Sie sind auf Seite 1von 21

The Thermodynamics of Molecular Recognition

Corinne L. D. Gibb and Bruce C. Gibb


University of New Orleans, New Orleans, LA, USA

1 Preamble
2 Basic Concepts
3 Practicalities
4 Conclusion
References

1
1
10
20
20

PREAMBLE

If the birth of supramolecular chemistry is taken to be


Fischers lock and key hypothesis,1 then there was an
eighty-year induction period between this genesis and the
pioneering work of Cram, Lehn, and Pedersen, which ultimately led to their winning the Nobel Prize in 1987.24 As
Schalley has eloquently pointed out in his excellent book,
Analytical Methods in Supramolecular Chemistry,5 two of
the more important reasons for this induction period were
the time required for chemists to appreciate the importance
of weak intermolecular forces, and the time required for
chemists to develop techniques that could actually measure them. By the late 1960s, both these obstacles were
beginning to fall, and this was instrumental in heralding
in what can be called the first phase in the development
of supramolecular chemistry. In more ways than one, this
first phase has been a time of equilibration; it has been
a period during which the field has equilibrated and figuratively, found its feet, and a period in which the field
has literally been dominated by the measurement of twoentity systems at equilibrium. Now that it has found its

feet, the field has been busy working towards phase two:
applying what it now knows to topics such as sensing,6
while simultaneously exploring many-bodied systems that
may or may not be at equilibrium; the latter in the short term
is fundamental to the development of systems chemistry7
and in the long term may contribute to the development of
synthetic complex systems (i.e., those possessing emergent
phenomena).811 Regardless of exactly how this second
phase evolves, the ability to measure equilibria will always
be a part of supramolecular chemistry, and hence a thorough
understanding of the underlying concepts and the practical
aspects of measuring equilibria is of fundamental importance. This chapter summarizes the important concepts,
techniques, and protocols in studying systems at equilibrium. We begin with the basic concepts, and, where appropriate, relate these to the common methods of determining the thermodynamics of complexation [NMR (nuclear
magnetic resonance), UVvis spectroscopy, and isothermal titration calorimetry (ITC)]. In the second section, we
examine practical aspects of determining this data using the
aforementioned techniques.

2
2.1

BASIC CONCEPTS
Association constants and Gibbs free energy

The study of how noncovalent forces can lead to the


association of two or more molecules is the embodiment
of supramolecular chemistry. The simplest supramolecular
systems therefore involve two molecules interacting with
each other, and frequently these are defined as host and
guest. The host, the larger of the two chemical entities, possesses curvature or concavity, which results in the inward
projection of functional groups or moieties (the recognition
motif) that define the binding site. In contrast, the smaller

Supramolecular Chemistry: From Molecules to Nanomaterials, Online 2012 John Wiley & Sons, Ltd.
This article is 2012 John Wiley & Sons, Ltd.
This article was published in the Supramolecular Chemistry: From Molecules to Nanomaterials in 2012 by John Wiley & Sons, Ltd.
DOI: 10.1002/9780470661345.smc005

Concepts

guest possesses a convex array of complementary structure


which is recognized by the binding site. However, to understand the fundamentals behind the interaction of a host and
guest, we must first discuss simpler systemsspecifically
the simple case where only one species is dissolved in solutionand then the case were two species in solution are
in equilibrium. Once we have accomplished this, we will
be able to discuss the case of a host and guest being in
equilibrium with a hostguest complex. Readers who wish
to go deeper than the following discussion are directed to
two excellent texts.12, 13
For the simplest of cases, where solute A is dissolved
in solvent S, the Gibbs free energy (GFE) is the energy of
the entire system at constant pressure. The GFE of such a
solution is an intricate sum of the free energy of the solute
and solvent, as well as factors that take into account the
combining of the two, such as solvation of the former and
the overall entropy of mixing. To determine the GFE of a
solution, let us first consider solute A. The chemical potential of this solute (A ) is the extent to which the total GFE
(Gt ) of the solution changes as a function of the amount of
solute A. Mathematically, this is expressed by (1):
A =

Gt
nA

(1)

where nA is the mole fraction of solute A. Chemical potential is analogous to potential energy, and thus assuming
there is no (kinetic) barrier of impediment, a system of
higher chemical potential will change so that a lower potential can be attained. Note that (1) relates how the free energy
of a system changes upon a change in its composition, that
is, Gt = A nA . Building on this, the total GFE (Gt ) of
the solution of A is the sum of the products of the mole
fraction and chemical potential of each of the constituents.
Thus in the case of solute A dissolved in solvent (S), we can
write (2), where nA and nS are the mole fractions of solute
and solvent, and A and S are their respective chemical
potentials:
Gt = nA A + nS S

(2)

In an ideal dilute solution, that is, one that obeys Henrys


Law, each solute particle A is fully solvated, and there
is no aggregation occurring that could otherwise influence
the behavior of the solution. In such cases, the chemical
potential of solute A is given by (3):

A = A + RT ln nA

(3)

where A is the chemical potential of the standard (reference) state, R is the gas constant, T is the temperature, and
nA is, again, the mole fraction of A. However, most solutions do not behave ideally, and in these cases (3) can be

modified (4) to express the chemical potential as a function


of the activity of solute A (aA ) rather than its mole fraction:

A = A + RT ln aA

(4)

Essentially, the activity of a species is its effective mole


fraction. It is introduced in order to preserve the form of
equations derived from ideal solution in nonideal situations.
All of the deviations from ideality are contained within the
activity. For nonideal solutions, we can further isolate
this deviation from ideality by invoking the activity coefficient ( A ), which relates the activity and the mole fraction
of the solute (5):
aA = A nA

(5)

All the deviations from ideality are now captured in the


activity coefficient, and since solutes obey Henrys law as
their concentrations approach zero, aA nA and A 1
as nA 0. The activity coefficients of a solute can be estimated by DebyeHuckel theory using interionic forces to
estimate aggregation. That said, this is an exceedingly difficult if not impossible process for solutions containing multiple species. As a result, it is normally assumed that A = 1,
and the conditions under study amount to an ideal dilute
solution. This assumption also negates the practical difficulties with measuring activities of solutions, because under
ideal, dilute conditions the activity of the solute is numerically equal to its (more readily measured) concentration.
Moving up to a slightly more complicated system,
consider now a system composed of two solutes (A and
B) in equilibrium (6).
AB

(6)

Consider also what happens when a small amount (d ) of A


turns into B. In such a situation, the change in the amount
of A is dnA = d , and the change in the amount of B is
dnB = +d . The quantity is called the extent of reaction
(units, moles), and when the extent of reaction changes, the
initial amount of A (nA,0 ) decreases to nA,0  and the
amount of B increases (nB,0 ) to nB,0 +  . The reaction
Gibbs energy Grct of this system is defined as the slope
of the line in the plot of Gibbs energy against the extent of
reaction [(7) and Figure 1]:



= Grct

(7)

p,T

Note that normally when we see a G term, we think


of a difference in free energies. Here, however, Grct
represents the derivative, that is, the slope of the line at a
particular value of . There is, however, a close relationship

Supramolecular Chemistry: From Molecules to Nanomaterials, Online 2012 John Wiley & Sons, Ltd.
This article is 2012 John Wiley & Sons, Ltd.
This article was published in the Supramolecular Chemistry: From Molecules to Nanomaterials in 2012 by John Wiley & Sons, Ltd.
DOI: 10.1002/9780470661345.smc005

The thermodynamics of molecular recognition

between two energy states? Let us now define the standard


state molar free energy of reaction, Grct , the change in
free energy when reactants in their molar standard state
change to products in their standard states (11):

G r < 0, reaction is
spontaneous

Grct = GB GA = 0B 0A
Gt

G r > 0, reverse
reaction is
spontaneous

Slope = G rct

B A = 0B 0A + RT (ln aB ln aA )
Grct =

nB = 1
Extent of reaction (z)

Grct

Figure 1 Graph of total Gibbs free energy (Gt ) against extent


of reaction ( ).

between this derivative and the normal usage of G terms.


Building from (2), while appreciating that dnA = d and
dnB = +d , for a small change in the amount of A ( d )
the corresponding change in free energy (dG) is given by
dG = A dnA + B dnB = A d A + B d B
= d (B A )
which can be rearranged to


G
= B A
p,T

Grct

+ RT (ln aB ln aA )
 
aB

= Grct + RT ln
aA

(12)
(13)
(14)

Importantly, (14) allows us to relate the free energy of


reaction to the activities of the two solutes. Furthermore, the
assumption that the system under study is an ideal and dilute
solution allows us to relate the concentration of solutes to
the free energy of reaction (15):


[B]

(15)
Grct = Grct + RT ln
[A]

(8)
At any point of a reaction, the ratio of products to reactants,
in this case [B]/[A], is defined by the reaction quotient Q.
Formally, and more generally, we can define Q as
(9)

and hence
Grct = B A

(11)

where 0B and 0A are the chemical potentials of A and


B in their standard molar states. If we expand (4) to
accommodate the additional solute B, we can write (12),
and noting (11) and (12), (13) and (14) readily follow:

G rct = 0

nB = 0

(10)

Thus, Grct can be viewed as the difference in the chemical


potentials (or partial molar free energies) of A and B:
but only at the particular composition of the mixture.
For contemporary supramolecular chemistry, determining
Grct at a particular point on the way to equilibrium is
not vital. However, in general terms it is important to note
the following three different cases for Grct = B A
in which A is equilibrating into a mixture of A and B:
(i) the forward reaction (A B) is spontaneous when
A > B (Gr is negative and the reaction is exergonic);
(ii) it is spontaneous in the reverse direction (B A) when
A < B , (Gr is positive, and the forward reaction is
endergonic); and (iii) the reaction is spontaneous in neither
direction when A = B (Grct = 0, i.e., the reaction is at
equilibrium).
So what of our more usual understanding of G terms,
that is, that they represent the free energy differences

Q = J aJ J

(16)


where  denotes the product of what follows it (as
denotes the sum), aJ is the activity of component J, and
J is the stoichiometric number. The reader will recall
that stoichiometry numbers are different from stoichiometry
coefficients, and are negative for the reagents and positive
for the products of a reaction. Thus, in the case at hand,
Q = aA1 aB1 = aB /aA . Hence, (14) can be written in the
more general form (17)

Grct = Grct + RT ln Q

(17)

Furthermore, the reaction quotient at equilibrium is defined


as the equilibrium constant K (K = Qeq ), and as Grct = 0
at equilibrium, we can write

Grct = RT ln K

(18)

In this case, Grct does reflect the free energy difference


between a solution of solute A in its standard state and
a solution of solute B in its standard state. Hence, with
the assumption that the system under study is an ideal

Supramolecular Chemistry: From Molecules to Nanomaterials, Online 2012 John Wiley & Sons, Ltd.
This article is 2012 John Wiley & Sons, Ltd.
This article was published in the Supramolecular Chemistry: From Molecules to Nanomaterials in 2012 by John Wiley & Sons, Ltd.
DOI: 10.1002/9780470661345.smc005

Concepts

dilute solution, that is, we can replace solute activities with


concentrations, (18) allows us to determine the free energy
difference between A and B in their standard states by
measuring the equilibrium concentration ratios of the two
solutes.
We can now take what we have learned from these simple
systems and apply it to the case when a host and guest are
in equilibrium with their complex (19).
H + G  HG

(19)

As we have just discussed, we can use (18) to determine


the free energy difference between the reactants (H and G)
and the product (the HG complex) in this equilibrium. Here,
however, the thermodynamic equilibrium (association) constant Ka , again defined by (16), is formally the product
() of the activities (aJ ) of the host, guest, and hostguest
complex raised to the respective stoichiometric number
( J ). Thus, in the case at hand, Ka = aH 1 aG 1 aHG .
Again, assuming we are dealing with an ideal dilute solution, we can replace these activities with concentrations and
define the very familiar (20) and (21):
[HG]
[H ][G]
[H ][G]
Kd =
[HG]
Ka =

(20)
(21)

where Ka is again the association constant between host


and guest, and Kd the reciprocal of Ka is the dissociation constant for the same process. For historical reasons, association constants (units, M1 ) have been preferred
by supramolecular chemists, whereas dissociation constants
(units, M) are routinely used by biochemists and medicinal
chemists. For Ka values, the larger the number, the stronger
the binding of the guest to the host, while for Kd values the
opposite is true. One of the reasons that many researchers
prefer dissociation constants is that a Kd value is numerically equal to the concentration of the guest [G] at which
the binding site of the host is half occupied, that is, when
[H ] = [HG]. In other words, a Kd value equates to the
concentration of a drug required to begin to significantly
inhibit an enzymatic target.
A major component of determining the thermodynamic
parameters for a complexation event is the determination
of the association constant (or dissociation constant). A
number of spectroscopic techniques (NMR, UVvis, fluorescence), as well as ITC, are used to make these determinations, and as we continue to expand on the basic concepts
and subsequently discuss the practicalities involved in such
determinations, we intrude these analytical techniques in
more detail.

One fundamental difference between (6) and (19) is the


molecularity of their respective forward and backwards
reactions. In a simple A  B system, the equilibrium constant is defined by two unimolecular processes, whereas in
the H + G  HG system the forward process is bimolecular and the reverse is unimolecular. This reaction asymmetry has ramifications that are both theoretical and practical.
Regarding the former, if concentrations are used to determine the Ka for an A  B system, the Ka value has no
units and (18) can be directly applied to determine Grct
(G as we refer to it from now on) for the equilibrated
system. However, by the same approach the Ka for the system H + G  HG has the units of reciprocal molar, and so
by the application of (18) we find ourselves trying to take
the natural logarithm of units! This issue stems from the
assumption that the system is behaving as an ideal dilute
solution and that we can replace the dimensionless activities of the solutes with concentrations (which in the case
of molarity has units M). As this assumption is generally
valid in most hostguest systems (or at least assumed to
be!), the only option is to ignore the units of Ka values
when applying (18). It is, however, important to appreciate
this very convenient omission.
The practical issue arising from the molecularity difference between the A  B and H + G  HG systems
concerns the composition of the systems at equilibrium and
how this varies with concentration. In the more straightforward A  B system, the Ka value is defined as the ratio
of the two solutes, and this is independent of the actual
concentrations of the solutes. For example, if Ka = 5 and
the initial concentration of solute A is 6 mM, the concentration of solutes A and B at equilibrium would be 1 and
5 mM, respectively. If, however, the starting concentration
of A is 12 mM, then the concentration of solutes A and B at
equilibrium would be 2 and 10 mM, respectively. Again, a
5 : 1 ratio is observed between the concentration of reactants
and products. In other words, the Ka value is the gradient
of the straight line obtained in a graph of [B] against [A].
It is more complicated in the H + G  HG system. Here,
the association constant is defined by a ratio in which the
numerator is a concentration to the first power, and the
denominator is related to concentration squared. The net
mathematical effect of this is that, with Ka a constant in
(20), for a series of complexations with increasingly dilute
initial concentrations of H and G, the change in [H ] and
[G] terms must decrease more slowly than the change in the
numerator [HG] term. In other words, if [H ] = [G], then
the Ka value is the gradient of the straight line obtained
in the plot of [HG] against [H ]2 . We discuss the practical
aspects of this phenomenon in more detail later, but a specific numerical example may illuminate this important point
further. For a system with Ka = 10, at 1 M starting concentrations of H and G, the equilibrium values of [H ], [G],

Supramolecular Chemistry: From Molecules to Nanomaterials, Online 2012 John Wiley & Sons, Ltd.
This article is 2012 John Wiley & Sons, Ltd.
This article was published in the Supramolecular Chemistry: From Molecules to Nanomaterials in 2012 by John Wiley & Sons, Ltd.
DOI: 10.1002/9780470661345.smc005

The thermodynamics of molecular recognition


and [HG] are respectively 0.27, 0.27, and 0.73 M. The host
and guest are 73% complexed. However, the same system is
only 1% complexed at 1 mM (approximate values for [H ],
[G], and [HG] are 0.99, 0.99, and 0.01 mM, respectively).
These changes can also be rationalized thermodynamically
(see above), but to do so, we need to embrace two topics
we have avoided so far: enthalpy and entropy. We discuss
these next.

2.2

Enthalpy and entropy

The free energy of the interaction of a host and guest


is of course just the tip of the thermodynamic iceberg.
Under the surface are (among other things) the important
thermodynamic parameters: enthalpy and entropy. A lot of
chemical insight can be gained by determining the standard
enthalpy change (H ) and standard entropy change (S )
associated with a binding event. The link between free
energy, enthalpy, and entropy of a reaction is of course
the famous GibbsHelmholtz equation (22):

G = H TS

(22)

The standard change in enthalpy of a chemical process is


defined as the change in heat between reactants and products in their standard (1 M) state at constant pressure if no
work is done. The easiest way to view a change in enthalpy
is to associate it with changes in bond strength. From
a supramolecular perspective, we are primarily involved
with how noncovalent bonds change. However, because
of small changes in bond angle, strain, and torsion angle
arising through steric interactions between host and guest,
there are always subtle changes to the strengths of covalent
bonds within both host and guest as a complex is formed.
Although enumerating these different effects is difficult, at
a more general level we can of course classify a hostguest
complexation (or any reaction) as being endothermic (H
is positive) or exothermic (H is negative).
From a thermodynamic perspective (as opposed to a
statistical viewpoint), the entropy of a system is a measure of its disorder. The more the disorder, the lower the
overall free energy of the system. The standard change
in entropy for a chemical process is a measure of the
differences in atomic movements, that is, degrees of freedom, for reactants and products. The term degree of freedom encompasses three types of molecular motion, the
symmetry (point group) of the molecule(s), as well as
entropy changes arising through the mixing of chemical
entities. For a small molecule at 1 M concentration, these
entropic contributions can be ranked in the order of impor
)  rotational entropy
tance: translational entropy (Strans

(Srot ) > vibrational entropy (Svib ) > entropy of symmetry

(Ssym
) entropy of mixing (Smix
). As a rule of thumb,

Strans and Srot for a small solute are approximately 30 eu


(entropy units, kcal mol1 K1 ). Both will be lost if a
small molecule undergoes a reaction. In contrast, a loss of

Svib
which can be mostly attributed to changes in internal
bond rotationsis usually an order of magnitude smaller.
It should be noted that of these three, only translational
entropy is concentration dependent. The additional entropy

and Smix
are generally quite small. Thus, the
factors Ssym
entropy of symmetry is defined by Ssym = R ln , where
is the symmetry number characteristic of the point group
of the molecule. For molecules of low symmetry (e.g.,
the C1 point group) = 1, whereas for higher symmetry molecules, for example, dodecahedrane (Ih point group)

usually lies between zero and


= 60. In other words, Ssym
8.3 eu, with a distinct bias toward zero. The entropy
 of mix
ing of i components is defined by, Smix
= R i ni ln ni ,
where n is the mole fraction. Thus, for an equimolar two
component system Smix
= R(0.5 ln 0.5 + 0.5 ln 0.5) =
R ln 2 = 1.38 eu. So again, any change in these types of
entropy as a result of reaction (or complexation) is usually
small. Hence, for a hostguest complexation event, it is the
changes in translational and rotational entropy that dominate, although in some cases a loss of the other forms of
entropy for the complexed host and guest may also play a
role. As a final note on entropy, it is also worth recalling
that as a solution is made more dilute, the entropy of the
system increases. In the case of the dilution of a solution
of host, guest, and hostguest complex (19), this entropic
change will be larger if the distribution of species shifts
toward free host and guest (two species) rather than the
hostguest complex. Hence the observed decomplexation
of a hostguest complex as a solution is diluted.
How do we determine the enthalpy and entropy contributions to the overall free energy change of a binding event?
We should recall that there are two general approaches.
The most accurate one is to determine the enthalpy change
directly using a calorimetric approach. ITC measures the
amount of heat liberated by a binding event as aliquots of
the guest are added to the host (or vice versa). As the titration proceeds, the amount of free host decreases and so the
amount of heat liberated with each addition of guest also
decreases. The result of an ITC experiment is therefore the
overall enthalpy change for complexation and, equally as
important, a curve of how the amount of heat liberated or
consumed decreases as a function of the host/guest ratio.
This latter curve defines the equilibrium constant for the
process, and hence using (18) and (22) the complete thermodynamic profile (G , H , S ) at constant pressure
is obtained. Errors in H can be as low as 1%, with
attendant errors in free energy and entropy changes ideally
<5%.

Supramolecular Chemistry: From Molecules to Nanomaterials, Online 2012 John Wiley & Sons, Ltd.
This article is 2012 John Wiley & Sons, Ltd.
This article was published in the Supramolecular Chemistry: From Molecules to Nanomaterials in 2012 by John Wiley & Sons, Ltd.
DOI: 10.1002/9780470661345.smc005

Concepts

A less accurate approach is to use a technique that


directly provides Ka , such as NMR, UVvis, or fluorescence spectroscopy, and takes advantage of the fact that
the binding constant changes as a function of temperature.
The relationship between Ka and temperature is apparent if
we (replacing K with Ka ) rearrange (18) in terms of ln Ka ,
and combine this with (22) giving (23):
S
H
+
RT
R

Guest 5

H (kcal mol1)

ln Ka =

+ve

(23)

Hence, if the equilibrium constant for a binding event


is determined at several different temperatures, plotting
ln Ka versus 1/T gives a so-called vant Hoff plot, which
is a straight line whose slope is H /R and intercept
is S /R. With ITC being a relatively recent addition
to the list of tools commonly utilized in supramolecular
chemistry, spectroscopically derived vant Hoff plots are
very common. It should, however, be noted that the errors
inherent in this approach are larger than those using
ITC, primarily because the enthalpy change is determined
directly using ITC but is a first derivative when determined
using a vant Hoff plot. Additionally, in the latter there is
an inherent assumption that the enthalpy change does not
vary with temperature and that the obtained straight line
is just thata straight line. In other words, there is no
change in heat capacity associated with complexation. In
many systems studied in organic solvents, this is often a
reasonable assumption, but in water this is not normally
the case (see above). Hence, to err on the safe side, the
range of temperature used to determine the association
constant should be kept small (20 C). Correspondingly,
the number of determinations of K should be quite high
(five or more) so that the error in the extrapolation of the
straight line to the intercept is minimal.
There is another important issue that can arise with vant
Hoff plots. It is frequently the case that, when enthalpy
and entropy changes for guest complexation are carried out
for a series of guests, enthalpyentropy compensation is
observed.14a By way of example, it is frequently observed
that for a series of guests, as the enthalpy of complexation
becomes more exothermic, there is a corresponding increase
in the entropy penalty for each binding. The thermodynamic
data for a series of hypothetic guests is shown in Figure 2.
This plot fits the simple equation y = mx + C, where
y = H and x = S , and the slope of this line (m),
usually referred to as or the compensation temperature
Tc , is the temperature at which any variation in the H
for a series of guests is compensated by a change in S
such that the overall free energy change remains constant.
In Figure 2, G corresponds to the point on the line when
x = 0: that is, G is negative. There are two possibilities
as to the cause of such a trend: they reflect a real chemical

Guest 4
Guest 3

ve

+ve
Guest 2

Guest 1

ve
S (kcal mol1 T1)

Figure 2 Graph of H versus S for a series of hypothetic


guests binding to a common host.

phenomenon, or they are purely artifactuala result of


the fact that in a vant Hoff plot H and S are
not measured independently. Krug and coworkers have
defined two important tests that differentiate between these
two possibilities.14b First, if there is a chemical cause to
the observed trend, then the compensation temperature Tc
must be significantly different from the average of the
experimental temperatures used to determine Ka . Second,
a true chemical effect can also be assumed when the vant
Hoff plots for the different reactions (in this case the
reaction between host and guest) are graphed together
and intersect at a common temperature. If either or both
of these scenarios are not observed, then the observed
enthalpyentropy compensation is most likely caused by
the statistical handling of the obtained data.
Both these approachesITC and the use of vant
Hoff plots in spectroscopic determinationsare commonly
used to measuring the enthalpy and entropy changes of
complexation; therefore, a practitioner of supramolecular
chemistry needs to be aware of the strengths and limitations
of the different approaches.

2.3

Some kinetic considerations

Thermodynamics and kinetics are intimately linked, and so


although the focus here is on thermodynamics we must
also touch on the topic of kinetics. In particular, we have
to discuss (i) the exchange rate between the free and bound
states in the system in question and how this relates to
the timeframe of the analytical technique being used to
investigate it; (ii) the timeframe by which equilibrium in
a system is attained, which is closely related to (i); and

Supramolecular Chemistry: From Molecules to Nanomaterials, Online 2012 John Wiley & Sons, Ltd.
This article is 2012 John Wiley & Sons, Ltd.
This article was published in the Supramolecular Chemistry: From Molecules to Nanomaterials in 2012 by John Wiley & Sons, Ltd.
DOI: 10.1002/9780470661345.smc005

The thermodynamics of molecular recognition


(iii) the overall timeframe (lifetime) of the system under
investigation. These points illustrate three different kinds
of systems that cover the gamut of chemical equilibria
possibilities: those that are equilibrated, those that not yet
equilibrated, and nonequilibrium systems. Considering all
three phenomena gives different perspectives on the notion
and importance of timeframes. The major fundamental to
note here is that all systems are tending toward equilibrium,
and it is just a matter of how long the system takes to attain
this state in relation to the experiment we are performing.
Let us first briefly consider the second and third
examples. In 10 min or so it may take to prepare a sample and run a NMR experiment, most hostguest systems
have attained equilibrium. The rate of the forward reaction
and the rate of the reverse process are so fast relative to
our actions that what is observed when the data is collected
is an equilibrated system. This is not, however, always the
case, particularly with the increasing structural complexity
of contemporary supramolecular systems. Regardless of the
structure of the host and/or guest, it is therefore prudent to
double-check that the system has equilibrated by running
the same experiment on the original sample after another
time period to confirm that the system has not changed. If
the system is still changing as a function of time, then the
system has not yet equilibrated and it may be possible to
use an analytical technique to study how the system changes
in a temporal manner. The term nonequilibrium systems
is normally reserved for those that are not at equilibrium
because some input of potential (chemical or otherwise)
is holding the system out of, or far from, equilibrium. In
other words, the system is subject to the laws of nonequilibrium thermodynamics.15 In chemistry, these types of systems are exemplified by oscillation reactions such as the
BelousovZhabotinsky (BZ) reaction.16, 17 The BZ reaction was the first of now many known nonlinear chemical
reactions that display periodic or chaotic temporal oscillation and spatial pattern formation. Although still often
viewed as a chemical oddity, we should note in passing
that nonequilibrated systems are apparent everywhere, from
the smallest life form to the largest hurricane (and beyond).
All nonequilibrium systems come into existence, exist for
a defined period of time, and then come to an end; and
in all cases the timeframe of our analytical measurements
is generally much shorter than the timeframe (lifespan) of
the system itself. Nonequilibrium systems are antithetical
to contemporary supramolecular system, but their appreciation bolsters the notion that we need to be mindful when
determining a binding constant that the system in question
is in equilibrium.
Closely related to the timeframe in which equilibrium is
established is the exchange rate between the free and bound
states in the system in question and how this compares to
the timeframe of the experimental technique being used.

To pose a question, are the exchange processes between


the free and the bound states under examination faster or
slower than the timescale of the analytical technique? The
actual rate constants for the forwards and reverse reactions
of a simple hostguest system, k1 and k1 (24), correspond
to the bimolecular complexation (units = M1 s1 ) and the
unimolecular decomplexation (units = s1 ).
H +G

k1

k1

HG

(24)

However, when we spectroscopically examine a hostguest


system, we are observing its free and the bound states. For
example, for an NMR analysis we may be examining the
free and bound signals for a particular guest proton. Thus,
we are directly concerned with the observed rate constants
(ka and kb ) for the exchange between the free and bound
states (25).
Free

ka

kb

Bound

(25)

The rate constants for both the forward and reverse processes are those of unimolecular processes (units s1 ), and
it is the slowest of these two processes that we must compare with the rate at which the system is interrogated. For
spectroscopic determinations, the timescale of the analytical
technique is dictated by the Heisenburg uncertainty principle. The energy gap (E) between the two states connected
by the absorption is related to the time difference t by (26)

E t = 2

(26)

Hence, with large energy differences in UV spectrometry,


t is small, whereas the very small energy differences
between spin states in NMR means that t is relatively
large. To appreciate the timescale of NMR, consider the
situation in which two resonance signals from a proton in
the guestfor the bound and free statesare separated by
1 ppm. We must first recall that chemical shift () and the
change in chemical shift () are usually expressed as parts
per million so that the quoted values are independent of the
frequency of spectrometer. For our purposes here, however,
we must consider the resonance frequency (, units Hz) of
a nucleus, and recall that this is dependent on the applied
field strength (27).
=

B0
2

(27)

where is the magnetogyric ratio and B0 is the strength


of the applied magnetic field. The value of in parts
per million is given by (28), where S and TMS are

Supramolecular Chemistry: From Molecules to Nanomaterials, Online 2012 John Wiley & Sons, Ltd.
This article is 2012 John Wiley & Sons, Ltd.
This article was published in the Supramolecular Chemistry: From Molecules to Nanomaterials in 2012 by John Wiley & Sons, Ltd.
DOI: 10.1002/9780470661345.smc005

Concepts

the respective frequencies of the sample (guest) signal in


question and the reference tetramethylsilane.
=

s TMS
Operating Frequency (MHz) 106

(28)

Hence, a 1 ppm difference between two signals corresponds


to 100 Hz on a 100 MHz instrument but 800 Hz on an
800 MHz instrument. As we will see, this is important
to note because the timescale of a 100 MHz instrument
is not the same as that of an 800 MHz. Focusing for
now on the smaller spectrometer, two peaks separated by
100 Hz are separated by 100 s1 . Returning to (25), if kb
(which for hostguest complexation formation is less than
ka ) is greater than this, say 400 s1 , then the exchange
process between the free and bound states is faster than the
100 MHz NMR timescale. In such a situation, the weighed
average of the two signals is observed. If, on the other hand,
kb = 20 s1 , then exchange is slow on the NMR timescale.
In this scenario both the free and the bound states are
distinct, that is, separate signals are observed. This situation
is more informative than when exchange is fast on the NMR
timescale because it is possible to determine the chemical
shift difference between the free and bound state of each
proton. Furthermore, with a slow exchanging system, the
Ka value can be determined directly from integration of the
two signals, whereas when the process is fast on the NMR
timescale, it is necessary to perform a titration experiment
in which the guest is added to the host (or vice versa) and
the shift in the target signal monitored as a function of
host/guest ratio (see below). Fast exchanging systems not
only require more work when it comes to spectroscopic
determinations, but the attendant errors are also larger: a
fact that should not be forgotten if a vant Hoff plots is
being constructed.
It is also possible that the exchange process being investigated occurs on the NMR timescale. Between the situations
when two peaks are observed (slow exchange) and one peak
is observed (fast exchange), signal coalescence occurs. This
scenario can lead to very broad signals, or signals disappearing into the baseline, a very uninformative situation.
Thankfully, the exchange rate is temperature dependent and
so by increasing or decreasing the temperature it is possible
to speed up or slow down exchange so that NMR data can
be gathered. Indeed, this is a common approach by which
kinetic barriers to exchange can be probed. Returning to
our larger NMR spectrometer, with a peak separation of
800 s1 , both the exchange processes discussed above will
be slow relative to this spectrometer. Hence, moving up to a
larger spectrometer is one, albeit expensive, way in which a
relatively fast exchanging system can be interrogated more
successfully.
NMR spectrometry is quite distinct from other spectroscopic techniques in regard to its timescale. Fluorescence,

UVvis, and IR spectrometry all operate on much faster


timescales because they interrogate electronic absorption
(emission) and nuclear vibrational motion: processes that
are much faster than the complexation or decomplexation of a hostguest system. In other words, all binding
events are slow on the timescale of these techniques. Unfortunately, although all hostguest exchange processes are
slow, experiments with these techniques are not normally
as structurally informative as NMR.
Another way to visualize exchange processes, literally
the reciprocal of the viewpoint just given, is to consider the
lifetime of the species under consideration. The lifetime of
a species is defined as the reciprocal of the rate constant
for the first-order reaction forming it. Thus, in the case
of bound species in (25), its lifetime is defined by (1/ka ).
Hence, in the most general of terms, it can be asked
if the lifetime of a species is long enough such that it
holds together for the duration of the measurement being
carried out. This viewpoint allows us to move beyond
spectroscopy and ask, for example, if the lifetime is larger
than the high performance liquid chromatography (HPLC)
timescale (about 30 min). If it is, then HPLC will reveal
separate peaks for the free and bound species. If not, then
only one peak will be observed if host and guest were
combined in equimolar amounts. Supramolecular exchange
processes that are slow on the HPLC timescale are relatively
rare in synthetic systems,18 because the corresponding
association constant for the formation of the complex must
be very high (typically Ka > 1010 M1 ).
As we have discussed, ITC does not spectroscopically
examine the free and the bound states in a hostguest
system but instead determines the overall enthalpy change
upon addition of multiple aliquots of guest to a solution of
the host. Hence, the aforementioned discussion of experimental timeframes and their relationship to the exchange
rates do not apply to ITC. Instead, it is important to determine whether upon addition of each aliquot the mixture has
equilibrated before an addition aliquot has been added. In
other words, in ITC it is important that the time to equilibration for the system be faster than the pause between
each injection.

2.4

Medium of study: organic solvent or water

The role that solvent plays in chemistry can never be


overstated, and in supramolecular chemistry we are particularly interested in how solvent interacts with the free
and bound solutes, and consequently how this influences
the noncovalent interactions between them. The bulk of
research in the field carried out thus far has been in organic
solvents, with a bias toward nonpolar and aprotic solvents that maximize electrostatic interactions between the

Supramolecular Chemistry: From Molecules to Nanomaterials, Online 2012 John Wiley & Sons, Ltd.
This article is 2012 John Wiley & Sons, Ltd.
This article was published in the Supramolecular Chemistry: From Molecules to Nanomaterials in 2012 by John Wiley & Sons, Ltd.
DOI: 10.1002/9780470661345.smc005

The thermodynamics of molecular recognition


molecules. This has given practitioners a firm understanding of both the relationship between molecular structure
and noncovalent interactions, and how the different properties of solvents influence noncovalent forces. That stated,
in the natural world most supramolecular chemistry occurs
in water. Consequently, aqueous supramolecular chemical
research that has also focused on water as a solvent, particularly with cyclodextrins,1923 cyclophanes,24 and, most
recently, dynamic molecular capsules.25 This research has
highlighted many differences between aqueous and nonaqueous supramolecular chemistry which are important to
appreciate.
At a very general level, working in water rather than
organic solvent affects the thermodynamics of binding in
two important ways. First, all other things being equal,
binding in water is stronger than binding in organic
solvents.14a Second, whereas heat capacity changes of binding processes in organic solvents are usually small or not
observed at all, this is generally not the case in water. In
other words, in aqueous solution it is frequently observed
that the H and S for complexation changes as a function of temperature. We will expand on these points in due
course, but for now let us briefly discuss the root cause of
these two phenomena, the hydrophobic effect. Most simply stated, the hydrophobic effect2630 is the reason why
oil and water do not mix. Many details of the hydrophobic effect are still to be identified and enumerated, but the
key noncovalent interaction behind it is hydrogen bonding.
More specifically, the strength of the waterwater interaction gives it a high cohesive energy and a high surface
tension, which leads to a sizable energetic penalty when
forming a cavity in water. Hence, the dissolution of a
solute, particularly a hydrophobic one, involves significant
changes to the local dynamical structure of the water. Key
to the hydrophobic effect is therefore the dynamical structure of the hydration shell around solutes, how this changes
according to the size, shape, and nature of the solute, and
how these changes alter the enthalpy and entropy of the
solute/solvation shell. Broadly speaking, these hydration
shells vary from solute to solute, but at a fundamental level
the hydration of solutes is still poorly understood. Some
of the best insight has come about from in silico studies
examining how the solvation shell changes as a function
of the size27 and shape31, 32 of the hydrophobe. In addition,
empirical studies have also been of immense importance.
For example, the state of the art in studying solvation shells
is quite advanced for protons and small hydrocarbons such
as methane,33, 34 and is improving both for inorganic salts35
and organic molecules such as -cyclodextrin.36
Whatever the rules that govern the hydration of solutes, a
key point is that when considering a binding event in water
we should be mindful that in water the hostguest equilibrium depicted in (19) is a considerable oversimplification.

The host, guest, and hostguest complex each has a particular solvation shell, and the change in solvation in forming
the complex plays a large part in the overall thermodynamics of binding. A much more accurate equation would therefore account for the hydrating water molecules. Regardless
of our poor understanding of these hydration shells, a comparison of many hostguest complexes reveals that the
complexation of most organic molecules receives a thermodynamic boost from the hydrophobic effect. That said,
because of the shielding properties of highly polar water,
noncovalent interactions that primarily involve electrostatic
forces cannot be relied upon to the same extent as they
can be in organic solvents. Hence, although the strongest of
noncovalent interactionsmetal coordinationhas proven
to be effective drivers of complexation (and assembly) in
water, hydrogen bonding has proven so far to be of limited
utility. However, more often than not, the thermodynamic
boost from the hydrophobic effect more than compensates
for any loss of attractive electrostatic interactions between
molecules and, as a result, binding constants in water tend
to be at least 12 orders of magnitude larger. That this is
true is perhaps not so interesting in its own right. After
all, there are many situations where strong binding is not
a requirement, and many systems where strong binding
is a detriment. However, that binding is usually stronger
means that by and large any particular host is able to
bind a wider range of potential guests; and if selectivity is required, it is always easier to prevent binding than
create it.
In addition to stronger binding, the desolvation of surfaces on the host and guest as they form a complex also
leads to a characteristic decrease in the heat capacity of
the solution. The standard heat capacity of a substance at
constant pressure (Cp ), is the amount of energy a substance
absorbs per unit change in temperature (29):

Cp =

H
T

(29)

The characteristic decrease in heat capacity for a binding


event in water demonstrates that the free host and the guest
are able to absorb more energy per unit temperature than
the corresponding hostguest complex. Much is still to be
learned why this is so, but the current understanding is that
the ordered (and this word is used in the loosest possible
terms) solvation shells around the hydrophobes can act as
heat sinks because, as the temperature is raised, many less
ordered states become available in which energy can be
stored. However, desolvation of the hydrophobic surfaces
on the host and guest reduces the total number of salvationshell waters and attenuates this sink.29, 37 As a result,
the heat capacity of the solution decreases. Indeed, this
decrease in the heat capacity is one of the best hallmarks

Supramolecular Chemistry: From Molecules to Nanomaterials, Online 2012 John Wiley & Sons, Ltd.
This article is 2012 John Wiley & Sons, Ltd.
This article was published in the Supramolecular Chemistry: From Molecules to Nanomaterials in 2012 by John Wiley & Sons, Ltd.
DOI: 10.1002/9780470661345.smc005

10

Concepts

of the hydrophobic effect, and is a much more reliable


hallmark than the often observed increase in entropy upon
binding usually attributed to the release of ordered water
molecules from the solvation shell.
A change in heat capacity for a binding event indicates
that its associated enthalpy as well as entropy change
changes as a function of temperature, and this can lead
to considerable inaccuracy in vant Hoff plots if the
thermodynamic parameters for complexation are being
sought. The relationship between the standard enthalpy
change (H ) and the standard heat capacity change (Cp )
of a reaction or a complexation is given by (30), where H
is the reference enthalpy at 298 K:

H = H + Cp

(30)

A similar equation (31) can be derived for the relationship between the standard entropy change (S ) and the
standard heat capacity change:

S = S + Cp ln T

(31)

where S is the reference entropy at 298 K. Combining


(30) and (31) with (23) leads to (32):

R ln Ka = H (1/T ) + Cp ln T + (So Cp )


(32)
This equation allows us to carry out vant Hoff plots even
when H and S change as a function of temperature.
By fitting the equation using the three variables H , S ,
and Cp and then using (30) and (31) we obtain H and
S at the temperatures sought.
It is useful to recall that Cp is a second derivative when
determined with NMR and other spectroscopic techniques;
Ka values must be determined at different temperatures and
vant Hoff plot performed in order to get the enthalpy
change, and it is how this enthalpy change changes as
a function of temperature that gives Cp . On the other
hand, Cp is a first derivative of an ITC experiment.
The enthalpy change is determined directly, and a series
of experiments at different temperatures yields Cp as
the gradient of the graph of H versus T . Hence,
ITC is a much more accurate technique for determining
Cp .
We have now introduced the necessary basics for determining the association constant (Ka ), standard free energy
change (G ), standard enthalpy change (H ), standard
entropy change (S ), and standard heat capacity change
(Cp ) for the complexation of a host and a guest. In
subsequent sections, we detail practical aspects of these
measurements.

PRACTICALITIES

Beyond the basic thermodynamics that we have just discussed, there are many considerations regarding how we
actually perform experiments to determine Ka , G , H ,
S , and Cp of complexation in 1 : 1 hostguest systems
(19). In determining thermodynamic data for a complexation event, there are, as there are with any physical chemistry problem, two goals. The first goal is to define the
system mathematically with a basic mathematical model;
the second is to fit the obtained data to the mathematical
model. The first goal is, of course, independent of the analytical technique we are going to use, whereas the second
is very much dependent on it. Irrespective of the technique
used, the overall aim is to quantify the formation of the
hostguest complex for a given initial concentration of host
and guest. This can be accomplished by directly measuring
the amount of hostguest complex, or indirectly by measuring the remaining free host or guest and using mass balance
equations to calculate the concentration of the complex.
Many analytical techniques are available to the experimentalist for this task, but we focus here on the most widely
used techniques, NMR, UV, and fluorescence spectroscopy,
and ITC, and give a succinct account on how to conduct
such experiments with these techniques. For more detailed
descriptions of the individual techniques, as well as details
of other techniques used, the reader is directed to some of
the many excellent reviews available in the literature.5, 3841
Each technique has its advantages and limitations, and our
intent here is to provide enough information to allow the
experimentalist to choose the most suitable technique for
his/her particular research.
This section begins with highlights of how the timeframe
of a technique and the concentration of a sample have
important practical ramifications. Subsequently, we discuss
the base mathematical model for 1 : 1 complexations before
looking at how this model is tailored to each analytical
technique. Finally, we discuss a common approach sometimes used to confirm 1 : 1 binding, as well as very briefly
highlight higher stoichiometry systems.

3.1

Timeframe of analysis

We have discussed the importance of timeframe of analysis with regard to both the timeframe by which equilibrium
in a system is attained and the timeframe of the exchange
process of the complexation under investigation. Practical
aspects of the former simply involve double-checking that
equilibrium has in fact been attained. Practical aspects of
the latter are a little bit more complex. As we have discussed, different techniques operate at different timeframes,
and it is important to appreciate how this relates to the

Supramolecular Chemistry: From Molecules to Nanomaterials, Online 2012 John Wiley & Sons, Ltd.
This article is 2012 John Wiley & Sons, Ltd.
This article was published in the Supramolecular Chemistry: From Molecules to Nanomaterials in 2012 by John Wiley & Sons, Ltd.
DOI: 10.1002/9780470661345.smc005

The thermodynamics of molecular recognition


exchange rates between the free and bound states (25). Is
the instrumental technique detecting an average signal of
the species involved, or are the individual signals from
each species apparent? In most spectroscopic techniques
including UVvis and fluorescence, the exchange between
free and bound species is much slower than the timeframe
of the technique. In other words, the binding event under
study is slow on the instrument timescale and the instrument
detects each of the species individually. In contrast, this is
not the case with NMR spectrometry where complexations
that are slow as well as fast on the NMR timescale are both
observed. We will return to this point when discussing the
individual techniques.

3.2

Concentration range

Another important practical aspect is the concentration


under which the complexation is being studied. Each technique has its own limitations defined by its sensitivity, and
we discuss these as we examine the individual approaches.
In addition, there is the common factor arising from the
asymmetry inherent to the hostguest equilibrium (19) and
how, for a fixed binding constant, the initial concentration
of the host and guest dictates the extent of complexation.
Figure 3 shows the state of affairs with a total (initial)
concentration of host (Ht ) and guest (Gt ) of 1 mM and
a range of binding constants (Ka ) from 10 to 108 M1 . Ideally, for accurate measurements, equilibrium should result
in somewhere between 20 and 80% complexation. If the

11

concentration chosen is too high, then the concentrations


of host and guest will be too low and sizable errors will
ensue. Likewise, if the initial concentrations are too low,
then there will be an insufficient amount of the complex
formed and again large errors will arise. In the case shown
in Figure 3, the optimum range of Ka is observed to be
between 5 102 and 2 104 M1 . Of course, the distribution of species can be shifted easily by changing the ratio
of Ht /Gt ; for example, if Ka = 10, and Ht = 1 mM and
Gt = 100 mM, then 50% complexation is attained (compare
with first column in Figure 3).
The best way to determine the concentration at which an
experiment needs to be run is to know the binding constant;
but this is of course the first step of a circular argument.
The only option therefore is to take an educated guess at
the strength of association and then to perform the required
experiment. A second experiment can subsequently be run
if binding was weaker or stronger than anticipated.
An alternative viewpoint is expressed in Figure 4, which
shows percentage complexation against total host and guest
concentration (Ht and Gt ) for a hostguest complexation
(Ka = 1 103 ). This graph represents the effect of dilution upon complex formation. At high guest concentration
(1 M), the system is close to full saturation with 97% complexation. In contrast, if the working concentration is too
low (Ht = Gt < 1 104 M) essentially no complexation
is observed. We can define a concentration range for this
model system of between 0.5 to 50 mM, and whether a particular technique is suitable for the task at hand depends on
its inherent sensitivity. Furthermore, there may be solubility

1E 03

Concentration (M)

1E 03

8E 04

6E 04

4E 04

2E 04

0E +00

1E + 01

1E + 02

1E +03

1E +04

1E +05

1E + 06

1E + 07

1E + 08

K a (M1)

Figure 3 Graph of the concentration of host [H ] (blue), guest [G] (red), and hostguest complex [HG] (green) against equilibrium
constant (Ka ) where the total (initial) concentration of host and guest (Ht = Gt ) = 1 mM.
Supramolecular Chemistry: From Molecules to Nanomaterials, Online 2012 John Wiley & Sons, Ltd.
This article is 2012 John Wiley & Sons, Ltd.
This article was published in the Supramolecular Chemistry: From Molecules to Nanomaterials in 2012 by John Wiley & Sons, Ltd.
DOI: 10.1002/9780470661345.smc005

12

Concepts
100

% complexation

80

60

40

20

0
1E + 00

1E 01

1E 02

1E04
G t (M)

1E 03

1E 05

1E 06

1E 07

Figure 4 Graph of the percentage complexation against total guest concentration (Gt = Ht ) where the equilibrium constant (Ka ) is
set to 1 103 M1 .

limits of the host, guest or complex that may also need to


be considered.

3.3

Mathematical model for 1 : 1 complexation

With the exception of just one scenario, all of the techniques commonly utilized by supramolecular chemists to
determine thermodynamic data yield too many unknowns
if just a single host/guest ratio is studied. The exception,
slow exchanging systems studied by NMR, allows the direct
measurement of the individual components of the mixture (see below) from just one Ht : Gt ratio. For the other
approaches, we must carry out a titration in which the
Ht : Gt ratio is varied systematically and construct a mathematical model based on the mass balance equations for
the equilibrium and (20). The corresponding mass balance
equations are (33) and (34):
Ht = [H ] + [HG]

(33)

Gt = [G] + [HG]

(34)

where Ht and Gt are again the total amount of host and


guest in solution. For each of the analytical techniques
we are going to discuss, we define equations that relate a
measurable to three unknowns: the equilibrium constant;
a constant specific to the complex and the technique
used ( max , , and H for NMR, UVvis, and ITC
respectively); and the concentration of free guest [G]. We
therefore need an expression of [G] that relates it to a series
of known quantities and the equilibrium constant for the
process. To do this we will combine (20), (33), and (34) to
give (35):
Ka [G]2 + (1 Ka Gt + Ka Ht )[G] Gt = 0

(35)

The solution of this quadratic, that is, (36), is our base


equation for 1 : 1 complex formation, which we will combine with equations tailored for each analytical technique
relating a measurable to Ka , max //H , and the concentration of free guest [G]. It is useful to note that the
right-hand side of (36) is composed of only Ht and Gt
(which can be calculated) and the unknown Ka .
[G] =

(1 Ka Gt + Ka Ht )


(1 Ka Gt + Ka Ht )2 + 4Ka Gt
2Ka

(36)

3.4

NMR spectroscopy

Most binding constant determinations in supramolecular


chemistry have been performed using NMR, and, in particular, 1 H NMR. Protons are almost ubiquitous to organic
chemistry, and the 1 H nucleus is of high abundance. This
means a particularly fast analysis time relative to other
popular nuclei such as 13 C. In general, NMR will permit
the determination of binding constants of between 0.1 and
104 M1 , although stronger binding constants can be determined if a longer acquisition time is possible or a competition experiment is performed (see below). In these cases,
binding constants of up to 1 106 M1 represent the absolute upper limit of the technique before analyses are beset
by large errors. The normal limit of Ka 104 M1 in NMR
spectroscopy arises from the techniques relative insensitivity. As a result, most proton NMR spectra will be recorded
at sample concentrations of 15 mM. If the binding constant is high, then it will be necessary to run the sample
at concentrations of 100 M or less, and in such cases the
signal-to-noise ratio is small and an extended acquisition

Supramolecular Chemistry: From Molecules to Nanomaterials, Online 2012 John Wiley & Sons, Ltd.
This article is 2012 John Wiley & Sons, Ltd.
This article was published in the Supramolecular Chemistry: From Molecules to Nanomaterials in 2012 by John Wiley & Sons, Ltd.
DOI: 10.1002/9780470661345.smc005

The thermodynamics of molecular recognition


time is required. This may be possible in slow exchanging
systems (see below), but in fast exchanging systems this
approach becomes unwieldy.
A critical point in NMR determinations of thermodynamic data is to have a proton whose chemical environment changes sufficiently upon binding to lead to a change
in chemical shift, that is, Hfree = Hbound . As has been
discussed above, binding determinations by NMR can be
divided into two cases: those that are slow on the NMR
timeframe and those that are fast. The observed complexation and decomplexation rates (25) are those of unimolecular processes (units s1 ), and it is the slower of
these two processes that we must compare with the rate of
NMR data acquisition. Our previous discussion emphasized
how the timeframe of the NMR experiment was dependent on the external field strength of the instrument, but
it is also important to note that whether a complexation is
observed to be fast or slow on the NMR timescale depends
on the chemical shift difference between the free and the
bound state. The key equation is (37), which defines the
boundary between fast and slow timescales, that is, the
observed rate constant for coalescence (kcoal ) of the signals for the proton in question in the free and the bound
state:
kcoal = 2.22

(37)

Thus, the larger the  the higher the observed rate


constant [kb in (25)] needs to be to switch from slow to
fast on the NMR timescale. In certain cases, where there
is a spread in  values ranging from the very small to
very large (e.g., 3 ppm or more), it is entirely possible for
some pairs of signals to be fast on the NMR timeframe
while others appear slow. Furthermore, in such cases,
intermediately shifted signals may be close to coalescence,
resulting in a broad or unobserved signal. In most cases,
however, all the signals are usually either fast or slow on
the NMR timeframe.
If an equilibrium process is slow on the NMR timescale,
the spectrum will show distinct peaks for the free and the
bound state. Hence, by knowing the initial concentration
of Ht and Gt , it is then straightforward to determine the
concentration of HG, H , and G by integration. In such
cases, the Ka values (and using (18) the G value) are
quickly determined, and if enthalpy and entropy change
or, indeed, heat capacity changes are also sought, it is
simply a matter of recording NMR spectra of the same
sample at different temperatures. That said, for a number
of reasons it is advisable to perform the determination of the
Ka and G at two or three different ratios of Ht and Gt .
Changing this ratio will help confirm which signals belong
to the formed complex (increase in intensity with increased
titrant), as well as confirm slow kinetics (the peaks will

13

change in intensity but not shift). Doing so also avoids any


major error since the Ka and G values obtained should
be identical. Determinations obtained this way and repeated
3 times with new stock solutions will give determinations
with errors of less than 10%.
During the course of the development of the field, hosts
(as well as guests) have tended to become structurally
more elaborate, in which case they often exchange slowly
on the NMR timescale. Nevertheless, structurally more
open hosts that exchange guests rapidly on the NMR
timescale still account for the majority of host molecules.
The determination of binding constants in these systems
involves more effort, both in terms of data collection and
fitting. Regarding the latter, we will need (36) to build our
mathematical model for Ka determinations with NMR.
We begin by first noting that, in fast exchanging systems,
the observed frequency Hobs of the proton of interest
becomes the weighed average of the free and bound
states (38):
Hobs = xfree Hfree + xbound Hbound

(38)

where xfree and xbound are the mole fractions of the


free component and the complex, respectively. In our
determination, we are going to change the ratio of host
and guest by titrating in the guest (or host) and monitor
how Hobs changes. In other words, we are going to plot a
binding isotherm. Typically, a host solution is titrated with
a stock guest solution in the NMR tube, and the shift of
the proton most affected by binding and not obscured by
other signals is then recorded for a minimum of 10 different
ratios. It is critical to cover a large range of host/guest ratios
to ensure that the system is close to saturation at the end of
the titration. In other words, the resulting binding isotherm
should reach a plateau. Now that we have our data, if we
set  obs = Hobs Hfree and  max = Hbound Hfree ,
then from the mass balance equations (33) and (34) and the
equilibrium equation (20), the NMR binding isotherm can
be expressed as
 obs =

 max Ka [G]
1 + Ka [G]

(39)

which can be rearranged to (40):


 obs =

 max
1
Ka [G]

+1

(40)

where  obs is the shift in parts per million of the observed


proton, and  max is the maximum shift of the observed
proton at full complexation. Note that we have too many
unknowns in this equation (Ka , [G], and  max ) to solve
for Ka with a single Ht : Gt ratio. To solve this problem,

Supramolecular Chemistry: From Molecules to Nanomaterials, Online 2012 John Wiley & Sons, Ltd.
This article is 2012 John Wiley & Sons, Ltd.
This article was published in the Supramolecular Chemistry: From Molecules to Nanomaterials in 2012 by John Wiley & Sons, Ltd.
DOI: 10.1002/9780470661345.smc005

14

Concepts

we must combine (40) with our quadratic expression for


[G] (35) to get a theoretical expression of  obs versus
Gt (41):

 obs = 

resulting binding isotherm is too sharp and approximates


to a step rather than a curve. In such cases, too few
points define the binding event and a large error ensues.
 max


2

+1
Ka Gt Ka Ht 1 + (1 Ka Gt + Ka Ht )2 + 4Ka Gt

A simple iteration process of Ka and  max using the


spreadsheet software such as Origin or the solver in
Excel will fit the experimentally derived binding isotherm
for  obs versus Gt and will yield the remaining two
unknowns Ka and  max .
As mentioned, the titration must cover the full range of
complexation and ultimately bring about the saturation of
the host. In addition, to avoid large errors, the obtained
binding isotherm must be neither too shallow nor too sharp.
Figure 5 shows three theoretical binding isotherms, where
a guest is titrated to a host solution of 1 mM, inducing
a  max of 0.2 ppm. The binding isotherm obtained for
the Ka = 1000 M1 system is ideal. It covers the full
complexation range, and will allow a good fit of the data.
With the weaker binding guest (Ka = 50 M1 ) the binding
isotherm is more akin to a straight line and insufficient
guest has been added to saturate the host. In such cases,
there are two options: (i) titrate the host further with more
guest to reach a plateau close to the  max = 0.2 ppm,
in which case unless a highly concentrated guest solution
can be made the dilution of the host must be taken into
account39 ; (ii) work at higher concentrations of the host,
which may or may not be possible depending on its
solubility. On the other hand, when the binding is too
strong for the working concentration (Ka = 104 M1 ), each
addition of guest results in it being fully complexed. The

(41)

The easiest way to obtain an ideal curve is to lower the


working concentration, which is often not possible with
NMR because of its relatively low sensitivity. Alternatively,
a less direct approach with larger attendant errors is to
perform a competition experiment in which a complex of
the host and a relatively weakly bound guest is titrated with
the stronger binding guest (see below).
Once a Ka value has been determined, this leads directly
(18) to the corresponding G of complexation, but to
ascertain the H and S values of a fast exchanging
system it is necessary to determine a vant Hoff plot by
repeating the titration procedure at different temperatures.
Note that, because measured Ka values in fast exchanging
systems have larger intrinsic errors than analogous determinations in slowly exchanging systems, errors in ascertaining
the H and S values via a vant Hoff plot can be quite
substantial.
In summary, the relative insensitivity of NMR spectroscopy means that the solution under study must be
relatively concentrated, a factor that attenuates the range
of binding constants that can be determined. In NMR
examinations of supramolecular events, the exchange process can be faster than, slower than, or on the NMR
timescale. Slow exchanging systems provide more structural information and allow a more straightforward and
accurate determination of Ka , G , H , and S than

0.20

d (ppm)

0.15

0.10

0.05

0.00
0

0.002

0.004

0.006

0.008
G t (M)

0.01

0.012

0.014

0.016

Figure 5 Three theoretical binding isotherms for 1 mM host titrated with guests of association constants, Ka = 50 ( ), 1000 ( ), and
104 M1 ( ).
Supramolecular Chemistry: From Molecules to Nanomaterials, Online 2012 John Wiley & Sons, Ltd.
This article is 2012 John Wiley & Sons, Ltd.
This article was published in the Supramolecular Chemistry: From Molecules to Nanomaterials in 2012 by John Wiley & Sons, Ltd.
DOI: 10.1002/9780470661345.smc005

The thermodynamics of molecular recognition


fast exchanging systems. Finally, it should be noted the sensitivity weakness of NMR is countered by the large amount
of structural information provided by the technique, particularly when the exchange process between free host and
guest and the hostguest complex is slow on the NMR
timescale.

Furthermore, for common 1 cm cells we can write (43),


which expresses the total observed absorbance as a function
of the respective molar absorptivities of the host, guest, and
complex at a given wavelength (H , G , and HG ) and their
respective concentrations:
Aobs = H [H ] + G [G] + HG [HG]

3.5

UVvis and fluorescence spectrometry

(43)

It is helpful to now define the change in absorbance 


upon complex formation:

After NMR, UVvis and fluorescence spectrometry are


probably the most widely used techniques for thermodynamic data collection. Both require UVvis absorbance
(and in the case of fluorescence, emission) of the host
and/or the guest. It is also important that upon complex
formation the absorbance or emission change. Both analytical techniques can be approached in a very similar fashion:
because of its wider popularity, we emphasize here UVvis
spectroscopy.
In UV spectrometry, the excitation of electrons by
absorption is accompanied by changes in vibrational and
rotational quantum numbers. As a consequence, a broad
signal of vibrational and rotational fine structure is produced rather than a single absorption line corresponding
to a particular electronic transition. Furthermore, interactions between the solute and the solvent further complicate
the data and smooth this broad absorption band. As a result,
the amount of structural information provided by these techniques is less than that provided by NMR. One consequence
of broad absorption bands is that there is inevitably extensive band overlap from the different species in solution: a
fact that can sometime complicate data interpretation. Even
if this is not the case, data gathering must rely on titration
studies analogous to those described for the NMR analysis of fast exchanging systems; there is no direct analogy
to systems exchanging slowly in the NMR timescale even
though exchange is slow on the UVvis timescale. The reason for this is that, with UVvis determinations, it is signal
intensity that we are measuring as a function of hostguest
ratio and this amplitude is dependent on the unknown
of the solution. Hence, there are too many unknowns to
determine Ka values from a single mixture of the host and
the guest.
For a hostguest complexation event, there are potentially three species that can absorb and, because the
absorbance of each is additive, we can write (42):

 = HG H G

(42)

where Aobs is the total observed absorbance, and AH , AG ,


and AHG are the respective absorbances of H , G, and HG.

(44)

Equation (44) actually pertains to a relatively rare system, because in most cases we select the absorption to
monitor so that one of the species does not absorb. In
rare cases where all species do absorb at the wavelength
examined, it is necessary to determine G and H in a
separate experiment to reduce the number of unknowns.
Hiroses excellent practical guide on binding constant
determinations describes the methods to treat the collected data (Aobs vs Gt ) of this more complex regression
analyses.39
For the majority of cases where only one species
absorbs at the observed wavelength (H or G = 0), we
can write simplified equations (45 and 46). In the case of
these equations, we assume it is the guest that does not
absorb:
Aobs = AH + AHG

(45)

 = HG H

(46)

Combining the mass balance equations (33) and (34) with


(45), we can express (47):
Aobs = H Ht + [HG]

(47)

which when combined with (20) gives (48):


A = Ka [H ][G]

(48)

And using the mass balance equations again, we can then


obtain the binding isotherm (49):
A =

Aobs = AH + AG + AHG

15

Ka Ht [G]
1 + K[G]

(49)

We now have an expression similar to that obtained by


NMR (39) and we can insert (49) into our expression of
[G] for 1 : 1 complexations (36) to give (50):

Supramolecular Chemistry: From Molecules to Nanomaterials, Online 2012 John Wiley & Sons, Ltd.
This article is 2012 John Wiley & Sons, Ltd.
This article was published in the Supramolecular Chemistry: From Molecules to Nanomaterials in 2012 by John Wiley & Sons, Ltd.
DOI: 10.1002/9780470661345.smc005

16

Concepts

Aobs = 

Ht
 2
+1
Ka Gt Ka Ht 1 + (1 Ka Gt + Ka Ht )2 + 4Ka Gt

Equation (50) is the mathematical expression that we


must iterate with the experimentally determined data. We
can see that we are dealing with an equation similar to that
obtained for NMR in fast exchanging systems (41). The
experimental data will in this case be A versus Gt (rather
than  obs vs Gt ), but the fitting of the binding isotherm
is exactly the same except that now Ka and  are iterated
to fit the curve.
Remember that we are dealing with changes in absorption
intensity rather than a shift in absorption (as we were with
NMR). As a result, we cannot change the host/guest ratio
simply by adding one to the other because this will change
the concentration of the host and hence the absorption
intensity. Rather, the experimentalist must prepare separate
solutions with increasing amounts of guest (or host) at a
constant concentration of host.
Although these determinations require more effort,
because we are typically working at lower concentrations
(about 0.01 mM), UVvis spectrometry allows the determination of thermodynamic data that cannot be measured
by NMR because the binding constant is too high. The
typical range of Ka that UVvis experiments cover is
11 106 M1 , but higher binding constants can be determined if the extinction coefficient of the molecules is
high. As an example, if = 3 105 , this would give an
absorbance of 0.3 at 1 M, allowing Ka values as high as
1 107 M1 to be ascertained. On the other hand, the determination of small binding constants by UVvis is facilitated by weakly absorbing species. Of course, the general
point that studies with UVvis require working at lower
concentrations also means that each set of experiments
require less host and guest: a fact that reduces pressure
on synthesis to form the molecules in question.
The study of binding by fluorescence spectrometry is
performed with the same experimental and data analyses as
UVvis spectrometry. Thus experiments usually involve
either a fluorescent host or a fluorescent guest, and the
change in fluorescence is monitored at different host/guest
ratios. The UVvis equations described above can easily be
adapted to fluorescence to provide the association constant
for a binding event.5
In summary, UVvis and fluorescence spectroscopies
do not provide as much structural detail of a formed
hostguest complex. They do, however, allow a wider
range of binding constants to be determined. In addition,
although each determination requires more practical work,
the amount of material required for a determination isall

(50)

other things being equalless that that required for NMR


experiments.

3.6

Isothermal titration calorimetry (ITC)

The development of sensitive isothermal titration calorimeters first impacted the biological sciences. However, over
the last decade or so the technique has become increasingly utilized within supramolecular chemistry. One of the
major reasons behind its proven popularity is the fact that
Ka , G , H , and S are obtained in a single automated experiment. In a typical ITC titration, small aliquots
of a concentrated solution of the guest are added to a solution of the host in the ITC cell. Upon each addition, a
measured amount of heat is given off, and this decreases
during the titration and reaches zero upon saturation of
the host. The total heat liberated in this titration yields
the enthalpy change, while the shape of the curve for
heat release as a function of host/guest ratio provides the
equilibrium constant and hence the free energy of binding. As a result, the entropy change for complexation can
be directly calculated. This more direct approach to garnering thermodynamic data has much smaller associated
errors, particularly for H , and avoids the issues associated with vant Hoff plots such as the possibility that H
varies with temperature, or whether the obtained data has
chemical roots or is simply artifactual. ITC also allows
a more direct and accurate determination of heat capacity
changes. For straightforward cases of 1 : 1 complex formation, there is a linear relationship between H and T ,
the gradient of which is the Cp associated with binding. Thus, a series of five or more experiments run at
different temperatures accurately yields any heat capacity
change.
An ITC titration yields an amount of heat released or
absorbed (Q) for each aliquot of added guest solution.
The sum of these heats (Q) can be defined by (51):

Q = [HG]V0 H = (Gt [G])V0 H

(51)

where V0 is the volume of the reaction (ITC cell), and H


is the molar heat of the ligand binding (in cal mol1 ).
This expression is similar to that obtained by NMR (39)
and we can insert it into our expression of [G] for 1 : 1
complexations (36) to give (52):

Supramolecular Chemistry: From Molecules to Nanomaterials, Online 2012 John Wiley & Sons, Ltd.
This article is 2012 John Wiley & Sons, Ltd.
This article was published in the Supramolecular Chemistry: From Molecules to Nanomaterials in 2012 by John Wiley & Sons, Ltd.
DOI: 10.1002/9780470661345.smc005

The thermodynamics of molecular recognition

Q = Gt V0 H V0 H

(1 Ka Gt + Ka Ht )

Modern instruments generate a plot of Q (cal mol1 )


versus molar ratio of host and guest using a spreadsheet
software that takes into account the change in volume in
the cell as the titrant is added. This plot is equivalent to
the binding isotherm of NMR or UV. Thus, the theoretical expression of Q (52) can be fitted to a number of
preset models of differing stoichiometry to generate Ka
(and hence G ) and H and calculate S .40 In addition to the goodness of the fit of the obtained curve, ITC
also generates the stoichiometry number (N), which, in a
1 : 1 complex, should be close to 1 (0.95 < N < 1.05). Any
major deviation from this, for example, N = 0.5, suggests
another binding stoichiometry, in which case the fitting
model should be reconsidered. If the stoichiometry of complexation can be confirmed by another method, then the
ITC instrument allows the N value to be numerically fixed,
which can reduce the error of the obtained thermodynamic
data, especially in weakly binding systems. If the stoichiometry is in doubt, it is often a good idea to confirm it by NMR
or another spectroscopic technique.
Ideally an S-shaped titration curve should be obtained.
If binding is too weak, then the obtained curve will tend
toward linear, whereas if it is too strong, a step will
instead be observed. The steepness in the titration curve
can be described by the Wiseman parameter c, defined as
c = Ka [X], where [X] is the concentration of titrate.40
It is often quoted that, in order to obtain an isotherm
corresponding to a smooth S-shaped curve and minimal
errors in both H and Ka (23%), the Wiseman
parameter should lie in the range of 100500. However,
most recently it has been shown that it is possible to obtain
accurate determinations of Ka with c values as low as
5.42
ITC is capable of determining a wide range of binding constants, typically between Ka = 5 and 109 M1 .
Of course, the spectroscopic character of the molecules
involved is not important in such experiments. Instead, the
ability to measure Ka is dictated by the extent that the
complexation liberates or absorbs heat. For the lower limit,
the closer a complexation is to being enthalpically neutral,
the larger the error in both H and Ka , and events that
are truly enthalpically neutral across a temperature range
cannot be measured. Such events are relatively rare, and
consequently the lower practical limit of ITC is more often
than not defined by insufficient solubility of the host and
guest, which limits the amount of heat generated and results
in a flat curve. The upper limit of Ka determination in ITC

(1 Ka Gt + Ka Ht )2 + 4Ka Gt
2Ka

17


(52)

pertains to the shape of the titration curve. If the binding is


so strong that a smooth S-curve in not obtained but rather a
step is observed, and this cannot be remedied by dilution
because insufficient heat is given out by complexation, then
although the H of complexation can be determined, the
obtained Ka will have an unacceptably large error. For c
values >500, the steepness of the curve is such that errors
in Ka ensue.
In summary, ITC is capable of determining a wider range
of binding constants than either NMR or UVvis spectrometry, and generally does so with much smaller errors
in Ka , G , H , S , and Cp : the only prerequisite
being that the binding event is not enthalpically neutral at
the temperature range being studied. One drawback of the
approach that has perhaps inhibited the growth of the technique in supramolecular chemistry is the relatively large
volume of the ITC cell (1.4 ml). However, recently sample requirements have decreased considerably, with most
recent instruments having a cell volume of only 200 l.
Automated ITC instruments can also be purchased, which
makes an in-house ITC a rapid and accurate means to determine thermodynamic data.

3.7

Should NMR, UVvis, and ITC


measurements give the same results?

It is important to note that the study of a single binding


event using multiple analytical techniques will not necessarily give the same thermodynamic data. Yes, they are
measuring the same complexation process, but their viewpoints are not the same. In ITC, for example, even if heats
of dilution or heats of protonation are taken into account
by the subtraction of data from reference titrations, what
is being measured is a global change in enthalpy. Thus,
the data obtained will include, for example, changes in
the solvation of the host and guest. In contrast, a technique such as NMR relies on measuring subtle changes
in electron distribution in and around one atom. This more
local viewpoint is sometimes apparent from thermodynamic
data generated by monitoring the shift of different protons in a host (or guest). In such instances, the obtained
binding constants may differ somewhat from proton to proton. The different viewpoints of the various techniques
noted, the obtained thermodynamic data should be similar.
If not, problems with the determinations must be examined
for.

Supramolecular Chemistry: From Molecules to Nanomaterials, Online 2012 John Wiley & Sons, Ltd.
This article is 2012 John Wiley & Sons, Ltd.
This article was published in the Supramolecular Chemistry: From Molecules to Nanomaterials in 2012 by John Wiley & Sons, Ltd.
DOI: 10.1002/9780470661345.smc005

18

3.8

Concepts

Does the complexation possess 1 : 1


stoichiometry?

3.9

The focus of this chapter has been on 1 : 1 binding events


summarized by (19), and the tacit assumption has been
that the stoichiometry of the binding event is known. For
binding events that are slow on the NMR timescale, it
is trivial to determine the ratio of host and guest in the
complex. However, in fast exchanging systems, and when
using UVvis spectrometry and ITC, we rely on fitting
programs to confirm the stoichiometry of complexation
from the binding isotherm. But what do we do if a poor
fit is obtained? Are we dealing with a 1 : 1 complexation
and poor data, or are we dealing with a higher order
process? One way for ascertaining the stoichiometry of
binding is to perform a so-called Job plot (also called
the continuous variation method). The theory behind, and
practice of, this approach has been described elsewhere,38, 39
so we only briefly discuss the practicalities of the method.
The approach is straightforward: a series of at least 10
experiments are carried out in which a metric proportional
to [HG]chemical shift, absorption, enzyme activityis
measured as a function of different ratios of host and guest
at constant concentration. The plot of this metric against
the mole fraction (of either host or guest) is hyperbolic in
the case of 1 : 1 complexation, with a maximum at the 1 : 1
ratio of the host and guest (Figure 6).
If a higher stoichiometry is under investigation, the
hyperbolic curve will be asymmetric, with a maximum
for the plot of the mole fraction of the host (xH )
at xH /[xH + xG ]. Thus for a 1 : 2 complex or a 2 : 1
complex, the maxima will be at 0.666 and 0.333,
respectively.

Arbitrary units

3E 05
2E 05
2E 05
1E 05
5E 06

Competition experiments

Each of the analytical technique discussed above has an


upper limit for determining a binding constant (104 , 106 ,
and 109 M1 for NMR, UVvis, and ITC, respectively).
What can be done if a host and guest are found to associate
more strongly than this limit? The strategy to overcome this
problem is to perform a competition experiment, whereby
the guest that binds too strongly to the host (GA ) is titrated
to a complex between the host and a more weakly binding
guest (GB ) of known association constant (Ka(B) ). Such a
titration will allow an apparent reduction of the binding
constant of GA (Ka(A) ). We can write (53) and (54):
H + GA

ka(A)

HGA

(53)

H + GB

ka(B)

HGB

(54)

where Ka(A) > Ka(B) and the latter is known. The mass balance equations then become: Ht = [H ] + [HGA ] + [HGB ],
GAt = [GA ] + [HGA ], and GBt = [GB ] + [HGB ], where Ht
is the total host and GAt and GBt are the totals of GA
and GB , respectively. If we take the example of slow
exchanging system by NMR, the determination of Ka(A)
is accomplished using these mass balance equations, the
relative integration of HGA and HGB , and the following
equation (55):
[HGA ][GB ]
Ka(A)
=
Ka(B)
[HGB ][GA ]

(55)

The mathematics necessary to fit a binding isotherm arising


from a competition experiment utilizing a fast exchanging
process in NMR or UVvis spectroscopy or ITC is
considerably more involved. These have been extensively
described in the literature,43, 44 particularly in regard to
the use of displacement assays for the determination of
association constants of spectroscopically silent guests.45
As a rule of thumb, competition experiments will extend
the upper binding constant limit of a technique by at
most 3 orders of magnitude. Put another way, they define
the upper limits of NMR, UVvis, and ITC respectively
to 107 , 109 , and 1012 M1 , respectively. The down side
of competition experiments is that the errors associated
with the determination of the weaker binding guest are
propagated into the competition experiment such that errors
in the determination of the larger association constant are
typically at least 1015% (and often higher).

0E + 00
0.0

0.2

0.4

0.6
Mole ratio

0.8

1.0

Figure 6 Continuous variation method (Job Plot) for a hypothetical 1 : 1 complex formation.

3.10

A word about higher order systems

Higher order systems, such as ternary binding/assemblies


involving a host and two guests, are also frequently

Supramolecular Chemistry: From Molecules to Nanomaterials, Online 2012 John Wiley & Sons, Ltd.
This article is 2012 John Wiley & Sons, Ltd.
This article was published in the Supramolecular Chemistry: From Molecules to Nanomaterials in 2012 by John Wiley & Sons, Ltd.
DOI: 10.1002/9780470661345.smc005

The thermodynamics of molecular recognition


encountered in supramolecular chemistry. Assuming one is
interested in the microscopic (individual) binding events
rather than the macroscopic (overall) binding process,
these systems are intrinsically more complicated than
1 : 1 binding systems: not only because of the higher
stoichiometries but also because there are variations within
each system of defined stoichiometry. For example, in a 1 : 2
hostguest system the two binding sites may be the same or
different, may be independent of each other or coupled (the
system possesses cooperativity), and may bind their guest
at the same time or sequentially. This means that there is
no standard mathematical model for each binding/assembly
of defined stoichiometry, but rather models have to be
customized for each variation. The result is a considerable
broadening of the range of base mathematical models which
is beyond the scope of this introduction. Nevertheless, it is
instructive to consider some of the general principles.
In the 1 : 1 binding systems, we devised a base mathematical model (36) that gave an exact solution to the association
constant determination. Such a mathematical model is often
called a closed-solution approach. There is in fact an alternative to this, and that is to carry out a open-form approach.
We have not mentioned this up until this point because
the mathematics of the 1 : 1 complexation is simple enough
that an exact solution can be readily sought. In the case of
higher stoichiometries, however, an exact (closed) solution
is often difficult or impossible. As we discuss below, the
only option therefore is to take an open approach to such
systems.
Let us briefly consider stoichiometry of higher order
systems. Not unexpectedly, the overarching theme here is
that, the greater the number of entities involved in the
final chemical entity, the more complicated the mathematics
needed to describe the system in question. Let us begin by
expanding (19) to give a general equilibrium equation for
higher order assembly:
mH + nG  Hm Gn

(56)

where Ka = [Hm Gn ]/[H ]m [G]n . The corresponding mass


balance equations are: Ht = [H ] + m[Hm Gn ] and Gt =
[G] + n[Hm Gn ], where Ht and Gt are respectively the total
host and guest present in solution. It is possible to again
obtain an expression of free guest ([G]) such as was carried
out for the 1 : 1 complexes (36), but as the stoichiometry
increases, the order of the polynomial generally increases as
well. For example, in most ternary hostguest systems, the
base mathematical model, or closed-form expression of [G],
takes the form of a cubic equation (general form: f (x) =
ax3 + bx2 + cx + d), whereas a binding event involving
four species generally takes the form of a quartic function
(general form: f (x) = ax4 + bx3 + cx2 + dx + e). Hence,

19

even for a simple ternary complex, the exact (or closedform) solution of the required cubic equation is relatively
involved.
As just mentioned, in a host with two binding sites, the
nature of the binding sites, whether there is cooperativity
in binding, and whether there is an order to guest complexation, all modify the base mathematical model. The nature
of the recognition sites is straightforward; they can either
be identical or different. Regarding cooperativity in the system, there are three possibilities: (i) the two binding pockets
are independent of one another and there is no cooperativity; (ii) the net free energy change of binding for the
overall process is more negative than the sum of the individual free energy changes arising from binding each single
guest: in other words the system displays positive cooperativity; (iii) the system is negatively cooperative system,
that is, the net free energy change for the overall process
is less negative than the sum of the free energy changes
of the individual binding events. Finally, binding can occur
randomly or sequentially. The simplest combination of all
these variations is where the two binding sites are identical and independent of each other, in which case the exact
solution base model simplifies to a quadratic equation. If
the binding sites are different, and/or there is cooperativity
involved, and/or there is an order to complexation, then the
base mathematical models are generally cubic equations. As
with 1 : 1 complexations, the next step after a base model
has been chosen is to modify it in order for the model to fit
the chosen analytical technique. Software that models the
different systems arising from combinations of these factors is available in modern ITC instruments, and a number
of researchers have provided their own in-house software
for those techniques such as NMR that are not specifically
designed for binding constant determinations.46, 47
As alluded to above, an alternative to closed solutions of
higher polynomials is to use an open, or iterative, approach.
This more general strategy to parsing out the thermodynamic data for each microscopic binding event within
higher order systems requires that each equilibrium constant expression and corresponding mass balance equation
be identified. From this, the concentration of each species at
set Ka values can be formulated by iteration, and this process layered on top of the normal iteration process that
fits the species distribution to the experimentally obtained
data.48 The advantage of an open solution approach is that
it can be readily expanded for higher stoichiometric systems that are too difficult, if not impossible, to analyze
via closed-form solutions. It is worth noting that many
researchers make available in-house software for these calculations.4951 A significant disadvantage of this approach,
however, is that this dual iterative approach requires an
initial estimate of the association constants at each microscopic step; and the more complicated the system is, the

Supramolecular Chemistry: From Molecules to Nanomaterials, Online 2012 John Wiley & Sons, Ltd.
This article is 2012 John Wiley & Sons, Ltd.
This article was published in the Supramolecular Chemistry: From Molecules to Nanomaterials in 2012 by John Wiley & Sons, Ltd.
DOI: 10.1002/9780470661345.smc005

20

Concepts

more the accuracy required for these estimations. The reason for this increased accuracy requirement is that different
initial Ka estimates can lead to different final Ka values. It
is therefore prudent to begin with a good estimate of the
Ka values and make sure that the final Ka values represent
a global minimum by changing the starting point to see
if this affects the outcome. It is also prudent to use chemical intuition when considering the final data. For example,
the garnering of four association constants from a binding isotherm possessing only one inflection point should be
treated with considerable suspicion. Likewise, if the magnitude of the obtained data seems overtly high or low relative
to published data, the practitioner should be cautious.
Of the numerous practical issues that can arise in
studying higher order systems, perhaps the most important
is the strength of cooperativity. The degree or strength of
cooperativity can be determined using the Hill equation.12
For practical purposes, however, we simply need to be
mindful of cases where binding is strongly cooperative
because this will affect our ability to observe selected
intermediates. For example, if it is strongly positively
cooperative (Ka2  Ka1 ), then the addition of host to guest
may not allow the observation of the intermediate HG
because the excess guest present will promote the full
conversion of HG to HG2 . In cases where exchange is
slow on the NMR timescale, this would mean that no
peaks corresponding to HG would be apparent; whereas
if exchange is fast on the NMR timescale (or we are
using UVvis or ITC), no inflection point corresponding
to this complex would be apparent in the binding isotherm.
If this is the case, then only the overall (macroscopic)
thermodynamic data can be obtained. One possible way to
obtain the microscopic data is to perform a reverse titration
of guest into host, but only if the positive cooperativity is
not too strong. Similarly, if strong negative cooperativity is
observed, it may not be possible to observe the formation
of HG2 even if the host is titrated into excess guest. These
influences of cooperativity mean that it is often instructive
to perform both a forward and a reverse titration when
studying higher order systems.49

CONCLUSION

We have focused here on the theoretical and practical


aspects of determining the thermodynamic profiles of 1 : 1
hostguest complexations. The determination of binding
constants, the change in free energy, the change in enthalpy,
and the change in entropy for complexation processes has
had a profound influence on our understanding of how
structure and solvent effects influence intermolecular reactions. Quantifying these thermodynamic parameters has

also allowed researchers to identify previously underappreciated noncovalent forces, and how they can play a role in
the properties of molecules. As a result, en masse, these
studies have been responsible for defining and uniting the
field of supramolecular chemistry. As the field moves forward and begins to address new challenges of a second
phase, it is likely that binding constant determinations will
continue to play a significant role. That role may change,
but the need to quantify thermodynamic parameters of association is of such importance in both synthetic and biological contexts that a good grounding will always be essential.
With that in mind, we hope that readers have found this
introduction to the topic useful.

REFERENCES
1. E. Fischer, Ber. Deutsch. Chem. Ges., 1894, 27, 29852993.
2. D. J. Cram, Angew. Chem. Int. Ed. Engl., 1988, 27,
10091020.
3. J.-M. Lehn, Angew. Chem. Int. Ed. Engl., 1988, 27, 89112.
4. C. J. Pedersen, Angew. Chem. Int. Ed. Engl., 1988, 27,
10211027.
5. C. A. Schalley, Analytical Methods in Supramolecular
Chemistry, Wiley-VCH, Weiheim, 2007.
6. E. V. Anslyn, J. Am. Chem. Soc., 2010, 132, 1583315835.
7. R. F. Ludlow and S. Otto, Chem. Soc. Rev., 2008, 37,
101108.
8. B. C. Gibb, Nat. Chem., 2009, 1, 1718.
9. B. C. Gibb, Nat. Chem., 2009, 1, 252253.
10. A. C. Balazs
16321634.

and

I. R. Epstein,

Science,

2009,

325,

11. G. M. Whitesides and R. F. Ismagilov, Science, 1999, 284,


8992.
12. E. V. Anslyn and D. A. Dougherty, Modern Physical
Organic Chemistry, University Science Books, Sausalito,
2006.
13. P. Atkins and J. de Paula, Physical Chemistry, W. H. Freeman and Company, New York, 2006.
14. (a) K. N. Houk, A. G. Leach, S. P. Kim, and X. Zhang,
Angew. Chem. Int. Ed., 2003, 42, 48724897;
(b) R. R. Krug, W. G. Hunter, and R. A. Grieger, Nature,
1976, 261, 566567.
15. S. R. de Groot and P. Mazur, Nonequilibrium Thermodynamics, Dover, New York, 1984.
16. I. R. Epstein and K. Showalter, J. Phys. Chem., 1996, 100,
1313213147.
17. I. R. Epstein, J. A. Pojman, and O. Steinbock, Chaos, 2006,
16, 037101.
18. J. Rao, J. Lahiri, L. Isaacs, et al., Science, 1998, 280,
708711.
19. K. A. Connors, Chem. Rev., 1997, 97, 13251358.

Supramolecular Chemistry: From Molecules to Nanomaterials, Online 2012 John Wiley & Sons, Ltd.
This article is 2012 John Wiley & Sons, Ltd.
This article was published in the Supramolecular Chemistry: From Molecules to Nanomaterials in 2012 by John Wiley & Sons, Ltd.
DOI: 10.1002/9780470661345.smc005

The thermodynamics of molecular recognition

21

20. M. V. Rekharsky and Y. Inoue, Chem. Rev., 1998, 98,


18751917.

36. D. Harries, D. C. Rau, and V. A. Parsegian, J. Am. Chem.


Soc., 2005, 127, 21842190.

21. E. Engeldinger, D. Armspach, and D. Matt, Chem. Rev.,


2003, 103, 41474586.

37. K. A. Dill, Biochemistry, 1990, 29, 71337155.

22. Y. Liu and Y. Chen, Acc. Chem. Res., 2006, 39, 681691.
23. H. Dodziuk, Cyclodextrins and their Complexes, WileyVCH, Weinheim, 2006.
24. C. S. Wilcox, N. M. Glagovich, and T. H. Webb, Designing
Synthetic Receptors for Shape-Selective Hydrophobic Binding, ACS Symposium Series, ACS (American Chemical Society), Washington, DC, 1994, vol. 568 (Structure and Reactivity in Aqueous Solution), pp. 282290.
25. Z. R. Laughrey and B. C. Gibb, Chem. Soc. Rev. 2011, 40,
363386.
26. S. C. B. S. Granick, Science, 2008, 322, 14771478.
27. D. Chandler, Nature, 2005, 437, 640647.
28. Y. Levy and J. N. Onuchic, Annu. Rev. Biophys. Biomol.
Struct., 2006, 35, 389415.
29. N. T. Southall, K. A. Dill, and A. D. J. Haymet, J. Phys.
Chem., 2002, 106, 521533.
30. J. Israelachvili, Intermolecular & Surface Forces, Academic
Press, San Diego, CA, 1991.
31. P. Setny and M. Geller, J. Chem. Phys., 2006, 125, 144717.
32. J. Ewell, B. C. Gibb, and S. W. Rick, J. Phys. Chem. B,
2008, 112, 1027210279.
33. E. S. Stoyanov, I. V. Stoyanova, and R. A. Reed, J. Am.
Chem. Soc., 2010, 132, 14841485.
34. S. F. Dec, K. E. Bowler, L. L. Stadterman, et al., J. Am.
Chem. Soc., 2006, 128, 414415.

38. K. A. Connors, Binding Constants: The Measurement of


Molecular Complex Stability, 1st edn, John Wiley & Sons,
Inc., New York, 1987.
39. K. Hirose, J. Incl. Phenom. Macrocycl. Chem., 2001, 39,
193209.
40. M. M. Pierce and B. T. Nall, Methods, 1999, 19, 213221.
41. A. Valazquez-Compoy, S. A. Leavitt, and E. Freire, Methods
Mol. Biol., 2004, 261, 3554.
42. W. B. Turnbull and A. H. Daranas, J. Am. Chem. Soc., 2003,
125, 1485914866.
43. B. W. Sigurskjold, Anal. Biochem., 2000, 277, 260266.
44. N. J. Buurma and I. Haq, Methods, 2007, 42, 162172.
45. E. V. Anslyn, J. Org. Chem., 2007, 72, 687699.
46. P. J. Munson and D. Rodbard, Anal. Biochem., 1980, 107,
220239.
47. A. P. Bisson, C. A. Hunter, J. C. Morales, and K. Young,
Chem.Eur. J., 1998, 4, 845851.
48. J. Huskens, A. Mulder, T. Auletta, et al., J. Am. Chem. Soc.,
2004, 126, 67846797.
49. J. C. D. Houtman, P. H. Brown, B. Bowden, et al., Protein
Sci., 2007, 16, 3042.
50. P. Gans, A. Sabatini, and A. Vacca, Talanta, 1996, 43,
17391753.
51. O. Raguin, A. Gruaz-Guyon, and J. Barbet, Anal. Biochem.,
2002, 310, 114.

35. K. J. Tielrooij, N. Garcia-Araez, M. Bonn, and H. J. Bakker,


Science, 2010, 328, 10061009.

Supramolecular Chemistry: From Molecules to Nanomaterials, Online 2012 John Wiley & Sons, Ltd.
This article is 2012 John Wiley & Sons, Ltd.
This article was published in the Supramolecular Chemistry: From Molecules to Nanomaterials in 2012 by John Wiley & Sons, Ltd.
DOI: 10.1002/9780470661345.smc005

Das könnte Ihnen auch gefallen