Copyright WILEY-VCH Verlag GmbH & Co. KGaA, 69469 Weinheim, Germany, 2017.
Supporting Information for Adv. Funct. Mater., DOI: 10.1002/adfm.201700259
Modulating the Ferromagnet/Molecule Spin Hybridization Using an Artificial Magnetoelectric Micha# Studniarek, Salia Cherifi-Hertel, Etienne Urbain, Ufuk Halisdemir, Rémi Arras, Beata Taudul, Filip Schleicher, Marie Hervé, Charles-Henri Lambert, Abbass Hamadeh, Loïc Joly, Fabrice Scheurer, Guy Schmerber, Victor Da Costa, Bénédicte Warot-Fonrose, Cécile Marcelot, Olivia Mauguin, Ludovic Largeau, Florian Leduc, Fadi Choueikani, Edwige Otero, Wulf Wulfhekel, Jacek Arabski, Philippe Ohresser, Wolfgang Weber, Eric Beaurepaire, Samy Boukari, and Martin Bowen*
Supporting Information Modulating the ferromagnet/molecule spin hybridization using an artificial magnetoelectric
M. Studniarek1,2, S. Cherifi-Hertel1, E. Urbain1, U. Halisdemir1, R. Arras3, B. Taudul1, F. Schleicher1, M. Herve4, C.-H. Lambert5, A. Hamadeh5, L. Joly1, F. Scheurer1, G. Schmerber1, V. Da Costa1, B. Warot-Fonrose3, C. Marcelot3, O. Mauguin6, L. Largeau6, F. Leduc2, F. Choueikani2, E. Otero2, W. Wulfhekel4, J. Arabski1, P. Ohresser2, W. Weber1, E. Beaurepaire1, S. Boukari1, M. Bowen1* 1
Institut de Physique et Chimie des Matériaux de Strasbourg UMR 7504 CNRS, Université de Strasbourg, 23 Rue du Loess, BP 43, 67034 Strasbourg Cedex 2, France 2
Synchrotron SOLEIL, L’Orme des Merisiers, Saint-Aubin, BP 48, 91192 Gif-sur-Yvette, France
3
CEMES, Université de Toulouse, CNRS-UPR 8011, UPS, 29 rue Jeanne-Marvig, F-31055 Toulouse, France 4
Physikalisches Institut, Karlsruhe Institute of Technology, Wolfgang-Gaede-Str. 1, 76131 Karlsruhe, Germany 5
Institut Jean Lamour UMR 7198 CNRS, Université de Lorraine, BP 70239, 54506 Vandoeuvre les Nancy cedex, France 6
CNRS - C2N / Site de Marcoussis, Route de Nozay, 91460 Marcoussis, France
*Correspondence to:
[email protected]
Fabrication and processing of studied device To make our devices, we used commercially available Si/SiO2/Ti/Pt/ PbZr0.2Ti0.8O3 (PZT) substrates of size 4 × 4 mm2. Here, the tetragonal PZT, grown by spin-coating, is 150-nm-thick. The PZT substrates were partially covered with technological SiO2 in order to decrease the contact area and thus leakage current. After annealing the substrate to desorb contaminants, 3 nm of Co was thermally evaporated in UHV (Po ~ 1010 mbar) through a shadow mask with a circular opening of 2 mm diameter. The magnetic properties of the ferromagnetic (FM) layer were verified by magneto-optic Kerr effect measurements. Then, 1 ML of FePc was deposited from Knudsen cell heated up to 315 oC, and capped by 3 nm of Pt so as to prevent the organic and FM material from degradation in atmosphere. After that, the sample was removed from vacuum and transferred for a thermal deposition of Au top contact. The thickness and roughness of the Co layer was ascertained using X-Ray reflectivity measurements performed using a Rigaku Smartlab X-ray diffractometer equipped with a monochromatic source (Ge(400)x2) delivering a Cu Kα1 incident beam (45 kV, 200 mA, λ = 0.154056 nm) (see Fig. S1). Numerous fitting attempts were made so as to widely sweep parameter space. The best fit, with a quite good χ2 = 0.015 figure of merit given the roughness of the films, was found when the Ti and Pt underlayers were kept
constant to the 56 nm and 124 nm values found previously, while the PZT, Co and Pt layer thicknesses were iteratively used as free parameters, and the FePc molecular monolayer was omitted since it is too thin. We find a Co layer thickness of 3.4 nm with a statistical roughness σ = 2.0 nm. This roughness isn’t surprising given the similar roughness σ = 1.6 nm of the underlying spin-coated PZT layer, and the 3D growth mode expected when depositing Co onto the PZT oxide surface.
Figure S1. X-Ray reflectivity measurement of Ti/Pt/PZT/Co/FePc/Pt stacks. Experimental data points are shown in black, while the fit (red) with a figure of merit χ2 = 0.015 was found for Ti(56 nm), Pt(124 nm), PZT(145.9 nm; σ = 1.6 nm), Co(3.4 nm; σ = 2.0 nm) and Pt(3.1 nm; σ = 2.2 nm). The FePc molecular monolayer was omitted in the fit since it is too thin. The local structure of Co/PZT bilayers was further probed by means of transmission electron microscopy (TEM) using a Jeol JEM- 2100FS operated at 200 keV. The cobalt film was protected with a 3-nm-thick chromium layer to avoid damage to the bilayer during the thinning procedure. Figure S2 shows that the Co layer forms a continuous film that perfectly follows the granular topography of the textured tetragonal PZT film. The top Au pad served as the positive electrode, while the Pt layer below the substrate as the negative one. The PZT in this architecture exhibits an out-of-plane ferroelectric (FE) polarization, which we denote as P↑ (P↓) for ferroelectric polarization vector pointing towards (away from) the positive electrode. The FE polarization can be reversed upon applying a bias voltage between the electrodes. Note that our large (~ 3 mm2) device area inevitably leads to large leakage current (~ 20 mA) which eventually causes a PZT break down after several poling sequences. Yet such a large pillar size is necessary for a macroscopic synchrotron study given the presently available beam size ~ 800 × 800 µm2 when the electrical insert V2TI is used on beamline DEIMOS.
Figure S2. Low (a) and high (b) magnification transmission electron microscopy (TEM) image highlighting the continuity of the cobalt layer. The dotted red lines represent the boundaries of the film. The TEM study was conducted on a continuous Cr(3 nm)-capped Co(3 nm)/PZT(150 nm) bilayer. The system was in the as-grown state (i.e., no electric voltage applied prior to the experiment).
Ferroelectric properties of the device Prior to in situ x-ray absorption experiment, the complete ferroelectric (FE) switching of the device was tested with a custom-developed probe station connected to a ferroelectric tester (AixACCT TF 2000E analyzer). Both the dynamic hysteresis and pulsed measurements indicate intense peaks in the I(V) curves. This provides evidence for the existence of the FE polarization and its switching. We relied on the pulsed PUND (Positive Up Negative Down) measurement [1] since recording the dynamic P-loops is known to overestimate the polarization amplitude (strong contribution of the leakage current) which may lead to misleading results.[2] The obtained I(V) curves (Fig. S3) manifest typical peaks corresponding to two FE states of the device with the polarization of about 12 µC/cm2. In that way, we proved that the complete device preserves good FE properties and switching ability when ±10V are applied to the system. After testing procedure, the device was left in the initial P↓ state.
15
10
10 5
5
0
0
-5
-5
Current (mA)
Polarization (C/cm2)
15
-10
-10
-15 -15 -10
-5
0
5
10
Applied voltage (V)
Figure S3. Testing the Ferroelectricity of the device. Results of PUND measurement with the peaks in the I(V) characteristics corresponding to two FE states of the PZT substrate denoted later as P↑ and P↓.
XAS/XMCD experiment with in situ ferroelectric switching Experimental setup and FE poling procedure On the fabricated device we performed the XAS/XMCD measurements at the DEIMOS beamline[3] of the SOLEIL synchrotron with in situ electrical access for PZT FE polarization reversal provided by the V2TI insert.[4] The experimental geometry is presented in Fig. S4a. The reversal of the PZT FE state was provoked by a sequence of voltage pulses provided by the Agilent 33512B function/arbitrary waveform generator programmed and triggered via a LabView-based software. Each sequence consisted of two pre-poling pulses -10 V and +10 V followed by one +10 V pulse (two -10 V pulses) for the desired up polarization P↑ (down polarization P↓) state (see Fig. S4b). The electrical connection between the V2TI and the poling bench was established exclusively for polarization reversal and removed during the XAS measurement so as to minimize the electrical noise in the total electron yield signal.
(b) (a)
Figure S4. XAS/XMCD measurements at the DEIMOS beamline. (a) Experimental geometry of the XAS/XMCD measurements performed at the DEIMOS beamline.[3] The x-ray absorption spectra were recorded at room temperature with 45o x-ray incidence. Both total electron and fluorescence yield detections were recorded. The out-of-plane ferroelectric polarization of PZT was reversed with use of an Agilent 33512B wave function generator connected via the V2TI insert.[4] (b) Voltage pulses sequences applied to reverse the FE polarization of PZT in order to obtain the state with the polarization vector pointing towards the positive (P↑) or negative (P↓) electrode.
Ti L edge: validation of effective FE polarization reversal It was demonstrated in ex situ experiments that the FE switching of the PZT state should be accompanied by the reversible alterations observed in the XAS spectra at Ti L3,2 edges.[5] We employed this fact and used the XAS and x-ray linear dichroism (XLD) techniques to provide evidence for the effective FE polarization reversal with our protocol. Figure S5a and S5d present the XAS and XLD spectra acquired at the Ti L3,2 edges for the device in the initial P↓ state. The absorption curves reflect a typical XAS of Ti4+, i.e., dipolar transitions from d0 to 2p53d1 configuration. The spin-orbit coupling of the 2p states results in a 5.45 eV splitting[6] further subjected to a 1.62 eV octahedral crystal field splitting, which is FE polarization dependent, into eg and t2g symmetry groups.[7,8] The eg orbitals of Ti are pointing towards oxygen anions, meanwhile t2g are oriented in between them.[5,8] A stronger broadening of the transition assigned to eg orbitals results from the distorted Ti octahedral environment. This in turn leads to a further splitting of eg(O) into b1 = d(x2-y2) and a1 = d(z2), and of t2g(O) into b2 = d(xy) and e = d(xz, yz). This symmetry reduction is the reason of the unequal x-ray absorption for a linear horizontal and vertical polarization (Fig. S5b), reaching a maximum of ~20% at 2p3/23d(eg) peak. Note that the used PZT has a textured structure, resulting in a random orientation of the tetrahedral unit cells in the film plane, yet still with a distinguished c axis pointing out of the sample plane.
2p3/2 e g t2g
2.0
2p1/2
eg
lin. horizontal lin. vertical
t2g
1.5
XLD (a.u.)
1.0 0.2 0.1 0.0 -0.1
(b)
455
460 465 470 475 Photon energy (eV)
480
XAS intensity (a.u.) Diff. (a.u.) XAS intensity (a.u.)
(a) Ti L3,2-edges
Diff. (a.u.)
XAS intensity (a.u.)
2.5
2,5 (c) Ti L3,2-edges
P P
2,0 1,5 1,0
LH/LV x-ray pol. aver.
0,03 (d) 0,00
PP
-0,03 2,5
(e)
P P2
2,0 1,5 1,0
LH/LV x-ray pol. aver.
0,01 (f) 0,00 -0,01
P P2
455
460 465 470 475 Photon energy (eV)
480
Figure S5. XAS and XLD at Ti L3,2 edges of PZT. (a) XAS spectra at Ti L3,2 edges of PZT for the initial P↓ state. X-ray linear dichroism (XLD) of ~20% (b) results from the symmetry reduction of Ti4+ eg and t2g orbitals. The x-ray polarization averaged spectra (c), (e) for two consecutive PZT polarization reversals with the resulting differences (d) and (f) reveal the reversible change of the Ti electronic structure due to the displacement of the Ti atoms along the c axis. The data were acquired in total fluorescence yield mode at 0.1 T external magnetic field applied at 45º with respect to the sample surface, and at 300 K.
In the next step, we investigated the alterations at the Ti L3,2 edges of PZT upon the ferroelectric polarization reversal. We compared the x-ray polarization averaged spectra for the corresponding FE polarization states of PZT (Fig. S5c-f). First, we discuss the difference between the curves obtained for the initial P↓ state and the P↑ state (Fig. S5d), i.e. after the first poling sequence. We observe the most prominent feature at 2p3/23d(t2g) reaching ~3% which results from the energetic shift of the orbital. That is followed by a smaller variation of the intensity and position also for states at higher energies. This can be explained by an interplay of an increased electrostatic potential within the PZT unit cell for P↑ with respect to P↓ due to a deformation along the c axis,[5] and ensuing modification of the octahedral crystal field strength.[8] By comparing the spectra after the second polarization reversal from P↑ to P2↓ (Fig. S5f), we observe the same set of features as for the first poling. Knowing that our PZT is textured, and thus there is no in-plane organization, we infer that the observed alterations result from a displacement of the Ti atoms along the c axis of the unit cell. This is in agreement with recently reported observations at the Ti
K edge of PZT.[9] A further polarization reversal was not possible. This is most likely the result of the large device surface required for the experiment. Although the intensities of changes to the Ti L edges upon polarization reversal differ between Fig. S5d and S5f, we find a change to Co XMCD amplitudes that is similar (see Fig. 2d). We therefore claim to switch the PZT substrate polarization 3 times (per the ferroelectric testing procedure).
Impact of ferroelectric poling on the background of XAS at the Fe L3,2 edges We observed the FE poling to strongly affect the background of the x-ray absorption at the Fe L3,2 edges of FePc. Figure S6 presents these XAS spectra for both circular polarization of photons and resulting XMCD for the P↓, P↑, and P2↓ states of the PZT substrate. The evolution of the background at the tail of the scans is attributed to a ferroelectric-state dependent charge distribution within the system. Note, however, that the resulting XMCD spectra are not affected and exhibit a flat baseline such that the XMCD curves could be directly compared. (a)
(b)
3,0 2,5
Fe L3,2 edges
CR CL
P
2,5
2,0 XAS (a.u.)
1,5 1,0 0,5
CR CL
Fe L3,2 edges
2,0
P
1,5 1,0 0,5 XMCD
XMCD 0,0 -0,5
0,0
700
710
720
730
740
-0,5
700
(c)
710
3,0 2,5
CR CL
Fe L3,2 edges P
2,0 1,5 1,0 0,5 XMCD
0,0 -0,5
720
Photon energy (eV)
Photon energy (eV)
XAS (a.u.)
XAS (a.u.)
3,0
700
710
720
Photon energy (eV)
730
740
730
740
Figure S6. Impact of PZT FE switching on the XAS background at the Fe L3,2 edges of FePc. XAS spectra at the Fe L3,2 edges of FePc acquired for (a) P↓, (b) P↑, and (c) P2↓ state of the PZT substrate. The strong evolution of the background with the FE poling is attributed to alterations of the charge distribution within the system. However, the baseline of the XMCD remains flat, thereby justifying the comparison discussed in the main text. The data were acquired in total electron yield mode at 0.1 T external magnetic field applied at 45º with respect to the sample surface, and at 300 K.
Difference of XMCD at the Fe L3,2 edges upon ferroelectric polarization switching of PZT In the main text, and referring to Fig. 2c, we discuss changes to three features, labelled A, B, C, in the XMCD at the Fe L3 edge upon ferroelectric switching of PZT. We attribute these changes to the impact of the PZT/Co artificial magnetoelectric on the magnetic properties of the Co/FePc organic spinterface. Considering the low amplitude of these features against the overall L3 edge XMCD amplitude, and in order to buttress this link, we use the data of Fig. 2c to present in Figure S7 the XMCD differences (i) P↑ - P↓, (ii) P↑ - P2↓, as well as (iii) P↓ P2↓ as a control dataset. We find a very similar photon energy dependence of the difference spectra (i) and (ii) that is, in particular, characterized by similar modulations around features A, B and C. In comparison with the amplitude of these modulations, the XMCD difference of the control dataset (iii) is much flatter, indeed more featureless as expected, and overall does not mimic the XMCD differences (i) and (ii). This confirms the link between features A, B and C and the ferroelectric poling of PZT, and underscores how the same molecular magnetic properties at the Co/FePc interface arise upon successive P↓ and P2↓ ferroelectric states of PZT.
XMCD difference (a. u.)
Fe L3 edge A
B
C
P P P P2 P P2
709
710 711 Photon energy (eV)
712
Figure S7. XMCD difference spectra at the Fe L3 edge of FePc upon ferroelectric switching of the PZT ferroelectric. The differences were calculated from the data presented in Fig. 2c of the main text. The spectra P↑ - P↓ (black) and P↑ - P2↓ (red) exhibit a similar trend throughout the photon range considered, and in particular around energies A, B and C at which features in the XMCD spectra for the P↓ and P2↓ were observed (see Fig. 2c). The similar XMCD modulations around features A, B and C in these difference spectra contrast with the comparatively flatter XCMD difference obtained for the control dataset P↓ - P2↓ (blue), as expected.
Ab-initio study First-principles calculations based on the density functional theory (DFT) were performed using the Vienna Ab initio Simulation Package (VASP)[10,11] with projector augmented wave (PAW) pseudopotentials[12] and by applying the generalized gradient approximation (GGA-PBE).[13] The energy cut-off of the plane wave basis was fixed at 500 eV, and the first Brillouin zone was sampled with a 8 × 8 × 3 Monkhorst-Pack grid.[14] We used a 2 × 2 × 5.5 PbZr0.25Ti0.75O3(001) cell with a ZrTi3O8 surface termination, which corresponds to 11 atomic layers in the (001) direction. Three atomic layers of Co have been added at the ZrTi3O8-terminated PZT surface. The Co atoms in the first layer near the interface were initially located on top of the oxygen atoms of the interfacial ZrTi3O8 layer; the Co atoms of the second and the third layer are then distributed following a face-centered cubic lattice. The coordinates of the first 5 PZT atomic layers farthest away from the interface were kept fixed at their bulk value for a given electric polarization state, while the top layers were allowed to relax as described by Fechner et al.[15] The relaxed equilibrium structure near the interface was calculated for both upward (P↑) and downward (P↓) polarized PZT in order to study the electric polarization orientation effect on the electronic structure of the interface (density of states and spin magnetic moments). Our findings, summarized in Table S1, in particular show that small changes to the Co magnetic moment are still expected for atomic layers farther away from the PZT interface.
mS (µB/atom)
d(Co1-O) (Å) d(Co1-Ti) (Å)
O Zr Ti Co1 Co2 Co3
P↓ 0.077 0.001 -0.032 1.827 1.775 1.856
P↑ 0.026 -0.146 -0.388 1.600 1.769 1.867
Ti-O-Ti
1.871
2.016
Zr-O-Ti
1.884
2.091
Ti-O-Ti
3.007
2.584
Zr-O-Ti
3.020
2.584
Table S1. Variation of the spin magnetic moments (mS), averaged in each atomic layer, and of the interatomic Co−O and Co−Ti distances for two different environments (Ti-O-Ti and Zr-O-Ti)
upon polarization reversal from downward (P↓) to upward (P↑), obtained by means of firstprinciples calculations. References [1]
K. M. Rabe, C. Ahn, J.-M. Triscone, Physics of Ferroelectrics: A Modern Perspective, Springer-Verlag, Berlin, 2007.
[2]
J. F. Scott, J. Phys. Condens. Matter 2008, 20, 21001.
[3]
P. Ohresser, E. Otero, F. Choueikani, K. Chen, S. Stanescu, F. Deschamps, T. Moreno, F. Polack, B. Lagarde, J.-P. Daguerre, F. Marteau, F. Scheurer, L. Joly, J.-P. Kappler, B. Muller, O. Bunau, P. Sainctavit, Rev. Sci. Instrum., 2014, 85, 13106.
[4]
L. Joly, B. Muller, E. Sternitzky, J.-G. Faullumel, A. Boulard, E. Otero, F. Choueikani, J.P. Kappler, M. Studniarek, M. Bowen, P. Ohresser, J. Synchrotron Radiat. 2016, 23, 652.
[5]
E. Arenholz, G. van der Laan, A. Fraile-Rodríguez, P. Yu, Q. He, R. Ramesh, Phys. Rev. B 2010, 82, 140103.
[6]
J. C. Jan, H. M. Tsai, C. W. Pao, J. W. Chiou, K. Asokan, K. P. K. Kumar, W. F. Pong, Y. H. Tang, M.-H. Tsai, S. Y. Kuo, W. F. Hsieh, Appl. Phys. Lett. 2005, 87, 12103.
[7]
G. van der Laan, I. W. Kirkman, J. Phys. Condens. Matter 1992, 4, 4189.
[8]
S. C. Ray, H. C. Hsueh, C. H. Wu, C. W. Pao, K. Asokan, M. T. Liu, H. M. Tsai, C. H. Chuang, W. F. Pong, J. W. Chiou, M.-H. Tsai, J. M. Lee, L. Y. Jang, J. M. Chen, J. F. Lee, Appl. Phys. Lett. 2011, 99, 42909.
[9]
W. Kempet, B. Marungsri, R. Yimnirun, W. Klysubun, T. Fangsuwannarak, S. Rattanachan, N. Pisitpipathsin, M. Promsawat, S. Pojprapai, Ferroelectrics 2016, 492, 35.
[10] G. Kresse, J. Hafner, Phys. Rev. B 1994, 49, 14251. [11] G. Kresse, J. Furthmüller, Phys. Rev. B 1996, 54, 11169. [12] Blöchl, Phys. Rev. B. Condens. Matter 1994, 50, 17953. [13] J. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 1996, 77, 3865. [14] H. J. Monkhorst, J. D. Pack, Phys. Rev. B 1976, 13, 5188. [15] M. Fechner, S. Ostanin, I. Mertig, Phys. Rev. B 2008, 77, 94112.
