Thermodynamics of halogen bonding in solution: Substituent, structural, and solvent effects

Article abstract of DOI:10.1021/ja9086352

A detailed study of the thermodynamics of the halogen-bonding interaction in organic solution is presented. ¹⁹F NMR titrations are used to determine association constants for the interactions of a variety of Lewis bases with fluorinated iodoalkanes and iodoarenes. Linear free energy relationships for the halogen bond donor ability of substituted iodoperfluoroarenes XC ₆F₄I are described, demonstrating that both substituent constants (σ) and calculated molecular electrostatic potential surfaces are useful for constructing such relationships. An electrostatic model is, however, limited in its ability to provide correlation with a more comprehensive data set in which both halogen bond donor and acceptor abilities are varied: the ability of computationally derived binding energies to accurately model such data is elucidated. Solvent effects also reveal limitations of a purely electrostatic depiction of halogen bonding and point to important differences between halogen bonding and hydrogen bonding.

Full text of DOI:10.1021/ja9086352

Published on Web 01/15/2010
Thermodynamics of Halogen Bonding in Solution:
Substituent, Structural, and Solvent Effects
Mohammed G. Sarwar, Bojan Dragisic, Lee J. Salsberg, Christina Gouliaras, and
Mark S. Taylor*
Department of Chemistry, Lash Miller Chemical Laboratories, UniVersity of Toronto, Toronto,
Ontario M5S 3H6, Canada
Received October 10, 2009; E-mail: mtaylor@chem.utoronto.ca
Abstract: A detailed study of the thermodynamics of the halogen-bonding interaction in organic solution is
presented. 19F NMR titrations are used to determine association constants for the interactions of a variety
of Lewis bases with fluorinated iodoalkanes and iodoarenes. Linear free energy relationships for the halogen
bond donor ability of substituted iodoperfluoroarenes XC6F4I are described, demonstrating that both
substituent constants (σ) and calculated molecular electrostatic potential surfaces are useful for constructing
such relationships. An electrostatic model is, however, limited in its ability to provide correlation with a
more comprehensive data set in which both halogen bond donor and acceptor abilities are varied: the
ability of computationally derived binding energies to accurately model such data is elucidated. Solvent
effects also reveal limitations of a purely electrostatic depiction of halogen bonding and point to important
differences between halogen bonding and hydrogen bonding.
Introduction
able promise in medicinal chemistry, represents another sig-
nificant development.4
Halogen bonding (XB), the interaction between electron-
deficient halogen compounds and electron donors, has emerged
in recent years as a powerful and broadly useful noncovalent
force relevant to such diverse fields as medicinal chemistry and
organic materials (Figure 1).1 Although the first systematic
studies of halogen-bonding-based self-assembly were carried
out decades ago,2 it is only in recent years that chemists have
begun to understand and exploit the full potential of this
interaction. Metrangolo, Resnati, and co-workers have demon-
strated convincingly that XB is a general strategy for crystal
engineering and for the noncovalent assembly of new materials,
sparking new levels of interest in this interaction. Halogen
bonding has been employed for a wide range of applications,
including liquid crystalline and magnetic materials, chiral
discrimination, ion pair recognition, supramolecular polymer
formation, porous material design, and chemical separation.3
The recognition that halogen-bonding interactions are surpris-
ingly widespread in biological systems, and may hold consider-
Research reported over the past several years has provided
considerable insight into the fundamental structural and energetic
properties of the halogen-bonding interaction. The detailed
investigations by Legon and co-workers of XB in the gas phase
using rotational spectroscopy have provided a wealth of data,
including geometries, directional preferences, and trends in
binding energy.5 Numerous theoretical studies of XB have been
undertaken,6 and it is generally accepted that the halogen-
bonding interaction results from the “σ-hole”,6j a site of electron
deficiency created when polarizable halogen atoms are bound
to electronegative groups (Figure 1). Calculations suggest that
(4) (a) Auffinger, P.; Hays, F. A.; Westhof, E.; Ho, P. S. Proc. Natl. Acad.
Sci. U.S.A. 2004, 101, 16789–16794. (b) Trogdon, G.; Murray, J. S.;
Concha, M. C.; Politzer, P. J. Mol. Model. 2007, 13, 313–318. (c)
Voth, A. R.; Hays, F. A.; Ho, P. S. Proc. Natl. Acad. Sci. U.S.A. 2007,
104, 6188–6193. (d) Voth, A. R.; Khuu, P.; Oishi, K.; Ho, P. S. Nat.
Chem. 2009, 1, 74–79.
(1) For reviews, see: (a) Metrangolo, P.; Meyer, F.; Pilati, T.; Resnati,
G.; Terraneo, G. Angew. Chem., Int. Ed. 2008, 47, 6114–6127. (b)
Halogen Bonding: Fundamentals and Applications; Metrangolo, P.,
Resnati, G., Eds.; Springer: Berlin, 2008. (c) Risannen, K. Cryst. Eng.
Commun. 2008, 10, 1107–1113. (d) Politzer, P.; Lane, P.; Concha,
M. C.; Ma, Y.; Murray, J. S. J. Mol. Model. 2007, 13, 305–311. (e)
Metrangolo, P.; Neukirch, H.; Pilati, T.; Resnati, G. Acc. Chem. Res.
2005, 38, 386–395.
(5) Legon, A. C. Angew. Chem., Int. Ed. 1999, 38, 2686–2714.
(6) For a review, see: (a) Karpfen, A. In ref 1b, pp 1-15. (b) Lommerse,
J. P. M.; Stone, A. J.; Taylor, R.; Allen, F. H. J. Am. Chem. Soc.
1996, 118, 3108–3116. (c) Alkorta, I.; Rozas, J.; Elguero, J. J. Phys.
Chem. A 1998, 102, 9278–9285. (d) Valerio, G.; Raos, G.; Meille,
S. V.; Metrangolo, P.; Resnati, G. J. Phys. Chem. A 2000, 104, 1617–
1620. (e) Karpfen, A. J. Phys. Chem. A 2000, 104, 6871–6879. (f)
Wang, W. Z.; Wong, N.-B.; Zheng, W. X.; Tian, A. M. J. Phys. Chem.
A 2004, 108, 1799–1805. (g) Zou, J. W.; Jian, Y. J.; Guo, M.; Hu,
G. X.; Zhang, B.; Liu, H. C.; Yu, Q. S. Chem.sEur. J. 2005, 11,
740–751. (h) Wang, Y. H.; Zou, J. W.; Lu, Y. X.; Yu, Q. S. J. Theor.
Comput. Chem. 2006, 4, 719–721. (i) Riley, K. E.; Merz, K. M. J.
Phys. Chem. A 2007, 111, 1688–1694. (j) Clark, T.; Hennemann, M.;
Murray, J. S.; Politzer, P. J. Mol. Model. 2007, 13, 291–296. (k) Lu,
Y.-X.; Zou, J.-W.; Wang, Y.-H.; Jiang, Y.-J.; Yu, Q.-S. J. Phys. Chem.
A 2007, 111, 10781–10788. (l) Cavollotti, C.; Metrangolo, P.; Meyer,
F.; Recupero, F.; Resnati, G. J. Phys. Chem. A 2008, 112, 9911–9918.
(2) (a) Bent, H. A. Chem. ReV. 1968, 68, 587–648. (b) Hassel, O. Science
1970, 170, 497–502. (c) Dumas, J.-M.; Peurichard, H.; Gomel, M. J.
Chem. Res. 1978, 54–57.
(3) For representative examples, see ref 1a and the following references: (a)
Sun, A.; Goroff, N. S.; Lauher, J. W. Science 2006, 312, 1030–1034.
(b) Nguyen, H. L.; Horton, P. N.; Hursthouse, M. B.; Legon, A. C.;
Bruce, D. W. J. Am. Chem. Soc. 2004, 125, 16–17. (c) Metrangolo,
P.; Carcenac, Y.; Lahtinen, M.; Pilati, T.; Rissanen, K.; Vij, A.;
Resnati, G. Science 2009, 323, 1461–1464.
9
1646 J. AM. CHEM. SOC. 2010, 132, 1646–1653
10.1021/ja9086352  2010 American Chemical Society

Halogen Bonding: Structural and Solvent Effects
A R T I C L E S
Figure 1. Representative applications of halogen bonding and a depiction of the molecular electrostatic potential of C6F5I demonstrating the “σ-hole” at
iodine (B3LYP/6-31+G**-LANLdp, Gaussian 03 (see the text for details); blue indicates a region of positive electrostatic potential).
the origin of the halogen-bonding interaction (electrostatic vs
charge transfer) may depend on the structure of the components
involved.6k
Despite the progress achieved through the studies mentioned
above, our understanding of XB remains limited in important
respects. In particular, quantitative descriptions of halogen
bonding in solution lag far behind those available for other
noncovalent forces such as hydrogen bonding7 and interactions
of π systems.8 Only a handful of association constants for
halogen-bonding interactions in the solution phase have been
determined.9-12 Data of this type are essential for developing
and evaluating metrics for predicting the strength of halogen-
bonding interactions. As a first step toward our long-term goal
of applying halogen bonding for the development of new
molecular receptors and catalysts, we have undertaken a
systematic study of structural effects on the energies of halogen-
bonding interactions. The goals of this study are (1) to determine
the binding energies of halogen bond donor-acceptor pairs and
(2) to evaluate the ability of empirically derived parameters (for
example, substituent constants or solvent parameters) and
computationally derived quantities to model trends in halogen
Figure 2. Structures of the substituted iodoperfluoroarenes employed in
the substituent effect study.
bond energies, with the aim of identifying quantities useful for
predicting halogen-bonding affinities in solution. In addition to
being important to our own research efforts, information of this
type will be of value to emerging applications of halogen
bonding in medicinal chemistry and materials design.
Results and Discussion
Experimental Determinations of Ka Values for Halogen-
Bonding Interactions. Perfluorinated iodoalkanes and -arenes
have been identified as among the strongest donors of halogen
bonds (Figure 1). 19F-NMR-based methods are ideally suited
for probing halogen-bonding interactions of these species due
to the high signal dispersion and sensitivity of 19F NMR
(7) (a) Laurence, C.; Berthelot, M. Perspect. Drug DiscoVery Des. 2000,
18, 39–60. (b) Abraham, M. H.; Platts, J. A. J. Org. Chem. 2001, 66,
3484–3491. (c) Cook, J. L.; Hunter, C. A.; Low, C. M. R.; Perez-
Velasco, A.; Vinter, J. G. Angew. Chem., Int. Ed. 2007, 46, 3706–
3709.
(8) Meyer, E. A.; Castellano, R. K.; Diederich, F. Angew. Chem., Int.
Ed. 2003, 42, 1210–1250.
techniques. Metrangolo, Resnati, and co-workers have used 19
F
(9) XB of haloalkynes: (a) Laurence, C.; Queignec-Cabanetos, M.;
Dziembowska, T.; Queignec, R.; Wojtkowiak, B. J. Chem. Soc. Perkin
Trans. 2 1982, 1605–1610. (b) Laurence, C.; Queignec-Cabanetos,
M.; Wojtkowiak, B. Can. J. Chem. 1983, 61, 135–138.
NMR as the basis for a semiquantitative scale of halogen bond
acceptor ability: the difference in chemical shift of the CF2I
group in a basic solvent of interest relative to that in a nonbasic
(hydrocarbon) solvent is a measure of the acceptor ability of
the solvent.13 Using 19F NMR titrations, XB association
constants for a number of donor-acceptor pairs have been
determined by the Milan group11c and by the groups of
Brammer10b and Hunter.11e Results from this method are in
agreement with data obtained by other means, including infrared
and Raman spectroscopy, as well as calorimetric measurements.
Likewise, we have found that this technique provides accurate
(10) XB of haloarenes: (a) Wash, P. L.; Ma, S.; Obst, U.; Rebek, J. J. Am.
Chem. Soc. 1999, 121, 7973–7974. (b) Libri, S.; Jasim, N. A.; Perutz,
R. N.; Brammer, L. J. Am. Chem. Soc. 2008, 130, 7842–7844.
(11) XB of haloalkanes: (a) Larsen, D. W.; Allred, A. L. J. Phys. Chem.
1965, 69, 2400–2401. (b) Blackstock, S. C.; Lorand, J. P.; Kochi, J. K.
J. Org. Chem. 1987, 52, 1451–1460. (c) Metrangolo, P.; Panzeri, W.;
Recupero, F.; Resnati, G. J. Fluorine Chem. 2002, 114, 27–33. (d)
Rosokha, S. V.; Neretin, I. S.; Rosokha, T. Y.; Hecht, J.; Kochi, J. K.
Heteroat. Chem. 2006, 17, 449–459. (e) Cabot, R.; Hunter, C. Chem.
Commun. 2009, 2005–2007.
(12) In other studies, the relative acceptor abilities of bases or donor abilities
of iodo compounds have been quantified by NMR or IR spectral
changes, without measurement of association constants. See, for
example: (a) Webb, J. A.; Klijn, J. E.; Hill, P. A.; Bennett, J. L.; Goroff,
N. S. J. Org. Chem. 2004, 69, 660–664. (b) Laurence, C.; Queignec-
Cabanetos, M.; Dziembowska, T.; Queignec, R.; Wojtkowiak, B.
J. Am. Chem. Soc. 1981, 103, 2567–2573.
(13) (a) Messina, M. T.; Metrangolo, P.; Panzeri, W.; Ragg, E.; Resnati,
G. Tetrahedron Lett. 1998, 39, 9069–9072. (b) Lunghi, A.; Cardillo,
P.; Mesina, M. T.; Metrangolo, P.; Panzeri, W.; Resnati, G. J. Fluorine
Chem. 1998, 91, 191–194.
9
J. AM. CHEM. SOC. VOL. 132, NO. 5, 2010 1647

A R T I C L E S
Sarwar et al.
Table 1. Halogen-Bonding Association Constants of para-Substituted Iodotetrafluorobenzenes (Tri-n-butylphosphine Oxide Acceptor,
Cyclohexane Solvent), with Substituent Constants (σpara and σmeta) and Computed Values of the Molecular Electrostatic Potential Surface at
the Iodine Atom
b
b
substituent
Kaa (M-1
)
log Ka
σpara
σmeta
surface potential (AM1)d (kcal/mol)
surface potential (DFT)e (kcal/mol)
F
12 ( 2.5
5.6 ( 1.1
3.9 ( 0.8
3.3 ( 0.7
1.3 ( 0.3
1.0
0.06
0.16
-0.01
-0.27
-0.83c
0.34
0.14
0.06
24.2
20.8
20.6
20.7
20.0
25.8
22.9
22.0
22.5
18.9
Ct CPh
Ph
OCH3
N(CH2)5
0.75
0.59
0.52
0.11
0.12
-0.15c
a Association constant Ka with tri-n-butylphosphine oxide in cyclohexane at 298 K determined by curve-fitting of 19F NMR titration data to a 1:1
binding isotherm. See the Supporting Information for full details. b Values of σ from Hansch, C.; Leo, A.; Unger, S. H.; Kim, K. H.; Nikaitani, D.;
c
Lien, E. J. J. Med. Chem. 1973, 16, 1207-1216. σpara and σmeta values for the dimethylamino substituent are listed. d Maximum value of the molecular
electrostatic potential surface at the iodine atom, calculated at the AM1 level using Spartan 06. e Maximum value of the molecular electrostatic potential
surface at the iodine atom, calculated at the DFT/B3LYP level of theory with the 6-31+G**-LANLdp basis set using Gaussian 03. See the text and
Supporting Information for details.
and reproducible Ka determinations for halogen-bonding interac-
tions of fluorinated aromatic and aliphatic organoiodides.
Titrations were carried out by recording 19F NMR spectra of
the halogen bond donor in the presence of increasing concentra-
tions of acceptor: the signal corresponding to the fluorine atoms
adjacent to the iodo substituent underwent a dramatic downfield
shift upon increasing acceptor concentration, diagnostic of the
halogen-bonding interaction.11c,e The changes in chemical shift
as a function of acceptor concentration were well-modeled by
1:1 binding isotherms using standard curve-fitting methods. All
experiments described were carried out in duplicate or in
triplicate and are reported with an estimated error of (20% to
reflect the weak nature of the interactions being studied.
Figure 3. Molecular electrostatic potential surfaces (B3LYP/6-31+G**-
LANLdp, Gaussian 03) of iodopentafluorobenzenes 1a (left) and 1b (right).
Substituent Effects on the Halogen Bond Donor Ability of
Iodoperfluorobenzenes. We began our investigation of substitu-
ent effects on XB by studying the effect of para substitution
on the halogen bond donor ability of iodotetrafluorobenzenes
(Figure 2). Linear free energy relationships based upon sub-
stituent effects represent a powerful tool for the study of
noncovalent interactions. Insight into hydrogen-bonding,14 cat-
ion-π,15 arene edge-face,16 and other interactions17 has been
obtained through investigations of substituent effects. While
Laurence and co-workers have shown that the strengths of
halogen bonds involving substituted iodoacetylenes XCt CI are
linearly correlated with the Hammett substituent parameter σ,9b
substituent effects on XB of haloarenes have not been reported
to date.
Red indicates negative charge density and blue positive charge density: the
plots have been set to the same color scale so that a visual comparison can
be made. Maximum values of the surface electrostatic potential at the I
atom are indicated beneath each structure.
alcohols or amines.18 Despite these widespread applications, the
question of the electronic effect of a para oxygen or nitrogen
substituent on the halogen bond donor ability of the C6F4I group
has not been addressed.
Piperidine- and methoxy-substituted iodotetrafluorobenzenes
1b and 1c were prepared from iodopentafluorobenzene by
nucleophilic aromatic substitution according to literature pro-
tocols.18 Halogen bond donors 1d19 and 1e were prepared from
1,4-diiodotetrafluorobenzene by palladium-catalyzed Sonogash-
ira and Suzuki reactions, respectively. The association constants
for the XC6F4I-tri-n-butylphosphine oxide (Bu3PO) halogen
bonds were determined in cyclohexane by 19F NMR as described
above and are assembled in Table 1. As expected, the presence
of electron-donating substituents (1b, 1c) resulted in a lower
halogen bond donor ability: the Ka value of 1a is almost 10
times that of amine-substituted 1b. Electrostatic potential
surfaces of 1a and 1b show that the σ-hole, the site of electron
deficiency at iodine where halogen bonding occurs, is consider-
ably less pronounced upon introduction of the piperidine
substituent (Figure 3).
We undertook measurements of the strengths of the halogen
bonds of iodotetrafluorobenzenes 1a-1e with a common
acceptor, with the goal of evaluating the value of established
substituent constants and computationally obtained quantities
as predictors of XB donor ability. It should be noted that para-
substituted iodotetrafluorobenzenes have been applied exten-
sively as components of complex building blocks for crystal
engineering, largely because of their ease of synthesis by
nucleophilic aromatic substitution of iodopentafluorobenzene by
(14) (a) Wilcox, C. R.; Kim, E.; Romano, D.; Kuo, L. H.; Burt, A. L.;
Curran, D. P. Tetrahedron 1995, 51, 621–634. (b) Deans, R.; Cooke,
G.; Rotello, V. M. J. Org. Chem. 1997, 62, 836–839. (c) Deans, R.;
Cuello, A. O.; Galow, T. H.; Ober, M.; Rotello, V. M. J. Chem. Soc.,
Perkin Trans. 2 2000, 1309–1313.
A plot of log Ka vs the aromatic substituent constant σ shows
a poor degree of correlation (F ) 0.74, r2 ) 0.82, Figure 4a).
The σmeta constant frequently correlates well with phenomena
that respond primarily to inductive/field effects,20 and indeed
(15) Zhong, W.; Gallivan, J. P.; Zhang, Y.; Li, L.; Lester, H. A.; Dougherty,
D. A. Proc. Natl. Acad. Sci. U.S.A. 1998, 95, 12088–12093.
(16) (a) Carver, F. J.; Hunter, C. A.; Livingstone, D. J.; McCabe, J. F.;
Seward, E. M. Chem.sEur. J. 2002, 8, 2847–2859. (b) Kim, E.;
Paliwal, S.; Wilcox, C. S. J. Am. Chem. Soc. 1998, 120, 11192–11193.
(17) Hof, F.; Scofield, D. M.; Schweizer, W. B.; Diederich, F. Angew.
Chem., Int. Ed. 2004, 43, 5056–5059.
(18) Guardigli, C.; Liantonio, R.; Mele, M. L.; Metrangolo, P.; Resnati,
G.; Pilati, T. Supramol. Chem. 2003, 15, 177–188.
(19) Collings, J. C.; Burke, J. M.; Smith, P. S.; Batsanov, A. S.; Howard,
J. A. K.; Marder, T. B. Org. Biomol. Chem. 2004, 2, 3172–3178.
(20) Reynolds, W. F. J. Chem. Soc., Perkin Trans. 1 1980, 985–992.
9
1648 J. AM. CHEM. SOC. VOL. 132, NO. 5, 2010

Halogen Bonding: Structural and Solvent Effects
A R T I C L E S
Figure 4. Correlation of the 4-X-C6F4I-Bu3PO halogen bond strength log
Ka with the (a) σpara and (b) σmeta substituent constants of X.
Figure 5. Correlation of the 4-X-C6F4I-Bu3PO halogen bond strength log
Ka with the electrostatic potential at the iodine atom, calculated with (a)
AM1 and (b) DFT (B3LYP/6-31+G**-LANLdp) computational methods
(see the text and Supporting Information).
the plot of log Ka vs σmeta shows a measurably higher correlation
coefficient (F ) 1.8, r2 ) 0.94, Figure 4b) than that of log Ka
vs σpara. These data are consistent with the idea that the effect
of substituents on this halogen-bonding interaction is predomi-
nantly electrostatic in origin.
electrostatic potential at the bromine atom.25 However, experi-
ments permitting quantitative correlations of calculated elec-
trostatic potentials of halocarbons with measured XB thermo-
dynamics have not been reported. Molecular electrostatic
potential surfaces were calculated for 1a-1e at the AM1 level
of theory with Spartan 06, and the maximum values at the iodine
atom are assembled in Table 1. These correlate poorly with the
experimental binding energies (r2 ) 0.73, Figure 5a). Speculat-
ing that this poor degree of correlation might be a result of
limitations of the ability of the AM1 semiempirical method to
describe the electrostatic potential of iodoperfluoroarenes, we
calculated the same quantity at a higher level of theory. Density
functional theory (DFT) calculations were carried out with the
B3LYP functional, using the 6-31+G** basis set for all atoms
except iodine: the LANL2DZ effective core potential,26 aug-
mented with polarization functions of d symmetry and diffuse
functions of p symmetry,27 was employed for iodine. We denote
this basis set 6-31+G**-LANLdp. Values of the electrostatic
potential at iodine calculated at this level of theory provide
excellent correlation with the experimental halogen bond donor
abilities (r2 ) 0.97, Figure 5b).
In search of more general predictors of trends in halogen-
bonding ability, we turned to computational methods. It has been
shown that the molecular electrostatic potential at a hydrogen
atom of interest may be correlated with hydrogen bond donor
ability14a,21,22 and that a similar calculation of potential at the
centroid of an aryl group is a reliable measure of cation-π
binding strength.23 Given the low cost of such calculations, a
similar relationship for halogen bonding could be valuable in
the context of medicinal chemistry or in the design of new
halogen bond donors. The relationship between molecular
electrostatic potentials and the XB phenomenon has been
discussed in some detail,4a,6j,24 and a recent computational study
of acetone-bromoarene complexes revealed a linear relationship
between calculated halogen-bonding energies and calculated
(21) Hagelin, H.; Murray, J. S.; Brinck, T.; Berthelot, M.; Politzer, P. Can.
J. Chem. 1995, 73, 483–488.
(22) (a) Deans, R.; Cooke, G.; Rotello, V. M. J. Org. Chem. 1997, 62,
836–839. (b) Deans, R.; Cuello, A. O.; Galow, T. H.; Ober, M.;
Rotello, V. M. J. Chem. Soc., Perkin Trans. 2 2000, 1309–1313.
(23) Mecozzi, S.; West, A. P., Jr.; Dougherty, D. A. Proc. Natl. Acad. Sci.
U.S.A. 1996, 93, 10566–10571.
(25) Riley, K. E.; Murray, J. S.; Politzer, P.; Concha, M. C.; Hobza, P.
J. Chem. Theory Comput. 2009, 5, 155–163.
(26) Hay, P. J.; Wadt, W. R. J. Chem. Phys. 1985, 82, 270–283.
(27) Check, C. E.; Faust, T. O.; Bailey, J. M.; Wright, B. J.; Gilbert, T. M.;
Sunderlin, L. S. J. Phys. Chem. A 2001, 105, 8111–8116.
(24) Zordan, F.; Brammer, L.; Sherwood, P. J. Am. Chem. Soc. 2005, 127,
5979–5989.
9
J. AM. CHEM. SOC. VOL. 132, NO. 5, 2010 1649

A R T I C L E S
Sarwar et al.
Table 2. Experimental and Predicted Binding Energies of C6F5I
and C8F17I with Representative Halogen Bond Acceptors
halogen bonda
Kab (M-1
)
∆Gc (kcal/mol)
predicted ∆Gd (kcal/mol)
C6F5I-Et3N
1.3 ( 0.2
2.0 ( 0.4
2.8 ( 0.6
6.2 ( 1.2
12 ( 2.5
18 ( 4
-0.2 ( 0.1
-0.4 ( 0.1
-0.6 ( 0.1
-1.1 ( 0.1
-1.5 ( 0.1
-1.7 ( 0.1
-1.8 ( 0.1
-2.1 ( 0.1
0.3
0.1
C6F5I-Bu2SO
C8F17I-Et3N
C8F17I-Bu2SO
C6F5I-Bu3PO
C8F17I-Bu3PO
C6F5I-quinuclidine
C8F17I-quinuclidine
-0.7
-1.1
-0.2
-1.6
0
20 ( 4
34 ( 7
-1.2
a Halogen bond acceptors are abbreviated as follows: Et3N
)
triethylamine; Bu2SO ) di-n-butyl sulfoxide; Bu3PO ) tri-n-butyl-
phosphine oxide. b Association constant Ka in cyclohexane at 298 K,
determined by curve-fitting of 19F NMR titration data to a 1:1 binding
isotherm. c Free energy of binding calculated from the association
constant. d Free energy of binding predicted by the Hunter pairwise
electrostatic interaction model (see the text).
Figure 6. Molecular electrostatic potential surfaces (B3LYP/6-31+G**-
LANLdp, Gaussian 03) of (iodoethynyl)benzene (left) and 5-iodoracil (right).
Red indicates negative charge density and blue positive charge density: the
plots have been set to the same color scale as those shown in Figure 3 so
that a visual comparison can be made. Maximum values of the surface
electrostatic potential at the I atom are indicated beneath each structure.
The experimental data described above confirm that molecular
electrostatic potentials calculated at an appropriate level of
theory are of considerable value in predicting the relative
strengths of halogen-bonding interactions, in agreement with
the computational work of the Politzer group.6j We calculated
electrostatic potential surfaces for two compounds known to
form halogen bonds: (iodoethynyl)benzene, a representative
alkynyl iodide of the type studied by the groups of Laurence9
and Goroff,3a,12a and 5-iodoracil, a halogenated nucleobase
of the type that forms XB in biological systems (Figure 6).4a
At the level of theory described above, the maximum values of
the σ-hole for these compounds are 24.6 and 19.1 kcal/mol,
respectively. These fall within the range of values explored in
this substituent effect study and suggest that, to a first ap-
proximation, these structurally distinct organohalides may have
halogen bond donor ability similar to that of the compounds
studied in detail here.
Figure 7. Experimental binding data (-∆Gbinding) against values predicted
by the model of Hunter and co-workers. The dotted line represents y ) x.
predictions of noncovalent bond energies in solution, may be
useful as a predictive tool for halogen bonding. Indeed, the
Hunter group recently measured the strengths of the halogen
bonds of iodoperfluorohexane with a variety of Lewis bases in
three solvents and showed that their electrostatic model, as well
as other parameters including donor nucleophilicity, correlated
quite well with the experimental data.11e While the relative
contributions of electrostatic, charge-transfer, and dispersion
components to XB continue to be debated, computational studies
generally suggest that the charge-transfer and dispersion com-
ponents may not be negligible, particularly as the strengths of
the interactions are varied. It is thus unclear to what extent a
purely electrostatic model should be useful for predicting the
association constants of halogen bonds involving partners that
vary widely in structure.
To probe this issue, we took a distinct approach from that of
Hunter and co-workers by selecting two different halogen bond
donors (iodoperfluorooctane and iodopentafluorobenzene) and
measuring the strengths of their interactions with four electroni-
cally distinct electron donors (triethylamine, di-n-butyl sulfoxide,
tri-n-butylphosphine oxide, and quinuclidine) in cyclohexane.
The binding data are assembled in Table 2.
Structural Effects on Halogen Bonding: A Combined
Experimental and Computational Approach. The data discussed
above demonstrate that, within the series of substituted iodop-
erfluoroaromatics 1a-1e, the effect of substituents on halogen
bond donor ability is well-modeled by electrostatic effects. This
observation suggests that the model of Hunter and co-workers,28
which uses pairwise electrostatic effects as the basis for accurate
(28) For a review, see: Hunter, C. A. Angew. Chem., Int. Ed. 2004, 43,
5310–5324. In brief, the Hunter model treats noncovalent interactions
in solution as a competition between pairwise electrostatic interactions
of solute and solvent molecules. Solvents and solutes are characterized
by donor ability R and acceptor ability ꢀ. The values of R and ꢀ are
obtained either from experimental measurements of hydrogen-bonding
association constants or from molecular electrostatic potential calcula-
tions. The latter are generally carried out at the AM1 level, and the
resulting maximum and minimum values of the electrostatic potential
surface Emax and Emin are used to estimate R and ꢀ using the following
equations:
R ) Emax/(52 kJ/mol)
ꢀ ) Emin/(52 kJ/mol)
(1)
(2)
The energy of interaction of donor X and acceptor Y in solvent S is
estimated by the following equation
A plot of the eight experimentally determined free energies
of binding against those predicted by the Hunter model is shown
in Figure 7. The predicted values in cyclohexane (Table 2) were
obtained using parameters for hydrocarbon solvents (R ) 1.2
and ꢀ ) 0.6) developed recently by Hunter and co-workers.29
∆G ) -(RX - RS)(ꢀY - ꢀS) + 6 kJ/mol
(3)
where RX is the donor ability of X, RS is the donor ability of the
solvent, ꢀY is the acceptor ability of Y, and ꢀS is the acceptor ability
of the solvent. The cost of bringing two molecules together in solution
is estimated to be 6 kJ/mol.
9
1650 J. AM. CHEM. SOC. VOL. 132, NO. 5, 2010

Halogen Bonding: Structural and Solvent Effects
A R T I C L E S
Table 3. Computed Halogen-Bonding Energies for Comparison to
These parameters are based on determinations of a large set of
hydrogen-bonding association constants in nonpolar solvents.
While the magnitudes of the free energies of binding are
predicted quite successfully, the correlation between the pre-
dicted and experimental values is poor (r2 ) 0.3). Values of ꢀ
suggest that XB acceptor ability should decrease in the order
Bu3PO > Bu2SO ∼ quinuclidine > Et3N, whereas the experi-
mental acceptor abilities follow the order quinuclidine > Bu3PO
> Me2SO > Et3N. We note that assigning values of R to the
halogen bond donors is not straightforward: the predictions in
Table 2 are based on values derived from AM1 electrostatic
potential surfaces for consistency with Hunter’s approach (R
) 1.9 for C6F5I and 2.5 for C8F17I). However, our substituent
effect studies (see the previous section) indicate that charge
distributions in iodoperfluorocarbons are not accurately de-
scribed by this semiempirical method. DFT calculations suggest
values of R for C6F5I and C8F17I of 2.1 and 2.0, respectively.
These parameters provide poorer correlation still with experi-
mental data because they improperly order the donor abilities
of the two compounds.
Experimental Data From Table 2
b
c
halogen bonda
∆EB3LYP (kcal/mol)
∆ECP
(kcal/mol)
B3LYP
C6F5I-Et3N
C6F5I-Me2SO
C4F9I-Et3N
C4F9I-Me2SO
C6F5I-Me3PO
C4F9I-Me3PO
C6F5I-quinuclidine
C4F9I-quinuclidine
-2.0
-2.9
-2.5
-3.4
-3.6
-4.0
-4.4
-3.5
-1.3
-2.5
-2.1
-2.9
-3.1
-3.6
-3.8
-3.3
a Halogen bond acceptors are abbreviated as follows: Et3N
)
triethylamine; Me2SO ) dimethyl sulfoxide; Me3PO ) trimethylphosphine
oxide. b DFT-calculated (B3LYP/6-31+G**-LANLdp) binding energy (see
above and the Supporting Information). c DFT-calculated binding energy,
corrected for the basis set superposition error by the CP method (see the
text).
In search of approaches that would provide better predictive
power for the thermodynamics of halogen bonding, we turned
to gas-phase calculations of binding energies. Halogen bonding
has been the subject of many computational studies, and both
DFT and MP2 calculations (particularly the latter) have been
shown to model the experimental geometries and spectral
properties6l,30 of XB complexes with good accuracy. On the
other hand, systematic comparisons of computationally predicted
halogen-bonding energies with experimental data are lacking:
this is a reflection of the small set of experimental binding data
available to date. The accuracy with which computational
methods are able to describe noncovalent interactions is the
subject of much active research,31,32 and XB represents an
interesting test case because it may differ from hydrogen
bonding in terms of the importance of electrostatic, charge-
transfer, and dispersion contributions. Moreoever, identifying
calculated quantities with the predictive power of the thermo-
dynamics of XB has clear practical relevance.
For the purposes of time efficiency, truncated models of the
halogen-bonded complexes shown in Table 2 were used in the
computational studies: C8F17I was simplified to C4F9I, dibutyl
sulfoxide to dimethyl sulfoxide, and tributylphosphine oxide to
trimethylphosphine oxide. Geometry optimizations of the XB
complexes were first carried out using DFT (B3LYP), with the
6-31+G**-LANLdp basis described in the previous section. The
gas-phase binding energies (∆EB3LYP) of the complexes were
obtaining by subtracting the sum of the calculated energies of
the isolated components from the calculated energy of the
complex. Corrections for the basis set superposition error were
carried out by the counterpoise (CP) method of Boys and
Bernardi:33 counterpoise-corrected binding energies are denoted
∆ECPB3LYP. The calculated results are assembled in Table 3.34
Figure 8. Experimental binding data (-∆Gbinding, Table 2) against DFT-
calculated (-∆ECPB3LYP, Table 3) gas-phase binding energy.
LANLdp binding energies of the corresponding simplified
complexes (-∆ECPB3LYP, Table 3) is shown in Figure 8. Given
that limitations of the B3LYP functional have been identified
in terms of its ability to model energies and geometries of
noncovalent interactions in general,32 and halogen-bonding
interactions in particular,25 the agreement between calculated
and experimental binding data is perhaps surprisingly good.
The DFT calculations systematically predict binding energies
higher than those measured experimentally, but correspon-
dence between these two sets of data is not to be expected
given the fundamental differences between the computed gas-
phase energies of interaction (which were not corrected for
zero-point-energy differences here) and the free energy
changes measured in solution. Nonetheless, there is a
significant correlation between calculated and experimental
data (r2 ) 0.81). The interaction posing the most significant
problem for this level of theory is the strongest association
studied (the C8F19I-quinuclidine halogen bond): this point
lies significantly above the trendline shown in Figure 8.
Previous work by Lu and co-workers suggests that a covalent
contribution to halogen bonding becomes important as the
strengths of these interactions increase;6k similar models for
hydrogen bonding have been put forward by Gilli and co-
workers.35 We speculate that the strong quinuclidine-
iodoperfluoroalkane halogen bond may involve a contribution
from charge-transfer, dispersion, or covalent character that
A plot of the experimental free energies of binding (-∆Gbinding
,
Table 2) against the counterpoise-corrected B3LYP 6-31+G**-
(29) Cabot, R.; Hunter, C. A.; Varley, L. M. Org. Biomol. Chem., in press.
(30) (a) Rege, P. D.; Malkina, O. L.; Goroff, N. S. J. Am. Chem. Soc.
2002, 124, 370–371. (b) Glaser, R.; Chen, N.; Wu, H.; Knotts, N.;
Kaupp, M. J. Am. Chem. Soc. 2004, 126, 4412–4419.
(31) Mu¨ller-Dethlefs, K.; Hobza, P. Chem. ReV. 2000, 100, 143–167.
(32) Zhao, Y.; Truhlar, D. G. J. Chem. Theory Comput. 2005, 1, 415–432,
and references therein.
(33) Boys, S. F.; Bernardi, F. Mol. Phys. 1970, 19, 553.
(34) Zero-point-energy-corrected binding energies were also calculated and
provide poorer correlation with the experimental data. See the
Supporting Information.
(35) Gilli, P.; Bertolasi, V.; Ferretti, V.; Gilli, G. J. Am. Chem. Soc. 1994,
116, 909–915.
9
J. AM. CHEM. SOC. VOL. 132, NO. 5, 2010 1651

A R T I C L E S
Sarwar et al.
is not well-modeled by DFT with the B3LYP functional.32
Hunter and co-workers also noted the failure of their
electrostatic model to accurately predict the halogen-bonding
energies of strong acceptors such as quinuclidine.11e Higher
level calculations are under way to shed light on this point.
However, we note that, with the exception of the C8F19I-
quinuclidine pair, B3LYP/6-31+G**-LANLdp calculations
show useful degrees of correlation with experimental data and
may be useful in guiding the design of new halogen-bonding
motifs.
Solvent Effects on Halogen Bonding. The solvent dependence
of noncovalent interactions is of considerable interest: not only
are fundamental properties of an interaction revealed by its
medium dependence, but the information gathered from such
studies is also valuable as a predictive tool for applications in
self-assembly or medicinal chemistry. While solvent effects on
hydrogen bonding,7c,36 aromatic edge-face and face-face
interactions,37 and other noncovalent forces have been analyzed
in detail, studies of solvent effects on halogen bonding remain
limited in comparison. Laurence and co-workers noted that the
association between pyridine and 1-cyano-2-iodoacetylene is
more favorable in carbon tetrachloride than in benzene
solution,9b and the group of Kochi observed that Ka values for
Table 4. Association Constants for the
Iodoperfluorooctane-Triethylamine Complex in Various Solvents
as a Function of the Solvent Properties
solvent
Kaa (M-1
)
log Ka
ET(30)b (kcal/mol)
pyc
cyclohexane
benzene
acetonitrile
dichloromethane
acetone
tetrahydrofuran
dioxane
tert-butyl alcohol
chloroform
2-propanol
2.8 (0.6
2.6 ( 0.5
1.9 ( 0.4
1.8 ( 0.3
1.3 ( 0.3
1.2 ( 0.2
1.1 ( 0.2
0.7 ( 0.3
0.6 ( 0.4
0.3 ( 0.7
0.45
0.41
0.20
0.20
0.11
0.08
0.04
-0.15
-0.2
-0.4
31
34
46
41
42
37
36
43
35
48
0.58
1.05
1.79
1.35
1.64
1.35
1.5
0.92
1.25
1.09
a Association constant Ka for the Et3N-IC8F17 halogen bond (298 K)
obtained by curve-fitting 19F NMR titration data to a 1:1 binding
isotherm. b Values of ET(30) from ref 39. c Values of py from ref 40.
the CBr4-DABCO complex followed the trend CHCl3
>
CH3CN > CH3OH.11b These small data sets do not enable
correlation of halogen bond strengths with known metrics of
solvent properties. In their recent study, Hunter and co-workers
measured the strengths of halogen-bonding interactions of
iodoperfluorohexane with a variety of donors in three solvents,
chloroform, benzene, and carbon tetrachloride, and obtained data
that were largely consistent with their electrostatic model of
noncovalent interactions (see above).11e While this data set was
useful in evaluating this specific model, it represents a limited
selection of commonly employed solvents and does not allow
comparisons to other, empirically developed solvent scales in
current use. In particular, Lewis basic donor solvents were not
investigated.
We thus assembled a complementary set of data, mea-suring
the association constants for a single XB interaction (iodoper-
fluorooctane-triethylamine) in 10 diverse solvents. Association
constants in methanol and dimethyl sulfoxide were too low to be
determined by NMR, while those in chloroform and 2-propanol
represent the lower limits of the NMR technique. For the range of
solvents for which accurate determinations of Ka were possible,
the iodoperfluorooctane-triethylamine association constant showed
a detectable variation in magnitude (Table 4).
Figure 9. Experimental binding data (log Ka for the C8F17I-triethylamine
interaction) plotted against binding data predicted by the Hunter electrostatic
model. The dotted line represents y ) x.
in Figure 9.38 In agreement with previous results,11e the data
for chloroform and benzene lie on the line y ) x within
experimental error, as do those for cyclohexane and 2-pro-
panol. However, the situation is dramatically different in the
case of the Lewis basic solvents acetone, acetonitrile, and
tetrahydrofuran. In all three cases, the experimental binding
constants are significantly higher than the predicted values.
The triethylamine-iodoperfluorooctane interaction is more
faVorable in competitiVe, Lewis basic solVents than is
predicted by an electrostatic model. Whether this behavior
results from limitations of the description of the Et3N-IC8F17
halogen bond as a purely electrostatic interaction, the unusual
solubility properties of the perfluoroalkanes, or some other
phenomenon is not clear. Nonetheless, this observation may
be important in applications of halogen bonding in solution,
particularly where competition with other types of interactions
is a concern.
A traditional way to characterize solvent effects on nonco-
valent interactions is to employ empirically developed solvent
polarity parameters.39 We found that graphs of log Ka against
values of commonly used polarity scales such as ET(30) (Figure
10a) and π* (see the Supporting Information) led to a common
conclusion: the Et3N-IC8F17 interaction shows a relatively
minor response to solvent polarity within the range of solvents
tested, but is significantly weakened in solvents that can serve
as competitive hydrogen bond donors. The py scale, based on
the intensities of vibronic bands of pyrene fluorescence, il-
lustrates this point well (Figure 10b): while many solvent scales
that employ solvatochromic dyessincluding ET(30)sactually
A graph of the experimental binding data (log Ka) against
the predictions of the Hunter electrostatic model is shown
(36) Beeson, C.; Pham, N.; Shipps, G., Jr.; Dix, T. A. J. Am. Chem. Soc.
1993, 115, 6803–6812, and references cited therein.
(37) (a) Cockroft, S. L.; Hunter, C. A. Chem. Commun. 2006, 3806–3808.
(b) Cubberley, M. S.; Iverson, B. L. J. Am. Chem. Soc. 2001, 123,
7560–7563. (c) Breault, G. A.; Hunter, C. A.; Mayers, P. C. J. Am.
Chem. Soc. 1998, 120, 3402–3410. (d) Smithrud, D. B.; Diederich, F.
J. Am. Chem. Soc. 1990, 112, 339–343.
(38) The values of R ) 2.5 for iodoperfluorooctane and ꢀ ) 7.5 for
triethylamine reported in ref 11d were used. Values of R and ꢀ reported
in ref 7c were used for acetonitrile, acetone, tetrahydrofuran, and
chloroform. New parameters developed by Hunter and co-workers (ref
29) were employed for cyclohexane (R ) 1.2, ꢀ ) 0.6), benzene (R
) 1.1, ꢀ ) 1.6), and dichloromethane (R ) 1.7, ꢀ ) 1.5). Values for
2-propanol (R ) 2.7, ꢀ ) 5.5) were provided by Prof. C. A. Hunter
(personal communication).
(39) Reichardt, C. SolVents and SolVent Effects in Organic Chemistry, 3rd
ed.; Wiley-VCH: Weinheim, Germany, 2003.
(40) Dong, D. C.; Winnik, M. Can. J. Chem. 1984, 62, 2560–2565.
9
1652 J. AM. CHEM. SOC. VOL. 132, NO. 5, 2010

Halogen Bonding: Structural and Solvent Effects
A R T I C L E S
bond donor are varied in a systematic, incremental fashionsas
in the measurements of the interactions of para-substituted
iodoperfluoroarenes presented hereselectrostatic models may
be of significant value. The fact that XB interactions are subject
to linear free energy relationships with empirically derived
substituent parameters (σ) and quantities derived from calcula-
tions (molecular electrostatic potentials) represents a basis for
systematically tuning halogen bond donor ability in the context
of receptor or drug design.
A wider variation of the structure of halogen-bonding partners
also reveals limits to a purely electrostatic approach: a data set
that includes interaction energies of two halogen bond donors
with four halogen bond acceptors that vary widely in structure
is poorly modeled by an electrostatic model. Instead, DFT-
calculated binding constants show moderate correlation with
experimental data, except in the case of the strongest halogen
bond of the data set (quinuclidine-iodoperfluoroalkane). This
represents the most extensive comparison of experimentally
determined halogen bond energies with computed data to date,
and it demonstrates that DFT methods may have utility in
modeling the types of halogen bonds currently applied in
supramolecular settings. It also will represent a useful data set
for testing the ability of newer DFT functionals and higher level
calculations to describe trends in the thermodynamics of halogen
bonding.
Limitations to an electrostatic model are also revealed through
studies of solvent effects: while hydrogen bond donor solvents
are detrimental to XB in solution, Lewis basic solvents compete
with triethylamine for halogen bonds less well than would be
expected on the basis of electrostatic interactions alone. Thus,
while this study provides support for the characterization of XB
as an interaction with a dominant electrostatic component, it
also reveals subtle differences between this interaction and
hydrogen bonding. These differences may represent interesting
opportunities for exploiting halogen bonding in competition or
cooperation with other noncovalent interactions.
Figure 10. Experimental binding data (log Ka for the C8F17Istriethylamine
interaction) as a function of solvent polarity as measured by (a) ET(30) and
(b) the py scale.
measure a combination of polarity and hydrogen bond donor
ability, the py scale quite efficiently separates these effects,
providing a measure of polarity without a substantial H bond
donor contribution.40 The hydrogen bond donor solvents tert-
butyl alcohol, chloroform, and 2-propanol are clear outliers from
a data set that otherwise shows relatively minor variation of
the strength of the halogen-bonding interaction as a function of
py. These data suggest that hydrogen bond donor solvents are
able to compete with iodoperfluorooctane for binding to the
common acceptor triethylamine. Competition between halogen
bond and hydrogen bond donors for a common acceptor has
also been observed in the solid state.41
Acknowledgment. This work was supported by NSERC
(Discovery Grants Program), the Canadian Foundation for Innova-
tion (Leaders Opportunity Fund), the Ontario Ministry of Research
and Innovation (Research Infrastructure Program), the University
of Toronto, and Merck Research Laboratories (Canadian Academic
Development Program). We thank Prof. Mitch Winnik for his advice
and valuable comments on this study and Prof. Datong Song for
assistance with the DFT calculations. We also thank Prof. Chris
Hunter (Department of Chemistry, University of Sheffield) for
helpful discussions and for sharing unpublished data on new values
of R and ꢀ for nonpolar solvents.
Conclusions
We have measured association constants for the relatively
strong halogen bonds of iodoperfluoroalkanes and iodoperfluo-
roarenes employed most frequently in modern applications of
the XB interaction. When the electronic properties of a halogen
Supporting Information Available: Complete experimental
procedures, data analysis, and computational details. This
material is available free of charge via the Internet at http://
pubs.acs.org.
(41) (a) Metrangolo, P.; Resnati, G. Science 2008, 321, 918–919. (b)
Corradi, E.; Meille, S. V.; Messina, M. T.; Metrangolo, P.; Resnati,
G. Angew. Chem., Int. Ed. 2000, 39, 1782–1785. (c) Aakero¨y, C. B.;
Fasulo, M.; Schultheiss, N.; Desper, J.; Moore, C. J. Am. Chem. Soc.
2007, 129, 13772–13773.
JA9086352
9
J. AM. CHEM. SOC. VOL. 132, NO. 5, 2010 1653