J. Phys.: Condens. Matter 11 (1999) 9365–9385. Printed in the UK
`
`PII: S0953-8984(99)05098-5
`
`Growth of thin films
`
`E Bauer
`Department of Physics and Astronomy, Arizona State University, Tempe, AZ 85287-1504, USA
`
`Received 10 June 1999
`
`Abstract. The results of epitaxial growth studies of ferromagnetic metals for a selected class of
`nonmagnetic substrates is reviewed. The reverse sequence is also discussed for some systems of
`importance in double layer and sandwich studies. The interrelation between film structure and
`magnetic properties is pointed out for several examples.
`
`1. Introduction
`
`The rapid evolution of magnetic thin film sensors and memories during the past decade has
`not only led to an explosive growth of the literature on magnetic properties of thin films but
`to a similar development of studies of the growth of these films. This is in part due to the
`increasing number of experimental techniques suitable for growth studies and in part due to
`the recognition that the magnetic properties of the films depend strongly on the film structure
`which is largely determined by the growth. This article can, therefore, cover only a small
`fraction of the work in this field. It will not discuss the practical aspects of thin magnetic films.
`They can be found, for example, in the proceedings of the conferences on magnetic recording
`media [1] and of other conferences on magnetism and magnetic materials. Likewise, films on
`amorphous substrates, polycrystalline, sputtered and electrolytically deposited films will not
`be included. This leaves epitaxial layers grown under ultrahigh vacuum (UHV) conditions of
`which only a few could be selected. For a more general discussion of metal epitaxy on metals
`the reader is referred to older reviews [2, 3] and monographs, in particular to the articles in [4]
`which cover the various processes involved in epitaxy.
`The review is organized as follows. After a brief description of the experimental methods
`growth on densely packed surfaces, mainly bcc (110) and fcc (111) surfaces will be discussed.
`Growth on fcc (100) surfaces is treated jointly with growth on bcc (100) surfaces. Films on
`other surfaces such as bcc (111), bcc (211), fcc (110) and hcp (10N10) as well as quasi-one-
`dimensional crystal growth on vicinal surfaces are touched only briefly. A discussion of the
`processes which allows us to tailor films such as nucleation control, use of misfit anisotropy,
`surfactants etc concludes this contribution.
`
`2. Experimental methods
`
`Of the many techniques in the arsenal of surface science the laterally averaging methods of
`reflection high energy electron diffraction (RHEED), low energy electron diffraction (LEED)
`and Auger electron spectroscopy (AES) have been most widely used in growth studies. Work
`function change (18) measurements have also made valuable contributions, in particular in
`
`0953-8984/99/489365+21$30.00 © 1999 IOP Publishing Ltd
`
`9365
`
`TDK Corporation Exhibit 1010 Page 1
`
`

`
`9366
`
`E Bauer
`
`the monolayer range. More recently ultraviolet photoelectron spectroscopy (UPS) and x-ray
`photo-electron diffraction (XPD) have also been shown to be very useful. Although RHEED is
`still used for the determination of the film orientation, the specular beam intensity oscillations
`due to periodic atomic roughness changes has become its major application. Similarly, LEED
`not only gives the lateral periodicity and—with a dynamical intensity analysis—the atomic
`positions in the unit cell but in its SPALEED version—which has a high resolution in reciprocal
`space—also laterally averaged information on the surface topography. UPS allows us to
`monitor film growth via the sensitivity of the electronic structure to atomic environment up to
`several monolayers (ML) and XPD becomes an important tool for structural analysis beyond
`that thickness.
`Our understanding of the growth of epitaxial films would be rudimentary were it not for the
`contributions which laterally resolving techniques, foremost scanning tunnelling microscopy
`(STM) and low energy electron microscopy (LEEM), have made. The initial problems of
`STM, thermal drift and shadowing by the tip, which made it impossible to monitor film
`growth, have been largely overcome so that growth processes can be studied now quasi-
`continuously by intermittent tip retraction. LEEM does not have these problems but at the
`expense of much lower lateral resolution than STM. However, many aspects of the growth do
`not need the resolution achievable with STM and the possibility to combine LEEM with LEED,
`spin-polarized LEEM (SPLEEM) and x-ray photo-emission electron microscopy (XPEEM)
`allows a comprehensive characterization of the topography, the crystal structure, the magnetic
`domain structure and the chemical composition of the film [5, 6]. The results reported below
`were obtained with a combination of several of these techniques. Others, of course, have
`contributed too but cannot be included here for lack of space.
`
`3. Growth on densely packed surfaces
`
`3.1. The bcc (110) surface
`
`Although not close packed, these is the most densely packed surface of bcc metals and has
`long been a favourite substrate for epitaxy starting with the early studies of the growth of
`Cu on W(110) [7–9]. W(110) and Mo(110) substrates are so attractive because due to their
`high melting point they can be cleaned easily by flashing off deposited layers—with some
`precautions (see below)—so that they can be frequently reused. The (110) surfaces of the
`other bcc metals (Ta, Nb, Fe, Cr and V) are much harder to clean and have been used much
`less.
`
`The surface energies of the ferromagnetic metals of interest here, Ni, Co and Fe are
`sufficiently lower than those of W and Mo so that Stranski–Krastanov growth is expected at
`elevated temperatures and quasi-Frank–van der Merwe growth at lower temperatures. On the
`basis of van der Merwe’s structural phase diagram initially 1–2 pseudomorphic monolayers
`should form with the subsequent growth in the Nishiyama–Wassermann orientation of the fcc
`Ni and equivalent orientations of the hcp Co and the bcc Fe [10]. This growth sequence had
`already been observed earlier in the first AES/LEED/18 study of Ni on W(110) [11] and was
`confirmed by the subsequent studies of this system [12–22] and of Ni on Mo(110) [23] with
`minor differences.
`
`3.1.1. Ni on W(110) and Mo(110) The present state of understanding of the growth of Ni
`on W(110) and Mo(110) may be summarized as follows. At room temperature and below
`where nucleation rate is high and diffusion limited, two-dimensional (2D) pseudomorphic (ps)
`islands form initially with more or less dendritic shape. The ‘dentricity’ depends not only upon
`
`TDK Corporation Exhibit 1010 Page 2
`
`

`
`Growth of thin films
`
`9367
`
`deposition temperature and rate but also upon residual gas coadsorption, a problem encountered
`in all experiments in which the time between flashing the crystal and start of the deposition is
`long and/or the deposition is made cumulatively, for example in STM studies. Due to limited
`mobility, site blocking and possible Ehrlich–Schwoebel barriers at these temperatures, atoms
`landing on top of the islands cannot be incorporated at their edges and ‘squeeze’ into the
`island, transforming them into a close-packed (cp) structure with increasing coverage long
`before completion of the ps layer [24]. Alternatively the transition may occur spontaneously
`when two islands meet [11]. Only under very clean conditions and at elevated temperatures
`can the ps ML fully develop before the transition to the cp layer occurs. This is clearly evident
`in 18 measurements which are very sensitive to atomic roughness. Figure 1(a) shows in
`the upper part such 18 measurements for Ni on Mo(110) at temperatures at which kinetic
`limitations are negligible. Due to the loose packing in the ps ML 8 decreases initially nearly
`linear with coverage until the ps ML is completed and rises thereafter rapidly due to the
`formation of the cp layer. The subsequent change is slow because the 3D islands forming
`cover only a small fraction of the surface (600 K and 790 K) and because of alloying (880 K).
`The AES measurements show only a small change of slope of the Ni signal at the ps ! cp
`transition—which has been overlooked in less accurate work [13]—caused by changes in the
`electronic structure. These are clearly evident in the intensities of UPS spectra at energies
`which are particularly sensitive to structural changes as illustrated in figure 2 [17]. At a
`binding energy of 0.64 eV (b) the intensity suddenly increases strongly at the transition from
`ps to cp packing. For comparison also AES data are shown which were acquired at an emission
`) more sensitive to structural changes than that used in figure 1(a) (42(cid:6)5
`(cid:14)
`(cid:14)
`). Figure 2
`angle (50
`also shows that the ps ! cp transition occurs earlier at 300 K due to kinetic limitations. The
`effect of these limitations can be seen particularly well in figure 1(b) which shows the 18
`change during heating after cumulative depositions of Ni on Mo(110) at 365 K. Above about
`1=3 ps ML heating causes initially an increase of 8 due to the incorporation of atoms on top of
`the ps islands and resulting ps ! cp transition followed by a strong decrease to the equilibrium
`ps structure. The decrease is strongest around 1 ps ML as one would expect. The additional
`subtle features in these curves are explained in [23].
`Recent STM studies [20–22] are in apparent contradiction with the picture of the ps ! cp
`tranistion derived from the laterally averaging studies discussed above. In the STM studies
`the cp (7 (cid:2) 1) structure at 0.9 ML coverage could not be converted into the ps structure by
`annealing at 900 K. Future work has to show whether this is due to the higher step density
`on the surface used in the STM study or due to contamination. The second possibility is not
`unlikely as LEEM studies of the initial growth of Co on W(110)—which is very similar to that
`of Ni—have also shown holes in the ML similar to those observed in the STM work whenever
`the surface was contaminated.
`The growth beyond the first ML depends strongly upon temperature. At 300 K RHEED
`specular beam intensity oscillations can be seen over about 10 ML, depending upon the angle
`of incidence. Growth in poor vacuum increases the range of the oscillations significantly and
`so does deposition in good vacuum at 100 K [12, 15, 16]. This does not necessarily mean
`monolayer-by-monolayer growth as frequently assumed but indicates only that the surface
`roughness varies periodically during growth. Nevertheless, quasi-Frank–van der Merwe
`growth occurs at least initially as the Auger data up to three ML show.
`In recent LEED
`work the film strain along the W[1N10] direction could be observed up to 8 ML before the
`surface became rough [25]. Although the temperature at which alloying sets in has not been
`determined accurately, there is little doubt on the basis of AES and energy loss spectroscopy
`that alloying between Ni and Mo starts definitely below 900 K and levels off at the approximate
`composition NiMo at about 1100 K in layers more than 4 ML thick [23]. The situation on W
`
`TDK Corporation Exhibit 1010 Page 3
`
`

`
`9368
`
`E Bauer
`
`Figure 1. (a) Work function change 18, Ni and Mo Auger signals as a function of deposition time
`(bottom) and coverage (top) on Mo(110) at various temperatures, measured at temperature. (b)
`Work function change during heating after cumulative depositions at 365 K; coverage in ps units
`[23].
`
`is believed to be similar [11]. It should be noted that the solubility of Ni in W and Mo is low
`but that W and Mo dissolve in Ni up to more than 20 at.%. Thus the Ni islands have to be
`large enough before alloying can start. No alloying occurs below 1 ML. The alloying problem
`is much clearer in Co films and will be discussed there in more detail. At lower temperatures
`the Ni layers grow in the Stranski–Krastanov mode as judged by the comparison with the AES
`and 18 data obtained at 300 K. For example, the slope of the AES signal in figure 1(a) beyond
`
`TDK Corporation Exhibit 1010 Page 4
`
`

`
`Growth of thin films
`
`9369
`
`Figure 2. Ni AES and UPS signals as a function of deposition time (bottom) and coverage (top)
`on W(110). (a) Amplitude of differentiated M2;3M4;5M4;5 (61 eV) Auger signal at a polar angle
`(cid:14)
`of 50
`in the [001] azimuth. (b)–(d) UPS intensities in normal emission at 1.46 eV, 0.64 eV and
`0.16 eV binding energy, respectively. The curves are shifted for clarity [17].
`
`Figure 3. LEED patterns (a), (c), (d) and reciprocal lattice unit mesh (b) of pattern (a). (a), (b)
`(8 (cid:2) 1), (b) complex and (c) p(2 (cid:2) 2) structure. Energies 141 eV, 104 eV and 128 eV, respectively.
`(c) and (d) are from an alloy phase [23].
`
`the first ML is significantly smaller at 600 K than at 300 K (not shown) and decreases further
`with increasing deposition temperature. More detailed data for Ni are not available.
`
`3.1.2. Co on W(110) and Mo(110). The growth of Co on W(110) and Mo(110) has been
`studied in much more detail, though not with STM but with LEEM. First the results of the
`
`TDK Corporation Exhibit 1010 Page 5
`
`

`
`9370
`
`E Bauer
`
`Figure 4. UPS intensity at a binding energy of 0.03 eV as a function of coverage measured during
`(cid:14)
`deposition at 300 K with parallel polarized light in the W [001] azimuth at a polar angle of 30
`[27].
`
`laterally averaging studies [23, 26–29] will be discussed. Co grows on both surfaces initially
`in the same manner as Ni does. Unless deposited or annealed at sufficiently high temperature
`the transition from the ps to the cp structure occurs as early as about 1=2 ML. The cp layer
`has the same characteristic (8 (cid:2) 1) or (8 (cid:2) 2) pattern—depending upon the definition of the
`substrate unit mesh—as the cp Ni layer (figure 3(a), (b)) [23]. Deposition at 750 K leads
`to very sharp changes in the structure-sensitive quantities such as AES, 18, UPS signals
`and LEEM contrast which allow a very accurate coverage and deposition rate calibration.
`However, even during room temperature deposition the ps ! cp transition is usually clearly
`visible, for example in the UPS intensity versus coverage curve taken at a binding energy
`which is particularly sensitive to changes in the first monolayer (figure 4 [28]). The transition
`occurs here before the completion of the ps layer for reasons discussed above. The rounded
`transitions between the second and the third and the following monolayers shows that these
`layers are not completed before the start of the subsequent layer.
`The growth of the first monolayer has been studied extensively by LEEM as a byproduct
`of SPLEEM [30] studies of thin Co films on W(110) for cleanness checks and deposition rate
`calibration. Figure 5 shows the influence of contamination on the step flow growth ((a),(b)
`[31]) and the transition from the ps to the cp monolayer ((c),(d) [32]). On a clean surface in
`good UHV the growth front is always smooth, at least at high temperatures. A rough growth
`front is a good indicator of contamination unless growth is crystallographically anisotropic,
`for example due to anisotropic strain. Anisotropic strain is responsible for the much higher
`growth rate of the cp layer in the [001] direction. This is seen in figures 5(c),(d) which show
`the monolayer shortly after the completion of the ps layer and before the completion of the
`cp layer. The layer is growing preferentially along the [001] direction because it is a ‘floating
`layer’ in this direction due to the large misfit while in the low misfit direction .[1N10]/ it is
`locked into the substrate.
`Growth after completion of the monolayer is very similar to that of Ni. High resolution
`LEED measurements [29] indicate that the strain in the layer is constant up to about 9 ML and
`subsequently decreases inversely with thickness. AES [23], UPS and RHEED specular beam
`intensity oscillations [28] show that monolayer-by-monolayer growth rapidly deteriorates with
`increasing thickness at 300 K while at 100 K many RHEED oscillations can be seen. The
`layer is, however, less ordered at 100 K. LEEM studies have shown that films with minimum
`
`TDK Corporation Exhibit 1010 Page 6
`
`

`
`Growth of thin films
`
`9371
`
`Figure 5. LEEM images of the early stages of the growth of the pseudomorphic Co layer on
`W(110) ((a), (b) [31]) and near the beginning and end of the close-packed monolayer ((c), (d) [32])
`at 750 K. Field of view: 10, 8, 14 and 14 (cid:22)m, electron energy 1.5, 1.5, 7 and 8.2 eV, in (a), (b), (c)
`and (d), respectively.
`
`roughness, that is a three atomic layer level system, and good order can be obtained by
`depositing at temperatures just below the coalescence temperature. This increases diffusion on
`the existing film surface, reduces nucleation of the next monolayer on it and overcomes possible
`Ehrlich–Schwoebel barriers. The LEEM image of figure 6 [33] shows such a film surface at a
`mean thickness of about 5 ML. The film consists of regions which are 4, 5 and 6 ML thick.
`Coalescence is very clearly evident in AES annealing experiments as illustrated in figure 7
`[23]. The layer is stable up to about 450 K and then agglomerates strongly within a small
`temperature range. Above 750 K alloying starts, which leads to wetting and re-spreading of
`the Co alloy crystal over the surface. Annealing at 800 K is sufficient to produce the complex
`LEED pattern of figure 3(c). The AES extrema at about 850 K correspond to the composition
`Co3Mo. Between 850 K and about 1150 K the alloy crystals agglomerate and at still higher
`temperatures Co sublimes from them leaving substrate mesa structures of the type seen in
`figure 8 [34] behind. If their formation is to be suppressed then desorption has to be done
`very fast so that the transport processes necessary for their formation cannot be effective. In
`thicker layers (>10 ML) coalescence is delayed and pre-empted by alloying which produces
`the p(2 (cid:2) 2) pattern shown in figure 3(c).
`Growth above 450 K is clearly of the Stranski–Krastonov type with a stable monolayer
`between the 3D crystals. This is illustrated in figure 8 [34] in which the mean film thickness
`
`TDK Corporation Exhibit 1010 Page 7
`
`

`
`9372
`
`E Bauer
`
`Figure 6. LEEM image of a 5 ML thick Co layer deposited on W(110) at about 400 K. Field of
`view 6 (cid:22)m, electron energy 3 eV. Quantum size contrast [33].
`
`increases from 4.5 ML to 11 ML. From the surface coverage by the 3D crystals the mean
`crystal thickness can be estimated to increase from 28 ML to 48 ML in the thickness range
`shown.
`
`Fe as the fundamental ferromagnetic metal has been
`3.1.3. Fe on W(110) and Mo(110).
`studied most extensively with all methods listed at the beginning except LEEM, both on
`W(110) [35–46] and Mo(110) [23, 47]. In contrast to Ni and Co not only one ps ML grows
`at room temperature but also a metastable second one which is stable up to 650 K [37] or
`770 K [46] on W(110) but only up to about 500 K on Mo(110) [23]. Thus there is no ps ! cp
`transition in the first monolayer. Rather, ps islands grow up to about 0.6 ML at 300 K and then
`rapidly coalesce. The first 2D crystals of the second ML appear as early as 0.8 ML. Similarly,
`at higher coverages the next layer starts long before the preceding layer is completed [40, 42]
`which explains the pronounced rounding of the Auger signal–deposition time curves.
`In
`the first ML the islands grow rather isometrically, in the second ML preferentially along the
`W[001] direction. The preferential growth along this direction continues with further growth,
`leading to the 1D coarsening into roof-like facets along the [001] direction studied in detail on
`annealed Fe(110) films on W(110) [48]. This kinetic roughening phenomenon has also been
`used to grow corrugated Fe/Cr(110) superlattices whose morphology as a function of growth
`temperature and rate was studied in detail by RHEED and SEM [49].
`At elevated temperatures growth occurs via step flow. The second ML is ps only up to
`about 1.2–1.3 ML [40] and then develops misfit dislocations along the [001] direction with an
`average spacing of about 4.6 nm. In the third layer the misfit in the W[1N10] direction is still
`accommodated by dislocations along the W[001] direction but now with a more regular spacing
`of about 2.7 nm [40, 42]. Starting from the fourth ML onward the misfit in both substrate
`directions is accommodated by a 2D modulation of the lattice [40, 42–44] which had already
`been deduced earlier from LEED studies [35]. From 4 ML to about 10 ML, the thickest crystals
`
`TDK Corporation Exhibit 1010 Page 8
`
`

`
`Growth of thin films
`
`9373
`
`Figure 7. Co and Mo AES signals from 2.5 ML and 5.1 ML thick Co films on Mo(110) as a function
`of annealing temperature showing coalescence, alloying, coalescence and desorption [23].
`
`on which this modulation can be seen in STM, the periodicity of the modulation corresponds
`to a complete strain relaxation of the Fe lattice and the crystals grow across substrate steps
`with flat top [40] faces similar to the Co crystals shown in figure 8. Coalescence, which starts
`in the intermediate thickness range (3–6 ML) already at about 600 K, begins in thicker films
`(>10 ML) above 750 K and is more or less completed at 850 K [46], resulting in large fully
`relaxed thick crystals surrounded by 1 ps ML [44, 46]. As already mentioned, the growth of
`Fe on Mo(110), which has been studied in considerable detail by AES, LEED, 18 and TDS
`measurements [23] and more recently by STM [47], is very similar to that on W(110). The
`details can be found in the original literature. Additional recent data of the growth of Fe on
`W(110) may be found in [50] (RHEED, AES) and [51] and [52] (stress, STM).
`
`3.2. The hcp (0001) surface
`
`Much less work has been done on these surfaces than on the bcc (110) surface. The growth of
`Ni on Re(0001) was studied by AES and LEED with the result that growth at 325 K occurs in the
`quasi-Frank–van der Merwe mode at least up to 4 ML and that the first ML is pseudomorphic.
`The subsequent MLs up to about 10 ML form an approximate (10 (cid:2) 10) coincidence lattice
`which can be interpreted by a misfit dislocation network and double scattering [53]. These
`laterally averaged results are nicely confirmed by a more recent STM study [54]. Annealing
`above 600 K leads to coalescence [53]. The growth of Ni on Ru(0001) differs somewhat
`
`TDK Corporation Exhibit 1010 Page 9
`
`

`
`9374
`
`E Bauer
`
`Figure 8. LEEM images selected from a series during deposition of Co on W(110) just below the
`onset of alloying. The Co crystals nucleate preferentially on the W mesas formed in many preceding
`Co desorptions and grow preferentially in the W[001] direction without noticeable impediment by
`substrate steps and step bunches. Accompanying SPLEEM images in the original work show
`that this is also the easy magnetization axis and that the crystals are single domains and maintain
`their magnetization also upon coalescence. The local magnetization is also maintained when a
`continuous film breaks up into islands with increasing annealing temperature [33].
`
`from that on Re(0001) [55, 56]. A superstructure LEED pattern whose spot positions agree
`within the limits of error with those of an unstrained Ni(111) plane appears already above
`0.5 ML. Annealing layers with this pattern leads to a (1 (cid:2) 1) LEED pattern and reduces the
`work function considerably. This shows that the ps ! cp transition occurs already in the
`first ML similar to the growth on the bcc (110) surfaces. The further growth is very much like
`that on Re(0001). Annealing leads to agglomeration of all material in excess of 1 ps ML. The
`agglomeration temperature increases with thickness from about 600 K at 2–3 ML to more than
`1000 K at 5 ML.
`The growth of Fe on Ru(0001) has been studied by several groups [56–60]. Quasi-Frank–
`van der Merwe growth is observed at 300 K initially. After the completion of the third ML AES
`suggests double layer growth. The first ML is pseudomorphic, the second and third show the
`LEED pattern of a transition structure to the unstrained Fe(110) orientation in three equivalent
`azimuthal orientations (Fe[1N11] k Ru[10N10]) which becomes clearer with increasing thickness
`and slight annealing. The agglomeration temperatures are strongly coverage dependent: about
`400 K up to 4 ML, 520 K at 8 ML and 660 K at 12 ML. Between 760 K and 780 K alloying sets
`in which produces above 4 ML a LEED pattern attributed to the phase FeRu. The intensity
`
`TDK Corporation Exhibit 1010 Page 10
`
`

`
`Growth of thin films
`
`9375
`
`of this pattern decreases with increasing annealing temperature, accompanied by a decrease
`of the Fe Auger signal but a nearly constant Ru signal. This suggests thickening rather than
`agglomeration of the alloy. At coverages less than 4 ML only a (1 (cid:2) 1) pattern is seen above
`800 K [56].
`Before concluding this section a few growth studies of nonmagnetic films on ferromagnetic
`films should be mentioned which are of relevance for the understanding of the influence of
`overlayers on magnetic properties and for the understanding of interlayer coupling between
`two magnetic layers and in superlattices. Examples are Cr on Co(0001) [61], Cu on Co(0001)
`[62] and Pt on Co(0001) [63].
`
`3.3. The fcc (111) surface
`
`The literature of the growth of ferromagnetic metals on fcc (111) surfaces is extensive so that
`only a few examples can be discussed which illustrate the complexity of the film growth when
`the surface energy of the layer material is larger than that of the substrate.
`
`3.3.1. Au(111). Because of its 22(cid:2)p
`
`3 herringbone reconstruction this surface shows not only
`the complexities of the thermodynamically driven place exchange between film and substrate
`material but also those caused by the preferred nucleation at the point dislocations of the
`reconstruction. The very low coverage range, in which the preferred nucleation is best visible,
`has been studied extensively with STM, mainly in Co films [64–68] but also in Ni films [69–71]
`and Fe films [72]. Additional information for thicker films comes from AES [73, 74], helium-
`atom scattering [74], XPS [75] and magnetic measurements [68]. The three film/substrate
`systems have many things in common but differ also in some in part surprising aspects.
`Common features are (i) the preferred nucleation on the point dislocations at the herringbone
`corners on well ordered surfaces, (ii) the disappearance of the Auger signals above about
`600 K in Ni and Co layers, probably also in Fe layers, and (iii) the approximate monolayer-
`by-monolayer growth from about 2 to 4 ML followed by increasingly rougher growth.
`Differences exist in the manner in which the film crystals sink into the substrate and in
`how Au covers them with increasing temperature. The most striking difference is that the
`crystals formed initially are 1 ML thick in the case of Ni and Fe, but 2 ML thick in Co films,
`although the atomic diameters and surface energies (according to [76]) of the three materials
`differ very little. The references quoted reveal a great complexity of the place exchange
`between film and substrate material which depends strongly on material, coverage, deposition
`and annealing temperature and time. In any case, at temperatures above 400 K–450 K some
`Au has to be expected on top of the film material which is known to have a strong influence
`on the perpendicular anisotropy of the films, and at still higher temperatures very thin layers
`may be completely covered by Au. This restricts film growth on Au to considerably lower
`temperatures than on the refractory metals discussed in section 3.1.
`
`3.3.2. Cu(111). Ferromagnetic films on Cu(111) have a long and conflicting history [77]. The
`article by Heinz in this issue describes the structure of the system Co/Cu(111) in detail so that
`only some of the apparent discrepancies between the various growth studies will be discussed
`briefly. They are to a large extent due to different growth and substrate conditions but in part
`also caused by the limitations of the methods used to characterize the growth via the structure of
`the resulting film. Growth was characterized by RHEED [77, 78], LEED, AES, STM, XPD, ion
`scattering spectroscopy, secondary electron angular distribution measurements, helium atom
`scattering and other techniques. For references see the article by Heinz and some of the recent
`work on this subject [79–85]. In agreement with theoretical expectations [86] initially crystals
`
`TDK Corporation Exhibit 1010 Page 11
`
`

`
`9376
`
`E Bauer
`
`form which are 2–3 ML high, one ML of which is sometimes at the level of the surrounding
`substrate surface. However, there are also other data [87] which appear incompatible with
`double layer growth and rather suggest monolayer-by-monolayer growth.
`Cu moves easily onto the surface of these crystals, even at room temperature and rapidly
`around 400 K, covering the surface with a (sub)monolayer of Cu. Completion of the double
`layer depends strongly upon substrate temperature and deposition rate and occurs between 2
`and 4 ML. The layers are pseudomorphic up to 6–7 ML according to some reports but the
`fcc ! hcp transition in Co films occurs earlier. In Fe films the transition from fcc to bcc
`structure with Kurdyumov–Sachs starts as early as 2–3 ML in slow depositions while pulsed
`laser deposition which produces a much larger nucleation rate stabilizes the fcc structure up
`to about 6 ML [124, 125]. After the transition to the equilibrium crystal structure the films
`develop rapidly considerable roughness but Cu usually does not segregate to the surface of
`thick films unless they are annealed or deposited at elevated temperatures (400–500 K).
`
`3.3.3. Other related work. There are numerous other studies on fcc (111) surfaces, mostly
`in connection with magnetic measurements but not detailed enough to give a clear picture of
`the film growth. On substrates with low surface energy such as Ag and Al phenomena similar
`to those discussed in 3.3.2 should occur while on substrates with high surface energy such as
`Rh, Pt or Ir only the heat of mixing which can be quite large in some combinations can drive
`interdiffusion during film growth or annealing.
`The understanding of the reverse film growth, that is of nonmagnetic films on magnetic
`substrates, is also very important in connection with the influence of overlayers and of interlayer
`coupling as already mentioned. Cu has attracted most attention because of the long controversy
`over the conditions necessary for optimum antiferromagnetic coupling but Au has also been
`studied extensively. In the case of Cu STM [62] and LEEM [88] show in qualitative agreement
`that initially a monolayer forms but the second layer starts to grow before the completion of the
`first. Upon deposition at slightly elevated temperatures (e.g. 365 K) the second layer islands
`could be identified to be 2 ML thick [88]. Growth at elevated temperature (400 K) shows clear
`Stranski–Krastanov growth of Cu with one ps ML [89], causing a very rough interface for a
`subsequent Co film. Therefore, high deposition rates and/or low temperatures are necessary
`for the growth of sandwiches and superlattices with flat interfaces.
`Although Au has a lower surface energy than Co the large misfit precludes initial monolayer
`formation [86]. During room temperature deposition on a Co film a smoothing of its surface
`was reported but by the time that 2 ML have been deposited the Au layer is already so rough that
`it produces a transmission RHEED pattern. Deposition at elevated temperatures, e.g. at 400 K,
`produces from the very beginning a surface which is so rough that it does not show the peak
`in the perpendicular anisotropy observed at 1 ML during room temperature deposition [88].
`With increasing thickness the roughness increases to such an extent that roughness-induced
`biquadratic coupling becomes as strong as bilinear coupling in Co/Au/Co sandwiches [31].
`
`4. Growth on (100) surfaces
`
`4.1. The bcc (100) surface
`
`In contrast to the fcc (100) surface the bcc (100) surface is not densely packed. The surface
`has a strong potential corrugation and, therefore, influences the growth of thin films to a much
`larger extent than the surfaces discussed up to now. This is, in particular, true for refractory
`metals because of their strong interaction with the film. Although being an equilibrium surface,
`the bcc (100) surface tends to reconstruct, a tendency which is enhanced by adsorption. For
`
`TDK Corporation Exhibit 1010 Page 12
`
`

`
`9377
`Growth of thin films
`example, it is now well established that the c(2 (cid:2) 2) reconstruction observed upon annealing
`of 1=2 ML of many metals is caused by formation of an ordered 2D surface alloy. Although
`there are many studies of the growth of nonmagnetic metals on W(100) and a few on Mo(100),
`Nb(100) and Ta(100) not