research papers
the additional 10 min of reaction time, a comparative PXRD
study was performed as shown in Fig. S1 (all tables, figures and
sections with the prefix ‘S’ are available in the supporting
information). The powder diffraction patterns were identical,
except for a slight difference in the absolute intensities of the
peaks, suggesting improved crystallinity for the product
obtained after 15 min of reaction. These PXRD patterns were
also used in a search–match procedure to identify a possible
impurity or unexpected solid in the reaction products. The
search–match procedure using the 2013 version of the Inter-
national Centre for Diffraction Data, Powder Diffraction File
(PDF-2) database yielded no phases consistent with the
observed powder pattern. A search for the crystal structure of
divanillin or related compounds in the 2018 edition of the
Cambridge Structural Database (CSD; Groom et al., 2016),
yielded three divanillin derivatives with the same C and O
substituents in the biphenyl ring system, but also additional
moieties. Moreover, a search in CAS SciFinder did not result
in any crystallographic characterization of this material. A
peak extraction procedure performed from the 15 min powder
pattern using Winplotr 2004 (Roisnel & Rodriguez-Carvajal,
2012) afforded 20 peaks that were indexed with DICVOL04
lographic study and the impossibility of obtaining suitable
single crystals for structural analysis motivated us to pursue
the crystal structure determination of divanillin from powder
diffraction data.
Divanillin recrystallized from DMSO as a powder showed
the same diffraction pattern but higher crystallinity than the
original 5 and 15 min samples; therefore, a batch of this solid,
thoroughly ground, was used for high-quality powder
diffraction data collection using a Rigaku ULTIMA IV
powder diffractometer. Data collection details are summar-
ized in Table 1 and Section S2. Unit-cell refinement with the
Le Bail method using the full high-quality pattern confirmed
the previous unit cell, and the determination of systematic
absences suggested the space groups Pba2 (No. 32) or Pbam
(No. 55), applying the unit-cell transformation c,ꢄb,a to the
3
˚
above unit cell. Considering 18 A as the average atomic
volume per non-H atom in organic molecules and the above
unit-cell volume, an estimated value of Z = 2 was obtained.
Taking into account the internal twofold symmetry of the
divanillin molecules, the space group Pba2, with four half-
molecule fragments per unit cell and Z0 = 21, was deemed most
probable and crystal structure determination was first
attempted using the latter space-group symmetry. Crystal
data, data collection and structure refinement details are
summarized in Table 1.
¨
(Boultif & Louer, 2004), to obtain an initial unit cell in the
orthorhombic system, with a = 3.9344, b = 12.2995, c =
3
˚
˚
13.9841 A and V = 676.707 A , with figures of merit M(20) =
81.8 and F(20) = 85.5. This unit cell was confirmed by fitting
the powder pattern using the Le Bail method (Le Bail et al.,
1988) implemented in the GSAS/EXPGUI software suite
(Larson & Von Dreele, 2004; Toby, 2001) using both the 5 and
the 15 min powder patterns. The confirmation of a single-
phase divanillin powder, the lack of a previous crystal-
2.3. Crystal structure determination, Rietveld refinement and
DFT-D simulation
Crystal structure determination was simultaneously
attempted using direct methods implemented in the software
EXPO2013 (Altomare et al., 2013) and a direct-space
approach using the software WinPSSP (Pagola et al., 2017) in
1
the space group Pba2 and with Z0 = . The chemical formula
2
and powder diffraction data were input into EXPO2013,
allowing it to perform peak-search, indexing, space-group
symmetry and crystal structure determination using the
default calculation options. Unit-cell determination confirmed
the previously indicated space group. The crystal structure
determination algorithm detected the possible presence of
preferential orientation effects in the pattern and successfully
used this information, finding a structural model with all non-
H atoms located and the expected atomic connectivity, after
one run using default options and one EXTRA recycling of
the initial solution. WinPSSP was fed with a planar model of
half a divanillin molecule (one phenyl ring with –OCH3,
–CHO and –OH substituents) and was able to determine its
position in the unit cell with enough precision to afford a
correct crystal structure refinement; however, the starting
model before Rietveld refinement was not as good as that
from EXPO2013, likely due to the lack of a preferred orien-
tation correction. Both equivalent structural models were
refined to the same final model within the s.u. values reported
by the Rietveld method (Loopstra & Rietveld, 1969) using the
GSAS/EXPGUI suite. A linear interpolation function with 36
parameters was used to describe the background, while a
pseudo-Voigt function corrected for axial divergence was used
Figure 2
A divanillin molecule as obtained from the Rietveld fit. Note the atom-
numbering scheme. Atoms are represented as spheres of arbitrary radii.
The twofold axis between C5 and C5i is shown as a green line. [Symmetry
code: (i) ꢄx + 1, ꢄy + 1, z.]
ꢂ
1770 Imer et al. Interpenetrating hydrogen-bonding 3D networks in divanillin
Acta Cryst. (2018). C74, 1768–1773