PRL 102, 097002 (2009)

PHYSICAL REVIEW LETTERS

week ending 6 MARCH 2009

Scanning Tunneling Spectroscopy and Vortex Imaging in the Iron Pnictide Superconductor BaFe1:8 Co0:2 As2
Yi Yin,1 M. Zech,1 T. L. Williams,1 X. F. Wang,2 G. Wu,2 X. H. Chen,2 and J. E. Hoffman1,*
Department of Physics, Harvard University, Cambridge, Massachusetts 02138, USA Hefei National Laboratory for Physical Science at Microscale and Department of Physics, University of Science and Technology of China, Hefei, Anhui 230026, People’s Republic of China (Received 29 October 2008; published 3 March 2009)
2 1

We present an atomic resolution scanning tunneling spectroscopy study of superconducting BaFe1:8 Co0:2 As2 single crystals in magnetic ﬁelds up to 9 T. At zero ﬁeld, a single gap with coherence " peaks at Á ¼ 6:25 meV is observed in the density of states. At 9 and 6 T, we image a disordered vortex lattice, consistent with isotropic, single ﬂux quantum vortices. Vortex locations are uncorrelated with strong-scattering surface impurities, demonstrating bulk pinning. The vortex-induced subgap density of states ﬁts an exponential decay from the vortex center, from which we extract a coherence length  ¼ # 27:6 Æ 2:9 A, corresponding to an upper critical ﬁeld Hc2 ¼ 43 T.
DOI: 10.1103/PhysRevLett.102.097002 PACS numbers: 74.25.Qt, 68.37.Ef, 74.25.Jb, 74.50.+r

The recently discovered iron-arsenic based superconductors (pnictides) break the monopoly that copper oxides (cuprates) have held over high temperature superconductivity for more than two decades [1–5]. There is new hope that investigation of the pnictides as a ‘‘foil’’ for cuprates will ﬁnally lead to a fundamental understanding of high-Tc superconductivity. An immediate challenge is therefore to catalog and understand the similarities and differences between cuprates and pnictides. Both materials become superconducting after chemical doping of antiferromagnetic parent compounds and both have dome-shaped phase diagrams. Spin-ﬂuctuation-mediated pairing, a promising mechanism for cuprate superconductivity, has also been proposed for the pnictides [6,7]. However, signiﬁcant differences are thought to exist between these two families of materials. For example, the parent compounds of pnictide superconductors are computed to be semimetals [8] instead of Mott insulators. Of more technological relevance, the pnictides are found to be more isotropic in a magnetic ﬁeld [9], which may facilitate application due to more effective pinning of quantized magnetic vortices [10]. A second challenge is therefore to experimentally characterize vortex pinning. To investigate the pnictides, both in comparison to the cuprates, and for applications requiring vortex pinning, we use a home-built cryogenic scanning tunneling microscope (STM) capable of tracking atomically resolved locations as the magnetic ﬁeld is swept up to 9 T. We choose to study optimally doped single crystal BaFe1:8 Co0:2 As2 . Since the Co dopants are incorporated into the strongly bound FeAs layer, the topmost layer is more like the bulk and remains more stable while tunneling. Our samples, grown with FeAs ﬂux to avoid contamination by other elements [11], show a sharp resistive transition at Tc ¼ 25:3 K with width ÁT ¼ 0:5 K. 0031-9007=09=102(9)=097002(4)

Figure 1 shows an atomic resolution topographic image of the surface of BaFe1:8 Co0:2 As2 , cleaved in situ at $25 K and recorded at 6.15 K in zero magnetic ﬁeld. From the Fourier transformation we ﬁnd an interatomic spacing consistent with x-ray diffraction measurements in either # the Ba or As layer, a ¼ 3:96 A [12]. The surface shows a complex stripelike structure in which alternate rows appear shifted with respect to each other. Similar stripes were observed in a previous STM study on Sr1Àx Kx Fe2 As2

FIG. 1 (color online). (a) A 20 Â 20 nm2 constant-current topographic image of the cleaved surface of BaFe1:8 Co0:2 As2 . # The interatomic spacing is a ¼ 3:96 A. (b) Zoom in (Â2 magniﬁcation) on a 5 Â 5 nm2 area within (a). The black circles denote the positions of the 2 Â 1 stripe structure. (c) Fourier transformation of (a). Peaks A and A? represent the atomic spacing between and within rows, respectively. Peaks B and C represent the alternating row characteristics: B shows intensity modulations attributable to height or density of states variations, while C shows in-plane shifts along the row direction. Both images are recorded using Vsample ¼ À20 mV and I ¼ 40 pA, at T ¼ 6:15 K in zero magnetic ﬁeld.

097002-1

Ó 2009 The American Physical Society

PRL 102, 097002 (2009)

PHYSICAL REVIEW LETTERS

week ending 6 MARCH 2009

[13]. The stripelike feature is oriented at 45 to the onedimensional spin order observed in the parent compound of these materials [14] and is therefore not likely related to this bulk order. In addition to the stripelike feature, there is another sparse, perpendicular, one-dimensional feature in the topographic image which is a possible dislocation to relieve surface strain. Neither feature has shown any effect on the local superconducting gap. We focus ﬁrst on electron pairing, the fundamental origin of superconductivity, which is characterized by the superconducting energy gap Á in the electronic density of states (DOS). We record differential conductance dI=dV as a function of sample-tip voltage V, which is proportional to ~ the local density of states gðr; eVÞ. For each dI=dV spectrum in a dense array of locations, we extract the magnitude of the energy gap Á, one half the distance between coherence peaks. A single energy gap Á $ 6 meV is observed in all spectra, independent of junction resistance. This gap energy is comparable to the nodeless gap on the À-centered Fermi surface ( FS) reported by angle resolved photoemission spectroscopy (ARPES) on similar electron doped BaFe1:8 Co0:2 As2 [15]. A map of the spatial dependence of Á [Fig. 2(b)] at zero magnetic ﬁeld and 6.25 K shows a total range of Á from 4.5 to 8.0 meV. From the $16 000 measured spectra, we ex" tract the average Á ¼ 6:25 meV and standard deviation "  ¼ 0:73 meV, leading to a fractional variation =Á of 12%. Spectra with smaller gap energies tend to exhibit

FIG. 2 (color online). (a) A series of dI=dV spectra taken along an 11 nm line, illustrating the shape of the superconducting gap and the low differential conductance at the Fermi energy. To reduce noise, each spectrum shown is the average of all ˚ spectra acquired within a 5 A radius. The vertical dashed lines are guides to the eye located at Æ6 meV. (b) A 20 Â 20 nm2 gap map, revealing the spatial variation of the gap magnitude Á. The " gap has Á ¼ 6:25 meV and fractional variation only 12%. A color-coded histogram of Á is shown below the gap map. The six spectra in (a) (from top to bottom) are taken at the locations of the black points indicated in (b) (from upper left to lower right). Data were acquired at T ¼ 6:25 K in zero magnetic ﬁeld.

weaker coherence peaks and higher zero bias conductance, as exempliﬁed in Fig. 2(a). This effect may be explained by impurity scattering in unconventional superconductors " [16]. In Bi2 Sr2 CaCu2 O8þx (Bi2212), gaps with Á $ " $ 21%) are typically re33 meV and  $ 7 meV (=Á ported [17]. In Bi-based cuprates, the larger intrinsic electronic inhomogeneity and the observed (opposite) relation between coherence peak height and gap width may be complicated by the coexistence of a pseudogap at nearby energy [18,19]. From the average of the gap map we calculate the " reduced gap 2Á=kB Tc ¼ 5:73. This exceeds the values for weak-coupling s-wave or d-wave BCS superconductors, which are 3.5 and 4.3, respectively [20,21]. Experiments which decouple the pseudogap by coherent tunneling [22] or normalization [18] ﬁnd reduced gap values in the cuprates between 6 and 10. Our result suggests that although pnictides are in a strong coupling regime, they are not as strongly coupled as cuprates. While bulk studies have indicated high critical ﬁelds in the pnictides [23], single vortex imaging gives the most stringent bounds on vortex pinning [24]. Static vortices have been imaged magnetically by scanning SQUID microscopy at 33 mG in the related compound NdFeAsO0:94 F0:06 [25], but it remains important to understand pinning at the higher ﬁelds which will be of interest for applications such as superconducting magnets. We image vortices electronically by mapping the conductance at an energy where a vortex alters the density of states. Figure 3(b) shows such a conductance map, recorded in a 9 T magnetic ﬁeld at an energy corresponding to the ﬁlled state coherence peak. A second conductance map at the same location, recorded at 6 T, is shown in Fig. 3(c). In both maps, the vortices appear as broad areas of depressed conductance. Impurities, possibly single Fe or Co vacancies, appear as sharper minima in the conductance, also visible as white spots in the simultaneously recorded topography in Fig. 3(a) (at 9 T). As in YBa2 Cu3 O7Àx (Y123) and Bi2212, the vortex lattice is disordered [26,27]. We ﬁnd no correlation between the locations of the vortices and the visible impurities [28]; therefore bulk pinning must play a dominant role in this material. These observations are not compatible with pancake vortices [29], which would be pinned by the surface impurities. The isotropy of the Fourier transformation of the vortex lattice also indicates that vortex locations are not signiﬁcantly affected by the stripelike surface feature, or by any residual one-dimensional spin or structural order [14] within the bulk. After ﬁltering out the impurities, a peak-ﬁtting algorithm was used to extract the vortex center locations from the conductance maps. Voronoi cells [27] were overlaid onto the vortex image [Figs. 3(d) and 3(e)] and the " average per-vortex ﬂux  was calculated from the Voronoi " cell size. We observe ð9 TÞ ¼ 2:05 Â 10À15 T m2 and

097002-2

PRL 102, 097002 (2009)

PHYSICAL REVIEW LETTERS

week ending 6 MARCH 2009

FIG. 3 (color online). (a) A 100 Â 100 nm2 topographic image recorded in 9 T magnetic ﬁeld with Vsample ¼ À5 mV and I ¼ 10 pA. The bright spots are impurities. (b) Differential conductance map recorded simultaneously with the topographic image in (a), revealing the sample DOS at 5 meV in 9 T magnetic ﬁeld at T ¼ 6:15 K. Vortices are visible as dark features due to the local suppression of the coherence peaks. The vortices form a disordered lattice. Impurities, also visible in (a), appear as smaller, sharper dark features. (c) Differential conductance map recorded at 6 T in the same ﬁeld of view and with the same parameters as (b), except T ¼ 6:21 K. (d),(e) A Voronoi overlay on impurity-ﬁltered data emphasizes the vortices and demonstrates the procedure used to compute the average area associated with each vortex.

FIG. 4 (color online). (a) A 40 Â 40 nm2 map of the zero bias conductance in 9 T magnetic ﬁeld at T ¼ 6:15 K, showing $8 vortices (broader, lighter objects) and 8 strong-scattering impurities (sharper, brighter objects). (b) A series of spectra along a 14.5 nm trajectory through one of the vortices, indicated by the red or gray line in (a). The superconducting gap is completely extinct inside the vortex core; only the V-shaped background remains. The three thick red or gray lines emphasize the spectra at the vortex core and far from it on both sides. The 75 spectra shown are offset by a total of 3 nS for clarity. (c) Azimuthally averaged radial dependence of the differential conductance gðr; 0Þ around a single vortex as measured (black squares) and ﬁt by an exponential decay (red or gray line). The constant background g1 has been removed in order to emphasize the exponential decay on the logarithmic scale of the y axis. The # exponential ﬁt leads to a coherence length of  ¼ 30:8 Æ 1:4 A # for this vortex and to an average of  ¼ 27:6 A with standard ˚ deviation 2.9 A for all vortices investigated.

" ð6 TÞ ¼ 2:17 Â 10À15 T m2 , in good agreement with the single magnetic ﬂux quantum, È0 ¼ 2:07 Â 10À15 Tm2 . We gain insight into a phenomenon by probing the conditions under which it breaks down: local measurements in high magnetic ﬁeld allow us to observe the destruction of superconductivity in vortex cores. Figure 4(a) shows the zero bias conductance (ZBC), corresponding to the Fermi level density of states, in a region of Fig. 3(b). Here the vortices appear as enhanced subgap density of states, in

contrast to Figs. 3(b)–3(e), where the vortices appear as depressed conductance due to the local suppression of the coherence peaks. Figure 4(b) shows a series of spectra on a trajectory through one of the vortices in 9 T magnetic ﬁeld. No ZBC peaks, which would be a signature of quasiparticle bound states at energies 1 Á2 =EF [30], are observed within 2 the vortex core as would be expected in a conventional s-wave superconductor; only the V-shaped background remains. In Y123 [26] and Bi2212 [27], particle-hole symmetric subgap peaks have been reported within vortex cores, with energies approximately ÆÁ=4. We do not observe such particle-hole symmetric subgap states, although the corresponding low energy scale in this material, Á=4 $ 1:5 meV would likely appear as a weak ZBC peak as well. Our observed pnictide vortices are isotropic and lacking internal structure, in contrast to predictions of d-wave vortices [31] and observations of fourfold symmetric internal vortex structure in d-wave cuprates [32].

097002-3

PRL 102, 097002 (2009)

PHYSICAL REVIEW LETTERS
[2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24]

week ending 6 MARCH 2009

To extract the superconducting coherence length , we use the azimuthally averaged radial dependence of the vortex-induced ZBC. We ﬁt an exponential decay gðr; 0Þ ¼ ˚ g1 þ A expðÀr=Þ within a distance of r ¼ 20 to 100 A from the vortex core, as shown in Fig. 4(c). From several # vortices, we ﬁnd an average coherence length  ¼ 27:6 A ˚ with standard deviation 2.9 A. Using the Ginzburg-Landau expression Hc2 ¼ 0 =22 , we compute the upper critical ﬁeld Hc2 ¼ 43 T. With residual resistivity 0 ¼ 0:23 m cm and Hall coefﬁcient RH ¼ 11 Â 10À9 m3 =C, the electronic mean ˚ free path ‘ ¼ @ð32 Þ1=3 =e2 n2=3 0 is approximately 81 A. # is almost Given that our coherence length  ¼ 27:6 A 3 times smaller than ‘, the superconductor under investigation is not in the dirty regime in which the ZBC peak would be suppressed due to signiﬁcant scattering. In contrast to the cuprates, where substitutions into the critical superconducting CuO2 plane lead to strong scattering and the reduction of Tc [33,34], the 10% of Co atoms doped directly into the superconducting FeAs plane in this material do not provide a strong scattering signature. Strong scatterers are dilute in this material, corresponding to only 0.029% of the Fe sites, leading to an average # spacing comparable to ‘ $ 81 A. To summarize, we have presented the ﬁrst atomically resolved STM/STS studies on a single crystal pnictide superconductor in magnetic ﬁelds up to 9 T. We observe " a single gap with Á ¼ 6:25 meV and relative standard deviation 12% in zero magnetic ﬁeld. We have imaged a stationary but disordered vortex lattice at 6 and 9 T, uncorrelated with the locations of surface impurities. This demonstrates that vortices experience bulk pinning at ﬁelds up to 9 T, even in a clean single crystal pnictide superconductor with Tc only 25.3 K. The nonobservation of subgap peaks within the vortex cores contrasts with observations in both d-wave cuprates and conventional s-wave superconductors. We acknowledge Eric Hudson, Subir Sachdev, and Ophir Auslaender for valuable discussions. This work is supported by the NSF (DMR-0508812) and the AFOSR (FA9550-05-1-0371). T. L. W. acknowledges support from the NDSEG.

[25] [26] [27] [28]

[29] [30] [31] [32] [33] [34]

*jhoffman@physics.harvard.edu [1] Y. Kamihara, T. Watanabe, M. Hirano, and H. Hosono, J. Am. Chem. Soc. 130, 3296 (2008).

X. H. Chen et al., Nature (London) 453, 761 (2008). G. F. Chen et al., Phys. Rev. Lett. 100, 247002 (2008). Z.-A. Ren et al., Europhys. Lett. 83, 17002 (2008). M. Rotter, M. Tegel, and D. Johrendt, Phys. Rev. Lett. 101, 107006 (2008). I. I. Mazin, D. J. Singh, M. D. Johannes, and M. H. Du, Phys. Rev. Lett. 101, 057003 (2008). F. Wang, H. Zhai, Y. Ran, A. Vishwanath, and D.-H. Lee, Phys. Rev. Lett. 102, 047005 (2009). D. J. Singh and M. H. Du, Phys. Rev. Lett. 100, 237003 (2008). N. Ni et al., Phys. Rev. B 78, 014507 (2008). M. R. Norman, Physics 1, 21 (2008). X. F. Wang et al., arXiv:0806.2452 [Phys. Rev. Lett. (to be published)]. A. S. Sefat et al., Phys. Rev. Lett. 101, 117004 (2008). M. C. Boyer et al., arXiv:0806.4400. C. de la Cruz et al., Nature (London) 453, 899 (2008). K. Terashima et al., arXiv:0812.3704. G. Preosti, H. Kim, and P. Muzikar, Phys. Rev. B 50, 1259 (1994). K. McElroy et al., Phys. Rev. Lett. 94, 197005 (2005). M. C. Boyer et al., Nature Phys. 3, 802 (2007). J.-H. Ma et al., Phys. Rev. Lett. 101, 207002 (2008). T. Y. Chen, Z. Tesanovic, R. H. Liu, X. H. Chen, and C. L. Chien, Nature (London) 453, 1224 (2008). H. Won and K. Maki, Phys. Rev. B 49, 1397 (1994). G. Deutscher, Rev. Mod. Phys. 77, 109 (2005). A. Yamamoto et al., Appl. Phys. Lett. 94, 062511 (2009). E. W. J. Straver, J. E. Hoffman, O. M. Auslaender, D. Rugar, and K. A. Moler, Appl. Phys. Lett. 93, 172514 (2008). C. W. Hicks et al., J. Phys. Soc. Jpn. 78, 013708 (2009). I. Maggio-Aprile, C. Renner, A. Erb, E. Walker, and O. Fischer, Phys. Rev. Lett. 75, 2754 (1995). S. H. Pan et al., Phys. Rev. Lett. 85, 1536 (2000). At 9 T, 10.6% of the total ﬁeld of view is covered by vortex cores, and 4 out of 49 impurities lie within these cores. The difference between 10.6% and 4 out of 49 (8.2%) is not statistically signiﬁcant. A. Grigorenko, S. Bending, T. Tamegai, S. Ooi, and M. Henini, Nature (London) 414, 728 (2001). C. Caroli, P. G. de Gennes, and J. Matricon, Phys. Lett. 9, 307 (1964). M. Franz, C. Kallin, P. I. Soininen, A. J. Berlinsky, and A. L. Fetter, Phys. Rev. B 53, 5795 (1996). J. E. Hoffman et al., Science 295, 466 (2002). E. W. Hudson et al., Nature (London) 411, 920 (2001). J. M. Tarascon et al., Phys. Rev. B 36, 8393 (1987).

097002-4