PHYSICAL REVIEW B VOLUME 59, NUMBER 16 15 APRIL 1999-II Surface alloy formation of Fe on Cr 100... studied by scanning tunneling microscopy Y. J. Choi, I. C. Jeong,* J.-Y. Park, S.-J. Kahng, J. Lee, and Y. Kuk National Creative Research Initiatives, Center for Science in Nanometer Scale and Department of Physics, Seoul National University, Seoul 151-742, Korea Received 2 March 1998; revised manuscript received 28 October 1998 Surface alloy formation of Fe on a Cr 100 surface was studied using a UHV scanning tunneling micro- scope. As we deposited Fe of less than 1 ML at room temperature and subsequently annealed the substrate at temperatures between 200 °C and 300 °C, we observed that Fe atoms incorporate into the Cr 100 surface, forming a well-ordered surface alloy of Fe0.5Cr0.5. These two elements were differentiated by imaging at the bias voltages near their surface states. The surface alloying was also confirmed by imaging at the bias voltages of their Shockley image states. By annealing the Fe layer of 1 ML at 250 °C, we found that the solubility of Fe into the Cr 100 surface is 25% at that temperature. Fe overlayer shows kinetically roughened mound structure at room temperature while it shows equilibrium two-dimensional islands at 300 °C. S0163-1829 99 06115-9 INTRODUCTION magnetoresistance GMR sensor as a form of a Fe/Cr/Fe multilayer.16 Some magnetic anomalies, such as a surpris- In the last decade, there has been great interest in the ingly high magnetic moment of Cr at 0.1 ML,17 the unex- study of thin metal films to understand their versatile chemi- pected phase of the antiferromagnetic ordering in the Cr cal, electronic, and magnetic properties. It has been known layer, and the delayed onset of the antiferromagnetic order- that the structure of a film interface plays an important role ing in the Fe/Cr/Fe sandwich structure,18 were explained in defining its property. Many metallic films are known to with interface alloying between the Cr overlayer and the Fe grow in three different modes, depending on lattice mis- substrate at the growth temperature.19 Although Cr can dis- match and lattice structure: the Frank van der Merwe mode solve into the Fe matrix only up to 3% at 300 °C in the layer-by-layer growth , the Volmer-Weber mode three di- bulk-phase diagram,20 it was reported that the solubility of mensional 3D island growth , and the Stranski Krastanov Cr in Fe at the interface is higher than that in the bulk.19 mode layer growth, followed by 3D growth . These three There have been many studies on the Cr film on a single modes are classified under the assumption of negligible solu- crystal Fe 100 substrate, but there have been few studies on bility between the overlayer and the substrate.1 But, in many the opposite case, Fe film on a single-crystal Cr 100 sub- other cases of metallic films, their interface structures are not strate, because of the difficulty in cleaning a Cr substrate.21 abrupt because of the intermixing between the overlayers Although the bulk-phase diagram shows very similar solubil- and the substrates, although they are immiscible or their ity in Fe rich and Cr rich dilute alloys at 300 °C, the surface solubilities are quite low in their binary bulk-phase alloying was predicted and confirmed only when Cr is grown diagrams.2­9 Recently, scanning tunneling microscope on Fe substrate. Despite this prediction, there are some cases STM and total-energy calculation have revealed that metal- where the surface alloying occurs in both dilute alloys.22 on-metal growth modes are much more complex than three In this paper, we show that the surface-alloy formation growth modes.2,4,6,7,10 Especially using STM, the initial can also occur at the low Fe coverage on the Cr 100 surface stages of alloy formation at the atomic level could be ob- after Fe is deposited at room temperature and subsequently served directly in a number of metal-on-metal annealed the sample at temperatures between 200 °C and systems.2­8,11­15 300 °C. Incorporated Fe atoms are identified by the differ- STM, with scanning tunneling spectroscopy, can be uti- ence in tunneling currents at the bias voltage of the Cr sur- lized to identify alloy species during the formation process, face state and at that of the Fe surface state. The alloyed area as well as to image the surface geometrical structure. In the can also be differentiated by measuring the change of the cases where the lattice mismatch between two materials is work function at the alloyed region from that at a bare Cr somewhat large or their valence band structures are different, surface. In addition to the alloy formation at the low Fe each species in the alloyed phase could be identified with coverage, we observed that layer-by-layer growth of Fe lay- apparent difference in the STM images.2­7,11­15 However, in ers can be achieved at the Fe coverage of 1 ML by anneal- cases where the lattice mismatch between two materials is ing the sample at 250 °C. not large, it is difficult to identify two species from the simple STM images. Recently, Davies et al. reported that EXPERIMENT different chemical species could be identified with the local- ized surface states of the two elements in the Cr on Fe 001 A clean Cr surface was prepared by repeated cycles of Ar system, where the lattice mismatch is less than 0.6%.8 The ion sputtering and annealing in an ultrahigh vacuum chamber Cr on Fe 100 system has been widely investigated for its with the base pressure of 1.5 10 10 Torr. It is well unusual magnetic properties and its application for a giant known that obtaining an adsorbate-free especially nitrogen , 0163-1829/99/59 16 /10918 5 /$15.00 PRB 59 10 918 ©1999 The American Physical Society PRB 59 SURFACE ALLOY FORMATION OF Fe ON . . . 10 919 FIG. 1. Change of the morphology of the Fe overlayer as a function of subsequent annealing temperature in the low coverage of Fe. The density and total area of islands decrease as the islands grow to rectangular shape with increasing annealing temperature, T. a T room temperature, b T 200 °C, c T 300 °C. All the images were taken at Vs 2 V and the image sizes are 350 350 Å. clean Cr surface is difficult.21 Insufficient sputtering and an- Fe coverage. The density and the total area of islands de- nealing results in a nitrogen-induced superstructure. We were creases and the islands grow to rectangular shapes as the able to decrease impurity levels below 0.3% at the surface by annealing temperature is raised. There are two possibilities to sputtering and annealing up to 900 °C for several months, as explain this phenomenon; 1 Fe atoms reevaporate at the monitored by low energy electron diffraction LEED , Auger annealing temperature, or 2 Fe atoms incorporate into the electron spectroscopy, and STM. After the clean surface was Cr substrate. Since the Fe Auger signal was nearly un- obtained, Fe was deposited at room temperature by directly changed during and after annealing, it was concluded that Fe heating a 99.99% Fe wire with the deposition rate of 1 incorporation into Cr is far more probable. Since Fe has a ML/min. In order to study alloying of Fe into the Cr sub- localized surface state just above the Fermi level and Cr has strate, the sample was annealed subsequently at the various one just below the Fermi level, one can distinguish Fe atoms temperatures for 10 min each time. The temperature of the from the Cr substrate if the tunneling voltage is set at the Fe sample was measured by an optical pyrometer and the un- surface state. Since the images in Fig. 1 were taken at the certainty in the temperature measurements is 10 °C. The sample bias voltage (V STM images were obtained after cooling down the sample to s) of 2 V and those in Fig. 2 were taken at 0.3 V close to the Fe surface state , more visible the room temperature over a period of an hour. The detailed atomic rows in Fig. 2 are believed to be incorporated Fe design of UHV STM used in this paper can be found atoms. Although they are clearly visible, their heights are elsewhere.23 But briefly, the system is equipped with UHV much lower than those of 1 ML Fe islands, indicating that STM, LEED, and Auger electron spectroscopy. STM images the rows are not from the physical protrusion but from their were obtained at various tunneling voltages in order to iden- electronic property. The atomic rows start from island or step tify chemical species on the surface. edges and there is no region where they are nucleated on the terrace. If steps or island edges may work as the reaction RESULTS AND DISCUSSION sites for incorporation of adatoms into the substrate, the al- loying would start from those sites. Based on many observed When Fe was deposited on the Cr 100 surface at room STM images, there is a strong tendency that the ratio of the temperature, heterogeneous nucleation of Fe was observed total amount of Fe atoms as pure Fe islands to incorporated near defects. Figure 1 shows the change of the morphology Fe atoms decreases with decreasing terrace width. It con- as a function of subsequent annealing temperature at the low firms the fact that alloying starts from step edges because FIG. 2. Surface alloy by incorporation Fe atoms into Cr matrix. There can be seen the Fe atomic rows, which are brighter than Cr bare terraces and darker than Fe islands with 1-ML height. The incorporation of Fe atoms started from island or step edges. When the annealing temperature is 300 °C, the alloyed area was found on the terrace. a T 200 °C, b T 300 °C, c close-up view of the alloyed area in b . All the images were taken at the Fe surface state (Vs 0.3 V) and the image sizes are 180 180 Å except c . Additional black and white lines appear along the step edges in the process of enhancing the image contrast. 10 920 CHOI, JEONG, PARK, KAHNG, LEE, AND KUK PRB 59 FIG. 4. The depressed tunneling probability of the alloyed re- FIG. 3. Identification of different elements in the alloyed re- gions when the image is taken above the vacuum level of Cr at a gions using surface states. a V Vs 5.5 V, compared to the Fe surface state at b Vs 0.3 V. The s 0.3 V Fe surface state , b Vs 0.2 V Cr surface state . The contrast of the atomic row marked image sizes are 350 350 Å and T 300 °C. by an arrow is reversed as the sample bias is changed from 0.3 to 0.2 V. The image sizes are 80 90 Å and T 200 °C. ered with care in each case. It should be also noted that surface-alloy formation is promoted only above 200 °C in the there are effectively more defect sites in narrow terraces for Cr/Fe 100 case from the former results.25 Therefore, it may alloying at the step edges. As we raised the annealing tem- be inferred that Fe atoms with enough thermal energy will perature up to 300 °C, the alloyed region on the terrace can incorporate into the Cr matrix near the island edges and Cr be found. The region is thought to be the region where Fe atoms substituted by Fe atoms diffuse towards the edges of islands existed at lower temperature and the Fe island is now the upper terrace to form the surface alloy near the step fully incorporated into the Cr substrate at 300 °C. edges. From the STM images at two surface states, we pro- For incorporated Fe atomic rows, their contrast would be pose a structural model of the surface Fe reversed as the sample bias is changed from the Fe surface 0.5Cr0.5 alloy in which Fe and Cr rows alternate. state (Vs 0.3 V) to the Cr surface state (Vs 0.2 V). Fig- It was reported that elemental contrast in an STM can be ure 3 shows this phenomenon dramatically. Figures 3 a and obtained from the different local work function of the differ- 3 b show the STM images taken at sample bias voltages of the Fe and Cr surface states, respectively. The contrast of the ent elements.26 As the bias voltage of STM is set near the atomic row marked by an arrow is reversed as the sample vacuum level, a Rydberg-type series of resonance states may bias is changed from 0.3 to 0.2 V. From this observation, enhance the tunneling probability. They can be regarded as we can conclude that Fe atoms incorporate into the Cr sub- image states modified by the field of the tip. While the low- strate to form a well-ordered surface alloy at the annealing est level of the image states is known to be bound by 0.85 eV temperature of 200 °C. below vacuum level from a simple model,27 the field applied When Cr atoms diffuse into the Fe matrix, they form a by the tip shifts the lowest level of the image states slightly disordered isomorphic alloy.8 However, the observed Fe above the work function. Since the image states are tied to atomic rows indicate an ordered alloy formation as we grow the local work function of a material, it is possible to obtain Fe on a Cr 100 substrate. The driving forces for the surface- elemental contrast in STM by switching the tip bias between alloy formation have been attributed to their lattice mismatch image states of different materials. Recently, the chemical and difference in their surface-free energies. Since the lattice identification of Cu stripes formed along the surface steps of mismatch between Fe 2.87 Å and Cr 2.88 Å is small,24 vicinal Mo 110 was achieved by using the different image the difference in the surface-free energies between Fe and Cr states resulting from the different work functions.26 Since Fe may play a significant role in this case. Although simple and Cr are very similar in their work functions, it is expected thermodynamic arguments have been successful in explain- that their chemical identification may not be possible. At the ing surface-alloy formation in many binary cases, recent same time, atomically resolved STM images are hardly ob- STM studies and total energy calculations show that the tained when the tunneling bias is set near the vacuum level nai¨ve surface-free energy argument is inadequate for some due to the large tunneling gap. However, the work function cases.2,6,7,10 From the coherent-potential approximation and of an ordered alloy is known to be varied as a function of the effective-medium theory EMT calculation, Christensen compositional ratio between two elements and may have a et al. constructed surface-phase diagrams for all transition- local minimum at a specific composition.28 Therefore, the and noble-metal combinations for close-packed surfaces of work function of the well-ordered alloy in Fe/Cr 100 can be the equilibrium structure of the host metal as a function of considerably different from that of Fe or Cr. Figure 4 a the surface composition.10 Their results successfully explain shows the depressed tunneling probability of the alloyed re- many STM results of alloys including the Cr/Fe 100 system, gion when the image was taken just above the Cr vacuum although the present face 100 is not a close-packed plane. level. While the alloyed regions look brighter than the bare In the view of their results, it was proposed that Fe cannot Cr surface in Fig. 4 b , which was taken at the Fe surface alloy with the Cr substrate. However, since their general cal- state, the contrast is reversed when the image was taken culation is restricted to the zero temperature limit, the tem- above the Cr vacuum level. From these observations, it is perature dependence of the phase diagram should be consid- proven that Fe makes an alloy with the Cr substrate and the PRB 59 SURFACE ALLOY FORMATION OF Fe ON . . . 10 921 is changed from the Fe surface state to the Cr surface state, as shown in Fig. 5. When 1 ML Fe is deposited at room temperature, the Fe island grows into the mound shape of similar size, which resembles the feature during the Fe 001 homoepitaxy.30 The origin of the mound shape of islands was explained by the difference of energy barrier between the transport of diffus- ing atoms downward over the island edge and the usual sur- face diffusion on the island. If the thermal energy is large enough to overcome the energy barrier for the descending of the diffusing atoms over the island edge, a two-dimensional island will be formed. In this paper, the kinetic energy of Fe FIG. 5. The nitrogen-induced structure of Cr 100 surface, at a atoms can be increased by the postannealing. As the sample Vs 0.2 V, b Vs 0.3 V. The image sizes are 110 110 Å. is postannealed at 250 °C for 10 min, the coalescence of Fe islands is observed in addition to the flattened shape of Fe local barrier height of the alloy is quite different from that of islands. While the coverage calibrated from the deposition the substrate. rate is 0.8 ML, the coverage estimated from the STM images when the coalescence occurs is about 0.55 ML, suggesting The structure of the alloy resembles that of the nitrogen- that the solubility of Fe into the Cr 100 surface at 250 °C is induced Cr 100 surface, which was recently reported by about 25%. The enhancement of the solubility in the surface Schmid et al.29 They observed that nitrogen induces the re- alloying compared to the bulk case is consistent with the Cr construction of the Cr 100 surface and the reconstructed on Fe 100 case.19 When the annealing temperature is raised structure changes as a function of the concentration of seg- up to 300 °C, the diffusivity of Fe atoms on the Cr surface is regated nitrogen on the Cr surface. The proposed structure large enough and the boundary of the Fe layer goes along changes from the clean Cr (1 1) to the N-rich (1 1) via highly symmetric 100 directions. The seemingly negative c(2 2) and c(3& &) R 45° with increasing coverage islands on the terrace in Fig. 6 c are the exposed areas of the of N. When the N coverage is above 0.5 ML, line-shaped Cr substrate, which is surrounded by the Fe overlayer. Un- domain boundaries are observed. Similar results were also expectedly, there is no indication of well-ordered alloy for- observed in the process of cleaning the Cr surface. When the mation in that area, as well as there is no evidence of alloy nitrogen is not fully depleted from the Cr surface, which can formation on the Fe overlayer. From these facts, it can be be monitored from Auger electron spectroscopy, there appear inferred that incorporation of Fe into the Cr surface at the atomic rows similar to the Fe atomic rows in the surface high-coverage regime mainly occurs at the interface between alloy of the Fe/Cr 100 system. In spite of the similarity in the Fe overlayer and the Cr substrate. their shapes, there still remain a couple of remarkable differ- ences. One is the ordering of the spacing of atomic rows and the other is the dependence of its contrast on the bias volt- CONCLUSION ages. While the rows in the surface alloy are well ordered, those in the N-induced structure are arbitrarily arranged with While Cr alloys with an Fe substrate without any ordering a great deal of disorder. Although there are also atomic rows at the interface of Cr/Fe 100 ,8 we found that Fe atoms in- similar to those in the Fe/Cr alloy, their spacings are not corporate into the Cr surface and form a well-ordered alloy at uniform compared to the case of the Fe/Cr alloy. The bias the elevated sample temperature at the low-coverage regime voltage dependence of the contrast of the rows is quite dif- of Fe. The compositional ratio between Fe and Cr in the ferent from that in the Fe/Cr alloy. The contrast of the atomic surface alloy is 1:1. We distinguished the Fe atomic rows rows is enhanced when the image is taken at the Cr surface from the Cr rows in the Fe/Cr surface alloy, by imaging at state, and there is no contrast inversion when the bias voltage the bias voltages of the surface states of two elements. In FIG. 6. Change of the morphology of the Fe overlayer from the kinetically roughened mound structure to the equilibrium two- dimensional island by raising annealing temperature. a T room temperature, b T 250 °C, c T 300 °C. All the images were taken at Vs 2 V. The image sizes are 180 180 Å for a and 350 350 Å for b and c . 10 922 CHOI, JEONG, PARK, KAHNG, LEE, AND KUK PRB 59 addition, the work function at the alloyed regions is some- on Fe 100 case,19 the alloying is observed to be confined what different from that at a bare Cr surface, which can be only to the interface of Fe/Cr 100 . inferred from the lower tunneling probabilities observed at the Cr image states. When we deposit Fe around 1 ML and anneal the sample ACKNOWLEDGMENTS up to 300 °C, Fe atoms are dissolved into the Cr surface about 25% and the boundaries of the Fe overlayer go along This paper was partially supported by the Ministry of the crystalline symmetric axis 100 . Contrary to the low- Education, Korea through Grant No. BSRI-97-2416, the Ko- coverage regime, there is no indication of well-ordered alloy rean Science and Engineering Foundation through Grant No. formation in the exposed Cr surface. While an alloyed ap- 961-0207-072-2, and the national creative research initiative pearance is observed at the Cr coverage of 2­3 ML in the Cr in Korea. *Present address: Hyundai Electronics Industries Co. LTD, Ichon, 14 P. W. Murray, S. Thorshaug, I. Stensgaard, F. Besenbacher, E. Kyoungki-do, 467-701, Korea. Lægsgaard, A. V. Ruban, K. W. Jacobsen, G. Kopidakis, and H. Author to whom correspondence should be addressed. Electronic L. Skriver, Phys. Rev. B 55, 1380 1997 . address: ykuk@phya.snu.ac.kr 15 T. Flores, S. Junghans, and M. Wuttig, Surf. Sci. 371, 14 1997 . 1 E. Bauer and Jan H. van der Merwe, Phys. Rev. B 33, 3657 16 L. M. Falicov, Phys. Today 45 x..., 46 1992 . 1986 . 17 C. Turtur and G. Bayreuther, Phys. Rev. Lett. 72, 1557 1994 . 2 I. Pleth Nielsen, F. Besenbacher, I. Stensgaard, E. Lægsgaard, C. 18 J. Unguris, R. J. Celotta, and D. T. Pierce, Phys. Rev. Lett. 67, Engdahl, P. Stoltze, K. W. Jacobsen, and J. K. Nørskov, Phys. 140 1991 . Rev. Lett. 71, 754 1993 . 19 A. Davies, J. A. Stroscio, D. T. Pierce, J. Unguris, and R. J. 3 C. Nagl, O. Haller, E. Plazgummer, M. Schmid, and P. Varga, Celotta, J. Magn. Magn. Mater. 165, 82 1997 . Surf. Sci. 321, 237 1994 . 20 4 O. Kubaschewski, Iron-Binary Phase Diagrams Springer- C. Nagl, E. Plazgummer, M. Schmid, P. Varga, S. Speller, and W. Verlag, Berlin, 1982 . Heiland, Phys. Rev. Lett. 75, 2976 1995 . 21 5 Markus B. Hugenschmidt, Andreas Hitzke, and R. Ju¨rgen Behm, J. S. Foord and R. M. Lambert, Surf. Sci. 115, 141 1982 . 22 Phys. Rev. Lett. 76, 2535 1996 . For example, Refs. 2 and 5 are the Au-Ni case, and Refs. 3 and 4 6 P. T. Sprunger, E. Lægsgaard, and F. Besenbacher, Phys. Rev. B are the Pb-Cu case 23 54, 8163 1996 . Y. Kuk and P. J. Silverman, Rev. Sci. Instrum. 60, 165 1989 . 7 24 H. Ro¨der, R. Schuster, H. Brune, and K. Kern, Phys. Rev. Lett. C. Kittel, Introduction to Solid State Physics, 6th ed. Wiley, New 71, 2086 1993 . York, 1991 , p. 23. 8 A. Davies, J. A. Stroscio, D. T. Pierce, and R. J. Celotta, Phys. 25 D. T. Pierce, J. A. Stroscio, J. Unguris, and R. J. Celotta, Phys. Rev. Lett. 76, 4175 1996 . Rev. B 49, 14 564 1994 . 9 J. Tersoff, Phys. Rev. Lett. 74, 434 1995 . 26 T. Jung, Y. W. Mo, and F. J. Himpsel, Phys. Rev. Lett. 74, 1641 10 A. Christensen, A. V. Ruban, P. Stoltze, K. W. Jacobsen, H. L. 1995 . Skriver, J. K. Nørskov, and F. Besenbacher, Phys. Rev. B 56, 27 N. V. Smith, Phys. Rev. B 32, 3549 1985 . 5822 1997 . 28 A. Kiejna and K. F. Wojciechowski, Metal Surface Electron 11 D. D. Chambliss and S. Chiang, Surf. Sci. Lett. 264, L187 1992 . Physics Alden, Oxford, 1996 , p. 94. 12 C. Nagl, E. Plazgummer, O. Haller, M. Schmid, and P. Varga, 29 M. Schmid, M. Pinczolits, W. Hebenstreit, and P. Varga, Surf. Surf. Sci. 331­333, 831 1995 . Sci. 377-379, 1023 1997 . 13 R. G. R. van der Kraan and H. van Kempen, Surf. Sci. 338, 19 30 J. A. Stroscio, D. T. Pierce, M. D. Stiles, A. Zangwill, and L. M. 1995 . Sander, Phys. Rev. Lett. 75, 4246 1995 .