Free Standing Large Area Metal-Organic Nanosheets Encapsulating Periodically Spaced Single-Ion Magnets

The high demand for electronic components from limited natural resources is forcing manufacturers to focus on thin-lm technology but they are facing problems - limits of operation, uniformity of the lms and expensive manufacturing tools. Chemistry can provide solutions to the last two problems with ease of manufacture and uniformity of two-dimensional materials retaining all functionalities within one single layer. Here, we present one such layered compound, [Co II (terpy)(H 2 O) 2 @Co II2 (µ 2 -H 2 O) 3 {Ir III (ppy-COO) 3 } 2 ]·6H 2 O (1). Each charge-balanced layer is an anionic triple-ply holding [Co II (terpy)(H 2 O) 2 ] 2+ cations. Treating the crystals with warm acetone or by sonication breaks the weak interlayer supramolecular π-π and H-bonds leading to free standing single- and double layers (>30´30 µm 2 ). Each Co II (terpy)(H 2 O) 2 , at crystallographic positions 14.5 Å apart, behaves as a single-ion magnet (SIM). This is the rst observation of periodic positioning of SIM shielded from their neighbours amounting to a staggering 350´10 12 Co II /in 2 capable of functioning as a 350 TeraBit/in 2 magnetic memory device.


Introduction
There are considerable advantages in miniaturisation of electronic components -cost, light weight, energy consumption, increasing the time availability of a limited natural resource, etc…, while also balancing between socio-economic and geopolitical problems. 1,2 But there are limiting factors to contemporary materials, in particular those needed for memory devices, due to reliance on collective properties -electricity and magnetism -needing a large ensemble of atoms. 3 To overcome these problems manufacturers are tending to materials with properties controlled on a molecule basis -thus, only few atoms . [4][5][6] Present usable materials are built from the bulk, often metals or metal-oxides, and cut to the processable sizes and shapes using high-cost engineering tools and most importantly, there is considerable loss in cuttings. This process is often called top-to-bottom and it is physical. The size of each sensor is limited to ca. 25 nm in one direction and the spacing dominated by the width of the cutting tools, often that of the laser beam used in lithography. 7 Due to the control by chemistry in working from the bottom up, materials made of molecules are challenging the existing extended-solid materials for logical reasons -the device density will be increased, cost will be slashed, life-time of the resources will be extended, etc… -all due to the size reduction to 1 nm or less. 8 While these are possible using different synthetic chemical processes, it is hampered at the processing stage to organise them in periodic arrays required to ease the device manufacture and usage. 9 Thus, an in-between method is very appealing and consequently, many researchers have been focusing on materials that can be separated into sheets. 10,11 One of the most beautiful examples is the use of the Scotch-tape technique to separate natural graphite down to single layer graphene. 12 This was possible due to the moderate interlayer supramolecular π-π interaction energy. Layered materials have been converted into 2D-sheets by mechanical methods (e.g. sonication) and chemical methods through solvation, functionalisation, etc… [13][14][15][16] On this front there have been considerable developments involving graphene, 17, 18 graphene oxide, 19 hexagonal boron nitride, 20 metal oxides, 21 metal chalcogenides 22,23 and metal hydroxides. 24 The major problem of this approach is that the active component for the device is the structure forming unit -for a magnetic memory device it will be the metal ions used as the moment carrier. 25 So it still requires to be cut to sizes in excess of a magnetic domain. 26,27 That means we still need a cutter and inevitably suffer of loss in cuttings as well as requiring expensive tools. Facing the wall, this is where Chemists appreciate that accident happens and this serendipity leads to a novel material that one can then use to design others. Our work is one of these accidental discoveries while our original aim was focused on photocatalysis, magnetism and luminescence using ordered mixed Ir-Co alloy. 28,29 Here, we present such a case where the choice of the racemic mixture of a metalloligand [Iridium(III)tris(pyridine-benzoate), Ir(ppy-COOH) 3  between them in addition to existing π-π ones. When these crystals are treated with warm acetone or by sonication the weak interlayer supramolecular bonds are broken leading to free standing single-and double-layers. The exfoliated 2D-nanosheet (>30´30 µm 2 ) house single-ion magnetic (SIM) cations, Co II (terpy)(H 2 O) 2 , at crystallographic positions of 14.5 Å apart. This is the rst observation of regular periodic positioning of SIM shielded to their neighbours amounting to 350´10 12 Co II /in 2 or a potential 350 TeraBit/in 2 magnetic memory device. We propose further where it can lead to with some suggestions.

Results And Discussion
Syntheses and crystal structures. The structure can be broken down to its components to allow us, with additional chemical knowledge, to understand the solvothermal reaction leading to the formation of the compound. From a chemical reactivity point of view the meridian terpy is the most potent chelate and is expected to form rst; thus, possibly acts as template. The presence of a racemic mixture of D-and L-Ir(ppy-COOH) 3 is strictly required and the water molecules are as important in the formation of the crystal.
Several hexagonal single crystals of 1 ( Figure S1), suitable for diffraction intensity data collections gave similar lattice parameters and were isostructural. The experimental PXRD patterns of several batches match well with that calculated from the single crystal structure data ( Figure S2). There is no indication of any other crystalline material present. The copper(II) analogue (2) has been prepared and structurally characterised but the yield has been low to performed all the required measurements (Table S1).
The organometallic iridium centres adopt distorted-octahedral geometries [Ir-C: 1.998, Ir-N: 2.118 Å] (Table   S2), similar to those reported in other heterometallic Ir-M coordination polymers. 30 Figure 1). The resulting layer is neither chiral nor polar and a centrosymmetric cell is formed. Due to the crystal symmetry the carboxylate is coplanar with the phenylpyridine moiety. More interesting is that the cations sitting within the cavities of the layers are crystallographically disordered over three symmetry-equivalent positions. Therefore, it looks like a three-blade propeller with one unit consisting of one Co2, one terpy and two water molecules ( Figure S5). While Co2 is located at a special position with the ve sites occupied by three nitrogen atoms as meridian and two water molecules (Co-N, 1.76 -1.81 Å; Co-O, 2.56 Å). The unique geometry of [Co II (terpy)(H 2 O) 2 ] 2+ cations compared to other previously reported pentagonal compounds can be attributed to high symmetry and limited space. 31 Each cavity houses one unit in a triangular array at a centre-to-centre distance of the crystallography a-axis of 14.5 Å (Figure 1).
The [Co II (terpy)(H 2 O) 2 ] 2+ cations have no direct chemical bond with the layers, and are only locked in position by two weak C···O (pyridine to carboxylate) H-bonds of 3.12 (C14-H14···O1) and 3.35 Å (C15-H15···O1). Because each anionic layer consists of edge-sharing bipyramids, it has apices and troughs and for e cient packing the apices of adjacent layers are translated in the ab-plane to t in the troughs of the centre one ( Figure S6). Consequently, the structure adopts an ABABA sequence with two layers per unit cell along the c-axis. The layers are locked to each other via a bond-over-ring π-π interaction between adjacent phenyl groups, characterized by C···C distance as short as 3.48 Å ( Figure S7). A similar picture to graphite though limited in numbers. The remaining space between the layers is occupied by an octahedron of water molecules with a short (3.09 Å) and a long (3.37 Å) contacts. They form two carboxylate O···O contacts (3.61 and 3.64 Å) and one with the phenyl C···O (3.56 Å).
The chemistry of 2D-materials is dominated by two families; those consisting of neutral or charged layers. 32 For both families, solvents can be intercalated or exchanged. It all depends on the balance of electrostatic energies of the host and the intercalant. Additionally, the layer is also amenable to doping with metal-cations or metal-complexes or halogen anions to tune their electronic properties. For example ionic layered materials, such as 2D metal-oxalate network, the compensating charged ions between the sheets can be exchanged by functional ones. 33,34 Consequently, there is only one 2D material reported that has charge-compensated cations in cavities within its two-dimensional network, and thus allowing free standing neutral monolayer. 35 As mentioned by the authors, this type of network can be easily exfoliated into atomically-thin layers by using the Scotch tape method. Since our compound is the second one, as well as being the rst with magnetic cation, this report prompted us to explore the production of 1 monolayer with a periodic array of a potential SIM. Following literature reported techniques, we rst used ultrasonic treatment which operates by local microscopic high-temperature heating to break up the layered structure and produce single-or few aggregated-sheets. 36,37 The successful exfoliation was demonstrated by the observation of Tyndall effect of a laser pointer through acetone suspension ( Figure 2) over long periods. Indeed, the uniform sheet morphology of 1 is supported by a SEM and TEM images (Figures 3a-3b, S8). Atomic force microscopy (AFM) images reveal that the sheets are quite uniform with a lateral dimension up to 1.5 μm and thickness of 2.5 ± 0.2 nm (Figure S9), which is equivalent to that of a monolayer (2.2 nm) in the single crystal. However, conventional ultrasonic exfoliation damages the planar structure of the sheets and results in large amount of fragmentation with a lateral dimension of less than 2 μm or riddled with holes. So, it is di cult to produce sheets for making devices. To address these problems, soft-physical processes, such as freeze-thaw method 38 and solvent-induced delamination, 39,40 was used to prepare tens of micron-scale sheets. The latter gives the most uniform, complete and large area single-and double-sheets. AFM analyses on samples of different sizes range from freeze-thaw method using acetone as supporting solvent suggest different distributions of multilayers segments where the majority consists of doublelayer (3.7 ±0.2 nm, experimental thickness is 3.5 nm, Figure S11). We hypothesize that the penetration of acetone between the layers impose different constraints on the structure leading to a preferential exfoliation of the crystal into double-sheets. The challenge to exfoliate bulk samples into a single layer structure by this soft physical method remains a real challenge to understand given the balance of external forces versus that of the intrinsic π-π and H-bonding interactions between the sheets. Unlike ultrasonic exfoliation to lateral sizes of 2 µm, sheets of up to 30 μm can be obtained by this nonmechanical method ( Figure S11). By increasing the number of cycles, further exfoliation follows to bilayers and eventually monolayers (Figures 3c and S12). A homogeneous single-layer state is achieved after 90 repetitions, but at a cost of reducing the lateral dimensions (Figures 3d and 3e).
Going to the softest non-mechanical method where the crystals (average lateral size of 200 µm) were suspended in acetone and applying a gentle heat to 35 °C for few days and we found the Tyndall effect is enhanced with time. AFM shows principally double-layer segments with very clean surfaces extended over 30 µm (Figures 3f and S13). The quantity of single-layer sheets is low and this can be increased by shaking the suspension using a vibrating-mixer with little damage to the double-layer segments. The high-quality double-layer sheet obtained by self-delamination has rarely been reported. 41 The PXRD pattern of the as-synthesised crystals shows the Bragg re ections expected from the crystal structure data. No other re ection was observed suggesting the crystals were the only diffracting material and the 2D material maintains good stability during the exfoliation process ( Figure 4). Interestingly, the PXRD pattern of the exfoliated sample displays dominant re ections at 2θ = 6.26, 12.47 and 18.66°w hich are assigned to the (002), (004) and (006) re ections, suggesting high preferential orientation which can only be from the re-structuration upon drying. From the width of the Bragg re ections, we can assume the ordering is long-range after re-structuration. Moreover, the morphology and the multilayer crystalline structure are retained in the residual particles formed during the preparation of 1-NS by different methods (Figure S14-S18).

Magnetic properties
The temperature dependence of the magnetic moment is presented as the product of susceptibility and temperature (c M T) in gure 5a. The data can be discussed as being composed of three parts and since the two different cobalt atoms are quite far apart and not connected through bonds we can consider the total moment as being the sum from these two parts. On lowering the temperature from 300 to 150 K a gradual drop of the magnetic moment from 0.82 to 0.42 cm 3 K/mol is observed. This part is associated to the face-sharing Co 2 (µ 2 -OH 2 ) 3 dimer, that is Co1, which is appropriate for very strong antiferromagnetic exchange through the three oxygen atoms of the water molecules. Consequently, the value of 0.82 cm 3 K/mol at 300 K is only one tenth of ca. 7.5 cm 3 K /mol expected.
The second part, a constant plateau at ca. 0.4 cm 3 K/mol, is valued to an anisotropic S = ½ moment carrier. We associated this part to the [Co II (terpy)(H 2 O) 2 ] 2+ and assume from known examples in the literature that this ve-coordinated cation stabilises in the Kramers S = ± ½ state due to a very strong axial anisotropy (D). 31 This is projected in the observation of EPR signals starting ca. 100 K and increasing in intensity down to 5 K, as well as tting quite well to an anisotropic g-tensor of low-spin Co II where simulation gives g x = 2.208(2), g y = 2.132(2), and g z = 2.017(2) (Figure 5b). 42 The third part is concerned with the increase below 50 K. Its gradual rise to a peak corresponds to a magnetic long-range ordering (LRO) of an impurity, whose quantity is very small (see below) considering the value of 0.8 cm 3 K/mol at the peak. The LRO is further characterized by bifurcation in the ZFC-FC magnetisation in a eld of 10 Oe and the presence of both ac-susceptibility components independent on frequency ( Figure S20). EPR spectroscopy shows a broad signal for all frequencies at 300 K at g = 2.082(2) ( Figure S21). Its width originating from magnetic coupling and weak dependence of temperature are consistent with a strongly coupled magnetic impurity.
Although the crystals appear very clean under the microscope and by PXRD, the magnetic properties suggest a small amount of a highly magnetic material is present. As mentioned in the synthetic section, a green coloured powder appeared before the reaction mixture was subjected to solvothermal reaction. The high pH condition of the reaction should be the reason which may lead to magnetic layered hydroxide specially in the presence of an organic carboxylate. 43 The green colour of the particles is known to originate from structures containing both octahedral and tetrahedral coordinated cobaltous ions. 44 Surprisingly, both SEM and PXRD measurements do not show two phases. The magnetic properties of the green powder, isolated before the solvothermal treatment, exhibit long-range-ordering characteristics of the ferrimagnetic material and also con rms the above hypothesis ( Figure S22). Comparison of the moment in 1 to those of two samples of the green impurity those of collected before hydrothermal treatment gives an estimate of less than 1/500 in weight.
Furthermore, ac-susceptibility measurements in applied elds suggest an underlying moment displaying slow magnetic relaxation ( Figure S23). As the dimer will be silent at low temperatures due to it being in the S = 0 ground-state and the magnetic impurity will be fully saturated in a eld of 3 kOe, the only varying ac-susceptibility can only originate from superparamagnetism (Figure 6a). This can only come from [Co II (terpy)(H 2 O) 2 ] 2+ within the cavities. We believe their segregation within each layer diminish any exchange between nearest neighbours so that the frequency dependence can be observed. The presence of both ac-susceptibility components are exacerbated by the Cole-Cole behaviour when plotted in an Argand diagram (Figure 6b). This behaviour persists up to 8 K and analyses of the temperature dependence data using standard procedure gave a barrier to magnetisation reversal of 28 K and a relaxation time of 3.72´10 -5 s ( Figure S24). If one consider that each [Co II (terpy)(H 2 O) 2 ] 2+ is behaving as a memory site within the layer, this particular free-standing layer is one with the highest ever density known.
A calculation using the crystallographic dimension of 14.5 Å spacing gives a staggering 350´10 12 Co II /in 2 , translating to a potential 350 TeraBit/in 2 magnetic memory device. All the more important, the ac-susceptibility of a sample consisting principally of double layers would also exhibit similar results (Figures 6c, 6d and S25).

Conclusions
In self-assembling paramagnetic cobalt(II) with the metalloligand [Ir(ppy-COOH) 3 ] in the presence terpyridine to study the combine luminescence and magnetic properties and catalytic effect of Co-Ir, a unique layered compound is obtained which is easily delaminated to double and single nanosheets. Its uniqueness is in its content, which consists of anionic tri-ply, Ir-Co-Ir, encapsulating the cation [Co(terpy) (H 2 O) 2 ] 2+ within its cavities. Applying three different methods -ultrasonication, freeze-thaw, and soaking in acetone -the layers were separated to nanosheets of lateral dimension exceeding 30×30 µm 2 . Each cation behaves as an independent magnet -single-ion magnet (SIM) -due to the negligible dipolar interaction favoured by their periodic spacing of 14.5 Å apart. Each layer is ordered like bottles in a rack.
This amounts to 350 SIM/in 2 ; each relaxes in a time of 3.72×10 -5 s. While the density compares to ~ 300 times the currently available maximum density, the technological advances on the read/write head are lagging behind. Assuming a collection of say 5×5 SIM is used per bit it should be possible to theoretically put over ten pocket hard discs on the current market into a matchbox sized device.

Methods
Materials and physical measurements. All starting materials were of reagent grade quality and were obtained from commercial sources without further puri cation. Powder X-ray diffraction (PXRD) data were recorded on a Bruker D8 ADVANCE X-ray powder diffractometer (Cu-K a ) at room temperature.
Infrared spectra were measured as KBr pellets on a Bruker Tensor 27 spectrometer in the range of 400 -4000 cm -1 . Thermogravimetric analyses (TGA) were performed under a ow of nitrogen in the temperature range 25 -600 °C at a heating rate of 5 °C / min using a METTLER TOLEDO TGA/DSC 1 instrument. Elemental analyses for C, H and N were obtained with a Perkin Elmer 240C elemental analyser. Atom force microscope (AFM) measurements were carried out on VEECO Inc Multimode V. Scan electron microscopy (SEM) was performed on SHIMADZU SSX-550. The magnetic data were obtained on polycrystalline and powder samples using a Quantum Design SQUID VSM system. The diamagnetic contributions of the samples were estimated from Pascal's constants. 45 Synthesis of [Co II (terpy)(H 2 O) 2 @Co II 2 (H 2 O) 3 {Ir III (ppy-COO) 3 } 2 ]·6H 2 O (1). Ir(ppy-COOH) 3 (0.04 mmol, 31.5 mg) was suspended in 12 mL of distilled water and dissolved by adjusting the pH to 12.5 with 6 M aqueous NaOH. Then terpyridine (0.03 mmol, 7 mg) in 0.4 mL MeOH was added followed by 0.08 mmol Co(ClO 4 ) 2 ·6H 2 O in 4 mL water. An immediate precipitation of a green powder appeared. After keeping the vial at 100 °C for 2 days, red block crystals were obtained. The crystalline product of 1 was separated by ltration and washed with distilled water ten times to completely remove the residual ligand and metal salt. Anal  Figure S26). Thermal analysis reveals that the water molecule can be completely removed below 140 °C, corresponding to the release of nine (25 -80 °C) and two (80 -140 °C) water molecules ( Figure  S27). The total weight loss is 9.87 %, in agreement with the removal of six lattice and ve coordinated water molecules (calcd. 9.10 %). According to this result, the elemental analysis sample may have lost three water before tested, and the results are consistent with formula C 87 H 69 N 9 Co 3 Ir 2 O 20 (calcd. (%) C, 49.24; H, 3.28; N, 5.94).
Exfoliation into sheets. For all the exfoliation experiments the following was adopted. Crystals of 1 (2 mg) were suspended in acetone (4 mL) followed by different methods of exfoliation and nally separated from the large particles by centrifugation at 14,000 rpm for 15 minutes and further free-standing the colloidal suspension for at least 2 days before use. Three methods with different degrees of energy for the exfoliation were applied, (a) strong: treatment in an ultrasonic bath (300 W) at room temperature for 3 hours; (b) mild: the crystals in acetone were frozen in a liquid nitrogen bath (-196 °C) until solid followed by immediate heated in a warm water bath (55 °C) for 5 min and repeated for the desired number of times, and (c) soft: keeping in warm acetone at 35 °C for several days leads to low quantity of large double-sheets and using a vibrating mixer increases the quantity. The suspensions of 1-NS in acetone were deposited by being drop-cast onto mica substrates (20 μL), and carbon-coated copper grids (5 μL) for AFM and TEM measurements, respectively. Considering the e ciency of preparation, the nanosheet samples which prepared by freeze-thaw with 20 cycles were collected for other measurements.
Single-crystal structure determination. Crystals were selected for indexing and intensity data collection on a Bruker SMART APEX CCD diffractometer using graphite-monochromated Mo Kα radiation (λ = 0.71073 Å) at 153 K. The data collected on one selected crystal were integrated using the Siemens SAINT program, 46 with the intensities corrected for Lorentz factor, polarization, air absorption, and absorption due to variation in the path length through the detector face plate. Absorption corrections were applied.
The structures were solved by direct methods and re ned on F 2 by full-matrix least squares using SHELXTL. 47 All of the non-hydrogen atoms were located from the Fourier maps and were re ned anisotropically. All H atoms were re ned isotropically with the isotropic thermal parameters related to the non-hydrogen atoms to which they are bonded. Crystallographic and re nement details are listed in Table  S1.

Data availability
We declare that the data supporting the ndings of this study are available within the article and Supplementary Information le or from the corresponding author upon reasonable request.