The direct epitaxial growth of multilayer BN by atomic layer deposition is of critical significance for two dimensional device applications. To date, however, epitaxial growth has only been reported on graphene or on transition metal surfaces. X-ray photoelectron spectroscopy (XPS) and low energy electron diffraction (LEED) demonstrate layer-by-layer BN epitaxy on a monolayer of RuO2(110) formed on a Ru(0001) substrate. Growth was accomplished with BCl3/NH3 cycles at 600 K substrate temperature and subsequent annealing in ultrahigh vacuum. This yielded stoichiometric BN layers, Cl impurities levels of ≲1 at. %, and an average BN film thickness linearly proportional to the number of BCl3/NH3 cycles. XPS data indicate negligible charge transfer or band bending for the BN/RuO2 interface. LEED data indicate a 30° rotation between the coincident BN and oxide lattices. The atomic layer epitaxy of BN on an oxide surface suggests new routes to the direct growth and integration of graphene and BN with industrially important substrates, including Si(100).

The direct growth of multilayer h-BN(0001) films without exfoliation or physical transfer is critical for industrial scale development of 2D-based devices. This includes a range of beyond-CMOS electronic or spintronic applications, both as a spin filter1–4 and as substrate for graphene epitaxy.5,6 Atomic layer deposition (ALD), with precise control of epitaxial BN film thickness, has obvious advantages for some of these applications.4,5 To date, however, most investigations of single or few-layer BN growth, whether by ALD, chemical vapor deposition, or magnetron sputter deposition have involved BN growth on clean transition metal surfaces,5–10 although polycrystalline BN growth on Si(100) has recently been reported.11 The ability to grow large area h-BN(0001) films epitaxially on oxide substrates would significantly increase the types of suitable substrates for direct growth of h-BN and h-BN/graphene heterostructures for device applications. Of additional interest are the possible effects of an interfacial oxide layer—however thin—on BN/metal interactions, including interfacial orbital hybridization.8,12–14 Yang et al.15 showed that oxygen intercalation at elevated temperatures between Ru(0001) and BN can modify or eliminate the strong interfacial orbital hybridization observed between a BN monolayer and clean Ru(0001) substrate, although effects on the relative orientations of the BN and Ru lattices were not discussed. The results reported here demonstrate for the first time the direct epitaxial growth of h-BN(0001) on a RuO2(110) substrate, using an ALD process. The results further demonstrate that a monolayer of RuO2 disrupts the strong orbital hybridization that occurs at the BN/Ru metal interface.

The ALD process employed here involves alternate BCl3 and NH3 exposures at 600 K, similar to that employed for multilayer h-BN(0001) growth on Co(0001).5,16 This chemistry has been shown capable of producing conformal, amorphous BN films on ZrO2 particles.17 The relatively mild temperatures involved and the precise control over film thickness inherent in the ALD process make this method attractive for device fabrication. Since multilayers of BN, even if physically transferred, are superior to monolayers for spin injection,18 the advantages of ALD over the often-used but monolayer self-limiting borazine pyrolysis method10,14 are evident. ALD also has inherent advantages over MBE (Ref. 19) or sputter magnetron deposition,6 particularly for proposed spin filter applications,2,4 where precise, atomic level control of film thickness, and azimuthal alignment between layers, are important for device performance.

In this paper, x-ray photoelectron spectroscopy (XPS) and low energy electron diffraction (LEED) data are presented, demonstrating layer-by-layer epitaxy of h-BN(0001) on a ∼monolayer of RuO2(110) formed on Ru(0001). The data indicate that the BN lattice is aligned with that of the Ru(0001) substrate, with an absence of the strong BN 2p/Ru(3d) orbital hybridization observed at the BN interface with metallic Ru(0001).8,10,15 XPS and LEED also indicate that the ALD process itself leaves intact the RuO2(110) monolayer, demonstrating that this ALD process is suitable for preserving the surface structures and thicknesses of ultrathin oxide films during deposition.

The Ru(0001) film used was deposited on sapphire [Al2O3(0001)] by magnetron sputter deposition at 700 K in an Ar plasma. Film roughness was evaluated ex situ by AFM after magnetron sputter deposition, and after further annealing (to 1000 K) in UHV. The final rms roughness of the film used was <1 nm.

The system used for BN ALD growth and characterization has been described previously.5 Briefly, the system included a chamber for ALD (base pressure 3 × 10−6 Torr) and a chamber for XPS and LEED analyses in UHV (base pressure 3 × 10−10 Torr). Pressures in the ALD chamber were recorded by a combination of nude ion and baratron gauges, while pressures in UHV were recorded with a nude ion gauge calibrated for N2. Sample temperatures were varied by resistive heating and measured by a K type thermocouple mounted on the sample stage near the sample. Samples were transported between the chambers without exposure to ambient conditions by use of a magnetically coupled feedthrough.

XPS data were acquired with a 100 mm mean radius hemispherical analyzer operating at a constant pass energy mode (44 eV), with nonmonochromatic Al Kα x-ray source operated at 15 keV, 300 W. XPS spectra were analyzed with commercially available software. Relative atomic concentrations and thicknesses were calculated by standard methods20 using inelastic mean free paths through BN; 30.42 and 27.15 Å for Ru 3d and O 1s photoelectrons, respectively.21 The XPS-derived thickness of a BN monolayer was taken to be 3.3 Å, in agreement with previous STM studies of multilayer BN on Ru(0001)22 and of multilayer graphene.23 Spectra were corrected for any sample charging by placing the peak of the metallic Ru 3d5/2 feature at 280.1 eV.24 

LEED images were obtained using a commercial three grid reverse-view LEED apparatus. LEED images were captured at electron beam energies from 300 to 60 eV before significant charging of the sample (due to ∼5 ML insulating BN) obscured results at lower energies. After that, LEED images were captured at electron beam energies from 400 to 220 eV.

Prior to BN ALD, Ru was annealed at 1000 K in UHV for 2 h, and then cooled to room temperature. The sample was subsequently annealed at 1000 K in oxygen (1 × 10−7 Torr) for 30 min to remove contamination resulting from ambient exposure. The sample was then further annealed at 1000 K for 4.5 h more in UHV to order the surface region of the film.

BCl3 and NH3 gases were electronic grade, obtained from commercial vendors and used without further purification. The turbopump to the deposition chamber was closed off during each exposure, and a dedicated roughing pump was used to create a flow through system during exposures. BCl3 pressure during an exposure was 250 mTorr for 5 min, and NH3 pressure was 350 mTorr for 2 min, all at 600 K. Between BCl3 and NH3 exposures, the rough pump was closed off and the turbopump reopened for 2 min to the system to allow any excess gas to be pumped away.

Space filling models of the coincident BN(0001) and RuO2(110) direct space lattices were generated using commercially available software25 using literature values for atomic diameters and bond lengths.26 

XPS Ru 3d and O 1s spectra of the sample prior to ALD are shown in Figs. 1(a) and 1(b), respectively. The Ru 3d5/2 feature can be decomposed into a main peak at 280.1 eV, corresponding to metallic Ru,24 and a smaller feature near 281.0 eV, consistent with RuO2 (Ref. 27) (supplementary material, Fig. S1).28 The Ru and O XPS data therefore confirm the presence of a thin Ru oxide overlayer, and the relative intensities of the O 1s and metallic Ru 3d5/2 features indicate an average RuO2 thickness of ∼2 Å.20 Corresponding LEED data are shown in Fig. 2(a). The LEED data before ALD are consistent with RuO2(110)29 although a smearing of some oxide-related LEED spots suggests a certain amount of disorder in the oxide layer. Such disorder is also suggested by the breadth of the O 1s spectrum of the surface prior to ALD [Fig. 1(b)], possibly due to the presence of adsorbed OH and/or H2O species on the RuO2 layer.30–32 The XPS and LEED data therefore indicate that, prior to ALD, the substrate surface is an ordered RuO2(110) film, perhaps partially hydroxylated, with an average thickness of ∼2 Å, approximately that of an Ru–O bond length.33 The data therefore indicate the presence of an RuO2(110) film ∼1 monolayer thick.

Fig. 1.

(Color online) XPS spectra of Ru and O core levels. (a) Ru 3d and (b) O1s: (black trace) before h-BN(0001) deposition, (red trace) after 20 BCl3/NH3 exposures, (blue trace) after 800 K anneal in UHV.

Fig. 1.

(Color online) XPS spectra of Ru and O core levels. (a) Ru 3d and (b) O1s: (black trace) before h-BN(0001) deposition, (red trace) after 20 BCl3/NH3 exposures, (blue trace) after 800 K anneal in UHV.

Close modal
Fig. 2.

(Color online) LEED of evolution of ALD BN on RuO2(110)/Ru(0001). (a) RuO2(110)/Ru(0001); (b) One monolayer BN (4 BCl3/NH3 cycles) on RuO2(110)/Ru(0001); (c) Two monolayers BN (8 BCl3/NH3 cycles) on RuO2(110)/Ru(0001). All LEED acquired at 80 eV beam energy.

Fig. 2.

(Color online) LEED of evolution of ALD BN on RuO2(110)/Ru(0001). (a) RuO2(110)/Ru(0001); (b) One monolayer BN (4 BCl3/NH3 cycles) on RuO2(110)/Ru(0001); (c) Two monolayers BN (8 BCl3/NH3 cycles) on RuO2(110)/Ru(0001). All LEED acquired at 80 eV beam energy.

Close modal

ALD induces only a small shift in the Ru 3d feature to higher binding energy [Fig. 1(a)], although the intensity is reduced, as expected, by the presence of the BN overlayer. The small shift to higher binding energy is consistent with a slight additional oxidation, or chlorination, of the surface during the ALD process, but this small shift is largely reversed upon annealing the BN/Ru surface to 800 K in UHV [Fig. 1(a)].

ALD induces significant change in the O 1s feature [Fig. 1(b)], including the attenuation of the low binding energy feature below 530 eV, and the broadening of the O1s feature to higher binding energies at ∼533.5–530 eV, centered near 532.5 eV. That the total O 1s intensity—unlike the Ru 3d intensity—is not reduced upon formation of the BN overlayer, and indeed increases after 20 BCl3/NH3 cycles [Fig. 1(b)], indicates that a portion of the total O1s signal originates from the BN overlayer, rather than at the BN/RuO2 interface. The observed O 1s binding energies are consistent with the formation of B suboxide or hydroxide.34 Annealing of the BN/Ru sample to 800 K in UHV significantly reduces the width and intensity of the O1s feature. After annealing, the O 1s peak maximum is observed to shift back toward 532 eV. However, the overall O 1s intensity remains greater than the observed prior to ALD. These observations demonstrate that the BN ALD process, under these experimental conditions, results in some oxygen content within the growing BN film, and this is significantly but not completely reversed by annealing in UHV.

A comparison of the surface sensitive (80 eV beam energy) LEED spectrum before and after the deposition of one monolayer of BN(0001) (4 BCl3/NH3 cycles) is given in Figs. 2(a) and 2(b), respectively. The unit cells of the BN(0001) and RuO2(110) reciprocal lattices are outlined. The persistence of the Ru oxide layer in the LEED spectrum upon deposition of the first BN monolayer is evident. Also evident is the coincidence between the BN(0001) and RuO2(110) reciprocal lattices. Extra spots are apparent in the 1 ML BN/RuO2/Ru(0001) spectrum, which are attributable to double layer scattering.8 These spots are not observed after the deposition of the second BN layer [Fig. 2(c)], consistent with the extra spots in Fig. 2(b) being due to double layer scattering. The data in Figs. 2(b) and 2(c) indicate that the BN lattice is rotated 30° with respect to the oxide substrate, and is azimuthally aligned with the lattice of the Ru(0001) metallic substrate. The data in Figs. 1 and 2 therefore indicate that ALD of BN on a monolayer of RuO2(110)/Ru(0001) preserves the structure of the oxide substrate lattice while yielding highly ordered h-BN(0001) overlayers.

XPS B 1s and N 1s spectra are shown in Figs. 3 after 4 and 20 BCl3/NH3 ALD cycles, and after 800 K UHV anneal, corresponding to the formation BN films (before anneal) with XPS-derived average thicknesses of ∼1 and 5 ML. Corresponding B 1s and N 1s peak binding energies are also listed in Table I and compared to relevant binding energies from the literature. The data in Fig. 3(a) and Table I demonstrate that upon formation of the first ALD monolayer, a B 1s binding energy of 191.8 eV is observed, but that this value relaxes toward 191.1 eV as the BN film thickens. In contrast, no such change is observed in the N 1s peak binding energy with increasing film thickness [Fig. 3(b), Table I]. Upon annealing in UHV, however, both B 1s and N 1s peak binding energies shift toward larger binding energies—191.6 and 398.7 eV, respectively—coincident with loss of Cl and some O [Figs. 3(b) and 1(a), respectively] and a visible narrowing of the B 1s feature. The observed B 1s and N 1s binding energies after the anneal are close in value to those reported5 for an annealed BN film of comparable thickness on Co(0001). The energy difference between the N 1s and B 1s features (Δ, Table I) is, however, ∼0.6 eV smaller than that reported for a borazine-derived BN monolayer on Pt(111)35 or for the borazine-derived BN monolayer on Ru with an intercalated oxide layer.15 

Fig. 3.

(Color online) XPS spectra of (a) B 1 s and (b) N 1 s for h-BN(0001) film on RuO2(110)/Ru(0001): (black trace) after 4 BCl3/NH3 cycles; (red trace) after 20 BCl3/NH3 cycles; (blue trace) after 800 K UHV anneal. Cl 2p feature is marked by arrow in (a).

Fig. 3.

(Color online) XPS spectra of (a) B 1 s and (b) N 1 s for h-BN(0001) film on RuO2(110)/Ru(0001): (black trace) after 4 BCl3/NH3 cycles; (red trace) after 20 BCl3/NH3 cycles; (blue trace) after 800 K UHV anneal. Cl 2p feature is marked by arrow in (a).

Close modal
Table I.

Binding energies of B 1s, N 1s, and Δ after 4 BCl3/NH3 exposures, after 8 BCl3/NH3 exposures, after 20 BCl3/NH3 exposures, after a 30 min 800 K anneal, and relevant data from literature.

BNΔ
4 BCl3/NH3 cycles 191.8 398.1 206.3 
8 BCl3/NH3 cycles 191.5 398.1 206.6 
20 BCl3/NH3 cycles 191.1 398.0 206.9 
After 30 800 K 191.6 398.7 207.1 
6 ML BN on Co(0001) (Ref. 5191.2 398.3 206.8 
1 ML BN on Ru(0001) (Ref. 15190.94 398.99, 398.02 208.05, 207.26 
1 ML BN + oxygen intercalation, 100 °C (Ref. 15190.66 398.46 207.8 
Monolayer h-BN on Pt foil (Ref. 35189.9 397.6 207.7 
BNΔ
4 BCl3/NH3 cycles 191.8 398.1 206.3 
8 BCl3/NH3 cycles 191.5 398.1 206.6 
20 BCl3/NH3 cycles 191.1 398.0 206.9 
After 30 800 K 191.6 398.7 207.1 
6 ML BN on Co(0001) (Ref. 5191.2 398.3 206.8 
1 ML BN on Ru(0001) (Ref. 15190.94 398.99, 398.02 208.05, 207.26 
1 ML BN + oxygen intercalation, 100 °C (Ref. 15190.66 398.46 207.8 
Monolayer h-BN on Pt foil (Ref. 35189.9 397.6 207.7 

The Cl 2p intensity [Fig. 3(a), arrow] indicate a Cl atomic concentration of ∼5 at. % after 20 BCl3/NH3 cycles, but after a brief 800 K anneal in UHV, the Cl concentration in the BN layer decreased to <1 at. %, in line with previous ALD BN results on Co(0001).5 These data also indicate a somewhat broad B 1s feature, prior to anneal, which is not inconsistent with a small amount of oxidized or chlorinated B near ∼193 eV.34,36 The narrowing of the B 1s feature after the anneal suggests the removal/reduction of such species. However, the proximity of the Cl 2p feature prevents a firm conclusion on this point.

The evolution of XPS-derived BN average film thickness and the corresponding XPS-derived B:N atomic ratio with the number of BCl3/NH3 cycles are displayed in Fig. 4. The increase in average thickness is linear with the number of BCl3/NH3 cycles, as expected for an ALD process. The y-intercept of the linear regression of the average thickness of BN does not go through 0, indicating that the initial sticking coefficient of BN on RuO2 is higher than the sticking coefficient on BN, consistent with what was observed for BN grown by magnetron sputtering on clean Ru(0001).6 The data in Fig. 4 also show that during the ALD process, the B:N atomic ratio is consistently 1.1:1(±0.1), with the estimated uncertainty related to the measurement of the XPS intensities. After annealing in UHV, however, the B:N atomic ratio exhibits a slight reduction to the expected value of 1.0:1(±0.1). This is consistent with the annealing-induced removal of a small B impurity, either B-O or B-Cl. In summary, the data in Figs. 2–4 demonstrate the deposition of an epitaxial BN(0001) layer on Ru(0001), with an average BN film thickness that is linearly proportional to the number of BCl3/NH3 cycles.

Fig. 4.

(Color online) (black squares) Thickness after each deposition and after 800 K anneal as a function of the number of BCl3/NH3 exposures. (green squares): B:N atomic ratio after each deposition and after 800 K anneal as a function of the number of BCl3/NH3 exposures. Straight lines are least-square fits to data.

Fig. 4.

(Color online) (black squares) Thickness after each deposition and after 800 K anneal as a function of the number of BCl3/NH3 exposures. (green squares): B:N atomic ratio after each deposition and after 800 K anneal as a function of the number of BCl3/NH3 exposures. Straight lines are least-square fits to data.

Close modal

The data in Figs. 1 and 2 indicate that ALD BN on monolayer RuO2(110) on Ru(0001) yields an epitaxial BN layer rotated 30° with respect to the oxide substrate and azimuthally aligned with the lattice of the metallic Ru(0001) underlayer. Data in Fig. 1(b), however, demonstrate that the ALD process under these conditions is accompanied by inclusion of an oxygen impurity. The degree of impurity can be estimated by calculating the total O 1s intensity due to the interfacial oxide layer, attenuated by ∼5 ML of BN, and subtracting this amount from the total O 1s intensity observed after ALD of BN but prior to annealing in UHV [Fig. 1(b)]. The O 1s intensity after this subtraction is then the approximate intensity due to noninterfacial O—i.e., to impurities incurred during ALD. Using an estimated O 1s photoelectron inelastic mean free path of 27.20 Å thru BN, the O 1s “impurity intensity” is then the total O 1s intensity less the corrected intensity due to interfacial oxygen. This results in an estimated B:O atomic impurity ratio of 8:1 prior to annealing and 10:1 B:O after annealing in UHV.

The data in Figs. 3(a) and 4 also indicate that the ALD process at the temperatures reported here yields an observable Cl impurity, which is greatly reduced upon annealing in UHV. The reduced width of the B 1s spectrum after annealing [Fig. 3(a)] and the corresponding change in observed B:N stoichiometry from 1.1:1 to 1.0:1 (Fig. 4) suggests that this Cl impurity is associated with volatile BClx species, which are desorbed during the annealing step.

The decrease in impurity O, Cl, and excess B upon annealing demonstrate that while the growth temperature chosen allows the ALD reaction to proceed, the inclusion of impurities is observable by XPS. As a practical matter, therefore, the production of device-worthy BN multilayers by this process requires some further optimization—possibly alteration of exposure conditions, substrate temperatures, or more frequent annealing steps, in order to produce films with levels of impurities too small to be observed by XPS.

The LEED and XPS data (Figs. 2 and 3 and Table I) provide further insight into how a monolayer of oxide alters BN/Ru electronic interactions. STM and XPS data demonstrate that a BN monolayer produced by either borazine pyrolysis10,15 or by ALD with BCl3 and NH3 (Ref. 8) results in strong Ru/BN orbital hybridization. This correlates with the observation of two N 1s bonding environments at the BN/Ru interface produced by borazine pyrolysis10,15 shifting to one broad peak at lower binding energy upon oxygen intercalation at elevated temperatures15 (Table I). In contrast, the N 1s feature in the growth process reported here exhibits a constant binding energy (prior to anneal in UHV) of ∼398.1 eV at all BN film thicknesses. In contrast, the B 1s binding energy [Fig. 3(a), Table I] exhibits a monotonic decrease in binding energy with film thickness prior to annealing. Given that the N 1s binding energy remains constant as the B 1s binding energy decreases, this shift cannot be ascribed to band bending at the interface, but instead reflects an interfacial B–substrate interaction that is attenuated as additional BN layers are formed. We suggest that the increased B 1s binding energy observed in the first BN layer on Ru (Table I) reflects some B interaction with the surface oxide. Thus, the XPS data for BN/RuO2(110) suggests a weak but observable B-surface oxide interaction, and negligible N-RuO2(110) interaction. This is in contrast to BN interactions with clean Ru, where strong N/Ru orbital hybridization is apparent.8,15

The shift of both B 1s and N 1s binding energies to higher energy upon annealing in UHV (Fig. 3, Table I) coincides with removal of Cl impurity, and therefore suggests some band bending due to the potential of Cl impurities at or near the interface. After the anneal, the B 1s and N 1s binding energies are in good agreement with those previously reported for multilayer BN(0001) film of comparable thickness on Co(0001).5 

The difference in N1s and B1s binding energies (Δ, Table I) is independent of band bending or energy scale calibration, and allows comparison of values for both multilayer and single layer BN films. The Δ value observed here, ∼207 eV (Table I), is in very good agreement with that reported for an ALD 6 ML BN(0001) film on Co(0001) formed under similar reaction conditions.5 This value, however, is ∼0.6–0.8 eV smaller than those generally reported for monolayer BN films derived by borazine pyrolysis on transition metal substrates.10,13,15,35 Whether this reflects on some fundamental differences between films produced by borazine pyrolysis versus BCl3/NH3 ALD, or points to some difference in electronic structures for epitaxial multilayer or single layer films, requires further study.

Finally, the monolayer of interfacial oxide alters the orientation of the BN lattice relative to the substrate. Previous ALD of BN on Ru(0001) produced a monolayer of BN rotated 30° with respect to the Ru lattice.8 In contrast, the data in Fig. 2 demonstrate that the BN lattice is rotated 30° with respect to the oxide lattice, and aligned with the metal Ru(0001) lattice beneath the surface oxide. The data in Fig. 2(b) demonstrate that this rotation produces a close coincidence between the BN(0001) and RuO2(110) lattices. The growth of epitaxial BN(0001) on RuO2(110)/Ru(0001) is therefore an example of BN growth on a coincident substrate. This is illustrated in Fig. 5, which shows the overlay and unit cells of the BN(0001) and RuO2(110) direct space in-plane lattices. Since graphene, Co(0001), and Ni(111) in-plane lattice constants are all similar to BN, this suggests multiple schemes for the epitaxial integration of graphene and BN with Co and Ni on RuO2(110) for spintronic applications.2,4 Indeed, the ability to grow smooth RuO2(110) on oxide films such as MgO(100),31 which can in turn be grown epitaxially on Si(100)37 suggests a route toward the integration of 2D-based nanoelectronics or spintronics devices with Si(100).

Fig. 5.

(Color online) Simulated space filling model of h-BN(0001) deposited on RuO2 (110). The substrate is RuO2 (110) (black = Ru and red = O). The overlayer is BN (green = N and blue = B). Unit cells shown overlaid in bottom left corner.

Fig. 5.

(Color online) Simulated space filling model of h-BN(0001) deposited on RuO2 (110). The substrate is RuO2 (110) (black = Ru and red = O). The overlayer is BN (green = N and blue = B). Unit cells shown overlaid in bottom left corner.

Close modal

As a final note of caution, the data reported here provide no definitive conclusion regarding the continuity of the BN layer with respect to the presence/absence of pinholes or islands, etc. Transport/tunneling measurements on fabricated structures are therefore of obvious significance. However, the quenching of Ru metal-related LEED spots and spots related to double layer scattering after the deposition of only a second BN monolayer [Figs. 2(b) and 2(c)], as well as the linearity of XPS-derived film thicknesses with ALD precursor cycles (Fig. 4), strongly suggest that this process produces uniform, macroscopically continuous BN films suitable for device applications, consistent with previous applications of this process on Co(0001)5 and on Ru(0001).8 

XPS and LEED data demonstrate the epitaxial growth of stoichiometric h-BN(0001) layers on RuO2(110)/Ru(0001) using a BCl3/NH3 ALD process at a growth temperature of 600 K, followed by annealing in UHV at 800 K. The oxide monolayer remains intact during the deposition process, and the interfacial oxide inhibits the strong interfacial interactions observed for BN on clean Ru(0001). The BN average film thickness is linearly proportional to the number of BCl3/NH3 cycles. The BN layers are also in azimuthal alignment with each other and coincident with the RuO2(110) lattice. The data demonstrate that this ALD BCl3/NH3 process has potential for the formation of multilayer epitaxial BN films for a variety of beyond-CMOS applications.

This work was supported by C-SPIN, a funded center of STARnet, a Semiconductor Research Corporation (SRC) program sponsored by MARCO and DARPA under task IDs 2381.001 and 2381.006; and by National Science Foundation under Grant No. ECCS-1508991. B.B. also gratefully acknowledges a STARnet undergraduate internship. B.P. was supported by the National Science Foundation Research Experience for Undergraduates program under Grant No. CHE-S1461027.

1.
L.
Ci
 et al,
Nat. Mater.
9
,
430
(
2010
).
2.
S. V.
Faleev
,
S. S. P.
Parkin
, and
O. N.
Mryasov
,
Phys. Rev. B
92
,
235118
(
2015
).
3.
L.
Britnell
 et al,
Nano Lett.
12
,
1707
(
2012
).
4.
V. M.
Karpan
,
P. A.
Khomyakov
,
G.
Giovannetti
,
A. A.
Starikov
, and
P. J.
Kelly
,
Phys. Rev. B
84
,
153406
(
2011
).
5.
M. S.
Driver
,
J. D.
Beatty
,
O.
Olanipekun
,
K.
Reid
,
J. A.
Kelber
,
A.
Rath
, and
P. M.
Voyles
,
Langmuir
32
,
2601
(
2016
).
6.
P.
Sutter
,
J.
Lahiri
,
P.
Zahl
,
B.
Wang
, and
E.
Sutter
,
Nano Lett.
13
,
276
(
2013
).
7.
A.
Nagashima
,
N.
Tejima
,
Y.
Gamou
,
T.
Kawai
, and
C.
Oshima
,
Phys. Rev. Lett.
75
,
3918
(
1995
).
8.
C.
Bjelkevig
,
Z.
Mi
,
J.
Xiao
,
P. A.
Dowben
,
L.
Wang
,
W.-N.
Mei
, and
J. A.
Kelber
,
J. Phys. Condens. Matter
44
,
302002
(
2010
).
9.
M. V.
Kamalakar
,
A.
Dankert
,
J.
Bergsten
,
T.
Ive
, and
S. P.
Dash
,
Sci. Rep.
4
,
6146
(
2014
).
10.
A.
Goriachko
,
Y.
He
,
M.
Knapp
,
H.
Over
,
M.
Corso
,
T.
Brugger
,
S.
Berner
,
J.
Osterwalder
, and
T.
Greber
,
Langmuir
23
,
2928
(
2007
).
11.
A.
Haider
,
C.
Ozgit-Akgun
,
E.
Goldenberg
,
A. K.
Okyay
, and
N.
Biyikli
,
J. Am. Ceram. Soc.
97
,
4052
(
2014
).
12.
L.
Kong
 et al,
J. Phys. Chem. C
114
,
21618
(
2010
).
13.
A. B.
Preobrajenski
,
A. S.
Vinogradov
, and
N.
Mårtensson
,
Surf. Sci.
582
,
21
(
2005
).
14.
A. B.
Preobrajenski
,
A. S.
Vinogradov
, and
N.
Mårtensson
,
Phys. Rev. B
70
,
165404
(
2004
).
15.
Y.
Yang
,
Q.
Fu
,
M.
Wei
,
H.
Bluhm
, and
X.
Bao
,
Nano Res.
8
,
227
(
2015
).
16.
J.
Beatty
,
Y.
Cao
,
I.
Tanabe
,
M. S.
Driver
,
P. A.
Dowben
, and
J. A.
Kelber
,
Mater. Res. Express
1
,
46410
(
2014
).
17.
J. D.
Ferguson
,
A. W.
Weimer
, and
S. M.
George
,
Thin Solid Films
413
,
16
(
2002
).
18.
T.
Yamaguchi
,
Y.
Inoue
,
S.
Masubuchi
,
S.
Morikawa
,
M.
Onuki
,
K.
Watanabe
,
T.
Taniguchi
,
R.
Moriya
, and
T.
Machida
,
Appl. Phys. Express
6
,
73001
(
2013
).
19.
Z.
Zuo
,
Z.
Xu
,
R.
Zheng
,
A.
Khanaki
,
J.-G.
Zheng
, and
J.
Liu
,
Sci. Rep.
5
,
14760
(
2015
).
20.
M. P.
Seah
,
Practical Surface Analysis
, edited by
D.
Briggs
and
M. P.
Seah
, 2nd ed. (
Wiley
,
West Sussex
,
1997
), pp.
200
255
.
21.
S.
Tanuma
,
C. J.
Powell
, and
D. R.
Penn
,
Surf. Interface Anal.
35
,
268
(
2003
).
22.
J. A.
Kelber
,
Graphene, Carbon Nanotubes and Nanostructures: Techniques and Applications
, edited by
J. E.
Morris
and
K.
Iniewski
(
CRC
,
Boca Raton, FL
,
2013
), pp.
89
113
.
23.
M.
Xu
,
D.
Fujita
,
J.
Gao
, and
N.
Hanagata
,
ACS Nano
4
,
2937
(
2010
).
24.
J.
Moulder
,
W.
Stickle
,
P.
Sobol
, and
K.
Bomben
,
Handbook of X-Ray Photoelectron Spectroscopy
(
Perkin-Elmer Corporation
,
Eden Prairie
,
1992
).
25.
K.
Momma
and
F.
Izumi
,
J. Appl. Crystallogr.
44
,
1272
(
2011
).
26.
O. O.
Kurakevych
and
V. L.
Solozhenko
,
Acta Crystallogr., Sect. C
63
,
i80
(
2007
).
27.
K. S.
Kim
and
N.
Winograd
,
J. Catal.
35
,
66
(
1974
).
28.
See supplementary material at http://dx.doi.org/10.1116/1.4972784 for a deconvolution of the Ru(3d)5/2 spectrum prior to ALD, and after ALD and annealing to 800 K in UHV.
29.
Y. D.
Kim
,
A. P.
Seitsonen
, and
H.
Over
,
Surf. Sci.
465
,
1
(
2000
).
30.
W. E.
O'Grady
,
L.
Atanasoka
,
F. H.
Pollak
, and
H. L.
Park
,
J. Electroanal. Chem
178
,
61
(
1984
).
31.
D.
Rochefort
,
P.
Dabo
,
D.
Guay
, and
P. M. A.
Sherwood
,
Electrochim. Acta
48
,
4245
(
2003
).
32.
Y.
Gao
,
G.
Bai
,
Y.
Liang
,
G. C.
Dunham
, and
S. A.
Chambers
,
J. Mater. Res.
12
,
1844
(
1997
).
33.
Q.
Sun
,
K.
Reuter
, and
M.
Scheffler
,
Phys. Rev. B
70
,
235402
(
2004
).
34.
Y.
Wang
and
M.
Trenary
,
Chem. Mater.
5
,
199
(
1993
).
35.
J.
Park
,
J. C.
Park
,
S. J.
Yun
,
H.
Kim
,
D. H.
Luong
, and
S. M.
Kim
,
ACS Nano
8
,
8520
(
2014
).
36.
Y.
Wang
,
J.
Fan
, and
M.
Trenary
,
Chem. Mater.
5
,
192
(
1993
).
37.
X. Y.
Chen
,
K. H.
Wong
,
C. L.
Mak
,
J. M.
Liu
,
X. B.
Yin
,
M.
Wang
, and
Z. G.
Liu
,
Appl. Phys. A
74
,
703
(
2002
).

Supplementary Material