Cr silicate as a prototype for engineering magnetic phases in air-stable two-dimensional transition-metal silicates

Identifying environmentally inert, ferromagnetic two-dimensional (2D) materials with high Curie temperatures (T c) down to the single layer limit has been an obstacle to fundamental studies of 2D magnetism and application of 2D heterostructures to spin-polarized devices. To address this challenge, the growth, structure and magnetic properties of a 2D Cr-silicate single layer on Pt(111) was investigated experimentally and theoretically. The layer was grown by sequentially depositing SiO and Cr followed by annealing in O2. Scanning tunneling microscopy (STM), low-energy electron diffraction (LEED), and low energy electron microscopy all indicated a well-ordered layer that uniformly covered the surface, with STM and LEED indicating that the silicate relaxed to its favored lattice constant. Further experimental characterizations demonstrated that the Cr was nominally 3+ but with a lower electron density than typical trivalent Cr compounds. Comparison with theory identified a Cr2Si2O9 structure that resembles a single layer of a dehydrogenated dioctahedral silicate. Magnetic circular dichroism in x-ray absorption spectroscopy revealed a ferromagnetically ordered state up to at least 80 K. Theoretical analysis revealed that the Cr in a dehydrogenated Cr-silicate/Pt(111) is more oxidized than Cr in freestanding Cr2Si2O9H4 layers. This greater oxidation was found to enhance ferromagnetic coupling and suggests that the magnetism may be tuned by doping. The 2D Cr-silicate is the first member of a broad series of possible layered first-row transition metal silicates with magnetic order; thus, this paper introduces a new platform for investigating 2D ferromagnetism and the development of magnetoelectronic and spintronic devices by stacking 2D atomic layers.


Introduction
Two-dimensional (2D) materials have attracted tremendous research interest for over a decade. Much ongoing progress on 2D materials has been driven by the vast reservoir of properties these low-dimensional materials possess [1]. One of the most exciting recent developments has been the discovery of intrinsic magnetic order in atomically thin layers [2][3][4][5][6][7]. This unique feature provides an incredible possibility for understanding atomically thin magnetic materials and opens the way for applications in future ultrathin spintronic devices and other quantum phenomena-based technologies [8][9][10][11][12]. Similar to bulk materials, there are four main types of magnetism in 2D materials: ferromagnetism; antiferromagnetism; ferrimagnetism; and paramagnetism [4,6,13,14]. The 2D ferromagnets show parallel spins on atoms, and thus possess large net magnetization and susceptibility, which are most interesting for spintronic applications. Two-dimensional magnets are few-atom-thick and thus could ideally be stacked with layers with different functionalities as Lego blocks, leading to possibilities to create new materials with diverse magnetic properties and physical phenomena. Since the first experimental discovery of the 2D ferromagnets Cr 2 Ge 2 Te 6 [3] and CrI 3 [4] in 2017, many new 2D magnetic materials have been theoretically predicted and experimentally discovered [6]. Two-dimensional ferromagnetism has been demonstrated in both layered and non-layered materials such as CrX 3 (Cl, Br, I) and Cr 2 Se 3 respectively [4,[15][16][17]; however, the ferromagnetic order in these materials is fragile, suffering from low curie temperature (T c ), and limited stability under ambient conditions [3,4,18,19]. For example, CrI 3 has a T c below 60 K and is unstable in air due to reactions with water [4,18,20]. Meanwhile transition metal oxides display rich magnetic properties and are stable in air, making them great alternative candidates for 2D and reduceddimension spintronics applications. A recent theoretical study of layered first row transition metal silicates explored the potential for magnetic ordering in the single layer limit [21]. The study revealed that single transition metal silicates tended to favor antiferromagnetic order with Cr silicate the most likely to display ferromagnetic ordering. Therefore, we have begun our exploration of the magnetic properties of 2D transition metal silicates with Cr silicate. In this paper it will be shown that a highly ordered, single layer 2D Cr-silicate grown on Pt(111) displays ferromagnetic order up to at least 80 K.
Among the wide variety of transition metal oxides, transition metal silicates can adopt layered structures that are inherently stable in air. Each 2D layer is non-centrosymmetric perpendicular to the layer, dictating a piezoelectric response that opens the possibility to couple magnetic, elastic and electrostatic responses which would be unique in single layer 2D materials [21]. A recent theoretical report on the piezoelectric coefficients of single layers of transition metal silicates reveals that their piezoelectric coefficients are on par with other 2D piezoelectric materials such as TMD and transition metal dioxides [21][22][23][24]. Ferromagnetic order is known to exist in bulk layered silicates, e.g. the Fe containing mineral greenalite [25], but they have not been experimentally demonstrated in the single layer limit. At the same time, in the crossover to low dimensions, the magnetic ground state of these materials can be distinct from their bulk counterparts, thus opening a plethora of opportunities for emergent new magnetic phases. More importantly, transition metal silicates comprise a series of possible materials with different magnetic transition metal cations (e.g. Cr, Fe, Ni), which may bring great opportunities to explore magnetic properties in the 2D limit, where spin fluctuations  [27][28][29][30][31][32].
are expected to be strongly enhanced [21,[26][27][28][29][30][31]. Figure 1 illustrates the palette of demonstrated and predicted stable, single layer ternary transition metal silicates. Indeed, the wide flexibility of doping these 2D materials to form mixed metal-O layers suggests straightforward tuning of the magnetic anisotropy, which is crucial for reducing or strengthening spin fluctuations and thus various forms of order [9]. Exploring new kinds of 2D magnetic materials with enhanced stability and developing facile and scalable synthetic routes to controllably prepare them are of considerable importance for further development of this field. Therefore, the 2D transition metal silicate family is the focal point of our study.
Crystalline 2D transition metal silicates, formed from a layer of corner-sharing six-membered rings of SiO 4 tetrahedra atop an octahedrally-coordinated metal-oxygen layer, have recently been grown on metal substrates [26][27][28][29][30]. Figure 1 highlights the range of potential metal substrates [31]. The first representatives of these 2D materials were Fe-and Tisilicate films grown on Ru(0001) [26,27], followed by others such as Ni-silicate on Ni x Pd 1-x (111) alloys [28,29]. Traditional van der Waals (vdW) compounds have chemically saturated bonds at their surfaces; however, the aforementioned silicates bind relatively strongly to their metallic substrates, which hinders their exfoliation. In a recent theoretical report, the authors adopted a hydrogenation strategy to weaken the interaction between 2D transition metal silicate overlayers and the underlying substrate to the exfoliable regime which allows the materials to be considered for scientifically and technically interesting for vdW heterostructures [32]. To achieve fast growth and controlled phase formation, efforts have been made to generalize recipes for their synthesis [31]. The challenge of fabricating stoichiometric transition metal silicate thin films lies in controlling stoichiometry and crystallinity. To address this issue, we recently developed a method for preparing stoichiometric 2D transition metal silicates using the controlled solid-state reaction of SiO and metals (M: Ni, Fe) under UHV conditions [28,30], paving the way for the fabrication of 2D oxides with structural compositions not seen in bulk materials [27][28][29].
In this work, we report the preparation of single layer, air-stable 2D Cr-silicate and the investigation of its structure and magnetic properties. To determine the growth mode, atomic, and electronic structure, we combined diverse surface analytical methods including scanning tunneling microscopy (STM), low-energy electron diffraction (LEED), x-ray photoelectron spectroscopy (XPS), reflection-absorption infrared spectroscopy (RAIRS) and x-ray absorption near edge spectroscopy (XAS) with first principles theory. We found that the Pt surface is completely covered by a wetting Cr-silicate layer that maintains high crystallinity across the Pt terraces. X-ray magnetic circular dichroism (XMCD) unveiled a ferromagnetically ordered state at 80 K. Meanwhile, theory supported these findings via ground state magnetism calculations benchmarked over a wide range of Hubbard-U parameters (the DFT + U method [33]). We calculated the magnetic critical temperature using linear spinwave theory which indicated long-range ordered ferromagnetism with T c in the 50-150 K range depending on the DFT functional used. Simulated RAIRS, core-level emission and absorption results all reasonably agree with experiment, confirming our structural model. Both experiment and theory indicate that the Pt substrate inefficiently screens the charge on the Cr cations thereby enhancing their ferromagnetic coupling and opening the door to controlling the magnetic order in transition metal silicates by chemical or electrostatic doping [34].

Growth and structure of Cr-silicate on Pt(111)
The Cr-silicate layer was grown on Pt (111) by first depositing 1 ML SiO, then depositing enough Cr to obtain a 1:1 Cr:Si ratio; both components were deposited in UHV with the sample at 300 K. The sample was then annealed in 2-5 × 10 −6 Torr O 2 to successively higher temperatures until a well-ordered LEED pattern was obtained. We found that order emerged near 825 K and that the patterns did not significantly change past 975 K. With the assistance of STM and LEED, we found that the Cr-silicate on Pt formed under these conditions is a highly crystalline 2D overlayer with a low defect density. Figure 2 reveals details of the Cr-silicate structure on Pt(111), as derived from LEED and STM measurements. For clean Pt(111), figure 2(a) shows only six spots, indicating the 3-fold symmetry (unit cell indicated in red) as well as the cleanliness of the Pt(111) surface. In contrast, figure 2(b) shows a moiré pattern once the Cr-silicate/Pt(111) formed; such patterns are commonly observed for relaxed oxide layers due to lattice mismatch with the support [26,[28][29][30]. The LEED pattern of the Cr-silicate structure features sharp reflections confirming an overall longrange order. Compared with the pattern for clean Pt(111), it reveals a 30 • angle between the Cr silicate and the Pt (111) surface lattices (see dashed lines in figures 1(a) and (b)). This behavior has also been observed by LEED for Fe silicate monolayers (MLs) grown on Ru(0001) and Pd(111) surfaces [26,28].
The Cr-silicate growth was further characterized by STM, as shown in figure 2(c). Figure 2(c) presents a large-scale topographic image, showing a highly ordered hexagonal superstructure consistent with the moiré pattern expected from the LEED data. The measured step height in image 1(c) is 2.3 ± 0.1 Å, which matches the height difference between Pt terraces, confirming the sample surface is completely covered by a wetting Cr-silicate layer. Notably, the step edges are straight when they parallel a closepacked direction of the moiré lattice (white arrow lower, center of figure 2(c)) but facet when the mean step direction deviates from a close-packed direction, as highlighted by the blue arrows in figure 2(c). As shown in the model in figure 2(d), the closepacked direction of the moiré lattice follows that of the underlying Pt(111). Therefore, the implication is that the overlayer does not alter the favored step orientation despite the 30 • rotation between the overlayer and substrate but rather limits step motion leading to observable faceting at room temperature.
The moiré pattern was observed on all terraces with a period of 49 ± 3 Å. The high resolution STM image in the inset (figure 2(c)) clearly shows two different regions, termed valley (dark) and hill (bright) as a reflection of the moiré pattern. The apparent height difference varied with the sample bias from 70 pm (for 1.3 V) to 50 pm (for 0.7 V), indicating that the height differences arise at least in part from an electronic state modulation; such bias-dependent height differences have been previously observed for Ni-silicate and Fe-silicate layers [28,30]. In addition to the moiré structure, the high resolution STM image (inset, figure 2(c)) shows a honeycomb structure with a lattice constant of 5.4 ± 0.15 Å with the unit cell marked in green. The schematic drawing of the honeycomb structure of the Cr-silicate layer atop Pt(111) (figure 2(d)) is consistent with experimental observations and confirms the expected epitaxial relationship. The smallest repeating cell is highlighted by the black dashed line, with a lattice constant of 47 Å. The high resolution STM image in the figure 2(c) inset also shows considerable variation in the local contrast across the moiré pattern, which has also been seen for 2D Ni-silicate on Ni x Pd 1-x (111) [28]. The LEED analysis coupled with the local STM imaging cannot rule out the possibility that there might be some local regions that are not covered by the Cr-silicate layer. As an additional and separate verification that the surface is uniformly covered by Cr silicate, we carried out ex-situ photoemission electron microscopy (PEEM)/low energy electron microscopy (LEEM) measurements. The LEEM image (figure S1(a) in the supporting information (SI)) acquired with a field of view of 30 µm suggested that the Cr-silicate consists of a single phase and the film thickness is uniform. The µLEED pattern shown in figure S1(b) of the SI, obtained from the center of the sample using a 1.5 µm selected-area aperture shows a single hexagonal phase with a moiré structure with the same periodicity seen in figure 2(b). Additionally, these observations corroborate the in-situ LEED and STM results, confirming that the Cr-silicate structure is not sensitive to exposure to the ambient environment. Figure 3 presents the RAIRS spectrum of the Crsilicate layer on Pt(111). The spectrum exhibits a single sharp band at 997 cm −1 , virtually identical to the observed vibrational stretching mode for Fesilicate layers on Ru(0001) and Pd(111) (1005 and 1050 cm −1 ) [26,30]. There was no evidence of the peak due to 2D bilayer SiO 2 which appears at 1294 cm −1 for SiO 2 /Pt(111) [35]. The peak is in the region expected for Si-O-metal links but at a lower energy than seen for Fe-silicate on Pd(111) or Ru(0001) [26,30]. Theory reproduces this downward shift, as detailed below. This finding straightforwardly leads to the conclusion that the film is a 2D metal silicate layer. We next utilized ex-situ core-level spectroscopies, XPS and XAS, to study the chemical composition and electronic structure of the Cr-silicate layer. Figure 4 shows detailed Cr 2p (a), O 1s (b), Si 2p (c) XPS and Cr L 2,3 edge (d) XAS spectra recorded at 300 K. In figure 4(a), the Cr 2p core level spectrum of the sample as put into the XPS chamber (bottom curve), is compared to the spectrum after annealing in O 2 (P (O2) = 5 × 10 −5 Torr) at 850 K to remove adsorbates such as water (top curve). The Cr 2p spectra present similar intensity, which excludes the possibility of desorption and/or diffusion of Cr into the bulk during the oxidation treatment. The Cr 2p spectrum shows a maximum of the Cr 2p 3/2 core level peak at 577 eV. This peak position and its narrow line shape with a relatively low full width at half maximum (FWHM) of 2.7 eV suggests that Cr exists in a single oxidation state. As a result, the Cr 2p core level region was fit with two peaks centered at 586.8 ± 0.05 and 576.95 ± 0.05 eV (figure 4(a), bottom curve), ascribed respectively to Cr 2p 1/2 and Cr 2p 3/2 . The shoulder on the 2p 3/2 peak at 581 eV is a satellite peak. For Cr 3+ , literature values for the Cr 2p 3/2 peak range from 576.5 to 576.8 eV while the Cr 2p 1/2 peak appears between 586.0 and 586.7 eV [36,37]. The peak positions observed here for Cr-silicate are slightly above the upper ends of these ranges. On the other hand, other stable Cr oxidation states of can be ruled out. For CrO 3 , the Cr 2p 3/2 peak position is 1.3 eV higher than that of the Cr-silicate overlayer [38]. At the same time, Cr 4+ is unique in that its Cr 2p peaks are at lower binding energies than those of Cr 3+ despite the higher oxidation state; reported Cr 2p 3/2 peak positions for CrO 2 range from 576.3 to 576.4 eV [36,38]. Therefore, the data suggest Cr is nominally 3+ with a somewhat lower electron density than typical trivalent Cr compounds.
The Si 2p XPS spectrum (figure 4(b)) shows that all of the Si is in the 4+ oxidation state [39]. Only a single Si 2p peak was observed indicating that all of the Si atoms have the same chemical surrounding and bonding configuration. Take-off angledependent XPS data (not shown) demonstrate that Cr ions occupy the metal-oxide interface region while Si ions reside close to the surface. The deconvolved O 1s spectra (figure 4(c)) reveal three components peaks at 529.5, 531.0, and 531.6 eV. The two higher binding energy components are assigned to oxygen within the Cr-silicate structure and correspond to oxygen bound only to Si and oxygen bound to at least one Cr [26,30], respectively. Comparing the position of the lower binding energy component with data for Pt-(2 × 2)O [40], we attribute the peak at 529.5 eV to oxygen bonded to Pt(111). The stoichiometry of the Cr-silicate overlayer was determined by Cr 2p, Si 2p, and O 1s core levels spectral analysis. The relative areas of the total Cr 2p and Si XPS peaks (corrected by relative sensitivity factors) show nearly equal concentration of Cr and Si species (table S1). The O content was evaluated quantitively using the Pt(2 × 2)−O structure for calibration purposes (table S1 and figure S2 in the SI). Together, the data indicate a Cr 2 Si 2 O 9 stoichiometry, consistent with a dehydrogenated dioctahedral kaolinite-like layer [21].
To gain further insight into the Cr environment, element-specific XAS measurements were performed at the Cr L 2,3 edges at 300 K by the total photoelectron yield method; these measurements were performed after ambient exposure and without further treatment. The clear assignment of the Cr 3+ oxidation state in Cr 2 O 3 via its L 2,3 edge can be used as a guide for the evaluation of the oxidation state of the Cr ions in the Cr-silicate structure. For comparison, in figure 4(d) we show our spectrum (black curve) with the published experimental L-edge spectrum of a Cr 2 O 3 layer (blue curve) [41]. The L 3 peak positions for the Cr 2 O 3 and Cr-silicate single layers are 576.5 eV and 577.4 ± 0.2 eV, respectively. This indicates an energy shift to higher energy in the Cr-silicate by ∼1 eV as compared to the Cr 2 O 3 (Cr 3+ , d 3 ). In figure 4(d) (top curve), the CrO 2 XAS spectrum is also shown for comparison [42]. The positions of the L 3 and L 2 peaks for the Cr-silicate overlayer relative to those of Cr 2 O 3 (Cr 3+ , 3d 3 ) and CrO 2 (Cr 4+ , 3d 2 ) can be used as an indication of the Cr valence. Qualitatively, this chemical shift implies a lower d-electron count in the Cr-silicate overlayer compared to Cr 2 O 3 but higher than that of CrO 2 . This is further investigated below via time-dependent density function theory (TDDFT) calculations, which refine this trend and provide additional insights on the effects of d orbital occupancy of Cr on the XAS and XMCD spectra. The oxidation states of transition metal elements can be also be characterized by comparing the L 3 /L 2 integrated intensity ratio (the L 3 /L 2 ratio) of unknown samples with standard reference samples [43]. For the Cr-silicate spectrum in figure 4(d), the L 3 /L 2 ratio is 1.6 while the L 3 /L 2 peak ratio for the C 2 O 3 and CrO 2 references is 1.679 and 1.484, respectively [43], thus, there is a good match of the L 3 /L 2 ratio for the Cr-silicate structure and the Cr 3+ reference. Alternatively, the nature of the Cr atoms in the Cr-silicate layer can be characterized by comparing the XAS line shapes with reference spectra. The L 3 and L 2 profiles for Cr-silicate show a good overall resemblance to the XAS Spectrum of C 2 O 3 [41] − as highlighted by blue arrows in figure 4(d). From these three criteria, we conclude that the oxidation state of the Cr in the Crsilicate structure is nominally the 3+ state but with a shift to higher energies consistent with a somewhat lower electron density around the Cr compared to typical Cr 3+ compounds, in agreement with our XPS data.

Theoretical analysis of Cr-silicate structure and thermodynamic stability
Combining the experimental data with the results of the theoretical calculations, we were able to develop a full picture of the Cr-silicate/Pt(111) system. Prior work on 2D Fe silicate on Ru(0001) [26,32] and Pd(111) [30] found that the dehydrogenated form of the dioctahedral silicate Fe 2 Si 2 O 9 preferred to transfer one oxygen to the metal substrate leading to Fe 2 Si 2 O 8 atop a chemisorbed oxygen layer. Because the adsorbed oxygen may readily be removed from Pt by reaction with background H 2 or CO, here we consider both Cr 2 Si 2 O 8 and Cr 2 Si 2 O 9 on Pt(111); their minimum energy structures are shown in figure 5. Unlike Fe-silicate on Pd(111) [30], we find that none of the oxygen atoms fully transfer to the Pt(111) substrate. Although the oxygen adsorption energies on Pd and Pt(111) are similar, Cr has a much higher formation energy for the trivalent sesquioxide, 11.75 eV versus 8.54 eV for Fe, indicating that Cr 3+ -O bonds are considerably stronger. Hence, all of the oxygens at the interface remain bound to Cr, though as shown in the side view, there is one longer Cr-O bond

Simulated spectra
DFT phonon simulations were applied to determine the consistency between the structural models and the experimental observations. RAIRS spectra were calculated for Cr 2 Si 2 O 8 and Cr 2 Si 2 O 9 on Pt(111), where the focus was on the vertical Si-O stretching mode which yields vibrational frequencies in the 900-1300 cm −1 range for 2D silica and silicates on metal surfaces [28,30,31]. Figure 6 shows the simulated RAIRS spectra for Cr 2 Si 2 O 8 and Cr 2 Si 2 O 9 as a function of the U value in the DFT + U method. There is an interesting U dependence whose origins is explained below, but the main findings are that: (a) the spectra vary little once U ⩾ 2 eV, (b) the key vibrational peak is just above 1000 cm −1 for the Cr 2 Si 2 O 8 and slightly above that for Cr 2 Si 2 O 9 , and (c) in accord with experiment, the simulation predicts lower vibrational frequencies for Cr silicates compared to Fe silicates on metal surfaces [26,30]. We expect U = 2 eV to be ideal for this system, considering that d-electrons of Fe silicates on a metal substrate presents an interpolation between the more localized d electrons of bulk Fe-oxides and delocalized d electrons of Fe metal. These two limits are associated with U values that are near 4 and 0 respectively [44]. Compared to the experimental spectrum in figure 2, there is better agreement on the peak position for Cr 2 Si 2 O 9 , but the small difference in the predicted peak position for Cr 2 Si 2 O 8 and Cr 2 Si 2 O 9 , makes a definitive assignment based on RAIRS alone challenging. (Note that the smaller peak predicted at 840 cm −1 is near the cutoff of the photoelastic modulator used to collect the experimental spectrum, and one should not read too much into its absence in experiment.) There are multiple reasons for the U dependence of the RAIRS spectra. From a chemical viewpoint, increasing U on the Cr 3d manifold pushes the unoccupied Cr 3d states up in energy, weakens Cr-O bonding hybridization, weakens the Cr-O bonds, and the Si-O bonds strengthen in response: this explains the stiffening of the vibrational mode with increasing U. Intermixed into this chemical effect is a strain effect: the simulated Cr silicate is an epitaxial (2 × 2) superlattice on the Pt substrate, so changing the Cr U modifies the preferred bond lengths in the Cr silicate and thus the tensile strain felt by the Cr-silicate layer which modifies its out-of-plane structure and vibrational properties. Some of the complexities are exemplified by the case of the Cr 2 Si 2 O 9 RAIRS spectrum which shows a very strong change from U = 0 to U ⩾ 1 eV. Here, the ground state structure changes with U significantly: specifically, the three interfacial oxygens forming O-Pt bonds have a different evolution versus U (see figure S3 in the SI). One of the oxygens shortens its O-Pt bond length from 2.53 Å for U = 0 to 2.06 Å for U = 4 eV while the other two interfacial oxygens only change their bonds lengths from 2.18 to 2.11 Å and 2.11 to 2.03 Å. Hence, the U = 0 case has strongly differing tethering of oxygens to the Pt substrate leading to a split vibrational spectrum, while the larger U cases have an essentially uniform O-Pt tethering.
Moving on to simulated XPS data, we also calculated the core level binding energies of the Cr 2p, Si 2p and O 1s levels for Cr 2 Si 2 O 8 and Cr 2 Si 2 O 9 on Pt(111) using an approach described previously [32] (see figure S4 in the SI). For Cr 2p, the extra oxygen in Cr 2 Si 2 O 9 induces a significant 1 eV shift to higher binding energy compared to Cr 2 Si 2 O 8 . Meanwhile the Si 2p peak is scarcely affected by the extra oxygen, consistent with the expectation that the Si remains 4+ in all of the silicate compounds. In contrast, theory predicts significant differences in the O 1s region between the O 8 and O 9 compounds. Figure S4(c) of the SI shows that the highest O 1s binding energy component assigned to Si-O-Si bonds is unaffected by the total oxygen content, again consistent with the tetrahedral environment surrounding the Si atoms being essentially the same in the two structures. On the other hand, the Si-O-Cr and Cr-O-Cr oxygens appear at ≈0.5 eV lower binding energies in the O 9 compound. More significantly, the longer Cr-O bond and concomitantly shorter Pd-O bond when the ninth oxygen is added induces a lower energy peak in the O 1s region of the O 9 compound, similar to that expected for oxygen bound to the Pt surface. Thus, the experimental observation of the Cr peak towards the higher end of the Cr 3+ range and a component in the O 1s region consistent with O-Pd bonds reinforces the conclusion that we formed the Cr 2 Si 2 O 9 structure. Both experiment and theory indicate that the Cr 2 Si 2 O 9 layer induces at most small changes in the Pt work function. Experimentally, a work function of 6.1 ± 0.2 eV for the Cr-silicate covered Pt surface was determined from the high energy photoelectron cutoff, which is within the experimental uncertainty of the bare Pt value [45], while theory suggests that a Cr 2 Si 2 O 9 layer would increase the Pt(111) work function by only 0.14 eV.
To deepen the analysis of the Cr-silicate/Pt(111) system electronic structure, TDDFT was implemented to calculate of the Cr L 2,3 x-ray absorption edges of Cr-silicates on Pt(111). More details about the TDDFT calculations can be found in SI section S5. The calculated Cr L 2,3 spectra for Cr-silicates with the candidate structures of Cr 2 Si 2 O 8 /Pt(111) and Cr 2 Si 2 O 9 /Pt(111) are shown with the experimental spectrum in figure 7. It is interesting to see that the Cr L 3 edges of Cr 2 Si 2 O 8 /Pt(111) and Cr 2 Si 2 O 9 /Pt(111) are quite similar in terms of peak shape, and in good agreement with experiment. However, the Cr L 2 edges of Cr 2 Si 2 O 8 /Pt(111) and Cr 2 Si 2 O 9 /Pt(111) show obvious differences, especially the relative intensity of the second sub-peak. The shape of L 2 edges in the XAS spectra of Cr 2 Si 2 O 9 /Pt(111) is closer to the experimental results, also suggesting that Cr 2 Si 2 O 9 is the more likely experimental structure.

Magnetic properties
The XMCD technique was employed to examine the magnetic order in the Cr-silicate layer. XAS spectra acquired at normal incidence and with a 1.9-T magnetic field applied in opposite directions (µ + and µ − ) at 80 K and 300 K are shown in figure 7 together with the resulting XMCD spectra. The Cr L 2,3 edge XAS lineshapes (figure 4(d), middle curve) are similar at the two temperatures. As detailed above, the lineshape, L 3 /L 2 ratio and XPS data indicate that the Cr is nominally 3+. On the other hand, at 300 K there is no difference between the µ + and µ − spectra while at 80 K the data in figure 7(b) show a noticeably weaker absorption for µ + at the L 3 edge and the opposite at the L 2 edge. The XMCD signal is taken as the difference between the absorption for the two magnetic field orientations and is shown in the lower panels of figures 7(a) and (b). At 80 K, the non-zero XMCD signal with opposite signs for the L 3 and L 2 edge is indicative of ferromagnetic order. The loss of this signal at 300 K indicates that a ferromagnetic-toparamagnetic phase transition took place between 80 and 300 K.
The dichroism at 80 K indicates that the magnetic moments of the Cr ions have a component parallel to the photon helicity vector, in this case perpendicular to the basal plane of the 2D layer. Applying so-called 'sum rules' to the XMCD spectrum collected at 80 K allowed us to identify the spin (m s ) and orbital (m l ) moment contributions to the total net magnetization of the Cr-silicate layer. The calculated m s and m l magnetic moments were 0.27 ± 0.02 µ B /Cr atom and 0.05 ± 0.01 µ B /Cr atom, respectively. Thus, the total magnetic moment (m s + m l ) is 0.32 µ B /Cr atom.
For comparison, the theoretically calculated Cr L 2,3 XAS and XMCD for Cr 2 Si 2 O 8 /Pt(111) and Cr 2 Si 2 O 9 /Pt(111) are also presented in figures 7(c) and (d). Again, it is noted that the calculated Cr L 2,3 XAS and XMCD spectra for Cr 2 Si 2 O 9 (figure 7(d)) better fit the measured spectra. The calculated XAS and XMCD of Cr 2 Si 2 O 9 reproduce the experimental spectra well except for the fine structure of the L 2 edge-the calculated spectra has the relative intensities of the two peaks reversed. To address this issue, the effects of the d orbital occupancy of Cr on the XAS and XMCD spectra of Cr 2 Si 2 O 9 /Pt(111) was investigated using the TDDFT method. By moving some electrons from the Cr d orbital to the large radius 4p orbitals, one can simulate the spectra with different d orbital occupancy with less d electrons corresponding to greater oxidation. As shown in figure S8 of the SI, the intensities of the second sub-peak (as indicated by the arrow) slightly decrease with decreasing d orbital occupancy (increased oxidation) and approach the experimentally observed sub-peak intensity ratio at the L 2 edge. It becomes clear that to obtain a better agreement between the measured and calculated spectra requires treating the ground state of Cr as nominal 3+ valence (see figure S8). On the other hand, this indicates that the Cr-silicate, if it crystalizes as Cr 2 Si 2 O 9 /Pt(111), may have slightly larger oxidation state than typical for the Cr 3+ formal oxidation state.

Origin of magnetic order
The magnetic ordering in the Cr silicate/Pt(111) system was analyzed theoretically. As a starting point we considered the three different magnetic configurations of Cr 2 Si 2 O 9 in a (4 × 2) supercell on Pt(111) illustrated in figure 8. Understanding the magnetic   2) supercell, as shown in figure 8. We calculate the energy differences between three magnetic configurations possible in this supercell: FM, AFM-1 and AFM-2 as indicated in the figure. Using PBE calculations (U = 0 eV) and the lattice parameter of Cr 2 Si 2 O 9 H 4 (5.24 Å), we find that the FM (ferromagnetic) configuration is stable compared to the other two AFM (antiferromagnetic) configurations by more than 50 meV/Cr atom as shown in figure 8.
Noting that the AFM-1 and AFM-2 energies only differ by 2 meV/Cr atom, we can benchmark the effect of lattice parameter on the magnetic ground states using the FM and AFM-1 states only. We find again using PBE, but the lattice parameter of Pt(111) (5.64 Å), that the energy difference between FM and AFM-1 phases is 56 meV/Cr atom. In our second benchmark to understand the effect of U on the energy differences between different magnetic phases, we use the latter simulation setup. We used the lattice parameter of Pt (111)  While the collinear calculations point to a ferromagnetic ground state, due to the Mermin-Wagner theorem [46], out of plane magnetic anisotropy is required to obtain long range magnetic order in 2D materials. Therefore, magnetic anisotropy was benchmarked for both Cr 2 Si 2 O 8 and Cr 2 Si 2 O 9 on Pt(111) using PBE + U. Here the magnetic anisotropy energy (MAE) is defined as MAE = E(z)-E(x) where z is normal to the (111) surface and x is within the surface plane, so that negative MAE means out-of-plane magnetic moments are preferred and thus long-range (Ising-like) ordering is possible. In table 2, we present site-resolved MAEs for the Cr-silicate layer as well as the layers of the Pt substrate. In Cr 2 Si 2 O 8 /Pt(111), Table 2. Site-resolved magnetic anisotropy (in meV) of (a) Cr2Si2O8/Pt(111) and (b) Cr2Si2O9/Pt(111). Pt slabs contain a total of 9 layers, and the total MAE per layer is reported for each of the top four layers. For layers 5-9 the average MAE per layer is reported for the respective layers. we see that the MAE contribution from the Pt substrate is larger than that from the Cr-silicate overlayer. This is in part because the small magnetic moments of the interface Pt atoms (less than 0.06 µ B compared to 2.36 µ B for the Cr atoms) go together with nearly an order of magnitude larger spin-orbit coupling for Pt compared to Cr due to the heavier atom effect [9]. In Cr 2 Si 2 O 9 /Pt(111), the situation is again very similar with magnetic moments on Pt sites less than 0.04 µ B , while the two distinct Cr atoms in the unit cell have magnetic moments of 1.74 and 2.43 µ B . As shown in table 2, the MAE of the Cr-silicate is more negative for Cr 2 Si 2 O 8 /Pt(111) than Cr 2 Si 2 O 9 /Pt(111), both Crsilicate layers have negative MAE and thus prefer easyaxis (out-of-plane) moments, and both show increasing (less negative) MAE with increasing U. While the two compositions have opposite sign net MAE, we explain below that this is not the most relevant aspect of the magnetic stabilities. Linear spinwave theory [47,48] was used to estimate the Curie temperature (T c ) of the two prospective 2D Cr silicate phases on Pt(111). E(FM-AFM1) values in table 1 and the MAE values under Cr-silicate columns in tables 2(a) and (b) were used to calculate magnetic exchange and onsite parameters that are utilized in the linear spinwave theory models as explained in [48]. Despite the large MAE contribution from the substrate, we expect small magnetic exchange interactions between the overlayer and the substrate due to the small Pt magnetic moment, so that to zeroth order, the two magnetic subsystems are effectively decoupled. Thus, only the MAE contributions from the overlayer were considered in constructing the spinwave model. Given that the two AFM phases in figure 7 are virtually the same, we assume that all Cr atoms are identical and occupy a perfect honeycomb lattice. Using linear spinwave theory, we can calculate the partition function of spin excitations and thus generate the magnetization vs. temperature curves as in figure 9. For both Cr 2 Si 2 O 8 and Cr 2 Si 2 O 9 on Pt(111) the anisotropic exchange interaction parameter decreases with increasing U, as can be inferred from table 2. Thus, based on only the magnetic anisotropy, T c would be expected to fall with increasing U. However, figure 9 shows that T c for Cr 2 Si 2 O 8 increases with increasing U. Despite the monotonic changes in magnetic anisotropy with U, we find that the changes in T c are better correlated to the changes in isotropic exchange which can be obtained from the energy differences between collinear magnetic phases; since we assume a perfect honeycomb lattice, the energy differences in table 1 could be used. Overall, using reasonable values of U (1-2 eV), we estimate that T c should fall between 130-270 K for Cr 2 Si 2 O 8 /Pt(111) and 45-130 K for Cr 2 Si 2 O 9 /Pt(111); both are consistent with the experimental observation of a ferromagnetic to paramagnetic phase transition between 80 and 300 K.
Constraining the Cr silicate to a Pt(111) supercell in the simulations presents some technical limitations. First, the in-plane lattice parameter is set to that of the Pt(111) (2 × 2) supercell which yields ≈5% tension on the Cr silicate based on the lattice parameters of isolated freestanding Cr 2 Si 2 O 8,9 which in the experiments is alleviated by forming an incommensurate layer. However, freestanding Cr 2 Si 2 O 8,9 are unstable compounds due to having 4+ and 5+ Cr oxidation states, respectively. The freestanding layers can be stabilized by adding hydrogens which donate charge to Cr, reaching a 3+ Cr formal oxidation state for Cr 2 Si 2 O 9 H 4 ; adding hydrogen also increases the Cr silicate lattice parameter. Therefore, in the actual heterostructure, charge transfer from Pt to the Cr silicate can increase the lattice parameter of the overlayer, and hence the mismatch should be smaller than 5%. Nevertheless, the substantial imposed strain in the calculations might still alter the predicted magnetic and vibrational properties of the material.
To isolate the effect of varying degrees of charge transfer from the substrate to the overlayer from the strain effect, we study freestanding Cr 2 Si 2 O 9H4 layers as a surrogate system, whereH are pseudohydrogens that can adopt a fractional atomic number (Z) between 0.25 and 2. This theoretical surrogate system offers two benefits: (i) its lattice parameter is fully optimized due to its freestanding nature; and (ii) the Cr formal oxidation state is precisely controlled by varying the atomic number ofH. Including Cr 2 Si 2 O 9 , this gives access to the range of Cr formal charges between 5+ and 2+ without significantly altering the geometric structure or applying variable and artificial strain.
The XPS and XAS data in figures 4 and 7 along with typical Cr oxidation states found in oxides argue that the Cr oxidation states of interest are between 3+ and 4+. In table 3 we summarize the results of calculations withH pseudocharges of 0.5, 0.75 and 1.0. Using supercells to calculate the FM and two AFM configurations (same as figure 7), we calculate the magnetic coupling parameters by fitting the energy changes to a Heisenberg spin model SiSj where we include only nearest neighbor Cr-Cr exchange interactions. ForH values of 0.5, 0.75 and 1, the corresponding S i,j values are 2, 2.5 and 3 µ B . We find two distinct nearest neighbor interactions within the Cr sublattice leading to two distinct magnetic coupling parameters J1 and J2: even for freestanding layers, the Si-O tetrahedra distorts the Cr-O octahedra, creating alternating distances between the Cr atoms with the J1 coupling related to the shorter Cr-Cr distance (figure 10). The edge sharing oxygens that govern the J1 coupling have no chemical bonds to the Si atoms, while the oxygens between the farther Cr-Cr pair (associated with J2) are also part of the Si-O polyhedra. The isotropic exchange parameters listed in table 3 as a function ofH pseudocharge. Clearly, the sign of the magnetic interaction changes from AFM to FM when Z ofH is below 1.0, then the magnitude increases monotonically below 1.0. The preference for ferromagnetic order for the supported layer is explained by this finding and means that the Pt(111) substrate only partially passivates the Cr 2 Si 2 O 9 layer compared to four hydrogen atoms. This is evident from the XPS spectra which show Cr peaks at the high limit for Cr 3+ and by the fact that the Cr 2 Si 2 O 9 subsystem has a metallic density of states on Pt(111) (see figure S9 in the SI), whereas freestanding Cr 2 Si 2 O 9 H 4 has an insulating gap (figure S10 in the SI) for U = 0 eV [21]. We also find that decreasing the Z ofH increases the width of the t 2g band, in agreement with the substantially increased width of Cr t 2g bands in Cr 2 Si 2 O 9 /Pt(111) (Figs. S10-13 in the SI).
The results outlined above suggest that the Pt substrate promotes ferromagnetic order in the silicate by pushing the Cr towards a more oxidized state, even if the Cr is not formally 4+. This means that ferromagnetic order can also be enhanced by doping that increases the Cr oxidation state (hole doping), for example by substituting trivalent Al for tetravalent Si, or by applying an electric field on an appropriate substrate that would draw holes into the Cr-silicate layer. A similar magnetic gating has been demonstrated for CrI 3 layers [34]. On the other hand, the Crsilicate layer lacks inversion symmetry perpendicular to the layer dictating a piezoelectric response, suggesting that electrostatic doping may be achieved by compressing or stretching the layer thereby coupling the elastic and magnetic responses.

Conclusion
A well-ordered 2D Cr-silicate layer was grown on Pt(111) and its structure, chemical and magnetic properties were characterized using a combination of microscopy, electron diffraction, and vibrational, electronic and magnetic spectroscopies. The results were compared with first principles theory which provided a detailed structural model and understanding of the magnetic ordering in the layer and the factors that enhance ferromagnetic coupling. The findings indicate that the silicate forms a Cr 2 Si 2 O 9 layer that resembles a single layer dehydrogenated dioctahedral silicate with O-Pt substrate bonds replacing the O-H bonds that stabilize vdW trioctahedral silicates in layered kaolinite-type structures. Compared to freestanding vdW Cr 2 Si 2 O 9 H 4 , the Pt substate is less efficient at passivating the silicate layer, leading to Cr ions that are more electron deficient than typically observed in Cr 3+ compounds. The Crsilicate layer was found to display ferromagnetic order at 80 K which transitioned to a paramagnetic state by 300 K. The Pt substrate's inefficient screening of the Cr-silicate layer was found to enhance ferromagnetic ordering, suggesting that the magnetic properties of the silicate layer may be tuned statically by chemical doping or dynamically by electrostatic hole doping on an appropriate substrate. The intrinsic piezoelectricity of the Cr-silicate layer suggests that an electric field may also be generated by compressing or stretching the layer, creating the possibility to couple elastic and magnetic responses. Although the Cr-silicate layer was prepared using UHV techniques, it maintained its structure and magnetic properties after sustained ambient exposure making it an attractive material for applications of 2D magnetic heterostructures. The Cr-silicate structure is similar to other single layer first row transition metal silicates spanning Ti to Ni that have recently been grown, but it is the first to have its magnetic order analyzed. Thus, there is great promise that much higher Curie temperatures may be achieved by changing the transition metal cation or combining multiple transition metal cations; coupled with their atmospheric stability, this makes transition metal silicates a compelling new system for exploring and engineering 2D magnetism.

Experimental details
The layers were grown and initially characterized using a UHV system equipped with a high-speed variable temperature scanning tunneling microscope [49], a double-pass cylindrical mirror analyzer for Auger electron spectroscopy (AES) measurements, and reverse-view LEED optics, as well as the usual facilities for sample manipulation, surface cleaning and thin film evaporation. The sample temperature was measured using a Type K thermocouple fixed to the sample surface and was cross-checked with an infrared pyrometer. All STM images were recorded at room temperature in constant current mode (Pt/Ir tips made by the cut-and-pull method) at positive sample biases and with the feedback current set between 0.05 and 1.0 nA during scanning.
Chromium was sublimed from a resistively heated tungsten basket (Alfa Aesar), while the silicon oxide was deposited via sublimation of SiO granules from an effusion cell (DCA Instruments) at 1300 K. The evaporation flux (0.15-1.2 ng·s −1 cm −2 ) was monitored with a water-cooled quartz crystal microbalance (Inficon). The evaporated amount of Cr and silicon oxide is given in MLs of Cr atoms and SiO units, whereby 1 ML is defined as 8.22 × 10 14 SiO (or Cr atoms)/cm 2 , a coverage sufficient to form a sheet of corner-sharing SiO 4 tetrahedra.
Core-level x-ray photoelectron spectra (XPS) were recorded after transferring the sample through air to a system equipped with a SPECS PHOIBOS NAP 150 hemispherical analyzer and a monochromatic Al Kα x-ray source. A Shirley background was subtracted from all core level spectra. The fitting was performed using a mixed Lorentzian-Gaussian line shape with the Gaussian percentage held at 25% which was found to consistently give the best fits. Polarization modulation RAIRS data were recorded using a Thermo Fisher Nicolet iS50 FTIR spectrometer with the incident light at a grazing angle of 81 • in a dry nitrogen environment. Following the exsitu measurements, the sample was returned to the UHV system where the sample was grown and initially characterized; after heating at 925 K in 2 × 10 −6 Torr O 2 , no observable changes in the film composition or structure could be detected by AES, LEED, or STM.
Micro-spot µLEED, and LEEM measurements were carried out at the x-ray photoemission electron microscopy/low-energy electron microscopy (XPEEM/LEEM) end station of the ESM beamline (21-ID), National Synchrotron Light Source II at Brookhaven National Laboratory. A selected-area aperture with a diameter of 1.5 µm was used to acquire µLEED patterns. LEEM images were processed using the standard basic contrast enhancement within the software ImageJ [50]. XAS and XMCD characterizations were performed at beamline 6.3.1 of the Advanced Light Source (Lawrence Berkeley National Laboratory). The Cr L 3,2 -edge spectra were recorded at 300 K and 80 K in total electron yield mode, corresponding to transitions into Cr 3d states near the Fermi level from the 2p 3/2 and 2p 1/2 core levels. The temperature was stabilized within ±1 K. All spectra were recorded at normal incidence, i.e. the incoming x-ray beam parallel to the sample surface normal. XMCD measurements were conducted by switching the magnetization parallel and antiparallel to the propagation direction of the circularly polarized x-rays at normal incidence; thus, the difference between the two XAS spectra is defined as XMCD. From the acquired XMCD spectra, a detailed calculation of orbital (m orb ) and spin (m spin ) magnetic moments in 2D Cr-silicate was made based on previously described 'sum rules'; the equations are as follows [51]: Here m orb and m spin are in units of µ B /Cr atom; and L 3 and L 2 the integration regions at the corresponding to each respective Cr L edge while L 2,3 indicates integration over the entire Cr L 3,2 edge region. The XAS spectra are represented by µ + and µ − when the magnetization is parallel and antiparallel to the x-ray helicity, respectively. E, ⟨T Z ⟩, and n h are the photon energy, the magnetic dipole operator, and the number of 3d band holes, respectively. ⟨T Z ⟩ is expected to be small for transition metal compounds (typically smaller than 5%) [52], and therefore, neglected in this work.

Computational details
All density functional theory [53][54][55] (DFT) calculations were performed using the plane-wave Vienna ab initio Simulation Package [56,57] (VASP) using the Perdew-Burke-Ernzerhof [58] (PBE) exchange correlation functional with D3 [59] semi-empirical vdW correction (PBE + D3) and Hubbard-U [33] corrections on the d-orbitals of Cr atoms. We used a kinetic energy cutoff of 520 eV, projector-augmented wave [60] pseudopotentials, and dipole correction [61]. A vacuum region of minimum 15 • A was used to perform slab and 2D layer calculations. Geometric relaxations were performed until all of the force components were smaller than 0.01 eV/ • A and an energy convergence of 10 −5 eV is used at every step of the geometric relaxations.
The simulations of XAS for Cr L 2,3 edges were carried out using the full multiple-scattering theory as implemented in the FDMNES code [62,63]. The calculation is fully relativistic including the spin orbit, to take into account the Cr magnetic features. Although the three-dimensional periodicity of unit cell was used for the construction of crystal structure, the final states were calculated by the Dyson equation in a cluster centered around the different absorbing atoms.

Data availability statement
The data cannot be made publicly available upon publication because the cost of preparing, depositing and hosting the data would be prohibitive within the terms of this research Project. The data that support the findings of this study are available upon reasonable request from the authors.