Characterization of sites of different thermodynamic affinities on the same metal center via isothermal titration calorimetry

Article abstract of DOI:10.1016/j.jcat.2013.02.020

We investigate the binding thermodynamics of a series of phosphorus ligands to a model compound, PdCl₂(solv)₂, where solv refers to a molecule of solvent, using isothermal titration calorimetry (ITC). ITC allows for the quantification of the equilibrium binding constant, the binding enthalpy, and the binding stoichiometry all in a single experiment. For systems in which two equivalents of ligand were able to bind to the Pd center, the binding sites on each Pd center in solution showed a different thermodynamic affinity for the same ligand. Changes in binding modes between different phosphorus ligands were due to steric bulk and poor electron-donating ability of such ligands. Our results demonstrate ligand binding was strongly enthalpy-driven due to solvent reorganization, which is the rearrangement of solvent molecules in the bulk solvent and the solvent molecules surrounding the solvated species.

Full text of DOI:10.1016/j.jcat.2013.02.020

Journal of Catalysis 302 (2013) 1–9
Contents lists available at SciVerse ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Characterization of sites of different thermodynamic affinities on the same
metal center via isothermal titration calorimetry
⇑
Eric G. Moschetta, Kristina M. Gans, Robert M. Rioux
Department of Chemical Engineering, 165 Fenske Laboratory, The Pennsylvania State University, University Park, PA 16802, United States
a r t i c l e i n f o
a b s t r a c t
Article history:
We investigate the binding thermodynamics of a series of phosphorus ligands to a model compound,
PdCl2(solv)2, where solv refers to a molecule of solvent, using isothermal titration calorimetry (ITC).
ITC allows for the quantification of the equilibrium binding constant, the binding enthalpy, and the bind-
ing stoichiometry all in a single experiment. For systems in which two equivalents of ligand were able to
bind to the Pd center, the binding sites on each Pd center in solution showed a different thermodynamic
affinity for the same ligand. Changes in binding modes between different phosphorus ligands were due to
steric bulk and poor electron-donating ability of such ligands. Our results demonstrate ligand binding
was strongly enthalpy-driven due to solvent reorganization, which is the rearrangement of solvent mol-
ecules in the bulk solvent and the solvent molecules surrounding the solvated species.
Received 4 December 2012
Revised 14 February 2013
Accepted 25 February 2013
Keywords:
Palladium
Phosphorus ligands
Calorimetry
Thermodynamics
Homogeneous catalysis
Solvent effects
Ó 2013 Elsevier Inc. All rights reserved.
1. Introduction
relationship). The main issue with these parameters is that corre-
lations between the observed kinetic and thermodynamic data
The most important factors when considering the activity and
selectivity of a homogeneous catalyst are how electronic and steric
effects of ligands affect the coordination and electronic character-
istics of the metal center. As such, the choice of ligands greatly im-
pacts the types of reactions for which the catalyst is best suited.
Phosphorus ligands are ubiquitous in organometallic chemistry
and are studied extensively to relate ligand properties to catalytic
activities and selectivities [1–3]. Experimental data, such as enthal-
pies of reaction, rate and equilibrium constants, and spectroscopic
data, are frequently correlated with the electronic and steric prop-
erties of phosphorus ligands in order to predict trends in reactivity
for other phosphorus ligands. Perhaps the most famous and useful
parameters for describing phosphorus ligands are the Tolman cone
and the ligand properties themselves are indirect. It would be more
informative to measure the interactions between metal centers
and ligands in solution directly such that contributions from other
factors, namely the solvent, may be considered in the metal–ligand
binding equilibria. Solvent is often implicated in catalytic mecha-
nisms due to the fact that solvent molecules interact with metal
centers and ligands in solution, but its exact role in a given mech-
anism is usually unknown and merely implied [6,7].
Pd chemistry is an integral part of homogeneous catalysis,
particularly for cross-coupling reactions. C–C bond formation is
an overwhelmingly popular synthetic technique, as evidenced by
the development of such Pd chemistries as the Heck reaction [8],
Suzuki coupling [9], and Sonogashira coupling [10]. These chemis-
tries, particularly the Heck reaction, rely on the in situ generation of
Pd(0) species from starting Pd salts, such as palladium acetate,
Pd(OAc)2, and PdCl2. These Pd(II) species are reduced to Pd(0),
the active species, most commonly by means of exogenous
phosphine ligands added to the reaction mixture. The ligands bind
to the Pd centers in order to form the active species. The thermo-
dynamic stability that the ligands impart to the Pd(0) species is
instrumental in determining the activity and selectivity of the
organometallic complex. PdCl2 is an easily accessible starting
material and is used in cross-coupling reactions as an alternative
to Pd(OAc)2. Pd(II) species such as PdCl2 have special use in
oxidation reactions, such as the famous Wacker process (oxidation
of ethylene to acetaldehyde) for which Pd(0) species are often
angle (h) and the Tolman electronic parameter (TEP, mCO). The cone
angle is an empirical measure of the overall steric bulk of a phos-
phorus ligand while the TEP is an empirical measure of the ability
of a phosphorus ligand to donate electrons to a metal [1]. Models
such as QALE [4] (Quantitative Analysis of Ligand Effects) and
ECW [5] (named for the E, C, and W parameters in the model itself)
emphasize the r- and p-acidities and basicities as well as the steric
bulk of the tested ligands such that the inherent properties of the
ligands may be used to understand trends in kinetics and thermo-
dynamics of organometallic processes (i.e., a structure–function
⇑
Corresponding author. Fax: +1 814 865 7846.
E-mail address: rioux@engr.psu.edu (R.M. Rioux).
0021-9517/$ - see front matter Ó 2013 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.jcat.2013.02.020

2
E.G. Moschetta et al. / Journal of Catalysis 302 (2013) 1–9
Table 1
for ligand properties) and to determine the effect of solvent on
the thermodynamics of binding triphenylphosphine, PPh3, to
PdCl2(solv)2 (2), where solv is a molecule of solvent. We use ITC
to characterize the metal–ligand interactions for each solvated sys-
tem and discuss our results in terms of the known electronic and
steric properties of the ligands and the inherent properties of the
solvents. We supplement our calorimetric analysis with 31P NMR
characterization of the reaction intermediates and products and
UV–Vis spectroscopy to verify reaction stoichiometry, so that we
may adequately describe the binding equilibria occurring in the
ITC and validate the appropriate choice of binding model. Further-
more, we consider the solvation and desolvation of the solutes (1, 2,
and the phosphorus ligands) and solvent reorganization manifest-
ing themselves in metal–ligand binding equilibria.
Cone angle (h) and Tolman electronic parameter (TEP) data for the phosphorus ligands
used in this study. There is no reported value for the TEP for PFu3 in the literature. All
other values are taken from Ref. [1].
Phosphine
Abbreviation
Cone angle, h (degrees)
TEP, mCO (cmꢀ1
)
P(C6H5)3
PPh3
P(p-FC6H5)3
PFu3
P(OPh)3
P(o-tolyl)3
145
145
133
128
194
2068.9
2071.3
–
2085.3
2066.6
P(p-FC6H5)3
P(C4H3O)3
P(OC6H5)3
P(o-CH3C6H4)3
unsuitable [11,12]. Specifically, Pd(II) is a better reagent for
oxidation reactions with substituted and cyclic olefins and alkynes,
though catalytic processes involving Pd(II) frequently generate
Pd(0) intermediates, which necessitates the use of a reoxidizing
agent [13,14]. Thus, there is considerable interest in understanding
the thermodynamic stability, and potential regeneration, of Pd spe-
cies in homogeneous catalytic processes.
2. ITC theory
ITC theory is well-developed in the literature, but due to the fact
that there are very few studies of solution-phase calorimetry that
attempt to discern different binding modes between metals and li-
gands in organometallic chemistry using thermodynamic data, we
briefly introduce the main concepts behind ITC theory and empha-
size the mathematics that are the most pertinent to our work. First,
we consider the binding of a single type of ligand, L, with a metal
receptor, M, in solution:
Calorimetry is an excellent medium for ascertaining data with
respect to understanding catalytic activity and metal–ligand
binding equilibria from a thermodynamic perspective. Obtaining
thermodynamic data in the liquid phase for a variety of metal–
ligand interactions allows organometallic chemists to design new
chemical syntheses because the data reveal vital information
regarding the stability of the resulting organometallic complexes
and how the sterics and electronics of the metal center may influ-
ence the reaction pathway for a particular class of substrates [15].
Isothermal titration calorimetry (ITC) is a technique capable of
characterizing receptor–ligand interactions in solvated systems,
to measure the equilibrium binding constants, enthalpies of reac-
tion, and reaction stoichiometries in a single experiment [16].
Briefly, ITC experiments consist of a sample cell containing a
solution of the receptor (in our case, this is the metal complex
PdCl2(solv)2) and reference cell containing the solvent that are held
at the same temperature. The titrant solution, a solution of ligand,
is then titrated into the sample cell once the power rating supplied
to the sample cell is constant (note: the power rating is the thermal
compensation of the calorimeter that is applied to the sample cell
to keep it at the same temperature as the reference cell). Heat is
then evolved or absorbed due to the metal–ligand binding and
the controller compensates for these heat effects to maintain the
cell at a constant temperature. Ligand is injected into the cell until
the system is saturated, that is, no additional heat is evolved or ab-
sorbed due to ligand binding, only heat of mixing is present, and
the resulting peaks are integrated to determine the total amount
of heat released per injection. These integrated heats are then fit
to an appropriate binding model in order to obtain the equilibrium
binding constant, enthalpy of reaction, and reaction stoichiometry.
ITC is widely used in biological and supramolecular systems
because the thermodynamic information reveals how the guests
(ligands) interact with the hosts (receptors) such that optimal
ligands and receptors may be designed for specific processes.
Modern calorimeters are capable of measuring equilibrium con-
stants as high as 109 Mꢀ1 due to their nW level of sensitivity
[17]. This information is critical in any field where molecular rec-
ognition is a central topic, such as drug design, metalloenzymes,
and protein–ligand interactions, among others, in which the objec-
tive is to maximize a host’s affinity for a specific guest. There are
extensive reviews of the ITC literature over the last decade, includ-
ing how ITC is used to obtain kinetic data via heat evolution as well
as new types of systems, such as zeolites and nanoparticles, that
are being investigated calorimetrically [18–25].
M þ nL $ MLn
ð1Þ
where n represents the total number of ligands (i.e., the binding
stoichiometry) that bind to the metal and MLn is the final complex
that forms between the metal and n ligands. In our systems, M
represents PdCl2(solv)2 and L represents the series of phosphorus
ligands. We will only consider two types of binding within the
scope of this work, though interested readers are referred elsewhere
for additional information with respect to the establishment of dif-
ferent binding models as well as full derivations of these models
[26,27]. The first type of binding is independent binding, in which
the metal may have several binding sites, but each site is thermody-
namically identical and has the same thermodynamic affinity for
the ligand. We can write the general equilibrium binding constant,
Ki, for each binding step as
½MLiꢁ
Ki ¼
ð2Þ
½MLiꢀ1ꢁ½Lꢁ
where the terms in brackets represent concentrations of the
respective species, MLi represents a metal center with i ligands
bound to it, and MLiꢀ1 represents a metal center with i ꢀ 1 ligands
bound to it, extending from i = 0 to i = n. The second type of binding
is multiple-site binding, in which the metal has two thermodynam-
ically different types of sites. Specifically, each site has its own
affinity for the same ligand, but the occupancy of one site does
not affect the affinity of the other, that is, the sites do not exhibit
cooperative binding behavior [26]. For our systems, we can model
this behavior by considering the sequential binding of the same
type of ligand to the same metal center twice (n = 2):
M þ L $ ML
ð3Þ
ML þ L $ ML2
ð4Þ
For Eqs. (3) and (4), the equilibrium constants are given by Eqs. (5)
and (6), respectively
½MLꢁ
K1
K2
¼
¼
ð5Þ
ð6Þ
½Mꢁ½Lꢁ
This study aims to determine the thermodynamics of ligand
binding to a model compound, PdCl2(MeCN)2 (1), in acetonitrile
(MeCN) using various phosphorus ligands (see Table 1 in Section 3.1
½ML2ꢁ
½MLꢁ½Lꢁ

E.G. Moschetta et al. / Journal of Catalysis 302 (2013) 1–9
3
With full expressions for the respective equilibrium constants,
it is now possible to combine these expressions with mass balances
on each component:
experimentally determine the heat of mixing of the phosphorus li-
gands with the pure solvent. These blanks were subtracted from
the experiments with PdCl2(solv)2 in the cell and integrated to iso-
late the heat evolved from metal–ligand interactions. Data analysis
was performed using NanoAnalyze v2.1 from TA Instruments using
the Independent Sites algorithm (see Section 2 for derivations of
the appropriate models) [26]. We used a modified version of the
Multiple Sites model in Microsoft Excel in order to account for
the inability of NanoAnalyze to fit integrated heat data near zero
accurately. We used the Solver function in Excel to minimize the
sum of the squares of the differences between the measured heat
and the calculated heats. The first integrated heat point in each
data set is omitted from each fit because the syringe allows for a
miniscule amount of ligand to mix before the experiment starts
which makes the first data point unreliable. Error analysis was also
performed using NanoAnalyze v2.1 via its Statistics function. The
uncertainties in the parameters obtained from the fits are
calculated by adding perturbations to the optimized fits and then
refitting the models for a set number of trials. Each perturbation
obeys a Gaussian distribution with the same standard deviation
generated from the original fit. The error in each data set was
determined within one standard deviation for 1000 trials. The error
values reported in the main text for the K values are also multiplied
by the factor outside of the parentheses. For example, a K value of
(5.99 0.01) ꢂ 105 means that the error is 0.01 ꢂ 105, or 1000.
Independent verification of K values was unsuccessful due to poor
detection of species by UV–Vis and 31P NMR spectroscopies over
the most appropriate ranges of concentrations. Specifically, as Hir-
ose details, higher K values require lower concentrations of metal
and ligand in order to detect the resulting complexes reliably,
and for our systems, we could not detect appreciable signals at
½MꢁT ¼ ½Mꢁ þ ½MLꢁ þ ½ML2ꢁ
ð7Þ
½LꢁT ¼ ½Lꢁ þ ½MLꢁ þ 2½ML2ꢁ
ð8Þ
Eqs. (7) and (8) can be extended to any binding system of n ligands,
noting that [M]T and [L]T are the total concentrations of metal and
ligand in the calorimeter cell. These mass balances can be substi-
tuted into the expressions for the equilibrium constants, such that
only the total concentrations (i.e., measurable quantities) appear
in the final ITC equations. This flexibility allows the experimenter
to track the progress of the binding equilibria by calculating the
molar ratio, the total amount of ligand in the calorimeter cell to
the total amount of metal in the cell, as the independent variable.
The dependent variable in ITC experiments is the total amount of
heat released per injection of ligand, dQ:
X
dQ ¼ V
D
Hid½MLiꢁ
ð9Þ
i
where V is the volume of the calorimeter cell,
D
Hi is the enthalpy of
binding for the formation of MLi, and d[MLi] is the incremental
amount of complex, MLi, formed during the injection. Eq. (9) may
be extended to any number of complexes in solution. Substituting
Eqs. (5)–(8) into Eq. (9) allow dQ to be written explicitly in terms
of Ki,
can be fit to a statistical model as a function of the molar ratio that
determines the binding parameters (Ki, Hi, and ni) in a single
DHi, [M]T, and [L]T, meaning that the heats from each injection
D
experiment (see Supporting Information for the full equations and
discussion).
the recommended concentrations (less than 1 lM for metal and li-
3. Experimental
gand for K values in excess of 105 Mꢀ1) [30].
3.1. Materials and solution preparations
3.3. 31P NMR characterization of reaction intermediates and products
PdCl2 and the phosphorus ligands P(o-tolyl)3, PFu3, and P(p-
FC6H4)3 were obtained from Alfa Aesar. P(OPh)3 was obtained from
TCI and PPh3 was obtained from Sigma. MeCN and pyridine were
obtained from EMD while DMSO was obtained from BDH and
DMF was obtained from Sigma. All chemicals were used without
further purification.
All 1–5 mM solutions of PdCl2(solv)2 were prepared by dissolv-
ing 8.9–44.3 mg of anhydrous PdCl2 in 50 mL of the appropriate
degassed solvent and stirred vigorously overnight with gentle
heating. The structures of PdCl2(solv)2 for each solvent are trans
as reported in the literature [28,29]. All phosphorous ligand solu-
tions were prepared immediately before each titration by dissolv-
ing the appropriate amount of ligand in the desired degassed
solvent followed by vigorous stirring.
Mixtures of one or two equivalents of phosphorus ligand to one
equivalent of PdCl2(solv)2 were prepared in advance and allowed
to stir at room temperature for 1 h. An aliquot of the Pd solution
was mixed with an equal volume of deuterated solvent (Cambridge
Isotope Laboratories) in an NMR tube. Samples were run on a Bru-
ker AV-360 at room temperature with proton decoupling. Each
sample was run for 128 scans and each spectrum was observed
for the chemical shifts of the Pd complexes.
3.4. Solution calorimetry
Solution calorimetry experiments were performed in a TAM III
microcalorimeter (TA Instruments) at 25 °C. Samples of PPh3 and
P(OPh)3 were placed into glass ampoules and sealed with wax to
prevent premature mixing of solvent with the ligands. Sealed am-
poules were immersed in 25 mL of MeCN in a reaction cell and stir-
red at 600 rpm. Ampoules were then broken and the ligand (solute)
was allowed to mix with the solvent for 1 h while monitoring the
heat flow. An electronic heat pulse was applied to the reaction cell
before and after dissolution to calibrate the heat capacity in each
instance. An empty ampoule (blank) was broken to account for
the heat evolved due to breaking the ampoule. The total heat
evolved or absorbed during each experiment was obtained using
the Analyze Experiment function in the SolCal v1.2 software (TA
Instruments). Each heat value was then corrected for the blank
experiment and then normalized to the total amount of moles of
solute dissolved to calculate the enthalpy of dissolution. The uncer-
tainty for each value originated from the average difference be-
tween the evolved heats for multiple experiments.
3.2. ITC experimental procedure for binding phosphorus ligands to
PdCl2(solv)2
ITC experiments were performed using a NanoITC III calorime-
ter (TA Instruments, New Castle, DE) equipped with hastelloy cells
(V = 1.056 mL). All titrations were carried out at 25 °C using a 250-
lL syringe at a stirring rate of 250 rpm. The sample cell contained
PdCl2(solv)2 and the reference cell contained the chosen solvent.
All solvents were degassed prior to titration. The ‘‘heat flow’’ base-
line was allowed to equilibrate once the reference and sample
solutions were loaded into the cells. The titrations began after
equilibration of the power rating. Titrations were run as an incre-
mental series of injections of the appropriate phosphorus ligand
into the PdCl2(solv)2 solution. Blank experiments were conducted
under identical conditions with only solvent in the sample cell to

4
E.G. Moschetta et al. / Journal of Catalysis 302 (2013) 1–9
of the mono-ligated complex PdCl2(PPh3)(py), and one at
d = ꢀ6.8 ppm, indicative of free PPh3. Dissolution of pure PdCl2
(PPh3)2 into py yielded an identical spectrum to the solution of
PdCl2(py)2 and PPh3 while a solution of pure oxidized triphenyl-
phosphine, OPPh3, produced a singlet at d = 24.8 ppm, leading to
the conclusion that only one equivalent of PPh3 binds to PdCl2(py)2
in py. A broad resonance was observed for mixtures of PdCl2
(MeCN)2 and P(o-tolyl)3 at d = 21.7 ppm, which is shifted from free
P(o-tolyl)3 at d = ꢀ31.6 ppm. ITC results show that only one equiv-
alent of P(o-tolyl)3 was able to bind, so we attribute this broad res-
onance to the existence of PdCl2(P(o-tolyl)3)(MeCN). No evidence
of chloro-bridged Pd dimers or ionization isomers was observed
for any of the tested ligands and solvents. For one equivalent of li-
gand, PPh3 (d = 31.2 ppm), P(p-FC6H4)3 (d = 28.8 ppm), and PFu3
(d = ꢀ24.7 ppm) have different resonances than their respective
two equivalent (cis or trans) complexes. The PPh3 resonance corre-
sponds to coordinated ligand with a bound MeCN molecule trans to
it, meaning that the difference in shifts for the one equivalent com-
plexes compared to the two equivalent complexes is due to the
coordination of the second ligand as originally reported by Colacot
et al. [33]. Based on these 31P NMR results, we present Scheme 1 as
a description of the binding equilibria that occur in the ITC exper-
iments. Scheme 1 is a specific representation of Eqs. (3) and (4) and
describes both independent (just the K1 equilibrium reaction) and
multiple-site binding. For our systems, we observed one equivalent
of bound ligand for the independent cases while we observed two
equivalents of bound ligand for the multiple site cases.
4.2. ITC results for binding phosphorus ligands to PdCl2(MeCN)2
Fig. 2 shows thermograms for binding ligands to 1 in MeCN at
25 °C and the integrated heats and the best-fit binding models
for each system as measured by ITC; best-fit parameters are
compiled in Table 2. Note that n1 and n2 refer to the step-by-step
stoichiometries presented in Eqs. (3) and (4) and the overall bind-
ing equilibria presented in Scheme 1 (n = n1 + n2). Additionally, we
provide the ITC experimental conditions in Table A1 of the Sup-
porting Information. The ligands exhibited two binding modes
with Pd: either two ligands were able to bind to the same Pd atom,
each ligand with its own thermodynamic affinity for the Pd center
as expressed in Eqs. (3) and (4), or only one ligand was able to bind
to Pd as described by Eq. (3) only (independent binding). The
difference in affinities between sites on the same Pd atom is due
to the presence of bound phosphorus ligand after the first equilib-
rium step, which changes the ground-state thermodynamics of the
intermediate complex when compared to 1, resulting in two
distinct equilibrium constants for identical ligands (K1 and K2)
[34]. As seen in Section 4.1, two equivalents of PPh3, PFu3, and
P(p-FC6H4)3 were able to bind to 1 in MeCN, while only one equiv-
alent each of P(OPh)3 and P(o-tolyl)3 were able to bind. 31P NMR
confirmed that PPh3 and P(p-FC6H4)3 formed the trans product only
while PFu3 formed the cis product only in MeCN (Fig. 1a). Redfield
and Nelson studied the thermodynamics of the cis–trans isomeri-
zation of Pd(II)–phosphine complexes and found the cis isomer
was generally the most stable by as many as two orders of magni-
tude in K [35]. For the trans complexes, this instability is reflected
in lower K2 values, while the more stable cis complex has a higher
K2 value. We are not ascribing these differences in K2 between the
cis and trans complexes to the well-known trans influence because
both chloride ligands remain bound in the cis complex. The trans
influence is the ability of a ligand already bound to a metal center
to weaken the bond of the ligand trans to it, altering the ground-
state thermodynamics of the complex itself [34]. Empirically, phos-
Fig. 1. (a) 31P NMR spectra of PdCl2(solv)2 and 2 equivalents of phosphorus ligand
as synthesized in the ITC. (b) 31P NMR spectra of PdCl2(solv)2 and 1 equivalent of
phosphorus ligand for verification of the intermediates is presented in Scheme 1.
The resulting Pd–P complexes, along with the reaction solvents, are listed for each
spectrum.
4. Results and discussion
4.1. 31P NMR results for phosphorus ligand binding to PdCl2(solv)2
Fig. 1a shows the 31P NMR data for the final bis-ligated com-
plexes and Fig. 1b shows 31P NMR data for the intermediate
mono-ligated complexes formed (applicable for the complexes in
which no free ligand was observed in the presence of two equiva-
lents of ligand). For two equivalents, PdCl2(PPh3)2 (d = 23.4 ppm)
and PdCl2(P(p-FC6H4)3)2 (d = 21.2 ppm) exhibited downfield sing-
lets characteristic of the trans complex while PdCl2(PFu3)2
(d = ꢀ22.2 ppm) exhibited the upfield singlet characteristic of the
cis complex [31,32]. A mixture of PdCl2(MeCN)2 and P(OPh)3 dis-
played singlets at d = 83.0 ppm indicative of the complex PdCl2
(P(OPh)3)(MeCN) and at d = 128.2 ppm indicative of free P(OPh)3,
which agreed with the ITC results. Both PdCl2(PPh3)2 in DMSO
(d = 23.0 ppm) and DMF (d = 23.0 ppm) showed singlets character-
istic of the trans complex. The addition of two equivalents PPh3 to
PdCl2(py)2 produced two singlets: one at d = 28.1 ppm, indicative
phorus ligands are better trans directors than Cl ligands, but our 31
P
NMR data confirm that the Cl ligands are not substituted, so we
attribute the formation of both cis and trans complexes to the

E.G. Moschetta et al. / Journal of Catalysis 302 (2013) 1–9
5
K1
Solv
Cl
Solv
Cl
Cl
Pd
Pd
+ PR3
Pd
Cl
PR3
Solv
R3P
Cl
Cl
K2
PR3
Solv
Cl
Cl
Pd
+ PR3
PR3
K2
R3P
Cl
PR3
Cl
Pd
Scheme 1. Illustration of metal–ligand equilibria in solution. For each equilibrated step, a PR3 ligand displaces a bound solvent molecule (solv). For cases where two
equivalents of ligand bind, either the cis (PFu3) or trans (PPh3 and P(p-FC6H4)3) product is formed exclusively.
Fig. 2. Real-time ITC thermograms for (a) PPh3 and (b) P(OPh)3 binding to 1 in MeCN at 25 °C with (c) and (d) as the respective integrated heat data with fitted models.
‘‘Power’’ refers to the thermal compensation of the calorimeter to keep the sample at a constant temperature (positive peaks are exothermic – the heat evolved reduces the
power compensation resulting in a positive value when the baseline is subtracted). See Fig. A1 in the Supporting Information for thermograms of the other ligands.
inherent weakness of the coordinated MeCN molecules. We con-
firmed metal–ligand binding ratios using UV–Vis spectroscopy
and we provide the data in Fig. A3 of the Supporting Information
[30]. As discussed in Section 2, the binding sites on the Pd center
are thermodynamically independent and the state of one site,
bound or unbound, does not affect the affinity of the other site
[26]. This trait is reflected in the shapes of the isotherms as two
different sigmoidal regions (see Fig. 2a), for the three ligands for
which two equivalents of ligand bind, characteristic of multiple-
site binding. Conversely, the two ligands for which only one equiv-
alent binds display a single inflection point, characteristic of inde-
pendent binding. The different binding modes are consistent with
the Tolman cone angles and electronic parameters of phosphites as
compared to phosphines. P(o-tolyl)3 has the largest cone angle,
194°, so it is logical that only one ligand binds to the Pd centers
in solution [1]. P(OPh)3 is the poorest electron donor, though being
a triarylphosphite, it is a better
p
-acceptor than the triarylphos-
-acceptor than P(o-tolyl)3
phine, PPh3 [1,34]. It is also a better
p

6
E.G. Moschetta et al. / Journal of Catalysis 302 (2013) 1–9
Table 2
Thermodynamic parameters for the binding of PR3 ligands to 1 in MeCN at 25 °C. Errors were calculated using the statistics function in NanoAnalyze (see Section 3.2).
Ligand
K1
D
D
G1
G2
D
D
H1
H2
T
T
D
D
S1
S2
n1
n2
K2
(Mꢀ1
)
(kJ/mol)a
(kJ/mol)b
(kJ/mol)
PPh3
(1.20 0.01) ꢂ 109
ꢀ51.8 8.6
ꢀ41.8 5.6
ꢀ45.8 3.2
ꢀ36.2 8.7
ꢀ30.6 1.4
ꢀ42.8 4.2
ꢀ62.4 0.2
ꢀ37.0 0.2
ꢀ64.1 0.1
ꢀ37.5 0.1
ꢀ36.7 0.5
ꢀ46.2 0.5
ꢀ10.6 8.6
1.03 0.01
1.32 0.01
(2.13 0.01) ꢂ 107
4.8 5.6
P(p-FC6H4)3
PFu3
(1.06 0.01) ꢂ 108
(2.20 0.01) ꢂ 106
ꢀ18.3 3.2
ꢀ1.3 8.7
ꢀ6.1 1.5
ꢀ3.4 4.2
1.02 0.01
1.34 0.01
(2.27 0.01) ꢂ 105
1.46 0.01
1.03 0.01
(3.15 0.01) ꢂ 107
P(OPh)3
P(o-tolyl)3
(2.91 0.01) ꢂ 106
ꢀ36.9 10.9
ꢀ123.6 1.0
ꢀ86.7 11.0
0.82 0.01
–
–
–
–
–
(2.49 0.01) ꢂ 105
ꢀ30.8 10.5
ꢀ49.7 0.2
ꢀ19.0 10.5
1.16 0.01
–
–
–
–
–
a
The values for
DG and TDS were calculated from the given K and DH values.
b
The integrated heats for each titration were analyzed using either the multiple sites model or the independent model to obtain binding constants (Ki), enthalpies of
Hi), and binding stoichiometries (ni).
binding (
D
Fig. 3. Real-time ITC thermograms for PPh3 binding to 2 at 25 °C in (a) DMSO and (b) py with (c) and (d) as the respective integrated heat data with fitted models. See Fig. A2
for the thermogram of DMF.
and has a smaller cone angle, which would explain its larger affin-
ity for binding with 1 [4,36,37]. The Tolman electronic parameter
(Table 1) for PPh3 is less than that of P(p-FC6H4)3, indicating that
the overall electron-donating ability of PPh3 is greater [1,34]. These
two ligands have identical cone angles and the difference in elec-
tron-donating ability explains why PPh3 binds more strongly to 1.
best-fit parameters for all tested solvents are in Table 3. The struc-
tures of 2 are trans as reported in the literature and were prepared
as reported (see Section 3.1) [28,29]. Multiple-exchanged
Pd(DMSO)n species (n > 1) have been observed, such as
[Pd(DMSO)4]2+, but their formation requires silver perchlorate,
AgClO4 (our synthesis does not use silver perchlorate) [38], while
our preparation for PdCl2(py)2 is identical to that of Gupte and
Chaudhari [39] and the formation of PdCl2(py)2 was structurally
verified via XRD by Liao and Lee [40]. Therefore, we do not expect
formation of either Pd(DMSO)n or Pd(py)n (n > 1). For DMSO and
N,N-dimethylformamide (DMF), two equivalents of ligand bind as
4.3. ITC results for binding PPh3 to PdCl2(solv)2
Fig. 3 shows the ITC results for the binding of PPh3 to 2 in di-
methyl sulfoxide (DMSO) and pyridine (py) at 25 °C while the

E.G. Moschetta et al. / Journal of Catalysis 302 (2013) 1–9
7
Table 3
Thermodynamic properties for the binding of PPh3 to 2 in various solvents at 25 °C. Errors were calculated using the statistics function in NanoAnalyze (see Section 3.2).
Solvent
K1
D
D
G1
G2
D
D
H1
H2
T
T
D
D
S1
S2
n1
n2
K2
(Mꢀ1
)
(kJ/mol)
(kJ/mol)
(kJ/mol)
MeCN
DMSO
DMF
py
(1.20 0.01) ꢂ 109
ꢀ51.8 8.6
ꢀ41.8 5.6
ꢀ48.0 3.3
ꢀ29.9 9.8
ꢀ42.2 1.3
ꢀ27.9 1.4
ꢀ62.4 0.2
ꢀ37.0 0.2
ꢀ41.0 0.1
ꢀ33.2 0.3
ꢀ62.8 0.2
ꢀ35.5 0.1
ꢀ10.6 8.6
1.03 0.01
1.32 0.01
(2.13 0.01) ꢂ 107
4.8 5.6
(2.55 0.01) ꢂ 108
(1.75 0.01) ꢂ 105
6.9 3.3
ꢀ3.3 9.8
ꢀ20.5 1.4
ꢀ7.7 1.4
12.1 7.0
–
1.06 0.02
1.17 0.01
(2.51 0.01) ꢂ 107
0.90 0.01
1.09 0.02
(7.64 0.01) ꢂ 104
(2.08 0.01) ꢂ 104
ꢀ24.6 7.0
ꢀ12.5 0.3
1.29 0.02
–
–
–
–
seen in MeCN. In py, only one ligand binds, which is attributed to
two factors. It has a high donor number (DN) which is an empirical
measure of the ability of a solvent to donate electrons to a solute,
the displacement of the last bound solvent molecule. It is notewor-
thy that py has nearly equal enthalpy and entropy contributions.
Regarding the large RDE of py, it makes sense that the observed
PdCl2 [41]. This evidence indicates that
a
py molecule
D
H decreased in magnitude relative to the other solvents because
(DN = 138.5 kJ/mol, AN = 14.2 (unitless), see below for definition
of AN) binds to the Pd center, occupying one of the sites as seen
by ITC. The other factor can be traced to a calorimetric study of
the thermodynamics of ligand exchange for PdCl2(C7H5N)2 to
determine relative displacement energies (RDEs) comparing the
intrinsic binding strength of one ligand to another, which found
that py has an RDE less than that of PPh3, which means that dis-
the displacement of the bound py contributes unfavorably to DG
[42,47]. For MeCN and DMF, the entropic contributions increase
after the second ligand binds, in agreement with the trend from
the different ligands. The DMSO system decreases in entropy after
the second ligand binds because it has the highest AN of the tested
solvents; it is the best electron acceptor and the desolvation of the
PPh3 ligand, which has a lone pair of electrons on the P atom, is not
placing PPh3 has a more unfavorable contribution to
D
G [42]. We
as favorable as it is in the other solvents. Our
the DMSO system, are all smaller in magnitude than their respec-
tive S1 values because of fluxionality at the Pd centers (with re-
spect to solvent molecules). Certainly, we expect the Pd–P
complexes to be fluxional in their own right, but the PdCl2(solv)2
complexes rapidly exchange bound solvent molecules as ligands
with bulk solvent molecules. As ligands bind to the Pd center, this
solvent fluxionality is lost, which manifests itself as a decrease in
entropy. We presume this loss in fluxionality to be less dramatic
for mono-ligated Pd–P complexes than for PdCl2(solv)2, which ex-
DS2 values, except for
tested py binding to 1 in MeCN using ITC and found that only
one equivalent of py binds (shown in our other study [43]), agree-
ing with the RDE study. For our system, PdCl2(py)2 is in excess py,
whereas the previous study was conducted in dichloromethane, so
it is reasonable to attribute the presence of only one bound PPh3
ligand to the excess of py relative to unbound PPh3 (Fig. 3). The
binding is strongest in MeCN (highest K values) and weakest in
DMF (lowest K values). In terms of DN and acceptor number (AN,
which is an empirical measure of the ability of a solvent to accept
electrons from solutes), MeCN (DN = 59.0 kJ/mol, AN = 18.9) is the
weakest electron donor and a better electron acceptor than DMF
(DN = 111.3 kJ/mol, AN = 16.0), so bound MeCN molecules are more
easily displaced by free PPh3 [41]. The best-fit values of binding in
DMSO (DN = 124.7 kJ/mol, AN = 19.3) and DMF are similar in terms
of the orders of magnitude of the obtained K values due to having
similar DN and AN values (note: all DN and AN values are taken
from Refs. [41,44]).
D
plains why
DS2 values are smaller in magnitude than DS1 values.
Solvent molecules can form adducts with the ligands, which would
change their overall polarity and subsequently change the enthal-
py of solvation of the products (Pd–P complexes). Partenheimer
et al. studied this effect for acid–base systems in cyclohexane
and CCl4 and noticed more cyclohexane molecules formed adducts
with the base, dimethylacetamide, than did CCl4 which explained
the differences in the observed enthalpies of acid–base reactions
between the solvents [48]. Overall, ligand binding is enthalpy-dri-
ven, that is, the large enthalpy values indicate that solvent reorga-
nization contributes greatly to the observed enthalpies, as
originally noted by Chervenak and Toone [49]. Solvent reorganiza-
tion includes the loss of solute–solvent interactions (in the solva-
tion shells around the solutes) during the binding event that is
necessarily accompanied by new solvent–solvent interactions that
occur once the expelled solvent molecules reenter the bulk [50]. In
their study, Chervenak and Toone used ITC to measure the binding
thermodynamics of several systems, such as protein–carbohydrate
and protein–peptide binding equilibria, in both H2O and D2O and
4.4. Understanding the contributions to the enthalpy and entropy of
binding
The obtained
DH and DS values provide insight as to what the
dominant effects are for a given set of metal–ligand interactions.
The obtained ITC parameters are observed values, rather than
intrinsic values. The observed enthalpy includes the loss of sol-
ute–solvent bonds in the form of solvent reorganization (PdCl2
and ligand are solutes), the loss of van der Waals forces, solvation
and conformational changes at the binding site, and metal–ligand
binding [45,46]. The observed entropy change includes contribu-
tions from losses of translational, rotational, and vibrational de-
grees of freedom and from the solvation and desolvation of the
solutes; however, the main positive contribution to the observed
entropy is expulsion of solvent molecules bound to the metal,
which normally manifests itself as an increase in entropy [45,46].
From Table 2, ligand binding is enthalpy-driven because of the
large, exothermic enthalpies and small entropies. The ligands for
which two equivalents bind to 1 all have an increase in entropy
after the first ligand is bound, which is attributed primarily to
determined that decreased
D
H values in D2O were compensated
G constant. They assumed
by changes in S, practically leaving
D
D
that intrinsic ligand–receptor binding was invariant of the nature
of the solvent, meaning that solvent reorganization accounted for
25–100% of the observed binding enthalpies for the systems stud-
ied [49].
Examination of theoretical Pd–P binding energies helps to
explain how the observed binding enthalpies differ greatly from
one P ligand to another in solution. Fey et al. determined bond
dissociation energies (BDEs) from [PdCl3L]ꢀ where L represents a

8
E.G. Moschetta et al. / Journal of Catalysis 302 (2013) 1–9
specific P ligand [51]. For the five P ligands in our study, the BDEs
range from ꢃ108 to 148 kJ/mol, so it is unsurprising to learn that
ligand binding is enthalpy-driven from a theoretical perspective.
Their calculations do not account for the effect of solvent and their
model Pd structures are obviously different from ours, but these
values are a reasonable estimate for theoretical Pd–P binding ener-
gies. With the exception of P(OPh)3, our observed enthalpies are a
factor of 2–3 smaller for binding one equivalent of ligand to Pd,
which we attribute to the fact that the ligands and Pd complexes
are solvated. Additionally, their model Pd compound is negatively
charged, while ours is neutral.
We found that these metal–ligand interactions are enthalpy-
driven. These large, exothermic enthalpies indicate that solvent
reorganization likely contributed greatly to the observed enthal-
pies and played an active role in the metal–ligand equilibria. We
attributed the ability of the solvent molecules to interact with
the solutes as another contributor to the observed enthalpies of
binding. All ligands for which two equivalents bind to 1 had an in-
crease in entropy after the second ligand was bound, likely due to
displacement of a bound solvent molecule. The largest decrease in
observed entropy was seen for P(OPh)3 and was attributed to a loss
of translational and rigid-body rotational entropies.
The drastic variation in thermodynamic values obtained for
P(OPh)3 relative to the other P ligands warrants further discussion.
The other ligands in this study have rigid aryl groups connected di-
rectly to the P atom, whereas the phenyl groups in P(OPh)3 are
linked to the P atom by O atoms, which allows for more rotational
freedom when the ligand is unbound. All ligands will lose transla-
tional and rigid-body rotational entropy upon binding, but P(OPh)3
would seem to suffer the largest penalty. The enthalpies of dissolu-
tion for both ligands, PPh3 (26.21 0.04 kJ/mol) and P(OPh)3
(6.86 0.01 kJ/mol), are endothermic. In both cases, there are
intermolecular forces that must be overcome in order for dissolu-
tion to occur. The enthalpy of solution for PPh3 is larger in magni-
tude because its lattice energy must be overcome in order for it to
dissolve (it is a solid) while P(OPh)3 is a liquid and has no such en-
ergy. If the enthalpy of fusion for PPh3 (19.69 kJ/mol) [52] is taken
into account as a crude approximation of its lattice energy, then
dissolution of a hypothetical PPh3 liquid would be the difference
between the enthalpy of dissolution and the enthalpy of fusion,
which is 6.52 kJ/mol, slightly smaller in magnitude than the en-
thalpy of solution of P(OPh)3, which is 6.86 kJ/mol. Since the en-
thalpy of dissolution for P(OPh)3 is positive, overcoming its own
intermolecular forces and breaking up solvent–solvent intermolec-
ular forces are larger in combined magnitude than solute–solvent
interactions. The inability of P(OPh)3 to overcome these forces
which, along with its enormous rotational entropy penalty, ex-
plains why P(OPh)3 has such a large, negative entropy of binding
[53]. Searle and Williams estimate an entropy loss of 8.8–44.8 kJ/
Acknowledgments
The work was supported by The Pennsylvania State University
and The Penn State Institutes of Energy and Environment (PSIEE)
through start-up funds provided to RMR and a 3M Non-Tenured
Faculty Grant. K.G. acknowledges support from the NASA WISER
program for undergraduate research.
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.jcat.2013.02.020.
References
[1] C.A. Tolman, Chem. Rev. 77 (1977) 313–348.
[2] R.S. Drago, Coord. Chem. Rev. 33 (1980) 251–277.
[3] O. Kuhl, Coord. Chem. Rev. 249 (2005) 693–704.
[4] M.R. Wilson, D.C. Woska, A. Prock, W.P. Giering, Organometallics 12 (1993)
1742–1752.
[5] G.C. Vogel, R.S. Drago, J. Chem. Educ. 73 (1996) 701–707.
[6] V.M. Nosova, Y.A. Ustynyuk, L.G. Bruk, O.N. Temkin, A.V. Kisin, P.A.
Storozhenko, Inorg. Chem. 50 (2011) 9300–9310.
[7] J.A. Osborn, F.H. Jardine, J.F. Young, G. Wilkinson, J. Chem. Soc. A (1966) 1711–
1732.
[8] R.F. Heck, J.P. Nolley Jr., J. Org. Chem. 37 (1972) 2320–2322.
[9] A. Suzuki, J. Organomet. Chem. 576 (1999) 147–168.
[10] K. Sonogashira, Y. Tohda, N. Hagihara, Tetrahedron Lett. (1975) 4467–4470.
[11] R. Jira, Angew. Chem. Int. Ed. 48 (2009) 9034–9037.
[12] J.A. Keith, P.M. Henry, Angew. Chem. Int. Ed. 48 (2009) 9038–9049.
[13] K.S. Yoo, C.H. Yoon, K.W. Jung, J. Am. Chem. Soc. 128 (2006) 16384–16393.
[14] C. Martinez, R. Alvarez, J.M. Aurrecoechea, Org. Lett. 11 (2009) 1083–1086.
[15] T.J. Marks, M.R. Gagne, S.P. Nolan, L.E. Schock, A.M. Seyam, D. Stern, Pure Appl.
Chem. 61 (1989) 1665–1672.
mol (for TDS) at 25 °C for the binding of small molecules to pro-
teins in solution due to lost rigid-body entropy, noting that the lar-
ger losses in entropy are associated with larger increases in
enthalpy, that is, enthalpy–entropy compensation [54]. The desolv-
[16] T. Wiseman, S. Williston, J.F. Brandts, L.N. Lin, Anal. Biochem. 179 (1989) 131–
137.
ation of the p-acidic P(OPh)3 upon binding would decrease entropy
[17] E. Freire, O.L. Mayorga, M. Straume, Anal. Chem. 62 (1990) A950–A959.
[18] M.J. Cliff, J.E. Ladbury, J. Mol. Recogn. 16 (2003) 383–391.
[19] M.J. Cliff, A. Gutierrez, J.E. Ladbury, J. Mol. Recogn. 17 (2004) 513–523.
[20] A. Ababou, J.E. Ladbury, J. Mol. Recogn. 19 (2006) 79–89.
[21] A. Ababou, J.E. Ladbury, J. Mol. Recogn. 20 (2007) 4–14.
[22] O. Okhrimenko, M. Jelesarov, J. Mol. Recogn. 21 (2008) 1–19.
[23] S. Bjelic, I. Jelesarov, J. Mol. Recogn. 21 (2008) 289–311.
[24] R.J. Falconer, A. Penkova, I. Jelesarov, B.M. Collins, J. Mol. Recogn. 23 (2010)
395–413.
[25] R.J. Falconer, B.M. Collins, J. Mol. Recogn. 24 (2010) 1–16.
[26] E. Freire, A. Schon, A. Velazquez-Campoy, in: M.L. Johnson, J.M. Holt, G.K.
Ackers (Eds.), Methods in Enzymology: Biothermodynamics, vol. 455, Part A,
Elsevier Academic Press Inc., San Diego, 2009, pp. 127–155.
[27] G.M.K. Poon, Anal. Biochem. 400 (2010) 229–236.
[28] B.B. Wayland, R.F. Schramm, Inorg. Chem. 8 (1969) 971–976.
[29] K. Nakamoto, Infrared and Raman Spectra of Inorganic and Coordination
Compounds, Wiley; Wiley-Blackwell, distributor, Hoboken, NJ, 2009.
[30] K. Hirose, J. Inclusion Phenom. Macrocyclic Chem. 39 (2001) 193–209.
[31] S.O. Grim, R.L. Keiter, Inorg. Chim. Acta 4 (1970) 56–60.
[32] C.M. Hettrick, W.J. Scott, J. Am. Chem. Soc. 113 (1991) 4903–4910.
[33] T.J. Colacot, W.A. Carole, B.A. Neide, A. Harad, Organometallics 27 (2008)
5605–5611.
due to the nature of MeCN being a good electron acceptor.
5. Conclusions
We have provided detailed analysis of the thermodynamics of
the binding of P ligands to 1 and the effect of different solvents
on the thermodynamics of binding to PPh3 to 2. We observed via
31P NMR spectroscopy that one or two equivalents of ligand bind
to the Pd centers. For bulkier ligands and poor electron-donating
ligands, only a single ligand was able to bind to 1. Our ITC experi-
ments demonstrated that ligands interact with the Pd center in
two binding modes: either two equivalents of ligands bind to sites
of different affinity on the same Pd center or only one equivalent of
ligand binds. For the bis-ligated complexes, the addition of the sec-
ond ligand proved to be more stable when forming a cis complex
when compared to the trans complexes. The electronic influences
of the different solvents also affected the binding ability of PPh3.
MeCN was the most easily displaced due to its weak-coordinating
ability coupled with its strong electron-accepting ability. The
strong electron-donating ability and RDE of py restricted binding
to a single PPh3 ligand while two equivalents of PPh3 were able
to bind in the other three solvents (MeCN, DMSO, and DMF).
[34] R.H. Crabtree, The Organometallic Chemistry of the Transition Metals, Wiley,
Hoboken, NJ, 2009. xiii, 505p.
[35] D.A. Redfield, J.H. Nelson, Inorg. Chem. 12 (1973) 15–19.
[36] A.L. Fernandez, C. Reyes, A. Prock, W.P. Giering, Perkin 2 (2000) 1033–1041.
[37] A.L. Fernandez, M.R. Wilson, A. Prock, W.P. Giering, Organometallics 20 (2001)
3429–3435.
[38] B.B. Wayland, R.F. Schramm, Chem. Commun. (1968) 1465–1466.
[39] S.P. Gupte, R.V. Chaudhari, J. Mol. Catal. 24 (1984) 197–210.

E.G. Moschetta et al. / Journal of Catalysis 302 (2013) 1–9
9
[40] C.Y. Liao, H.M. Lee, Acta Crystallogr. Sec. E: Struct. Rep. Online E62 (2006)
m680–m681.
[47] W. Partenheimer, B. Durham, J. Am. Chem. Soc. 96 (1974) 3800–3805.
[48] W. Partenheimer, T.D. Epley, R.S. Drago, J. Am. Chem. Soc. 90 (1968) 3886–
3888.
[49] M.C. Chervenak, E.J. Toone, J. Am. Chem. Soc. 116 (1994) 10533–10539.
[50] E. Grunwald, C. Steel, J. Am. Chem. Soc. 117 (1995) 5687–5692.
[51] N. Fey, B.M. Ridgway, J. Jover, C.L. McMullin, J.N. Harvey, Dalton Trans. 40
(2011) 11184–11191.
[41] Y. Marcus, J. Solution Chem. 13 (1984) 599–624.
[42] W. Partenheimer, E.F. Hoy, Inorg. Chem. 12 (1973) 2805–2809.
[43] E.G. Moschetta, K.M. Gans, R.M. Rioux, Personal communication.
[44] R.W. Taft, N.J. Pienta, M.J. Kamlet, E.M. Arnett, J. Org. Chem. 46 (1981) 661–
667.
[45] I. Jelesarov, H.R. Bosshard, J. Mol. Recogn. 12 (1999) 3–18.
[46] A. Bronowska, in: J.C. Moreno-Pirajan (Ed.), Thermodynamics – Interaction
Studies – Solids, Liquids, and Gases, InTech, 2011.
[52] D.R. Kirklin, E.S. Domalski, J. Chem. Thermodyn. 20 (1988) 743–754.
[53] C.W. Murray, M.L. Verdonk, J. Comput.-Aided Mol. Des. 16 (2002) 741–753.
[54] M.S. Searle, D.H. Williams, J. Am. Chem. Soc. 114 (1992) 10690–10697.