La 0.7 Sr 0.3 MnO 3 - based Spintronics

Transcription

1 La 0.7 Sr 0.3 MnO 3 - based Spintronics Investigation on fundamental issues and applications Umberto Scotti di Uccio DiMSAT, Università di Cassino CNR Coherentia Napoli MR (%) µ 0 H (mt) 1

2 Research group & Institutions where Monte Cassino Abbey Founded in 529 DC Destroyed in 1944, feb 15 th CNR - Coherenthia labs. University of Cassino 2

3 Research group & Institutions who R. Vaglio (Coherentia Labs leader) Di.M.S.A.T. - University of Cassino U. Scotti di Uccio P. Perna CNR Coherentia Labs. F. Miletto N. Lampis M. Radovic N. Russo A. Sambri Cooperations: devices G. Pepe A. Ruotolo M. Salluzzo R. Di Capua STM 3

4 Research group Scientific interests and background Structural and transport characterization XRD RSM R(T) STM Physics of the film surface Oxide films growth I (a. u.) O1s LSMO GXRD LEED BE (ev) PES HT c Josephson Devices fabrication and characterization The core activity of the group is the investigation of oxide films, mainly perovskites. At present we can resort to different techniques to characterize the structural and transport properties of samples, including x ray and electron diffraction, scanning tunneling microscopy, and photoemission spectroscopy. As far as devices are concerned, our experience mainly come from the fabrication and characterization of high Tc Josephson junctions, 4

5 Research group Scientific interests and background Structural and transport characterization XRD RSM R(T) STM Physics of the film surface Oxide films growth I (a. u.) O1s LSMO GXRD LEED BE (ev) PES 0.6 MR (%) µ 0 H (mt) Magnetoelectronics Devices fabrication and characterization but since a couple of years we started to work on magnetoelectronics devices based on LSMO, that are the topic of this talk. 5

6 Why LSMO? La 1-x Sr x MnO 3 is a perovskitic manganite Mn O La, Sr La 1-x Sr x MnO 3 is a ferromagnet Optimal doping x = Normal FM Half Metal La 1-x Sr x MnO 3 has a large spin polarization at the Fermi level N P = N N + N 1 So, why LSMO. LSMO is a perovskitic material. The content x in Strontium acts as a doping that controls the number of carriers, actually holes, at tha Fermi level. At the optimal doping that we considered in this work LSMO is a robust ferromagnet with Curie temperature well above room temperature. Most importantly LSMO is an almost perfect half metal, that is, the conduction band is mostly filled up with one orientation of spin, opening the door to application to spin injection. 6

7 La 0.7 Sr 0.3 MnO 3 - based Spintronics Investigation on fundamental issues and applications Outline LSMO films Overview on spintronics Preliminary experiments on LSMO Device fabrication and characterization Conclusions and future plans 7

8 La 0.7 Sr 0.3 MnO 3 - based Spintronics Investigation on fundamental issues and applications Outline LSMO films Overview on spintronics Preliminary experiments on LSMO Device fabrication and characterization Conclusions and future plans 8

9 LSMO films deposition conditions Laser ablation AFM STM Eccimer laser KrF - λ = 248 nm Target substrate 41 mm P(O 2 ) = 0.1 mbar LEED XPS-UPS PLD Effective Fluency: 80 mj /2.6 mm 2 Repetition rate: 2 Hz Deposition temperature: 850 C This work: STO substrate Coherentia Labs. Napoli UHV base P < mbar Modular system for Oxide Deposition and Analyses This is a picture of our lab. It shows the MODA system, that is a Modular system for Oxide Deposition and Analysis. It mainly consists of a chamber devoted to pulsed laser deposition, that is connected to several analysis chambers, that is, X ray photoemission Spectroscopy, Low Energy Electron Diffraction, Scanning Tunneling Microscopy. The deposition condition of our films are quite standard. In this work we only employed Srontiun Titanate substrates. 9

10 Films structure and morphology log(i/i o ) 10 1 STO (002) LSMO (002) θ Counts (a.u.) 1.0 (002) LSMO 0.5 (002) STO FWHM = ω (deg) Epitaxial, high structural quality Smooth surfaces Cube-on-cube (001) STO substrate epitaxy 3 µm These are some data for LSMO grown on (001) STO. The films are smooth, and show a high crystal quality. They also are fully strained at least up to 100 nm. 10

11 Films structure and morphology Epitaxial, high structural quality Smooth surfaces (110) STO substrate Cube-on-cube epitaxy In device fabrication we mostly employed films grown on (110) STO. The LSMO grows in the usual cube on cube fashion also in this case, and also in this case it shows a very high crystal quality and smooth surfaces 11

12 Films structure and morphology High Curie Temperature Nice transport properties (110) STO substrate M/M(0) T C 350 K I FM M PM T (K) T C R (Ω)..and nice transport and magnetic properties, with Curie Temperature well above room temperature. By the way, the comparison between magnetization and resistivity that I m showing here demonstrate the well known fact that in the ferromagnetic phase LSMO is a good metal, and in the paramagnetic phase it is a bad conductor. 12

13 Magnetic properties M(emu) 5x T = 100 K (001) (1-10) easy _ (110) -5x10-4 in plane µ o H (T) (110) STO substrate (001) hard In this viewgraph I show the magnetic properties of LSMO with (110) orientation. As known, the LSMO has a strong magnetostriction, so that the easy axis in films depend on the strain. In the case of STO we have tensile strain, and the easy axis lies parallel to the substrate. The standard characterization of the magnetic hysteresis cycles was performed in a vibrating sample magnetometer. We see that there is a strong anisothropy between the two in-plane directions, and that the easy axis is aligned to the (001) direction. 13

14 La 0.7 Sr 0.3 MnO 3 - based Spintronics Investigation on fundamental issues and applications Outline LSMO films Overview on spintronics Preliminary experiments on LSMO Device fabrication and characterization Conclusions and future plans So it is time to introduce the research on devices with a brief overview on spintronics devices 14

15 The GMR effect Hard disk read heads Normal FM Half Metal Pictures from Gary A. Prinz, SCIENCE 282, 1660 (1998) Review data Spin-polarized transport occurs in ferromagnetic half metals. The orientation of the majority spins determines the magnetization of the sample and it is controlled by the external field. When two half metals are put in contact through a spacer, there is a state with low resistance when the magnetic moments are parallel, and a state with high resistance when the magnetic moments are antiparallel. This is the magnetoresistance effect that is exploited in the read heads of our hard disks. 15

16 TMR devices 300 K R TMR = M. Coey, et al., nature materials, 9 (2005) ( 0) R( H) R( H) Stuart S. P. Parkin, et al., nature materials 862, (2004) Review data In the TMR devices the spacer between the two magnetized layers is a tunnel barrier. The quality factor for the device is the TMR ratio, defined as the percent variation of resistivity in presence of an external magnetic field. Tunnel devices demonstrate a very high TMR ratio and are serious candidates for many commercial devices. At present the best results are based on MgO barriers separating Cobalth-Iron alloys electrodes. 16

17 LSMO based TMR devices Actual materials are not ideal half metals, but LSMO is attractive because of its high spin polarization leading to high expected MR N P = N N + N M. Bowen, et al., APL 82, 233 (2003) 4.2 K R R P1 P 2 1 P P 1 2 M. Julliere, Phys. Lett. 54A, 225 (1975) F. Pailloux, PRB 66, (2002) Review data The actual materials are not ideal half metals but they instead have some population of minirity spins at the Fermi level, that results in a reduction of the TMR. In this context LSMO is attractive because of its high degree of spin polarization leading to high magnetoresistance. This was observed for instance in TMR junctions with STO barrier. 17

18 LSMO based TMR devices Actual materials are not ideal half metals, but LSMO is attractive because of its high spin polarization leading to high expected MR N P = N N + N M. Bowen, et al., APL 82, 233 (2003) R R P1 P 2 1 P P 1 2 M. Julliere, Phys. Lett. 54A, 225 (1975) The problem is that the TMR drops with temperature Usual interpretation: degraded properties at LSMO interface Review data The problem with LSMO however is that the TMR drops fastly with temperature and it is small at room temperature. This effect is generally interpreted in terms of some degradation of the magnetic properties at the surface of LSMO, or in the case of trilayers at the interface with barrier. 18

19 La 0.7 Sr 0.3 MnO 3 - based Spintronics Investigation on fundamental issues and applications Outline LSMO films Overview on spintronics Our experience on: Preliminary experiments on LSMO Degradation of surface Intrinsic properties Device fabrication and characterization Conclusions and future plans Now I would like to show what is our experience on the properties of LSMO in connection to the decrease of magnetoresistance. I will try to demonstrate that this effect may be connected both to degradation of the surface and to more intrinsic properties of LSMO. 19

20 The dead layer M x t (emu m -2 ) (110) LSMO MODA lab t 5 nm t (nm)? top bottom First of all, I would like to comment on the so called dead layer. The existence of the dead layer was demonstrated both for LSMO and for common ferromagnetic metals and alloys by comparing the magnetic moment or also the electrical conductance of samples with different thickness. For instance, in our measurements on (110) LSMO grown on STO we find that the extrapolated magnetization of a layer as thin as about 5 nm is zero, so that we infer a non magnetic layer of about 5 nm in the samples. Reported values are similar and are slightly dependent on the substrate and on the growth technique. However bulk measurements don t distinguish between sample surface and interface with the substrate. 20

21 The metallic state at the surface 1 1 LEED on as-deposited samples demonstrate metallicity at the surface Not probing for magnetic properties: LSMO is not a real insulator in the PM state La 0.7 Sr 0.3 MnO 3 on (001) STO So let s consider surface sensitive analyses. In this picture I m showing the electron diffraction pattern taken on an as-deposited LSMO sample. Since the electron diffraction is only observed on conducting surfaces, the measurements demonstrate that the LSMO surface of as-deposited samples is conducting. This is important, but it is not probing, because LSMO is not a real insulator even when in the paramagnetic state. 21

22 Degradation of the surface properties of LSMO: evidence by XPS Experimental technique: XPS at different emerging angles x ray source e - detector x ray source e - detector Shallow angle emission Normal emission PS sampling depth 1-2 nm Shallow angle: still lower l l cosθ We can get some more information by photoemission. The photoemission has a sampling depth of a few nanometers, because the mean free path of photoelectrons is short. Photoemission at shallow angle has a still lower sampling depth. Thus, in order to get information on the very last layer I will compare between measurements taken at different emission angles. 22

23 Degradation of the surface properties of LSMO: evidence by XPS Mn 3d e g t 2g CB O 2p pσ pπ VB La 0.7 Sr 0.3 MnO 3 : 2/3 filling e g First let me remind that in LSMO the conduction band is the lower Mn3d eg band, that is partially filled, while the valence band is made by overlapping Mn t2g and O 2p states. 23

24 Degradation of the surface properties of LSMO: evidence by XPS Counts (a.u.) Mn e g normal emission shallow emission Mn t 2g O 2p difference At the surface: Higher BE of e g Low conducting state E F BE (ev) difference Spectral weight transfer: There is a reduction in the DOS close to the Fermi edge 0.0 bulk CB VB E F surface These are the results. We see that the density of states at the LSMO at the surface is different. We interpret the weight transfer as a sign of bad metallicity because the e g states are pushed far from the Fermi level. (In this case there is also an enhanced occupation, that is possibly due to the presence of oxygen vacancies, because oxygen vacancies are electron donors.). 24

25 The inhomogeneous phase transition La 0.7 Sr 0.3 MnO 3 on STO(110) R (Ω) LT HT dρ/dt (a. u.) ρ (a. u.) T (K) T (K) T M The behaviour of ρ(t) has been described in terms of an inhomogeneous transition with phase separation LT : conducting FM phase HT : insulating PM phase HT HT LT HT ρ = f ρ LT coexistence + ( 1 f ) ρht R changes smoothly The magnetic properties are not well established and not uniform Transport properties in manganite thin films Phys. Rev. B 71, (2005) Let me present a different point of view. This is again the ρ(t) plot of a sample deposited on (110) STO, and on the right I show a magnification of the region of the metal-insulator transition. The red curve is the derivative dr/dt and T M is Curie temperature. We see that there is a region of about 100 K where the resistivity changes dramatically. This is the crossover between the conducting ferromagnetic phase at low temperature, to the insulating paramagnetic phase at high temperature. In a recent paper, we demonstrated that this crossover can be described in terms of a phase separation as shown in this formula where f is the fraction of conducting phase. So we have a large region where the magnetic properties are not well established and not uniform. 25

26 The inhomogeneous phase transition R (Ω) LT dρ/dt (a. u.) ρ (a. u.) T (K) La 0.7 Sr 0.3 MnO 3 on STO(110) La 0.7 Sr 0.3 MnO 3 on STO(110) STM in conductance map mode K T (K) V o = 2 V, feedback disconnected nm 2 J. Phys.: Condens. Matter (2006) This mechanism influences surface properties. Here I m showing a conductance map taken by low temperature STM, that is highly sensitive to surface. In this mode of operation, the false colors indicate the conductance of the junction between tip and sample, that in a first approximation is the density of states of the sample. We see that the map is homogeneous in spite of the details of sample morphology. 26

27 The inhomogeneous phase transition R (Ω) LT T (K) La 0.7 Sr 0.3 MnO 3 on STO(110) La 0.7 Sr 0.3 MnO 3 on STO(110) STM in conductance map mode dρ/dt (a. u.) T (K) coexistence ρ (a. u.) 77 K 300 K Dark regions : low conductivity Light regions : high conductivity nm nm 2 J. Phys.: Condens. Matter (2006) At room temperature the situation changes. The phase separation separation mechanism is at play 27

28 The inhomogeneous phase transition 1000 coexistence R (Ω) LT T (K) La 0.7 Sr 0.3 MnO 3 on STO(110) La 0.7 Sr 0.3 MnO 3 on STO(110) STM in conductance map mode Dark regions : low conductivity Light regions : high conductivity 77 K nm nm K and we see the coexistence of insulating and conducting regions that are spatially separated. Now, this is an intrinsic effect and there is no reason to suspect that it should not be at play not only at surfaces but also at interfaces, reducing the performance of TMR junctions. Concluding this section, I would like to comment that non intrinsic mechanisms such as oxygen loss or contamination can be perhaps controlled by a suitable technology, but the phase separation mechanism seems a more fundamental problem with LSMO. 28

29 The inhomogeneous phase transition Concluding this section Non intrinsic mechanisms such as oxygen loss or contamination can be perhaps controlled by a suitable technology PS is an intrinsic, thermodynamical effect PS can be at play also at interfaces (but how?) PS can reduce TMR performances because PM regions are present 77 K 300 K nm nm 2 and we see the coexistence of insulating and conducting regions that are spatially separated. Now, this is an intrinsic effect and there is no reason to suspect that it should not be at play not only at surfaces but also at interfaces, reducing the performance of TMR junctions. Concluding this section, I would like to comment that non intrinsic mechanisms such as oxygen loss or contamination can be perhaps controlled by a suitable technology, but the phase separation mechanism seems a more fundamental problem with LSMO. 29

30 La 0.7 Sr 0.3 MnO 3 - based Spintronics Investigation on fundamental issues and applications Outline LSMO films Overview on spintronics Preliminary experiments on LSMO Device fabrication and characterization Conclusions and future plans Now, let us come to our activity on devices 30

31 La 0.7 Sr 0.3 MnO 3 - based Spintronics Investigation on fundamental issues and applications What about interfaces between LSMO and conventional FM? Can the degraded interface act as a tunnel barrier? Will it lead to tunneling magnetoresistance? Surface layer substrate Metal LSMO O migration? A. Plecenick, et al., Appl. Phys. Lett. 81, 859 (2002) Our first work was devoted to this issue: can a degraded interface between LSMO and a magnetic alloy act as a tunnel barrier? Will it lead to tunneling magnetoresistance? There were some indication in literature that a metallic contact on LCMO can show an insulating interlayer due to oxygen migration from LCMO, but little about the magnetic properties. 31

32 La 0.7 Sr 0.3 MnO 3 - based Spintronics Investigation on fundamental issues and applications What about interfaces between LSMO and metals? Ni 80 Fe 20 deposition DC Sputtering O.5 Pa room temperature First experiment: stencil mask Surface layer Permalloy (Ni Fe 0.20 ) LSMO (110) STO In a first experiment we realized a simple bilayer LSMO/Permalloy 32

33 Magnetic properties of LSMO Py bilayers 5x10-7 La 0.7 Sr 0.3 MnO 3 /Py on STO(110) 30 nm LSMO + 10 nm PY 30 nm LSMO 2x10-7 8x10-8 difference m (A m 2 ) 0 m (A m 2 ) 0-8x x µ o H (mt) -2x µ o H (mt) LSMO+Py H C = K M S = 620 emu/cm 100 K LSMO H c = K M S = 515 emu/cm 100 K Py H c = K M S = 830 emu/cm 3 No evidence of exchange coupling Relatively large saturation of Py The hysteresis loops indicate that the two layers are magnetically decoupled, and that the permalloy has a relatively high saturation field. 33

34 Low temperature MR effect CPP I + V + V - I V - + V I + I - 30 nm LSMO 10 nm Py on STO(110) Patterned La 0.7 Sr 0.3 MnO 3 on STO(110) 4.2 K H r r 4.2 K R/R Hc (%) µ o H (mt) R/R Hc (%) times smaller µ 0 H(mT) U. Scotti di Uccio, et al.et al., APL 88, (2006) Here I show the MR of the bilayer at 4.2 K. The MR of the bilayer is quite small. However it is ten times higher than a single LSMO film, and it cannot either be ascribed to Py... 34

35 Low temperature MR effect I + V + V - I - 30 nm LSMO 10 nm Py on STO(110) 4.2 K 5x10-7 CPP H r r J V + V - I + I - Patterned La 0.7 Sr 0.3 MnO 3 on STO(110) 4.2 K R/R Hc (%) x µ o H (mt) H C LSMO M (A m 2 ) R/R Hc (%) times smaller µ 0 H(mT) U. Scotti di Uccio, et al.et al., APL 88, (2006) because the peaks in the MR plot correspond to the coercive field of LSMO. So there is some indication that the device works based on some spin scattering mechanism, but it is still unclear the role of the interface, that on this basis could well be something as a metallic spacer. Moreover, the MR of the device doesn t show a flat baseline, probably due to the MR of LSMO itself. 35

36 LSMO Py TMR devices 2 % MR 1 Ion milling etching µm µ o H (mt) µm 2 Ohmic behavior R 580 Ω tunnel barrier Reduced role of film resistance U. Scotti di Uccio, et al.et al., APL 88, (2006) Flat baseline no MR from LSMO Broad peaks high H S of Py To get more information we patterned devices with a small contact area. Here I show the MR of a 70 x 100 micron square junction. The I vs V plot is linear, with a resistance that is much higher than the resistance of the electrodes. This is an indication of a tunneling mechanism. Moreover we have the suppression of the MR due to the LSMO film resulting in a flat baseline. The peaks are broadened due to the not well established antiparallel state between LSMO and Py, due to the large saturation field of Py. 36

37 LSMO Py TMR devices 2 % MR µ o H (mt) Simulation based on the Julliere model: Measured M(H) on a similar bilayer R P1 P2 Computation of R 1 P P 1 2 % MR 2 1 simulation ~10% spin polarization µ o H (mt) A computation, that is not actually a fit, based on the Jullier model reproduces quite well the general behavior and indicates a very small spin polarization at the interface. Of coarse the device is much worse than all-lsmo TMR devices. Nevertheless, the interesting thing is that the degraded layer at the surface of LSMO can act as a tunnel barrier. 37

38 LSMO nanoconstrictions A completely different concept Devices based on the DW resistance nanoconstrictions The DW are trapped at the costrictions LSMO Top view substrate Picture adapted from A. M. Haghiri, Half metallic devices for spintronics DW thickness Costriction width P. Bruno Phys. Rev. Lett. 83, 2425 (1999) Thinner DW higher MR The principle of operation is the following. The domain walls between regions with different orientation of magnetization scatter the electrons. So we can make a device if we can control the formation of domain walls. This is achieved by realizing nanocostrictions, because it was demonstrated that nanocostrictions pin the domain walls, and that when located at nanocostrictions the walls become thinner and scatter electrons more efficiently. 38

39 LSMO nanoconstrictions APPL. PHYS. LETT. 87, C. Rüster, et al., PRL 91, (2003) Nanoconstrictions in (Ga,Mn)As J. Appl. Phys., Vol. 89, (2001) arxiv:cond-mat/ v1 12 Oct 2006 Review data This is not a new concept, and it was already exploited in several studies, among which I remind those made at the CRISMAT by prof. Raveau team and recently by prof. Mercey team. 39

40 LSMO nanoconstrictions A completely different concept Devices based on the DW resistance M M MR Low MR Picture adapted from A. M. Haghiri, Half metallic devices for spintronics 0 Hc 1 Hc 2 H So this is a cartoon of device operation. At the beginning the resistance il low 40

41 LSMO nanoconstrictions A completely different concept Devices based on the DW resistance M M MR High MR The central part has higher coercivity Picture adapted from A. M. Haghiri, Half metallic devices for spintronics 0 Hc 1 Hc 2 H When the field is increased, the side arms flip earlier, because the central region has higher coercivity for geometrical reasons. 41

42 LSMO nanoconstrictions A completely different concept Devices based on the DW resistance M M MR Low MR Picture adapted from A. M. Haghiri, Half metallic devices for spintronics 0 Hc 1 Hc 2 H and only at higher field the magnetic moments are again parallel. 42

43 Fabrication process a) Standard lithography + ion milling 100 nm b) FIB with Ga + ions 10 pa, 30 KeV c) FIB with Ga + ions 1 pa, 30 KeV d) SEM photograph Easy axis (001) LSMO 43

44 Fabrication process a) Standard lithography + ion milling 100 nm b) FIB with Ga + ions 10 pa, 30 KeV c) FIB with Ga + ions 1 pa, 30 KeV d) SEM photograph Easy axis (001) LSMO nm constrictions 44

45 30 nm constrictions I (µa) V (V) Strong non-linearity in a wide range of T Fowler-Nordheim eq Legend: V V ' = 2 ψ = barrier height w = barrier thickness These are the IV curves for a 30 nm constriction at different temperatures. The behavior is well fitted by the Fowler Nordheim equation, that is a model for electron tunneling. 45

46 30 nm constrictions I (µa) V (V) The MR decreases with V (as in TMR) 4.2 K Moreover, the MR decreases with voltage, that is typical of TMR devices. 46

47 30 nm constrictions Why tunneling? Not to scale a) FIB damage Lateral straggling 10 nm b) Artificial AF DW AF Sharp Twist of magnetization Stabilization of AF state in LSMO So the question is, why are the constrictions behaving like tunnel devices. One possibility is that the lateral straggling of FIB damages the whole nanocostriction area. The second, more intriguing effect that may play a role in the antiparallel state is that at the constriction the twist of magnetization is so sharp that it results in a thin antiferromagnetic insulating region. This is an open issue. 47

48 50 nm constrictions 0.6 I bias = 2µA The depinning is not simultaneous MR (%) The hysteresis loop is not symmetric µ 0 H (mt) MR (%) 0.5 H - H + IV characteristics almost linear 0.0 No evidenec of tunneling µ 0 H (mt) The behavior of the wider constrictions is different because the IV characteristics are linear and the MR is lower. The plot of magnetoresistance shows that for this device the depinning of the domain walls at the two nanocostrictions is not simultaneous. Moreover the hysteresis loop is not symmetric. This means that the direction of the current has influence on the pinning and depinning mechanisms 48

49 50 nm constrictions MR (%) 0.5 H - H + The depinning is not simultaneous Spin transfer torque is at play? r τ = µ 0 H (mt) dm dm M b M + c J dx dx H f (mt) bias current (µa) as also the current intensity does. This is not a classical behavior, and it is a plausible manifestation of a spin torque acting on the domain walls because of spin injection. 49

50 50 nm constrictions The current is able to switch the state of the device without the application of an external magnetic field. switch induced by the current di/dv (1/MΩ) H = I (ma) current density (H = 0) J = A/m Finally, in these constriction we can have a switch completely determined by the current, at zero field. 50

51 Conclusions & perspectives 1. Both intrinsic and non-intrinsic mechanisms can lead to reduction of MR in devices based on LSMO 2. A damaged interface between LSMO and Py acts as a tunnel barrier 3. The devices based on domain wall resistance at nanoconstrictions are promising. Physical nature of the constrained region? Different materials? Next step: electron beam lithography 51