VOLUME 76, NUMBER 22 P H Y S I C A L R E V I E W L E T T E R S 27 MAY 1996 Atomic-Scale Observations of Alloying at the Cr-Fe(001) Interface A. Davies, Joseph A. Stroscio, D. T. Pierce, and R. J. Celotta Electron Physics Group, National Institute of Standards and Technology, Gaithersburg, Maryland 20899 (Received 13 December 1995) While much progress has been made using epitaxial growth of Fe Cr Fe structures to study magnetic exchange coupling, a number of anomalies have arisen in studies of this model system. Using scanning tunneling microscopy and spectroscopy to investigate Cr growth on Fe(001), we have identified a potential structural cause of these anomalies. We show that Cr growth under layer-by-layer conditions on Fe(001) leads to the formation of a Cr-Fe alloy. We exploit a Cr and Fe surface state to identify single Cr impurities in Fe and evaluate the alloy concentrations with increasing Cr coverage. [S0031-9007(96)00285-2] PACS numbers: 61.16.Ch, 68.55.­a, 75.70.Cn Significant progress has been made in our understand- louin light scattering (BLS) [10] measurements observe ing of the phenomena of exchange coupling between ferro- a ferromagnetic alignment between the two Fe layers at magnetic layers separated by a nonmagnetic layer. In part, even Cr layer thickness where oscillatory coupling is ob- this is due to the use of Fe Cr Fe as a model system where, served below 20 monolayers. under the right conditions, structures produced experimen- In SEMPA measurements of bare Cr on Fe(001), the tally can closely approximate those presumed in theoretical surface magnetization does not show reversals indicative calculations [1]. This is advantageous because of the great of antiferromagnetic ordering until the third or fourth influence physical structure has on magnetic properties [2] layer [1]. Such a delayed onset of clear antiferromagnetic coupled with the general difficulty in growing perfect tran- ordering is also observed in inelastic polarized electron sition metal structures. High quality Fe Cr Fe structures scattering measurements, but unlike the SEMPA or BLS are possible because of the excellent Fe-Cr lattice match measurements, odd layers of Cr are aligned antiparallel to (within 0.6%) and the availability of near perfect, single the Fe substrate magnetization [11]. crystal Fe whisker substrates [3]. Such Fe Cr Fe struc- Finally, estimates of the average Cr moment for a tures have revealed more than 70 alignment reversals in the monolayer of Cr on Fe(001) from a variety of experiments coupling of Fe layers with increasing Cr thickness, allow- range from nonexistent [12] to values near that of the bulk ing the experimental confirmation of the calculated cou- [8] to 3 times the bulk value [13]. pling oscillation period to 0.5% accuracy [1]. Thus, there are conflicting or unexplained results in four Recent progress not withstanding, a number of impor- areas: indications of a surprisingly high Cr moment for the tant anomalies in the magnetic properties of the Cr Fe and first 0.1 ML, the unexpected phase of the antiferromag- Fe Cr Fe systems have arisen, specifically with regard to netic ordering of the Cr layers, the delay in the onset of the change in surface magnetization at low Cr coverages, this ordering, and a lack of consistency in measurements the antiferromagnetic ordering of the Cr layers, and the of the average magnetic moment of a Cr monolayer on Fe. size of the Cr moment. Scanning electron microscopy We have investigated a structural cause of these anoma- with polarization analysis (SEMPA) measurements of the lies by carrying out scanning tunneling microscopy (STM) average magnetization of the topmost surface layers of studies of Cr growth on Fe(001). We show that, in contrast Cr Fe(001) show a dramatic decrease during the deposi- to the assumed formation of a chemically abrupt interface, tion of the first 0.1 monolayer (ML) of Cr [4]. In recent layer-by-layer growth at 300 ±C leads to the formation of alternating gradient magnetometer measurements [5] a a Cr-Fe alloy that is observed as a distribution of single similar initial rapid decrease in average surface moment atomic Cr impurities dispersed in the Fe substrate in the is observed, from which it was deduced that the moment submonolayer-coverage regime. In contrast to other STM of the initially deposited Cr atoms is 4.5mB [5]. This is studies of surface alloys where the source of the image more than 7 times the bulk moment of 0.59mB, and even contrast is unknown [14], we use tunneling spectroscopy to more than the enhanced values of 3.1mB [6] and 3.6mB identify the density-of-states variations which lead to the [7] calculated for a Cr surface. atomic-scale chemical contrast. A surface state, seen in Photoemission measurements of the Cr Fe(001) inter- conductance spectra on Fe(001) and Cr(001) [15], is used face demonstrate that the first layer of Cr moments align to interpret spectra on the Cr-Fe alloyed surface, leading antiparallel to the Fe substrate magnetization [8,9]. With to a clear chemical identification in the submonolayer Cr antiferromagnetic stacking of the Cr layers and antiparal- coverage regime and an estimate of the chemical compo- lel coupling between Cr and Fe, two Fe layers separated sition as the alloy evolves with increasing Cr coverage. by an odd number of Cr layers would be expected to align The experiments were performed in an ultrahigh ferromagnetically. In contrast, both SEMPA [1] and Bril- vacuum system with facilities for thin film growth and 4175 VOLUME 76, NUMBER 22 P H Y S I C A L R E V I E W L E T T E R S 27 MAY 1996 surface characterization by STM and reflection high- curves). When the tip is away from an impurity (over the energy electron diffraction, as described previously [16]. smooth grey regions in the images) we see a strong conduc- Cr was deposited at a rate of approximately 0.8 ML min21. tance peak at the voltage corresponding to the Fe surface Figure 1(a) shows an image of the surface after deposit- state. When the tip is over an impurity atom, the Fe sur- ing 0.4 ML of Cr at a substrate temperature of 290 6 10 ±C face state peak is attenuated and a weaker broad peak ap- [17]. The layer-by-layer quality of the deposition is clear pears below the Fermi level. The latter conductance peak from the presence of only single atomic-step islands (the leads to the positive height contrast of the impurities when lighter grey regions in the image). The area of the islands imaging the filled states of the sample (see Fig. 1). These provides a calibration of the Cr coverage. spectral features were reproducibly observed in numerous A closer look at the surface reveals small-scale features measurements with different W(111) tips. The two types on both the substrate and island levels as shown in of spectra are seen on both the exposed regions of the sub- Fig. 1(b). The dark features also appear in STM images strate and the islands for submonolayer coverages, indicat- of clean Fe whiskers and we believe these are due to ing that the smooth grey regions in the images are Fe, on contamination from residual gases in the chamber. The both the exposed substrate and island levels, and that the small white features on both the substrate and island levels impurity atoms must therefore be Cr. This implies that are only seen after depositing Cr. These features are some of the deposited Cr atoms have replaced Fe atoms, identical, appearing rotationally symmetric with atomic- resulting in a growth layer in the low coverage limit that scale widths as small as 0.5 nm (FWHM). When imaging contains mostly Fe instead of pure Cr (see the schematic the filled (empty) states of the sample, the features have a in Fig. 2). maximum apparent positive (negative) height contrast of In the low-coverage regime where the individual Cr approximately 0.01 nm. The maximum height contrast of atoms can be resolved, the spatial correlation can be evalu- these features compared to the atomic-step islands can be ated from the Cr-pair distribution function shown in appreciated from the rendered image shown in Fig. 1(c). Fig. 3(b). The coordinates of each Cr impurity are deter- The height of the atomic-step islands corresponds to the mined from the image in Fig. 3(a). The relative position interplanar separation of 0.14 nm for both Cr and Fe of each Cr pair is found and plotted in Fig. 3(b). Where along the [001] direction. The size, shape, and voltage multiple pairs have the same relative position (within some dependence of the dotlike features lead us to identify each bin size) the area of the symbol is increased to reflect the feature as a single substitutional impurity atom in the number of such pairs. The relative lattice sites correspond- surface layer. ing to first, second, and third nearest neighbors [18] are We use tunneling spectroscopy [15] to chemically iden- labeled in the figure. Surprisingly, we find no occurrences tify the atoms in the surface alloy by taking advantage of a of first nearest-neighbor Cr pairs. The occupation proba- bcc (001) surface state that lies near the Fermi energy for bility for second nearest neighbors is 0.038 6 0.007 ML many transition metals. This surface state leads to a strong which is only 0.6 6 0.2 times the value expected for a and narrow conductance peak at a sample bias of 10.17 V random distribution 0.059 6 0.003 ML . These obser- for Fe(001) and 20.05 V for Cr(001), as shown by the vations, particularly the absence of first nearest-neighbor dashed curves in Fig. 2 [15]. Spectra taken on the Cr-Fe Cr pairs, can be appreciated by comparing the impurity alloyed surface are also shown in the figure (the solid-line features in Fig. 3(a) to simulated images of near-neighbor FIG. 1. STM images of Cr growth on Fe(001) at a sample bias of 21.1 V. (a) Large area scan showing the layer-by-layer quality of the growth of 0.4 ML Cr deposited at 290 6 10 ±C. (b) Small area scan of the surface shown in (a). The grey scale has been used twice through the height range of the image, each range covering approximately 0.1 nm. The island levels are surrounded by the thick black lines and the Fe substrate level by the thin white line. (c) Rendered perspective of the image in (b). 4176 VOLUME 76, NUMBER 22 P H Y S I C A L R E V I E W L E T T E R S 27 MAY 1996 FIG. 2. Tunneling conductance versus sample bias voltage. The top two dashed curves are representative tunnel- ing conductance spectra from clean Cr(001) and Fe(001) surfaces [15]. These spectra have been offset and scaled for comparison to the bottom spectra. (The scale of these spectra is lower because they were taken at a larger tip-sample sepa- ration distance.) The solid-line spectra are repre- sentative of spectra taken on the submonolayer Cr Fe(001) alloyed surface. A model of the Cr-Fe alloy is shown in the lower part of the figure. impurities shown in the inset. A suppression of nearest- neighbor occupation is indicative of an effective repulsive interaction between the Cr impurities, which is an inter- esting contrast to many other surface alloys studies where clustering is observed [19]. The variation of the Cr impurity concentration with submonolayer Cr coverage is shown in Fig. 4 for growth at 300 ±C. The Cr concentrations on the exposed regions FIG. 3. (a) High-resolution STM image of the exposed sub- of the substrate (islands) are indicated by empty (filled) strate level for submonolayer Cr coverage showing the alloyed Cr impurity atoms. Simulated images of Cr pairs separated by circles. The Cr concentration on the islands could only first, second, third, and fourth nearest-neighbor (nn) distances be clearly determined for low coverages, and we find are shown in the inset for comparison. The edges of both the no significant difference between the substrate and island real and simulated image are approximately along 110 direc- concentrations in this limit. The initial slope in Fig. 3(b) tions so the orientation of features can be directly compared. is approximately 0.25, indicating that only one out of every (b) Plot of the distribution of relative Cr-pair coordinates corre- sponding to the impurity distribution in part (a). The axes are four deposited Cr atoms remains in the surface layers. along 100 directions, which is rotated relative to the orienta- Beyond a coverage of 0.2 ML, the Cr concentration of tion in part (a). The area of each symbol is proportional to the islands is difficult to evaluate, but qualitatively appears to number of Cr pairs with relative coordinates x 6 d 100 and increase [see Fig. 1(b)]. For coverages from 0.2 to 1 y 6 d 010 , using a bin size of d a 32 where a 0.29 nm ML, the Cr accumulation rate in the exposed substrate layer is the in-plane lattice constant. The scatter in the plot is due to the uncertainty 60.07 nm in identifying the relative coordi- decreases, and consequently the Cr concentration in these nates of each impurity pair. Regions corresponding to the first, regions approaches a constant value of approximately second, and third nearest-neighbor separations are indicated in 0.10 ML. the figure. Beyond a Cr coverage of 1 ML, the Cr concentration in the topmost surface layer appears to continue increasing. The images still have an alloyed appearance at coverages of the peak maxima fall at the Cr surface state voltage of of 2­3 ML with all conductance spectra still showing a 20.05 V, instead of 20.3 V corresponding to the single peak near the Fermi energy. Now, however, the majority impurity spectra (see Fig. 2), and no peak maxima are 4177 VOLUME 76, NUMBER 22 P H Y S I C A L R E V I E W L E T T E R S 27 MAY 1996 [1] J. Unguris, D. T. Pierce, R. J. Celotta, and J. A. Stroscio, in Magnetism and Structure in Systems of Reduced Dimension, edited by R. F. C. Farrow et al. (Plenum, New York, 1993), p. 101; J. Unguris, R. J. Celotta, and D. T. Pierce, Phys. Rev. Lett. 67, 140 (1991). [2] D. T. Pierce, J. A. Stroscio, J. Unguris, and R. J. Celotta, Phys. Rev. B 49, 14 564 (1994). [3] A. S. Arrott, B. Heinrich, and S. T. Purcell, in Kinetics of Ordering and Growth at Surfaces, edited by M. G. Lagally (Plenum, New York, 1990), p. 321. [4] D. T. Pierce, R. J. Celotta, and J. Unguris, J. Appl. Phys. 73, 6201 (1993); J. Unguris, R. J. Celotta, and D. T. Pierce, Phys. Rev. Lett. 69, 1125 (1992). [5] C. Turtur and G. Bayreuther, Phys. Rev. Lett. 72, 1557 FIG. 4. Plot of the Cr concentration in the surface layers versus Cr deposition for the 300 ±C growth condition. The (1994). empty and filled circles represent the concentrations on the [6] C. L. Fu, A. J. Freeman, and T. Oguchi, Phys. Rev. Lett. substrate and the first growth layer, respectively. 54, 2700 (1985). [7] R. H. Victora and L. M. Falicov, Phys. Rev. B 31, 7335 (1985). [8] F. U. Hillebrecht, Ch. Roth, R. Jungblut, E. Kisker, and A. Bringer, Europhys. Lett. 19, 711 (1992). observed at the Fe surface state voltage. We therefore infer [9] P. D. Johnson, N. B. Brookes, and Y. Chang, Mater. Res. that the surface is predominantly Cr at these coverages. Soc. Symp. Proc. 231, 49 (1992). To begin to understand the implications of the inter- [10] B. Heinrich, Z. Celinski, J. F. Cochran, and M. From, facial alloy on the magnetic properties, we must know in Magnetism and Structure in Systems of Reduced the final Cr-Fe concentration profile across the interface. Dimensions, edited by R. F. C. Farrow, B. Dieny, M. We can only evaluate the surface concentrations from the Donath, A. Fert, and B. D. Hermsmeir, NATO Advanced STM data. By comparing surface concentrations to the Study Institute Ser. B, Vol. 309 (Plenum, New York, Cr deposition, we know that growth at 300 ±C leads to 1993), p. 101. significant Cr interdiffusion below the surface. Very re- [11] T. G. Walker, A. W. Pang, and H. Hopster, Phys. Rev. cent angle-resolved Auger electron spectroscopy experi- Lett. 69, 1121 (1992). [12] M. Donath, D. Scholl, D. Mauri, and E. Kay, Phys. Rev. ments on this system show Cr as deep as the third layer B 43, 13 164 (1991). below the surface for submonolayer growth at 300 ±C [13] P. Fuchs, K. Totland, and M. Landolt, in Proceedings [20]. The tunneling spectroscopy measurements suggest of the 14th International Colloquium on Magnetic Films the first predominantly-Cr layer occurs at a Cr coverage of and Surfaces, E-MRS Symposium on Magnetic Ultrathin no more than 2­3 ML. Films and Multilayers and Surfaces, Dusseldorf, Germany, A diffuse rather than chemically abrupt Cr Fe inter- 1994 (Elsevier, Amsterdam, 1995). face could account for many of the magnetic anomalies [14] See J. L. Stevens and R. Q. Hwang, Phys. Rev. Lett. 74, in Cr Fe systems. The phase of the antiferromagnetic or- 2078 (1995), and references therein. dering of Cr growth on Fe(001) [4] as well as the suppres- [15] J. A. Stroscio, D. T. Pierce, A. Davies, R. J. Celotta, and sion of clear antiferromagnetic ordering up to the third or M. Weinert, Phys. Rev. Lett. 75, 2960 (1995). fourth layer may result from the spatially varying Cr con- [16] J. A. Stroscio, D. T. Pierce, and R. A. Dragoset, Phys. Rev. Lett. 70, 3615 (1993). centration and the uncertainty in the position of an effec- [17] All errors reported in this paper represent 1 standard tive Cr Fe interface. The large Cr moments deduced from deviation and include both statistical and systematic the rapid decrease in the net magnetization in depositing Cr errors. on Fe, as observed in SEMPA [4], and other measurements [18] Nearest-neighbor sites in this paper refer to the surface [5], may be a consequence of the changes in Cr and neigh- 1 3 1 lattice. The surface first and second nearest boring Fe moments in the dilute Cr-Fe alloy [21]. Ex- neighbors corresponds to the second and third nearest perimental determinations of Cr moments which assume a neighbors in the bulk, respectively. chemically abrupt interface should be reinterpreted given [19] L. P. Nielson, F. Besenbacher, I. Stensgaard, E. Lægs- the Cr-Fe alloying shown here. We believe much of the gaard, C. Engdahl, P. Stoltze, K. W. Jacobson, and J. K. variation in experimental estimates of the Cr moment in Nørskov, Phys. Rev. Lett. 71, 754 (1993); H. Röder, the first Cr layer is due in part to varying degrees of inter- R. Schuster, H. Brune, and K. Kern, Phys. Rev. Lett. 71, facial alloying. 2086 (1993). [20] D. Venus and B. Heinrich, Phys. Rev. B 53, R1733 (1996). We would like to thank M. D. Stiles for many helpful [21] V. N. Gittsovich, V. G. Semenov, and V. M. Uzdin, discussions and A. Zangwill for the suggestion that J. Magn. Magn. Mater. 146, 165 (1995). interfacial alloying may be operative in the Cr Fe system. This work was supported in part by the Office of Naval Research. 4178