Home
Search
Collections
Journals
About
Contact us
My IOPscience
Epitaxy of MgO magnetic tunnel barriers on epitaxial graphene
This content has been downloaded from IOPscience. Please scroll down to see the full text. 2013 Nanotechnology 24 475708 (http://iopscience.iop.org/0957-4484/24/47/475708) View the table of contents for this issue, or go to the journal homepage for more
Download details: IP Address: 152.14.136.96 This content was downloaded on 07/11/2013 at 09:53
Please note that terms and conditions apply.
IOP PUBLISHING
NANOTECHNOLOGY
Nanotechnology 24 (2013) 475708 (6pp)
doi:10.1088/0957-4484/24/47/475708
Epitaxy of MgO magnetic tunnel barriers on epitaxial graphene Florian Godel1 , Emmanuelle Pichonat2 , Dominique Vignaud2 , Hicham Majjad1 , Dominik Metten1 , Yves Henry1 , St´ephane Berciaud1 , Jean-Francois Dayen1 and David Halley1 1
Institut de Physique et Chimie des Mat´eriaux de Strasbourg (IPCMS), 23 rue du Loess, BP 43, F-67034, Strasbourg Cedex 2, France 2 Institut d’Electronique, de Micro´electronique et de Nanotechnologie (IEMN), Avenue Poincar´e, BP 60069, F-59652 Villeneuve D’Ascq, Cedex, France E-mail: [email protected]
Received 18 July 2013, in final form 26 September 2013 Published 5 November 2013 Online at stacks.iop.org/Nano/24/475708 Abstract Epitaxial growth of electrodes and tunnel barriers on graphene is one of the main technological bottlenecks for graphene spintronics. In this paper, we demonstrate that MgO(111) epitaxial tunnel barriers, one of the prime candidates for spintronic application, can be grown by molecular beam epitaxy on epitaxial graphene on SiC(0001). Ferromagnetic metals (Fe, Co, Fe20 Ni80 ) were epitaxially grown on top of the MgO barrier, thus leading to monocrystalline electrodes on graphene. Structural and magnetic characterizations were performed on these ferromagnetic metals after annealing and dewetting: they form clusters with a 100 nm typical lateral width, which are mostly magnetic monodomains in the case of Fe. This epitaxial stack opens the way to graphene spintronic devices taking benefits from a coherent tunnelling current through the epitaxial MgO/graphene stack. S Online supplementary data available from stacks.iop.org/Nano/24/475708/mmedia (Some figures may appear in colour only in the online journal)
1. Introduction
much recent progress in spintronics, by increasing the spin polarization of electrons: coherent tunnelling in epitaxial magnetic tunnel junctions such as Fe/MgO/Fe(001) [10] or (La, Sr)MnO3 /SrTiO3 /(La, Sr)MnO3 [11] turned out to select the tunnelling electrons as a function of the symmetry of their Bloch wavefunction. For instance, the tunnelling transport in epitaxial Fe/MgO/Fe(001) barriers is dominated by electrons with the 11 symmetry. Those electrons are theoretically 100% spin polarized in Fe, at the Fermi level, when their wavevector is along [001], i.e. perpendicular to the barrier. This high spin polarization led to huge tunnelling magneto-resistance values up to 180% at room temperature [12] and even 600% in the case of FeCoB electrodes [13]. In the case of graphene, theoretical works predicted [14, 15] a large spin polarization for epitaxial graphene on fcc Ni(111) or hcp Co(0001). Indeed, the coherent injection of electrons from the metal into the graphene should select
Graphene appears as an appealing material for spintronic applications. Recent studies indeed proved a very long spin diffusion length in graphene layers, either exfoliated [1, 2], grown by chemical vapour deposition [3] or on SiC substrates [4]. These values, up to 100 µm, would enable new device architectures and logics based for instance on pure spin currents [5, 6]. Different works [4, 7] demonstrated that spin injection from a ferromagnetic material into graphene must be performed through a tunnel barrier to solve the conductance mismatch issue [8] and subsequent spin depolarization. Satisfying results were obtained on graphene devices with MgO [1] or Al2 O3 polycrystalline barriers [9]: they exhibited giant magneto-resistive signals up to 10% in the so-called local geometry [4], with graphene channel length in the micrometre range. Moreover, the fabrication of monocrystalline inorganic tunnel barriers has resulted in 0957-4484/13/475708+06$33.00
1
c 2013 IOP Publishing Ltd Printed in the UK & the USA
Nanotechnology 24 (2013) 475708
F Godel et al
Figure 1. (a) Scheme of the deposited stack with the ferromagnetic (FM) clusters. (b) RHEED diagram of the MgO(111) surface grown on ¯ direction of the MgO graphene, (c) of the 10 nm (110) Fe layer before annealing, and (d) after annealing. The zone axis is along the [112] layer. The electron energy is 20 keV. (e) RHEED diagram of the permalloy(Py) layer after annealing.
electrons having their in-plane wavevector component close to the high-symmetry K point of graphene in the reciprocal space—i.e. energy which is close to the Fermi level in neutral graphene. Those electrons on Ni(111) or Co(0001) surfaces are only minority electrons—defined relative to the orientation of the Ni or Co magnetization—which implies a large spin polarization. This spin filtering effect was investigated in recent experiments on the graphene/fcc Ni(111) interfaces [16, 17], or in Co/Al2 O3 /graphene/Ni magnetic tunnel junction [18] with amorphous alumina barriers. We demonstrate in this paper how epitaxial MgO(111) tunnel barriers can be grown directly on graphene on SiC, and show strategies to superpose a Co(0001) or FeNi(111) epitaxial ferromagnetic electrode over it. These results open the way to coherent tunnelling and enhanced spin injection in graphene, and offer new opportunities for graphene spintronic device engineering.
deposition of the tunnel barrier by MBE, thanks to an e-beam heated cell. Following Wang et al [1], we deposited a 0.12 nm thick Ti layer on graphene at room temperature, in order to promote the sticking of MgO and reduce the roughness of the tunnel barrier. This small amount of metal—probably forming small clusters on the graphene surface—is expected to oxidize at the interface with MgO [1] as was also demonstrated for instance in the case of Cr or V layers below a MgO barrier [21]. Then, a 3 nm thick MgO layer was grown at 100 ◦ C at a 0.8 nm min−1 deposition rate (see the heterostructure scheme in figure 1(a)). The pressure increases to a few 10−8 mbar during deposition, due to the oxygen partial pressure. The reflexion high energy electron diffraction (RHEED) pattern (figure 1(b)) shows a slightly rough surface, with a clear signature of epitaxial MgO: the cubic MgO—rock salt structure—grows on a (111) plane, thus keeping the three-fold symmetry of the underlying substrate. The crystallographic orientation of MgO is confirmed by x-ray diffraction analyses (figure 2(b)): the observation of MgO(111) and (222) peaks confirms the epitaxy of MgO along the (111) growth direction. No other peaks related to MgO are observed. Due to the thickness of the MgO layer (3 nm) the peaks are quite broad, with a full width at half maximum equal to 0.42 ± 0.05◦ for the (111) peak and 0.98 ± 0.06◦ for the (222) peak. The out-of-plane lattice parameter given by the peak position can be assessed at 0.426 ± 0.001 nm. This value is 1.1% larger than in the case of MgO bulk (0.420 nm). This out-of-plane strain involves an in-plane compressive strain in MgO. The Poisson ratio for MgO is ν = 0.187 [22] which gives an in-plane deformation of 0.032 ± 0.005 nm—from the equation 1x = 1y = 1z/ν, where x and y are in-plane directions—corresponding to 7.6 ± 2.5% compressive strain in MgO.
2. Growth of an epitaxial MgO tunnel barrier on graphene Growth was achieved on n-doped SiC:6H polytype monocrystal substrates, on the (0001) Si-face. After degassing and surface flattening [19] graphene was grown by annealing at ∼1280 ◦ C for 6 min in an ultra-high vacuum chamber. This led to a graphene thickness of ∼3 monolayers, as evaluated by x-ray photoemission spectroscopy measurements [20] (for AFM and Raman data, see supplementary information available at stacks.iop.org/Nano/24/475708/mmedia). The samples were then transferred into the molecular beam epitaxy (MBE) chamber with a 10−10 mbar base pressure. They were first degassed in vacuum at 800 ◦ C before 2
Nanotechnology 24 (2013) 475708
F Godel et al
˚ Figure 2. 2θ x-ray diffraction spectra—λ = 1.5406 A—recorded on samples after deposition of the ferromagnetic layer: θ –2θ cans performed on annealed samples with Co, Fe or Py (a). The spectra are vertically offset for clarity. (b) X-ray diffraction spectra zoom on MgO(111) and MgO(222) peaks. The asterisk indicates a weak reflexion due to the sample holder.
The characteristics of the MgO growth on graphene obtained in this study are similar to those reported for direct growth at 140 ◦ C on bare 6H-SiC(0001) substrates [23], which yielded good quality insulating barriers. In our case, the graphene layer on top of the SiC substrate does not seem to affect the growth conditions of the MgO layer. Notice that the MgO(111) planes are chemically pure O or Mg planes and should match well with the hexagonal graphene lattice. Indeed the distance between nearest neighbours, i.e. 0.298 nm in MgO(111) planes, is close to twice the interatomic distance in graphene (2 × 0.142 = 0.284 nm): the associated misfit would thus be close to 4.7% in compression for MgO. This hypothesis of a matching between graphene and MgO(111) hexagonal lattices is coherent with the x-ray measurements where a 7.6 ± 2.5% compression is observed in MgO.
spin injectors into graphene and deserve more systematic study. In order to improve the crystalline quality of the Fe electrode, we annealed this layer in situ, at 500 ◦ C, for 1 h. Annealing leads to longer and thinner rods on the RHEED diagram, together with well-defined Bragg spots (figure 1(d)). This diagram suggests the presence of both flat and rough parts on the Fe surface. This can be understood thanks to AFM (atomic force microscopy) images showing the dewetting of Fe on MgO, which occurred during annealing. It yielded Fe mesa in the shape of ribbons with a 200 nm characteristic width and 35 nm characteristic height—this height can reach 60 nm on some clusters. The tops of these mesa are smooth (see figure 3(a) inset), explaining the rods observed by RHEED, whereas the reflexion on the cluster edges leads to the 3D features on the RHEED diagram. The MgO layer in between the clusters remains as before, depositing Fe, with a lower than 0.7 nm r.m.s roughness. We also checked that the graphene was not modified by the 500 ◦ C annealing under the MgO layer. The Raman spectrum (figure 3(d)), recorded through the Fe/MgO top layer, is indeed consistent with that of a few-layer graphene sample grown on the Si-face of 6H-SiC(0001) SiC. It shows clear G and 2D-mode features that are blueshifted relative to the reference spectra measured on freestanding graphene layers [26]. Such shifts can be attributed to compressive strain, as well as to changes in the electronic and phononic dispersion relations for graphene layers on a SiC substrate [27, 28]. Importantly, the intensity of the defect-related D-mode feature remains weak relative to that of the G-mode (ID /IG < 10), suggesting that the lattice structure of graphene is well preserved after deposition of MgO and Fe. Magnetic force microscopy (MFM) observations were performed on the Fe clusters on MgO. The in-plane magnetization image shows that most of these clusters are magnetic monodomains (figure 4(a)): there is a bright and dark contrast at ribbon edges, whereas the remaining clusters appear as uniform. This rules out the presence of intermediate
3. Growth and characterization of epitaxial magnetic clusters 3.1. Fe clusters A 10 nm thick Fe electrode was then grown at 100 ◦ C on the MgO barrier. The RHEED diagram (figure 1(c)) shows an epitaxial growth of bcc (110) Fe, but the spotty pattern reveals a large roughness and mosaicity. The direction of growth is confirmed by x-ray diffraction θ –2θ scans (inset figure 2(a)) which exhibit an intense (110) peak of Fe. The epitaxial growth of Fe on MgO(111) has already been observed by Hauch et al [24]. They showed that the epitaxial relationship is as follows: Fe(110)[111] k MgO(111)[212] with three possible variants. This growth of Fe(110) on MgO(111) is appealing as band structure calculations predicted a large value of the spin polarization of electrons on the Fe(110) surface [24]. Indeed, an 80% spin polarization was measured near the Fermi energy on this surface using spin and angle-resolved photoelectron spectroscopy [25]. These studies show that epitaxial Fe(110) electrodes are potential 3
Nanotechnology 24 (2013) 475708
F Godel et al
Figure 3. AFM images of the ferromagnetic electrodes after annealing and Au capping. (a) Fe, (b) permalloy and (c) Co. The smaller clusters are attributed to Au capping deposited at room temperature after annealing. The average mesa height is close to 30–35 nm as shown on the profiles in the insets. (d) Micro-Raman spectrum of graphene obtained on the sample with Fe, after annealing. The spectrum was measured using a laser photon phonon energy of 2.33 eV and a laser power of less than 1 mW. The G and 2D peak positions and widths are: ωG = 1593 cm−1 , 0G = 35 cm−1 , ω2D = 2735 cm−1 , 02D = 60 cm−1 .
The presence of Co(0002) or Py(111) peaks associated to hcp Co or fcc Py respectively, without any other related peak, confirms the growth directions. We again performed in situ annealing at 500 ◦ C to improve the crystalline quality of these layers: RHEED diagrams evolved after annealing (figure 1(e)) show thinner rods, and AFM images (figures 3(b) and (c)) again exhibit dewetting of the metallic layers on top of the MgO barrier. We also observe a flat surface on top of the clusters, which are clearly facetted in the case of permalloy. Their three-fold symmetry comes from their epitaxial growth on MgO(111). Some Co clusters also appear as facetted, though less strikingly than in the case of permalloy. Notice that the permalloy clusters, with long connected ribbons forming ‘Y’ shapes, differ from the convex shape of Co clusters. Additional structural data on cluster orientation are provided in supplementary materials (available at stacks.iop.org/Nano/ 24/475708/mmedia). We also checked by AFM, in the case of Co, that for thick deposited layers—60 nm—the magnetic film is continuous. From a magnetic point of view, MFM images obtained on annealed permalloy and Co clusters (figures 4(b) and (c))—in the case of a 10 nm deposited nominal thickness—show magnetic domain walls on individual clusters, which are thus magnetic multidomains. Notice that, despite the reduced
magnetic domain walls in those clusters and proves the existence of magnetic monodomains. They could be exploited to achieve nanometric epitaxial magnetic electrodes for spin current injection/detection into graphene. Note that the electronic band structure of the metal is not modified by confinement effects in clusters of this size—larger than 100 nm. The formation of clusters will thus not be detrimental to the expected high spin polarization from the clusters into graphene. 3.2. Cobalt and permalloy clusters We extended our study to other ferromagnetic electrodes by growing Co or Fe20 Ni80 (permalloy) layers with a nominal 10 nm thickness on similar MgO(111)/graphene/SiC samples at 100 ◦ C. As in the case of Fe, RHEED diagrams (see supplementary materials available at stacks. iop.org/Nano/24/475708/mmedia) prove an epitaxial growth with nevertheless a low structural quality. Consistent with the (111) growth observed by Ohtake et al [29], on MgO(111), a permalloy (111)[110] k MgO(111)[110] epitaxial relationship is observed for our fcc permalloy. Co on the other hand grows in an hcp structure with its [0001] axis along MgO[111]. The inset in figure 2(a) shows the x-ray diffraction θ –2θ scan obtained on samples with Co or Py. 4
Nanotechnology 24 (2013) 475708
F Godel et al
Figure 4. MFM images of the ferromagnetic electrodes after annealing. (a) Fe, (b) Py and (c) Co. (d) Magneto-optical Kerr hysteresis measurement of the in-plane magnetization on the permalloy annealed layer.
way to new studies within the field of spintronics in graphene, taking advantage of the expected high spin polarization on Ni(111) or Co(0001) surfaces and of the coherent electrons tunnelling through the MgO(111) epitaxial barrier. Indeed, the MgO(111) planes have the three-fold symmetry of the underlying graphene and of the Co or permalloy electrodes: this should enable the electrons to tunnel while keeping the symmetry of their Bloch function, and thus their high spin polarization. The ability to grow directly on graphene such a well-defined interface is also important to make clearer the spin injection process through graphene, especially to discriminate the different contributions of elastic and non-elastic channels [31]. Finally, as MgO(111) films are polar [32], they might offer new opportunities to tune the density and polarity of the charge carrier in graphene devices, together with influencing the spin-polarized density of electronic states at the interface with the magnetic electrode.
dimensions of the clusters and their shape and size dispersion, the hysteresis cycle obtained on the whole sample (figure 4(d)) by in-plane Kerr magnetometry shows, in the case of permalloy, an abrupt reversing, with a coercive field of about 100 Oe. This reveals a similar magnetic behaviour of the assembly of clusters, which have a small dispersion of their magnetic coercive fields. The difference in magnetic configurations between permalloy and Co on the one hand and Fe on the other hand can tentatively be attributed to two origins. First, the aspect ratio of the Fe long thin ribbons favours, due to the demagnetizing field, a magnetization along the ribbons, which disfavours the formation of magnetic domain walls. Indeed the length/width ratio is on average close to 5 in Fe clusters, but no more than 2 in Co, for similar cluster thickness. The case of permalloy is more complex, as the clusters are often connected with one another. Second, the magneto-crystalline anisotropy in permalloy or in the hexagonal plane of Co is weaker than along the [110] direction of Fe [30]. This easy magnetization axis (Ku # 3 × 105 erg cm−3 ) is in the plane of the film due to the epitaxial relationship, and could also disfavour the formation of magnetic domain walls.
Acknowledgments We thank S Boukari for Kerr measurements and B. Doudin for fruitful discussions. Financial support from the CNRS (Nano 2012: G3N) is gratefully acknowledged.
4. Conclusion References
In conclusion, we showed the epitaxial growth of MgO(111) barriers on top of graphene on SiC substrates. Moreover, good quality ferromagnetic electrodes were grown on this oxide layer, with well-defined epitaxial relationship. This paves the
[1] Wang W H, Han W, Pi K, McCreary K M, Miao F, Bao F, Lau C N and Kawakami R K 2008 Appl. Phys. Lett. 93 183107 5
Nanotechnology 24 (2013) 475708
F Godel et al
[17] van ’t Erve O M J, Friedman A L, Cobas E, Li C H, Robinson J T and Jonker B T 2012 Nature Nanotechnol. 7 737–42 [18] Dlubak B et al 2012 ACS Nano 6 10930–4 [19] Ferrer F J, Moreau E, Vignaud D, Godey S and Wallart X 2009 Semicond. Sci. Technol. 24 125014 [20] Moreau E, Godey S, Ferrer F J, Vignaud D, Wallart X, Avila J, Asensio M C, Bournel F and Gallet J J 2010 Appl. Phys. Lett. 97 241907 [21] Halley D, Majjad H, Bowen M, Najjari N, Henry Y, Ulhaq-Bouillet C, Weber W, Bertoni G, Verbeeck J and Van Tendeloo G 2008 Appl. Phys. Lett. 92 212115 [22] Madelung O, R¨ossler U and Schulz M (ed) 1999 Magnesium oxide (MgO) Young’s, shear and bulk moduli, Poisson’s ratio II–VI and I–VII Compounds; Semimagnetic Compounds vol 41B (Berlin: Springer) pp 1–3 [23] Posadas A, Walker F J, Ahn C H, Goodrich T L, Cai Z and Ziemer K S 2008 Appl. Phys. Lett. 92 233511 [24] Hauch J O, Fonin M, Fraune M, Turban P, Guerrero R, Aliev F G, Mayer J, R¨udiger U and G¨untherodt G 2008 Appl. Phys. Lett. 93 083512 [25] Turner A M and Erskine J L 1982 Phys. Rev. B 25 3 [26] Berciaud S, Li X, Htoon H, Brus L E, Doorn S K and Heinz T F 2013 Nano Lett. 13 3517–23 [27] Ni Z H, Chen W, Fan X F, Kuo J L, Yu T, Wee A T S and Shen Z X 2008 Phys. Rev. B 77 115416 [28] Lee D S, Riedl C, Krauss B, von Klitzing K, Starke U and Smet J H 2008 Nano Lett. 8 4320–5 [29] Ohtake M, Tanaka T, Matsubara K, Kirino F and Futamoto M 2011 J. Phys.: Conf. Ser. 303 012015 [30] Yu J, R¨udiger U, Kent A D, Thomas L and Parkin S S P 1999 Phys. Rev. B 60 7352 [31] Wehling T O, Grigorenko I, Lichstein A I and Balatsky A V 2008 Phys. Rev. Lett. 101 216803 [32] Goniakowski J and Nogura C 2002 Phys. Rev. B 66 085417
[2] Tombros N, Jozsa C, Popinciuc M, Jonkman H T and van Wees B J 2007 Nature 448 571–4 [3] Avsar A et al 2011 Nano Lett. 11 2363–8 [4] Dlubak B et al 2012 Nature Phys. 8 557–61 [5] Wu D H, Ciftcioglu B, Huang M, Song Y, Kawakami R K, ˇ Shi J, Krivorotov I, Zutic I and Sham L J 2012 IEEE Trans. Electron Devices 59 259 [6] Zeng M, Shen L, Su H, Zhang C and Feng Y 2011 Appl. Phys. Lett. 98 092110 [7] Han W, McCreary K M, Pi K, Wang W H, Li Y, Wen H, Chen J R and Kawakami R K 2012 J. Magn. Magn. Mater. 324 369–81 [8] Schmidt G, Ferrand D and Molenkamp L 2000 Phys. Rev. B 62 R4790 [9] Dlubak B, Martin M B, Deranlot C, Bouzehouane K, Fusil S, Mattana R, Petroff F, Anane A, Seneor P and Fert A 2012 Appl. Phys. Lett. 101 203104 [10] Butler W, Zhang X G, Schulthess T and MacLaren J 2011 Phys. Rev. B 63 1–12 [11] Ishii Y, Yamada H, Sato H, Akoh H, Ogawa Y, Kawasaki M and Tokura Y 2006 Appl. Phys. Lett. 89 042509 [12] Yuasa S, Nagahama T, Fukushima A, Suzuki Y and Ando K 2004 Nature Mater. 3 868–71 [13] Ikeda S, Hayakawa J, Ashizawa Y, Lee Y M, Miura K, Hasegawa H, Tsunoda M, Matsukura F and Ohno H 2008 Appl. Phys. Lett. 93 082508 [14] Karpan V M, Giovannetti G, Khomyakov P, Talanana M, Starikov A, Zwierzycki M, van den Brink J, Brocks G and Kelly P J 2007 Phys. Rev. Lett. 99 176602 [15] Karpan V M, Khomyakov P A, Starikov A A, Giovannetti G, Zwierzycki M, Talanana M, Brocks G, van den Brink J and Kelly P J 2008 Phys. Rev. B 78 195419 [16] Dedkov Yu S, Fonin M and Laubschat C 2008 Appl. Phys. Lett. 92 052506
6
Search
Collections
Journals
About
Contact us
My IOPscience
Epitaxy of MgO magnetic tunnel barriers on epitaxial graphene
This content has been downloaded from IOPscience. Please scroll down to see the full text. 2013 Nanotechnology 24 475708 (http://iopscience.iop.org/0957-4484/24/47/475708) View the table of contents for this issue, or go to the journal homepage for more
Download details: IP Address: 152.14.136.96 This content was downloaded on 07/11/2013 at 09:53
Please note that terms and conditions apply.
IOP PUBLISHING
NANOTECHNOLOGY
Nanotechnology 24 (2013) 475708 (6pp)
doi:10.1088/0957-4484/24/47/475708
Epitaxy of MgO magnetic tunnel barriers on epitaxial graphene Florian Godel1 , Emmanuelle Pichonat2 , Dominique Vignaud2 , Hicham Majjad1 , Dominik Metten1 , Yves Henry1 , St´ephane Berciaud1 , Jean-Francois Dayen1 and David Halley1 1
Institut de Physique et Chimie des Mat´eriaux de Strasbourg (IPCMS), 23 rue du Loess, BP 43, F-67034, Strasbourg Cedex 2, France 2 Institut d’Electronique, de Micro´electronique et de Nanotechnologie (IEMN), Avenue Poincar´e, BP 60069, F-59652 Villeneuve D’Ascq, Cedex, France E-mail: [email protected]
Received 18 July 2013, in final form 26 September 2013 Published 5 November 2013 Online at stacks.iop.org/Nano/24/475708 Abstract Epitaxial growth of electrodes and tunnel barriers on graphene is one of the main technological bottlenecks for graphene spintronics. In this paper, we demonstrate that MgO(111) epitaxial tunnel barriers, one of the prime candidates for spintronic application, can be grown by molecular beam epitaxy on epitaxial graphene on SiC(0001). Ferromagnetic metals (Fe, Co, Fe20 Ni80 ) were epitaxially grown on top of the MgO barrier, thus leading to monocrystalline electrodes on graphene. Structural and magnetic characterizations were performed on these ferromagnetic metals after annealing and dewetting: they form clusters with a 100 nm typical lateral width, which are mostly magnetic monodomains in the case of Fe. This epitaxial stack opens the way to graphene spintronic devices taking benefits from a coherent tunnelling current through the epitaxial MgO/graphene stack. S Online supplementary data available from stacks.iop.org/Nano/24/475708/mmedia (Some figures may appear in colour only in the online journal)
1. Introduction
much recent progress in spintronics, by increasing the spin polarization of electrons: coherent tunnelling in epitaxial magnetic tunnel junctions such as Fe/MgO/Fe(001) [10] or (La, Sr)MnO3 /SrTiO3 /(La, Sr)MnO3 [11] turned out to select the tunnelling electrons as a function of the symmetry of their Bloch wavefunction. For instance, the tunnelling transport in epitaxial Fe/MgO/Fe(001) barriers is dominated by electrons with the 11 symmetry. Those electrons are theoretically 100% spin polarized in Fe, at the Fermi level, when their wavevector is along [001], i.e. perpendicular to the barrier. This high spin polarization led to huge tunnelling magneto-resistance values up to 180% at room temperature [12] and even 600% in the case of FeCoB electrodes [13]. In the case of graphene, theoretical works predicted [14, 15] a large spin polarization for epitaxial graphene on fcc Ni(111) or hcp Co(0001). Indeed, the coherent injection of electrons from the metal into the graphene should select
Graphene appears as an appealing material for spintronic applications. Recent studies indeed proved a very long spin diffusion length in graphene layers, either exfoliated [1, 2], grown by chemical vapour deposition [3] or on SiC substrates [4]. These values, up to 100 µm, would enable new device architectures and logics based for instance on pure spin currents [5, 6]. Different works [4, 7] demonstrated that spin injection from a ferromagnetic material into graphene must be performed through a tunnel barrier to solve the conductance mismatch issue [8] and subsequent spin depolarization. Satisfying results were obtained on graphene devices with MgO [1] or Al2 O3 polycrystalline barriers [9]: they exhibited giant magneto-resistive signals up to 10% in the so-called local geometry [4], with graphene channel length in the micrometre range. Moreover, the fabrication of monocrystalline inorganic tunnel barriers has resulted in 0957-4484/13/475708+06$33.00
1
c 2013 IOP Publishing Ltd Printed in the UK & the USA
Nanotechnology 24 (2013) 475708
F Godel et al
Figure 1. (a) Scheme of the deposited stack with the ferromagnetic (FM) clusters. (b) RHEED diagram of the MgO(111) surface grown on ¯ direction of the MgO graphene, (c) of the 10 nm (110) Fe layer before annealing, and (d) after annealing. The zone axis is along the [112] layer. The electron energy is 20 keV. (e) RHEED diagram of the permalloy(Py) layer after annealing.
electrons having their in-plane wavevector component close to the high-symmetry K point of graphene in the reciprocal space—i.e. energy which is close to the Fermi level in neutral graphene. Those electrons on Ni(111) or Co(0001) surfaces are only minority electrons—defined relative to the orientation of the Ni or Co magnetization—which implies a large spin polarization. This spin filtering effect was investigated in recent experiments on the graphene/fcc Ni(111) interfaces [16, 17], or in Co/Al2 O3 /graphene/Ni magnetic tunnel junction [18] with amorphous alumina barriers. We demonstrate in this paper how epitaxial MgO(111) tunnel barriers can be grown directly on graphene on SiC, and show strategies to superpose a Co(0001) or FeNi(111) epitaxial ferromagnetic electrode over it. These results open the way to coherent tunnelling and enhanced spin injection in graphene, and offer new opportunities for graphene spintronic device engineering.
deposition of the tunnel barrier by MBE, thanks to an e-beam heated cell. Following Wang et al [1], we deposited a 0.12 nm thick Ti layer on graphene at room temperature, in order to promote the sticking of MgO and reduce the roughness of the tunnel barrier. This small amount of metal—probably forming small clusters on the graphene surface—is expected to oxidize at the interface with MgO [1] as was also demonstrated for instance in the case of Cr or V layers below a MgO barrier [21]. Then, a 3 nm thick MgO layer was grown at 100 ◦ C at a 0.8 nm min−1 deposition rate (see the heterostructure scheme in figure 1(a)). The pressure increases to a few 10−8 mbar during deposition, due to the oxygen partial pressure. The reflexion high energy electron diffraction (RHEED) pattern (figure 1(b)) shows a slightly rough surface, with a clear signature of epitaxial MgO: the cubic MgO—rock salt structure—grows on a (111) plane, thus keeping the three-fold symmetry of the underlying substrate. The crystallographic orientation of MgO is confirmed by x-ray diffraction analyses (figure 2(b)): the observation of MgO(111) and (222) peaks confirms the epitaxy of MgO along the (111) growth direction. No other peaks related to MgO are observed. Due to the thickness of the MgO layer (3 nm) the peaks are quite broad, with a full width at half maximum equal to 0.42 ± 0.05◦ for the (111) peak and 0.98 ± 0.06◦ for the (222) peak. The out-of-plane lattice parameter given by the peak position can be assessed at 0.426 ± 0.001 nm. This value is 1.1% larger than in the case of MgO bulk (0.420 nm). This out-of-plane strain involves an in-plane compressive strain in MgO. The Poisson ratio for MgO is ν = 0.187 [22] which gives an in-plane deformation of 0.032 ± 0.005 nm—from the equation 1x = 1y = 1z/ν, where x and y are in-plane directions—corresponding to 7.6 ± 2.5% compressive strain in MgO.
2. Growth of an epitaxial MgO tunnel barrier on graphene Growth was achieved on n-doped SiC:6H polytype monocrystal substrates, on the (0001) Si-face. After degassing and surface flattening [19] graphene was grown by annealing at ∼1280 ◦ C for 6 min in an ultra-high vacuum chamber. This led to a graphene thickness of ∼3 monolayers, as evaluated by x-ray photoemission spectroscopy measurements [20] (for AFM and Raman data, see supplementary information available at stacks.iop.org/Nano/24/475708/mmedia). The samples were then transferred into the molecular beam epitaxy (MBE) chamber with a 10−10 mbar base pressure. They were first degassed in vacuum at 800 ◦ C before 2
Nanotechnology 24 (2013) 475708
F Godel et al
˚ Figure 2. 2θ x-ray diffraction spectra—λ = 1.5406 A—recorded on samples after deposition of the ferromagnetic layer: θ –2θ cans performed on annealed samples with Co, Fe or Py (a). The spectra are vertically offset for clarity. (b) X-ray diffraction spectra zoom on MgO(111) and MgO(222) peaks. The asterisk indicates a weak reflexion due to the sample holder.
The characteristics of the MgO growth on graphene obtained in this study are similar to those reported for direct growth at 140 ◦ C on bare 6H-SiC(0001) substrates [23], which yielded good quality insulating barriers. In our case, the graphene layer on top of the SiC substrate does not seem to affect the growth conditions of the MgO layer. Notice that the MgO(111) planes are chemically pure O or Mg planes and should match well with the hexagonal graphene lattice. Indeed the distance between nearest neighbours, i.e. 0.298 nm in MgO(111) planes, is close to twice the interatomic distance in graphene (2 × 0.142 = 0.284 nm): the associated misfit would thus be close to 4.7% in compression for MgO. This hypothesis of a matching between graphene and MgO(111) hexagonal lattices is coherent with the x-ray measurements where a 7.6 ± 2.5% compression is observed in MgO.
spin injectors into graphene and deserve more systematic study. In order to improve the crystalline quality of the Fe electrode, we annealed this layer in situ, at 500 ◦ C, for 1 h. Annealing leads to longer and thinner rods on the RHEED diagram, together with well-defined Bragg spots (figure 1(d)). This diagram suggests the presence of both flat and rough parts on the Fe surface. This can be understood thanks to AFM (atomic force microscopy) images showing the dewetting of Fe on MgO, which occurred during annealing. It yielded Fe mesa in the shape of ribbons with a 200 nm characteristic width and 35 nm characteristic height—this height can reach 60 nm on some clusters. The tops of these mesa are smooth (see figure 3(a) inset), explaining the rods observed by RHEED, whereas the reflexion on the cluster edges leads to the 3D features on the RHEED diagram. The MgO layer in between the clusters remains as before, depositing Fe, with a lower than 0.7 nm r.m.s roughness. We also checked that the graphene was not modified by the 500 ◦ C annealing under the MgO layer. The Raman spectrum (figure 3(d)), recorded through the Fe/MgO top layer, is indeed consistent with that of a few-layer graphene sample grown on the Si-face of 6H-SiC(0001) SiC. It shows clear G and 2D-mode features that are blueshifted relative to the reference spectra measured on freestanding graphene layers [26]. Such shifts can be attributed to compressive strain, as well as to changes in the electronic and phononic dispersion relations for graphene layers on a SiC substrate [27, 28]. Importantly, the intensity of the defect-related D-mode feature remains weak relative to that of the G-mode (ID /IG < 10), suggesting that the lattice structure of graphene is well preserved after deposition of MgO and Fe. Magnetic force microscopy (MFM) observations were performed on the Fe clusters on MgO. The in-plane magnetization image shows that most of these clusters are magnetic monodomains (figure 4(a)): there is a bright and dark contrast at ribbon edges, whereas the remaining clusters appear as uniform. This rules out the presence of intermediate
3. Growth and characterization of epitaxial magnetic clusters 3.1. Fe clusters A 10 nm thick Fe electrode was then grown at 100 ◦ C on the MgO barrier. The RHEED diagram (figure 1(c)) shows an epitaxial growth of bcc (110) Fe, but the spotty pattern reveals a large roughness and mosaicity. The direction of growth is confirmed by x-ray diffraction θ –2θ scans (inset figure 2(a)) which exhibit an intense (110) peak of Fe. The epitaxial growth of Fe on MgO(111) has already been observed by Hauch et al [24]. They showed that the epitaxial relationship is as follows: Fe(110)[111] k MgO(111)[212] with three possible variants. This growth of Fe(110) on MgO(111) is appealing as band structure calculations predicted a large value of the spin polarization of electrons on the Fe(110) surface [24]. Indeed, an 80% spin polarization was measured near the Fermi energy on this surface using spin and angle-resolved photoelectron spectroscopy [25]. These studies show that epitaxial Fe(110) electrodes are potential 3
Nanotechnology 24 (2013) 475708
F Godel et al
Figure 3. AFM images of the ferromagnetic electrodes after annealing and Au capping. (a) Fe, (b) permalloy and (c) Co. The smaller clusters are attributed to Au capping deposited at room temperature after annealing. The average mesa height is close to 30–35 nm as shown on the profiles in the insets. (d) Micro-Raman spectrum of graphene obtained on the sample with Fe, after annealing. The spectrum was measured using a laser photon phonon energy of 2.33 eV and a laser power of less than 1 mW. The G and 2D peak positions and widths are: ωG = 1593 cm−1 , 0G = 35 cm−1 , ω2D = 2735 cm−1 , 02D = 60 cm−1 .
The presence of Co(0002) or Py(111) peaks associated to hcp Co or fcc Py respectively, without any other related peak, confirms the growth directions. We again performed in situ annealing at 500 ◦ C to improve the crystalline quality of these layers: RHEED diagrams evolved after annealing (figure 1(e)) show thinner rods, and AFM images (figures 3(b) and (c)) again exhibit dewetting of the metallic layers on top of the MgO barrier. We also observe a flat surface on top of the clusters, which are clearly facetted in the case of permalloy. Their three-fold symmetry comes from their epitaxial growth on MgO(111). Some Co clusters also appear as facetted, though less strikingly than in the case of permalloy. Notice that the permalloy clusters, with long connected ribbons forming ‘Y’ shapes, differ from the convex shape of Co clusters. Additional structural data on cluster orientation are provided in supplementary materials (available at stacks.iop.org/Nano/ 24/475708/mmedia). We also checked by AFM, in the case of Co, that for thick deposited layers—60 nm—the magnetic film is continuous. From a magnetic point of view, MFM images obtained on annealed permalloy and Co clusters (figures 4(b) and (c))—in the case of a 10 nm deposited nominal thickness—show magnetic domain walls on individual clusters, which are thus magnetic multidomains. Notice that, despite the reduced
magnetic domain walls in those clusters and proves the existence of magnetic monodomains. They could be exploited to achieve nanometric epitaxial magnetic electrodes for spin current injection/detection into graphene. Note that the electronic band structure of the metal is not modified by confinement effects in clusters of this size—larger than 100 nm. The formation of clusters will thus not be detrimental to the expected high spin polarization from the clusters into graphene. 3.2. Cobalt and permalloy clusters We extended our study to other ferromagnetic electrodes by growing Co or Fe20 Ni80 (permalloy) layers with a nominal 10 nm thickness on similar MgO(111)/graphene/SiC samples at 100 ◦ C. As in the case of Fe, RHEED diagrams (see supplementary materials available at stacks. iop.org/Nano/24/475708/mmedia) prove an epitaxial growth with nevertheless a low structural quality. Consistent with the (111) growth observed by Ohtake et al [29], on MgO(111), a permalloy (111)[110] k MgO(111)[110] epitaxial relationship is observed for our fcc permalloy. Co on the other hand grows in an hcp structure with its [0001] axis along MgO[111]. The inset in figure 2(a) shows the x-ray diffraction θ –2θ scan obtained on samples with Co or Py. 4
Nanotechnology 24 (2013) 475708
F Godel et al
Figure 4. MFM images of the ferromagnetic electrodes after annealing. (a) Fe, (b) Py and (c) Co. (d) Magneto-optical Kerr hysteresis measurement of the in-plane magnetization on the permalloy annealed layer.
way to new studies within the field of spintronics in graphene, taking advantage of the expected high spin polarization on Ni(111) or Co(0001) surfaces and of the coherent electrons tunnelling through the MgO(111) epitaxial barrier. Indeed, the MgO(111) planes have the three-fold symmetry of the underlying graphene and of the Co or permalloy electrodes: this should enable the electrons to tunnel while keeping the symmetry of their Bloch function, and thus their high spin polarization. The ability to grow directly on graphene such a well-defined interface is also important to make clearer the spin injection process through graphene, especially to discriminate the different contributions of elastic and non-elastic channels [31]. Finally, as MgO(111) films are polar [32], they might offer new opportunities to tune the density and polarity of the charge carrier in graphene devices, together with influencing the spin-polarized density of electronic states at the interface with the magnetic electrode.
dimensions of the clusters and their shape and size dispersion, the hysteresis cycle obtained on the whole sample (figure 4(d)) by in-plane Kerr magnetometry shows, in the case of permalloy, an abrupt reversing, with a coercive field of about 100 Oe. This reveals a similar magnetic behaviour of the assembly of clusters, which have a small dispersion of their magnetic coercive fields. The difference in magnetic configurations between permalloy and Co on the one hand and Fe on the other hand can tentatively be attributed to two origins. First, the aspect ratio of the Fe long thin ribbons favours, due to the demagnetizing field, a magnetization along the ribbons, which disfavours the formation of magnetic domain walls. Indeed the length/width ratio is on average close to 5 in Fe clusters, but no more than 2 in Co, for similar cluster thickness. The case of permalloy is more complex, as the clusters are often connected with one another. Second, the magneto-crystalline anisotropy in permalloy or in the hexagonal plane of Co is weaker than along the [110] direction of Fe [30]. This easy magnetization axis (Ku # 3 × 105 erg cm−3 ) is in the plane of the film due to the epitaxial relationship, and could also disfavour the formation of magnetic domain walls.
Acknowledgments We thank S Boukari for Kerr measurements and B. Doudin for fruitful discussions. Financial support from the CNRS (Nano 2012: G3N) is gratefully acknowledged.
4. Conclusion References
In conclusion, we showed the epitaxial growth of MgO(111) barriers on top of graphene on SiC substrates. Moreover, good quality ferromagnetic electrodes were grown on this oxide layer, with well-defined epitaxial relationship. This paves the
[1] Wang W H, Han W, Pi K, McCreary K M, Miao F, Bao F, Lau C N and Kawakami R K 2008 Appl. Phys. Lett. 93 183107 5
Nanotechnology 24 (2013) 475708
F Godel et al
[17] van ’t Erve O M J, Friedman A L, Cobas E, Li C H, Robinson J T and Jonker B T 2012 Nature Nanotechnol. 7 737–42 [18] Dlubak B et al 2012 ACS Nano 6 10930–4 [19] Ferrer F J, Moreau E, Vignaud D, Godey S and Wallart X 2009 Semicond. Sci. Technol. 24 125014 [20] Moreau E, Godey S, Ferrer F J, Vignaud D, Wallart X, Avila J, Asensio M C, Bournel F and Gallet J J 2010 Appl. Phys. Lett. 97 241907 [21] Halley D, Majjad H, Bowen M, Najjari N, Henry Y, Ulhaq-Bouillet C, Weber W, Bertoni G, Verbeeck J and Van Tendeloo G 2008 Appl. Phys. Lett. 92 212115 [22] Madelung O, R¨ossler U and Schulz M (ed) 1999 Magnesium oxide (MgO) Young’s, shear and bulk moduli, Poisson’s ratio II–VI and I–VII Compounds; Semimagnetic Compounds vol 41B (Berlin: Springer) pp 1–3 [23] Posadas A, Walker F J, Ahn C H, Goodrich T L, Cai Z and Ziemer K S 2008 Appl. Phys. Lett. 92 233511 [24] Hauch J O, Fonin M, Fraune M, Turban P, Guerrero R, Aliev F G, Mayer J, R¨udiger U and G¨untherodt G 2008 Appl. Phys. Lett. 93 083512 [25] Turner A M and Erskine J L 1982 Phys. Rev. B 25 3 [26] Berciaud S, Li X, Htoon H, Brus L E, Doorn S K and Heinz T F 2013 Nano Lett. 13 3517–23 [27] Ni Z H, Chen W, Fan X F, Kuo J L, Yu T, Wee A T S and Shen Z X 2008 Phys. Rev. B 77 115416 [28] Lee D S, Riedl C, Krauss B, von Klitzing K, Starke U and Smet J H 2008 Nano Lett. 8 4320–5 [29] Ohtake M, Tanaka T, Matsubara K, Kirino F and Futamoto M 2011 J. Phys.: Conf. Ser. 303 012015 [30] Yu J, R¨udiger U, Kent A D, Thomas L and Parkin S S P 1999 Phys. Rev. B 60 7352 [31] Wehling T O, Grigorenko I, Lichstein A I and Balatsky A V 2008 Phys. Rev. Lett. 101 216803 [32] Goniakowski J and Nogura C 2002 Phys. Rev. B 66 085417
[2] Tombros N, Jozsa C, Popinciuc M, Jonkman H T and van Wees B J 2007 Nature 448 571–4 [3] Avsar A et al 2011 Nano Lett. 11 2363–8 [4] Dlubak B et al 2012 Nature Phys. 8 557–61 [5] Wu D H, Ciftcioglu B, Huang M, Song Y, Kawakami R K, ˇ Shi J, Krivorotov I, Zutic I and Sham L J 2012 IEEE Trans. Electron Devices 59 259 [6] Zeng M, Shen L, Su H, Zhang C and Feng Y 2011 Appl. Phys. Lett. 98 092110 [7] Han W, McCreary K M, Pi K, Wang W H, Li Y, Wen H, Chen J R and Kawakami R K 2012 J. Magn. Magn. Mater. 324 369–81 [8] Schmidt G, Ferrand D and Molenkamp L 2000 Phys. Rev. B 62 R4790 [9] Dlubak B, Martin M B, Deranlot C, Bouzehouane K, Fusil S, Mattana R, Petroff F, Anane A, Seneor P and Fert A 2012 Appl. Phys. Lett. 101 203104 [10] Butler W, Zhang X G, Schulthess T and MacLaren J 2011 Phys. Rev. B 63 1–12 [11] Ishii Y, Yamada H, Sato H, Akoh H, Ogawa Y, Kawasaki M and Tokura Y 2006 Appl. Phys. Lett. 89 042509 [12] Yuasa S, Nagahama T, Fukushima A, Suzuki Y and Ando K 2004 Nature Mater. 3 868–71 [13] Ikeda S, Hayakawa J, Ashizawa Y, Lee Y M, Miura K, Hasegawa H, Tsunoda M, Matsukura F and Ohno H 2008 Appl. Phys. Lett. 93 082508 [14] Karpan V M, Giovannetti G, Khomyakov P, Talanana M, Starikov A, Zwierzycki M, van den Brink J, Brocks G and Kelly P J 2007 Phys. Rev. Lett. 99 176602 [15] Karpan V M, Khomyakov P A, Starikov A A, Giovannetti G, Zwierzycki M, Talanana M, Brocks G, van den Brink J and Kelly P J 2008 Phys. Rev. B 78 195419 [16] Dedkov Yu S, Fonin M and Laubschat C 2008 Appl. Phys. Lett. 92 052506
6
