The disorder of perfluoroalkyl chains in crystals: Two case histories of interpretation and refinement

Article abstract of DOI:10.1016/j.jfluchem.2009.06.018

Perfluoroalkyl chains in solids are highly disordered in a wide range of temperatures. Poor attention is typically given to this problem in crystallographic studies to the point that no attempt is frequently made in order to model the collected data and d

Full text of DOI:10.1016/j.jfluchem.2009.06.018

Journal of Fluorine Chemistry 130 (2009) 816–823
Contents lists available at ScienceDirect
Journal of Fluorine Chemistry
journal homepage: www.elsevier.com/locate/fluor
The disorder of perfluoroalkyl chains in crystals: Two case histories of
interpretation and refinement
a,b,**
Archan Dey a, Pierangelo Metrangolo a, Tullio Pilati b, , Giuseppe Resnati
*
,
Giancarlo Terraneo a,***, Ivan Wlassics c
a NFMLab-Department of Chemistry, Materials, and Chemical Engineering ‘‘Giulio Natta’’, Politecnico di Milano, Via L. Mancinelli 7, 20131 Milan, Italy
b CNR-Institute of Molecular Science and Technology, University of Milan, Via Golgi 19, 20133 Milan, Italy
c Solvay-Solexis, Research & Development, Viale Lombardia 20, 20021 Bollate, Milan, Italy
A R T I C L E I N F O
A B S T R A C T
Article history:
Perfluoroalkyl chains in solids are highly disordered in a wide range of temperatures. Poor attention is
typically given to this problem in crystallographic studies to the point that no attempt is frequently made
in order to model the collected data and disorder remains even unmentioned in a large number of single
crystal reports. This paper presents a short analysis of the problems related with this disorder. Two
simple crystal structures, namely the halogen bonded adducts between N,N,N0,N0-tetramethyl-1,4-
Received 15 April 2009
Received in revised form 24 June 2009
Accepted 25 June 2009
Available online 5 July 2009
phenylendiamine and a,v-diiodoperfluorobutane or a,v-diiodoperfluorohexane are discussed in details
in order to suggest some simple and basic principles for the refinement of perfluoroalkyl chains in single
crystal structural studies.
Dedicated to Prof. Henry Selig
on the occasion of his receiving
the 2009 ACS Award for Creative
Work in Fluorine Chemistry.
ß 2009 Elsevier B.V. All rights reserved.
Keywords:
Perfluorocarbons
Fluoroalkanes
Disordered molecular crystals
Refinement problematic structures
Conformational flexibility
Rotational isomers
1. Introduction
in organic materials. A rational use of perfluoroalkyl residues in
order to tune the functional properties of a material requires a
The unique properties of the fluorine atom (high ionization
potential, electron affinity, and electronegativity, small atom
polatizability and radius) [1] are the basis for the unique properties
of perfluoroalkanes and their derivatives (high density, viscosity,
and chemical stability, low surface tension, refractive index, and
dielectric constants) [2]. The usefulness of these properties account
for the interest for the incorporation of perfluoroalkyl moieties into
small molecules and polymers when the obtainment of new
materials with modified physical characteristics is pursued (e.g.
visco-elastic, dielectric, and optical properties) [3]. The presence of
perfluoroalkyl residues is thus becoming more and more common
detailed knowledge of the structural characteristics of these
residues, both in solution and in the solid. This knowledge has to
span a wide length scale, from the Angstrom to the micrometer.
This paper gives a contribution to the full utilization of the
potential of single crystal X-ray analyses for determining the
structural details of perfluoroalkyl chains in the solid. An analysis
of the Cambridge Structural Database (CSD Version 5.30) proves
how perfluoroalkyl residues in crystals are highly disordered in a
wide range of temperatures. Poor attention is typically given to this
problem to the point that no attempt is frequently made in order to
model the collected data and disorder remains even unmentioned
in a large number of single crystal reports [4]. Some approaches to
model this disorder will be discussed thus favouring a deeper
insight into a key feature of perfluoroalky chains, namely their
conformational preferences. The halogen bonding (the non-
covalent interaction involving halogens as electrophilic species
[5]) has recently allowed for the easy formation of solid adducts on
self-assembly of mono- and diiodoperfluoroalkanes with a wide
diversity of electron donor species [6]. Nicely crystalline materials
*
Corresponding author. Tel.: +39 02 5031 4245; fax: +39 02 5031 3927.
** Corresponding author at: Department of Chemistry, Materials, Chemical
Engineering ‘‘Giulio Natta’’, Politecnico di Milano Via Mancinelli, 7, I-20131 Milan,
Italy. Tel.: +39 02 2399 3032; fax: +39 02 2399 3180.
E-mail addresses: tullio.pilati@istm.cnr.it (T. Pilati), giuseppe.resnati@polimi.it
(G. Resnati), giancarlo.terraneo@polimi.it (G. Terraneo).
*** Corresponding author. Tel.: +39 02 2399 3063; fax: +39 02 2399 3180.
0022-1139/$ – see front matter ß 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.jfluchem.2009.06.018

A. Dey et al. / Journal of Fluorine Chemistry 130 (2009) 816–823
817
Scheme 1. Halogen bonding driven self-assembly of infinite chains 3a,b.
are frequently obtained and two such adducts will be studied here.
Specifically, the single crystal X-ray analyses of the infinite chains
3a,b (obtained when N,N,N0,N0-tetramethyl-1,4-phenylendiamine
(TMPDA, 1) self-assembles with 1,l,2,2,3,3,4,4-octafluoro-1,4-
diiodobutane (PFDIB, 2a) or 1,1,2,2,3,3,4,4,5,5,6,6-dodecafluoro-
1,6-diiodohexane (PFDIH, 2b), respectively) will allow for the
description of some simple and basic protocols for the refinement
of perfluoroalkyl chains (Scheme 1).
(corresponding to 6925 datapoints) (Fig. 1a) shows clearly that the
subset still contains a large number of inaccurate data. In order to
eliminate the less accurate of them, we have excluded the single
datapoints in which at least one parameter p (either C–F distance or
F–C–F angle) differs, from the corresponding mean value (pmean
,
the white dot in Fig. 1a), for more than twice the corresponding
standard deviation [SD, derived by the CCDC VISTA program,
p = pmean ꢀ 2SD(p)]. Although the resulting intervals are quite large
˚
(0.12 A and 9.38), another 8% reduction of the set occurred (from 1083
to 1035 hits and from 6925 to 6339 datapoints).
2. Results and discussion
An analysis of this reduced subset was made by plotting the two
C–C–F angles of a CF2 group (Fig. 1b). Obviously, the two angles
being chemically equivalent in most cases, their mean values are
nearly equal (Fig. 1b, bottom). The scatterplot of Fig. 1b begins to
be sufficiently refined and shows a meaningful detail, namely the
presence of two partly overlapping populations (central and
bottom-left regions with high density of datapoints).
Fig. 2 is a further refinement of Fig. 1b and has been obtained by
excluding all datapoints from a structure wherein a parameter is
out of the limit p = pmean ꢀ 2SD(p) and also the datapoints from
structures where the parameter p = C–C–F angle is out of the interval
p = pmean ꢀ 3SD(p). The aim is to avoid the loss of the apparent bias of
Fig. 1b. The CSD hits are now reduced to 866 hits and give rise to 5079
datapoints. The presence of two populations is much more evident
(ellipsoids in Fig. 2), confirming that the subset is now quite clean. An
analysis of this bias shows that the group of datapoints in the bottom-
left ellipsoid are essentially due to –CF2– embedded in perfluor-
ocarbon chains, while those of the central ellipsoid contains mainly
the less hindered groups, like CH2–CF2–C and –CF3 moieties. It is
noteworthy that, on going from the 2753 initial hits to the ‘error-free’
866 hits of Fig. 2, more than 2/3 of the molecules containing the
CSD studies. Due to the strength of C–F bond and the steric
requirements of F–C–F groups, C–Csp3F2– groups are expected to be
characterized by very limited geometric parameters variability. In
order to confirm this expectation and to get an extensive overview
of these parameters in the solid state, an analysis of the Cambridge
Structural Database (CSD Version 5.30) was performed. The
analysis reveals that most of the terminal perfluorinated chains
are disordered and we will show how a fairly simple refinement
can solve some of the related problems. Specifically, we searched
for all structures, in the CSD, containing at least one acyclic C–
Csp3F2– group. To limit the inclusion of inaccurate data, or to elicit
biases due to different chemical environments, we imposed the
following restraints: 3D coordinates, no errors, no ions, not
polymeric, no powder structures. The search provided 2753 hits,
these structures becoming 2041 on adding the condition not
disordered, and being further reduced to 1083 on adding the
restraint R < 0.05. It might be expected the obtained dataset is
rather accurate and characterized by a very sharp distribution of
geometric parameters. On the contrary, the scatterplot of F–C–F
bond angles versus C–F bond lengths of the 1083 structures
˚
Fig. 1. (a) Scatterplot of F–C–F angle (8) versus C–F bond distance (A) in a group C–Csp3F2; data obtained from CSD Version 5.30; no limits established for F–C–F and C–F values.
The white spot represents the pmean value and the box delimitates the datapoints p = pmean ꢀ 2SD(p). (b) Scatterplot of C–C–F1 angle (8) versus C–C–F2 angle (8) for the more
˚
˚
accurate datapoints included in the box of (a) (namely: 102.1738 < F–C–F angles < 111.4778; 1.267 A < C–F distance < 1.387 A).

818
A. Dey et al. / Journal of Fluorine Chemistry 130 (2009) 816–823
Fig. 2. Scatterplot of datapoints of Fig. 1b, with the further exclusion of all datapoints
from a structure wherein a parameter is out of the limit p = pmean ꢀ 2SD(p) and of the
datapoints from structures where the parameter p = C–C–F angle is out of the interval
p = pmean ꢀ 3SD(p) (104.8008 < C–C–F1 (or C–C–F2) angle < 118.000). The two ellipses
evidence the bias between two groups of data with different hindrance (see text).
Fig. 3. Histogram of the numbers of C–C–C–C torsion angles with a given absolute
value (8) for the structures subset of Fig. 2 without any gauche conformation (mean
value = 168.7468).
corresponding 5079 –CF2– groups are probably disordered or at least
have very high anisotropic thermal parameters. In fact, the C–C–F1
and C–C–F2 angles are usually chemically equivalent, and they are
thus expected to be dispersed strictly along the diagonal of the plot.
Differently, this is not the case here and the dispersion of the data is
still very high despite the subset seems to be free from inaccurate
data. This observation indicates that perfluorinated residues are
frequently affected by disorder and that attempts are rarely made to
accurately refine these residues [4].
conformation, as we will see in one of the examples presented
below. In some cases, when the perfluorinated chain is short and
the deviation from the exact trans C–C–C–C torsion angles are
small, exclusively the ordered anisotropic standard refinement is
possible. The method gives good results and the only problem
encountered in these cases are the high values of the anisotropic
displacement parameters which causes an apparent small short-
ening of C–F covalent bonds and a wrong C–C–C–C torsion angle. In
contrast, when the perfluoroalkyl chain is long or the C–C–C–C
torsion angles markedly differ from 1808, or 608, the split model
refinement becomes essential as well as the use of restraints.
The modern codes for crystal structures refinement provide
facilities for refinement with restraints and constraints, easy to use
and very powerful. In the following discussion we will refer to
SHELX-97 and its facilities [11], the standard refinement program
in our laboratory, but procedures similar to those discussed here
may be found in other programs.
The single crystal X-ray structure determination, and a simple
refinement procedure, of the infinite chains 3a,b (Scheme 1)
(obtained on halogen bonding driven self-assembly of N,N,N0,N0-
tetramethyl-1,4-phenylendiamine (TMPDA, 1) with 1,4-diiodoper-
fluorobutane (PFDIB, 2a) or 1,6-diiodoperfluorohexane (PFDIH,
2b), respectively) is discussed below. These two case histories will
allow for the description of some simple and basic rules for the
refinement of perfluoroalkyl chains.
Fluorine atoms typically do not provide strong intermolecular
interactions with any atomic species situated at the periphery of
organic molecules (e.g. H, O, halogen atoms) [7]. In fact, at the bulk
materials level, perfluorinated chains impart hydrophobicity,
lipophobicity, and, even more remarkable, also the Fꢁ ꢁ ꢁF interac-
tions are very weak [1]. As their hydrocarbon counterparts,
perfluorinated residues can assume different conformations by
rotating around C–C single bonds, whenever possible. This is
particularly the case for trifluoromethyl groups [8]. In contrast,
residues with at least three consecutive –CF2– groups are stiffer
than the equivalent hydrocarbon chains and show ground state C–
C–C and F–C–F angles which are greater and smaller than the
tetrahedral angle, respectively. Few structures reported in the CSD
display a gauche conformation [9] as the trans arrangement is
largely preferred. In details, the distorted trans conformation is
favoured over the exact trans conformation due to the short
contacts and electrostatic repulsions between 1,3 positioned CF2
groups [10] (Fig. 3). If no severe packing constrains are present in
the crystal, the energy barrier between the conformation with all
positive (t+) and all negative (tꢂ) deviations from the exact trans
arrangement of the perfluoroalkyl chain is low and both
conformers of the perfluorocarbon chains are present, dynamically
or statically. This is often the case also at low temperature.
Generalities on the refinement of perfluoroalkyl chains. In most
cases an interpretation of perfluoroalkyl residues disorder less
inaccurate than that routinely practised is not time demanding and
is also very easy. It produces more accurate geometric parameters,
The case histories of 3a,b. Partial evaporation of equimolar
solutions of TMPDA 1 and PFDIB 2a, or PFDIH 2b, affords colourless
crystals 3a,b where starting modules are present in 1:1 ratio (as
confirmed with 1H/19F NMR analyses in the presence of 2,2,2-
trifluoroethyl ether as external standard [12]). Single crystal X-ray
analyses of these solids (Fig. 4) confirm the stoichiometry, reveal
that both TMPDA 1 and a,v-diiodoperfluoroalkanes 2a,b behave as
bidentate and telechelic modules and lie on crystallographic
˚
lower
R
values and residual electron density, and, more
symmetry centres. Short Nꢁ ꢁ ꢁI halogen bonds (2.862(5) A in 3a and
˚
importantly, it may give the correct molecular connectivity and
2.845(3) A in 3b) are present, which drive the self-assembly and

A. Dey et al. / Journal of Fluorine Chemistry 130 (2009) 816–823
819
Fig. 4. Ellipsoid representation of 1D unlimited linear chains of 3a (top) and 3b (bottom). Colours are as follow: carbon, grey; fluorine, green-yellow; nitrogen, light blue;
iodine, purple. Hydrogens are omitted for clarity. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of the article.)
give rise to unlimited chains where the complementary modules
alternate.
geometrical connection around the two independent C atoms of
PFDIB molecule, in fact C1–F2 and C2–F2 distances are 1.61(3) and
˚
While the supramolecular arrangements of 3a,b are quite
standard, these co-crystals present two significant and educational
aspects in terms of crystallographic resolution and refinement. The
ordinary refinement of 3a leads to incorrect molecular connectiv-
ity and the splitting model of 3b into two subsets, in spite of
SHELXL-97 suggestion, is not supported by the experimental
data.Fig. 5(top) shows the resulting peaks from SIR2002 resolution
[13] of 3a (purple circles for iodine and black circles for undefined
atoms), together with some peaks from the subsequent difference
Fourier map (green circles). Clearly, the TMPDA 1 moiety is well
defined and not disordered, but this is not the case for PFDIB 2a. A
quick and simple analysis based on geometrical considerations
provides the starting system for the refinement of 3a (Fig. 5,
bottom).
The four different refinement models M1–M4 are tested to
refine the perfluorinated chains. In the Model 1 (M1) the standard
anisotropic refinement was carried out to the end, but some
problems occurred with the PFDIB moiety (Table 1). In spite of the
reasonable disagreement factors, large correlation factors are
present, especially between anisotropic displacement parameters
(ADPs) components of fluorine atoms. This approach leads to poor
1.91(3) A, respectively.
Additionally, in M1 the ratios between maximum (Umax) and
minimum (Umin) principal mean square atomic displacements for
some atoms in the perfluorinated chain are very large, particularly
those of F atoms and C1 (Fig. 6a). If the displacements anisostropy
are only due to the libration around the molecular inertial axes, the
Umax/Umin ratios of I1, C1, and C2 are expected to be almost the
same, as their distance from the longest molecular axis is small and
nearly the same. On the contrary, C1 carbon shows an extremely
higher ratio than I1 and C2 (Table 1, second column). Only by the
superposition of the t+ and tꢂ conformers or of two opposite
conformers may explain this difference. Obviously, SHELXL97
suggests the splitting of C1 and all F atoms.
The second refinement model (M2) is slightly more sophisti-
cated than M1. Atoms are split as suggested by SHELXL97 but no
constraint is applied (Table 1, Fig. 6b and c). The correlations
between refined parameters given by M2 are higher than those of
M1 (0.893 vs 0.804) and the dispersion of the geometric
parameters in the perfluorinated chain is still very high (the
˚
intramolecular C2-F2B distance is as short as 1.85(3) A). The F4A–
F4B distance (3.005 A) is much longer than two times the expected
˚
˚
value for the canonical C–F distance (1.34 A) thus proving further
the presence of some problems in M2. This suggests that also the
C2 carbon atom must be split.
Clearly, the constrained refinement becomes mandatory and
new refinement models must be developed. The choice of the
restraints is somehow problematic as it does not follow strict, well-
defined, and pre-established rules and largely depends on
crystallographer’s preferences. In 3a, also C2 carbon atom has to
be split for imposing the correct restraints (Fig. 6d and e) and the
ambiguity problem of choosing the right model to describe the
perfluorinated chain has to be solved. PFDIB 2a can be interpreted
as the four different conformers C1A–C2A–C2B–C1B, C1A–C2A–
C2A0–C1A0, C1A–C2B0–C2B–C1A0, or C1A–C2B0–C2A0–C1B (Fig. 7).
The last two models can be excluded since the distance between
C1A–C2B0 is too short. C1A–C2A–C2B–C1B Sequence is treated
with the third method (M3) and C1A–C2A–C2A0–C1A0 sequence
with the fourth method (M4). The choice of the carbon atoms
sequence is a crucial factor and influences both the definition of the
geometric restraints and the modelling of the molecular con-
formation.
The conformation is not the only difference in the two backbone
sequences. In fact, the C1A–C2A–C2B–C1B chain (M3) is related to
C1A0–C2A0–C2B0–C1B0 by a crystallographic centre of symmetry,
this means that the population factor must be fixed to 0.5, while for
M4 chain the two split models are not correlated by any symmetry
element and their complementary population factors could need
an optimization parameter.
Fig. 5. Top: ball and stick model plot by SIR2002 solution of 3a (purple circles for
iodine, black circles for undefined atoms) and difference map residues (green circles).
Bottom:initialstructuretoberefined,obtainedbyeliminationofsomepeaksbasedon
pure geometric considerations. (For interpretation of the references to colour in this
figure legend, the reader is referred to the web version of the article.)

820
A. Dey et al. / Journal of Fluorine Chemistry 130 (2009) 816–823
Table 1
Proposed refinement models for 3a.
M1
M2
M3
M4
No. ref. parameters
No. restraints
118
0
164
0
168
170
169
181
R1 [Io > 2
s
(Io)]
0.0576
0.441
0.495
0.485
wR2 [Io > 2
s(Io)]
0.1484
0.804
0.1107
0.893
0.1336
0.770
0.1277
0.806
Max. correlation factors
No. correlatioꢂn3factors >0.7
9
23
4
5
˚
Drmin,max e A
ꢂ0.86, 1.52
ꢂ0.50, 1.52
ꢂ0.57, 1.52
ꢂ0.50, 1.52
No. reflections
Part A
2.20(2)
Part B
2.18(3)
Part A
Part B
2.180(17)
Part A
2.192(15)
Part B
C1–I1
2.148(14)
2.180(16)
2.196(17)
1.345(10)
1.339(10)
1.541(12)
1.523(12)
1.343(10)
1.344(10)
123(2)
C1-F1
1.201(19)
1.61(3)
1.49(3)
1.38(4)
1.43(4)
1.30(3)
–
1.362(11)
1.339(10)
1.519(11)
1.515(11)
1.355(10)
1.352(10)
1.356(11)
1.348(11)
1.523(12)
–
1.348(10)
1.327(10)
1.519(11)
1.530(11)
1.351(9)
1.344(9)
C1-F2
1.17(2)
C1–C2
C2–C2a
1.301(17)
1.577(19)
1.305(13)
1.539(16)
1.40(2)
1.590(14)
1.41(2)
C2–F3
1.278(18)
1.67(2)
98.9(16)
125(3)
85(2)
1.347(10)
1.363(10)
C2–F4
1.398(16)
F1–C1–F2
F1–C1–C2
F2–C1–C2
F3–C2–F4
F3–C2–C1
F4–C2–C1
F3–C2–C2a
F4–C2–C2a
C1–C2–C2a
Umax/Umin (I1)
Umax/Umin (C1)
Umax/Umin (C2)
Umax/Umin (F1)
Umax/Umin (F2)
Umax/Umin (F3)
Umax/Umin (F4)
105(2)
118(2)
117.6(19)
123(2)
119.7(17)
120.7(16)
78.0(15)
99.8(12)
119.3(12)
84.4(16)
112.5(12)
103.8(11)
125.3(14)
2.48
97(2)
109.8(2)
103.6(18)
104.5(16)
112.1(14)
111.2(14)
105.0(11)
119.4(13)
2.49
101.2(11)
105.0(11)
109.9(16)
105.6(10)
103.7(10)
114.1(11)
103.4(9)
119.4(12)
2.49
100.3(12)
101.4(12)
110.4(17)
105.2(11)
104.7(11)
114.0(11)
104.7(10)
118.3(13)
102.9(10)
103.1(10)
110.8(13)
106.1(9)
102.0(11)
103.2(12)
112.0(17)
104.6(11)
104.6(10)
111.3(11)
111.7(12)
112.4(14)
98.9(16)
117.3(18)
94(2)
103.8(9)
113.2(12)
98.1(9)
125.2(17)
2.49
111.8(10)
111.7(10)
112.2(12)
9.21
2.88
3.48
2.71
2.23
–
2.38
2.19
5.68
7.63
4.02
5.53
2.37
–
1.82
2.33
–
2.38
8.44
10.12
16.00
5.85
6.02
4.28
5.09
3.74
18.11
4.58
4.76
3.96
12.55
7.72
4.48
7.84
8.24
4.49
7.86
3.39
8.27
7.15
5.28
5.31
a
-x, -y, -z for M3, C2A–C2A and C2B–C2B bonds; for M4, C2A–C2B.
Fig. 6. The three different models of PFDIB refinement: (a) standard anisotropic M1; (b) and (c) unrestrained M2; (d) and (e) restrained M3.
In this case the chosen starting geometry was based on the
techniques summarized in Section 4.3. To refine appropriately the
3a system we decided to restrain the C–F and C–C distances to 1.34
Umax/Umin ratios compared with standard anisotropic and unrest-
rained models, a single exception is observed for F1B for M4.
Indeed, in our opinion the real situation, at room temperature, has
not to be ascribed to one single model but is a dynamic equilibrium
˚
˚
and 1.54 A, respectively, using
a 0.01 A estimated standard
deviation (e.s.d.). Moreover, considering that a 1,3 distance,
corresponding to a bond angle, is less rigid than a bond distance,
we have imposed the same values for chemically equivalent angles
˚
with a 0.03 A e.s.d. Additional restraints were adopted to describe
the perfluorobutane chain motion assuming similar ADPs with
e.s.d.s of 0.015. For C1A and C1B only, whose separation was lower
that the resolution data, we imposed equal ADPs. The applied
constraints and restraints were the same in both models M3 and
M4, apart from minor details due to the different symmetry of the
two models.
Both the M3 and M4 models provide a better geometrical
description of the perfluorinated chain than M1 and M2. The
refinement of M3 and M4 gives smaller correlation factors and
Fig. 7. Possible interpretations of molecular split: the exact trans conformation is
C1A–C2A–C2A0–C1A0, the distorted trans conformation is C1A–C2A–C2B–C1B.

A. Dey et al. / Journal of Fluorine Chemistry 130 (2009) 816–823
821
of the two models, despite M3 seems to be better on the basis of
geometric values, ADPs, and correlations as reported in Table 1.
The refinement of 3b is much simpler than of 3a. For 3b, we
applied a refinement based on a standard anisotropic model,
imposing very soft restraints only on C–F, C–C distances and F–C–F
and C–C–C angles. All the correlation parameters are smaller than
0.5 and the R and wR values are 0.0386 and 0.1120. The geometry
thus obtained for DIPFH is reasonable, with C–F and C–C–F
(s), 1042.5 (s), 934.6 (s), 817.3 (s), 763.4(s). 3b: M.p. (CHCl3, onset
temperature): 103 8C. IR (cmꢂ1, selected bands): 2894.4 (w),
2854.4 (w), 1514.6 (s), 1473.9 (m), 1301.2 (m), 1208.7 (s), 1171.0
(s), 1137.2 (s), 1078.8 (s), 936.3 (s), 819.8 (s), 819.8(s), 784.04 (m),
653.5 (s).
4.2. X-ray crystal structure determination
˚
distances and angles in the range 1.298–1.342 A and 107.18–
Crystal data for 3a: C10H16N2.C4F8I2, M = 618.09, monoclinic, C2/
˚
110.48, respectively. Notwithstanding this, SHELXL97 suggests to
split five of the six independent fluorine atoms, which we did. The
suggested separation between couples of split atoms is very low
and we constrained their ADPs to be equal. The results were very
disappointing as R and wR converged at 0.0387 and 0.1125,
respectively, these values being larger than the previous values. No
improvement in the description of the geometry was obtained: a
very large number of correlations between geometric parameters
were in the range 0.90–0.99 and, due to the increment of the
,
a = 22.340(4), b = 6.0763(10), c = 16.203(3) A,
b
= 113.46(3),
c
3
r
(calc) = 2.035 g/cm3, F000 = 1168, dimen-
˚
V = 2017.7(6) A , Z = 4,
sion: 0.06 ꢃ 0.34 ꢃ 0.40. Data collected by a Bruker SMART APEX
diffractometer Mo-K
a
radiation,
l
= 0.71073 A,
m
= 3.188 mmꢂ1
,
˚
T = 295(2) K; 16585 reflection collected, 1990 independent, 1575
with Io > 2s(Io), absorption corrections Tmin/Tmax = 0.724,
Rint = 0.038, 2umax = 528. Structure solved by SIR2002 and refined
on F2 by SHELX-97. CCDC M1: 726920; M2:726921; M3: 726922;
M4: 726923.
number of refined parameters,
a
substantial worsening of
Crystal data for 3b: C10H16N2.C6F12I2, M = 718.11, triclinic, P-1,
˚
geometric parameters’ e.s.d.s occurred. Besides, the maximum
a = 5.8258(10), b = 8.1756(14), c = 12.592(2) A,
a
= 96.16(2),
(calc) =
3
˚
˚
separation between split atoms was 0.35 A, below the resolution of
b
= 98.02(2),
g
= 98.77(2) V = 581.85(17) A , Z = 1,
r
our data. It is thus possible to conclude that the standard
refinement is completely adequate for 3b and that the large
anisotropy of fluorine atoms in the chain only reflects the large
thermal motion of these atoms.
2.049 g/cm3, F000 = 340, dimension: 0.05 ꢃ 0.08 ꢃ 0.38. Data col-
lected by a Bruker SMART APEX diffractometer Mo-K
= 2.80 mmꢂ1
˚
= 0.71073 A, m , T = 295(2) K; 18464 reflection
a radiation,
l
collected, 2682 independent, 2270 with Io > 2s(Io), absorption
corrections Tmin/Tmax = 0.790, Rint = 0.031, 2umax = 55.28. Structure
solved by SIR2002 and refined on F2 by SHELX-97, 145 parameters,
13 restraints. Final R = 0.0455 (0.0386 on observed reflections),
3. Conclusions
ꢂ3
˚
wR ¼ 0:1179 ð0:1120Þ, g.o.f. = 1.051, ꢂ0.49 < Dr < 1.48 e A
.
The complete analysis of the structure 3a, the refinement of the
various models included, did not require more than five working
hours. The analysis of 3b was even less time demanding. While
only few minutes are required to define the restraints on a
disordered difluoromethylene group and to set up an initial model
to be refined, the resulting improvements are substantial, as
presented for 3a. The approach does not allow for well refined
structures in all cases and sometimes the work must be very hard
and time consuming (as was the case for a calix[5]arene bonded to
five disordered –C7F15 chains [14]). Nevertheless, the disordered
refinement models should be adopted whenever needed, at least
for the simplest cases, in order to provide the scientific community
with much more information and consistent data.
CCDC 726919.
Copies of the 3a (all models M1–M4) and 3b data can be
obtained, free of charge via www.ccdc.cam.ac.uk/conts/retrie-
ving.html (or from the Cambridge Crystallographic Data Centre,
CCDC, 12 Union Road, Cambridge CB2 1EZ, UK; fax: 44 1223
336033 or e mail: deposit@ccdc.cam.ac.uk.
4.3. General technicalities for refining disordered perfluorated chains
As shown in the refinement of 3a, the presence of disorder in
perfluorinated chains may often be noticed from the very initial
stages. For example, SHELX resolution programs output a number
of peaks much greater than the independent atoms number, and
when disorder is present the excess often corresponds to
disordered fluorine atoms.
4. Experimental
Commercially available chemicals were used without further
purification. Chemicals were purchased from Sigma–Aldrich and
Apollo Scientific. IR spectra were recorded with a Nicolet Nexus FT-
IR spectrophotometer equipped with the Smart Endurance system
(UATR). The thermal analysis was recorded with a Linkam DSC600
Stage (temperature range: ꢂ196 8C to 600 8C) coupled with the
LN94 cooling system.
The definition of the starting model may follow different
protocols. The first protocol is suitable for the simplest cases (such
as the disorder in –CF3 groups) and simply requires all the F atoms
are refined anisotropically, the atoms are split (as suggested by
SHELXL), dividing the split peaks into two models as similar as
possible to an ideal geometry, and assigning a complementary
population factor to the two models.
A second protocol is to start directly from the peaks derived
from the structure resolution, possibly by adding the right peaks
derived from a Fourier map (calculated before starting the
refinement) as in the case shown for 3a structure. The peaks are
attributed to the two conformers (or parts in SHELXL terminology)
by simply resorting to geometric consideration, while the
population factor of each part may be initially the same also
when symmetry considerations does not add specific constrains.
Sometime, the initial model may be complete, as in the case of a
–CF3 split, where not only the disordered atoms are all defined, but
also the restraints may be imposed without ambiguity. In other
cases the initial model may be incomplete, due to the difficulty to
define the position of all the disordered atoms or to establish
adequate restraints. In any case, it is convenient to start the
refinement with isotropic atoms, and, above all, with a large
4.1. Formation of the co-crystals 3a and 3b
Equimolar amounts of N,N,N0,N0-tetramethyl-1,4-phenylendia-
mine (TMPDA, 1) and 1,l,2,2,3,3,4,4-octafluoro-1,4-diiodobutane
(PFDIB, 2a) or 1,1,2,2,3,3,4,4,5,5,6,6-dodecafluoro-1,6-diiodohex-
ane (PFDIH, 2b) were dissolved with chloroform in a vial of clear
borosilicate glass at room temperature. The open vial was placed in
a closed cylindrical wide-mouth bottle containing CCl4. CHCl3 was
allowed to diffuse at room temperature and after 24 h yellowish
co-crystals of 3a and 3b were obtained. 1H and 19F NMR spectra of
both adducts showed the signals of pure starting compounds with
minor chemical shifts changes, if any. 3a: M.p. (CHCl3, onset
temperature): 112 8C. IR (cmꢂ1, selected bands): 2890.5 (w),
2851.1 (w), 1516.5 (s), 1476.0 (m), 1301.7 (m), 1175.9 (s), 1116.5

822
A. Dey et al. / Journal of Fluorine Chemistry 130 (2009) 816–823
Fig. 8. Three different Fourier maps for the structure 3a: M1, M2 and M3 (projection down b*, a-axis vertical and c-axis horizontal).
damping factor, especially if the population factors are simulta-
neously refined with displacement parameters. In fact, at the
beginning, the least squares matrix may be very ill-conditioned
with very large correlation parameters. In our experience, 8 cycles
of refinement with a damping factor that reduces the parameters
changes by 0.5 is normally much better than 4 cycles without
damping factor.
After the initial cycles, the analysis of the results is mandatory.
Refined atoms with a too low population factor are probably to be
deleted and the same has to be done for atoms with very large
displacement parameter Uiso. Some caution is required for these
latter atoms as a large Uiso of an atom with a large population factor
may simply indicate that the atom must be split. In that case the
anisotropic refinement of this atom may help to decide how to
proceed. An initial, complete, and coherent model of disorder could
normally be found after some cycles of refinement and elimination
of incorrect atoms.
for 3a can be displayed showing the reduction of the disorder in
the three different models (Fig. 8).
Acknowledgments
The financial support by Fondazione Cariplo (Project ‘‘Self-
Assembled Nanostructured Materials: A Strategy for the Control of
Electrooptic Properties’’) and Istituto Italiano di Tecnologia
(Politecnico di Milano Unit, Research Line 1: Bioelectronics and
Biophotonic Interfaces between Cells and Artificial System) are
gratefully acknowledged.
References
[1] (a) J.A. Gladysz, D.P. Curran, I.T. Harvath (Eds.), Handbook of Fluorous Chemistry,
Wiley-VCH, Weinheim, Germany, 2004;
(b) R.D. Chambers, Fluorine in Organic Chemistry, Wiley, New York, 1973.
[2] (a) S.C. Sharma, D.P. Acharya, K. Aramaki, Langmuir 23 (2007) 5324–5330;
(b) G. Floudas, M. Antonietti, S. Fo¨rster, Chem. Phys. 113 (2000) 3447–3454;
(c) M. Hird, Chem. Soc. Rev. 36 (2007) 2070–2095;
At this point the model restraints may be defined. Firstly, we
have to assign the molecular part: any atoms must be attributed
to a part and any part must have its own population factor that
may be fixed or variable. In our experience, when the split
molecular model is completely defined, it is convenient to start
again with isotropic atoms and assigning a population factor for
all the models (normally they are only two, with population
factor 0.5). The initial restraints must be hard but must not be
too dependent on the molecular conformation. The assumption
of strong restraints for bond distances and softer for 1–3
distances is probably a good criterion. In our experience the
SHELXL function SAME, that automatically defines the similitude
between chemically equivalent groups of atoms, is difficult to
use because the initial model may be very poorly defined and
the use of this instruction may bring to inconsistent results.
DFIX or at least SADI routines, specifying all the distances to be
collectively refined, is much more useful, especially at the very
beginning. Also at this stage the use of a damping factor may
greatly help. After any group of least squares cycles, the
geometry of the independent models must be controlled. When
two atoms are too close to each other, they sometimes need to
be exchanged from a part to the other and, in that case, it is
convenient to reset their Uiso to a expected value. At this stage, if
the mean Uiso is very different for the different parts it may be
convenient to decide if the population factor must be refined.
When the refinement of the two parts is geometrically
stabilized, the anisotropic refinement can be applied. At this
stage, some cautions must be used because the correlations
between coordinates and ADPs could be very large and if so we
newly start with a damping factor. Similar ADPs for all the atoms
of a part have also to be imposed, but if the distance between
two atoms in different parts is too short (for example less than
half the data resolution), these atoms have to be constrained to
the same ADPs. At the last stage of the refinement, the damping
factor has to be eliminated and the restraints softened, playing
attention that the correlation factors between refined para-
meters remain sufficiently low. Additionally, the Fourier maps
(d) P. Metrangolo, C. Pra¨sang, G. Resnati, R. Liantonio, A.C. Whitwood, D.W. Bruce,
Chem. Commun. (2006) 3290–3292;
(e) D.W. Bruce, P. Metrangolo, F. Meyer, C. Pra¨sang, G. Resnati, G. Terraneo, New J.
Chem. 32 (2008) 477–482.
[3] (a) D.S.L. Slinn, S.W. Green, in: R.E. Banks (Ed.), Preparation, Properties and
Industrial Applications of Organofluorine Compounds, Ellis Horwood, Chichester,
1982, pp. 45–82;
(b) G. Sandford, Tetrahedron 59 (2003) 437–454, and literature cited therein;
(c) Z. Luo, Q. Zhang, Y. Oderaotoshi, D.P. Curran, Science 291 (2001) 1766–1769;
(d) I. Viola, G. Ciccarella, P. Metrangolo, G. Resnati, R. Cingolani, G. Gigli, J. Fluorine
Chem. 128 (2007) 1335–1339.
[4] (a) C.R. Clough, J.B. Greco, J.S. Figueroa, P.L. Diaconescu, W.M. Davis, C.C. Cum-
mins, J. Am. Chem. Soc. 126 (2004) 7742–7743;
(b) R.P. Hughes, S.M. Maddock, I.A. Guzei, G.M. Liable-Sands, A.L. Rheingold, J. Am.
Chem. Soc. 123 (2001) 3279–3288;
(c) Q.-M. Wang, T.C.W. Mak, J. Am. Chem. Soc. 123 (2001) 7594–7600;
(d) P. Rombke, A. Schier, H. Schmidbaur, S. Cronje, H. Raubenheimer, Inorg. Chim.
Acta 357 (2004) 235–237;
(e) D. Shirotani, T. Suzuki, S. Kaizaki, Inorg. Chem. 45 (2006) 6111–6113;
(f) H.V. Huynh, D. Le Van, F.E. Hahn, T.S.A. Hor, J. Organomet. Chem. 689 (2004)
1766–1770;
(g) K.W. Muir, R. Rumin, F.Y. Petillon, J. Organomet. Chem. 635 (2001) 110–118;
(h) Y.-Y. Lee, J.-H. Chen, H.-Y. Hsieh, F.-L. Liao, S.-L. Wang, J.-Y. Tung, S. Elango,
Inorg. Chem. Commun. 6 (2003) 252–258;
(i) R.P. Hughes, I. Kovacik, D.C. Lindner, J.M. Smith, S. Willemsen, D. Zhang, I.A.
Guzei, A.L. Rheingold, Organometallics 20 (2001) 3190–3197;
(j) C.J. Bourgeois, R.P. Hughes, T.L. Husebo, J.M. Smith, I.M. Guzei, L.M. Liable-
Sands, L.N. Zakharov, A.L. Rheingold, Organometallics 24 (2005) 6431–6439;
(k) J.F. Berrien, M. Ourevitch, G. Morgant, N.E. Ghermani, B. Crousse, D. Bonnet-
Delpon, J. Fluorine Chem. 128 (2007) 839–843;
(l) W.G. Hollis Jr., M.G. Poferl, M.D. Wolter, P.A. Deck, C. Slebodnick, J. Fluorine
Chem. 129 (2008) 119–124;
(m) I. Krossing, Chem. Eur. J. 7 (2001), 490–452.
[5] (a) P. Metrangolo, G. Resnati, Science 321 (2008) 918;
(b) P. Metrangolo, F. Meyer, T. Pilati, G. Resnati, G. Terraneo, Angew. Chem. Int. Ed.
47 (2008) 6114–6127;
(c) K. Rissanen, CrystEngComm. 10 (2008) 1107–1113;
(d) P. Metrangolo, T. Pilati, G. Resnati, R. Liantonio, F. Meyer, J. Polym. Chem. A 45
(2007) 1–15;
(e) P. Metrangolo, T. Pilati, G. Resnati, CrystEngComm. 8 (2006) 946–947;
(f) P. Metrangolo, T. Pilati, G. Resnati, Acc. Chem. Res. 38 (2005) 386–395;
(g) P. Metrangolo, T. Pilati, G. Resnati, A. Stevenazzi, Curr. Opin. Coll. Interface Sci.
8 (2003) 215–222;
(h) P. Metrangolo, G. Resnati, Chem. Eur. J. 7 (2001) 2511–2519.
[6] (a) P. Metrangolo, Y. Carcenac, M. Lahtinen, T. Pilati, K. Rissanen, A. Vij, G. Resnati,
Science 323 (2009) 1461–1464;

A. Dey et al. / Journal of Fluorine Chemistry 130 (2009) 816–823
823
(b) P. Metrangolo, F. Meyer, T. Pilati, D. Proserpio, G. Resnati, Chem. Eur. J. 13
(2007) 5765–5772;
(c) R. Liantonio, P. Metrangolo, F. Meyer, T. Pilati, W. Navarrini, G. Resnati, Chem.
Commun. (2006) 1819–1821;
(d) A. Casnati, R. Liantonio, P. Metrangolo, G. Resnati, R. Ungano, F. Ugozzoli,
Angew. Chem. Int. Ed. 45 (2006) 1915–1918;
(e) G. Gattuso, R. Liantonio, P. Metrangolo, F. Meyer, G. Resnati, A. Pappalardo,
M.F. Parisi, T. Pilati, L. Pisagatti, G. Resnati, Supramol. Chem. 18 (2006) 235–243;
(f) H. Neukirch, E. Guido, R. Liantonio, P. Metrangolo, T. Pilati, G. Resnati, Chem.
Commun. (2005) 1534–1536;
J. Mol. Struct. 524 (2000) 87–94;
(o) A. Farina, S.V. Meille, M.T. Messina, P. Metrangolo, G. Resnati, Angew. Chem.
Int. Ed. Engl. 38 (1999) 2433–2436;
(p) A. Lunghi, P. Cardillo, M.T. Messina, P. Metrangolo, W. Panzeri, G. Resnati, J.
Fluorine Chem. 91 (1998) 191–194;
(q) V. Amico, S.V. Meille, E. Corradi, M.T. Messina, G. Resnati, J. Am. Chem. Soc. 120
(1998) 8261–8262.
[7] K. Reichenbacher, H.I. Suss, J. Hulliger, Chem. Soc. Rev. 34 (2005) 22–30.
[8] J.M.S. Skakle, B. Gojdkab, J.L. Wardell, Acta Cryst. Sect. E 62 (2006) o1001–o1002.
[9] (a) G. Gattuso, A. Notti, S. Pappalardo, M.F. Parisi, T. Pilati, G. Resnati, G. Terraneo,
CrystEngComm. (2009), doi:10.1039/b904811c;
(b) S. Biella, G. Gattuso, A. Notti, P. Metrangolo, S. Pappalardo, M.F. Parisi, T. Pilati,
G. Resnati, G. Terraneo, Supramol. Chem. 21 (2008) 149–156.
[10] (a) C. Marega, A. Marigo, V. Causin, V. Kapeliouchko, E. Di Nicolo, A. Sanguineti,
(g) P. Metrangolo, T. Pilati, G. Resnati, A. Stevenazzi, Chem Commun. (2004)
1492–1494;
(h) M. Amati, F. Lelj, R. Liantonio, P. Metrangolo, S. Luzzati, T. Pilati, G. Resnati, J.
Fluorine Chem. 125 (2004) 629–640;
`
(i) R. Liantonio, P. Metrangolo, T. Pilati, G. Resnati, A. Stevenazzi, Cryst. Growth
Des. 3 (2003) 799–803;
(j) R. Liantonio, P. Metrangolo, T. Pilati, G. Resnati, Cryst. Growth Des. 3 (2003)
355–361;
Macromolecules 37 (2004) 5630–5637;
(b) M. Sprik, U. Ro¨thlisberger, M.L. Klein, J. Phys. Chem. B 101 (1997) 2745–2749.
[11] G.M. Sheldrick, SHELXL-97. Program for the Refinement of Crystal Structures,
University of Go¨ttingen, German, 1997.
(k) R. Bertani, P. Metrangolo, A. Moiana, E. Perez, T. Pilati, G. Resnati, L. Rico-
Lattes, A. Sassi, Adv. Mater. 14 (2002) 1197–1202;
[12] M.T. Messina, P. Metrangolo, S. Pappalardo, M.F. Parisi, T. Pilati, G. Resnati, Chem.
Eur. J. 6 (2000) 3495–4350.
(l) W.M.T. Navarrini, P. Metrangolo, T. Pilati, G. Resnati, New J. Chem. 24 (2000)
777–780;
[13] M.C. Burla, M. Camalli, B. Carrozzini, G.L. Cascarano, C. Giacovazzo, G. Polidori, R.
R. Spagna, J. Appl. Cryst. 36 (2003) 1103–1112.
(m) P. Cardillo, E. Corradi, A. Lunghi, S.V. Meille, M.T. Messina, P. Metrangolo, G.
Resnati, Tetrahedron 56 (2000) 5535–5550;
(n) M.T. Messina, P. Metrangolo, W.M.T. Navarrini, S. Radice, G. Resnati, G. Zerbi,
[14] S. Buscemi, A. Pace, A. Palumbo Piccionello, S. Pappalardo, D. Garozzo, T. Pilati, G.
Gattuso, A. Pappalardo, I. Pisagatti, M.F. Parisi, Tetrahedron Lett. 47 (2006) 9049–
9052.