Single-step synthesis and interface tuning of core–shell metal–organic framework nanoparticles†

Control over the spatial distribution of components in metal–organic frameworks has potential to unlock improved performance and new behaviour in separations, sensing and catalysis. We report an unprecedented single-step synthesis of multi-component metal–organic framework (MOF) nanoparticles based on the canonical ZIF-8 (Zn) system and its Cd analogue, which form with a core–shell structure whose internal interface can be systematically tuned. We use scanning transmission electron microscopy, X-ray energy dispersive spectroscopy and a new composition gradient model to fit high-resolution X-ray diffraction data to show how core–shell composition and interface characteristics are intricately controlled by synthesis temperature and reaction composition. Particle formation is investigated by in situ X-ray diffraction, which reveals that the spatial distribution of components evolves with time and is determined by the interplay of phase stability, crystallisation kinetics and diffusion. This work opens up new possibilities for the control and characterisation of functionality, component distribution and interfaces in MOF-based materials.


Contents
S6 Synthesis-structure prediction maps relating two-phase model refinements s32 S7 Composition gradient refinement fits to in situ XRD data s34 S8 In situ XRD data for Zn/Cd ZIF-8 crystallisation above 25 • C s35 S9 Kinetic analysis of parent Zn-and Cd-ZIF-8 crystallisation from in situ XRD data s39 S10 Bibliography s41 s2 S1 Experimental methods S1.1 Synthesis Zn 1−x Cd x (mIm) 2 samples were prepared using a modification of a previously reported solvothermal route. [1] The following solutions were prepared in methanol (Sigma Aldrich, ≥ 99 %): (i) 0.1 M Zn(NO 3 ) 2 ·6H 2 O (Sigma Aldrich, ≥ 99 %) (ii) 0.1 M Cd(NO 3 ) 2 ·4H 2 O (Sigma Aldrich, 98 %) (iii) 0.8 M 2-methylimidazole (HmIm) (Sigma Aldrich, 99 %), 0.8 M triethylamine (TEA) (Alfa Aesar, 99 %). Solutions containing the Zn and Cd salts were combined with the starting Cd mole fraction, x rxn , varied between 0 and 1 in 0.1 intervals to give a total volume of 5 cm 3 in 20 cm 3 PTFE autoclave liners. 5 cm 3 of the HmIm/TEA solution was added and the reaction mixture was stirred at 500 rpm for two minutes at room temperature. The liners were transferred to autoclaves and heated to a range of temperatures (20 • C-100 • C in 10 • C intervals) in an oven for 24 hr. The heating and cooling rates for all reactions were 200 • C/hr and 80 • C/hr respectively. The resulting solids formed suspensions of varying stability depending on the composition and synthesis conditions; in general the solid sedimented more quickly as Cd content and synthesis temperature increased. These suspensions were centrifuged at 9500 rpm for 10 min, the supernatant discarded and the sediments resuspended in fresh methanol to wash. In total, three cycles of washing were completed before samples were dried overnight in a vacuum oven at room temperature.

S1.2 X-ray diffraction
High resolution ex situ XRD patterns were obtained at the I11 beamline at the Diamond Light Source, UK. [2,3] Samples were finely ground and packed into 0.5 mm external diameter borosilicate capillaries (Capillary Tube Supplies Ltd.). Patterns were measured using a position sensitive detector (PSD) made up of Mythen-2 Si µ strip modules. Wavelengths used were 0.82460, 0.82444, 0.82503, and 0.82506Å, depending on beamtime dates. In situ PXRD measurements were taken at the I12 beamline at the Diamond Light Source, UK. [4] Monochromatic X-rays of wavelength λ = 0.22946Å were used, and patterns were recorded on a Pilatus 2M CdTe detector. Reactions for x rxn = 0.5, as well as the pure Zn and Cd end-members, were performed at synthesis temperatures between 25 • C and 65 • C in 10 • C steps, whilst high energy synchrotron XRD patterns were collected simultaneously and subsequently summed to give 20 s temporal resolution. Reactions were contained in culture tubes (Duran™) held at constant temperature in an Al heater block under constant stirring, with a thermocouple to measure the solution temperature. All solutions were prepared the same as for ex situ reactions. For each reaction, 1.5 cm 3 of the HmIm/TEA solution was injected into 1.5 cm 3 of the metal solution using a syringe pump. Pawley refinements [5] were implemented in TOPAS Academic (version 6.0). [6] Single phase refinements were carried out initially using a symmetric Thompson-Cox-Hastings-Pseudo-Voigt (TCH-PV) peak shape. [7] Two-phase refinements were performed by fixing the ratio of the intensities for peak hkl to h k l for both phases so the relative intensities of the peaks within each phase were consistent. To a first approximation this holds true because the variation in composition between the two isomorphic phases is small (∆x ≈ 0.2). Split peak refinements were implemented using the split Pseudo-Voigt (spv) peak shape, in which peak profiles are split at the maximum peak intensity and broadening is modelled as a crystallite size effect according to the Scherrer equation. We define h as the difference between the crystallite sizes, i.e., coherent scattering lengths, of the high-and low-angle components. Composition gradient refinements were performed as detailed in the main text, using 50 "phases", evenly spaced at r = 0.02, 0.04, ..., 0.98, 1. The scattering factor of each shell is weighted by the volume of its surface area element (Aδr ∝ r 2 ) and squared again to give its intensity contribution to the XRD data. We convolute a fixed contribution from the instrumental line broadening with sample broadening from the "phases" via the Crystallite Size macro in Topas Academic V6 [6] , using a single refinable term, D, according to the expression lor f whm = 0.1(180/π)(λ/D cos θ). D was not refined for in situ data because the low signal-to-noise ratio rendered the refinement unstable.

S1.3 Electron microscopy
An FEI Osiris microscope (Thermo Fisher) equipped with a high-brightness X-FEG electron source operated at 80 kV was used to acquire STEM data using a beam convergence semi-angle of 11.0 mrad. EDS measurements were collected by a Super-X detector system with four detectors symmetrically mounted about the optic axis. The beam current used was approximately 150 pA. The pixel size was 1 nm-2 nm and the dwell time per pixel was 120 ms. Under these conditions, the MOF crystals undergo loss of crystal structure and ionization damage. However, minimal migration of metal ions is observed in MOFs under these conditions [8] borne out by systematic differences observed between samples prepared under controlled synthetic parameters. STEM images were collected before, during, and after EDS acquisition using an annular dark field detector, in order to allow for the correction of the data for sample stage drift during STEM-EDS map aquisition.
Data were processed using the Hyperspy [9] open source software. Integration of K α (Zn) and L α (Cd) X-ray emission lines generated the EDS maps, which were corrected for sample drift using image registration routines in Matlab (Mathworks). The Cliff-Lorimer 'k-factor' approach [10] was implemented for quantification using constants supplied by Bruker (the detector manufacturer) with background subtraction by linear interpolation between adjacent energy windows applied to the selected X-ray lines. Due to errors in k-factors, especially for the L α emission lines, it is known that 10 % uncertainty or greater may be expected in quantification results. [11] However, the precision recorded in changes between a systematic series of similar samples is instead limited by counting statistics due to the particle size, electron beam current, and detector collection efficiency. This precision is estimated from the local fluctuations in composition as approximately 5 %. The maps and line profiles, which were extracted along paths traversing single particles identified from annular dark field STEM imaging, therefore reliably track changes in the projected core-shell composition under different synthesis conditions. It should be noted that the two-dimensional maps can be considered as projections of the three-dimensional composition along the electron beam trajectories (i.e., the map pixels). Therefore, core compositions are underestimated relative to the physical three-dimensional composition.
Particle size analysis was carried out using ImageJ software. One measurement of particle 'diameter' was taken for each discernible primary particle as a line from edge to edge across the centre of the projection of the particle in the ADF-STEM image. Care was taken to ensure this measurement was an intermediate estimate between short and long dimensions, a suitable choice to capture trends in size distributions given the particles were characterised by convex and compact features with minor deviations in aspect ratio.

S1.4 Nuclear magnetic resonance spectroscopy
Solid-state NMR spectra were recorded using a Bruker Avance III spectrometer equipped with a 9.4 T wide-bore superconducting magnet, giving Larmor frequencies of 400.1, 100.6 and 40.5 MHz, respectively, for 1 H, 13 C and 15 N. Samples were packed into standard 4 mm zirconia rotors and rotated at magic angle spinning (MAS) rates of 5 kHz ( 15 N) or 12.5 kHz ( 13 C) using a Bruker "low-γ" double-resonance probe. 13 C NMR spectra were recorded with cross polarisation (CP) from 1 H, with a contact pulse (ramped for 1 H) of 2.5 ms and signal averaging for 256-1024 transients with a recycle interval of 3 s. 15 N CP MAS NMR spectra were recorded with a contact pulse (ramped for 1 H) of 5 ms and signal averaging for 1024-22528 transients with a recycle interval of 3 s. High-power TPPM-15 decoupling of 1 H was carried out during acquisition. Chemical shifts are reported relative to (CH 3 ) 4 Si using l-alanine as a secondary reference (δ(CH 3 ) = 20.5 ppm) for 13 C and to CH 3 NO 2 , using 15 N-enriched glycine as a secondary reference (δ(NH 3 ) = −347.4 ppm) for 15 N.

S2 Refinements to high-resolution ex situ synchrotron XRD data
Representative fits to the XRD data obtained using the two-phase, single phase split peak, and composition gradient models, are shown below for samples synthesised at T = 20 • C, 60 • C, and 100 • C; x rxn = 0.1, 0.5 and 0.9.   Single phase split peak model T = 20 ℃ x rxn = 0.5 Single phase split peak model T = 20 ℃ x rxn = 0.9 Figure S4: Single phase split peak fits to high-resolution XRD data for samples synthesised at T = 20 • C; x rxn = 0.1, 0.5 and 0.9. Observed, calculated and difference data are shown as black dots, and orange and red lines, respectively. s8 Single phase split peak model T = 60 ℃ x rxn = 0.1 Single phase split peak model T = 60 ℃ x rxn = 0.5 Figure S5: Single phase split peak fits to high-resolution XRD data for samples synthesised at T = 60 • C; x rxn = 0.1 and 0.5. Observed, calculated and difference data are shown as black dots, and orange and red lines, respectively.

S3 Solid state nuclear magnetic resonance
Solid state MAS NMR spectra were obtained for Zn 0.5 Cd 0.5 (mIm) 2 samples synthesised at 40 • C, 60 • C, 80 • C, and 100 • C ( Figure S10). The aromatic 13 C signals, while sensitive to the metals present, do not provide any information on ordering, with three very closely-spaced resonances for the CH species at 124.9, 124.6 and 124.2 ppm and a single quaternary resonance at 151.4 ppm. The latter is mid-way between 151.1 ppm for the Zn end-member and 151.7 ppm for the Cd end-member. Unfortunately, the CH resonances are not sufficiently resolved to allow unambiguous decomposition and integration of the signal, especially since there is a poorly-defined broad resonance in the baseline of all four samples, centred at ∼126 ppm, which further complicates the deconvolution. The CH 3 resonances (inset, Figure S10) are better resolved and, in all cases, have three components at 13.7, 14.0 and 14.4 ppm. However, interpretation of these signals is complicated by issues with decomposition (including a broader, low-intensity resonance at ∼14.6 ppm) and the fact that the chemical shifts of the CH 3 groups in these ZIFs are more sensitive to disorder and magnetic (i.e., crystallographic) inequivalence than to the nature of the metal atoms three bonds away. Therefore, in this case, it is impossible to assign these resonances directly to differences in chemistry or structure.
The 15 N NMR spectra ( Figure S10, right), are far more sensitive to the nature of the N-bound metal. In this case, broadened 15 N resonances are observed at -173.5(2) ppm (corresponding to N-Cd) and -176.3(2) ppm (N-Zn) for the mixed-metal ZIFs. In addition, a minor resonance at -171 ppm can be attributed to one half of the 111/113 Cd-15 N J-coupled doublet (J 200 Hz), with the other half overlapping with the N-Zn signal. The spectra are sufficiently resolved that they can be integrated to obtain approximate Cd/Zn ratios. These ratios are plotted in Figure S11 and it can be seen that the Cd content of the product decreases significantly with increasing synthesis temperature, as found by XRD and STEM-EDS. We note that, in this case, it is not possible to use 15 N NMR spectroscopy to investigate the ordering of the metal cations.   S6 Synthesis-structure prediction maps relating twophase model refinements x rxn x rxn Figure S30: Variation of Cd mole fractions of the two phases, denoted x core and x shell , with synthesis conditions determined using two-phase refinements of high-resolution synchrotron XRD data. Note the different scales. Grey regions correspond to the pure end-members for which a two-phase refinement is not physically meaningful. The white region indicates conditions outside the synthesis window for ZIF-8; black dots indicate the data points.
x rxn x rxn xrxn Figure S31: Variation of coherent scattering domain size of the two phases, denoted D core and D shell , and the difference between them, ∆D, with synthesis conditions determined using two-phase refinements of high-resolution synchrotron XRD data. Note the different scales. Grey regions correspond to the pure end-members for which a two-phase refinement is not physically meaningful. The white region indicates conditions outside the synthesis window for ZIF-8; black dots indicate the data points.  Figure S36: Temporal evolution of total Bragg peak intensity, I (beige), interface diffuseness, ν (turquoise), and nominal interface radius, r c (blue) for mixed Zn/Cd ZIF-8 crystallisation at 45 • C, x rxn = 0.5, determined using composition gradient model fits to in situ XRD data. The discontinuities around t = 180 s could not be accounted for in data processing and may be due to product falling out of the beam path.
s36 Figure S37: In situ synchrotron XRD data for mixed Zn/Cd ZIF-8 crystallisation at 55 • C, x rxn = 0.5. A rolling ball background subtraction has been performed for clarity.  Figure S38: Temporal evolution of total Bragg peak intensity, I (beige), interface diffuseness, ν (turquoise), and nominal interface radius, r c (blue) for mixed Zn/Cd ZIF-8 crystallisation at 55 • C, x rxn = 0.5, determined using composition gradient model fits to in situ XRD data.
s37 Figure S39: In situ synchrotron XRD data for mixed Zn/Cd ZIF-8 crystallisation at 65 • C, x rxn = 0.5. A rolling ball background subtraction has been performed for clarity.  Figure S40: Temporal evolution of total Bragg peak intensity, I (beige), interface diffuseness, ν (turquoise), and nominal interface radius, r c (blue) for mixed Zn/Cd ZIF-8 crystallisation at 65 • C, x rxn = 0.5, determined using composition gradient model fits to in situ XRD data. Note that the oscillations in Bragg peak intensity could not be accounted for in data processing and may be due to product falling in and out of the path of the beam.

S9 Kinetic analysis of parent Zn-and Cd-ZIF-8 crystallisation from in situ XRD data
Total integrated peak intensity data were extracted by Pawley fitting of in situ synchrotron XRD for the parent Zn-and Cd-ZIF-8 materials under identical conditions used for the mixed-component syntheses, at T = 25 • C, 35 • C and 55 • C. In order to compare the different reactions, the rate constant for crystal growth, k G , and growth exponent, n G , were refined to the data as a function of time, according to the Avrami-Erofe'ev expression, α = 1 − exp(−(k G × t) n G ), where 0 < α < 1 is the extent of crystallisation. [12][13][14] Figure S41: Fit of Avrami-Erofe'ev kinetics to total Bragg peak intensity against reaction time for the formation of Zn-ZIF-8, left, and Cd-ZIF-8, right, at 25 • C. Figure S42: Fit of Avrami-Erofe'ev kinetics to total Bragg peak intensity against reaction time for the formation of Zn-ZIF-8, left, and Cd-ZIF-8, right, at 35 • C. Figure S43: Fit of Avrami-Erofe'ev kinetics to total Bragg peak intensity against reaction time for the formation of Zn-ZIF-8, left, and Cd-ZIF-8, right, at 55 • C.