Distinction of Nuclear States with the Scanning Tunneling Microscope

Fabian Donat Natterer, Fran¸coisPatthey, and Harald Brune Institute of Condensed Matter Physics (ICMP), Ecole Polytechnique F´ed´erale de Lausanne (EPFL), Station 3, CH-1015 Lausanne

We demonstrate rotational excitation spectroscopy with the scanning tunneling microscope for physisorbed H2 and its isotopes HD and D2. The observed excitation energies are very close to the gas phase values and show the expected scaling with moment of inertia. Since these energies are characteristic for the molecular nuclear spin states we are able to identify the para and ortho species of and , respectively. We thereby demonstrate nuclear spin sensitivity with unprecedented spatial resolution.

PACS numbers: 67.63.Cd, 67.80.ff, 67.80.F-, 33.20.Sn, 21.10.Hw, 68.43.-h, 68.37.Ef

Inelastic tunneling spectroscopy (IETS) (a) 15 Å (b) 15 Å probes the energies of atomic and molecular excitations in a tunnel junction. When the electron energy reaches the excitation threshold, a new conductance channel opens, leading to a step in the differential conductance (dI/dV ). IETS measurements were first carried out in planar tunnel junctions probing vibrations [1] and mag- netic excitations [2] of large ensembles of or atoms. A major breakthrough was achieved when per- forming IETS with the scanning tunneling microscope (STM). This has first been demonstrated for molecu- lar vibrations [3], identifying the molecules and their isotopes [4]. A few years later, this was followed by spin-excitations, revealing the Land´e g-factor, effective spin moment, and magnetic anisotropy energy [5,6]. (c) 5 Å (d) 1 Å-1 In either case, this information is retrieved for indi- vidual atoms and molecules of well known adsorption site, coordination-number, and -chemistry. These studies have significantly improved our understanding of surface chemistry and magnetism. The only process that could so far not be character- ized by IETS, neither in planar junctions nor in STM, are [011] true molecular rotations. Albeit, their excitation energies H2 B N Ni [211] contain manifold information, e.g., on chemical identity, FIG. 1. STM images of H2 superstructure on h-BN/Ni(111)– bond lengths, rotational degrees of freedom, and molec- (1 × 1). (a) Atomically resolved h-BN and circular areas of ular conformations. Notably, for homonuclear diatomics, hydrogen superstructure centered around Ti adatoms (expo- −6 the allowed rotational transitions depend on the nuclear sure 1 Langmuir H2, 1 L√ = 1.33√×10 mbar s, Vt = −10 mV, ◦ spin state. It = 20 pA). (b) Full ( 3 × 3)R30 H2 monolayer (100 L Here we demonstrate rotational excitation spec- H2, Vt = −20 mV, It = 20 pA). (c) Structure model su- perimposed on STM image of the H superstructure. (d) troscopy (RES) with the STM for physisorbed hydrogen, 2 Fourier√ transform√ of (a). The lozenges indicate the (1 × 1) deuterium, and deuterium-hydride. We observe sharp and ( 3 × 3)R30◦ unit cells. conductance steps in dI/dV at the energies correspond-

arXiv:1307.7046v1 [cond-mat.mes-hall] 26 Jul 2013 ing to the allowed rotational transitions of the respec- tive molecules in the gas phase. The ortho and para nuclear spin of hydrogen and deuterium entail STM-RES is proposed to involve a resonant molecular different rotational ground states [7]. We identify their ensemble state. In order to prevent its screening by the distinct excitation energies and thus demonstrate nuclear metal substrate, we introduced a monolayer of hexagonal spin sensitivity on ensembles containing by many orders boron nitride (h-BN) or graphene [15]. of magnitudes less molecules than probed in neutron We focus here on molecules that were physisorbed on diffraction [8,9], nuclear magnetic resonance [10, 11], h-BN/Ni(111)–(1 × 1) grown by chemical vapor depo- and high-resolution electron energy loss spectroscopy sition using borazine precursors [16]. The H2,D2, or (HREELS) [12–14]. The mechanism at the origin of HD molecules were subsequently dosed onto the surface 2 at 10 K and the STM measurements were performed at straints of the total molecular wavefunction [7]. It is a 4.7 K. The dI/dV spectra were measured with a Lock-In product of the nuclear, rotational, electronic, and vibra- amplifier using a bias modulation of 2 mV peak-to-peak tional wavefunctions. Hydrogen nucleons are , at 397 Hz. therefore this product must be antisymmetric with re- At low coverages, physisorbed hydrogen forms a two di- spect to permutation. For hydrogen, the vibra- 1 + mensional gas [8] that is transparent to the STM allowing tional and the electronic ( Σg ) ground states are sym- the imaging of the underlying h-BN with atomic resolu- metric. Consequently, the antisymmetric nuclear singlet tion as shown in Fig.1. The honeycomb lattice appears state (S = 0, para) requires a symmetric rotational wave- as hexagonally close-packed depressions. We adsorbed function (even J), whereas the symmetric nuclear triplet individual Ti atoms [17] in order to condense part of the state (S = 1, ortho) implies an antisymmetric rotational H2 gas in circular islands centered around the adatoms. wavefunction (odd J). The nucleons of deuterium are The H2 molecules are imaged as protrusions. They are√ requiring a symmetric molecular wavefunction. in registry with the h-BN depressions, however, at 3 Hence, the symmetric nuclear spin state is associated times their distance and rotated by 30◦. Upon satura- with a symmetric rotational state and the antisymmetric tion coverage,√ √ H2 forms a perfectly ordered monolayer nuclear spin configuration with an antisymmetric rota- of this ( 3 × 3)R30◦ superstructure, see Fig.1 (b). tional state. Transitions between the nuclear spin iso- Many weakly physisorbed adsorbates adopt this struc- mers are forbidden for free molecules, but are catalyzed ture, notably hydrogen on graphite [8, 10, 18] and on by paramagnetic impurities or inhomogeneous magnetic boron nitride [11]. and electric fields [21, 22]. The dI/dV spectra on the full monolayers of H2,D2, For the case of HD, the nucleons are distinguishable and HD are shown in Fig.2 (a). Each curve reveals two and the above symmetry constraints do not apply. The pairs of conductance steps. Their threshold energies are observed threshold energy for HD is in agreement with symmetric around zero bias, as expected for IETS. The a J = 0 → 2 excitation (Fig.2). For this , also numerical derivatives d2I/dV 2 in Fig.2 (b) are used to ∆J = 1 transitions are allowed, and the spectrum in- determine their values. We focus first on the high energy deed shows a little shoulder at (13 ± 1) meV close to the excitations that are located at (43.75±0.07), (32.8±0.4), reported J = 0 → 1 transition energy (TableI). Note and (20.89±0.07) meV, for H2, HD, and D2, respectively. that the RES steps are with 11 – 37 % [19] significantly Their ratios unambiguously identify them as rotational higher than the ones of vibrational-excitations for adsor- transitions since the energy of a rotational quantum state bates on metals [3], and they are comparable with the rot 2 J of a linear molecule is with EJ = J(J + 1)~ /2I spin-excitation step heights observed for magnetic atoms inversely proportional to its moment of inertia I. In on a decoupling monolayer [5,6]. addition, the absolute excitation energies of all three We attribute the low-energy conductance steps in molecules match the reported gas-phase values (cp. Ta- Fig.2 to the excitation of phonons in the molecular lay- bleI). Most importantly, the spectra identify H 2 in its ers. As can be seen from Fig.2 (c), the substrate po- para and D2 in its ortho nuclear spin configuration. tential creates a phonon gap at the Brillouin zone center reaching from zero to the energy where the weakly dis- persing bands are located [9, 20, 23, 24]. This creates a TABLE I. Comparison of gas phase and surface adsorbed rotational excitation energies ∆Erot(J → J 0) for the three narrow energy interval in which phonons can be excited molecules in their vibrational ground state. S labels the and thus meets the necessary condition for the observa- molecular spin and J the rotational ground state quantum tion of a distinct threshold energy in IETS. The excita- number. Note that even–odd transitions are forbidden for H2 tions are with (5.5±0.5), (5.1±0.5), and (4.4±0.5) meV, rot and D2, while they are allowed for HD. The gas phase ∆E for H2, HD, and D2, remarkably close to the centers of values are taken from Ref. [7]. The error bars for STM-RES the measured phonon bands [9, 24]. Notably, the H2/D2 indicate the standard deviation. energy ratio of 1.3 matches the one of the phonon gaps. p √ The deviation from mD2 /mH2 = 2 is caused by the Molecule SJErot (meV) STM-RES (meV) anharmonicity of the intermolecular and of the adsorp- ∆J = 1 ∆J = 2 tion potential [9, 20, 23, 25]. para 0 0 − 43.9 43.75 ± 0.07 We determined in Fig.3 the lateral range λ of excited H2 ortho 1 1 − 72.8 not observed molecules probed with STM-RES by recording dI/dV spectra on samples with partial h-BN coverage. This HD 0 11.1 32.9 13 ± 1 and 32.8 ± 0.4 √ gives access to physisorbed H2, once in the 3 phase on ortho 0/2 0 − 22.2 20.89 ± 0.07 D2 h-BN, for which we observe RES features, and once di- para 1 1 − 36.9 not observed rectly adsorbed on Ni(111), where rotational excitation conductance steps are absent, in agreement with former The distinct rotational excitation energies of the dis- low-T STM studies [26–29]. The dI/dV spectra recorded parate nuclear spin states are caused by symmetry con- across the transition between these two surface termi- 3

(a) (b) (c) 50

H2 H2 2 4 40

dI/dV HD HD 30

(arb. units) 2 2 I/dV 2 normalized

1 D2 d 20 Energy (meV)

0 D2 10

0 -2 0 -100 -50 0 50 100 -100 -50 0 50 100 Г K M Г Bias (mV) Bias (mV)

FIG. 2. Rotational excitation spectroscopy on molecular monolayers. (a) dI/dV of H2 ( ), HD (4), and D2 (). The conductance steps at 43.7, 32.8, and 20.9 meV represent rotational J = 0 → 2 transitions. The values for H2 and D2 are characteristic for the para-H2 and ortho-D2 configuration, respectively. The low-energy steps close to the Fermi energy are attributed to phonon gaps of the molecular layer. Spectra are averages of 615 ( ), 1100 (4), and 75 () dI/dV -curves. The spectra were vertically offset by 0.5 nA/V for clarity. (b) Numerical derivative of (a), d2I/dV 2. The black full lines show fits with a multi peak Lorentzian function [19]. The shoulder at (13 ± 1) meV (arrows) for HD is attributed to a J = 0 → 1 transition. (c) Roton and phonon bands calculated for p-H2 and o-D2 on graphite [20]. nations reveal the attenuation of the rotational excita- ies [26–29] and the absence of RES features, that we also tion when approaching the h-BN border to a distance of note in Fig.3 for H 2/Ni(111), must be due to screening λ = (1.7±0.5) nm. Within this radius there are (60±30) of the intermolecular interactions by the underlying sub- hydrogen molecules. This represents the ensemble size strate. Therefore a decoupling h-BN layer enables the probed for the present system by STM-RES. spectroscopy of molecular rotations with STM-IETS and The mechanism underlying STM-RES has to be differ- the collective nature of the resonant state explains the fi- ent from the one of HREELS, since the latter detected nite lateral range of excited molecules. According to this rotational excitations for physisorbed hydrogen on metal mechanism, STM-RES is an intrinsically multi-molecule surfaces [12, 13], while in STM there were no signs of such method, requiring for the case of H2 at least 60 interact- excitations for the same systems [26–29]. Instead, all ing molecules. Our collective rotational excitations can observed spectroscopic features were reminiscent of two be interpreted in terms of the calculated roton bands [20] state switching [26]. In EELS, a negative ion resonance shown in Fig.2 (c). We note that 60 molecules is an un- is populated by the primary and subsequently precedented small number in terms of the demonstrated decays into several inelastic channels, one of them being access to the nuclear spin state. the molecular rotation [30, 31]. For hydrogen this reso- With para-H2 and ortho-D2 we observe for each nance is at an energy accessible to the typical 5 eV inci- molecule only the nuclear spin with lowest en- dent electrons [12, 13, 30, 32], but evidently not to IETS ergy rotational ground state, J = 0. However, for ei- operating at electron energies near the excitation thresh- ther molecule both nuclear spin isomers are present in old of the molecular rotations. However, the existence the gas phase, with ortho/para ratios of 3/1 for H2 and of a rotational resonance for ultra-low electron energies 2/1 for D2 at room temperature. Therefore a fast con- was demonstrated by molecular density dependent elec- version to the lowest energy nuclear spin configuration tron drift velocity measurements [33, 34]. It has been at- must take place on the surface. In line, former EELS tributed to a collective resonance state originating from studies either observed only the spin isomer with even the electrostatic, polarization, and quadrupole interac- J [14], as in our case, or they reported the ortho to para- tions between neighboring hydrogen molecules [35]. This H2 conversion after a few minutes [13]. This conversion state requires molecular densities of comparable order of has been attributed to short range magnetic interactions 14 −2 magnitude than in the present work (6.2 × 10 cm ). with the surface [21]. We never observed ortho-H2 and Similar densities were also present in former STM stud- para-D2, neither on h-BN nor on graphene, both grown 4

(a) 3 nm (b) lution, as well as the intriguing ordering phenomena of Ni(111) ortho–para mixtures [10, 11]. Furthermore, the coupling 0.3 0.6 0.9 1.2 1.5 1.8 of the nuclear spin to the atomic environment and poten- dI/dV (arb. units) tially also nuclear processes become accessible on a local III 10 III scale. We finally note that even single molecule STM-

II II 8 RES might be feasible for those molecules with very low

Distance and broad negative ion resonance energy, thereby creat- 6 ing the necessary overlap with the rotational excitation threshold.

4 (nm) λ Funding from the Swiss National Science Foundation I I is greatly appreciated. 2

0 h-BN/Ni(111) -100 -50 0 50 100 Bias (mV) (c) (d) [1] J. Lambe and R. C. Jaklevic, Phys. Rev. 165, 821 (1968).

300 d I/dV (arb. units) [2] E. L. Wolf, Principles of Electron Tunneling Spectroscopy (Oxford University Press, New York, 1989). 2 [3] B. C. Stipe, M. A. Rezaei, and W. Ho, Science 280, 1732 (pm) 200 III II (1998). 100 I 1 [4] L. J. Lauhon and W. Ho, Phys. Rev. B 60, R8525 (1999).

Height [5] A. J. Heinrich, J. A. Gupta, C. P. Lutz, and D. M. Eigler, 0 0 Science 306, 466 (2004). 0 2 4 6 8 10 12 -100 -50 0 50 100 [6] C. F. Hirjibehedin, C.-Y. Lin, A. F. Otte, M. Ternes, Distance (nm) Bias (mV) C. P. Lutz, B. A. Jones, and A. J. Heinrich, Science 317, 1199 (2007). FIG. 3. The lateral extent λ of the rotational excitation. (a) [7] I. F. Silvera, Rev. Mod. Phys. 52, 393 (1980). STM image showing a step between the h-BN layer and the [8] M. Nielsen, J. P. McTague, and W. Ellenson, J. Phys. underlying Ni substrate, both are H2 covered. (b) dI/dV Colloques 38, C4 (1977). taken along the full red line perpendicular to the step. The [9] V. L. P. Frank, H. J. Lauter, and P. Leiderer, Phys. Rev. rotational excitation is attenuated at λ = (1.7 ± 0.5) nm from Lett. 61, 436 (1988). the step, see gray dashed line. (c) Line-profile across the step [10] P. R. Kubik, W. N. Hardy, and H. Glattli, Can. J. Phys. in (a). (d) dI/dV spectra at the indicated positions I, II, and 63, 605 (1985). III. Spectra represent averages of 16 dI/dV -curves. [11] K. Kim and N. S. Sullivan, Phys. Rev. B 55, R664 (1997). [12] S. Andersson and J. Harris, Phys. Rev. Lett. 48, 545 (1982). on non-magnetic substrates [15]. Therefore the conver- [13] P. Avouris, D. Schmeisser, and J. E. Demuth, Phys. Rev. sion to the lowest energy configuration must be driven Lett. 48, 199 (1982). by magnetic impurities. The electric fields of the tunnel [14] R. Palmer and R. Willis, Surf. Sci. 179, L1 (1987). junction are by at least one order of magnitude smaller [15] F. D. Natterer, F. Patthey, and H. Brune, in preparation. than required [22]. [16] A. Nagashima, N. Tejima, Y. Gamou, T. Kawai, and C. Oshima, Phys. Rev. B 51, 4606 (1995). There is pioneering STM work related to molecular [17] F. D. Natterer, F. Patthey, and H. Brune, Surf. Sci. 615, rotations. STM was used to induce and monitor the 80 (2013). rotation of O2/Pt(111) [36], the coupling of vibrational [18] J. Seguin and J. Suzanne, Surf. Sci. Lett. 118, L241 and rotational degrees of freedom was demonstrated for (1982). acetylene/Cu(100) [37], and hindered rotations were re- [19] Please consult the Supporting Material. ported for CO on two low-index Cu surfaces [4]. How- [20] W. B. J. M. Janssen, T. H. M. van den Berg, and ever, here we reveal for the first time the molecular rota- A. van der Avoird, Phys. Rev. B 43, 5329 (1991). [21] E. Ilisca, Prog. Surf. Sci. 41, 217 (1992). tional eigenvalues and thereby complete the meanwhile [22] T. Sugimoto and K. Fukutani, Nat. Phys. 7, 307 (2011). well established and widely used STM-IETS vibrational [23] A. D. Novaco, Phys. Rev. Lett. 60, 2058 (1988). and spin-excitation spectroscopy. [24] H. Lauter, V. Frank, P. Leiderer, and H. Wiechert, Phys- We demonstrated for H2 and D2 nuclear spin sensitiv- ica B 156, 280 (1989). ity and proposed a mechanism involving a collective low- [25] L. Mattera, F. Rosatelli, C. Salvo, F. Tommasini, U. Val- energy resonant state that emerges from molecular inter- busa, and G. Vidali, Surf. Sci. 93, 515 (1980). actions. STM-RES gives access to the eigenvalues of any [26] J. A. Gupta, C. P. Lutz, A. J. Heinrich, and D. M. Eigler, Phys. Rev. B 71, 115416 (2005). molecular rotor; particularly well suited are molecules 2 [27] R. Temirov, S. Soubatch, O. Neucheva, A. C. Lassise, with large rotational constants ~ /2I. For homonuclear and F. S. Tautz, New J. Phys. 10, 053012 (2008). diatomics, such as N2 and O2, the nuclear spin states [28] M. Sicot, O. Kurnosikov, H. Swagten, and B. Koopmans, can now be inspected with unprecedented spatial reso- Surf. Sci. 602, 3667 (2008). 5

[29] C. Weiss, C. Wagner, C. Kleimann, M. Rohlfing, F. S. [33] L. Frommhold, Phys. Rev. 172, 118 (1968). Tautz, and R. Temirov, Phys. Rev. Lett. 105, 086103 [34] R. W. Crompton and A. G. Robertson, Aust. J. Phys. (2010). 24, 543 (1971). [30] G. J. Schulz, Rev. Mod. Phys. 45, 423 (1973). [35] W. R. Garrett, Phys. Rev. A 16, 2305 (1977). [31] R. E. Palmer and P. J. Rous, Rev. Mod. Phys. 64, 383 [36] B. C. Stipe, M. A. Rezaei, and W. Ho, Science 279, 1907 (1992). (1998). [32] D. Teillet-Billy and J. Gauyacq, Surf. Sci. 502, 358 [37] B. C. Stipe, M. A. Rezaei, and W. Ho, Phys. Rev. Lett. (2002). 81, 1263 (1998).