Tunable magnetoresistance in an asymmetrically coupled single molecule junction
Ben Warner, Fadi El Hallak, Henning Prüser, John Sharp, Mats Persson, Andrew J. Fisher, Cyrus F. Hirjibehedin
TTunable magnetoresistance in an asymmetrically coupled singlemolecule junction ∗ Ben Warner,
1, 2
Fadi El Hallak, † Henning Pr¨user, John Sharp, MatsPersson,
3, 4
Andrew J. Fisher,
1, 2 and Cyrus F. Hirjibehedin
1, 2, 5, ‡ London Centre for Nanotechnology,University College London (UCL), London WC1H 0AH, U.K. Department of Physics & Astronomy, UCL, London WC1E 6BT, U.K. Surface Science Research Centre and Department of Chemistry,University of Liverpool, Liverpool, L69 3BX, U.K. Department of Applied Physics, Chalmers Universityof Technology, SE-412 96, G¨oteborg, Sweden Department of Chemistry, UCL, London WC1H 0AJ, U.K. ∗ Nature Nanotechnol. 10, 259 (2015); http: // dx. doi. org/ 10. 1038/ nnano. 2014. 326 † Present address: Seagate Technology, Derry BT48 0BF, U.K. ‡ email: [email protected] a r X i v : . [ c ond - m a t . m e s - h a ll ] O c t henomena that are highly sensitive to magnetic fields can be exploited insensors and non-volatile memories [1]. The scaling of such phenomena down tothe single molecule level [2, 3] may enable novel spintronic devices [4]. Herewe report magnetoresistance in a single molecule junction arising from nega-tive differential resistance that shifts in a magnetic field at a rate two ordersof magnitude larger than Zeeman shifts. This sensitivity to the magnetic fieldproduces two voltage-tunable forms of magnetoresistance, which can be selectedvia the applied bias. The negative differential resistance is caused by transientcharging [5–7] of an iron phthalocyanine (FePc) molecule on a single layer ofcopper nitride (Cu N) on a Cu(001) surface, and occurs at voltages correspond-ing to the alignment of sharp resonances in the filled and empty molecular stateswith the Cu(001) Fermi energy. An asymmetric voltage-divider effect enhancesthe apparent voltage shift of the negative differential resistance with magneticfield, which inherently is on the scale of the Zeeman energy [8]. These resultsillustrate the impact that asymmetric coupling to metallic electrodes can haveon transport through molecules, and highlight how this coupling can be used todevelop molecular spintronic applications.
Research into magnetoresistance [9, 10] has been driven by the widespread use of giantmagnetoresistance (GMR) sensors in hard drives as well as other applications such as mag-netoresistive random access memory (MRAM) [1]. To reach even higher storage densities,research has begun to concentrate on magnetoresistance at the atomic scale [2, 3, 11]. Fora single molecule, however, the small area for enclosing flux and modest energy scales as-sociated with electronic Zeeman shifts typically make it difficult to tune magnetoresistivephenomena with an external magnetic field.Another electron transport phenomenon with technological relevance is negative differ-ential resistance (NDR) [5, 7, 12–19], in which an increase in voltage causes a decrease incurrent. Commercial devices, such as the resonant tunnelling diode, utilise these regionsin specialised applications [20, 21]. Various mechanisms cause NDR at the atomic scale[5, 7, 12–19], though none are expected to have a magnetic field dependence that wouldshift the NDR on a scale larger than the Zeeman energy.Using low temperature scanning tunnelling microscopy (STM) (see Supplementary Meth-ods), we observe an NDR effect for FePc molecules placed in a vacuum junction on top of2 Cu(001) surface capped with a single layer of Cu N (Fig. 1). Cu N is a thin insulatorthat can decouple the spins of magnetic atoms from the underlying surface [22]; FePc is amagnetic molecule that can be easily sublimed [23–25] and is observed to have interestingmagnetic properties on thin insulating layers [26]. On Cu N, FePc is centred above both Cuand N sites. The two binding sites can be differentiated using atomically resolved imagingand spectroscopic measurements; typical spectra of both types are seen in Fig. 1c. Further-more, a broad distribution of binding angles is observed, with shallow peaks at 0 ◦ , 18 ◦ , and45 ◦ with respect to the crystallographic axes. Density functional theory (DFT) calculations(see Supplementary Methods) indicate relatively weak variations in the binding energy withangle (see Supplementary Table and Supplementary Data).Remarkably, when a magnetic field is applied the NDR minimum can shift by two ordersof magnitude more than the electronic Zeeman effect (Fig. 1c), here almost 0.1 V for anapplied field of 6 T. To our knowledge such magnetic sensitivity has not been observed forother systems exhibiting NDR. The NDR effect is observed in 12 .
5% of the molecules (23 outof 184), at both positive and negative bias, and at a variety of voltages for different moleculeson both Cu and N sites (Supplementary Fig. 2). NDR is observed at various binding anglesfor both sites, suggesting that NDR occurs on molecules with different binding geometries.In almost all cases (see Supplementary Methods), NDR was observed only at the centre ofthe FePc molecule (i.e. above the Fe atom, see Supplementary Fig. 3).A more detailed dependence of the changes in the NDR spectra with perpendicular mag-netic field is shown in Fig. 2a. As seen in Fig. 2b, the voltage of the NDR minimum shiftsapproximately linearly with a slope of -15 mV/T. An increase in | B | shifts the NDR mini-mum to lower (less positive / more negative) voltages, but the slope varies from molecule tomolecule, ranging from -2 mV/T up to -15 mV/T. Furthermore, our measurements suggestthat the shift of the NDR depends only on the magnitude of the field component perpen-dicular to the plane, with an in-plane field of 1 T and a reversal of the sign of the magneticfield having no impact. Additionally, on rare occasions we have observed sharp peaks in theconductance spectra in similar voltage ranges that exhibit a similar dependence on B (seeSupplementary Fig. 4).The ability to manipulate NDR with a magnetic field not only enables tuning of the volt-age of the NDR minimum [28] but also results in the creation of a junction that exhibits twonovel magnetoresistance effects. Figure 3 shows a model of the NDR where the differential3onductance line shape G ( V, B ) is represented by a Lorentzian dip that shifts linearly with | B | on top of a constant background. For voltages that are more positive than the voltage ofthe NDR minimum, the change in differential conductance ∆ G ( V, B ) = G (V , B ) − G (V , | B | until it saturates. Remarkably, however, the mag-netoresistance ratio ∆ G ( V,B ) G ( V, can become arbitrarily large as V approaches the value at which G ( V,
0) = 0. Furthermore, its sign is positive or negative depending on the sign of G ( V, G ( V, B )initially becomes increasingly negative with | B | until it reaches a minimum value; after this,it becomes more positive, crosses zero, and then saturates at a limiting value. The magneticfield at which the polarity of the differential conductance “crosses over” varies with voltage,creating a magnetic-field sensitive switch that is tunable with voltage. As seen in Fig. 2c,both of these effects are observed for FePc on Cu N.To explain this novel manifestation of magnetically sensitive NDR (Supplementary Dis-cussion), we suggest a mechanism based on transient charging that arises from the occupationof molecular resonances [29]. This results in a change in the tunnelling rates through themolecule that can increase or decrease the differential conductance, with the latter resultingin NDR. Sharp states corresponding to a two-step resonant tunnelling process between thetip, the molecule, and the substrate have been observed in studies of individual moleculeson thin insulators [6]. In the resonant tunnelling process, voltage is dropped across bothbarriers (vacuum and Cu N) in the tunnel junction, with most of the drop expected to occurin the vacuum between the tip and the molecule (Fig. 4b). The small fraction of the appliedbias voltage dropped across the thin insulator therefore shifts the molecular orbitals withrespect to the substrate Fermi energy.Because the fraction of the voltage dropped across the thin insulator varies with therelative size of the tip-molecule gap, the hallmark of this mechanism is a shifting of theNDR minimum with the height of the tip above the surface [6]. As seen in Fig. 4c, theNDR minimum clearly shifts closer to the Fermi energy as the set point current (tip height)is increased (decreased). Fig. 4d further shows that position of the NDR minimum shifts4inearly with tip height, as expected. DFT calculations in which an electric field has beenadded to the system also show that a finite potential drop exists between the molecule andthe substrate.In these asymmetric, double-barrier tunnel junctions, the strongest resonance occurs whenone of the molecular orbitals aligns with the Fermi energy in the substrate (Fig. 4b) becausethe molecule is more strongly coupled to the substrate than to the tip [29]. Depending onwhether the alignment occurs with an empty or filled orbital, the molecule can be transientlynegatively or positively charged respectively during the transport process; since these occurat negative and positive bias respectively [6] and can result in either increased or decreaseddifferential conductance [7], the NDR can occur in either polarity of bias voltage for differentmolecules. Note that this charging is a consequence of the extended lifetime of the tunnellingelectron on the molecule: if the tip were moved away the molecule would quickly return toits neutral state.Because the molecular levels shift with respect to the Fermi energy by much less thanthe applied bias voltage, the apparent voltage scale of the resonance is enhanced [6, 7]. Thiscan be quantified by considering the behaviour with temperature. As seen in Fig. 5a, theNDR minimum becomes dramatically more shallow and broad with increasing temperature.Figure 5c shows that the depth of the NDR minimum decreases with a 1 /T dependence,where T is the substrate temperature, as expected for thermal smearing. The full widthat half maximum (FWHM) is shown in Fig. 5b and is found to increase linearly with arate of approximately (225 ± k B /e , where k B is the Boltzmann constant and e is themagnitude of the electron charge. The expected broadening for thermal smearing from theFermi seas in the tip and the substrate is 3.5 k B /e , so for this molecule the enhancement is225 / . ∼ µ eV/T, which is of the order of the Zeeman energy. Zeeman splitting of such sharpmolecular resonances into doublets has been observed in the presence of a magnetic field [8].This shows that an asymmetric junction not only can significantly influence the electronicproperties of the junction [30] but also allows for the enhancement of energy scales, causingsmall shifts in energy to be magnified. 5n this case, the fact that NDR is observed only over the centre of FePc molecules suggeststhat the resonant levels are associated with the Fe d-orbitals. Furthermore, because of thelarge exchange splitting between the majority and minority Fe d-levels (see SupplementaryFig. 1), the resonant levels are spin-polarised and non-degenerate; they would therefore shiftin the presence of a magnetic field rather than splitting. Since most of the levels close to theFermi energy are minority spin states (see Supplementary Fig. 1), it is sensible that we haveonly observed resonances shifting in one direction with field. The lack of an observed shiftwith the application of a small in-plane magnetic field is consistent with an axial anisotropyfor the total d-electron moment, oriented out of the plane, as has been observed for FePcon CuO [26].In a simple parallel-plate capacitor model of the tunnel junction formed by the tip, themolecule, and the underlying metal, the enhancement factor can also be described by thefraction of the voltage that is dropped across the Cu N: d ∗ / ( d ∗ + z ), where d = d ∗ (cid:15) is thedistance between the molecule and the underlying metal, (cid:15) is the effective dielectric constantof the Cu N monolayer, and z is the distance between the molecule and the tip. Althoughthere are no existing direct measurements of (cid:15) for Cu N, we can estimate d ∼ (0 . ± .
05) nmand z ∼ (0 . ± .
10) nm (Supplementary Methods) to obtain (cid:15) ∼ (60 ± N play a role in creating an enhancement value that ismuch higher than that observed on other thin insulators [6].Furthermore, the large enhancement factor explains the low number of molecules forwhich NDR is observed. In principle, all of the molecules should exhibit this phenomenon.However, the spectroscopic window in which we can measure is limited to ∼ ± . ∼
60, this results in ourmeasurements only being sensitive to levels within ∼
40 mV of the Fermi energy. Becausethe spectroscopy on each molecule varies, we can only observe NDR for molecules in whichthe appropriate levels lie close enough to the Fermi energy to fall within our measurementwindow.In summary, we observe magnetically sensitive NDR in a single-molecule junction arisingfrom resonant tunnelling producing charging in the molecule. The effective shift of the NDRwith magnetic field is enhanced by the inherent voltage division across the two asymmet-6ic tunnelling barriers; this allows for the creation of novel magnetoresistance phenomena.Similar enhancement of the effective energy scale for other multi-step tunnelling phenom-ena, both magnetic and non-magnetic in origin, should be possible. Furthermore, the sizeof the enhancement can be controlled by tuning the asymmetry of the tunnelling barriers,which can be modified by making physical or chemical changes to the junction by usingdifferent thin insulators or molecules [7, 26, 32–34] . This highlights the prominent role thatthe junction itself can play in defining the properties of the smallest possible electronic andspintronic device architectures. [1] Wolf, S. A. et al.
Spintronics: A Spin-Based Electronics Vision for the Future.
Science et al.
Giant magnetoresistance through a single molecule.
Nature Nanotech. et al. Large Magnetoresistance through a Single Molecule due to a Spin-Split Hybridized Orbital.
Nano Lett.
NatureMater. Science
Phys. Rev. Lett.
Phys. Rev. Lett.
Phys. Rev. Lett.
Phys. Rev. B oom temperature in ferromagnetic thin film tunnel junctions. Phys. Rev. Lett.
Nature Mater.
Science et al.
Mechanism for Negative Differential Resistance in Molecular Electronic Devices:Local Orbital Symmetry Matching.
Phys. Rev. Lett. et al.
Negative differential resistance in the scanning-tunneling spectroscopy oforganic molecules.
Phys. Rev. B
R7852-7855 (1999).[15] Wang, B. et al.
Effects of discrete energy levels on single-electron tunneling in coupled metalparticles.
Appl. Phys. Lett.
Phys. Rev. Lett.
Appl. Phys. Lett.
Phys. Rev. Lett.
Phys. Rev. Lett. et al.
Oscillations up to 712 GHz in InAs/AlSb resonant-tunneling diodes.
Appl.Phys. Lett.
IEEEElectron Device Lett.
Science et al.
Scanning tunneling microscope study of iron(II) phthalocyanine growthon metals and insulating surfaces.
Surf. Sci. et al.
Orbital specific chirality and homochiral self-assembly of achiral moleculesinduced by charge transfer and spontaneous symmetry breaking.
Phys. Rev. Lett.
25] Minamitani, E. et al.
Symmetry-Driven Novel Kondo Effect in a Molecule.
Phys. Rev. Lett. et al.
Adsorption-Induced Switching of Magnetic Anisotropy in a Single Iron(II)Phthalocyanine Molecule on an Oxidized Cu(110) Surface.
Phys. Rev. Lett.
Science
The Physics and Applications of Resonant Tunnelling Diodes .(Cambridge Univ. Press, Cambridge, 1995).[29] Mikaelian, G., Ogawa, N., Tu, X. W., & Ho, W. Atomic scale control of single moleculecharging.
J. Chem. Phys. et al.
Spin coupling and relaxation inside molecule metal contacts.
Nat. Commun.
490 (2011).[31] Huang, J., et al.
Iron-phthalocyanine molecular junction with high spin filter efficiency andnegative differential resistance.
J. Chem. Phys. et al.
The Quantum Magnetism of Individual Manganese-12-Acetate MolecularMagnets Anchored at Surfaces.
Nano Lett.
Phys. Rev. Lett. et al.
Probing Superexchange Interaction in Molecular Magnets by Spin-Flip Spec-troscopy and Microscopy.
Phys. Rev. Lett.
ACKNOWLEDGEMENTS
We acknowledge Gabriel Aeppli, Vincent Crespi, Jeroen Elzerman, Joaqu´ın Fern´andez-Rossier, Mark Hybertsen, Peter Littlewood, Sebastian Loth, Christoph Mathieu, MarkusTernes, and Joris van Slageren for simulating discussions. B.W., F.E.H., H.P., A.J.F., andC.F.H. acknowledge financial support from the EPSRC [EP/H002367/1 and EP/D063604/1]and the Leverhulme Trust [RPG-2012-754]. M.P. is grateful for support from the EU project9RTIST and allocations of computer resources at HECToR through the Materials ChemistryConsortium funded by EPSRC [EP/L000202/1] and at PDC through SNIC.
AUTHOR CONTRIBUTIONS
F.E.H. and C.F.H. conceived of the experiments; B.W., F.E.H. and H.P. performed theexperiments and analysed the results; J.S. and M.P. performed the DFT calculations; allauthors discussed the results and contributed to the writing of the paper.
ADDITIONAL INFORMATION
COMPETING FINANCIAL INTERESTS
The authors declare no competing financial interests.10
IGURES d I/ d V ( n S ) -2.0 -1.0 0.0 1.0 2.0 Bias (V)
STM tip FePc Cu N a b c H e i gh t ( p m )
400 300 200 100 0
FIG. 1. FePc on Cu N/Cu(001). a) Schematic of the experimental configuration with the tunneljunction formed by the STM tip and Cu N/Cu with the FePc molecule sitting in between. Atomsin the molecule and Cu N/Cu are colour coded: grey=C, white=H, blue=N, orange=Fe, lightbrown=Cu. The tip (red and brown) is PtIr, however the last atoms are likely Cu because the tipis often indented into the surface to reshape it. b) STM topograph of the surface showing variousFePc molecules ( V set = − . I set = 0 . dI/dV spectroscopy measurements taken abovethe centre of different molecules ( V set = − . I set = 0 . B = 6 T are shown, but the features can vary significantly from molecule to molecule.Red spectra show a clear NDR feature, which appears in 12 .
5% of the molecules. This can shift byup to -15 mV/T, as seen in spectra taken at 0 T (red) and 6 T (black). The magnetic field onlymoves features in the NDR region: other features in the spectrum remain constant. Traces havebeen offset vertically for clarity; dI/dV = 0 is indicated by a dashed line for each trace. .01.00.0-1.0 ! G ( n S ) B (T) N D R m i n i m u m ( V ) B (T) !" d I/ d V ( n S ) Voltage(V)
OT1.5T3T4.5T6T
FIG. 2. Differential conductance changes caused by magnetic field sensitive NDR. a) Differentialconductance spectra ( V set = − . I set = 0 . B = 0 T, 1.5 T, 3.0 T, 4.5 T, and 6.0 T (as labelled). As B is increased theNDR region moves to lower voltages. Spectra are offset vertically for clarity. Vertical dashed linesindicate 1.85 V (red), 1.90 V (blue), and 1.95 V (black). b) NDR minimum vs. B , with the solidline showing a gradient of -15mV/T. Error bars show the uncertainty in defining the minimum foreach spectrum. c) ∆ G ( V, B ) = G (V , B ) − G (V ,
0) versus B at 1.95 V (black), 1.9 V (blue), and1.85 V (red). - - - D G - - - M RR a ti o - - - D G - - M RR a ti o - -
10 Voltage d I / d V a c b d e XMR+%%%%%%%%%%%%%%%XMR&%%%%%%%%%%%%%%%MR&%%%%%%%%%%%%%%%MR+%%%%%%%%%%%%%%%
B=0 B>0
FIG. 3. Model of magnetically sensitive NDR. a) Differential conductance versus voltage for con-stant differential conductance background with an NDR feature that shifts linearly with | B | ; twovalues of B , B = 0 (thick black curve) and | B | > G and the magnetoresistance ratio are labelled MR ± and XMR ± respectively. All units have arbitrary dimensions. b) ∆ G vs. B in XMR+ and XMR- regime. c)Corresponding magnetoresistance ratio from panel b. d) ∆ G vs. B in MR+ and MR- regime. e)Corresponding magnetoresistance ratio from panel d. " Tip Sample Molecule
Vacuum Barrier Thin Insulator V m-s V t-m -1.4-1.3-1.2-1.1-1.0 N D R m i n i m a ∆ Z (x10 -12 m) $" %" Thin Insulator Vacuum Barrier
Molecule Sample Tip -4-202 d I / d V ( A r b . ) -1.5 -1.0Voltage (V) FIG. 4. Tunnelling across a double barrier junction. a) Junction at V = 0, with the Fermi levelsof the tip and substrate aligned. Solid horizontal lines indicate the filled and empty states of themolecule; dashed horizontal line is a reference between these levels. b) Junction at V = V res > V m − s aligns the molecular orbital with the Fermi level of the substrate. The remainingvoltage is dropped across the vacuum barrier (i.e. between the tip and the molecule) V t − m , allowingthe molecular levels to shift with respect to the substrate Fermi level. Note that the potentialsaccount for the negative sign of the charge carriers. c) Selected dI/dV vs. voltage spectra ( V set = − . I set =25 pA (black), 250 pA (blue), and 500 pA (red). d) NDR minimumversus change in tip-substrate distance ∆ z , which is calibrated using I ( z ) spectroscopy. As the tipmoves towards the substrate (higher current setpoint), the NDR minimum shifts linearly closer tothe Fermi energy, as highlighted by the solid red line. .300.200.10 F W H M ( V ) Temperature (K) !" -2.0-1.5-1.0-0.50.00.51.0 d I/ d V ( n S ) Voltage(V) -2.0-1.5-1.0-0.50.0 d I/ d V ( n S ) Temperature (K)
FIG. 5. Temperature dependance of NDR. a) Zoom-in of the NDR minimum in differential con-ductance (V set = − . set = 0 . k B T /e . c)Differential conductance at the NDR minimum vs. temperature. Solid red line is a guide to theeye for a 1 /T decay.decay.