Atomic-Scale Structure and Local Chemistry of ... - ACS Publications

Feb 23, 2016 - WPI Advanced Institute for Materials Research, Tohoku University, 2-1-1 ... Department of Physics, University of York, Heslington, York...
0 downloads 0 Views 7MB Size
Download PDF
Letter pubs.acs.org/NanoLett

Atomic-Scale Structure and Local Chemistry of CoFeB−MgO Magnetic Tunnel Junctions Zhongchang Wang,† Mitsuhiro Saito,†,# Keith P. McKenna,*,†,‡ Shunsuke Fukami,§,∥ Hideo Sato,§,∥ Shoji Ikeda,§,∥,⊥ Hideo Ohno,*,†,§,∥,⊥ and Yuichi Ikuhara*,†,#,∇ †

WPI Advanced Institute for Materials Research, Tohoku University, 2-1-1 Katahira, Aoba-ku, Sendai 980-8577, Japan Department of Physics, University of York, Heslington, York YO10 5DD, United Kingdom § Center for Spintronics Integrated Systems, Tohoku University, 2-1-1 Katahira, Aoba-ku, Sendai 980-8577, Japan ∥ Center for Innovative Integrated Electronic Systems, Tohoku University, 468-1 Aramaki, Aza, Aoba-ku, Sendai 980-8577, Japan ⊥ Laboratory for Nanoelectronics and Spintronics, Research Institute of Electrical Communication, Tohoku University, 2-1-1 Katahira, Aoba-ku, Sendai 980-8577, Japan # Institute of Engineering Innovation, University of Tokyo, 2-11-16, Yayoi, Bunkyo-ku, Tokyo 113-8656, Japan ∇ Nanostructures Research Laboratory, Japan Fine Ceramics Center, 2-4-1 Mutsuno, Atsuta, Nagoya 456-8587, Japan ‡

S Supporting Information *

ABSTRACT: Magnetic tunnel junctions (MTJs) constitute a promising building block for future nonvolatile memories and logic circuits. Despite their pivotal role, spatially resolving and chemically identifying each individual stacking layer remains challenging due to spatially localized features that complicate characterizations limiting understanding of the physics of MTJs. Here, we combine advanced electron microscopy, spectroscopy, and first-principles calculations to obtain a direct structural and chemical imaging of the atomically confined layers in a CoFeB−MgO MTJ, and clarify atom diffusion and interface structures in the MTJ following annealing. The combined techniques demonstrate that B diffuses out of CoFeB electrodes into Ta interstitial sites rather than MgO after annealing, and CoFe bonds atomically to MgO grains with an epitaxial orientation relationship by forming Fe(Co)-O bonds, yet without incorporation of CoFe in MgO. These findings afford a comprehensive perspective on structure and chemistry of MTJs, helping to develop high-performance spintronic devices by atomistic design. KEYWORDS: CoFeB−MgO, magnetic tunnel junction, atomic structure, local chemistry, scanning transmission electron microscopy

M

if the CoFeB electrode is made sufficiently thin.9 Other major criteria for practical applications have also been proven to be satisfied by this system10,11 and with the latest improvements the CoFeB−MgO-based MTJ is now regarded as a critical building block for coming applications.12 However, the specific crystal structure, local chemistry, and local bonding of all stacking layers that make up the MTJs remain unclear, hindering the establishment of relationship between the material properties and nanoscale structures, although such information is of significance in enhancing performances of MTJs. In particular, the interfacial anisotropy, which originates from the hybridization of Fe (Co) 3d and O 2p orbitals,13,14 depends critically on the crystalline structure and bonding at the CoFeB/MgO interface. Moreover, the composition, chemical states, and local defects in CoFeB and MgO are known to be critical for transport and magnetic

agnetic tunnel junctions (MTJ) are at the heart of magnetic random access memory (MRAM) and spintronics-based integrated circuits, which serve as a novel paradigm of electronics owing to its nonvolatile feature and capability of fast operation with virtually unlimited endurance. The fundamental element of a MTJ consists of an atomically thin insulating tunnel barrier sandwiched in between two ferromagnetic electrodes. Since the demonstration of tunneling magnetoresistance (TMR) at room temperature,1,2 a number of breakthroughs have been made, leading to development of high-density spin-transfer torque (STT) MRAM that is currently being intensively pursued by industry. A giant TMR effect was identified in a (001)-oriented MgO MTJ with (001) CoFe electrodes due to coherent electron tunneling of the Δ1 Bloch states.3,4 A large TMR ratio up to 604% at room temperature was demonstrated in the CoFeB−MgO system5 with the easy axis of the magnetic electrodes lying in the plane of the film.5−7 Subsequently, the same CoFeB−MgO MTJs were found to exhibit a perpendicular easy axis with lower critical currents for STT switching for a given thermal stability8 © XXXX American Chemical Society

Received: September 8, 2015 Revised: January 12, 2016

A

DOI: 10.1021/acs.nanolett.5b03627 Nano Lett. XXXX, XXX, XXX−XXX

Letter

Nano Letters

Figure 1. Structure of the MTJ. Cross-sectional HRTEM image of (a) as-deposited and (b) annealed MTJ samples at 500 °C. The upper insets show selected-area diffraction patterns taken at the thin film and the lower ones show those taken at the Si substrate.

were deposited on thermally oxidized Si(001) substrates using RF magnetron sputtering with a base pressure of less than 10−7 Pa at room temperature. The structure of the MTJs comprised from the substrate are Si/SiO2/Ta(5 nm)/Ru(10 nm)/Ta(5 nm)/Co20Fe60B20(5 nm)/MgO(2.1 nm)/Co20Fe60B20(3 nm)/ Ta(5 nm)/Ru(5 nm) (the digit gives a nominal thickness), where the most essential portion is CoFeB/MgO/CoFeB, as shown schematically in Figure 1. The MgO layers were deposited from a high-purity MgO target at a pressure of 10 mTorr in Ar atmosphere. A portion of the MTJ samples were subsequently annealed at 500 °C in a vacuum of 10−4 Pa for 1 h under a magnetic field of 400 mT. TEM Specimen Preparation and Imaging Condition. Cross-section thin-foil specimens for TEM and STEM observations were prepared by cutting, grinding, and dimpling samples down to ∼20 nm. In the argon ion-beam thinning process, a gun voltage of 1−4 kV and an incident beam angle of 4−6° were used to decrease radiation damage. The SADP and HRTEM images were taken using the JEOL JEM-2010F conventional microscope and the FEI TITAN 80-300 aberration corrected microscope (CEOS GmbH) that were operated at 200 and 300 kV, respectively. HAADF and ABF images were taken using the ARM-200F (Cold FEG) STEM operated at 200 kV, which was equipped with a probe corrector (CEOS GmbH), providing an unprecedented opportunity to probe structures with a sub-angstrom resolution. For the HAADF STEM imaging, a probe convergence angle of ∼22 mrad and a detector with an inner semiangle of over 60 mrad were adopted. ABF STEM images were taken using a detector of 11−23 mrad, and EELS was recorded using a Gatan Enfina system equipped on the STEM with an energy resolution at full width at half-maximum of ∼0.3 eV. HAADF STEM image simulations were performed using the WinHREM STEM package (HREM Research Inc.) based on multislice method. Debye−Waller factors were considered for every element involved, yet were averaged over spatial directions, namely, anisotropy of the absorption of thermal diffuse scattering factors was ignored.

properties. Likewise, to design optimal stacks for the fabrication process, clarifying how each species of atom diffuses and structure changes via annealing is both timely and relevant. So far, although a few first-principles calculations have been conducted, they draw contradictory conclusions regarding structure of the formed MTJs following annealing.15−17 Experimentally, although the structural and chemical issues in the CoFeB−MgO MTJs have been addressed utilizing transmission electron microscopy (TEM), X-ray photoemission spectroscopy, and secondary ion mass spectroscopy,18−21 a consistent picture has not emerged yet, largely due to the intricacy of defect structure, the challenge of resolving such a highly complicated heterojunction, and the difficulty of probing local atomic details of the MTJs with a confined dimensionality. In particular, there have been many contradictory results on the fate of B following annealing including diffusion into Ta,22 segregation at the CoFe/MgO interface,16,18 B diffusion into MgO forming a magnesium boride phase,17,19 and diffusion of B into MgO as a dilute substitutional impurity.20,21,23 As a consequence, most of the theoretical models have been put forward based upon assumed structures only.24 Here, we combine aberration (Cs)-corrected TEM, highangle annular dark-field25−29 (HAADF) and annular brightfield30 (ABF) imaging in Cs-corrected scanning TEM (STEM), electron energy-loss spectroscopy (EELS),31 and first-principles calculation to uncover atomic-scale structure and local chemistry of confined stacking layers in the CoFeB−MgObased MTJ exhibiting a high TMR and offer direct evidence of how B diffuses in the MTJs via annealing. Such combined techniques allow us to demonstrate that B diffuses out of the crystalline CoFeB into Ta interstitial sites rather than into the MgO tunnel barrier after annealing, and CoFe atomically bonds to the textured MgO grains with an epitaxial orientation relationship by forming Fe(Co)−O bonds yet with no Co or Fe incorporation in MgO. The findings are of fundamental significance for facilitating development of spintronic devices for MRAMs and spintronics-based logics. Experimental and Calculational Details. Fabrication of Magnetic Tunnel Junctions. All magnetic tunnel junctions B

DOI: 10.1021/acs.nanolett.5b03627 Nano Lett. XXXX, XXX, XXX−XXX

Letter

Nano Letters

Figure 2. Local chemistry of the MTJ. A dark-field STEM image and its corresponding electron energy-loss spectroscopy (EELS) mapping of the Fe L2,3, Co L2,3, B K, and O K edges for (a) as-deposited and (b) annealed samples at 500 °C. The MgO tunnel barrier at the center of MTJ is highlighted by two dotted lines. The core-loss images were made at the same places as the ADF images.

Density Functional Theory (DFT) Calculations. Calculations were conducted using the Vienna ab initio simulation package (VASP) within the framework of DFT.32 The projector augmented wave (PAW) method was employed and the generalized gradient approximation (GGA) of Perdew−Burke−Ernzerhof (PBE) was used to describe exchange and correlation. The 2s and 2p electrons of O and B, 2p and 3p electrons of Mg, and 3d and 4s electrons of Co and Fe were treated as valence electrons and expanded in a plane wave basis with a cutoff energy of 400 eV. A Monkhorst− Pack grid of 9 × 9 × 9 k-points was adopted for calculations on primitive unit cells and equivalent k-point densities were used for supercells (up to 513 atoms). Interface models consisted of a 10-layer Fe3Co(001) slab connected to a MgO(001) slab of six layers. A vacuum gap of 10 Å was included in the supercell to avoid unwanted interactions between the slab and its periodic images. All atoms in the slabs were fully relaxed until the magnitude of the force on each atom fell below 0.05 eV/Å. For the charged defect calculations in MgO, total energies are corrected for potential alignment and artificial image-charge interactions.33 Results and Discussion. To characterize the MTJ, we first conducted a high-resolution TEM (HRTEM) imaging of the as-deposited sample that reveals clear and abrupt interfaces between the stacking layers with designated thickness (Figure 1a), meaning a successful fabrication of the MTJ. Analyses of diffraction patterns of the Si substrate and thin film (Figure 1a, inset) indicate that the view direction is along the [110] zone axis of Si substrate and some layers are of textured crystalline nature along the [100] axis. Upon closer inspection, we find that both the top and bottom Ta layers are amorphous, while the Ta layer on top of the bottom Ru layer is crystalline. In contrast, both the upper and lower Ru layers are of good crystallinity with a columnar structure, even though they are grown on amorphous Ta. Importantly, the two CoFeB layers are indeed amorphous and exhibit clear and sharp interfaces with the crystalline MgO barrier with a thickness of ∼2.1 nm, as can also be confirmed in the Cs-corrected HRTEM image (Supporting Information Figure S1a). Upon postannealing, the Ta, Ru, and MgO layers maintain the same structure as in the as-deposited sample but the two CoFeB electrodes undergo a phase transformation from

amorphous to crystalline CoFe (Figure 1b), consistent with previous reports.34 Bright and dark image contrasts are identified in the MgO and the two crystallized CoFe layers, implying that they form columnar grains with sharp grain boundaries in them. However, the CoFe/MgO interface remains clear and abrupt with no secondary-phase or transitional layers (Figure 1b), implying a direct bonding of the two materials, as also confirmed in a Cs-corrected HRTEM image (Supporting Information Figure S1b). To shed light on the texture of the nanocrystallites in the CoFe and MgO of the annealed MTJ, we show Fourier transformation (FFT) diffraction patterns (diffractogram) of a local area in both the MgO barrier and two CoFe electrodes (Supporting Information Figure S1b), uncovering that grains are highly (001)textured, which is critical to realizing a large TMR ratio. To extract the chemical information on the stacking layers in the MTJs before and after annealing, we conducted EELS mappings of the Fe L2,3, Co L2,3, B K, and O K edges by focusing on a region from the bottom to the top Ru layers (Figure 2). Interestingly, B in the as-deposited MTJ distributes uniformly in the two CoFeB layers with negligible amounts in the MgO layer, yet diffuses substantially to its neighboring Ta layers rather than to the MgO or Ru layers after postannealing (Figure 2),22 indicating a low solid solubility of B in highly crystalline CoFe and MgO. This also implies the absence of the B−O bonding and Mg3B2O6 compound in the MgO highlighting a fundamental impact of a clean and abrupt CoFe/ MgO interface on the TMR ratio and the interfacial anisotropy. Conversely, Co and Fe distribute homogeneously in the CoFeB layer in both as-deposited and annealed samples with no clear diffusion into either the MgO or Ta layers following annealing (Figure 2), which implies a sharp interface between CoFe and MgO with negligible interdiffusion. Likewise, O is confined to within the MgO barrier irrespective of the annealing of sample, implying that the Ta and Ru metallic layers are not oxidized via postannealing. To probe the orientational relationship between CoFe and MgO and to extract precise atomic information on the interface, we present HAADF STEM images of the CoFe/ MgO/CoFe heterojunction in both the as-deposited and annealed samples viewed from both [100] and [110] direction of a nanocrystallite in MgO (Figure 3). The CoFeB layers are C

DOI: 10.1021/acs.nanolett.5b03627 Nano Lett. XXXX, XXX, XXX−XXX

Letter

Nano Letters

Figure 3. Structural transformation of the MTJ by annealing. HAADF STEM images of the (a,b) as-deposited and (c,d) annealed samples taken from (a,c) [100] and (b,d) [110] direction of a nanocrystal in MgO.

Figure 4. Atomic-resolution imaging. HAADF STEM images of the CoFe/MgO interface in the annealed sample at 500 °C taken from (a) [100] and (b) [110] direction of MgO. The insets show the corresponding simulated images.

atomic columns,35 implying that brighter spots on the upper side represent Co/Fe columns, while darker ones on the lower side are either normal MgO columns (viewed from [100] direction of MgO) or individual Mg and O columns ([110] direction of MgO). One notices that Mg and O columns can be clearly discriminated in the HAADF image from the [110] direction of MgO. Clearly, this interface, which has a small lattice mismatch of 3.7%, is atomically abrupt and coherent without any secondary-phase layers, amorphous layers, or transition regions, confirming a clean and direct bonding of CoFe to MgO at the atomic scale. The Co/Fe atoms are found to sit on top of O atoms, forming Fe−O and Co−O bonds at the interface, in line with the variation in the interfacial magnetic anisotropy along the perpendicular direction via annealing. By measuring the magnetization hysteresis loop using vibrating sample magnetometry, we find that the saturation magnetization and interfacial perpendicular magnetic anisotropy increase by a factor of 1.3 and 2.7, respectively, after annealing (Supporting Information Figure S3 and Table S1). It has been reported that annealing can enhance interfacial anisotropy, and such an effect is more pronounced for the stack with increased B content,36 which can be attributed to the

confirmed to be of amorphous nature and the MgO barrier exhibits a domain structure with either [100] or [110] nanocrystallites in the as-deposited sample. Upon annealing, lattice fringes of the CoFe and MgO can be clearly identified, demonstrating crystallization of the CoFe electrodes. Interestingly, the domain size of MgO is much smaller than that of CoFe, but the MgO grains can bond atomically to CoFe, possessing an in-plane epitaxial orientation relationship (001) [100]MgO∥(001)[110]CoFe resulting from a rotation by 45° along the [001] direction (Figure 3c,d). These observations are also supported by the ABF STEM images, which are recorded simultaneously with the HAADF images (Supporting Information Figure S2). Formation of such a coherent and atomically abrupt CoFe/MgO interface is consistent with the experimentally observed enhancement of the TMR ratio of the CoFeB−MgO MTJs via postannealing.5 To extract exact bonding information, we present in Figure 4 high-magnification HAADF STEM images of the CoFe/MgO interface in the annealed sample viewed from both [100] and [110] direction of MgO. Because intensity of an atomic column in the HAADF imaging mode is proportional to ∼ Z1.7 (Z: atomic number), contrast is brighter for heavier atoms in D

DOI: 10.1021/acs.nanolett.5b03627 Nano Lett. XXXX, XXX, XXX−XXX

Letter

Nano Letters increased number of Fe−O and Co−O bonds at the interface as a result of the B diffusion and consequent crystallization. Note that although the termination species of the CoFe layer at the interface cannot be directly determined from the measurement of magnetic properties or structural characterization, the TEM study clearly identifies that the Co/Fe atoms sit on top of O atoms (rather than Mg atoms), which should facilitate the understanding of physics of magnetism localized at the interface. As a further confirmation, we established a CoFe/MgO atomic model by taking Fe 3 Co to represent CoFe 37 (Supporting Information Figure S4), because our atomicresolution energy-dispersive X-ray spectroscopy (EDS) reveals that the Co/Fe has a composition of one-third in the annealed sample (also consistent with the Co/Fe ratio in CoFeB sputtering target), and performed DFT calculations. The interface model consists of a 10-layer Fe3Co(001) slab connected to a MgO(001) slab of 6 layers with a 10 Å vacuum gap included (Supporting Information). We simulated the HAADF STEM images from both [100] and [110] directions of MgO (Figure 4, insets) using the fully optimized interface model (Supporting Information Figure S4) and compared them with their experimental counterparts (Figure 4), finding good agreement for both projections, which is in further support of a direct bonding of the electrodes to the MgO barrier. As a result of the direct bonding, we find interestingly that the interface induces no electronic states in the forbidden gap of MgO (Supporting Information Figure S5) that can open up preferential tunneling pathways, which should benefit device performances. To further explore which species of atom in CoFe electrodes is more favorable to bond MgO, we also calculated adhesion energies of two models of the Fe3Co/MgO interface (terminated with a 50% Co/50% Fe layer or a 100% Fe layer) as well as those of the Co/MgO and Fe/MgO. In all cases, we find that binding of Co to O is slightly more favorable than Fe (but only by 0.1 J/m2) for the Fe3Co/MgO interface (Supporting Information). We note that although Fe oxides usually have more negative heat of formation than Co oxides15 the preference for Co−O bonds may result from other effects such as the formation energy of the free metal surface, in-plane strain, and local structural, electronic, and magnetic perturbations near the interface. However, because the energy difference is so small one may expect either (or both) types of interface to be present depending on the growth conditions. An interesting area for further work would be to consider how the interfacial anisotropy and TMR is affected by these different terminations but this is beyond the scope of the present study, which is focused on atomic scale structure. To gain insights into physical mechanism of the diffusion process, we employed a first-principles thermodynamics method to assess the tendency for B, Co, and Fe to incorporate in MgO and Ta.38 The incorporation of B in MgO at interstitial sites, anion substitutional, and cation substitutional sites, B in Ta at interstitial sites, and Co and Fe in MgO at interstitial and cation sites is considered (Figure 5, inset). The formation energy Ef of a given defect can be expressed as39 E f = [Et(def) − Et(ideal) +

∑ ΔNiμi + ΔqEF] i

Figure 5. Defect formation energy. B related defect formation energy in MgO in (a) O-poor and (b) O-rich condition. A total of five types of defects are taken into account, including B at an interstitial site (Bi(MgO)), a cation site (BMg), and an anion site (BO) of MgO and at the interstitial sites of fcc (Bi(Ta fcc)) and bcc Ta (Bi(Ta bcc)). The numbers next to the lines indicate the B oxidation state for the most stable defects in each case. Defect formation energy of Fe and Co in MgO in (c) O-poor and (d) O-rich condition. A total of four species of defects are considered, including Fe at an interstitial (Fei(MgO)) and a cation site (FeMg), and Co at an interstitial (Coi(MgO)) and a cation site (CoMg) of MgO. The insets show the corresponding atomic models.

ideal and the defective supercell, μi is the chemical potential of each species, Δq is the difference in total charge between the two supercells, and EF is the Fermi energy. We considered Fe3Co with a low concentration of B as thermodynamic reservoir for B and the following two extremes of O chemical potential: the O-rich case, which corresponds to the chemical potential of half an oxygen molecule, and the O-poor case, which corresponds to the chemical potential of O when MgO is reduced to pure Mg metal. The Fe and Co bulk crystals are adopted to define Fe and Co chemical potentials. The electron chemical potential (Fermi energy) depends on the band offsets between CoFeB and MgO. For each possible B, Co, and Fe defect in MgO or Ta, we considered all possible charge states and calculated their defect formation energies as a function of Fermi energy (Figure 5). Figure 5a,b shows formation energies of B in MgO and Ta for the O-poor and O-rich cases. In the O-poor case (Figure 5a), we find that for the typical band offsets in the range 3−4 eV, the most stable B defect in MgO involves the incorporation

(1)

t

where E (def) is the total energy of the MgO (or Ta) supercells containing the B, Co, or Fe defects, Et(ideal) is the total energy of the corresponding ideal MgO (or Ta) supercell, ΔNi is the difference in number of atoms of each species i between the E

DOI: 10.1021/acs.nanolett.5b03627 Nano Lett. XXXX, XXX, XXX−XXX

Letter

Nano Letters

highly confined features that complicate characterizations, posing a significant hurdle to further improving TMR and functionality of the MTJs. By combining advanced TEM with first-principles calculations, we elucidate the fine structures and chemistry of a state-of-the-art MTJ film and offer direct evidence that B diffuses out of the crystalline CoFeB into the Ta interstitial sites rather than into the MgO tunnel barriers via annealing, and CoFe can bond to MgO gains with an epitaxial orientation relationship by forming Co(Fe)-O bonds yet with no incorporation of Co/Fe in MgO. Such capability of direct spatial and chemical analysis of the buried polycrystalline stacking layers in MTJs should open up new avenues to engineering material stacks at the atomic scale and substantially deepen our understanding of the physics of MTJs, thereby allowing us to accelerate the development of MRAM and integrated circuits with the MTJs.

at a vacant anion site (BO) in the 2− oxidation state. However, such defect formation energy is more than 4 eV, making B incorporation unlikely at typical annealing temperatures, which is in accordance with the experimental results. The B incorporation in MgO turns thermodynamically favorable only for very high or very low Fermi energies, indicating that it might be possible for B to electromigrate into MgO under a high field either as a 3+ interstitial ion for a high positive bias or a 3− substitutional anion for a high negative bias. Nevertheless, in the O-rich case B incorporation into vacant cation sites of MgO (BMg) in either the 3+ or the 1+ oxidation state becomes thermodynamically favorable for typical CoFe band offsets (Figure 5b). We notice that the 2+ state is absent for BMg, indicating that this is an example of a negative-U defect. As in the O-poor case, a high positive bias might drive the electromigration of B3+ interstitials yet in contrast a large negative bias would not result in electromigration owing to the absence of available anion vacancy sites. Our MTJ thin films deposited in high vacuum are underoxidized rather than overoxidized, corresponding more closely to the O-poor situation. In such case, the B is predicted not to be incorporated in MgO unless a high electrical bias is applied, consistent with the EELS mapping revealing the absence of B distribution in the MgO tunnel barrier (Figure 2). Supposing that MgO is overoxidized, B would readily incorporate at available cation sites thermally even without high electrical bias that hinders device performance because B substituting for Mg ion in MgO introduces electron trapping states at a similar position as oxygen vacancies. On the other hand, B incorporation into interstitial site (Bi) of either the body-centered cubic (bcc) (a = 3.306 Å) or face-centered cubic (fcc) (a = 4.207 Å) Ta exhibits a very low Ef for typical band offsets, irrespective of the O-poor or O-rich situation, implying that B prefers to reside at the Ta rather than MgO layer, which is in remarkable agreement with the EELS mapping (Figure 2). Over the entire range of EF, B is more favorable in fcc Ta interstitial sites than in bcc Ta interstitial sites, albeit that aberration-corrected TEM and diffractogram analyses (Supporting Information Figure S6) show that Ta nanocrystallites take a bcc structure with a lattice constant of 3.306 Å (Supporting Information). To probe the condition at which Co/Fe can incorporate into MgO, we also calculate the formation energies of Co and Fe at interstitial and cation sites of MgO in different charge states, as shown in Figure 5c,d. Interestingly, Co/Fe can incorporate as 3+ or 2+ ion in MgO cation site, yet such incorporation is only preferred for O-rich condition, implying that Co/Fe does not readily diffuse into the neighboring MgO tunnel barrier in our MTJ (close to O-poor situation), consistent with the EELS data showing the absence of Co/Fe in MgO (Figure 2). These results also suggest that although normally one tries to fully oxidize the insulating MgO tunnel barrier as oxygen vacancies can introduce gap states that open up preferential tunneling pathways and hinder device performance, in the presence of CoFeB overoxidization of MgO might give rise to the incorporation of B or Co/Fe in place of Mg, which may be equally detrimental to device performance. The role of the oxygen chemical potential in determining the thermodynamic tendency for B to incorporate into MgO may help explain some of the conflicting experimental results in the litereature.16−23 Despite its central importance, information on specific atomic-scale structures and local chemistry in the technological relevant CoFeB−MgO MTJs is still controversial due to their

■

ASSOCIATED CONTENT

* Supporting Information S

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.nanolett.5b03627. Characterizations: diffraction patterns, TEM and HRTEM images, annular bright-field imaging, and distance and line profile. Theoretical calculations: atomic model and density of states. (PDF)

■

AUTHOR INFORMATION

Corresponding Authors

*E-mail: [email protected] (K.P.M). *E-mail: [email protected] (H.O.). Tel: +81-22-2175931. Fax: +81-22-217-5930. *E-mail: [email protected] (Y.I.). Tel: +81-22-2175931. Fax: +81-22-217-5930. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This work was supported in part by the JSPS Grant-in-Aid for Scientific Research on Innovative Areas “Nano Informatics” (Grant 26106503) and A (15H02290), the “Nanotechnology Platform” (project no. 12024046) at the University of Tokyo from the Ministry of Education, Culture, Sports, Science, and Technology (MEXT) of Japan, and R & D Project for ICT Key Technology to Realize Future Society of MEXT, ImPACT Program of CSTI, and JSPS Core to Core Program. M.S. thanks support from Scientific Research (C) (Grant 26420662). K.P.M. thanks support from EPSRC (Grant EP/ K003151). This work made use of the facilities of Archer, the U.K.’s national high-performance computing service, via our membership in U.K. HPC Materials Chemistry Consortium, which is funded by the EPSRC (EP/L000202). A part of the calculations were conducted at Institute for Solid State Physics, University of Tokyo. Data relating to the density functional theory calculations performed during this research are available by request from the University of York Research database http://dx.doi.org/10.15124/b4750d52-dfc5-4188-8c9679d7359446ff.

■

REFERENCES

(1) Miyazaki, T.; Tezuka, N. J. Magn. Magn. Mater. 1995, 139, L231.

F

DOI: 10.1021/acs.nanolett.5b03627 Nano Lett. XXXX, XXX, XXX−XXX

Letter

Nano Letters (2) Moodera, J. S.; Kinder, L. R.; Wong, T. M.; Meservey, R. Phys. Rev. Lett. 1995, 74, 3273. (3) Yuasa, S.; Nagahama, T.; Fukushima, A.; Suzuki, Y.; Ando, K. Nat. Mater. 2004, 3, 868. (4) Parkin, S. S.; Kaiser, C.; Panchula, A.; Rice, P. M.; Hughes, B.; Samant, M.; Yang, S. H. Nat. Mater. 2004, 3, 862. (5) Ikeda, S.; Hyakawa, J.; Ashizawa, Y.; Lee, Y. M.; Miura, K.; Hasegawa, H.; Tsunoda, M.; Matsukura, F.; Ohno, H. Appl. Phys. Lett. 2008, 93, 082508. (6) Djayaprawira, D. D.; Tsunekawa, K.; Nagai, M.; Maehara, H.; Yamagata, S.; Watanabe, N.; Yuasa, S.; Suzuki, Y.; Ando, K. Appl. Phys. Lett. 2005, 86, 092502. (7) Hayakawa, J.; Ikeda, S.; Matsukura, F.; Takahashi, H.; Ohno, H. Jpn. J. Appl. Phys. 2005, 44, L587. (8) Mangin, S.; Ravelosona, D.; Katine, J. A.; Carey, M. J.; Terris, B. D.; Fullerton, E. E. Nat. Mater. 2006, 5, 210. (9) Endo, M.; Kanai, S.; Ikeda, S.; Matsukura, F.; Ohno, H. Appl. Phys. Lett. 2010, 96, 212503. (10) Ikeda, S.; Miura, K.; Yamamoto, H.; Mizunuma, K.; Gan, H. D.; Endo, M.; Kanai, S.; Hayakawa, J.; Matsukura, F.; Ohno, H. Nat. Mater. 2010, 9, 721. (11) Worledge, D. C.; Hu, G.; Abraham, D. W.; Sun, J. Z.; Trouilloud, P. L.; Nowak, J.; Brown, S.; Gaidis, M. C.; Osullivan, E. J.; Robertazzi, R. P. Appl. Phys. Lett. 2011, 98, 022501. (12) Sato, H.; Yamanouchi, M.; Ikeda, S.; Fukami, S.; Matsukura, F.; Ohno, H. Appl. Phys. Lett. 2012, 101, 022414. (13) Manchon, A.; Ducruet, C.; Lombard, L.; Auffret, S.; Rodmacq, B.; Dieny, B.; Pizzini, S.; Vogel, J.; Uhlir, V.; Hochstrasser, M.; Panaccione, G. J. Appl. Phys. 2008, 104, 043914. (14) Shimabukuro, R.; Nakamura, K.; Akiyama, T.; Ito, T. Phys. E 2010, 42, 1014. (15) Zhang, X. G.; Butler, W. H. Phys. Rev. B: Condens. Matter Mater. Phys. 2004, 70, 172407. (16) Burton, J. D.; Jaswal, S. S.; Tsymbal, E. Y.; Mryasov, O. N.; Heinonen, O. G. Appl. Phys. Lett. 2006, 89, 142507. (17) Bai, Z. Q.; Shen, L.; Wu, Q. Y.; Zeng, M. G.; Wang, J. S.; Han, G. C.; Feng, Y. P. Phys. Rev. B: Condens. Matter Mater. Phys. 2013, 87, 014114. (18) Karthik, S. V.; Takahashi, Y. K.; Ohkubo, T.; Hono, K.; Ikeda, S.; Ohno, H. J. Appl. Phys. 2009, 106, 023920. (19) Bae, J. Y.; Lim, W. C.; Kim, H. J.; Lee, T. D.; Kim, K. W.; Kim, T. W. J. Appl. Phys. 2006, 99, 08T316. (20) Takeuchi, T.; Tsunekawa, K.; Choi, Y. S.; Nagamine, Y.; Djayaprawira, D. D.; Genseki, A.; Hoshi, Y.; Kitamoto, Y. Jpn. J. Appl. Phys. 2007, 46, L623. (21) Komagaki, K.; Hattori, M.; Noma, K.; Kanai, H.; Kobayashi, K.; Uehara, Y.; Tsunoda, M.; Takahashi, M. IEEE Trans. Magn. 2009, 45, 3453. (22) Kozina, X.; Ouardi, S.; Balke, B.; Stryganyuk, G.; Fecher, G. H.; Felser, C.; Ikeda, S.; Ohno, H.; Ikenaga, E. Appl. Phys. Lett. 2010, 96, 072105. (23) Lu, Y.; Lépine, B.; Jézéquel, G.; Ababou, S.; Alnot, M.; Lambert, J.; Renard, A.; Mullet, M.; Deranlot, C.; Jaffrès, H.; Petroff, F.; George, J.-M. J. Appl. Phys. 2010, 108, 043703. (24) Yang, H. X.; Chshiev, M.; Dieny, B.; Lee, J. H.; Manchon, A.; Shin, K. H. Phys. Rev. B: Condens. Matter Mater. Phys. 2011, 84, 054401. (25) Muller, D. A. Nat. Mater. 2009, 8, 263. (26) Buban, J. P.; Matsunaga, K.; Chen, J.; Shibata, N.; Ching, W. Y.; Yamamoto, T.; Ikuhara, Y. Science 2006, 311, 212. (27) Nellist, P. D.; Chisholm, M.F.; Dellby, N.; Krivanek, O. L.; Murfitt, M. F.; Szilagyi, Z. S.; Lupini, A. R.; Borisevich, A.; Sides, W. H.; Pennycook, S. J. Science 2004, 305, 1741. (28) Ortalan, V.; Uzun, A.; Gates, B. C.; Browning, N. D. Nat. Nanotechnol. 2010, 5, 843. (29) Wang, Z. C.; Saito, M.; McKenna, K. P.; Gu, L.; Tsukimoto, S.; Shluger, A. L.; Ikuhara, Y. Nature 2011, 479, 380. (30) Findlay, S.; Shibata, N.; Sawada, H.; Okunishi, E.; Kondo, Y.; Yamamoto, T.; Ikuhara, Y. Appl. Phys. Lett. 2009, 95, 191913.

(31) Colliex, C. Nature 2007, 450, 622. (32) Kresse, G.; Furthmuller, J. Phys. Rev. B: Condens. Matter Mater. Phys. 1996, 54, 11169. (33) Lany, S.; Zunger, A. Modell. Simul. Mater. Sci. Eng. 2009, 17, 084002. (34) Yuasa, S.; Djayaprawira, D. J. Phys. D: Appl. Phys. 2007, 40, R337. (35) Pennycook, S. J.; Boatner, L. A. Nature 1988, 336, 565. (36) Ikeda, S.; Koizumi, R.; Sato, H.; Yamanouchi, M. K.; Mizunuma, K.; Gan, H. D.; Matsukura, F.; Ohno, H. IEEE Trans. Magn. 2012, 48, 3829. (37) Díaz-Ortiz, A.; Drautz, R.; Fähnle, M.; Dosch, H.; Sanchez, J. M. Phys. Rev. B: Condens. Matter Mater. Phys. 2006, 73, 224208. (38) Janotti, A.; Varley, J. B.; Rinke, P.; Umezawa, N.; Kresse, G.; Van de Walle, C. G. Phys. Rev. B: Condens. Matter Mater. Phys. 2010, 81, 085212. (39) Tanaka, I.; Mizoguchi, T.; Matsui, M.; Yoshioka, S.; Adachi, H.; Yamamoto, T.; Okajima, T.; Umesaki, M.; Ching, W. Y.; Inoue, Y.; Mizuno, M.; Araki, H.; Shirai, Y. Nat. Mater. 2003, 2, 541.

G

DOI: 10.1021/acs.nanolett.5b03627 Nano Lett. XXXX, XXX, XXX−XXX