`
`VOLUME 47, NUMBER 16
`
`15 APRIL 1993-II
`
`Growth and structure of Fe and Co thin films on Cu(lll), Cu(lOO), and Cu(llO):
`A comprehensive study of metastable film growth
`
`M. T. Kief and W. F. Egelhoff, Jr.
`Surface and Microanalysis Science Division, National Institute of Standards and Technology, Gaithersburg, Maryland 20899
`(Received 2 November 1992)
`The growth and structure of Fe and Co thin films on single-crystal Cu(lll), Cu(lOO), and Cu(llO) sub-
`strates have been investigated using x-ray-photoelectron and Auger electron forward scattering, CO-
`titration, low-energy electron diffraction, and reflection high-energy electron diffraction. The motivation
`for this study is to understand the role of surface structure and kinetics in the growth of metal films on
`metal substrates. The effect of varying substrate growth temperatures between 80 and 450 K plays a
`prominent role in determining both the film morphology and crystalline phase. Nonideal film growth,
`including agglomeration of Co and Fe and surface segregation of Cu, is the rule rather than the excep-
`tion. Simple considerations of surface diffusion and surface free-energy differences provide a basis for
`understanding why layer-by-layer growth is unlikely to occur in these systems and should not be expect-
`ed in many other metastable film-substrate systems.
`
`I. INTRODUCTION
`
`Substantial research has sought to understand the
`diverse structural and magnetic properties of thin films of
`Fe and Co grown on Cu. These systems are employed in
`the research fields of surface magnetism, low-dimensional
`magnetism, magneto-optics, giant magnetoresistance, and
`many others (see Ref. 1 and references therein). The ap-
`plications of this research span from magnetic recording
`media and recording heads to nonvolatile memory chips.
`Unfortunately, the results of these research studies have
`often been contradictory. This confusion is largely due to
`an inadequate understanding of the film growth, film
`morphology, and the delicate interplay between thin-film
`structure and magnetic properties. Therefore, it is timely
`to examine the building blocks of these structures name-
`ly, monolayer films of Fe and Co epitaxially grown on
`Cu(lll), Cu(lOO), and Cu(llO).
`We report the structure and morphology of Fe and Co
`films prepared by molecular-beam epitaxy on single-
`crystal Cu substrates. We interpret these results in terms
`of the film-growth dynamics. To examine the effects of
`substrate structure, the film-growth mode has been stud-
`ied on Cu(lOO), Cu(100), and Cu(111) with varying sub-
`strate preparations. To explore the effects of varying
`growth kinetics upon the system structure, films were
`grown at substrate temperatures ranging from 80 to 450
`K. Presented here is a systematic and comprehensive
`structural study of these metastable systems using several
`complementary techniques including x-ray-photoelectron
`and Auger electron forward scattering, low-energy elec-
`tron diffraction (LEED), reflection high-energy electron
`diffraction (RHEED) and CO titration.
`Epitaxial growth of a metal film on a metal substrate is
`often categorized according to three standard models:
`two-dimensional or Frank-van der Merwe (FM) growth,
`three-dimensional or Volmer-Weber (VW) growth, and
`two-dimensional
`followed by
`three-dimensional or
`
`47
`
`the
`Stranski-Krastanov (SK) growth. According to
`quasiequilibrium description by Bauer,2
`these
`three
`growth modes are governed by the surface free energies,
`the interface free energy, and the strain energy. The de-
`posited film will grow in two dimensions or layer-by-layer
`(FM) if
`
`(1)
`where a f is the deposited film surface free energy, a, is
`the substrate surface free energy, a; is the interface sur-
`face free energy, and a e is the strain energy. Otherwise,
`the film nucleates as three-dimensional clusters (VW). In
`the event that the inequality reverses with film thickness,
`layer-by-layer growth is followed by three-dimensional
`growth (SK).
`A common goal
`two-
`to produce
`is
`in epitaxy
`dimensional film structures with a particular crystallo-
`graphic phase and orientation. To attain this goal, we
`often desire FM growth. However, FM growth is
`difficult to obtain for Fe/Cu and Co/Cu because the sur-
`face free energies3 of Co (2.709 Jm- 2) and Fe (2.939
`J m - 2) are significantly larger than the surface free ener-
`gy of Cu (1.934 J m - 2). In addition, since the heats of
`mixing for both Fe-Cu and Co-Cu are endothermic,4 we
`can expect the interface free energies costs to be unfavor-
`able. According to Eq. (1), the initial equilibrium growth
`of Fe and Co on Cu should be similar to VW, not FM as
`has been frequently reported. 5- 24
`Furthermore, the quasiequilibrium VW growth mode
`predicted by Eq. (1) is frequently not obtained. Non-
`equilibrium growth can occur because kinetic factors
`(such as surface diffusion) are too slow. The actual film
`growth can result in departures from equilibrium struc-
`tures and crystallographic changes [e.g., Fe/Cu(111)). In
`addition, the three idealized growth modes neglect the
`possibility that substrate atoms can be mobile and may
`segregate to the surface during film growth. The impor-
`tance of surface segregation
`is exemplified by
`its
`Work of the U.S. Government
`Not subject to U.S. copyright
`
`10 785
`
`Lambeth Magnetic Structures, LLC Exhibit 2005
`
`LMBTH-000129
`
`
`
`10 786
`
`M. T. KIEF AND W. F. EGELHOFF, Jr.
`
`47
`
`widespread prediction and observation in metallic al-
`loys.4,25-27
`To understand whether equilibrium growth can be ex-
`pected, we must identify the controlling processes. Three
`important processes relevant to metal-film on metal-
`substrate growth are (i) surface adatom diffusion, (ii) sub-
`strate
`surface
`segregation, and
`(iii)
`film/substrate
`interdiffusion. These processes are activated by increas-
`ing the substrate temperature during film growth. For a
`process to be significant, its rate should be compared to
`the film deposition rate, which is typically around 1
`monolayer/min. In addition, it is essential to recognize
`that surface diffusion varies with crystal face and surface
`quality as well as element.
`Activation energy barriers for surface diffusion have
`been measured for some metal/metal systems.28 - 48 Typi-
`cally, the measured surface diffusion barrier ranges from
`near 0.1 to 0.9 e V for different metals and different crys-
`tal faces. The surface diffusion process is assumed to fol-
`law, D 0exp(-Ed/kBT),
`low an Arrhenius diffusion
`where Ed is the activation energy, kB is the Boltzmann
`constant, Tis the temperature, and with a typical preex-
`ponential D 0 of -10-3 cm2/s.28·29·32 Activation energies
`of0.1-0.9 eV translate to about 40-350 K for an adatom
`mobility of 1 hop/s. As an example of the crystal face
`dependence, Ir self-diffusion has been determined to have
`activation energies of 0.27 eV for Ir/Ir(111),30 0.7 eV for
`Ir/Ir(110),48 and 0.84 eV for Ir/Ir(lOO)Y
`The experimentally determined activation energy for
`self-diffusion on Cu(lOO), has been reported to be
`0.28±0.06 (Ref. 49), 0.39±0.06 (Ref. 50), and -0.48
`eV.51 Since other experimental surface diffusion data for
`Cu data are not available, we must rely solely upon
`theoretical estimates for self-diffusion on Cu(111) and
`Cu(110). Recent effective-medium calculations predict
`diffusion barriers for Cu(111), Cu(110), and Cu(lOO) of
`0.13, 0.18, and 0.21 eV, respectively.52·53 Since surface
`diffusion barrier energies are not available for many ma-
`terials, it is useful to have a simple means to approximate
`them. An estimate of the diffusion barrier can be ob-
`tained by scaling the activation energy of an unknown
`material to a known material using the cohesive energies.
`Using the cohesive energy ratio for Cu/Ir=O. 502 (Ref.
`54) gives activation energies of0.14, 0.35, and 0.42 eV for
`Cu on Cu (111), Cu(llO), and Cu(lOO), respectively. The
`Cu(lOO) estimate agrees reasonably with the experimental
`values. However, the theoretically calculated Cu(lOO)
`value of Hansen et al. 52 is roughly half the experimental
`values. 55
`Segregation of the substrate atoms may occur when the
`substrate surface free energy is lower than the deposited
`film surface free energy. It is difficult to estimate the ac-
`tivation energy for this process, since in addition to sur-
`face free-energy differences, heats of solution and the
`elastic size mismatch energy may also play a role.4 Fur-
`thermore, there may be more than one contributing
`segregation path (see Ref. 56 and references therein).
`However, we can expect that segregation may be impor-
`tant for Fe and Co on Cu for growth temperatures near
`and above room temperature because of experimental re-
`ports of significant segregation near 400 K for Fe and Co
`
`on Cu(100).16,57-59
`Finally, the upper temperature limit for layer-by-layer
`equilibrium growth is imposed by requiring an abrupt
`film/substrate interface. Assuming that interdiffusion
`occurs through bulk diffusion and with a typical growth
`time of - 100 s, a growth temperature upper limit is es-
`timated to be near -0.5 the melting temperature. The
`upper limit for Cu, Co, and Fe is about 675, 885, and 900
`K, respectively. Growth of films at temperatures above
`this threshold can often produce interdiffused or alloyed
`layers, even for systems that satisfy Eq. (1).
`The limits imposed by these mobilities, combined with
`the film deposition rate, create a temperature window
`where equilibrium FM growth may occur.60·61 For sys-
`tems, such as Fe/Cu and Co/Cu, which do not satisfy the
`Eq. (1) criteria for FM growth, layer-by-layer growth
`may not exist at all. Therefore, in systems of high-
`surface free-energy metals deposited on low-surface free-
`energy substrates, we have chosen an alternate approach:
`deposit the film at low temperature where thermal
`diffusion is minimal, then anneal the system to a tempera-
`ture that improves the lattice ordering but does not per-
`mit substrate segregation. This
`technique produces
`very-high-quality films.62 Nevertheless, the emphasis of
`the present study is upon examination of the growth pro-
`cess and characterization of the as-grown film/substrate
`system. Careful annealing of a film grown at low temper-
`ature should be considered an additional, valuable tool to
`optimize the film quality.
`In summary, it is essential to consider the growth ki-
`netics in these metastable thin-film systems. Simple con-
`siderations of surface free energies and atomic mobilities
`provide a foundation for understanding the nonideal
`growth modes and structures of metastable Fe and Co
`films on Cu(l11), Cu(lOO), and Cu(110). Furthermore,
`these ideas are expected to be equally valid and applicable
`to other thin-film systems.
`This paper is organized in six sections. The Introduc-
`tion (Sec. I) is followed by a review of the experimental
`methods (Sec. II). Growth on the different substrates is
`examined individually: Cu(lll) (Sec. Ill), Cu(IOO) (Sec.
`IV), and Cu(llO) (Sec. V). Within each of these three sec-
`tions, film growth is discussed (A) in detail for Fe, (B) in
`detail for Co, and (C) in general for both Fe and Co with
`an emphasis on common aspects. A brief, general con-
`clusion with specific highlights is given in Sec. VI.
`II. EXPERIMENT
`
`The molecular-beam epitaxy (MBE) system used in this
`work has two ultrahigh vacuum chambers. A long-stroke
`sample manipulator traverses the central axis of the
`chambers. The sample can be cooled to 80 K, heated to
`1000 K, and rotated by a stepper motor about the main
`axis of translation. The first chamber contains the
`sputter ion gun, quadrapole mass spectrometer, LEED
`system, RHEED system, and metal MBE sources. The
`system base pressure is 8 X 10-9 Pa. A second adjacent
`chamber is equipped with a commercial x-ray source and
`a 150-mm mean-radius hemispherical electron-energy
`analyzer with input electron optics. The x-ray source and
`analyzer axis are 90• apart in the plane defined by rota-
`
`LMBTH-000130
`
`
`
`47
`
`GROWTH AND STRUCTURE OF Fe AND Co THIN FILMS ON ...
`
`10 787
`
`tion of the sample normal. The x-ray photoelectron spec-
`troscopy (XPS) and Auger data were obtained with AI
`Ka (1486.6 eV) radiation. The analyzer has a 16-channel
`parallel detector for improved signal to noise. The
`geometric acceptance angle of the input electron-optics is
`±5•, which was satisfactory for routine XPS and CO-
`titration measurements (described below). Improved an-
`gular resolution was desirable for forward-scattering
`measurements and was obtained by inserting an addition-
`al aperture, which reduces the acceptance angle to ±2. 5•.
`The estimated absolute angular accuracy of these mea-
`surements is ± 1 • with significantly better relative accura-
`cy of ±0.25•. Further details of the MBE system are
`given in previously published work. 62
`All copper substrates were cut from a single-crystal
`boule using a wire-slurry saw. Each crystal was oriented
`to better than 0. 5• with a diffractometer. The mechani-
`cally damaged layer was then removed from both sides of
`the substrate using an acid polishing instrument. 63 The
`crystal orientation was
`then
`rechecked with
`the
`diffractometer. A final, near-mirror finish was obtained
`by a brief, manual acid polish. The acid polishing solu-
`tion64 is formulated to produce optically fiat metal sur-
`faces and is based upon a solution of HCI acid, po-
`lyethylene glycol, and 2-mercaptobenzimidazole saturat-
`ed with CuC12• The removal of much of the damaged
`layer at the crystal surface is demonstrated by the obser-
`vation of a weak LEED pattern without any further
`cleaning or annealing. To remove impurities, the crystal
`was sputtered using Ar+ orNe+ and annealed to -900
`K until no contamination was detectable with XPS and a
`sharp pattern was obtained with LEED.
`All films were grown in a vacuum of 1 X 10-8 Pa or
`better. Fe was evaporated from a metal oven described
`elsewhere.65 Cobalt was deposited from a tungsten-wire-
`filament evaporator. Typical film growth rates were -2
`monolayers/min as measured by an ion-gauge integration
`system.62 Film thicknesses were measured by two quartz
`crystal monitors, symmetrically adjacent to the sample.
`The calibration of these monitors was . done using
`RHEED oscillations. The average film thickness accura-
`cy is ±0.10 monolayers (ML) on Cu(lOO) and Cu(111)
`and ±0.15 ML on Cu(110).
`
`Films were routinely checked with XPS to monitor
`film purity. No attempts were made to correlate film-
`growth mode with XPS intensity or Auger kinks on ac-
`count of the dubious nature of this practice for many
`metal/metal epitaxial systems.66•67 LEED was used to in-
`vestigate film structure and morphology.
`The crystal structure of the film was investigated using
`x-ray photoelectron and Auger electron
`forward-
`scattering measurements. This technique has recently
`been used to study many metal-film/metal-substrate sys-
`tems by several groups. 57•68 - 71 The primary advantage of
`this method is its elemental specificity combined with its
`real-space correspondence to near-neighbor bond direc-
`tion
`through enhanced forward-scattering
`intensity.
`Strictly speaking, this interpretation of the electron for-
`ward scattering is accurate only for electron kinetic ener-
`gies of several hundred eV or greater. At lower kinetic
`energies, the electron scattering may be dominated by
`multiple-scattering effects that distort and obscure simple
`interpretation. A consistency check is provided by com-
`paring the XPS and Auger angular anisotropies for
`several different kinetic energies :::>: 0. 5 ke V: true bond
`directions will exhibit intensity enhancements indepen-
`dent of the kinetic energy. Therefore, it is generally
`straightforward to determine the crystal structure of the
`film from fast and simple XPS or Auger electron angular
`anisotropy measurements. For pertinent reviews on for-
`ward scattering as a diagnostic tool see Refs. 68, 69, and
`72.
`The crystal structure of the film can also be ascertained
`by a comparison of XPS and Auger angular anisotropies
`from deposited films with those observed from pure single
`crystals. This has the inherent advantage of including
`multiple-scattering and interference effects. Figure 1
`shows polar XPS and Auger angular anisotropies for Cu
`single crystals with surfaces oriented along the (100),
`(110), and (111) directions. The XPS and Auger angular
`anisotropy is defined as the angular-dependent intensity
`divided by the maximum intensity in the angular scan.
`The similarity between the anisotropies at the four kinet-
`ic energies, Cu 2p 312 (552.6 eV), Cu L 3M 45M 45 (916.6
`eV), Cu 3s (1362.6 eV), and Cu 3p 312 (1409.6 eV), for each
`Cu crystal provides an excellent demonstration of the
`
`FIG. 1. XPS angular anisotropy vs polar
`angle for Cu(lOO), Cu(llO), and Cu(lll) in the
`(001 ), (Oil), and ( 121) azimuths, respec-
`tively, using AI Ka (1486.6 eV) radiation.
`Spectra are shown for four Cu x-ray and
`(552.6 eV), Cu
`Auger energies: Cu 2p 312
`L 3M 45M 45 (916.6 eV), Cu 3s (1362.6 eV), and
`Cu 3p 312 (1409.6 eV). Nearest-neighbor direc-
`tions are indicated by vertical dotted lines.
`Anisotropy is defined for each energy as (angu-
`lar intensity)/(maximum peak intensity). The
`Cu(lOO) and Cu(llO) o• peaks appear asym-
`metric due to the grazing incidence of the AI
`K a radiation for this orientation.
`
`0 1 0 20 30 40 50 60 70 80 0 1 0 20 30 40 50 60 70 80 0 1 0 20 30 40 50 60 70 80
`Angle (deg)
`Angle (degl
`Angle (degl
`
`LMBTH-000131
`
`
`
`10 788
`
`M. T. KIEF AND W. F. EGELHOFF, Jr.
`
`47
`
`forward-scattering phenomena. The indicated directions
`correspond to the crystal normal and to the strong
`forward-scattering directions that are associated with
`nearest- and next-nearest-neighbor axes. Scattering from
`more distant neighbors typically shows weak or no
`detectable enhancement unless it is compounded by first-
`order constructive interference effects. 69 A good example
`of this is the (100) crystal that has consistently strong
`forward-scattering peaks for all energies along [100] (0°)
`and [101] (45°) and additional features near 20° and 70°,
`which disperse slightly with energy. The 20° peak has
`been shown by simulations 73 to result from both a weak
`forward-scattering intensity along the [103] ( 18.4°) direc-
`tion and a first-order interference maximum coincidental-
`ly near 20", which disperses with electron kinetic energy.
`The peak near 70° also corresponds to a weak forward-
`scattering peak combined with a first-order maximum
`and is symmetric with the 20° maximum about the [101]
`direction. Similar interference maxima can also be ob-
`served in the other crystals, symmetric about the ( 110)
`directions.
`Forward scattering serves to characterize the film crys-
`tallography, but does not provide direct elemental infor-
`mation about the surface layer. To answer difficult ques-
`tions on film agglomeration and Cu surface segregation, a
`technique was developed to measure what fraction of the
`surface was exposed Cu or was "surface Cu." This pro-
`cedure is referred to as CO titration and has been intro-
`duced previously.62 The procedure is based upon the sur-
`face core-level shift of the Cu 2p 312 state with adsorption
`of CO. Since only those Cu atoms exposed at the surface
`will have core-level shifts, the fraction of the surface that
`is Cu can be estimated by reference to a clean Cu sub-
`strate. To determine the amount of surface Cu for a par-
`ticular sample, we deposit the film of Fe or Co and then
`measure the Cu 2p 312 peak: (a) without CO, (b) with a
`saturation dose of CO at ~ 80 K, and (c) after warming
`to 300 K to desorb the CO from the Cu. The Cu 2p 312
`peak contains two contributions: (i) the signal from the
`surface Cu atoms, which shifts with CO adsorption, and
`(ii) the signal from nonsurface Cu atoms, which does not
`shift. To eliminate attenuation by the CO, the peaks are
`normalized to constant area, then the difference spectra
`(a)- (b) and (c)- (b) are calculated. (a)- (b) is called the
`adsorption cycle and (c)- (b) the desorption cycle. The re-
`sulting difference curves show a trough/peak shape that
`represents of the number of CO-shifted Cu surface atoms.
`The height of the difference curve for a given film is then
`compared to identical measurements on a clean Cu sub-
`strate with no film. The film/no-film height ratio corre-
`sponds to the fraction of the surface that is Cu. The ad-
`
`sorption cycle and desorption cycle estimates should be
`identical, within the estimated measurement uncertainty
`(±5%), if the number and kind of surface atoms remains
`constant after annealing to 300 K with adsorbed CO.
`Differences in the measurement cycles indicate undeter-
`mined instabilities, including film agglomeration and sub-
`strate segregation. We report both estimates to demon-
`strate the systematics of these measurements. However,
`when the adsorption and desorption measurements differ
`by more than the measurement uncertainty, we take the
`average as our best estimate of surface Cu and qualify
`these systems as metastable.
`Generally, CO-titration measurements are made by
`measuring the Cu 2p 312 intensity at a polar angle of 25°.
`For ideal flat films, measurements should show no change
`with detection angle in the fraction of the surface that is
`Cu surface. However, nonideal film growth can produce
`a variation with detection angle in estimated Cu. This is
`because measurements performed at 5° off the surface
`normal integrate contributions from all the Cu surface
`atoms equally, while measurements made near 80°, for ex-
`ample, will be less sensitive to Cu surface atoms at the
`bottom of cracks in the film. Therefore, additional
`angular-dependent CO-titration measurements were oc-
`casionally performed at 5", 45°, 65°, and 80°. The varia-
`tion in angle of the Cu estimate was then interpreted in
`terms of the distribution of the Cu surface atoms and the
`film morphology.
`III. GROWTH OF Fe AND Co ON Cu(lll)
`
`The growth of Fe and Co on Cu(lll) is a particularly
`rich system that has received considerable attention in
`the past. The structure of Fe films on Cu(lll) has been
`studied using electron microscopy/4- 78 field-ion micros-
`copy,70 LEED, and Auger electron
`spectroscopy
`(AES). 10•80·81 The magnetic properties have been studied
`torque magnetometry,5- 7·82 Mossbauer/6 and
`using
`electron-capture spectroscopy.8 Similar structural stud-
`ies of Co growth on Cu( Ill) have used LEED and AES, 18
`and nuclear magnetic resonance (NMR),83 - 87 surface-
`extended x-ray-absorption fine structure (SEXAFS),88·89
`x-ray scattering,90 and XPS forward scattering.91 The
`magnetic properties of Co/Cu(lll) have been measured
`by torsion magnetometry,92 ultraviolet photoelectron
`spectroscopy, 18•93 surface magneto-optic Kerr effect,94·95
`and torsion oscillation magnetometry.96
`Table I shows that there is a very close lattice match to
`the Cu(lll) surface net for both fcc and bee phases of Fe,
`and both fcc and hcp phases of Co. In addition to the
`lattice match, the crystalline phase of the film is con-
`trolled by the epitaxial strain in the film. The equilibrium
`
`Material/
`symmetry
`fcc Cu(111)
`fcc Fe(lll)
`bee Fe(110)
`fcc Co(lll)
`hcp Co(0001)
`
`TABLE I. Epitaxy of Fe and Co on Cu(111).
`Lattice
`Nearest
`Surface cell
`constant <A)
`neighbor <Al
`mismatch (%)
`3.61
`2.55
`3.59
`2.54
`2.87
`2.48
`3.54
`2.50
`2.51
`
`-0.8
`+3.4
`-3.9
`-3.2
`
`Interlayer
`spacing <A>
`2.08
`2.07
`2.03
`2.05
`2.03
`
`LMBTH-000132
`
`
`
`47
`
`GROWTH AND STRUCTURE OF Fe AND Co THIN FILMS ON ...
`
`10 789
`
`configuration of the strained film is predicted by minimiz-
`ing the system free energy for all possible film crystalline
`phases, orientations, strains, and dislocations,97•98 How-
`ever, even this model of lattice mismatch and film elasti-
`city is oversimplified, because it only considers continu-
`ous films. We will show below that there is no single epi-
`taxial phase for Fe and Co on Cu(lll). The epitaxial
`phase of the film depends on the growth temperature and
`the film thickness.
`
`A. Fe/Cu(lll)
`Fe films were prepared at substrate growth tempera-
`tures of 80 and 300 K. The structure of these films was
`examined for thicknesses up to 8 ML. The fraction of Cu
`in the exposed surface was measured with both the ad-
`sorption and desorption CO-titration sequence as de-
`scribed above. These values were typically found to agree
`within 5%, which we believe is near our experimental ac-
`curacy. Therefore, we only report the average values.
`Figure 2 shows the fraction of Cu in the exposed surface
`versus film thickness. The fraction of Cu in the exposed
`surface increases with growth temperature and decreases
`with deposited Fe thickness. The data show that after
`deposition of about 3 ML of Fe grown at 80 K and 5 ML
`of Fe grown at 300 K the surface is 5% and 12% Cu, re-
`spectively. This agrees well with scanning tunneling mi-
`croscopy (STM) estimates of 10% of the Cu substrate ex-
`posed for 4 ML of Fe deposited at 300 K. 99 Further-
`more, annealing these films shows that they are stable.
`This thermal stability is demonstrated by the observa-
`tions that a 2.3-ML grown at 80 K and a 5.6-ML film
`grown at 300 K show no increase in the fraction of Cu in
`the exposed surface for anneals of 300 K over the growth
`temperature. The large fraction of Cu at the surface indi-
`cates the Fe film growth is not an ideal layer-by-layer
`manner. This growth mode is in contrast to most of the
`
`100
`
`80
`
`60
`
`" u
`""
`u
`~ (f)
`c "' 40
`[!
`"' a_
`
`20
`
`0
`
`0
`
`IFe/Cu(111 >I
`
`0
`
`•
`80 K
`o 300 K
`
`--- ·- Poisson
`
`•..
`
`0
`
`0
`
`0
`
`0
`
`2
`
`3
`Deposited Fe (ML)
`
`4
`
`5
`
`6
`
`FIG. 2. The fraction of Cu in the surface for Fe deposited on
`Cu(lll) at 80 and 300 K. Coverage is determined by the CO-
`titration technique using the average of adsorption and desorp-
`tion measurements for 25• detection angle. Solid curves show
`an exponential fit to the drop in the measured fraction of Cu in
`the surface. The dotted line indicates the fraction of Cu in the
`surface for random substrate coverage according to Poisson
`statistics (see text for interpretation).
`
`reports room-temperature FM
`
`literature,5- 10 which
`growth of Fe/Cu(lll ).
`For comparison to the measured values, Fig. 2 shows
`the predicted values of the fraction of Cu in the exposed
`surface for the Poisson model. The Poisson model as-
`sumes random deposition on a simple cubic lattice. The
`direct comparison of the measured values with the Pois-
`son model should be done with caution because the sim-
`ple cubic lattice does not explicitly include the fcc(lll)
`threefold adsorption geometry or stacking faults. If the
`adatoms have zero mobility, this model implies unphysi-
`cal vacancies and overhangs. Relaxing this constraint
`and allowing the second layer adatoms to drop down into
`one of three possibly unfilled nearest-neighbor sites in-
`creases the substrate coverage to 52%, 21%, and 4% for
`depositions of 0.5, 1.0, and 2.0 ML, respectively. Com-
`paring these new estimates with the simple cubic model
`prediction (Fig. 2) decreases the agreement, particularly
`for depositions above 1.5 ML. Alternatively, if the ada-
`toms are somewhat mobile but cannot diffuse over a step
`and (or) the coverage is not random but locally correlated
`as occurs with small cluster nucleation, the simple cubic
`lattice Poisson estimate may be more appropriate. Exam-
`ples of these types of growth are Pt/Pt(lll) at 400 K
`(Refs. 100 and 101) and Fe, Co, and Cu on Cu(lOO) at 80
`K. 102 Keeping these possibilities in mind, we have chosen
`always to plot the Poisson model estimates with the CO-
`titration measurements. The Poisson model estimate
`serves as a guideline for comparison between different
`materials, symmetries, and growth temperatures.
`The crystalline structure of the Fe film was determined
`using forward scattering. Figure 3 plots the XPS angular
`anisotropies of Fe 3p 312 (1431.6 eV) for films grown at 80
`and 300 K and for depositions from about 1-7 ML.
`The films deposited at low temperature, 80 K, show the
`evolution of a bee Fe phase, which is indicated by a peak
`in the XPS angular anisotropy at 45•. The weak, broad
`rise near 45• with no o• feature show the 1.0-ML film is
`nearly flat. The rise of a peak at o• for a 2.3-ML film
`signifies the start of the third bee layer, which is con-
`sistent with nearly layer-by-layer growth. Increasing the
`film thickness shows increasing structure in the angular
`anisotropies that indicate a bee film.
`A 1.0-ML film deposited at 80 K has a LEED pattern
`that is p ( 1 X 1 ) and is threefold symmetric. The LEED
`spots alternate between fuzzy and sharp as a function of
`the beam energy. A 2.3-ML Fe film has a similar LEED
`pattern with a brighter background and very broad spots.
`These LEED patterns indicate that the atoms sit largely
`in lattice sites but many steps are present. Films that are
`annealed to 350 K have sharpened LEED spots and a
`clearer three-fold symmetry. In contrast, the XPS angu-
`lar anisotropy shows little change in structure for films
`that are annealed, indicating that short-range order in the
`films changed little. A film thicker than 4 ML has a
`LEED pattern that is sixfold symmetric with broad spots.
`In addition, the LEED pattern has new diffuse spots
`(these spots are similar to those labeled B in Fig. 4, which
`are observed in room temperature grown films). Anneal-
`ing the films sharpens the LEED pattern but less so for
`thicker Fe films. A 5.6-ML film grown at 80 K and
`
`LMBTH-000133
`
`
`
`10 790
`
`M. T. KIEF AND W. F. EGELHOFF, Jr.
`
`47
`
`fcc (111)
`
`(c)
`
`35°
`oor~J•o
`o• •o•
`eo• eoe
`•o•o•o
`ggqoq(g
`OJ~gJ~
`
`bee (110)
`oo
`
`450
`
`FIG. 3. XPS angular anisotropy for Fe films
`grown on Cu(l11) at (a) 80 K and (b) 300 K us-
`ing Fe 3p 312 photoelectrons. Schematic dia-
`grams
`indicating directions of scattering
`enhancement for fcc (111) and bee (110) are
`shown in (c). The data were recorded in the
`( 121) azimuth of Cu(l11). Note that the true
`bee Fe (001) azimuth is 5.3• off this ( 121) az-
`imuth. The shaded atoms lie in a layer below
`the plane of the page.
`
`0 1 0 20 30 40 50 60 70 80 0 1 0 20 30 40 50 60 70 80
`Angle (deg)
`Angle (deg)
`
`briefly annealed to 600 K has a LEED pattern similar to
`a 6-ML film grown at room temperature.
`The films deposited at room temperature, 300 K, show
`a more complex growth process than the films deposited
`at low temperature [see Fig. 3(b)]. The XPS angular an-
`isotropy of a 1.0-ML film (not shown) and a 1.4-ML film
`of Fe show a very strong peak near 37•. The strength and
`width of this peak suggest a substantial number of Fe
`atoms have atoms that lie above them. Figure 3(c)
`demonstrates that fcc(lll) should have a peak in the an-
`gular anisotropy at 35. 3•. Therefore, Fe deposited at 300
`K grows in a distorted fcc phase with an approximate in-
`terlayer compression of 3-6%. This relaxation is near
`the previously reported value of 2.5%. 10 A 2.1-ML film
`has a decreased angular anisotropy. The 3 7• fcc peak
`persists in the angular anisotropy, but there is also a
`broad peak at o•, which is not found in fcc(lll). A 3.8-
`ML film has an angular anisotropy lacking discernable
`structure, except a clear peak at o·. This implies the
`structure of the Fe film is neither purely fcc nor bee. It is
`conceivable that there could be domains of fcc and bee,
`and a summation of the anisotropies for the 2.1- and 7.8-
`
`ML films produces an angular anisotropy similar to the
`3.8-ML film. A 5.8-ML film of Fe (not shown) has an an-
`gular anisotropy that indicates the film has converted en-
`tirely to a bcc(llO) structure. This bee structure persists
`for thicker films. Comparing a 7.8-ML film grown at 300
`K to a 7.2-ML film grown at 80 K shows that the room-
`temperature film is similar to the cold film only better or-
`dered as expected.
`LEED patterns for room-temperature-grown films also
`show a gradual transition in structure with increasing Fe
`thickness. Films less than 2 ML thick have a LEED pat-
`tern that is p ( 1 X 1 ) with broadened spots and a strong
`threefold symmetry. This pattern corresponds to fcc Fe.
`Films more than 2 ML thick have a LEED pattern with a
`pair of weak, elongated spots (labeled A in Fig. 4), which
`develop inside the first-order Cu spots. Increasing the Fe
`film thickness further decreases the Cu LEED spot inten-
`sity and produces an additional pair of LEED spots out-
`side the first-order Cu spots (labeled B in Fig. 4) and a
`spot intermediate between the A spots. Films 8 ML
`thick have the LEED pattern shown in Fig. 4. This pat-
`tern can be interpreted as a bee (110) Fe str