Article
and polarization effects using Bruker AXS SAINT software^33. Absorption
corrections were applied using SADABS^34. Initial evaluation of the dif-
fraction data suggested that Zr 6 O 4 (OH) 4 (bpydc) 6 undergoes a change
of space group from Fm 3 m to P 21 3 (no. 225 and 198, respectively) upon
loading with NiBr 2 , NiCl 2 , CoCl 2 , or FeCl 2. Based on previous work^16 ,
attempts to solve and refine these structures in P 21 3 resulted in unsat-
isfactory refinement, thus solution and refinement in the space group
Pa 3 (no. 205) was instead attempted. In the end, the latter space group
gave the most satisfactory refinement. The structure was solved using
direct methods with SHELXS^35 ,^36 and refined using SHELXL^37 operated
in the OLEX2^38 interface. No significant crystal decay was observed
during data collection. Thermal parameters were refined anisotropi-
cally for all non-hydrogen atoms. Hydrogen atoms were placed in ideal
positions and refined using a riding model for all structures. Moving
from Fm 3 m to Pa 3 results in two twin domains related by the lost mir-
ror symmetry along the body diagonals of the unit cell. Consequently,
a twin law (TWIN 0 1 0 1 0 0 0 0 −1 2; BASF ≈ 0.50) was required for the
structural refinement.
The metal–organic framework Zr 6 O 4 (OH) 4 (bpydc) 6 is derived from
Zr 6 O 4 (OH) 4 (bdc) 6 or UiO-66, which has been known to have structural
defects where some of the linkers are absent^39. Therefore, the linker
occupancies in all structures were allowed to refine freely, resulting
in occupancies that range from 76.8% to 100%. When the ligand is not
present, water/hydroxide is known to replace it in the cluster^32. These,
however, could not be modelled in the structure due to their disorder
and low occupancy. The oxygen atoms of the oxo and hydroxo groups
on the zirconium clusters in the structure were disordered and, in cases
where this disorder could be modelled, the site occupancy factors of
these oxygen atoms were fixed to give a chemical occupancy of 50%.
Hydrogen atoms on the hydroxo groups could neither be found nor
placed and were omitted from the refinement but not from the formula.
Disorder of the linkers and the metal halides in some of the structures
required the use of geometric and displacement parameter restraints.
Voids in the structures that result from disordered solvent that could
not be modelled, large anisotropic displacement parameters that result
from linker and solvent disorder, and, in some cases, low data resolu-
tion gave rise to several A and B level alerts from checkCIF. Responses
addressing these alerts have been included in the crystallographic
information files (CIFs) and can be read in reports generated by check-
CIF. Extensive solvent disorder was found in the pores for most of the
structures and could not be modelled. Consequently, the unassigned
electron density in these structures was accounted for using SQUEEZE^40
as implemented in the PLATON^41 interface.
Powder X-ray diffraction
Powder X-ray diffraction patterns were collected on microcrystalline
powder samples of 1 (FeCl 2 ) 19 , 1 (CoCl 2 ) 18 , 1 (NiCl 2 ) 15 and 1 (NiBr 2 ) 15 , which
were loaded into 1.0 mm boron-rich glass capillaries inside a N 2 -filled
glovebox and then flame-sealed. High-resolution synchrotron X-ray
powder diffraction data were subsequently collected at 298 K with a
wavelength of 0.45220 Å at beamline 17-BM-B at the Advanced Photon
Source at Argonne National Laboratory. For all samples, a standard
peak search, followed by indexing through the Single Value Decompo-
sition approach^42 , as implemented in TOPAS-Academic^43 , allowed the
determination of approximate unit cell parameters. Analysis of the
patterns of all samples led to the assignment of the space group Pa 3
on the basis of systematic absences. The unit cells and space group
were verified by structureless Pawley refinements. In 1 (NiCl 2 ) 15 and
1 (NiBr 2 ) 15 it was observed that there was broadening of the hkl reflec-
tions arising from the lowering of symmetry to Pa 3 from Fm 3 m. Spe-
cifically, it was noted that the reflections corresponding to a mixture of
even and odd hkl values were broadened, such as the (0 2 1) and (2 1 1)
reflections at 2.2° and 2.4°. In the Pawley refinements, this broadening
could be modelled by defining one Lorentzian function for the peaks
corresponding to the Fm^3 m space group (all odd or all even hkl values)
and another Lorentzian convolution for the broadened peaks. Doing
so led to an excellent fit, and the parameters for the peak shapes were
implemented in later Rietveld refinements using the data of 1 (NiCl 2 ) 15
and 1 (NiBr 2 ) 15 , leading to improvements of ~8% and 4% in the weighted
profile R-factor (Rwp), respectively, when compared to refinements
performed without the broadening correction.
Subsequently, Rietveld refinements of all samples were attempted,
using the structural models determined by single-crystal X-ray diffrac-
tion as starting points. The atomic positions were initially not refined.
Occupancies of the metallated species (for example, Fe, Co, Ni, Cl and
Br) were allowed to vary relative to the full occupancy of Zr. These atoms
were also given isotropic atomic displacement parameters that were
individually refined. A separate occupancy factor and an isotropic
atomic displacement parameter were given to all atoms of the bipy-
ridine ligand, consistent with practices done for the single-crystal
structural model refinements. Finally, the cluster oxygen atoms were
given separate atomic displacement parameters and their occupancies
were refined relative to the Zr occupancy.
In the case of samples 1 (FeCl 2 ) 19 and 1 (CoCl 2 ) 18 , it was found that when
a Fourier difference map was generated from the single-crystal model
relative to the observed pattern, disordered electron density near the
extraneous Fe or Co species on the periphery of the sheet could be
observed. This disorder is postulated to arise from the presence of a
mixture of metal halide species and DME-solvated metal halide species
being present in the material, and was modelled in the single-crystal
model by reducing the occupancy of the chloride ligands of the periph-
eral species in order to accommodate partial oxygen occupancy. An
improved fit in both iron and cobalt cases was obtained when a DME
molecule was modelled as a rigid body and allowed to relax using a
simulated annealing approach, while keeping the rest of the structural
model constant. In both cases, a DME molecule could be found local-
ized near the peripheral metal species, with one of the oxygen atoms of
the DME molecule in bonding distance to the metal (~2.0 Å) and close
to a chloride position, consistent with the differences in typical bond
lengths between these ligands. Although disorder probably contrib-
utes to the high relative occupancies, the location is consistent with
unresolved electron density observed in the single-crystal models.
In the course of all refinements, the atom positions could not be
refined freely, as they resulted in chemically unreasonable positions
for numerous components of the structural model (particularly the
bipyridine linker). As a result, in the final stages of the refinement, soft
constraints were placed on the atomic positions (with the exception of
those for H, which were not refined). The thermal parameters, sample
and instrument parameters were then fit together with the background
parameters. The resulting calculated diffraction pattern for the final
structural models of 1 (FeCl 2 ) 19 , 1 (CoCl 2 ) 18 , 1 (NiCl 2 ) 15 and 1 (NiBr 2 ) 15 are
in excellent agreement with the experimental diffraction patterns
(Rietveld plots shown in Supplementary Figs. 5–8 and further crystal-
lographic details given in Supplementary Tables 10 and 11).
Finally, the refined occupancies of the metal(ii) halide sheet atoms
(Fe, Co, Ni, Cl, and Br) are within one standard error of the values
obtained by single-crystal X-ray diffraction of closely related sam-
ples, confirming that the structural models used are reasonable and
applicable to the bulk samples.Low-pressure gas adsorption measurements
Gas adsorption isotherms for pressures in the range 0–1.2 bar were
measured by a volumetric method using a Micromeritics ASAP2420
instrument. A typical sample, consisting of ~100 mg of material was
transferred to a pre-weighed analysis tube, which was capped with
a Micromeritics TranSeal and evacuated by heating at 120 °C for 1 or
80 °C for all samples loaded with metal(ii) halides at a ramp rate of
1 °C per min under dynamic vacuum until an outgas rate of less than
3 μbar min−1 was achieved. The evacuated analysis tube containing the
degassed sample was then carefully transferred to an electronic balance