Epitaxial growth of fcc Fe on Cu(100)

Surface Science 179 (1987) 219-229 North-Holland, Amsterdam

EPITAXIAL

GROWTH OF fee Fe ON Cu(100)

M. ONELLION, Department

M.A. THOMPSON,

J.L. ERSKINE

of Physics, University of Texas, Austin,

C.B. DUKE Xerox

219

TX 78712, USA

and A. PATON

Webster Research Center, Webster, NY 14627, USA

Layer-by-layer epitaxial growth of Fe on Cu(100) is reported. The epitaxy is characterized using Auger electron spectroscopy and low energy electron diffraction intensity analysis. Good quality epitaxiaf Fe films having thicknesses ranging from one to four monolayers are stabilized by the Cu(100) substrate. The overlayer structure is shown to be nearly identical to a continuation of the fee lattice of the substrate.

1. Introduction One of the most interesting aspects of modern materials science is the ability to stabilize new phases of matter by epitaxial growth. Molecular beam epitaxy (MBE) offers enormous opportunities for synthesizing metastable crystalline phases with novel physical properties [l]. Related to epitaxial growth and the possibilities of creating novel structures are issues pertaining to bulk and surface magnetism and the relationship between magnetic behavior and crystal structure [2]. The existence of metastable phases of 3d transition metals is of particular interest because these phases are intimately connected with the occurrence of ferromagnetism in certain elements in this series [3]. In the case of bulk Fe, the ferromagnetic bee a-Fe phase persists from low temperatures up to 910°C where a phase transformation to antiferromagnetic fee y-Fe occurs. This phase persists to 1390°C where it transforms to &Fe which is a nonmagnetic bee phase. Thus the ability to stabilize and characterize various phases of Fe offers interesting possibilities for studies of magnetism and magnetic properties. Dramatic effects have been predicted recently for thin epitaxial structures of transition metals on noble metal crystal surfaces [4]. The calculations suggest the possibility of inducing two-dimensional magnetism in a controlled manner based on the type of thin film materials engineering made possible by MBE. These all-electron local-spin-density calculations predict large ferromagnetic moments in various monolayer films (Cr, V, and Fe) on noble metal 0039-6028/87/$03.50 0 Elsevier Science Publishers B.V. (North-Holland Physics Publishing Division)

220

M. Onellion et al. / Epitaxral growth of Jcc Fe on Cu(I Gil)

substrates (Cu, Ag, and Au), and corresponding novel effects in modulated epitaxial layers. Based on these results, we have begun an extensive investigation, using angle resolved photoe~ssion, of thin epitaxial magnetic film systems. One of the key elements required for interpreting our results is precise knowledge of the structure of the films. This paper presents results we have obtained in our investigation of Fe layers deposited on Cu(100) surfaces. Our results show that good quality fee iron crystals can be stabilized at low coverages on the (100) face of Cu.

2. Ex~riment~

methods

Previous results suggest that Fe grows epitaxially on Cu substrates. Haase [5] reported that 10 A layers of Fe, deposited on single crystal films of Cu held at 35O”C, yielded reflection electron diffraction patterns which were indistinguishable from the uncoated Cu surfaces. This result indicated that the layers nucleated fee y-Fe on Cu, and that the lattice parameter of the film was nearly identical to that of the Cu substrate. Jesser and Matthews [6] repeated these experiments under better (5 X lo-* Torr) vacuum conditions, and used transmission electron diffraction techniques to verify that fee Fe formed with a strained lattice which exactly matched the Cu(100) substrate. Long straight misfit dislocations were observed at greater film thickness (2 1000 A) when the films relaxed to accommodate the difference in lattice constant between fee y-Fe and Cu. Our interest in the epitaxial growth of fee y-Fe on Cu(100) lies in the low coverage regime (one to five monolayers) corresponding to parameters of the available thin film calculations [7,8]. We are especially concerned with the possibility of island formation at coverages near one monolayer (ML), as well as the possibility of some degree of interfacial mixing. These issues have not been adequately dealt with in the previous studies [5,6,9] of the nucleation and growth of Fe on Cu(100). Our experiments were carried out in a spectrometer which has been described previously [lo]. The spectrometer combines the standard sample preparation and characte~zation tools (sputtering, annealing low-energy electron diffraction (LEED) and Auger analysis) with MBE capability in the preparation chamber which is coupled to an angle-resolving photoelectron spectrometer. The sample manipulator provides linear translation (to access multiple analysis points) as well as the two rotational degrees of freedom required to precisely align the crystal to yield identical intensity data for conjugate LEED beams. The evaporation source incorporates water cooled shields and electrodes to maintain low wall temperatures and low pressures during evaporation of the Fe layers. A quartz crystal microbalance, nude ion

M. Onellion et al. / Epiiaxial growth of fee Fe on Cu(lO0)

221

gauge and thermocouples attached to the substrate and to the evaporation filament permitted us to monitor important parameters during film growth. The Cu crystal was cut and aligned using X-ray Laue techniques to k lo of the (100) direction and cleaned in situ via Ne+ sputtering (1 keV, 1 PA/cm*) at (600” C) substrate temperature followed by annealing above 700°C. Clean Cu(100) surfaces were well ordered as shown by sharp LEED spots, and exhibited a total impurity level of < 1% ML (primarily oxygen) as determined by Auger analysis. Epitaxial Fe layers were grown under “MBE” conditions (p = lo- lo Torr, from a resistively heated W filament. The rate = few A/mm) by evaporation best epitaxial layers were obtained by evaporation onto a heated substrate (150-2OOOC). Substrate temperatures above 240°C tended to promote interdiffusion of Fe and Cu at the surface as shown by Auger analysis described later.

3. Results Fig. 1 illustrates the results obtained from a sequence of 17 evaporations in which LEED and Auger analysis was correlated with the change in quartz oscillator frequency. In these experiments the 47 eV Fe Auger line and 60 eV Cu Auger line were chosen to monitor the coverage because these are more surface sensitive than higher energy lines. The peak-to-peak intensity of the 47 eV Fe Auger line in the derivative of electron yield is designated F in the figure. The inset in fig. 1 indicates the LEED pattern observed at - 0.5-0.9 ML coverage. The (1 X 1) LEED pattern was observed to fade slightly in the coverage range 0 < 8 < 0.5 ML. In the coverage range of 8 = 0.5 ML, a new LEED pattern was observed (refer to inset). At 0 = 1.0 ML coverage, the LEED pattern reverts back to (1 X l), and remains (1 X 1) for all higher coverages. The new LEED pattern at coverages intermediate to 1 ML constituted one estimate of the thickness corresponding to the parameter F. An independent thickness calibration was established using Auger peak analysis and the observation in a change of slope in the parameter F as a function of coverage as described below. An analysis procedure was used to interpret our Auger data obtained during film growth. Our primary goal was to establish a layer-by-layer growth mechanism, to search for island formation at low coverages (i.e., in the 8 = 1.0 ML thickness range), and to investigate the substrate temperature range over which epitaxial growth occurred without being accompanied by detectable interdiffusion. Our Auger spectra were obtained using a Varian 4-grid LEED optics at beam voltages of 2000 V, beam currents of 100 PA and fixed modulation

222

M. Onellion et al. / Epitaxial growth of Jcc Fe on Cu(lOOj (10) l

0

0 l

Fe/Cu(lOO)

l

0

0 0

0

l

0

l

0 0

l

P(lXl)

0

10

CHANGE

(01)

0

/

P(lXl)

20

30

40

IN OSCILLATOR

60

70

FREQUENCY

50

(Hz)

Fig. 1. Height of the Fe(47 eV) Auger peak-to-peak intensity (parameter F) as a function of change in quartz crystal oscillator frequency (Hz) for a series of evaporations. The inset illustrates the LEED pattern observed at approximately 0.5-0.9 ML coverage.

amplitudes. Analysis of the Auger spectra [ll] defined as Fe(E,)

R=Fe,(E,)

Cu,(4)

1

CU(&) 5'

was based on a parameter R

(1)

where Fe(E,) (Cu(E,)) and Fe,(E,) (Cu,(E,)) are the Auger signal peak-topeak amplitudes for Fe (Cu) after an evaporation and for pure samples, respectively. These amplitudes are published for pure Cu and Fe [12]. In our analysis we have normalized the Auger amplitudes to account for the known differences between measurements obtained using a retarding field analyzer and the cylindrical mirror analyzer used in nonretarding mode [13]. We have used the density of fee Fe throughout our analysis since, as discussed below, the Fe grows as a fee phase for l-4 ML. E, ( E2) designate the energies of the Auger peaks. The two ratios we used were Fe(47 eV):Cu(60 eV) and Fe(703

223

M. Onellion et al. 7 Epitaxial growth of fee Fe on Cu(100)

eV): Cu(920 ev>. The parameter Q denotes evaporation and is calculated using:

Q= Fe/S,,

+ O/So + C/&

+ s/%

the

fraction

of iron

in

the

’

where Fe, 0, C, and S represent the Auger peak height of all constituents observed in the evaporated films, and Sx is the relative sensitivity of element X. Theoretical values for R were calculated in the same manner as by other investigators [11,14] with one exception. Instead of using an effective cos 8 normally used to account for various types of angle dependent analyzer configurations, an integration over angle was performed numerically using a five-point gaussian quadrature with limits of 6” and 60°. To calculate the theoretical R values, it is necessary to determine the inelastic mean free path (IMFP) for Cu and Fe Auger electrons. We also included an IMFP parameter for the incident (2000 ev> electrons (X(FLUX) to account for the incident beam attenuation. There is a wide range of IMFP values in the literature; the IMFP values we used are: X(Cu(60)) X(Cu(920))

= 3.50 A, = 8.25 A,

h(Fe(47)) X(Fe(703))

= 3.10 A, = 7.00 A,

x(FLUX) x(FLUX)

= 100 A; = 25 A;

the X(FLUX) is larger for the Cu(60) : Fe(47) ratio than for the Cu(920) : Fe(703) ratio since the mean energy of the incident electrons will decrease with penetration depth. Variations of up to f15% in the mean free paths did not affect our conclusions; however, larger variations did affect our analysis by introducting disagreements in the theoretical ratio versus thickness curves for the two ratios. Further, we did all Auger analysis and measurements at room temperature. We observed interdiffusion at 250°C as previously noted; this interdiffusion obviated our analysis. At reduced temperatures ( - 150°C) we still obtained significantly less self-consistency than at room temperature. Having obtained the theoretical ratio versus thickness curves for the two ratios, we analyzed each evaporation. For each evaporation, we measured the Fe(47 ev), Cu(60 eV), Fe(703 eV), and Cu(920 eV) Auger peak heights. Using the inelastic mean free paths noted above we calculated the thickness separately using the pairs Fe(47 eV) : Cu(60 eV) and Fe(703 eV) : Cu(920 eV). We used two models for the growth kinetics. The first assumed perfect layer-bylayer growth, with each layer completed before the next layer was begun. The second model assumed a small amount of island formation, in which the last layer was only 85% complete when the next layer was begun. Because of the large difference in inelastic mean free paths between the low energy and high energy Auger pairs, this was the most stringent test we could devise for

224

M. Onellion et ul. / Epitaxiol growth offa

4

-

.

LAYER-BY-LAYER

0

ISLAND

GROWTH

Fe on Cu(lO0)

MODEL

I I

FORMATION

GROWTH

MODEL , ; /I

0

3’

2

-

1 -

L 1

NUMBER Fig. 2. Number ratio (horizontal

OF MONOLAYERS

2

3

(Fe 47 I Cu 60)

of monolayers deduced for various evaporations from the Fr(47 eV) : Cu(60 eV) axis) and the Fe(703 eV) : Cu(920 eV) ratio (vertical axis). Layer-by-layer (Of and slight island (0) models used (see text).

determining the growth characte~stics. The results are illustrated in fig. 2 for a set of evaporations ranging from submonolayer to more than three monolayers coverage. The horizontal axis denotes the number of monolayers we deduced by using only the low energy Auger pair. The vertical axis denotes the corresponding number of monolayers deduced using only the high energy pair. The solid line denotes perfect agreement, which we should obtain if the data and model are flawless. Our limits of imperfection (+ 15%) are denoted by dashed lines. The results are denoted by solid or open circles. The solid circles correspond to using our perfect, layer-by-layer growth model. By taking the same data and applying our island formation model we obtained the results illustrated by open circles. Several noteworthy points emerge. Of the 25 evaporations, only two lie outside our f 15% limits if perfect, layer-by-layer growth is the model used.

M. One/lion et al. / Epitaxial growth offec Fe on Cu(1CQ)

225

At low coverages, less than 1 ML, the two models give essentially the same results. To avoid obscuring the results illustrated in fig. 2, we have not drawn in the island growth model (open circle) results for less than 1 ML. By contrast, above 1 ML, 9 of the I2 evaporations yield results outside our + 15% limits if we used the island growth model, including three evaporations (the thickest) for which results do not fit on the scale of fig. 2. Above 3.5 ML, the island growth model gave very bad agreement, and any model we devised that assumed a iarger amount of island formation consistently yielded worse agreement between what the low energy Auger pair and high energy Auger pair imply for the number of monolayers. The conclusion is clear: Fe grows on Cu(100) as epitaxial fee, predominantly in the layer-by-layer growth mode, up to 4 ML. We say “predominantly” because our data is not good enough to exclude a circa 5% deviation from layer-by-layer growth. The same conclusion was reached by Lee et al. [9] based on plotting the variation of the Fe (651 eV) Auger peak and Cu(920 eV) peak with time and comparing the result with an exponential decay model based on the assumption of layer-by-layer growth. In order to check the overlayer crystallographic structure, we measured LEED intensity versus voltage (I- I’) spectra for first and second order beams for clean Cu(100) and 1-4 ML overlayer of Fe. The results for clean Cu(100) illustrated in fig. 3, are in excellent agreement with those of other investigators [15]. We performed LEED structural calculations and determined that our i t

I

I

I ~(1x1)

I

I

I

CALCULATED EXPERIMENTAL

Fe/Cu(lOO)

I -_

I

(101 BEAM

2 MONOLAYERS

I 30

I

I

Fe

I

90

I

I

I

150 ELECTRON

ENERGY

210

I 270

(evt

Fig. 3. Comparison of LEED intensity calculations (dashed lines) with experimental results clean Cu (10 beam), 1 ML Fe, and 2 ML Fe. The calculation assumes unreconstructed ~rou~out and a single Debye temperature of 343 K.

for fee

226

M. Onellion et al. / Epituxial growth of Jcc Fe on Cu(lO0)

agreement with the calculations was limited by the number of LEED 1-V spectra measured and uncertainties in the background subtraction. As illustrated in fig. 3 for instance, we obtained a sufficient description of the measured intensities by varying only the non-structural parameters (the bulk and surface Debye temperatures) to conclude that to within the uncertainties in the data the Fe grows epitaxially on fee Cu with a vertical lattice constant within 0.05 A that of Cu. As is well known, the first monolayer of a bee or fee film will occupy the same site. There is a significant difference in the fee and bee iron interatomic spacing. That difference allows us. to conclude that the films, since they are in registry with the substrate for one or more monolayers, are fee even at just above 1 ML. In fig. 4 we illustrate that the agreement

RIGID

%

LAlTlCE

13~’ 343 K

L

I

I

I

ELECTRON

b

’ p(W)

I

I

I

I

I

I

I

150 ELECTRON

Fig. 4. Comparison Debye temperature

I

CALCULATED EXPERIMENTAL

90

I 270

(eV)

I

I

Fe/Cu(lOO)

I

I 210

ENERGY

i?-

30

I

150

90

30

I

ENERGY

210

I _ ____

-j

I 270

(ev)

between LEED intensity calculations assuming a rigid lattice or a single and experimental results (dashed lines) for p(1 xl) Fe (1 ML) on Cu(100). (a): (01) beam: (b): (10) beam.

M. Onellion et al. / Epitaxial growth of fee Fe on Cu(100)

~(1x1)

4 MONOLAYERS I 30

I

between

Cu(100) Cu(lO0)

DATA DATA _

Fe I

90

I 150

ELECTRON

Fig. 5. Comparison

CLEAN Fe ON

FdCu(100)

221

ENERGY

1

I

210

I 270

(eV)

LEED intensity data taken for clean Cu(100) (dashed 3 and 4 ML Fe on Cu(100).

lines) and 1, 2,

between experiment and theory is indeed improved by assuming a Debye temperature of 343 K rather than a rigid lattice. One can verify the structural results directly without recourse to calculated LEED intensities by noting that the atomic scattering factors of Fe and Cu are nearly identical. We illustrate in fig. 5 the excellent agreement between the spectra taken for clean Cu (dashed lines) and those taken for 1-4 ML of Fe. The similarity of these spectra reveals directly that fee iron overlayers of essentially the same structure as clean Cu(100) can be grown through 4 ML thickness on Cu(100). There is one final point to mention about our data. As noted, l-4 ML appears to grow as p(1 X 1) fee iron. For 0.5-1.0 ML, however, we observed the LEED pattern illustrated in the inset of fig. 1. Due to our equipment limitations, we were unable to obtain I-V profiles of the illustrated LEED pattern. Based on visual inspection, there are more than one possible ordered structures, including: (a) domains of (2 x 1) and (1 X 2); (b) plgl shifts away from the high symmetry, four-fold hollow site [16], as has been observed, e.g., for Ni(lOO) [17].

228

M. Onellion ef al. / Epitaxial growth of fee Fe on Cu(100)

What the spots indicate is a competition between fee and bee iron, which have different lattice constants. It is noteworthy that the spots disappear abruptly, for an additional exposure corresponding to less than 0.1 ML. This competition between crystal structures implies that the iron film may be under strain. Our LEED intensity data cannot seek out a small shift of the Fe atom location, from the bridge site, due to strain. This competition between crystal structures also has implications for the electronic structure, as discussed elsewhere [ 181.

4. Conclusions We have carried out an extensive investigation of the low coverage nucleation and growth of Fe layers on Cu(100). Our LEED analysis demonstrates that Fe grows epitaxially on Cu(100) as an fee lattice with the lattice constant of Cu(100) to within (*0.05 A). Our Auger analysis of the growth process suggests that the Fe grows predominantly in a layer-by-layer mechanism.

Acknowiedgemenis We wish to thank Dr. Alex Ignative for the use of a spot photometer. We also with to thank Craig Ballantine and Dave Anacker for carrying out some of the analysis of the Auger intensities. Three of us (MO, MT and JLE) were supported in our work by the National Science Foundation (Grant Number DMR 83-04368). We wish to thank the referee for pointing out the possibility of domains of (2 x 1) and (1 X 2) between 0.5 and 1.0 ML.

References [l] For recent reviews of MBE and other epitaxial techniques, see G.B. Stringfellow, Phys. 45 (1982) 469; E. Bauer, Appl. Surface Sci. 11/12 (1982) 479. [2] U. Gradman, Phys. Rev. B27 (1983) 1935, and references therein. [3] G.A. Prim, Phys. Rev. Letters 54 (1985) 1051. [4] C.L. Fu, A.J. Freeman and T. Oguchi, Phys. Rev. Letters 54 (1985) 2700. [5] 0. Haase, Z. Naturforsch. Al4 (1959) 920. [6] W.A. Jesser and J.W. Matthews, Phil. Mag. 15 (1967) 1097. [7] K.B. Hathaway, H.J.F. Jansen and A.J. Freeman, Phys. Rev. B31 (1985) 7603. [8] X. Zhu and J. Hermanson, Phys. Rev. B27 (1983) 2092. [9] Y.C. Lee, H. Min and P.A. Montano, Surface Sci. 166 (1986) 391. [lo] A.M. Turner, A.W. Donoho and J.L. Erskine, Phys. Rev. B29 (1984) 2986. [ll] S. Ossicini, R. Mimeo and F. Cicacci, I. Vacuum Sci. Technol. A3 (1985) 387.

Rept. Progr.

M. Onellion et al. / Epitaxinl growth of jcc Fe on Cu(lO0)

229

[12] L.E. Davis, N.C. MacDonald, P.W. Palmberg, G.E. Riach and R.E. Weber, Handbook of Auger Electron Spectroscopy, 2nd ed. (Perkin-Elmer, Eden Prairie, MN, 1971). [13] A. Septier, Ed., Focusing of Charged Particles, Vols. 1 and 2 (Academic Press, New York, 1967). [14] P.R. Webber, CD. Rojas, P.J. Dobson and D. Chadwick, Surface Sci. 105 (1981) 20. [15] H.L. Davis and J.R. Noonan, Phys. Scripta T4 (1983) 141. [16] See, e.g., M. Jo, M. On& and M. Nishijima, Surface Sci. 154 (1985) 417; B. Christman, in: Structure of Surfaces, Eds. M.A. Van Hove and S.Y. Tong (Springer, Berlin, 1985). [17] P.A. Dowben, private communication. [18] M. Onellion, C.L. Fu, M.A, Thompson, J.L. Erkine and A. Freeman, Phys. Rev. B33 (1986) 7322; B. Cooper and P. Montano, private communication, and to be published.