂Qj
s
∂ ad ad ∂ ad
− ad
ψ (r, Q) ψs (r)ψs (r ) ψ (r, Q) , (21.53)
∂Qj s ∂Qi s
which can be written as
& '
,i Vs ,s ,j − Vs,s ,j Vs ,s ,i .
nad nad nad nad
Vs,s (21.54)
s =s,s
and a potential function exists, which provides a solution to (21.50). For more
than two states, an integrability condition like (21.55) is not fulfilled and only
part of the nonadiabatic coupling can be eliminated.3 Often a crude diabatic
basis is used, which refers to a fixed configuration near the crossing point and
does not depend on the nuclear coordinates at all. In a diabatic representation,
the Hamiltonian of a two-state system has the general form
H11 (Qi ) H12 (Qi )
H= . (21.56)
H12 (Qi )† H22 (Qi )
Ead
Q2
Q1
Problems
21.1. Crude Adiabatic Model
Consider the crossing of two electronic states along a coordinate Q. As basis
functions, we use two coordinate-independent electronic wave functions which
diagonalize the Born–Oppenheimer Hamiltonian at the crossing point Q0 :
where χ(Q) and ζ(Q) depend on the coordinate Q whereas the basis func-
tions ϕ1,2 do not. Chose ζ(Q) such that Tel + V is diagonalized. Evaluate the
nonadiabatic interaction terms at the crossing point Q0 .
22
Dynamics of an Excited State
In this chapter, we would like to describe the dynamics of an excited state |s
which is prepared, for example, by electronic excitation, due to absorption of
radiation. This state is an eigenstate of the diabatic Hamiltonian with energy
Es0 . Close in energy to |s is a manifold of other states {|l}, which are not
populated during the short-time excitation, since from the ground state only
the transition |0 → |s is optically allowed. The l-states are weakly coupled
to a continuum of bath states1 and therefore have a finite lifetime (Fig. 22.1).
E
V
|s> |> Γ
|0>
Fig. 22.1. Dynamics of an excited state. The excited state |s decays into a manifold
of states |l which are weakly coupled to a continuum
The bath states will not be considered explicitly. Instead, we use a non-
Hermitian Hamiltonian for the subspace spanned by |s and {|l}. We assume
that the Hamiltonian is already diagonal2 with respect to the manifold {|l},
which has complex energies3
1
For instance, the field of electromagnetic radiation.
2
For non-Hermitian operators, we have to distinguish left eigenvectors and right
eigenvectors.
3
This is also known as the damping approximation.
230 22 Dynamics of an Excited State
G+
E
G−
Γl
El0 = l − i . (22.1)
2
This describes the exponential decay ∼ e−Γl t of the l-states into the continuum
states.
Thus, the model Hamiltonian takes the form
⎛ 0 ⎞
Es Vs1 · · · VsL
⎜ V1s E10 ⎟
⎜ ⎟
H = H0 + V = ⎜ . .. ⎟. (22.2)
⎝ .. . ⎠
VLs EL
For a Hermitian Hamiltonian, the poles of the Green’s operator are on the
real axis and the time evolution operator (the so-called propagator) is defined
by (Fig. 22.2)
= G+ (t) − G− (t),
G(t) (22.4)
∞±i
1
G± (t) = e−(iE/)t G(E)dE. (22.5)
2πi −∞±i
G(t) is given by an integral, which encloses all the poles En . Integration
between two poles does not contribute, since the integration path can be
taken to be on the real axis and the two contributions vanish. The integration
22.1 Green’s Formalism 231
G+
E
−G−
= G+ (t) − G− (t)
Fig. 22.3. Integration contour for G(t)
t<0
G+
E
t>0
along a small circle around a pole gives 2πi times the residual value which is
given by e−iEn t/ (Fig. 22.3).
Hence, we find
= t
G(t) |ne−(iE/)t n| = exp H . (22.6)
i
For times t < 0, the integration path for G+ can be closed in the upper half of
the complex plane where the integrand e−(iE/)t G(E) vanishes exponentially
for large |E| (Fig. 22.4).
Hence
G+ (t) = 0 for t < 0. (22.7)
We shift the integration path for times t > 0 into the lower half of the complex
plane, where again the integrand vanishes exponentially and we have to sum
over all residuals to find
G+ (t) = G(t) for t > 0. (22.8)
G+
E
G = G0 + G0 V (G0 + G0 V G) = G0 + G0 V G0 + G0 V G0 V G (22.13)
and
G0ss 1
Gss = = , (22.18)
1 − Gss l Vsl Gll Vls
0 0 E − Es0 − Rs
with the level shift operator
|Vsl |2
Rs = . (22.19)
E − l + i(Γl /2)
l
22.2 Ladder Model 233
The poles of the Green’s function Gss are given by the implicit equation
|Vsl |2
Ep = Es0 + . (22.20)
Ep − El0
l
The probability of finding the system still in the state |s at time t > 0 is
Ps (t) = |s|G(t)|s|2 ss (t)|2 ,
= |G (22.23)
l=∞
V2
Ep = , (22.26)
Ep − α − lΔ + i(Γ/2)
l=−∞
5
cot(z) has single poles at z = lπ with residues limz→lπ cot(z)(z − lπ) = 1.
234 22 Dynamics of an Excited State
as
V 2π π Γ
Ep = cot Ep − α + i . (22.28)
Δ Δ 2
For the following discussion, it is convenient to measure all energy quantities
in units of π/Δ and define
= απ/Δ,
α Γ = Γ π/Δ, (22.29)
p = Ep π/Δ,
E V = V π/Δ, (22.30)
to have # $
iΓp Γ
Ep = Ep −
r
= V cot Ep − α
2
+i , (22.31)
2 2
which can be split into real and imaginary part:
Epr pr − α
sin(2(E ))
= , (22.32)
V 2 *p − Γ) − cos(2(E
cosh(Γ r − α
p ))
For small molecules, the density of state 1/Δ is small. If we keep the lifetime
of the l-states fixed, then Γ is small in this limit. The poles are close to the real
axis.6 We approximate the real part by solving the real equation (Fig. 22.6)
and consider first the solution which is closest to zero. Expanding around E p(0) ,
we find
iΓ
Ep + ξ = V cot Ep + ξ − α
(0) 2 (0)
+
2
2 iΓ
= V cot E
2 p − α
(0)
− V 1 + cot E
2 p − α
(0)
ξ+ ,
2
(22.35)
6
We assume that α = 0.
22.2 Ladder Model 235
cot (x-1)
0
–4 –2 0 2 4
x
Fig. 22.6. Small molecule limit. The figure shows the graphical solution of
x = cot(x − 1)
which simplifies to
2 iΓ
ξ = −V 1 + cot E2
−α
(0)
ξ+ (22.36)
p
2
V2
Γp ≈ π 2
Γ. (22.39)
Δ2
The other poles are approximately undisturbed
pr = α
E + lπ, (22.40)
r2
V 2 +
E p
*p = Γ 2
Γ V
r2
→ Γ (22.41)
1 + V 2 +
E p
2
V
236 22 Dynamics of an Excited State
p + dE − α
(Ep + dE) − V cot E
2 + iΓ
2
1
=
,
p + dE)
− V 2 cot E
p − α 2
(E + iΓ
2 + V 2 dE(1
+ cot(E
p − α
+ iΓ
2 ) ) + ···
1 1 1
2 = , (22.42)
1 + V 2 1 + cot E p − α 2
dE V + E p
dE + iΓ
2
which shows that only contributions from poles close to zero are important.
Consider now the limit Γp > 1 and V > 1. This condition is usually fulfilled
in biomolecules (Fig. 22.7).
If the poles are far away enough from the real axis, then on the real axis
the cotangent is very close to −i and
V 2π π Γ V 2π
Rs = cot Ep − α + i ≈ −i . (22.43)
Δ Δ 2 Δ
Thus
1
Gss (E) = (22.44)
E + i(V 2 π/Δ)
4 4
2 2
0 0
−2 −2
−4 −4
−4 −4 −4 −4
−2 −2
−2 −2
0 0 0 0
Re(z) Im(z) 2 Re(z)
Im(z) 2 2 2
4 4 4 4
Fig. 22.7. Complex cotangent function. Left: imaginary part and right: real part
of cot(z + 3i) are shown together with (z) and (z), respectively
22.3 A More General Ladder Model 237
| = e−kt
2
P (t) = |e(−iV π/Δ)t 2
(22.45)
Es = E0 + ΔE, El = E0 + l . (22.49)
Es
ΔE εl
Eo
We introduce
V2
sl (il /)t
z= e (22.56)
2
l
This integral will now be evaluated approximately using the saddle point
(p. 339) method.7 The saddle point equation for the ladder model becomes
i d ln z(ts )
− ΔE + = 0. (22.58)
dt
From the derivative
d ln z 1 Vsl2 il itl /
= e , (22.59)
dt z 2
l
we find
1 Vsl2 itl /
ΔE = l e . (22.60)
z 2
l
Since ΔE is real, we look for a saddle point on the imaginary axis and write
its
= −β, (22.61)
where the new variable β is real. The saddle point equation now has a
quasithermodynamic meaning:
1 Vsl2 −βl
ΔE = l e = l , (22.62)
z 2
l
7
Also known as method of steepest descent or Laplace method.
22.3 A More General Ladder Model 239
where β plays the role of 1/kB T and gl = Vsl2 /2 that of a degeneracy factor.
The second derivative relates to the width of the energy distribution:
2
d2 z
d2 ln z d dz dz
= dt − dt
dt 2
2
=
dt dt z z z
2
1 −2l −βl 1 il −β
= gl 2 e − 2 gl e
z z
l l
2 8 9
2
=− + . (22.63)
2
We approximate now the integral by the contribution around the saddle point:
∞ %
−(i/)ΔE t+ln(z) βΔE 2π2
k= dt e = z(ts )e
∞ − 2
2
%
V2 2π
sl −β(l −ΔE)
= e . (22.64)
2 − 2
l
For comparison, let us consider the simplified model with Vsl = V and l =
lΔ. Then
∞
V 2 −βlΔ V 2 1
z= 2 e = 2 . (22.65)
1 − e−βΔ
l=0
The saddle point equation is
∂ Δ
ΔE = − ln z = βΔ (22.66)
∂β e −1
which determines the “temperature”
1 Δ
β= ln 1 + (22.67)
Δ ΔE
1
≈ if Δ ΔE. (22.68)
ΔE
Then
Δ
l = ≈ ΔE, (22.69)
−1
eΔ/ΔE
∂2 Δ2 eΔ/ΔE
2l − 2 = ln z = & '2 ≈ Δe ,
2
(22.70)
∂β 2 eΔ/ΔE − 1
V2 1 V 2 ΔE
z(ts ) = ≈ , (22.71)
2 1 − e−Δ/ΔE 2 Δ
and the rate becomes
V 2 ΔE 1 2π2 2πV 2 1 1
k= 2 e = e , (22.72)
Δ Δe2 Δ 2π
√
where the numerical value of e/ 2π = 1.08
240 22 Dynamics of an Excited State
We now want to discuss a displaced oscillator model (p. 205) for the transition
between the vibrational manifolds of two electronic states.
2π 2
k= V F CD(ΔE)
V2 ∞ (−it/)ΔE (it/)Hf −(it/)Hi V2 ∞
= 2 dt e e e = 2 dt e(−it/)ΔE F (t),
−∞ −∞
(22.73)
where Hf is the Hamiltonian of the nuclear vibrations in the final state and
the correlation function F (t) = exp(g(t)) for independent displaced oscillators
is taken from (31.12). The rate becomes
V2 ∞
k= 2 dt e(−it/)ΔE
−∞
# $
4 5
× exp gr2 (nr + 1)(eiωr t − 1) + nr (e−iωr t − 1) .
r
(22.74)
which gives for the rate in the classical limit the Marcus expression
2πV 2 1 (ΔE − Er )2
k= exp − . (22.77)
4πkB T Er 4Er kB T
which, after insertion into the saddle point equation, gives the approximation
ωeits ω
ΔE = S (22.87)
242 22 Dynamics of an Excited State
0
10
10–3
10–6
rate k
10–9
10–12
10–15
10–18
0 5 10 15
energy gap Δ
Fig. 22.9. Energy gap law. The rate (22.90) is shown as a function of the energy
gap in units of the average frequency for S = 0.2, 0.5, 1, 2
Problems
22.1. Ladder Model
Solve the time evolution for the ladder model approximately
⎛ ⎞
0 V ··· V
⎜ V E1 ⎟
⎜ ⎟
H =⎜ . .. ⎟ , Ej = α + (j − 1)Δ.
⎝ .. . ⎠
V En
22.4 Application to the Displaced Oscillator Model 243
First derive an integral equation for C0 (t) only by substitution. Then replace
the sum by integration over ω = j(Δ/) and extend the integration over the
whole real axis. Replace the integral by a δ-function and show that the initial
state decays exponentially with a rate
2πV 2
k= .
Δ
Part VII
Chlorophylls and carotenoids are very important light receptors. Both classes
of molecules have a large π-electron system which is delocalized over many
conjugated bonds and is responsible for strong absorption bands in the visible
region (Figs. 23.1 and 23.2).
O CH3 CH3
H CH3
2 3 H
CH3 1
I II 4
N N H
H Mg H
CH3 N N
IV III CH3
H
H V
CH2 H
O O
CH2 O
phytyl O O
74 CH3
Fig. 23.1. Structure of bacteriochlorophyll-b [62]. This is the principal green pig-
ment of the photosynthetic bacterium Rhodopseudomonas viridis. Dashed curves
indicate the delocalized π-electron system. Chlorophylls have an additional double
bond between positions 3 and 4. Variants have different side chains at positions 2
and 3
energy
unoccupied
orbitals
LUMO
HOMO
occupied
orbitals
2 ∂ 2
H=− + V (r) (23.3)
2me ∂x2
have to fulfill the boundary condition φ(0) = φ(L) = 0. Therefore, they are
given by
2 πsx
φs (x) = sin (23.4)
L L
with energies (Fig. 23.4)
2
2 L
2 πs
2 πsx
2 1 πs
Es = sin dx = . (23.5)
L 0 2me L L 2me L
Since there are 2N π-electrons, the energy of the lowest excitation is estimated
as
ΔE = En+1 − En
π 2 2 π 2 2
= ((N + 1)2
− N 2
) = . (23.6)
2me d2 (2N + 1)2 2me d2 (2N + 1)
V(x)
s=2 s=1
H d H
C C C C C C C C
H H H H H H H H
1
The end of the box is one bond length behind the last C-atom.
250 23 Photophysics of Chlorophylls and Carotenoids
which grows with increasing length of the polyene. The absorption coefficient
is proportional to
N 2 (N + 1)2
α ∼ μ2 ΔE ∼ , (23.8)
(2N + 1)3
which depends close to linear on the number of double bonds N .
where the atomic orbitals are centered on the nuclei and the coefficients are
determined from diagonalization of a certain one-electron Hamiltonian (for
instance, Hartree–Fock, Kohn–Sham, semiempirical approximations such as
AM1)
Hψ = Eψ. (23.10)
Inserting the LCAO wave function gives
Cs Hϕs (r) = E Cs ϕs (r) (23.11)
and projection on one of the atomic orbitals ϕs gives a generalized eigenvalue
problem:
0 = d3 r ϕs (r) Cs (H − E)ϕs (r)
s
= Cs (Hs s − ESs s ). (23.12)
s
23.4 Hückel Approximation 251
β β β β β β β
C C C C C C C C
The Hückel approximation [63] is a very simple LCAO model for the π-
electrons. It makes the following approximations (Fig. 23.5):
• The diagonal matrix elements Hss = α have the same value (Coulomb
integral) for all carbon atoms.
• The overlap of different pz -orbitals is neglected Sss = δss .
• The interaction between bonded atoms is Hss = β (resonance integral)
and has the same value for all bonds.
• The interaction between nonbonded atoms is zero Hss = 0.
The Hückel matrix for a linear polyene has the form of a tridiagonal matrix
⎛ ⎞
α β
⎜β α β ⎟
⎜ ⎟
⎜ .. .. .. ⎟
H =⎜ . . . ⎟ (23.13)
⎜ ⎟
⎝ β α β ⎠
β α
LUMO au
HOMO bg
au
bg
au
C2h E C2 σ i
Ag 1 1 1 1
Au 1 1 –1 –1
Bg 1 –1 –1 1
Bu 1 –1 1 –1
Fig. 23.7. Character table of the symmetry group C2h . All π-orbitals are antisym-
metric with respect to the reflection σ and are therefore of au - or bg -symmetry
to give2
π 1
ΔE = −4β sin ∼ for large n. (23.18)
4N + 2 N
This approximation can be improved if the bond length alternation is taken
into account, by using two alternating β-values [64]. The resulting energies
are
Ek = α ± β 2 + β 2 + 2ββ cos k, (23.19)
2
β is a negative quantity.
23.6 Cyclic Polyene as a Model for Porphyrins 253
S3 Ag
S2 Bu
S1 Ag
Ag Ag Ag Bu S0 Ag
1 iks
N
2π 2π 2π
φk = √ e ϕs , k = 0, , 2 , . . . , (N − 1) (23.22)
N s=1 N N N
254 23 Photophysics of Chlorophylls and Carotenoids
3 4 3 4
α β α β
2−
N N
2 5 2 5
2+
NH HN N Mg N
1 6 1 6
N N
δ γ δ γ
8 7 8 7
Fig. 23.9. Free base porphin and Mg-porphin. The bond character was assigned on
the basis of the bond lengths from HF-optimized structures
and eigenvalues
Ek = α + 2β cos k. (23.23)
This can be used as a model for the class of porphyrin molecules [70, 71]
(Fig. 23.9).
For the metal–porphyrin, there are in principle two possibilities for the
assignment of the essential π-system, an inner ring with 16 atoms or an outer
ring with 20 atoms, both reflecting the D4h -symmetry of the molecule. Since
it is not possible to draw a unique chemical structure, we have to count the
π-electrons. The free base porphin is composed of 14 H-atoms (of which the
peripheral ones are not shown), 20 C-atoms and 4 N-atoms which provide
a total of 14 + 4 × 20 + 5 × 4 = 114 valence electrons. There are 42 σ-
bonds (including those to the peripheral H-atoms) and two lonely electron
pairs involving a total of 88 electrons. Therefore, the number of π-electrons is
114 − 88 = 26. Usually, it is assumed that the outer double bonds are not so
strongly coupled to the π-system and a model with 18 π-electrons distributed
over the N = 16-ring is used. For the metal porphin, the number of H-atoms
is only 12 but there are four lonely electron pairs instead of two. Therefore,
the number of non-π valence electrons is the same as for the free base porphin.
The total number of valence electrons is again 114 since the Mg atom formally
donates two electrons.
9
−8 8
−7 7
−6 6
−5 5
−4 4
−3 3
−2 2
−1 1
k=0
1 −ik s iks
μ= e e ϕ∗s (r)(−er)ϕs (r)d3 r (23.24)
N
ss
which is approximately3
⎛ & '⎞
cos &s × 2π
1 −ik s iks N'
μ= e e δss (−eR) ⎝ sin s × π
N
⎠, (23.25)
N
ss 0
a b c
H y H y H
H H H
H x H x H
H H H
eV
y 8.6D
x y
2.0D (E) x x 10D
5 x,y
10D (E) x 8.1D
y 8.0D x 1.0D
4
x 0.5D
3 y 4.0D x 1.4D
0.9D (E)
y 7D
lower one is very weak and corresponds to the Q-band in the visible [75]. As
a result of ab initio calculations, the four orbitals of the Gouterman model
stay approximately separated from the rest of the MO orbitals [76]. If the
symmetry is disturbed as for chlorin, the Q-band splits into the lower Qy -band
which gains large intensity and the higher weak Qx -band. Figure 23.11 shows
calculated spectra for Mg-porphin, Mg-chlorin, and Mg-tetrahydroporphin.
S0
Chlorophyll Carotenoid
B
Bu
Ag
Q
T
T
S0 S0
the formation of triplet oxygen and for dissipation of excess energy [80]
(Figs. 23.12 and 23.13).
Problems
23.1. Polyene with Bond Length Alternation
C
C C
C C
C C
C
(d) Find a similar treatment for a linear polyene with odd number of C-atoms.
24
Incoherent Energy Transfer
Obviously, the last two determinants are the components of the triplet state
with Sz = ±1:
Linear combination of the first two determinants gives the triplet state with
Sz = 0:
1
|1, 0 = √ (| ↑↓ + | ↓↑) (24.5)
2
260 24 Incoherent Energy Transfer
excitonic exchange
D* A* D* A*
D A D A
Fig. 24.1. Intramolecular energy transfer. Left: the excitonic or Förster mech-
anism [81, 82] depends on transition intensities and spectral overlap. Right: the
exchange or Dexter mechanism [83] depends on the overlap of the wave functions
Here, energy can only be transferred by the exchange coupling [83]. Since
this involves the overlap of electronic wave functions, it is important at small
distances. In the singlet state, the excitonic interaction [81, 82] allows for
energy transfer also at larger distances.
262 24 Incoherent Energy Transfer
ρ1 ρ
2
R1 R2
r1 r2
D A
We take the position of the electrons relative to the centers of the molecules
1 1 1 R 1 3|R|(R)2 − |R|3 2
= − + + ··· (24.21)
|R + | |R| |R|2 |R| 2 |R|6
for
R = R1 − R2 , = 1 − 2 . (24.22)
The zero-order term vanishes due to the orthogonality of the orbitals. The
first-order term gives
e2
−2 R d3 1 d3 2 φ∗D∗ (r1 )φD (r1 )(1 − 2 )φ∗A (r2 )φA∗ (r2 ) (24.23)
4πε|R|3
and also vanishes due to the orthogonality. The second-order term gives the
leading contribution:
24.4 Energy Transfer Rate 263
e2
d3 1 d3 2 φ∗D∗ (r1 )φD (r1 )
4πε|R|6
×(3|R|(R1 − R2 )2 − |R|3 (1 − 2 )2 )φ∗A (r2 )φA∗ (r2 ). (24.24)
Only the integrals over mixed products of 1 and 2 are not zero. They can
be expressed with the help of the transition dipoles:
√
μD = 1 D|er|1 D∗ = 2 d3 1 φ∗D∗ (1 )e1 φD (1 ),
1 ∗
√ 3
μA = A|er| A = 2 d 2 φ∗A (2 )e2 φA∗ (2 ).
1
(24.25)
Finally1 the leading term of the excitonic interaction is the dipole–dipole term:
e2 & 2 '
Vexc = |R| μD μA − 3(RμD )(RμA ) . (24.26)
4πε|R|5
We neglect electron exchange between the two molecules and apply the Con-
don approximation. Then taking energies relative to the electronic ground
state |DA, we have
2π
k= Pi |i|H |f |2 δ(ωf − ωi )
i,f
1
= Pi i|H |f f |H |iei(ωf −ωi )t dt
2
i,f
1
= Pi i|H |f eiωf t f |H |eiωi t idt (24.31)
2
i,f
1
= i|Q−1 e−H0 /kB T H eitH0 / H eitH0 / |idt
2
i
1
= dt H (0)H (t) . (24.32)
2
Initially, only the donor is excited. Then the average is restricted to the
vibrational states of D∗ A:
1
k = 2 dt H (0)H (t)D∗ A . (24.33)
With the transition dipole operators
2ω 3
Ie (ω) = AD∗ D ge (ω) = |μD |2 ge (ω) (24.38)
3ε0 hc3
with the normalized lineshape function ge (ω).
The absorption cross section can be expressed with the help of the Einstein
ω
coefficient for absorption B12 as [86]
∗ ωga (ω)/c.
ω
σa (ω) = BAA (24.39)
π 2 c3 1 π
ω
BAA ∗ = AA∗ A = |μA |2
ωA∗ A
3 3 2 ε 0
3cε0 σa (ω)
α(ω) = = |μ̂A |2 ga (ω)
π ω
= Pi |i|μ̂A |f |2 δ(ωf − ωi − ω)
i,f
1
e−H/kB T
= dt e −iωt
i μ̂A f eiωf t f |μ̂A e−iωi t |i
2π i
Q
1 −iωt
= dt e μ̂A (0)μ̂A (t)A
2π
2
A detailed discussion of the relationships between absorption cross section and
Einstein coefficients is found in [86].
266 24 Incoherent Energy Transfer
EA*
0−0
α η
EA
Fig. 24.3. Absorption and emission. Within the displaced oscillator model, the
0 → 0 transition energy is located halfway between the absorption (α) and emission
maximum (η)
and3
3ε0 hc3 Ie (ω)
η(ω) = = |μD |2 ge (ω)
2ω 3
= Pf |f |μ̂D |i|2 δ(ωf − ωi − ω)
i,f
1
e−H/kB T
= dt e−iωt f eiωf t μ̂D i e−iωi t i|μ̂D |f
2π Q
f
1
= dt e−iωt μ̂D (t)μ̂D (0)D∗
2π
1
= dt eiωt μ̂D (0)μ̂D (t)D∗ . (24.40)
2π
Within the Condon approximation
8 9
μˆA (0)μ̂A (t)A = eiωA∗ t |μA |2 eitHA∗ / e−itHA /
A
= eiωA∗ t |μA |2 FA (t) (24.41)
and therefore the lineshape function is the Fourier transform of the correlation
function:
1
ga (ω) = dt ei(ω−ωA∗ )t FA (t). (24.42)
2π
Similarly,
8 9
μˆD (0)μ̂D (t)D∗ = e−iωD∗ t |μD |2 eitHD / e−itHD∗ /
D∗
= e−iωD∗ t |μD |2 FD∗ (t) (24.43)
3
Note the sign change which appears since we now have to average over the excited
state vibrations.
24.5 Spectral Overlap 267
and
1
ge (ω) = dt ei(ω−ωD∗ )t FD∗ (t). (24.44)
2π
With the inverse Fourier integrals
μ̂A (0)μ̂A (t)A = dωeiωt α(ω),
μ̂D (0)μ̂D (t)D∗ = dωe−iωt η(ω), (24.45)
(24.35) becomes
1 K2
k= dt dωe η(ω) dω e−iω t α(ω )
iωt
2 |R|6
1 K2
= dωdω η(ω)α(ω ) 2πδ(ω − ω )
2 |R|6
2π K 2
= dωη(ω)α(ω)
2 |R|6
2π K 2
= |μA | |μD |
2 2
dωge (ω)ga (ω). (24.46)
2 |R|6
This is the famous rate expression for Förster [81] energy transfer which
involves the spectral overlap of donor emission and acceptor absorption
[87, 88]. For optimum efficiency of energy transfer, the maximum of the accep-
tor absorption should be at longer wavelength than the maximum of the donor
emission (Fig. 24.4).
ω
ω
α(D)
D*
A*
α(A)
η(D)
D
A
η(A)
Fig. 24.4. Energy transfer and spectral overlap. Efficient energy transfer requires
good spectral overlap of donor emission and acceptor absorption. This is normally
only the case for transfer between unlike molecules
268 24 Incoherent Energy Transfer
LH1
RC
LH2
are mixed due to the excitonic interaction. The eigenstates are given by the
eigenvectors of the matrix
EA∗ B V
. (25.2)
V EAB∗
For a symmetric dimer, the diagonal energies have the same value and the
eigenvectors can be characterized as symmetric or antisymmetric
⎛ 1 ⎞ ⎛ 1 ⎞
√ − √1 √ √1
⎠ = EA∗ B − V
2 2 2 2
⎝ ⎠ EA∗ B V ⎝ .
V EA∗ B EA∗ B + V
√1
2
√1
2
− √1 √1
2 2
(25.3)
The two excitonic states are split by 2V .1 The transition dipoles of the two
dimer bands are given by
1
μ± = √ (μA ± μB ) (25.4)
2
and the intensities are given by
1 2
|μ± |2 = (μ + μ2B ± 2μA μB ). (25.5)
2 A
For a symmetric dimer μ2A = μ2B = μ2 and
C2
|+>
|A*B> 2V
|AB*>
|−>
absorption
Fig. 25.2. The “special pair” dimer. Top: the nearly C2 -symmetric molecular
arrangement for the reaction center of Rhodopseudomonas viridis (molekel graphics
[89]). The transition dipoles of the two molecules (arrows) are essentially antiparallel.
Bottom: the lower exciton component carries most of the oscillator strength
and therefore ⎛ ⎞
0
1 1
√ (μA + μB ) = √ ⎝ 0 ⎠ ,
2 2 2μ
Az
⎛ ⎞
2μ
1 1 ⎝ Ax ⎠
√ (μA − μB ) = √ 2μAy , (25.8)
2 2 0
which shows that the transition to the state |+ is polarized along the sym-
metry axis whereas the transition to |− is polarized perpendicularly. For the
special pair dimer, interaction with internal charge transfer states |A+ B− and
|A− B+ has to be considered. In the simplest model, the following interaction
matrix elements are important (Fig. 25.3).
The local excitation A∗ B is coupled to the CT state A+ B− by transferring
an electron between the two LUMOs
272 25 Coherent Excitations in Photosynthetic Systems
A+B − A−B +
UL UH UL
A*B AB*
Fig. 25.3. Extended dimer model. The optical excitations A∗ B and AB∗ are coupled
to the ionic states A+ B− and A− B+
A∗ B|H|A+ B−
1
= (A∗↑ A↓ − A∗↓ A↑ )B↑ B↓ H(B∗↑ A↓ − B∗↓ A↑ )B↑ B↓
2
= HA∗ ,B∗ = UL (25.9)
A∗ B|H|A− B+
1
= (A∗↑ A↓ − A∗↓ A↑ )B↑ B↓ H(A∗↑ B↓ − A∗↓ B↑ )A↑ A↓
2
= −HA,B = UH . (25.10)
AB∗ |H|A+ B−
1
= (B∗↑ B↓ − B∗↓ B↑ )A↑ A↓ H(B∗↑ A↓ − B∗↓ A↑ )B↑ B↓
2
= −HA,B = UH , (25.11)
AB∗ |H|A− B+
1
= (B∗↑ B↓ − B∗↓ B↑ )A↑ A↓ H(A∗↑ B↓ − A∗↓ B↑ )A↑ A↓
2
= HA∗ ,B∗ = UL . (25.12)
|A+B−> , |A−B+>
+
U− U+
|A*B> , |AB*>
−
⎛ ⎞
√1 √1
2 2
⎜ − √1 √1 ⎟
⎜ ⎟
S=⎜ 2 2
√1 √1 ⎠
⎟. (25.14)
⎝ 2 2
− √12 √12
PM 7.4 PL
10.8 10.5
BM BL
12.9 12.5
10.6 10.7
HM HL
Fig. 25.5. Transition dipoles of the reaction center Rps. viridis [90–93]. Arrows
show the transition dipoles of the isolated chromophores. Center–center distances
(as defined by the nitrogen atoms) are given in Å
Table 25.1. Excitonic couplings for the reaction center of Rps. viridis
HM BM PM PL BL HL
HM 160 33 −9 −11 6
BM 160 −168 −37 29 −11
PM 33 −168 770 −42 −7
PL −9 −37 770 −189 35
BL −11 29 −42 −189 167
HL 6 −11 −7 35 167
The matrix elements were calculated with the dipole–dipole
approximation for μ2 = 45 Debye2 for bacteriochlorophyll and
30 Debye2 for bacteriopheophytine [94]. All values in cm−1
16000
14000
PCT*(±)
H*(±)
12000
P*(+)
cm–1
B*(±)
P*(–)
10000
Fig. 25.6. Excitations of the reaction center. The absorption spectrum of Rps.
viridis in the Qy region is assigned on the basis of symmetric excitonic excita-
∗
tions [95]. The charge transfer states PCT are very broad and cannot be observed
directly
CN
μ n = SN
n
μ0 . (25.20)
The component parallel to the symmetry axis is the same for all monomers
μn,z = μ
(25.21)
276 25 Coherent Excitations in Photosynthetic Systems
φ + 2π
N
2π
N
φ
x
Without a magnetic field, the coupling matrix elements can be chosen real
and depend only on |m − n|. Within the dipole–dipole approximation, we
have furthermore3
3
A more realistic description based on a semiempirical INDO/S method is given
by [107].
278 25 Coherent Excitations in Photosynthetic Systems
|m − n| V|m−n| (cm−1 )
1 −352.5
2 −43.4
3 −12.6
4 −5.1
5 −2.6
6 −1.5
7 −0.9
8 −0.7
9 −0.6
The matrix elements are
calculated with the dipole–
dipole approximation (25.27)
for R = 26.5Å, φ0 = 66◦ , and
μ2 = 37 Debye2
with
2π
k= l, l = 0, 1, . . . , N − 1, (25.31)
N
N −1 N −1
1 ikn −ik n
k |H|k = e e n |H|n
N n=0
n =0
N −1 N −1
1 ikn −ik n
= e e H|n−n | . (25.32)
N n=0
n =0
We substitute
m = n − n (25.33)
to get
N −1 n−N +1
1 i(k−k )n+ik m
k |H|k = e H|m|
N n=0 m=n
N −1
= δk,k eik m H|m|
m=0
= δk,k (E0 + 2V1 cos k + 2V2 cos 2k + · · · ) . (25.34)
For even N , the lowest and highest states are not degenerate whereas for all
of the other states (Fig. 25.10)
Ek = EN −k = E−k . (25.35)
25.1 Coherent Excitations 279
E(k)
9
10 = – 8 8
11 = – 7 7
12 = – 6 6
13 = – 5 5
14 = – 4 4
13 = – 3 3
12 = – 2 2
11 = – 1 1
l=0
−π 0 π k
Fig. 25.10. Exciton dispersion relation. The figure shows the case of negative
excitonic interaction for which the k = 0 state is the lowest in energy
N −1 √
1
√ μz eikn = N μz δk,0 . (25.37)
N n=0
±i
μk,± = √12 √ 2
0 μk
⎛ ⎞⎛ ⎞
N ) − sin( N )
cos( 2πn 2πn
N−1
μx
1 ⎜ ⎟
= √ eikn √12 ± √i2 0 ⎝ sin( 2πn ) cos( 2πn ) ⎠ ⎝ μy ⎠
N n=0 N N
μz
1
⎛ ⎞
1 ikn √1 ±i2nπ/N √i ±i2nπ/N
⎝ x ⎠
N −1 μ
= √ e 2
e ± 2
e 0 μy
N n=0 μ z
1
N −1
±i μx
= √ ei(k±2π/N ) √12 √ 2
N n=0 μy
√ 1 1
= N δk,∓2π/N μ± with μ± = √ (μx ± iμy ) = √ μ⊥ e±iφ . (25.38)
2 2
280 25 Coherent Excitations in Photosynthetic Systems
9
10 = – 8 8
11 = – 7 7
12 = – 6 6
13 = – 5 5
14 = – 4 4
13 = – 3 3
12 = – 2 2
11 = – 1 1
0
ground state
Fig. 25.11. Allowed optical transitions. The figure shows the case of negative
excitonic interaction where the lowest three exciton states carry intensity
−i
cos Θ sin Θ 0 = √ eiΘ √12 √ 2
0 + √ e−iΘ √12 √i
2
0 (25.39)
2 2
couples to a linear combination of the k = ±2π/N states since
N
μk,Θ = μ⊥ ei(Φ−Θ) δk,−2π/N + e−i(Φ−Θ) δk,2π/N . (25.40)
2
The absorption strength does not depend on the polarization angle Θ
|μ−2π/N,Θ |2 + |μ2π/N,Θ |2 = N μ2⊥ . (25.41)
z
φβ θα θβ
Rβ
2ν α1 β1
φα
Rα
y
y x
Fig. 25.12. BChl arrangement for the LH2 complex of Rps. acidophila [99–103].
Arrows represent the transition dipole moments of the BChl molecules as calcu-
lated on the 631G/HF-CI level. Distances are with respect to the center of the four
nitrogen atoms. Rα = 26.1 Å, Rβ = 26.9 Å, 2ν = 20.7◦ , φα = 70.5◦ , φβ = −117.7◦ ,
θα = 83.7◦ , and θβ = 80.3◦
Vαβ1
Vββ1 Vαα1
β1
β2 α2 α1
βN
V βα 1V dim
αN
Table 25.3. Excitonic couplings for the LH2 complex of Rps. acidophila
−1 N −1
1 & ' N Hαα|m| Hαβm Csα
= Cs α Cs β ei(k−k )n+ik m ,
N Hβαm Hββ|m| Csβ
n=0 m=0
N −1 N −1
& ' m=0 eikm Hαα|m| m=0 eikm Hαβm Csα
= δk,k Cs α Cs β N −1 N −1 .
ikm ikm Csβ
m=0 e Hβαm m=0 e Hββm
(25.48)
The coefficients Csα and Csβ are determined by diagonalization of the matrix:
Eα + 2Vαα,1 cos k + · · · Vdim + eik Vαβ,1 + e−ik Vαβ,−1 + · · ·
Hk = −ik .
Vdim + e Vαβ,1 + e Vαβ,−1 + · · ·
ik
Eα + 2Vββ,1 cos k + · · ·
(25.49)
25.1 Coherent Excitations 283
are given by
1 1
√ , (25.51)
2 ±eiχ
where the angle χ is chosen such that
with
U (k) = sign(V )|V + eik W |
and the eigenvalues are
1
Ek,± = Eα ± U (k) = Eα ± sign(V ) Vdim
2 + W 2 + 2V
dim W cos k. (25.53)
Similar selection rules as for the simple ring system follow for the first factor
N −1
& ' μ ikn & '
0 0 1 μk,± = √ e cos θα ± eiχ cos θβ
2N n=0
N
= δk,0 μ (cos θα ± cos θβ ) , (25.55)
2
N −1
1 i(k±2π/N )n & '& '
√1 √i
2 2
0 μk,± = √ e 1 i 0 μα,1 ± eiχ μβ,1
2N n=0
284 25 Coherent Excitations in Photosynthetic Systems
N
= δk,∓2π/N μ sin θα ei(φα −ν) ± eiχ sin θβ ei(φ,Mβ +ν) ,
2
(25.56)
N −1
1 i(k±2π/N )n & '& '
−i
√1 √
2 2
0 μk,± = √ e 1 −i 0 μα,1 ± eiχ μβ,1
2N n=0
N
0= δk,∓2π/N μ sin θα ei(φα −ν) ∓ eiχ sin θβ ei(φβ +ν) .
2
(25.57)
The second factor determines the distribution of intensity among the + and −
states.
In the limit of 2N -equivalent molecules, V = W , 2ν +φβ −φα = π+(π/N ),
and θα = θβ and we have
hence
χ = k/2, U (k) = 2V cos k/2, (25.59)
and (25.56) becomes
1 i N
√ √ 0 μk,± = δk,∓2π/N μ sin θei(φα −ν) 1 ∓ ei(k/2+π/N ) ,
2 2 2
In the LH2 complex, the transition dipoles of the two dimer halves are
nearly antiparallel. The oscillator strengths of the N molecules are concen-
trated in the degenerate next to lowest transitions k = ±2π/N . This means
that the lifetime of the optically allowed states will be reduced compared
to the radiative lifetime of a monomer. In the LH2 complex, the transi-
tion dipoles have only a very small z-component. Therefore in a perfectly
25.2 Influence of Disorder 285
E(k)
circ circ
circ circ
−2π/N 0 2π/N π k
δE
k|H| − k = . (25.66)
N
The zero-order eigenstates are the linear combinations
1
|k± = √ (|k ± | − k) (25.67)
2
with zero-order energies (Fig. 25.15)
δE
E(k) = Ek + (nondegenerate states),
N
δE
E(k− ) = Ek , E(k+ ) = Ek + 2 (degenerate states). (25.68)
N
Only the symmetric states are affected by the perturbation. The interaction
between degenerate pairs is given by the matrix elements
(−) (+)
.. ..
CN .. .. . .
. .
−2 2
y
−1 1
δE
x l=0
Fig. 25.15. Local perturbation of the symmetry. This figure shows the case of
δE > 0
25.2 Influence of Disorder 287
δE
k+ |H|k+ = δk,k Ek + 2 ,
N
k− |H|k− = δk,k Ek ,
k+ |H|k− = 0, (25.69)
Optical transitions to the |1± states are linearly polarized with respect to
perpendicular axes. First-order intensity is transferred from the |1+ state to
the |0 and the other |k+ states and hence the |1− state absorbs stronger
than the |1+ state which is approximately
√
2δE 2δE
|1+ + |0 + |2+ + · · · (25.71)
N (E0 − E1 ) N (E2 − E1 )
Its intensity is reduced by a factor of
1
. (25.72)
1+ δE 2
N2
2
(E0 −E1 )2 + 4
(E2 −E1 )2 + ···
for all possible values of Δk. In this section, we discuss the most impor-
tant Fourier components separately. These are the Δk = ±4π/N components
which mix the k = ±2π/N states and the Δk = ±2π/N components which
couple the k = ±2π/N to the k = 0, ±4π/N states and thus redistribute the
intensity of the allowed states.
A general modulation of the diagonal energies with Fourier components
Δk = ±2π/N is given by
eiχ0 eiχ0 √
μ2π/N,+,Θ = √ μ−2π/N,Θ + √ μ2π/N,Θ = N μ⊥ cos(Φ + χ0 − Θ),
2 2
e−iχ0 eiχ0 √
μ2π/N,−,Θ = − √ μ−2π/N,Θ + √ μ2π/N,Θ = i Nμ⊥ sin(Φ + χ0 − Θ).
2 2
(25.77)
(−) (+)
.. .. .. ..
. . . .
−2 2
−1 1
l=0
.. .. .. ..
. . . .
−2 2
−1 1
l=0
we find
− pn E = −p⊥ E⊥ cos(Φ + n2π/N − ξ) (25.85)
and
1 t 1 −n
E αn E = (Et SN
n
)α0 (SN E)
2 2
E2 +
= ⊥ αxx cos2 (n2π/N − ξ) + αyy sin2 (n2π/N − ξ)
2
+ 2αxy sin(n2π/N − ξ) cos(n2π/N − ξ)}
+ E| E⊥ {αxz cos(n2π/N − ξ) + αyz sin(n2π/N − ξ)}
E
2
+ αzz . (25.86)
2
The quadratic term has Fourier components Δk = ±4π/N and can
therefore act as an elliptic perturbation.
Let us assume that the polarizability is induced by the coupling to charge
transfer states like in the special pair dimer. Due to the molecular arrange-
ment, the polarizability will be zero along the CN -symmetry axis and will
have a maximum along a direction in the xy-plane. We denote the angle of
this direction by η. The polarizability α0 simplifies to
⎛ ⎞
α⊥ cos2 η α⊥ sin η cos η 0
α0 = ⎝ α⊥ sin η cos η α⊥ sin2 η 0 ⎠ (25.87)
0 0 0
contains energy shifts δEn which are assumed to have a Gaussian distribution
function:
1
P (δEn ) = √ exp(−δEn2 /Δ2 ). (25.90)
Δ π
25.2 Influence of Disorder 291
and due to the disorder the k-states are mixed. The energy change of a non-
degenerate state (for instance k = 0) is in lowest-order perturbation theory
given by the diagonal matrix element
1
δEk = δEn (25.92)
N n
as the average of the local energy fluctuations. Obviously, the average is zero
and
1
δEk2 = δEn δEn . (25.94)
N2
n,n
5
This is known as exchange or motional narrowing.
292 25 Coherent Excitations in Photosynthetic Systems
δE 2
=2 . (25.99)
N
Consider now fluctuations also of the coupling matrix elements, for instance
due to fluctuations of orientation and distances. The Hamiltonian contains an
additional perturbation
1 −ik n +ikn
Hkk = e δVnn . (25.100)
N
nn
δVnn = 0, (25.101)
δVnn δVmm = (δnm δn m + δnm δn m − δnm δn m δnn )δV|n−n
2
| ,
(25.102)
it follows
Hkk = 0 (25.103)
If the dominant contributions come from the fluctuation of site energies and
nearest neighbor couplings, this simplifies to
1
|Hkk | =
2
(δE 2 + 2δV±1
2
(1 + cos(k + k )). (25.105)
N
6
We assume here that the fluctuation amplitudes obey the CN -symmetry.
25.2 Influence of Disorder 293
Problems
25.1. Photosynthetic Reaction Center
The “special pair” in the photosynthetic reaction center of Rps. viridis is
a dimer of two bacteriochlorophyll molecules whose centers of mass have a
distance of 7 Å. The transition dipoles of the two molecules include an angle
of 139◦.
C2
139º
μA 7Å μB
Calculate energies and intensities of the two dimer bands from a simple
exciton model
−Δ/2 V
H=
V Δ/2
as a function of the energy difference Δ and the interaction V . The Hamilto-
nian is represented here in a basis spanned by the two localized excited states
|A∗ B and |B∗ A.
294 25 Coherent Excitations in Photosynthetic Systems
9
H= {Eα |n; αn; α| + Eβ |n; βn; β|
n=1
+ Vdim (|n; αn; β| + |n; βn; α|)
+ Vβα,1 (|n; αn − 1; β| + |n; βn + 1; α|)
+ Vαα,1 (|n; αn + 1; α| + |n; αn − 1; α|)
+ Vββ,1 (|n; βn + 1; β| + |n; βn − 1; β|)} .
1 ikn 1 ikn
9 9
|k; α = e |n; α, |k; β = e |n; β,
3 n=1 3 n=1
k , α(β)|H|k, α(β) = 0 if k = k .
Hαα (k) = k; α|H|k; α, Hββ (k) = k; β|H|k; β, Hαβ (k) = k; α|H|k; β.
The total reorganization energy is the sum of all partial reorganization energies
(ωr Δqr )2
ER = = gr2 ωr (26.3)
r
2 r
1
Which have different equilibrium positions in the two states |i and |f .
298 26 Ultrafast Electron Transfer Processes
Fig. 26.1. Electron transfer processes in the reaction center of bacterial photosyn-
thesis. After photoexcitation of the special pair dimer (red ), an electron is transferred
to a bacteriopheophytine (blue) in few picoseconds via another bacteriochlorophyll
(green). The electron is then transferred on a longer time scale to the quinone QA
and finally to QB via a nonheme ferreous ion. The electron hole in the special pair
is filled by an electron from the hemes of the cytochrome c (orange). After a sec-
ond excitation of the dimer, another electron is transferred via the same route to
the semiquinone to form ubiquinone which then diffuses freely within the mem-
brane. This figure was created with Rasmol [106] based on the structure of Rps.
viridis [90–93] from the protein data bank [104, 105]
26 Ultrafast Electron Transfer Processes 299
2πV 2
ket = P ({nr }) (FCr (nr , nr ))2 δ ΔE + ωr (nr − nr )
n n r r
r r
2
2πV
= FCD(ΔE), (26.5)
where the Franck–Condon-weighted density of states FCD(ΔE) can be
expressed with the help of the time correlation formalism (18.2) as
1
FCD(ΔE) = dt e−itΔE/ F (t) (26.6)
2π
with
& '
F (t) = exp gr2 (eiωt − 1)(nr + 1) + (e−iωt − 1)nr . (26.7)
r
Quantum chemical calculations for the reaction center [116–119] put the elec-
tronic coupling in the range of V = 5–50 cm−1 for the first step P∗ → P+ B−
and V = 15–120 cm−1 for the second step P+ B− → P+ H− . For a rough esti-
mate, we take a reorganization energy of 2,000 cm−1 and a maximum of the
Franck–Condon-weighted density of states FCD ≈ 1/2,000 cm−1 yielding an
electronic coupling of V = 21 cm−1 for the first and V = 42 cm−1 for the
second step in quite good agreement with the calculations.
27
Proton Transfer in Biomolecules
Intraprotein proton transfer is perhaps the most fundamental flux in the bio-
sphere [120]. It is essential for such important processes as photosynthesis,
respiration, and ATP synthesis [121]. Within a protein, protons appear only
in covalently bound states. Here, proton transfer is essentially the motion
along a hydrogen bond, for instance peptide or DNA H-bonds (Fig. 27.1) or
the more complicated pathways in proton-pumping proteins.
a b
O C
N H O C H
N H O C H O
N N
N H
N
H O C N H N
N H O C
N H O C N N
N
H
Fig. 27.1. Proton transfer in peptide H-bonds (a) and in DNA H-bonds (b)
The energy barrier for this reaction depends strongly on the conforma-
tion of the reaction complex, concerning as well its geometry as its protonation
state. Due to the small mass of the proton, the description of the proton
transfer process has to be discussed on the basis of quantum mechanics.
ASP96
5.0
2.9
6.8
LYS216
ASP85
2.9 2.6
RET
2.6
2.8
ARG82
2.5
3.2
3.3 5.7
3.4
3.5
GLU194
2.3
GLU204
Fig. 27.2. X-ray structure of bR. The most important residues and structural waters
are shown, together with possible pathways for proton transfer. Numbers show O–O
and O–N distances in Å. Coordinates are from the structure 1c3w [122] in the protein
data bank [104, 105]. Molekel graphics [89]
BR ASP96
570
11 13 14
light
12 N
H+
O 5 ms
ASP85 4 ps
640 J/K
ASP96 ARG82 590
PRG ASP96
11 13 14
11 13 14
12 N 12
H+ NH+
ASP85
5 ms ARG82 H+
ASP85
H+ PRG 1 μs
ARG82
cytoplasmic PRG
side
ASP96
11 13 14 ASP96
11 13 14
12
+ 12
NH
NH+
N L
560 ASP85 550
ARG82 ASP85
10 μs ARG82
PRG ASP96 PRG
5 ms 11 13 14 H+
12
N
H+
extracellular
M ASP85
410 side
ARG82
PRG
Fig. 27.3. The photocycle of bacteriorhodopsin [123]. Different states are labeled
by the corresponding absorption maximum (nm)
11 13 +
14 Lys216
12 N
H
11 13 14
12 +
NH
Lys216
H = Tel + Tp + TN + V. (27.2)
(Tel + V (r; Rp , Q))ψs (r; Rp , Q) = Eel,s (Rp , Q)ψs (r; Rp , Q). (27.3)
In the second step, only the heavy atoms (Q) are fixed but the action of the
kinetic energy of the proton on the electronic wave function is neglected. Then
the wave function of proton state n obeys
(Tp + Eel,s (Rp ; Q))χs,n (Rp ; Q) = εs,n (Q)χs,n (Rp ; Q). (27.4)
The electronic energy Eel (Rp , Q) plays the role of a potential energy surface
for the nuclear motion. It is shown schematically in Fig. 27.5 for one proton
1.2
1
0.8
0.6 0.2
0.4
0.1
0.2
0 0 y
– 0.2 – 0.1
–1
–0.5 0
0.5 – 0.2
1
x 1.5 2
2.5
Fig. 27.5. Proton transfer potential. The figure shows schematically the potential
which, as a function of the proton coordinate x, has two local minima corresponding
to the structures · · · N–H · · · O · · · and · · · N · · · H–O · · · and which is modulated by
a coordinate y representing the heavy atoms
27.3 Nonadiabatic Proton Transfer (Small Coupling) 305
Rp Rp Rp
Fig. 27.6. Proton tunneling. The double well of the proton along the H-bond is
shown schematically for different configurations of the reaction complex
coordinate (for instance, the O–H bond length) and one promoting mode of
the heavy atoms which modulates the energy gap between the two minima.
For fixed positions Q of the heavy nuclei, Eel (Rp , Q) has a double-well
structure as shown in Fig. 27.6. The rate of proton tunneling depends on the
configuration and is most efficient if the two localized states are in resonance.
Finally, the wave function of the heavy nuclei in state α is the approximate
solution of
(TN + εs,n (Q))Φs,n,α (Q) = Es,n,α Φs,n,α (Q). (27.5)
(ΔG0 + ER )2
ΔG‡00 = (27.7)
4ER
and for the transition (s, n) → (s , m), we have approximately
εsm εs’n
εs0
εs’0
#
ΔGo ΔG00 Q
theory (Chap. 16) and the reaction rate is given by Borgis and Hynes [124] for
symmetrical proton transfer (ER
ΔG0 , ΔG‡00 = ER /4):
2πV02 1 ER
k= exp − . (27.15)
4πkT ER 4kT
In the derivation of (27.15), the Condon approximation has been used which
corresponds to application of (27.12) and approximation of the overlap integral
(27.14) by a constant value. However, for certain modes (which influence the
distance of the two potential minima), the modulation of the coupling matrix
element can be much stronger than in the case of electron transfer processes
due to the larger mass of the proton. The dependence on the transfer distance
(0) (0)
δQ = Qs − Qs is approximately exponential:
2 α2
ΔGtot = ΔG‡ + (27.19)
8M
and the additional factor χSC is given by
γSC kT
χSC = exp γSC 1 − (27.20)
4ΔGtot
with
αD
γSC = , (27.21)
M Ω2
where
∂ΔH
D= (Q0 ) (27.22)
∂Q
measures the modulation of the energy difference by the promoting mode Ω.
308 27 Proton Transfer in Biomolecules
If V el is large and the tunnel factor approaches unity (for s = s ), then the
application of the low-friction result (8.29) gives
ω ΔG‡nmad
knm = exp − , (27.23)
2π kT
A particular class of proteins (molecular motors) can move linearly along its
designated track, against an external force, by using chemical energy. The
movement of a single protein within a cell is subject to thermal agitation
from its environment and is, therefore, a Brownian motion with drift.
The following discussion is based on the pioneering work by Jülicher and
Prost [125–128].
we find
∂W ∂ ∂
= − S = − (W v) (28.6)
∂t ∂x ∂x
hence the drift velocity at x is
S(x)
vd (x) = (28.7)
W (x)
and the average velocity of systems in the interval 0, L is
L L
0 W (x)vd (x)dx 0 S(x)dx
v = L = L . (28.8)
0 W (x)dx 0 W (x)dx
Equation (28.6) shows that for a stationary solution, the flux has to be con-
stant. However, the stationary distribution is easily found by integration and
has the form
mγS x U(x )/kB T
Wst = e−U (x)/kB T C − e dx . (28.9)
kB T 0
From the periodic boundary condition, we find
L
−U (0)/kB T mγ U (x )/kB T
Wst (L) = e C− S e dx
kB T 0
which shows that in a stationary state, there is zero flux (Fig. 28.1)
S=0 (28.11)
U(x)
S>0
W(x)
F>0
F=0 S=0
W (0) = W (L),
L
−U (0)/kB T mγ
Ce =e (LF−U(0))/kB T
C− S e (U (x)−xF )/kB T
dx ,
kB T 0
(28.15)
From
kB T ∂ 1 ∂
Sst = − Wst (x) − Wst (x) (U (x) − xFext ) , (28.17)
mγ ∂x mγ ∂x
we find
1 ∂
mγS = − (kB T ln W (x) + U + Uext ), (28.18)
W (x) ∂x
which can be interpreted in terms of a chemical potential [129]
as
∂ mγS
μ(x) = − = −mγvd (x), (28.20)
∂x W (x)
where the right-hand side gives the energy which is dissipated by a particle
moving in a viscous medium.
kij
i j. (28.21)
kji
314 28 Continuous Ratchet Models
P M−ADP−P
3
M−ADP 4 M−ATP
2
M
ADP ATP
kB T ∂ 1 ∂
Si (x, t) = − Wi (x, t) − Wi (x, t) Ui (x). (28.25)
mγ ∂x γm ∂x
P M−ADP−P
M−ADP β1 β2 α2 α1 M−ATP
M
ADP ATP
U1,0
U2,1
U1,1
U2,2
The motor protein undergoes a chemical cycle which can be divided into four
states. In a simplified two-state model, this cycle has to be further divided into
two substeps. One option is to combine ATP binding and hydrolysis (Fig. 28.3)
α1
M+ATP M−ADP−P (28.28)
α2
Since each cycle consumes the energy from hydrolysis of one ATP, we consider
a sequence of states characterized by the motor state and the number of total
cycles (Fig. 28.4)
α1 β2 α1
· · · M1,0 M2,1 M1,1 M2,2 · · · (28.30)
α2 β1 α2
The potential energies Us,n and Us,n+m only differ by a vertical shift
corresponding to the consumption of m energy units
and therefore
k21 = α2 + β2 , (28.39)
k12 = α1 + β1 = α2 e(U1 −U2 +Δμ)/kB T + β2 e(U 1 −U2 )/kB T . (28.40)
k12 (x)
= e−(U2 −U1 )/kB T , (28.41)
k21 (x)
which indicates that the system is not chemically driven and is only subject
to thermal fluctuations. The steady state is then again given by
for which the flux is zero. For Δμ > 0, the system is chemically driven and
spontaneous motion with v
= 0 can occur. However, for a symmetric system
with U1,2 (−x) = U1,2 (x), the steady-state distributions are also symmetric
28.3 The Two-State Model 317
Wst
U(x)
WF F(x)
on the other hand, is generally not zero in this case since α1,2 are symmet-
ric functions. Therefore, two conditions have to be fulfilled for spontaneous
motion to occur: detailed balance has to be broken (Δμ
= 0) and the system
must have polar symmetry.
1−p2
p2
1−p1
p1
since
∂U ∂S
S dx = − U (x) dx = 0. (28.56)
∂x ∂x
Under the action of an external force Fext , the velocity and the hydrolysis rate
are functions of the external force and Δμ:
v = v(Fext , Δμ),
r = r(Fext , Δμ). (28.57)
Close to equilibrium, linearization gives
v = λ11 Fext + λ12 Δμ,
r = λ21 Fext + λ22 Δμ. (28.58)
The rate of energy dissipation is given by mechanical plus chemical work
2
Π = Fext v + rΔμ = λ11 Fext + λ22 Δμ2 + (λ12 + λ21 )Fext Δμ (28.59)
which must be always positive due to the second law of thermodynamics. This
is the case if
λ11 > 0, λ22 > 0, and λ11 λ22 − λ12 λ21 > 0. (28.60)
The efficiency of the motor can be defined as
−Fext v −λ11 Fext
2
− λ12 Fext Δμ
η= =
rΔμ λ22 Δμ2 + λ21 Fext Δμ
or shorter
λ11 a2 + λ12 a Fext
η=− , a= . (28.61)
λ22 + λ21 a Δμ
It vanishes for zero force but also for v = 0 and has a maximum at (Figs. 28.8
and 28.9)
√
1 − (λ12 λ21 /λ11 λ22 ) − 1 (1 − 1 − Λ)2 λ12 λ21
a= , ηmax = , Λ= .
λ21 /λ22 Λ λ11 λ22
(28.62)
320 28 Continuous Ratchet Models
0.8
velocity v/Δμ
0.6
0.4
0.2
0
–1 –0.8 –0.6 –0.4 –0.2 0
external force F/Δμ
0.4
efficiency η
0.2
0
–1 –0.5 0
external force F/Δμ
Problems
28.1. Deviation from Equilibrium
Make use of the detailed balance conditions
α1
= e(Δμ−ΔU)/kB T ,
α2
β1
= e−ΔU/kB T ,
β2
and calculate the quantity
k12 α1 + β1
Ω= − e−ΔU/kB T = − e−ΔU/kB T .
k21 α2 + β2
Use the approximation
C(ATP)
Δμ = Δμ0 + kB T ln
C(ADP)C(P)
28.4 Operation Close to Thermal Equilibrium 321
Δμ = 0 but Δμ kB T
and
Δμ
kB T.
29
Discrete Ratchet Models
For example, the four-state model of the ATP hydrolysis can be easily trans-
lated into such a scheme (Fig. 29.1).
M1
ATP
M2
ADP−P
M2
ADP−P
M1
ADP
M1
xn xn+1
G(x)
ΔG1
ΔG2 F=0
F< 0
x1 x2 x
d
Fig. 29.2. Influence of the external force. The external force changes the activation
barriers which become approximately ΔG1 (F ) = ΔG1 (0) − x1 F and ΔG2 (F ) =
ΔG2 (0) + x2 F
vo
−Fstal −F
α2 β20 e−(1−Θ)Fd/kB T
S = (eΔμ/kB T +Fd/kB T − 1) ,
α1 + α2 + β10 eΘFd/kB T + β20 e−(1−Θ)Fd/kB T
(29.13)
Γ − 1 = eΔμ/kB T +Fd/kB T − 1
Δμ
= e(1−F/Fstal )Δμ/kB T − 1 ≈ (1 − F/Fstal ) , (29.15)
kB T
Fd (β10 + (α1 + α2 )(1 − Θ))
ω ≈ ω0 1 − . (29.16)
kB T (α1 + α2 + β10 + β20 )
Part IX
Appendix
A
The Grand Canonical Ensemble
E,N T, μ
Fig. A.1. Ensemble of systems exchanging energy and particles with a reservoir
We search for the maximum of (A.5) under the restraints imposed by (A.2)
and (A.3). To this end, we use the method of undetermined factors (Lagrange
method) and consider the variation
⎛ ⎞
δ ⎝ln W − α nj − M − β nj Ej − Etot − γ nj Nj − Ntot ⎠
j j j
= δnj (− ln nj − α − βEj − γNj ) = 0. (A.6)
j
nj e−βEj −γNj
P (Ej , Nj ) = = (A.10)
j nj Ξ
and further
M
e−α = . (A.11)
Ξ
A.2 Connection to Thermodynamics 331
From
M
nj Ej = Ej e−βEj −γNj = Etot , (A.12)
j
Ξ j
which can be written as a sum over canonical partition functions with different
particle numbers
Ξ= eβμN e−βE(N ) = eβμN Q(N ). (A.20)
N E N
B
Time Correlation Function of the Displaced
Harmonic Oscillator Model
With the help of this relation, the single mode correlation function becomes
8 † † † †
9
Fr (t) = e−itωr br br e−gr (br −br ) eitωbr br egr (br −br ) . (B.3)
we find
& ' & '
Fr (t) = exp −gr (b†r e−iωr t − br eiωt ) exp gr (b†r − br ) . (B.5)
The commutator is
and we have
1
Fr (t) = exp − gr2 (e−iωt − eiωt )
2
& '
× exp −gr b†r (e−iωt − 1) + gr br (eiωt − 1) . (B.8)
dA 1 2 d2 A
A = A(0) + g + g + ··· (B.14)
dg 2 dg 2
The derivatives are
dA
= [b† b, b† − b] = b† [b, b† ] + [b† , −b]b = (b + b+ ),
dg g=0
d2 A
= [[b† b, b† − b], b† − b] = [b + b† , b† − b] = 2,
dg 2 g=0
dn A
= 0 for n ≥ 3, (B.15)
dg n g=0
Consider † †
A = eτ b b be−τ b b
with τ = −iωt. (B.18)
Make again a series expansion
dA
= [b+ b, b] = −b, (B.19)
dτ
d2 A
= [b+ b, −b] = b, etc., (B.20)
dτ 2
τ2 (itω)2
A=b 1−τ + − · · · = b 1 + itω + + · · · = b eiωt . (B.21)
2 2
Hermitian conjugation gives
† †
e−iωtb b b+ eiωtb b
= b† e−iωt . (B.22)
df (τ )
= e−Bτ (−A − B)e−Aτ e(A+B)τ + e−Bτ e−Aτ (A + B)e(A+B)τ
dτ
= e−Bτ [e−Aτ , B]e(A+B)τ . (B.24)
and therefore
∞ (−τ )n
(−τ )n n
[e−Aτ , B] = [A , B] = An−1 [A, B]
n=0
n! n
(n − 1)!
= −τ [A, B]e−Aτ (B.26)
df (τ )
= −τ [A, B]e−Bτ e−Aτ e(A+B)τ = −τ [A, B]f (τ ) (B.27)
dτ
which is for the initial condition f (0) = 1 solved by
2
τ
f (τ ) = exp − [A, B] . (B.28)
2
This derivation is based on [131]. For one single oscillator, consider the linear
combination
μb† + τ b = A + B, (B.30)
[A, B] = −μτ. (B.31)
Application of (B.29) gives
8 † 9 8 † 9
eμb +τ b = eμb eτ b eμτ /2 (B.32)
and due to the cyclic invariance of the trace operation, the right-hand side
becomes † †
Q−1 tr(eτ b e−βωb b eμb )eμτ (B.36)
which can be written as
† † † †
Q−1 tr(e−βωb b e+βωb b eτ b e−βωb b eμb )eμτ
8 † † †
9
= e+βωb b eτ b e−βωb b eμb eμτ . (B.37)
for large n. If the function φ(x) has a maximum at x0 , then the integrand
also has a maximum there which becomes very sharp for large n. Then the
integral can be approximated by expanding the exponent around x0
1 d2 (nφ(x))
(x − x0 ) + · · ·
2
nφ(x) = nφ(x0 ) + (C.2)
2 dx2 x0
as a Gaussian integral
%
∞
2π
e nφ(x)
dx ≈ e nφ(x0 )
. (C.3)
−∞ n|φ (x0 )|
If the integration contour is deformed such that the imaginary part is constant,
then the Laplace method is applicable and gives
nφ(z) in(φ(z0 ))
e dz = e en(φ(z)) dz. (C.5)
C C
φ (z0 ) = 0, (C.6)
Problems of Chap. 1
N
(b) rN − r0 = Δri ,
i=1
N
P Δri = R
i=1
∞ ∞
N
N
= d Δr1 · · ·
3 3
d ΔrN P (Δri )δ R− Δri
−∞ −∞ i=1 i=1
N
∞ ∞
1 1 27
= d3 keikR d Δr1 · · ·
3 3
d ΔrN
(2π)3 −∞ −∞ b3 8π 3
3Δr2
× exp − 2i − ikΔri
2b
1
= d3 keikR
(2π)3
342 Solutions
/
0N
∞
1 27 3Δr2
× d Δr exp − 2 − ikΔr
3
b3 8π 3 −∞ 2b
1 N b2 2
= d ke
3 ikR
exp − k
(2π)3 6
1 27 3R2
= 3 exp − .
b 8π 3 N 3 2N b2
1 f
(c) exp − Δri − κ
2
Δri
kB T 2
1 f
= exp − Δr2i − κΔri
i
kB T 2
f 3 3kB T
= 2 →f = .
2kB T 2b b2
Nκ
(d) xN − x0 = , yN − y0 = zN − z0 = 0,
f
N κb2
L = xN − x0 = .
3kB T
1.2. Three-Dimensional Polymer Model
(a) N b2 .
2 1 + x 2x(x2 − 1)
(b) b N + with x = cos θ
1−x (1 − x)2
1 + cos x
≈ N b2 .
1 − cos x
(1 + cos θ1 )(1 + cos θ2 )
(c) N b2 .
1 − cos θ1 cos θ2
(d) N
a/b with the coherence length
1 + cos x
a=b .
(1 − cos x) cos(x/2)
4 4 8
(e) N b 2
−1 −b 2
− 4 .
θ2 θ2 θ
1.3. Two-Component Model
1 M lα − L
(a) κ = −kB T ln
lα − lβ L − M lβ
1 1 lα − lβ lα − lβ
−kB T + + − .
2M lα − 2L 2M lβ − L 12(L − M lβ )2 12(M lα − L)2
Solutions 343
The exact solution can be written with the digamma function Ψ which is
well known by algebra programs as
1 L − M lβ M lα − L
κ = −kB T −Ψ +1 +Ψ .
lα − lβ lα − lβ lα − lβ
The error of the asymptotic expansion is largest for L ≈ M lα or L ≈
M lβ . The following table compares the relative errors of the Stirling’s
approximation and the higher-order asymptotic expansion for M = 1,000
and lβ /lα = 2:
M
(b) Z(κ, M, T ) = eκlα /kB T + eκlβ /kB T ,
Problems of Chap. 2
symmetric case
1
φc =
2
2
χc = can be small, demixing possible.
M
Problems of Chap. 4
4.1. Membrane Potential
W
ΦI = Beκx , ΦII = B 1 + κx , ΦIII = V − Be−κ(x−L) ,
M
V
B= ,
2 + (w /M )κL
Q/A = W κB per area A,
Q/A W κ 1
C/A = = = .
V 2 + (W /M )κL (2/w κ) + (L/M )
4.2. Ion Activity
a Z 2 e2 κ
kB T ln γ c = kB T ln =− ,
c 8π 1 + κR
1 Z 2 e2 κ
c
ln γ+ c
= ln γ− =− ,
kB T 8π 1 + κR
1 Z 2 e2 κ κ
c
ln γ± =− + ,
kB T 16π 1 + κR+ 1 + κR−
κ → 0 for dilute solution,
1 Z 2 e2 κ
c
ln γ± →− .
kB T 8π
Problems of Chap. 5
5.1. Abnormal Titration Curve
ΔG(B, B) = 0,
ΔG(BH+ , B) = ΔG1,int ,
ΔG(B, BH+ ) = ΔG2,int ,
ΔG(BH+ , BH+ ) = ΔG1,int + ΔG2,int + W1,2 ,
346 Solutions
Problems of Chap. 6
6.1. pH Dependence of Enzyme Activity
r 1
= ,
rmax 1 + (1 + (cH+ /K))(KM /cS )
c H+ c S −
K= , cs = cS− + cHS .
cHS
6.2. Polymerization at the End of a Polymer
(KcM )i
ciM = KcM c(i−1)M = · · · = ,
K
∞
i(KcM )i 1
i = i=1
∞ = for KcM < 1.
i=1 (Kc M )i 1 − KcM
6.3. Primary Salt Effect
r = k1 cX ,
cX Z A Z B e2 κ
K= exp − ,
cA cB 4πkB T
Z A Z B e2 κ
cX = KcA cB exp .
4πkB T
Problems of Chap. 7
7.1. Smoluchowski Equation
P (t + Δt, x) = eΔt(∂/∂t) P (t, x),
w± (x ± Δx)P (t, x ± Δx) = e±Δx(∂/∂x) w± (x)P (t, x),
eΔt(∂/∂t) P (t, x) = eΔx(∂/∂x) w+ (x)P (t, x) + e−Δx(∂/∂x) w− (x)P (t, x),
∂
P (t, x) + Δt P (t, x) + · · · = (w+ (x) + w− (x))P (x, t)
∂t
∂ Δx2 ∂ 2
+Δx (w+ (x)−w− (x))P (x, t)+ (w+ (x) + w− (x))P (x, t)+ · · · ,
∂x 2 ∂x2
Solutions 347
∂ Δx ∂ Δx2 ∂ 2
P (t, x) = (w+ (x) − w− (x))P (x, t) + P (x, t) + · · · ,
∂t Δt ∂x Δt ∂x2
Δx2 kB T +
D= , K(x) = − (w (x) − w− (x)).
Δt Δx
7.2. Eigenvalue Solution to the Smoluchowski Equation
kB T −U/kB T ∂ U/kB T
− e e W
mγ ∂x
kB T −U/kB T U/kB T ∂W 1 ∂U
=− e e +e U/kB T
W
mγ ∂x kB T ∂x
kB T ∂W 1 ∂U
=− − W = S,
mγ ∂x mγ ∂x
∂ ∂ kB T −U/kB T ∂ U/kB T
LFP W = − S = e e W ,
∂x ∂x mγ ∂x
kB T ∂ −U/kB T ∂ U/kB T
LFP = e e ,
mγ ∂x ∂x
kB T ∂ −U/kB T ∂ U/2kB T
L = eU/2kB T LFP e−U/2kB T = eU/2kB T e e ,
mγ ∂x ∂x
∂ ∂ kB T U/2kB T
LH = eU/2kB T − e−U/kB T − e = L,
∂x ∂x mγ
since
kB T
= constant.
mγ
For an eigenfunction ψ of L, we have
λψ = Lψ = eU/2kB T LFP e−U/2kB T ψ.
Hence
dN
M
dPN
M
N
0= = N =− kAB M N PN + (kAB − 2km )N 2 PN
dt dt
N =1 N =1 N =1
M
M
+ kAB M N PN −1 − kAB (N − 1)N PN −1
N =2 N =2
M−1
+ 2km N (N + 1)PN +1
N =1
dN 2 dPN
M
N
0= = N2 =− kAB M N 2 PN + (kAB − 2km )N 3 PN
dt dt
N =1 N =1
M
M
+ kAB M N PN −1 −
2
kAB (N − 1)N 2 PN −1
N =2 N =2
M−1
+ 2km N 2 (N + 1)PN +1
N =1
Problems of Chap. 9
9.1. Dichotomous Model
λ1 = 0,
⎛ ⎞
0
⎜0⎟ (L1 P0 ) 1
L1 = ( 1 1 1 1), R1 = ⎜ ⎟
⎝β ⎠, = ,
(L1 R1 ) α+β
α
λ2 = −(α + β),
⎛ ⎞
0
⎜ 0 ⎟ (L2 P0 )
L2 = ( α −β α −β), R2 = ⎜ ⎟
⎝ −1 ⎠ , = 0,
(L2 R2 )
1
k− + k+ + α + β
λ3,4 = −
2
1
± (α + β)2 + (k+ − k− )2 + 2(β − α)(k− − k+ ),
2
Fast fluctuations:
α β
λ3 = − k− − k+ + O(k 2 ),
α+β α+β
⎛ ⎞
β
⎜ α ⎟
L3 ≈ (1, 1, 0, 0), R3 ≈ ⎜ ⎟
⎝ −β ⎠ ,
−α
(L3 P0 ) 1
≈ ,
(L3 R3 ) α+β
α β
λ4 = −(α + β) − k+ − k− + O(k 2 ),
α+β α+β
⎛ ⎞
1
⎜ −1 ⎟ (L4 P0 )
L4 ≈ (−α, β, 0, 0), R4 ≈ ⎜ ⎟
⎝ 1 ⎠ , (L4 R1 ) ≈ 0,
−1
⎛ ⎞ ⎛ β
⎞
0 α+β
⎜ 0 ⎟ ⎜ α+β ⎜ α ⎟
⎟ λ3 t
P(t) ≈ ⎜
⎝ α+β
β ⎠ ⎟ + ⎜ β ⎟e → P (D∗ ) = eλ3 t .
⎝ − α+β ⎠
α
α+β − α+β
α
Solutions 351
Slow fluctuations:
λ3 ≈ −k+ − α,
⎛ ⎞
k+ − k−
⎜ −α ⎟
L3 ≈ (k+ − k− , −β, 0, 0), R3 ≈ ⎜ ⎟
⎝ −(k+ − k− ) ⎠ ,
α
L3 P0 β 1
≈ ,
L3 R3 α + β k+ − k−
λ4 ≈ −k− − β,
⎛ ⎞
β
⎜ k+ − k− ⎟ L4 P0 α 1
L4 ≈ (α, k+ − k− , 0, 0), R4 ≈ ⎜⎝
⎟,
⎠ L4 R4 ≈ α + β k+ −k− ,
−β
−(k+ − k− )
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
0 1 0
⎜ 0 ⎟ β ⎜ ⎟ ⎜ ⎟
P(t) ≈ ⎜ β ⎟ ⎜ 0 ⎟ −(k+ +α)t + α ⎜ 1 ⎟ e−(k− +β)t ,
⎝ α+β ⎠ + α + β ⎝ −1 ⎠ e α+β ⎝ 0 ⎠
α
α+β
0 −1
β α −(k− +β)t
P (D∗ ) ≈ e−(k+ +α)t + e .
α+β α+β
Problems of Chap. 10
10.1. Entropy Production
0 = dH = T dS + V dp + μk dNk ,
k
T dS = − μk dNk = − μk νkj dξj = Aj dξj ,
k j k j
dS Aj
= rj .
dt j
T
Problems of Chap. 11
11.1. ATP Synthesis
At chemical equilibrium,
0=A=− νk μk
352 Solutions
Problems of Chap. 15
15.1. Transition State Theory
K‡ k‡
‡
A+B [AB] → P,
k = k ‡ c[AB]‡ = k ‡ K ‡ cA cB ,
‡
c[AB]‡ q[AB]‡ −ΔH ‡ /kB T q[AB] ‡ ‡
‡
K = = e = qx e−ΔH /kB T ,
cA cB qA qB qA qB
√
2πmkB T
qx = δx,
h
v‡ 1 kB T
k‡ = = ,
δx δx 2πm
√ ‡
1 kB T 2πmkB T q[AB]‡ −ΔH ‡ /kB T
k= δx e cA cB
δx 2πm h qA qB
‡
kB T q[AB]‡ −ΔH ‡ /kB T
= e cA cB .
h qA qB
15.2. Harmonic Transition State Theory
∞ −mω2 x2 /2k T
‡ kB T −∞ e B
δ(x − x‡ )
k = vδ(x − x ) = ∞
2πm −∞
e−mω x /2kB T
2 2
2 ‡2
kB T e−mω x /2kB T
=
2πm 2πkB T /mω 2
ω −ΔE/kB T
= e .
2π
Solutions 353
Problems of Chap. 16
ωA −λA /4kB T
kA = e ,
2π
ωB −λB /4kB T
kB = e ,
2π
KAD = e−ΔG/kB T ,
ωAD (λAD + ΔG)2
kAD = exp −
2π 4λAD kB T
ωAD λA + λD ΔG ΔG2
= exp − − −
2π 8kB T 2kB T 4λAD kB T
%
2
ωAD ΔG2
= kA kB KAD exp − .
ωA ωD 4λAD kB T
Problems of Chap. 18
18.1. Absorption Spectrum
−βH
1 e
α= dt eiωt i| e−iωi t μf eiωf t f μi
2π Q
i,f
1
= dt eiωt e−iHt/ μeiHt/ μ
2π
1
= dt eiωt μ(0)μ(t)
2π
|μeg |2
≈ dt eiωt e−iHg t/ eiHe t/ g .
2π
354 Solutions
Problems of Chap. 20
Problems of Chap. 21
Im
−Δω/2 Δω/2
Re
2
∂2C 0 1 ∂2ζ ∂ζ
C† = − ,
∂Q2 −1 0 ∂Q2 ∂Q
∂2 2
†∂ C † ∂C ∂ ∂2
dr C † Φ† ΦC = C + 2C +
∂Q2 ∂Q2 ∂Q ∂Q ∂Q2
2
∂2 0 1 ∂2ζ ∂ζ 0 1 ∂ζ ∂
= + − + 2 ,
∂Q2 −1 0 ∂Q2 ∂Q −1 0 ∂Q ∂Q
dr C † Φ† (Tel + V0 + ΔV )ΦC
† † † † E(Q) − ΔE(Q)
2 V (Q)
= C EC + C dr Φ ΔV Φ, C=C C,
V (Q) E(Q) + ΔE(Q)
2
dr C † Φ† H Φ C
2
2 ∂ 2 E(Q) − ΔE(Q) V (Q) 2 ∂ζ
=− + C† 2 C+
2m ∂Q2 V (Q) E(Q)+ ΔE(Q)
2
2m ∂Q
2 0 1 ∂2ζ ∂ζ ∂
− +2 ,
2m −1 0 ∂Q2 ∂Q ∂Q
V (Q)
cos ζ sin ζ = ,
4V (Q)2 + ΔE(Q)2
ΔE(Q)
cos ζ 2 − sin ζ 2 = ,
4V (Q)2 + ΔE(Q)2
∂ ∂ζ 2V ΔE ∂ζ
(cs)2 = 2cs(c2 − s2 ) = ,
∂Q ∂Q 4V 2 + ΔE 2 ∂Q
356 Solutions
∂ ∂ V2
(cs)2 =
∂Q ∂Q 4V 2 + ΔE 2
2V ∂V V2 ∂ΔE ∂V
= − 2ΔE + 8V ,
4V 2 + ΔE 2 ∂Q (4V 2 + ΔE 2 )2 ∂Q ∂Q
∂ζ ΔE ∂V V ∂ΔE 1 ∂V
= − ≈ ,
∂Q 4V 2 + ΔE 2 ∂Q 4V 2 + ΔE 2 ∂Q ΔE ∂Q
2 2
∂2ζ ΔE ∂∂QV2 − V ∂∂QΔE
2 ∂V
∂ΔE
∂ΔE 2ΔE ∂Q + 8V ∂Q
∂V
= − ΔE − V
∂Q2 4V 2 + ΔE 2 ∂Q ∂Q (4V 2 + Δe2 )2
1 ∂2V 2 ∂ΔE ∂V
≈ 2
− ,
ΔE ∂Q ΔE 2 ∂Q ∂Q
√
≈ − ∂ + E − 4V + ΔE
2 2 2 2
H √
2m ∂Q2 E + 4V 2 + ΔE 2
2
2 1 ∂V
+
2m ΔE 2 ∂Q
2 0 1 1 ∂2V 2 ∂ΔE ∂V 2 ∂V ∂
− − + .
2m −1 0 ΔE ∂Q2 ΔE 2 ∂Q ∂Q ΔE ∂Q ∂Q
Problems of Chap. 22
22.1. Ladder Model
n
iĊ0 = V Cj ,
j=1
iĊj = Ej Cj + V C0 ,
Cj = uj e(Ej /i)t ,
2πV 2 2πV 2
Ċ0 = − C0 = − ρ(E)C0 ,
Δω
1 1
ρ(E) = = .
Δω ΔE
Problems of Chap. 23
23.1. Hückel Model with Alternating Bonds
β e−ik/2 + βeik/2
eiχ = ±e−ik/2 ,
β 2 + β 2 + 2ββ cos k
ββ e−ik + β 2 + β 2 + ββ eik
λ = α + βeiχ + β ei(k+χ) = α ±
β 2 + β 2 + 2ββ cos k
=α± β 2 + β 2 + 2ββ cos k.
β e−ik/2 + βeik/2
−ik/2
(c) 0= e iχ+i(N +1)k
= ±e e i(N +1)k
β 2 + β 2 + 2ββ cos k
β eiN k + βei(N +1)k
= ± ,
β 2 + β 2 + 2ββ cos k
0 = β sin(N k) + β sin(N + 1)k.
(d) For a linear polyene with 2N − 1 carbon atoms, use again eigenfunctions
c2n = sin(kn) = (eikn ),
c2n−1 = sin(kn + χ) = (ei(kn+χ) ),
and chose the k-values such that
(eiN k ) = sin(N k) = 0.
358 Solutions
Problems of Chap. 25
25.1. Special Pair Dimer
c −s −Δ/2 V c s
s c V Δ/2 −s c
is diagonalized if
(c2 − s2 )V = csΔ
or with c = cos χ, s = sin χ
2V
tan(2χ) = ,
Δ
1
c2 − s2 = cos 2χ = ≥ 0,
1 + (4V 2 /Δ2 )
1
2cs = sin(2χ) = ≥ 0.
1 + (Δ2 /4V 2 )
The eigenvalues are (Fig. 32.2)
Δ 2
E± = ± (c − s ) + 2csV
2
2
⎛ ⎞
Δ 1 1
= ±⎝ 1 +V 1 ⎠ = ±1 Δ2 + 4V 2 .
2 1 + 4V 2 1+ Δ2 2
Δ2 4V 2
6.0
4.0
relative energies E / 2V
2.0
0.0
–2.0
–4.0
–6.0
0 2 4 6 8 10
disorder Δ /2V
2.0
relative intensities
1.0
0.0
0 2 4 6 8 10
disorder Δ/2V
25.2. LH2
1 −ikn
|n; α = e |k; α,
3
k
9
9
1 −i(k−k )n
Eα |n; αn; α| = Eα e |k; αk ; α|
n=1 n=1
9
k,k
9
9
1 −i(k−k )n
Eα |n; βn; β| Eα e |k; βk ; β|
n=1 n=1
9
k,k
9
9
1 −i(k−k )n
Vdim |n; αn; β| = Vdim e |k; αk ; β|
n=1 n=1
9
k,k
9
9
1 −i(k−k )n
Vβα,1 |n; αn − 1; β| = Vβα,1 e−ik e |k; αk ; β|
n=1 n=1
9
k,k
9
9
1 −i(k−k )n
Vβα,1 |n; βn + 1; α| = Vβα,1 eik e |k; βk ; α|
n=1 n=1
9
k,k
9
Vαα,1 (|n; αn + 1; α| + h.c.)
n=1
9
1 −i(k−k )n
= Vαα,1 eik e |k; αk ; α| + h.c.
n=1
9
k,k
k=0
Eα+Eβ
2
k=0
by1
U 1 |U/Δ|
c2 − s2 = −sign , cs = .
Δ 1 + (4U 2 /Δ2 ) 1 + (4U 2 /Δ2 )
The eigenvalues are (Fig. 32.4)
1 2
E± (k) = E k ± sign V Δ + 4U 2
2
Eα + Eβ
= + (Vαα1 + Vββ1 ) cos k
2
% 2
Eα − Eβ + 2(Vαα1 − Vββ1 ) cos k
± sign V + V 2 + W 2 + 2V W cos k
2
1 ikn 1 ikn
μk,+ = c e μnα + seiχ e μnβ
3 n 3 n
1 ikn n 1 ikn n
=c e S9 μ0α + seiχ e S9 Rz (2ν + φβ − φα )μ0,α
3 n 3 n
⎛ ⎞
9 n − sin 9 n
cos 2π 2π
1 ikn ⎜ ⎟
= e ⎝ sin 2π n cos 2π n ⎠
3 n 9 9
1
⎛ ⎛ ⎛⎞⎞ ⎞
cos − sin sin θ
× ⎝c + seiχ ⎝ sin cos ⎠⎠ μ0 ⎝ 0 ⎠ .
1 cos θ
√
1
The sign is chosen such that for |Δ| |U |, the solution becomes c = s = 1/ 2.
362 Solutions
Problems of Chap. 28
105. F.C. Bernstein, T.F. Koetzle, G.J.B. Williams, E.F. Meyer Jr., M.D. Brice,
J.R. Rodgers, O. Kennard, T. Shimanouchi, M. Tasumi, J. Mol. Biol. 112, 535
(1977)
106. R. Sayle, E.J. Milner-White, Trends Biochem. Sci. 20, 374 (1995)
107. Y. Zhao, M.-F. Ng, G.H. Chen, Phys. Rev. E 69, 032902 (2004)
108. A.M. van Oijen, M. Ketelaars, J. Köhler, T.J. Aartsma, J. Schmidt, Science
285, 400 (1999)
109. C. Hofmann, T.J. Aartsma, J. Köhler, Chem. Phys. Lett. 395, 373 (2004)
110. S.E. Dempster, S. Jang, R.J. Silbey, J. Chem. Phys. 114, 10015 (2001)
111. S. Jang, R.J. Silbey, J. Chem. Phys. 118, 9324 (2003)
112. K. Mukai, S. Abe, Chem. Phys. Lett. 336, 445 (2001)
113. R.G. Alden, E. Johnson, V. Nagarajan, W.W. Parson, C.J. Law, R.G. Cogdell,
J. Phys. Chem. B 101, 4667 (1997)
114. V. Novoderezhkin, R. Monshouwer, R. van Grondelle, Biophys. J. 77, 666
(1999)
115. M.K. Sener, K. Schulten, Phys. Rev. E 65, 31916 (2002)
116. A. Warshel, S. Creighton, W.W. Parson, J. Phys. Chem. 92, 2696 (1988)
117. M. Plato, C.J. Winscom, in The Photosynthetic Bacterial Reaction Center, ed.
by J. Breton, A. Vermeglio (Plenum, New York, 1988), p. 421
118. P.O.J. Scherer, S.F. Fischer, Chem. Phys. 131, 115 (1989)
119. L.Y. Zhang, R.A. Friesner, Proc. Natl. Acad. Sci. USA 95, 13603 (1998)
120. M. Gutman, Structure 12, 1123 (2004)
121. P. Mitchell, Biol. Rev. Camb. Philos. Soc. 41, 445 (1966)
122. H. Luecke, H.-T. Richter, J.K. Lanyi, Science 280, 1934 (1998)
123. R. Neutze et al., BBA 1565, 144 (2002)
124. D. Borgis, J.T. Hynes, J. Chem. Phys. 94, 3619 (1991)
125. F. Juelicher, in Transport and Structure: Their Competitive Roles in Biophysics
and Chemistry, ed. by S.C. Müller, J. Parisi, W. Zimmermann. Lecture Notes
in Physics (Springer, Berlin, 1999)
126. A. Parmeggiani, F. Juelicher, A. Ajdari, J. Prost, Phys. Rev. E 60, 2127 (1999)
127. F. Jülicher, A. Ajdari, J. Prost, Rev. Mod. Phys. 69, 1269 (1997)
128. F. Jülicher, J. Prost, Progr. Theor. Phys. Suppl. 130, 9 (1998)
129. H. Qian, J. Math. Chem. 27, 219 (2000)
130. M.E. Fisher, A.B. Kolomeisky, arXiv:cond-mat/9903308v1
131. N.D. Mermin, J. Math. Phys. 7, 1038 (1966)
132. M.W. Schmidt, K.K. Baldridge, J.A. Boatz, S.T. Elbert, M.S. Gordon, J.J.
Jensen, S. Koseki, N. Matsunaga, K.A. Nguyen, S. Su, T.L. Windus, M. Dupuis,
J.A. Montgomery J. Comput. Chem. 14, 1347 (1993)
Index