BaFe2As2 Surface Domains and Domain Walls: Mirroring the Bulk Spin Structure
Guorong Li, Xiaobo He, Ang Li, Shuheng H. Pan, Jiandi Zhang, Rongying Jin, A. S. Sefat, M. A. McGuire, D. G. Mandrus, B. C. Sales, E. W. Plummer
aa r X i v : . [ c ond - m a t . s up r- c on ] J un BaFe As Surface Domains and Domain Walls: Mirroring the Bulk Spin Structure
Guorong Li , Xiaobo He , Ang Li , Shuheng H. Pan , Jiandi Zhang , Rongying Jin ,A. S. Sefat , M. A. McGuire , D. G. Mandrus , , B. C. Sales , and E. W. Plummer Department of Physics & Astronomy, Louisiana State University, Baton Rouge, Louisiana 70802, USA Department of Physics & Texas Center for Superconductivity,University of Houston, Houston, Texas 77204-5002, USA Materials Science & Technology Division, Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831, USA and Department of Materials Science & Engineering,The University of Tennessee, Knoxville, Tennessee 37996, USA
High-resolution scanning tunneling microscopy (STM) measurements on BaFe As − one of theparent compounds of the iron-based superconductors − reveals a (1 ×
1) As-terminated unit cell onthe (001) surface. However, there are significant differences of the surface unit cell compared to thebulk: only one of the two As atoms in the unit cell is imaged and domain walls between different(1 ×
1) regions display a C symmetry at the surface. It should have been C v if the STM imagereflected the geometric structure of the surface or the orthorhombic bulk. The inequivalent As atomsand the bias dependence of the domain walls indicate that the origin of the STM image is primarilyelectronic not geometric. We argue that the surface electronic topography mirrors the bulk spinstructure of BaFe As , via strong orbital-spin coupling. PACS numbers: 68.35.B-, 68.37.Ef, 73.20.-r, 74.40.Xa
The discovery of high-temperature superconductivityin the Fe-based compounds has generated enormous ex-citement and activity in the scientific community [1, 2].Not only is this a new class of materials exhibiting someform of unconventional superconductivity but at the firstglance the behavior resembles that of the cuprates [3],raising the expectation that the Fe-based superconduc-tors might offer an avenue to understand the inherentpairing mechanism responsible for superconductivity inboth systems. The ground state of the parent com-pounds in the cuprates and Fe-based superconductorsis antiferromagnetically (AFM) ordered and it appearsthat the magnetic ordering must be suppressed in orderto achieve superconductivity. Both sets of materials ex-hibit a superconducting “dome” as a function of eitherhole or electron doping. However, as more data becomesavailable for the Fe-based compounds, it is becoming in-creasingly clear that the members of this family behaverather differently from the cuprates. The AFM groundstate of the Fe-based parent compounds is metallic butMott insulating for the cuprates. The small magnetic mo-ments [4, 5] and the characteristic of electronic structureprobed by photoemission measurements [6] indicate thatthe Fe bands are like an itinerant metal not localized asin the cuprates. While cuprates such as La − x Sr x CuO undergo a structural transition [3], there is no evidencefor the coupling between structure and AFM ordering.In Fe-based compounds, there is complex coupling be-tween lattice and spin degrees of freedom: a structuraltransition from a high-temperature tetragonal (HTT) toa low-temperature orthorhombic (LTO) phase is alwaysaccompanied by a magnetic transition within a narrowtemperature window. It is also known that the applica-tion of pressure can drive some of the parent compounds into the superconducting state without chemical doping[7]. Naively, the creation of a surface can be viewed asthe application of a uniaxial pressure. In this Letter, weexplore the effect on the coupling between spin, latticeand electrons in one of the parent compounds, BaFe As ,caused by the creation of a surface.Figure 1 shows the bulk and surface structure for theLTO phase of BaFe As . It consists of alternativelystacking Ba and Fe-As layers in bulk (Fig. 1a). TheLTO phase ( <
140 K) is a collinear AFM ordering withthe spin structure shown in Fig. 1a. We know fromour previous study [8] that the ordered exposed surfaceof BaFe As is the As plane and there is no measurablesurface reconstruction. If the As atoms were buckledvertically, it would be detected by low energy electrondiffraction (LEED) [9]. On the surface, the As atoms(blue) are in the first plane and the Fe atoms (red) in thesecond plane for the (1 ×
1) (001) surface unit cell withthe bulk orthorhombic structure (see Fig. 1b). As shownin Fig. 1b, all As atoms are expected to be identicalin the surface unit cell exhibiting C v symmetry. Themystery is that the high-resolution scanning tunnellingmicroscopy (STM) image measure of electronic densitydistribution, only reveals half of As atoms that should bepresent for a bulk truncated surface (Fig. 1c) [8]. Thissuggests that there are two inequivalent As sites on thesurface not seen in bulk. According to our STM work re-ported in this Letter, it is plausible that the two inequiv-alent As sites result from the underneath spin structurethrough strong orbital-spin coupling. Given the fact thatthe Fe moments are aligned antiferromagnetically alongthe longer a axis and ferromagnetically along the shorter b axis [4], the relationship between the “visible” As atomsand the spin structure is illustrated in Figs. 1d (As2) and FIG. 1: (a) Bulk lattice and spin structures of BaFe As withFe magnetic moments indicated by red arrows; (b) Schematicview of As terminated surface with underneath Fe layer; (c)As atoms seen by STM (solid circles), the empty circles repre-sent “invisible” As atoms; (d-e) Possible relationship between“visible” (As2 in (d), As1 in (e)) and the spin structure of Featoms
1e (As1). While we cannot determine which of these twoconfigurations has the lowest energy, As1 and As2 areclearly surrounded by different spin environments.For our STM investigation, we use high-qualityBaFe As single crystals that were grown using self-fluxmethod [12]. The measurements were conducted on ahome-built variable temperature STM with a tungstentip. Single crystalline BaFe As was firstly pre-cooledto 80 K in an ultra-high vacuum environment with ba-sic pressure lower than 5 × − Torr. After the in - situ cleavage, the sample was immediately inserted into thepre-cooled STM head. Fig. 2a displays a typical STMtopographic image with atomic resolution. In additionto the (1 ×
1) surface structure, there are white spots ei-ther forming zigzag lines (small and clear) or randomlydistributed (large and fuzzy). The large and fuzzy whitespots can be manipulated by the tip, which are likelyBa atoms as discussed previously [8]. Similarly, thereare dark spots, some of which are randomly distributedand the others are aligned with white spots in the zigzaglines. Importantly, there are always “dark” spots (see
FIG. 2: (a) A 355 ˚A ×
355 ˚A low-bias constant-current STMtopography (V bias = 23 mV, I tip = 200 pA) on (001) surfaceBaFe As at 80 K; (b) An enlarged 50 ˚A ×
50 ˚A topographyfrom the box in (a). The a - and b -axis are identified by FT-STM (see the text); (c) A 700 ˚A ×
700 ˚A high-bias constant-current STM topography (V bias = 483 mV, I tip = 200 pA).
Figs. 2a, 2c) (impurities/defects/“invisible” As sites)at every corner, wherever the zigzag lines change thedirection. Large-scale topographic images (not shown)prove that the zigzag lines form closed loops. Fig. 2bshows that the periodicity is the (1 ×
1) surface structureof orthorhombic bulk, except that we only see half ofthe As atoms in the surface plane. Through careful cal-ibration including possible thermal drift-induced error,piezo scanner asymmetry and hysteresis by using Fourier-Transform(FT)-STM (see the inset of Fig. 2b), we areable to identify a and b directions of the orthorhombicunit cell which are labelled in Fig. 2b.Given the scenario outlined above (strong orbital-spincoupling), we examine in detail the zigzag lines − domainwalls. In solids, the origin of domains can be geometric,magnetic or electronic. Using the FT-STM, we find thatall domains in the image have the same a and b direc-tions so there is no rotation of the lattice when crossinga domain wall. This allows us to exclude the possibilitythat the domains are caused by a structural misorien-tations in the bulk. What we see are surface electronicdomains, which can be verified by looking at the bias de-pendence. Fig. 2c shows the topography taken with ahigh positive bias voltage (483 mV). Compared to thattaken at 23 mV (Fig. 2a), the domain boundaries haveswitched from bright to dark. However, we do not see theswitching of “visible/invisible” As atoms when changingpolarity in the low bias region, as predict by the the-ory [11]. Because the atomic resolution is lost, whetherthe switching occurs at high bias is unknown. It shouldalso be mentioned that Fig. 2a can be reproduced afterchanging the bias voltage from 483 mV back to 23 mV.Therefore the features seen in Fig. 2c are not due to thechange of tip or sample condition. The domain bound-aries seen by STM are primarily electronic in origin.Figures 3a and 3b show two different domains with twoboundaries in each. There are two boundaries with onealong ∼ ◦ and another along ∼ -45 ◦ (Fig. 3a) or ∼ ◦ (Fig. 3b) with respect to a direction. As can be seenin either figure the change in the (1 ×
1) domains, whencrossing a boundary, is an inversion of dark to bright inAs, i.e., there is a half electronic unit cell shift acrossthe boundary lines as indicated by lines with arrows inFigs. 3a and 3b. Such an inversion would not occur if theboundary is created simply due to the crystal structuraldislocation with half unit cell shift. In our picture, thereis no structural change across this boundary and theboundary is actually a spin domain wall. While all thebright white spots residing on both boundaries have el-liptical shape, a closer examination reveals that the whitespots along the -45 ◦ direction are more rounded and theones in 45 ◦ direction more elongated. If one examinesthe symmetry carefully it is clear that the domains ex-hibit C symmetry. Rotating the image in Fig. 3b by180 ◦ transforms the 135 ◦ boundary into a -45 ◦ boundarywhich is identical to the -45 ◦ boundary in Fig. 3a, asexpected if the boundary direction is unchanged. If wefurther reflect the rotated image about the a axis, whatwas the -45 ◦ boundary becomes a 45 ◦ boundary, whichdoes not look like the original 45 ◦ boundary shown inFig. 3a. Thus, the boundary symmetry is C not C v This is consistent with the theoretical proposal that C symmetry is induced by the magnetic ordering [10].The line profile Z(x) in Figs. 3c and 3d presentsa quantitative comparison on the boundaries shown inFigs. 3a and 3b, respectively. While it is expected thatZ(x) oscillates with the same periodicity along both di-rections, the amplitude for the spots in 45 ◦ direction (redlines) is more than double compared to that along -45 ◦ di-rection (blue lines). This proves that there are two typesof boundaries, reflecting the fact that, when spin config-uration is included, the symmetry is reduced from C v to C . Fig. 3e demonstrates how the spin structure changesacross the boundary, assuming that the spin structure inFig. 1d produces bright As atoms. All of our obser-vations indicate that there is an electronic order at thesurface that reflects coupling between orbits and spins.Using this model, it is natural to explain the half elec-tronic unit cell shift between adjacent domains by adopt- ing a π phase shift of the spin order along both AFM a axis and FM b axis when crossing boundaries. Thespots seen at the boundaries are enhanced local densityof states due to the orbital overlap between two “visible”As atoms. As illustrated in Figs. 3f and 3g, the boundaryalong 45 ◦ direction results in brighter spots compared tothat along the 135 ◦ direction, when taking into accountof spin contribution. This is consistent with our experi-mental observation (Figs. 3c and 3d). The spin structureshown in Fig. 3e suggests that there is a spin flip acrossthe surface boundary creating an anti-phase (Pi-phase)spin domain wall.The puzzle is why the electronic topography seen withSTM mirrors the spin structure, considering that the Felayer looks like an itinerant metal. Recent theoretical[13–21] and experimental studies [6, 22, 23] on the elec-tronic structures of iron-based compounds suggest thatorbital degree of freedom emerges in this multiband sys-tem with intimate coupling to lattice, charge and spin.It was proposed that the ferro-orbital Fe (3d xz ) orderleads to the structural and magnetic phase transitions[13, 14]. As a result [16], electrical conduction is higherin the AFM a direction than that in the FM b direc-tion as observed experimentally [22]. The recent laserangle-resolved photoemission spectroscopy and band cal-culations [6] indicate that the two Fermi surface pocketscentered at Γ point ( α α
2) have a predominant Fe3d xz orbital component which is polarized by AFM order[18]. As argued in Ref. [24], most of the detected elec-tronic contribution by STM comes from the Γ centered α α As is quite different from the bulk. The keyis how this difference is reflected in the physical prop-erties at or near the surface, and how the surface mayaffect the bulk. We have focused on the As-terminated(1 ×
1) structure because it corresponds to the bulk or-thorhombic phase. But a (1 ×
2) (tetragonal notation)surface reconstruction [26–31] has also been observed andassociated with the tetragonal bulk structure exhibiting C v symmetry. In these materials, these two structuresseem to coexist at the surface [28–30] through out thephase diagram of Ba(Fe − x Co x )As [25]. Naively, thiswould indicate that there are patches of tetragonal sur-face structure coexisting with orthorhombic phase. What FIG. 3: (a-b): Two 56 ˚A ×
56 ˚A low-bias constant-currentSTM topographies (V bias = 23 mV, I tip = 200 pA) showingboundary structures at 80 K. The arrows with dash lines in(a-b) indicate the half electronic unit cell shift in a and b di-rection respectively. The black rectangular box in (b) showsthe size of bright spot on boundary: length/width = 3/2 mea-sured from this image; (c-d): the line profiles for red and blueline in (a) and (b) respectively; (e) A model for domain walls.Here the blue solid circles denote the “visible” As atoms inthe LTO surface unit cell, and the red arrows indicate themagnetic moments of Fe atoms; Solid yellow ellipses repre-sent the bright spots at the boundary; Two black solid circlesrepresent impurities/defects/“invisible” As sites. (f-g): Twotypes of boundaries breaking C v to C symmetry. The light-blue clouds denote the polarized Fe 3d xz orbitals. Red arrowsrepresent spins and their orientations. Positive and negativesigns in small circles indicate the phase of orientation. Notethat each ellipse along 45 ◦ direction (g) includes 4 negativesigns most close to white spot (the overlapped two As atoms)by the boundary; while each along 135 ◦ boundary directionincludes 4 positive signs (f). is needed is a measurement of orthorhombicity as a func-tion of doping and temperature. It may well be thatthe (1 ×
1) surface structure loses its orthorhombicity asa function of temperature or doping, eventually turnninginto a C ( √ ×√ ◦ surface reconstruction of a tetrag-onal bulk. We already know that the measured supercon-ducting gap using STM is well behaved as a function of x in both surface structures [25]. It is possible that thesurface “pins” or “freezes” the magnetic or orbital fluctu- ations resulting in a much higher structural and magnetictransition temperature than the bulk. If this is true, thesurface may be a nucleation center for the bulk phasetransition.In summary, the imaged domains and domain wallsseen by STM are shown to be primarily electronic inorigin with strong electron/spin (orbital-spin) coupling,which is clearly reflected by the C symmetry. The inti-mate coupling between the spin and electron orbitals atthe surface enable us to observe the electronic structurethat mirrors the bulk spin structure. This offers great op-portunities for the investigation on the orbital-spin cou-pling. It also opens a new chapter in the long-standingissues of interplay between superconductivity and mag-netism which may only be present at the surfaces (orunder pressure) of this new class of superconductors.We would like to thank W. Ku, V. B. Nascimento, P.Phillips, and D. Singh for fruitful discussion. Research atLSU is partially supported by NSF DMR-1002622 (GL,RJ, EWP). Research at ORNL is sponsored by BES Ma-terials Sciences and Engineering Division (ASS, MAM,BCS, DGM), U. S Department of Energy. [1] Y. Kamihara et al , J. Am. Chem. Soc. , 3296 (2008).[2] David C. Johnston et al , arXiv:1005.4392v1.[3] Andrei Mourachkine, High − TempeartureSuperconductivity in Cuprates (Springer, 2002).[4] M.D. Lumsden et al , J. Phys.: Condens. Matter et al , Phys. Rev. Lett. , 257003(2008).[6] T. Shimojima et al , Phys. Rev. Lett. , 057002(2010).[7] Simon A. J. Kimber et al , Nature Mater. , 471(2009).[8] V. B. Nascimento et al , Phys. Rev. Lett. ,076104(2009).[9] Our LEED system can probe vertical buckling ofBaFe As with the sensitivity of 0.05 ˚A.[10] J. Knolle et al , Phys. Rev. Lett. , 257001(2010); Si-mon A. J. Kimber et al , ArXiv: 1005.1761v1.[11] Miao Gao et al , ArXiv: 0909.5136v1.[12] A. S. Sefat et al , Phys. Rev. Lett. , 117004(2008).[13] Chi-Cheng Lee et al , Phys. Rev. Lett. , 267001(2009).[14] Weicheng Lv et al , Phys. Rev. B , 224506 (2009).[15] Weicheng Lv et al , ArXiv: 1002.3165v1.[16] C.-C. Chen et al , ArXiv: 1004.4611v1.[17] Z. P. Yin et al , Phys. Rev. B , 174534 (2010).[18] M. Daghofer et al , Phys. Rev. B , 180514(R) (2010).[19] C.-C.Chen et al , Phys. Rev. B ,180418(2009).[20] F. Kr¨uger et al , Phys. Rev. B , 054504(2009).[21] I. I. Mazin et al , Nat. Phys. , 141(2009).[22] Jiun-Haw Chu et al , ArXiv: 0911.3878v1.[23] A. Akrap et al , Phys.Rev. B , 180502(2009).[24] Weicheng Lee et al , Phys. Rev. Lett. , 176101(2009).[25] S. H. Pan et al , APS march meeting Bulletin (Portland,USA, 2010).[26] Yi Yin et al , Phys. Rev. Lett. , 097002(2009).[27] T. -M. Chuang et al , Science , 181(2010).[28] F. Massee et al , Phys. Rev. B , 140507(2009) [29] R. Jin et al , Supercond. Sci. Technol. , 054005(2010).[30] F. C. Niestemski, et al , arXiv:0906.2761. [31] Hui Zhang et al , Phys. Rev. B81