Intermixing during Epitaxial Growth of van der Waals Bonded Nominal

GeTe and Sb2Te3 van der Waals bonded superlattices epitaxially grown on passivated .... Atomic structure and dynamic reconfiguration of layered defect...
0 downloads 0 Views 2MB Size
Article pubs.acs.org/crystal

Intermixing during Epitaxial Growth of van der Waals Bonded Nominal GeTe/Sb2Te3 Superlattices Ruining Wang,* Valeria Bragaglia, Jos E. Boschker, and Raffaella Calarco Paul-Drude-Institut für Festkörperelektronik, Hausvogteiplatz 5-7, 10117 Berlin, Germany S Supporting Information *

ABSTRACT: In the present work, GeTe and Sb2Te3 van der Waals bonded superlattices epitaxially grown by molecular beam epitaxy are investigated. These structures are grown on passivated Si substrates, resulting in one single epitaxial domain and its twinned domain, both sharing the same out-of-plane orientation. Supported by X-ray diffraction and Raman spectroscopy, attention is called to the thermodynamically driven tendency of GeTe and Sb2Te3 to intermix into a Ge− Sb−Te (GST) alloy at the interfaces. A growth model is proposed to explain how these GST structures are formed.

I

thus trading elastic energy for electronic stability. The weaker side of the bonds could then be accommodated near the silicon surface, thus reducing the interaction with the surface and mimicking van der Waals epitaxy.6 The present work shows that these improvements in the quality of the two materials directly translate into an improved CSL structure with both materials epitaxially stacked on top of each other. However, attention is drawn to the intermixing occurring at the interfaces between the sublayers. First, as a demonstration of the superior quality of the CSLs grown on passivated silicon substrates, compared to those grown on unpassivated surfaces, XRD φ-scans are performed in order to investigate the in-plane epitaxial relationship between the film and the substrate. Along with a reference φ-scan on the silicon substrate, and reference scans on Sb2Te3 films grown on Si(111)-(7 × 7) and Si(111)-(√3 × √3)R30°-Sb surfaces, scans on CSL 10 × [Sb2Te3(6 nm)/GeTe(4 nm)] structures grown on the same surfaces are shown in Figure 1a. These CSL structures start with Sb2Te3 as their first sublayer. In all cases, the reflections from the same set on planes are measured: Si{2 2 0} in cubic notation, Sb2Te3 {0 1 1̅ 5} from its hexagonal cell, and the equivalent planes in the CSL. When a superlattice structure is deposited on the Si(111)-(7 × 7) surface starting with a first layer of Sb2Te3, in-plane twisted domains and twinned domains are already formed in this first sublayer. As the GeTe sublayer is grown on top of the Sb2Te3, these same twisted domains are propagated into the GeTe layer and all subsequent sublayers, such that the whole CSL adopts the domains formed initially in the first layer. If the first Sb2Te3 sublayer is deposited on a Si(111)-(√3 × √3)R30°-Sb surface instead, in-plane twisted domains are

n the quest for the best material candidates in phase change memory (PCM) applications, GeTe/Sb2Te3 chalcogenide superlattice structures (CSL) have boasted better performances compared to their homogeneous Ge−Sb−Te (GST) counterparts.1 Not only did the CSL cells operate at a lower current, their cyclability was also improved; no significant changes were observed in their electric properties even after millions of cycles. Beyond their application as PCM, CSLs are also investigated for applications utilizing their topological insulator properties.2−4 In the latter case, an accurate control of the interfaces and the van der Waals epitaxy is mandatory to allow for the topological properties to be expressed. The superior phase change properties in CSLs have been interpreted as a switching mechanism occurring within the crystalline state. And in this perspective, it is beneficial to improve the crystalline quality in the CLSs to gain a clearer view of the phenomena at the core of these enhanced switching properties. It was recently reported that in-plane twisted domains were observed in Sb2Te3 and GeTe thin films grown by molecular beam epitaxy (MBE) on a Si(111)-(7 × 7) surface and that these twisted domains could be suppressed by growing on a Si(111)-(√3 × √3)R30°-Sb passivated surface instead, thus improving the crystalline quality.5,6 For Sb2Te3, this improvement is ascribed to the full passivation of the silicon surface, enabling van der Waals epitaxy instead of an epitaxial relationship dictated by the numerous dangling bonds of the Si(111)-(7 × 7) surface.5 Sb2Te3 is a two-dimensional (2D) lamellar material with resonantly bonded quintuple atomic layers separated by gaps, across which they interact through van der Waals forces. Therefore, the possibility for van der Waals epitaxy on passivated surfaces could be anticipated. GeTe on the other hand is not commonly considered as a 2D material; the fact that rotational domains could be similarly suppressed in the case of GeTe is more surprising. This is explained by resonant bonding and Peierls distortions, breaking the crystalline periodicity by splitting the highly delocalized p orbitals into longer weaker lobes and shorter stronger lobes, © XXXX American Chemical Society

Received: December 3, 2015 Revised: April 5, 2016

A

DOI: 10.1021/acs.cgd.5b01714 Cryst. Growth Des. XXXX, XXX, XXX−XXX

Crystal Growth & Design

Article

GeTe(4 nm)] structure grown on Si(111)-(√3 × √3)R30°-Sb is shown in Figure 2a. In the same figure, ω−2θ scans for epitaxial Sb2Te3 and GeTe thin films grown on the same surface are shown as a comparison. All spectra are plotted in reciprocal lattice units Qz = (4π/λ) × sin(θ). The sharpest peaks at Qz = 2.00, and 4.01 Å−1 are reflections coming from the Si(111) substrate. Next to those, there are two groups of reflections, centered on their most intense peak at Qz = 1.82 and 3.64 Å−1. These are the superlattice peaks, surrounded by their satellite peaks related to the superlattice periodicity.7 Because the second order occurs at exactly at twice the Qz value of the first order, these are Bragg reflections. From its position, it is determined that the first superlattice peak originates from the diffraction by a set of planes periodically spaced by Δ1 = 3.452 Å. This spacing is ascribed to the average distance between successive Te atomic layers in the growth direction. The exact distance between Te layers is dependent on what is inserted in between on the other sublattice: either Ge, Sb, or nothing in the case of a van der Waals gap.8,9 But XRD is an averaging technique, and the dominant peak results from the convolution of the Fourier transforms from each of the different Te−Te periods. To further comment on the satellite peaks around the superlattice peaks, they are spaced by a distance of 0.06 Å−1, which corresponds to a periodicity of 10.4 nm in real-space, in agreement with the nominal superlattice periodicity of 10 nm. No higher order satellite reflections are measured, indicating that random fluctuations and linear deviation in the periodicity are causing their loss of intensity and broadening.10 While the crystalline quality is demonstrated, the regularity of the structure at a longer range in the out-of-plane direction can still be improved. The peaks at Qz = 2.44 and 3.08 Å−1 in the CSL spectrum match with the Sb2Te3(0 0 0 12) and Sb2Te3(0 0 0 15) reflections, although there is a slight shift toward lower Qz values that could be due to the stacking of the material into the CSL. The other features to be investigated are the broad features between Qz = 3.1 and 3.4 Å−1. This part of the spectrum is more clearly shown in the magnified view presented in Figure 2b. Except from a clear peak at Qz = 3.08 Å−1, that has been assigned to Sb2Te3(0 0 0 15), the shoulder is very broad and difficult to resolve in a meaningful way. However, from the TEM investigation conducted on these CSLs,9 the intermixing of GeTe and Sb2Te3 into GST at the interfaces has been observed. Recent EXAFS measurements on these superlattices led to the same conclusion.12 In the following we show that these broad features could be assigned to the presence of GST of different compositions in the CSL. As a starting point, insights can be gained by establishing the link between the crystalline structure of Sb2Te3 and its XRD spectrum. As shown in Figure 2a, The Sb2Te3(0 0 0 12) and Sb2Te3(0 0 0 15) peaks at Qz = 2.44 and 3.08 Å−1 split the distance between Sb2Te3(0 0 0 9) and Sb2Te3(0 0 0 18) into three equal parts. For Sb2Te3 these superstructure reflections are caused by the van der Waals gaps found every three Te atomic layers. It is sensible that these van der Waals gaps diffract strongly, considering that they are regularly spaced and electronically depleted, in clear contrast from the rest of the resonantly bonded crystal. In the ordered phases of Ge1Sb2Te4, Ge2Sb2Te5, and Ge3Sb2Te6, the van der Waals gaps are found each 4, 5, and 6 Te layers, respectively. Therefore, because of the presence of GST in the CSL, additional features can be expected

Figure 1. (a) XRD φ-scans showing the {0 1 1̅ 5} reflections from Sb2Te3 films grown on Si(111)-(7 × 7) and Si(111)-(√3 × √3)R30°-Sb, compared with reflections from the equivalent planes in CSLs grown on the same surfaces, with Sb2Te3 as their first layer. (b) Similar comparison between films of GeTe, and CSL grown with GeTe as its first layer, on Si(111)-(√3 × √3)R30°-Sb. Substrate Si{220} reflections (the equivalent planes in silicon) are shown as a reference in (a) and (b).

suppressed from the first layer, and they are suppressed as well throughout the whole CSL. For a CSL starting with a first layer of GeTe, the results are similar: The whole CSL conforms to the in-plane orientations of the first GeTe layer, and twisted domains are still suppressed with the Si(111)-(√3 × √3)R30°-Sb surface [shown in Figure 1b]. The only difference is that twinned domains are strongly suppressed when GeTe is grown on the Si(111)-(√3 × √3)R30°-Sb surface,6 whereas they are formed again in the CSL as Sb2Te3 is grown on top of GeTe. They then propagate further into the rest of the CSL. A CSL with GeTe as its first layer is in fact nontrivial to engineer using other methods than MBE. For deposition techniques such as RF sputtering or physical vapor deposition, a buffer layer of Sb2Te3 is often applied on amorphous substrates, in order to utilize the intrinsic tendency of the material to texture itself in the out-ofplane direction, owing to its 2D nature. Starting the CSL with GeTe would mean to forego this advantage. Overall, these results show that the growth of CSL with one single in-plane orientation (and its twinned domain) has been achieved by starting with a Si(111)-(√3 × √3)R30°-Sb reconstruction. The crystalline quality of the CSL from the point of view of in-plane orientation is comparable to that of the Sb2Te3 single layers [fwhm of 2.8° in the φ-scans for the CSL and 1.8° for the Sb2Te3 film in Figure 1a]. To investigate the out-of-plane epitaxial relationship between the CSL film and substrate, symmetric ω−2θ XRD scans are performed. One such measurement on a 10 × [Sb2Te3(6 nm)/ B

DOI: 10.1021/acs.cgd.5b01714 Cryst. Growth Des. XXXX, XXX, XXX−XXX

Crystal Growth & Design

Article

Figure 2. (a) Symmetric ω-2θ XRD scan from the CSL 6/4, with reference spectra from GeTe and Sb2Te3 thin films. Δ1′ is the distance between CSL first order and second order reflections. (b) Zoomed viewgraph of (a) for CSL 6/4, 3/2, and 3/1. Additional features at distances Δ3′, Δ4′, Δ5′, and Δ6′ from CSL second order, correspond respectively to 1/3, 1/4, 1/5, and 1/6 of the separation Δ1′. (c) Schematic representation of the CSL in real space with Te atoms in gray, Sb in blue, and Ge in orange. Δ1 corresponds to the distance between two successive Te layers, Δ3, Δ4, Δ5, and Δ6 correlate to 3, 4, 5, and 6 time Δ1. Sb2Te3 and GST compounds are highlighted using colored frames (scale model realized with VESTA software11).

intercalated between Sb2Te3 blocks and never successively repeated, the envelope of their Fourier transform can only be very broad. This is also the reason why only one order of these superstructure peaks is observed; the lack of repetition of the GST blocks, coupled with the dispersion in the CSL periodicity, strongly suppress the higher order reflections. An idealized schematic representation of GST inside the CSL is shown in Figure 2c to help in the understanding of the grown structure. The Te atoms are colored in gray, highlighting the continuous Te sublattice spanning throughout the CSL. Sb atoms are shown in blue, and Ge in orange, the Sb2Te3 and GST blocks are highlighted using colored frames. Equipped with the knowhow to interpret the XRD spectra of these structures, CSLs grown with different stackings and sublayer thicknesses can be compared and the intermixing can be assessed in each case. Because the phase-change properties

approximately at intervals of 1/4, 1/5, and 1/6 of the distance between the CSL first order and CSL second order peaks. These GST peaks will not be located exactly at the Δ4′ = 0.459 Å−1, Δ5′ = 0.361 Å−1, and Δ6′ = 0.279 Å−1 distances from the CSL peaks, because the Te−Te distance changes depending on the species filling the other sublattice, and the average Te−Te distance inside the GST block is different from Δ1, the average Te−Te distance in the whole superlattice. In addition, the composition in the GST blocks observed in the CSL could deviate from that of the stoichiometric compounds. Up to a certain point, van der Waals gaps would still form each 4, 5, or 6 Te layers, but the distance between these layers would then change. The average composition in the CSL was however confirmed by TEM-EDX measurements.9 It is also important to point out that sharp XRD features for these GST blocks cannot be expected. Because they are C

DOI: 10.1021/acs.cgd.5b01714 Cryst. Growth Des. XXXX, XXX, XXX−XXX

Crystal Growth & Design

Article

vibrational modes;13 their shift can be explained by the fact that the Sb2Te3 is stacked in the CSL with another material, influencing its vibrational properties. The Eg(1) and Eg(2) inplane modes are not subjected to any shift. As for the two modes from GeTe at 80.1 and 121.4 cm−1, they are not observed at all in the CSL spectrum. Instead, the asymmetric shoulder at 103.8 cm−1 is assigned to the GST Raman mode found at the same position. In this specific case, with GeTe sublayers with a thickness of only 1 nm, the Raman data seems to indicate that all the GeTe deposited is intermixed into GST. Intermixing between GeTe and Sb2Te3 is a sensible outcome, considering how likely they are to form GST ternary compounds when the right conditions are met.17 In addition, GeTe and GST are both materials containing an important quantity of intrinsic vacancies18,19 that are predicted to provide convenient pathways for atomic migration.20,21 This is most likely thermally amplified by the fact that the CLS is typically grown at 230−250 °C substrate temperatures with the MBE. What is more surprising is that the atomic species have a clear tendency to segregate into separate layers,9 whereas a system entirely driven by diffusion would lead to a homogeneous alloy. STEM investigations also revealed that quasi pure Te−Sb−Te sequences were observed in the bottom half of the GST blocks, whereas the top part contained more intermixed Te−Ge−Te/ Te−Sb−Te sequences. These pure bottom layers are again in contradiction with a system that would be driven by diffusion only, and it would seem that intermixing only occurred from top to bottom, since only the top parts of the GST blocks were truly affected. On the basis of these observations, a new growth model is proposed, explaining step by step the formation of the observed natural GST structures during the growth of the CSL structures. Scheme 1 illustrates schematically the presented process. On the left-hand side, the diagram starts with two layers of Sb2Te3. As the deposition of GeTe is initiated, the impinging GeTe is able to bind with the topmost Sb2Te3 block. This could happen either directly, despite the passivated van der Waals surface, or possibly at defect sites or from the sides of Sb2Te3 layers that are not fully completed. Once bound, the top surface is most likely Te terminated, as the Te surface has been predicted to be much more stable than a Ge surface.22 At this point, a GST compound is already formed, but the Ge atoms sit at the edge of the GST structure, whereas Da Silva et al.8 have shown that it would be more favorable energetically for them to be gathered into pure layers near the middle of the blocks. In addition, a gradient in concentration is also formed, encouraging the Ge atoms to diffuse toward the lower layers. Considering that the impinging atoms possess a thermal budget and that growth is performed at a temperature of 230−250 °C, which is for GST above the metastable to stable transition temperature,23 enough thermal energy is provided for Ge and Sb atoms to exchange their positions, pushing Sb toward the newly formed surface. An analogy can be drawn with the use of Sb as a surfactant for the growth of pure Ge.24 This process continues as more GeTe is deposited, resulting into a natural GST structure at the end of the deposition of what was first intended as a GeTe sublayer. This structure then conveniently possesses a weakly interacting top surface that can host the growth of the following Sb2Te3 sublayer. The scenario described in Scheme 1 is an idealized case. In reality, the migration of Ge atoms toward the center of the GST blocks during the growth of a CSL is usually incomplete and

of CSLs are predicted to be linked to the interfaces between GeTe and Sb2Te3, an effort was devoted to reduce the thickness of each sublayer, increasing the proportion these interfaces occupy in the film. Three selected CSL are shown in Figure 2b, with sublayers thicknesses reduced from CSL 6/4 down to 3/1, all of which start with Sb2Te3 as their first layer. As intended, the superlattice satellite peaks shift away from the main peak with decreasing thickness, matching the respective superlattice periodicity. More surprisingly, although CSL 6/4 and 3/2 share the same GeTe/Sb2Te3 ratio, the position of the CSL second order peak changes, indicating that the average composition differs. Therefore, the relationship between the out of plane lattice spacing and the sublayer thicknesses may not be simply linear. As for the Sb2Te3 and GST features, their relative intensity change with the stacking sequence. GST reflections seem to become more intense as the sublayer thickness decreases, showing that intermixing is localized at the interfaces. The intensity of these reflections also seems to be further increased in CSL 3/2, which is richer in GeTe. And this is already a first hint that the intermixing is occurring primarily during the deposition of GeTe. To corroborate that intermixing is occurring during growth, the CSLs were investigated qualitatively using Raman spectroscopy. The spectrum acquired on a 15x[Sb2Te3(3 nm)/ GeTe(1 nm)] CSL (CSL 3/1) is shown in Figure 3 (black

Figure 3. Raman spectrum acquired on the CSL 3/1 with a 632.8 nm laser in z(y, xy)−z geometry. Spectra are acquired on GeTe, Sb2Te3, and Ge3Sb2Te6 films and shown as reference. The Raman modes are listed for Sb2Te3,13 GeTe,14,15 and GST.16

curve), along with reference spectra acquired on GeTe (orange), Sb2Te3 (blue), and Ge3Sb2Te6 (red) films grown by MBE as well. Strong similarities are immediately observed between the CSL and Sb2Te3, all four modes at 46.3, 69.2, 111.6, and 166.4 cm−1 are reproduced. These two spectra are distinguished from each other by overall broader modes for the CSL. The mode at 69.2 cm−1 is also visibly less intense and softened, while the mode at 166.4 cm−1 is strengthened to 170.4 cm−1. These A1g(1) and A1g(2) modes are out-of-plane D

DOI: 10.1021/acs.cgd.5b01714 Cryst. Growth Des. XXXX, XXX, XXX−XXX

Crystal Growth & Design

Article

Revealed by XRD, Raman, and TEM analysis,9 intermixing is observed between the GeTe and Sb2Te3 sublayers at their interfaces, forming different GST compounds. A growth model is drafted from these observations, attributing the intermixing to the growth of GeTe on top of Sb2Te3. Knowing that intermixing is bound to occur, it may be of interest to directly engineer Sb2Te3/GST superlattices, so that the higher concentration of Ge in the upper half of the GST blocks can be avoided. Fundamentally, despite the involvement of GeTe, a 3D material, the final structure of the CSL is a stacked heterostructure of 2D materials and should be treated as such in the investigation and simulation of these structures.

Scheme 1. Step by Step Schematic Representation of the Process by Which Natural GST Is Formed during the Deposition of GeTe onto Sb2Te3 in the CSL structure



ASSOCIATED CONTENT

* Supporting Information S

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.cgd.5b01714. Experimental details about sample preparation by MBE and characterization by XRD and Raman spectroscopy (PDF)



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This work was partially supported by EU within the FP7 Project PASTRY (GA 317746). We thank S. Behnke and C. Stemmler for technical support at the MBE system and M. Ramsteiner for support in Raman spectroscopy. J. Momand and B. J. Kooi are acknowledged for useful discussions and M. Hanke for careful reading of the manuscript.



imperfect, resulting in mixed Te−Ge−Te/Te−Sb−Te sequences in the top parts of the GST blocks. There is a thermodynamic advantage to form the ordered compound, but the dynamics have not allowed the equilibrium state to be reached. And there is certainly also some entropy introduced by diffusion. This growth model is most elegant in explaining why the bottom half of the GST blocks seems untouched by the intermixing: Because once the Ge atoms reach the center of the structure, they are in their most stable configuration. This stability barrier prevents diffusion from pushing the Ge atoms further down into the lower part of the GST block. In a metastable GST or in a growing GST layer, randomly distributed vacancies can provide pathways for diffusion. A different scenario might instead appear if one considers a natural GST, where the vacancies are already gathered into layers, so that diffusion of single atoms might be less probable. The exact mechanism for the formation of natural GST during growth or annealing should be further investigated with the help of theoretical calculations. In conclusion, the successful growth of Sb2Te3/GeTe superlattices has been demonstrated. To be mentioned in particular, the growth on the Sb passivated Si(111)-(√3 × √3)R30°−Sb surface ensures the whole CSL is textured with one single out-of-plane and in-plane orientation (plus its twinned domain). By MBE, growth can be initiated with GeTe just as well as with Sb2Te3, but lower interface roughness and narrower peaks were still obtained for the CSL starting with Sb2Te3.

ABBREVIATIONS MBE, molecular beam epitaxy; XRD, X-ray diffraction; GST, GeSbTe alloy; PCM, phase change memory; CSL, chalcogenide superlattice structures



REFERENCES

(1) Simpson, R. E.; Fons, P.; Kolobov, A. V. V; Fukaya, T.; Krbal, M.; Yagi, T.; Tominaga, J. Nat. Nanotechnol. 2011, 6, 501−505. (2) Bang, D.; Awano, H.; Tominaga, J.; Kolobov, A. V.; Fons, P.; Saito, Y.; Makino, K.; Nakano, T.; Hase, M.; Takagaki, Y.; Giussani, A.; Calarco, R.; Murakami, S. Sci. Rep. 2014, 4, 1−7. (3) Saito, Y.; Tominaga, J.; Fons, P.; Kolobov, A. V.; Nakano, T. Phys. Status Solidi RRL 2014, 8, 302−306. (4) Sa, B.; Zhou, J.; Sun, Z.; Tominaga, J.; Ahuja, R. Phys. Rev. Lett. 2012, 109, 096802. (5) Boschker, J. E.; Momand, J.; Bragaglia, V.; Wang, R.; Perumal, K.; Giussani, A.; Kooi, B. J.; Riechert, H.; Calarco, R. Nano Lett. 2014, 14, 3534−3538. (6) Wang, R.; Boschker, J. E.; Bruyer, E.; Di Sante, D.; Picozzi, S.; Perumal, K.; Giussani, A.; Riechert, H.; Calarco, R. J. Phys. Chem. C 2014, 118, 29724−29730. (7) Cummins, H. Z. Phys. Rep. 1990, 185, 211−409. (8) Da Silva, J.; Walsh, A.; Lee, H. Phys. Rev. B: Condens. Matter Mater. Phys. 2008, 78, 224111. (9) Momand, J.; Wang, R.; Boschker, J. E.; Verheijen, M. A.; Calarco, R.; Kooi, B. J. Nanoscale 2015, 7, 19136−19143. (10) Xiu, L.; Wu, Z. J. Appl. Phys. 1992, 71, 4892. (11) Momma, K.; Izumi, F. J. Appl. Crystallogr. 2011, 44, 1272−1276. E

DOI: 10.1021/acs.cgd.5b01714 Cryst. Growth Des. XXXX, XXX, XXX−XXX

Crystal Growth & Design

Article

(12) Casarin, B.; Caretta, A.; Momand, J.; Kooi, B. J.; Verheijen, M. A.; Bragaglia, V.; Calarco, R.; Chukalina, M.; Yu, X.; Robertson, J.; Lange, F. R. L.; Wuttig, M.; Redaelli, A.; Varesi, E.; Parmigiani, F.; Malvestuto, M. Sci. Rep. 2016, 6, 22353. (13) Sosso, G. C.; Caravati, S.; Bernasconi, M. J. Phys.: Condens. Matter 2009, 21, 095410. (14) Fons, P.; Kolobov, A. V.; Krbal, M.; Tominaga, J.; Andrikopoulos, K.; Yannopoulos, S.; Voyiatzis, G.; Uruga, T. Phys. Rev. B: Condens. Matter Mater. Phys. 2010, 82, 2−6. (15) Steigmeier, E. F.; Harbeke, G. Solid State Commun. 1970, 8, 1275−1279. (16) Němec, P.; Moreac, A.; Nazabal, V.; Pavlišta, M.; Přikryl, J.; Frumar, M. J. Appl. Phys. 2009, 106. (17) Deringer, V. L.; Dronskowski, R. J. Phys. Chem. C 2013, 117, 15075−15089. (18) Krbal, M.; Kolobov, A.; Fons, P.; Tominaga, J.; Elliott, S.; Hegedus, J.; Giussani, A.; Perumal, K.; Calarco, R.; Matsunaga, T.; Yamada, N.; Nitta, K.; Uruga, T. Phys. Rev. B: Condens. Matter Mater. Phys. 2012, 86, 045212. (19) Matsunaga, T.; Yamada, N. Phys. Rev. B: Condens. Matter Mater. Phys. 2004, 69, 104111. (20) Deringer, V. L.; Lumeij, M.; Stoffel, R. P.; Dronskowski, R. Chem. Mater. 2013, 25, 2220−2226. (21) Yu, X.; Robertson, J. Sci. Rep. 2015, 5, 12612. (22) Deringer, V. L.; Lumeij, M.; Dronskowski, R. J. Phys. Chem. C 2012, 116, 15801−15811. (23) Bragaglia, V.; Jenichen, B.; Giussani, A.; Perumal, K.; Riechert, H.; Calarco, R. J. Appl. Phys. 2014, 116, 054913. (24) Meyer, G.; Voigtländer, B.; Amer, N. M. Surf. Sci. 1992, 274, L541−L545.

F

DOI: 10.1021/acs.cgd.5b01714 Cryst. Growth Des. XXXX, XXX, XXX−XXX