ARTICLE IN PRESS
Journal of Crystal Growth 296 (2006) 165–173 www.elsevier.com/locate/jcrysgro
Growth by molecular beam epitaxy and interfacial reactivity of MnSb on InP(0 0 1) S.A. Hatfield, G.R. Bell Department of Physics, University of Warwick, Coventry CV4 7AL, UK Received 12 July 2006; received in revised form 9 August 2006; accepted 17 August 2006 Communicated by R. Kern Available online 11 October 2006
Abstract Growth by molecular beam epitaxy of MnSb on InP(0 0 1) has been studied over a range of substrate temperatures (250–425 C) and Sb:Mn flux ratios (2:1 to 8:1). Two growth phases were observed, their predominance mainly dependent on substrate temperature. These have been identified as MnSb (found preferentially at low temperatures) and InSb (found preferentially at high temperatures). The latter phase forms due to In out-diffusion from the substrate, leading to a granular film with increased total thickness and a very rough buried interface. This reaction only occurs in the presence of both Mn and Sb incident fluxes: when exposed to an Sb flux alone, very small islands of InSb are formed at the surface of InP due to residual In clusters but no disruption of the substrate occurs. No such effect was observed for growth on GaAs (0 0 1) and (1 1 1)B. Implications for the growth of hybrid epitaxial systems for semiconductor spintronics are discussed. r 2006 Elsevier B.V. All rights reserved. PACS: 68.35.Ct; 68.37.Hk; 68.55.Nq; 61.14.Hg Keywords: A3. Molecular beam epitaxy; B1. Antimonides; B2. Magnetic materials; B2. Semiconducting III–V materials
1. Introduction In recent years, manganese monopnictides have attracted a great deal of attention in regard to hybrid spintronic devices, where ferromagnetic materials are epitaxially combined with conventional non-magnetic semiconductors. The most widely investigated materials are MnAs and MnSb, both of which have a stable metallic ferromagnetic phase at room temperature with the NiAstype structure. This consists of ABAC hexagonal close packing as illustrated in Fig. 1. In the case of MnAs, the Curie temperature is T C ¼ 45 C, but between 45 and 125 C the structure changes from hexagonal NiAs-type to orthorhombic MnP-type [1]. High-quality MnAs has been grown on GaAs substrates even though the samples are normally cooled through this phase transition after growth Corresponding author. Tel.: +44 24 7652 3489.
E-mail address:
[email protected] (G.R. Bell). URL: http://www.warwick.ac.uk/go/grbell. 0022-0248/$ - see front matter r 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.jcrysgro.2006.08.031
and the lattice mismatches to both the a and c directions are rather large [2–4]. In contrast to the use of ferromagnetic simple metals to produce hybrid structures, both ferromagnet-on-semiconductor and semiconductor-on-ferromagnet epitaxy are possible [5,6], making MnAs and MnSb very attractive for hybrid device applications. MnSb also shows large magneto-optical effects [7] which may be advantageous for information storage devices. In the case of MnSb, there are no structural transitions below T C ¼ 314 C [1,8]. High-quality crystalline films have been grown on various III–V semiconductor substrates using techniques such as hot wall epitaxy [9–12] and molecular beam epitaxy (MBE) [13–15]. Although reported optimum growth conditions vary considerably, substrate temperatures are generally in the range 200–400 C with Sb-rich flux ratio. Various epitaxial orientations are possible between NiAs-structure MnSb and the ZnS-structure substrates. These are primarily dependent on the substrate crystal face, with epitaxial relationship found to be MnSbð1 1¯ 0 0ÞkGaAsð1 0 0Þ [13] and
ARTICLE IN PRESS S.A. Hatfield, G.R. Bell / Journal of Crystal Growth 296 (2006) 165–173
166
[1000] [0100]
[0101]
[0001] (0001) surface
(1101) surface [0010]
[0100] A [1000] [0010] B
(b)
(d)
A c
[0001] (1100) surface
C
[0010] Mn
A Sb Dangling bond a
(a)
(c)
Fig. 1. Exploded view of the MnSb structure and schematics for three low index surfaces: (b) (0 0 0 1); (c) ð1 1¯ 0 0Þ; (d) ð1 1¯ 0 1Þ.
MnSbð0 0 0 1ÞkGaAsð1 1 1Þ [9]. It is desirable to minimize the epitaxial strain which can be achieved by matching inplane lattice spacings. MnSb has lattice parameters a ¼ ˚ 4:128 A˚ and cp¼ ffiffiffi 5:789 A, while InP has a surface lattice spacing of a= 2 ¼ 4:150 A˚ for (1 0 0) and (1 1 1) surfaces. This leads to a mismatch of only 0.5% compared to 3:2% for GaAs, 10.1% for InSb and 7:0% for Si. For (0 0 0 1)oriented growth on (1 1 1) substrates, this mismatch occurs in all principal surface directions. Because of the usual epitaxial orientation on (0 0 1) substrates, this mismatch occurs in one principal direction while in the perpendicular direction the mismatch is much larger. MnAs has been epitaxially grown on InP(0 0 1), with the orientation MnAsð1 1¯ 0 0ÞkInPð1 0 0Þ [4], although to the best of our knowledge there have been no reports concerning MnSb on InP substrates. The related ternary material NiMnSb has been grown on InGaAs buffer layers lattice matched to InP(0 0 1) substrates [16], but again we know of no attempts at direct growth on InP. In this report, growth of MnSb on InP(0 0 1) by MBE is investigated, and the resulting over-layers examined using a range of techniques including reflection high-energy electron diffraction (RHEED), scanning electron microscopy (SEM) and X-ray photoelectron spectroscopy (XPS). 2. Experimental details MnSb/InP(0 0 1) samples were grown in a compact MBE chamber which contains shuttered Mn and Sb effusion cells (the latter having no thermal cracking stage), a retractable beam flux ionization gauge and a 15 keV electron gun and phosphor screen for RHEED. Connected to the MBE chamber is a preparation chamber, containing an ion gun
for surface preparation and a fast sample entry system. The MBE sample heater is calibrated to the melting points of In and InSb and uses two thermocouples to measure the temperature. Samples are mounted on stainless steel sample plates using In solder. InP(0 0 1) samples (Wafer Technology Ltd., UK) 10 mm 10 mm, were prepared ex situ by rinsing in acetone, propanol and de-ionized water to remove dust and debris. The samples were then blown dry with nitrogen and placed in the load chamber. Once inside the MBE system, the samples were cleaned by annealing at 425 C for 30 min, followed by argon ion bombarding at grazing incidence for 2 min. The ion energy was 500 eV and the low ion dose prevents the formation of large In clusters at the surface. The samples were then annealed at 425 C for a further 30 min and then cooled to the growth temperature. This process produced a sharp ð2 4Þ RHEED pattern. The n m notation corresponds to n periodicity in the ½1 1¯ 0 direction (observed with the RHEED beam aligned along the [1 1 0] direction) and m in the [1 1 0] (observed with the RHEED beam aligned along the ½1 1¯ 0). This follows the convention for InSb cð8 2Þ [17] and the mixed dimer model for InPð2 4Þ [18]. After allowing 5 min for the samples to reach thermal equilibrium, the growth was initiated by opening both Mn and Sb shutters simultaneously. The Mn beam pressure was kept constant throughout the experiments at 1 107 Torr, while the Sb pressure was varied to achieve different flux ratios. Here, flux ratio J Sb=Mn is defined as the ratio of directly measured beam pressures. The beam fluxes from both cells were observed to be constant over the growth period to better than 5%. The samples were grown for 90 min in all cases. Samples were analyzed in situ by
ARTICLE IN PRESS S.A. Hatfield, G.R. Bell / Journal of Crystal Growth 296 (2006) 165–173
RHEED, and ex situ by SEM (ZEISS SUPRA55VP) with energy dispersive X-ray analysis (EDX) and XPS (VSW ESCA, and Scienta ESCA300). 3. Results and discussion 3.1. RHEED MnSb was grown on InP(0 0 1) under a range of substrate temperatures and flux ratios, summarized in Fig. 2 which indicates the RHEED patterns observed postgrowth. Three different RHEED patterns are observed, termed diffuse, phase A and phase B. At low temperatures,
Fig. 2. RHEED patterns for MnSb growth on InP(0 0 1) as a function of substrate temperature and Sb/Mn flux ratio.
RHEED patterns are diffuse with no clear integer or fractional order streaks. This is associated with surface roughening, poor crystalline ordering and/or a weak epitaxial relationship to the substrate. At high temperatures (above 300 C), ð4 1Þ RHEED patterns are observed at both high and low flux ratios. The RHEED profiles for the two perpendicular high symmetry directions of this pattern, in addition to profiles from the InP(0 0 1) ð2 4Þ substrate are shown in Fig. 3(b). The lattice spacing for this reconstruction (labelled phase B) determined from the RHEED spacing (vertical arrows) was calculated to be 4:5 0:2 A˚ which is close to that of InSb. The RHEED patterns show no sign of rotated domains, indicating that the over-layer is epitaxial, and the orientation of 4 observed in the [1 1 0] beam direction is consistent for all the samples. Phase B has clear similarities to InSb(0 0 1). The cð8 2Þ=ð4 2Þ surface [17] can exhibit one-dimensional disorder leading to an apparent ð4 1Þ diffraction pattern. At intermediate substrate temperature ð300 CÞ and high flux ratios, a third RHEED pattern (phase A) is observed. This pattern is rather faint, and the quality varies considerably between samples, even those grown under identical growth conditions. Phase A is characterized by unequal integer order spacing in the two high symmetry directions, one approximately lattice matched to the InP substrate, the other with a smaller RHEED spacing, corresponding to a larger lattice. RHEED profiles for phase A in the two high symmetry directions are shown in Fig. 3(a). The lattice parameters determined for phase A ˚ a2 ¼ 6:9 0:2 A. ˚ The second lattice are a1 ¼ 4:3 0:2 A, parameter indicates a surface with a component in the
InP 2x
RHEED screen intensity (arb. units)
RHEED screen intensity (arb. units)
InP 2x
InP 4x
Phase A 2x
Phase A 2x
-15
(a)
-10
167
-5
0
5
Distance (mm)
10
15
InP 4x
Phase B 4x
Phase B 1x
-15
(b)
-10
-5
0
5
10
15
Distance (mm)
Fig. 3. RHEED intensity profiles for InP substrate and (a) low-temperature phase A, (b) high-temperature phase B. Beam directions from top to bottom are [1 1 0], ½1 1¯ 0, [1 1 0] and ½1 1¯ 0. The RHEED patterns correspond to InPð2 4Þ, phase A ð2 2Þ and phase B ð4 1Þ. Distances are measured horizontally across the RHEED screen.
ARTICLE IN PRESS 168
S.A. Hatfield, G.R. Bell / Journal of Crystal Growth 296 (2006) 165–173
Fig. 4. Plan view SEM for samples grown with flux ratio J Sb=Mn ¼ 6 at substrate temperatures: (a) 300 C; (b) 340 C; (c) 390 C and (d) 425 C.
c-axis such as MnSb ð1 1¯ 0 0Þ, which has primitive unit ˚ a2 ¼ 5:789 A, ˚ or MnSbð1 1¯ 0 1Þ, with vectors a1 ¼ 4:128 A, ˚ unit vectors a1 ¼ 4:128 A, a2 ¼ 7:1 A˚ and an angle of 73:1 between high symmetry directions. These planes are illustrated in Fig. 1. 3.2. Plan view SEM After growth, the samples were removed from the MBE system and imaged by SEM. Fig. 4 shows the SEM images for J Sb=Mn ¼ 6 samples with increasing growth temperature. For temperatures above 300 C, multiple phase growth can easily be observed. Image (c), corresponding to growth at 390 C, contains two phases with distinct grey levels. The darker phase becomes more prevalent at lower growth temperatures (b) and is dominant at 300 C (a). The lighter phase appears as a matrix surrounding micron-sized crystallites of the darker phase and is prevalent at the highest growth temperature employed, 425 C (d). By comparison with the RHEED results, the darker crystallites can be assigned to phase A and the lighter matrix to phase B. The SEM images were processed by thresholding to determine the relative abundance (area fraction) of the two phases. While this was not straightforward, due to the fairly low absolute contrast between the two phases, with care a reliable ratio could be extracted. The fraction of phase B present ðB% Þ is shown in Fig. 5 along with an inset processed image. For both flux ratios, B% increases linearly with growth temperature. For samples grown with a flux ratio of J Sb=Mn ¼ 2, the surface was covered in small droplets of diameter 100 mm or larger, which were identified by EDX to be predominantly In. These metallic droplets were not observed in samples grown at higher flux ratios.
Fig. 5. Coverage of phase B ðB% Þ as a function of growth temperature for J Sb=Mn ¼ 2 and 6, as determined from processed plan view SEM images (inset).
3.3. XPS In order to confirm the identities of the two phases, the samples were examined using XPS. Sample transfer between the MBE and XPS systems resulted in surface contamination. Since the samples could not be cleaned without potentially modifying the sample, an assumption is made that the spectrum obtained from the sample and oxide is representative of the surface before contamination. This is reasonable, since the sample is not annealed after contamination, and hence diffusion lengths of the atoms should be low. Mn, In, Sb, C and O were the only elements detected. Fig. 6 shows the relative surface concentrations of Mn, In and Sb for samples grown with J Sb=Mn ¼ 6 (solid
ARTICLE IN PRESS S.A. Hatfield, G.R. Bell / Journal of Crystal Growth 296 (2006) 165–173
169
B) and single phase MnSb (phase A) grown under similar conditions on GaAs(1 1 1). It is important to note that MnSb grown on both GaAs and InP with flux ratio J Sb=Mn p8 has a Mn-enriched surface as measured by XPS. On GaAs, we later confirmed that MnSb with surface stoichiometry close to Sb:Mn ¼ 1:1 could be achieved by growth at higher flux ratios J Sb=Mn X10. For this reason, the dotted line in the bottom panel of Fig. 6 is not horizontal although this does not reflect the bulk stoichiometry of the phases grown. All our EDX results indicate that the bulk stoichiometry phase A is close to Sb:Mn ¼ 1:1. Bearing in mind the Mn-rich surface layer of phase A, the agreement between the phase abundance B% and the elemental surface concentration is very good for both flux ratios. 3.4. Cross-sectional SEM
Fig. 6. Relative surface concentrations for Mn, In and Sb as a function of area fraction of phase B ðB% Þ for growth at J Sb=Mn ¼ 6 (solid squares) and J Sb=Mn ¼ 2 (open circles), as determined from the Mn 2p, In 3d and Sb 3d XPS peak intensities. The dotted lines are linear extrapolations for ‘ideal’ MnSb–InSb phase mixing.
squares) and J Sb=Mn ¼ 2 (open circles) as a function of area fraction of phase B (B% ). The concentrations are derived from the Mn 2p, In 3d and Sb 3d XPS peak areas, normalized according to atomic sensitivity and photoelectron mean free path [19]. These surface concentrations reflect the population of the three elements over the upper 2–3 nm of the samples. For the lowest growth temperatures where B% ¼ 0, the surface is highly Mn-rich. As the growth temperature and B% increase, the Mn surface concentration drops steadily to o5% at B% ¼ 100, while the In concentration increases from 5% (B% ¼ 0) to 50% (B% ¼ 100). The dotted lines in Fig. 6 show a linear extrapolation for ideal phase mixing between InSb (phase
The presence of In in the XPS spectrum after several hundred nm of growth leads to two possible conclusions. The first is that the growth layer is either much thinner than expected ðo4 nmÞ or is incomplete, meaning that parts of the substrate still contribute to the XPS and RHEED results. The second possibility is that the MnSb growth is disrupting the substrate, resulting in large scale In diffusion to the growing layer and the formation of an In-containing phase. In order to determine the source of the In signal in XPS, the cleaved edges of a number of samples were observed in SEM. This allows for direct observation of the growth layer thickness in addition to indications of the condition of the interface quality. This technique has been used to observe MnSb layers grown on GaAs (Fig. 7(d)) with good results. Fig. 7(a)–(c) shows a number of the images for MnSb/InP(0 0 1) samples ðJ Sb=Mn ¼ 6Þ, with growth temperatures from 250–425 C. Panel (a) shows a layer of approximately 300 nm, with no indications of multiple phases. From the plan view SEM images (Fig. 4), this phase may be identified as phase A. The interface region shows visible disruption compared to the sample grown on GaAs(0 0 1) (panel (d)). The overlayer in panel (b) (grown at 340 C) is thicker and now shows two clearly different phases: the crystallites as observed in Fig. 4 and the surrounding material. These are identified as phases A and B, respectively. There appear to be cavities in the lower sections of the growth layer, but these are probably caused by the removal of crystallites from the sample during the cleave. Panel (c) shows a thicker layer again, grown at 425 C, with significantly different morphology to previous samples. There is little indication of the crystallites (phase A), and the layer is dominated by phase B, with a very rough interface. Again, this is in agreement with the plan view SEM images. Panel (d) shows the cross-section SEM images from a MnSb layer grown on GaAs(0 0 1) under similar growth conditions to panel (a). The layer on GaAs can be seen to have a sharp interface region compared to all the layers grown on InP(0 0 1), highlighting the reactivity of the InP substrate.
ARTICLE IN PRESS 170
S.A. Hatfield, G.R. Bell / Journal of Crystal Growth 296 (2006) 165–173
Fig. 7. Cross-section SEM for MnSb samples grown on InP(0 0 1) with flux ratio J Sb=Mn ¼ 6 at substrate temperatures (a) 250 C, (b) 340 C, (c) 425 C and (d) MnSb grown on GaAs(0 0 1) at 300 C, J Sb=Mn ¼ 4.
values. As the temperature is increased, the total layer thickness increases and plateaus at 300 nm for J Sb=Mn ¼ 2. By contrast, for J Sb=Mn ¼ 6 the total layer thickness continues to increase up to 700 nm. This behavior is discussed quantitatively below. 3.5. InP annealed under Sb
Fig. 8. Growth layer thickness for J Sb=Mn ¼ 2 (open circles) and J Sb=Mn ¼ 6 (filled squares) as determined from cross-sectional SEM. Dotted lines indicate the total expected volume in the Sb-poor regime due to the difference in volume per formula unit between InSb and MnSb. The solid line shows the expected total layer thickness in the Sb-rich regime according to a model with DH eff ¼ 41 kJ mol1 .
The over-layer thickness for J Sb=Mn ¼ 2 and 6 samples, measured by cross-sectional SEM are shown in Fig. 8. The cross-sectional SEM images show that the growth layer is in excess of 180 nm for all samples, with no gaps, and so the In XPS signal cannot originate from the substrate. This is confirmed by the absence of P 2p peaks in all XPS spectra. The lack of P in the grown phases is also consistent with the low miscibility of Sb and P in InSbP alloys [20]: if an InSb-like phase is indeed formed, it is unlikely to be alloyed with InP. The In signal is therefore attributed to In diffusion from the substrate into the growing layer. The layer thicknesses predicted from growth studies of MnSb on GaAs(1 1 1)B are 200 and 325 nm for J Sb=Mn ¼ 2 and 6, respectively. For both flux ratios the thicknesses at low temperature are close to the predicted
In order to isolate the role of Mn in the development of the InSb phase, a clean InP(0 0 1)-ð2 4Þ sample was annealed at 400 C under Sb flux for 90 min. The RHEED was observed and the resulting surface examined using both plan view and cross-sectional SEM. Immediately after the Sb flux was initiated the ð2 4Þ RHEED pattern disappeared, leaving only weak integer order streaks in both high symmetry directions. After a few seconds the RHEED pattern became spotty, indicating transmission diffraction through small crystallites. After this change the RHEED pattern was unaltered for the remainder of the annealing time. The RHEED patterns in the two high symmetry directions are shown in Fig. 9 with three distinct features highlighted. The streaks labelled (i) have the largest spacing, which corresponds to a real-space distance of 4:2 0:2 A˚ as expected for the InP substrate. Feature (ii) shows streaks tilted at approximately 56 to vertical indicating crystalline facets, while feature (iii) shows the spotty transmission diffraction pattern. The facets and spots have the same periodicity, indicating a lattice spacing ˚ which matches well with the h1 1 0i spacing of 4:5 0:2 A, ˚ The facets were far more pronounced of InSb ð4:58 AÞ. observed with the beam aligned in the [1 1 0] direction. The sample was examined using SEM and the surface was mostly flat and featureless with a small number of small three-dimensional islands (p 250 nm in all dimensions). The islands were rectangular in shape with the edges aligned along h1 1 0i. There appears to be a preferential direction for the long axis of the rectangles along the [1 1 0] surface direction in Fig. 10. The insert of Fig. 10(a) shows
ARTICLE IN PRESS S.A. Hatfield, G.R. Bell / Journal of Crystal Growth 296 (2006) 165–173
171
Fig. 9. RHEED patterns after annealing InP(0 0 1) under Sb for 90 min. The beam is aligned along (a) [1 1 0] and (b) ½1 1¯ 0 surface directions. Features indicated are (i) substrate streaks, (ii) facets, and (iii) transmission spots.
cleave. Atomically smooth edges are never observed using this simple cleave technique because of irregularities and debris remaining at the corner defined by the intersection of {1 1 0} and {0 0 1} planes. Whatever the origin of the apparent undulations, there is clearly no development of large ð100 nmÞ crystallites and no large-scale disruption of the sub-surface region. Due to the low number density and size of the InSb islands it is unlikely that any large scale In out-diffusion has occurred, certainly not to the extent observed during MnSb growth for an equivalent time at 400 C. Instead, the islands are probably formed by the reaction of Sb with residual In left on the surface from preferential sputtering during the ion beam cleaning process. It is therefore apparent that the mechanism for In diffusion observed earlier in this paper is not caused by Sb alone via a group V exchange reaction [22]. A recent study of InP(0 0 1) exposed to trimethylantimony in a metalorganic vapor phase epitaxy reactor [20] showed that even at 600 C, a surface layer of InSb of maximum average thickness 7 A˚ was developed, with surface roughening of root mean square ˚ We conclude that the presence of Mn is amplitude o4 A. essential in promoting the growth of the thick InSb phase observed in the present study. 3.6. Layer thickness and growth kinetics Fig. 10. SEM images of an InP(0 0 1) sample annealed under Sb flux for 90 min. (a) Plan view showing aligned rectangular three-dimensional islands (the inset shows an enlarged view of one island), (b) cross-sectional SEM image. The preferential long axis for rectangular three-dimensional islands is along the [1 1 0].
an enlarged image of one of the 3D islands, which shows distinct facets, determined as {1 1 1} planes from the angles of the streaks in the RHEED (feature (ii), Fig. 9). These observations agree with those of Ferrer et al. [21] during heteroepitaxial growth of InSb islands on InP(0 0 1). Fig. 10(b) shows the cross-sectional SEM image of the sample, which shows a surface with apparent undulations of the order of 10 nm. These are very small compared to the disruption regions in Fig. 7, and are typical of an InP
It is clear than substantial disruption of the InP substrate occurs during MnSb growth and that this is stronger at higher substrate temperatures. Indium atoms are freed from the InP substrate and produce InSb in the form of epitaxial crystallites of the order of 100 nm in size. The effective incoming fluxes can be thought of as Mn ðf Mn Þ and Sb ðf Sb Þ from the vacuum and In ðf In Þ from the substrate, leading to a total layer thickness which depends on temperature and Sb flux (Fig. 8) and is larger than the nominal MnSb layer thickness. This behavior can be explained in terms of the formation of InSb and MnSb in two different growth regimes: the Sbpoor regime ðf Sb of Mn þ f In Þ and the Sb-rich regime ðf Sb 4f Mn þ f In Þ. In the Sb-poor regime the total growth thickness is limited by f Sb (which is not dependent on
ARTICLE IN PRESS 172
S.A. Hatfield, G.R. Bell / Journal of Crystal Growth 296 (2006) 165–173
substrate temperature) and MnSb and InSb must grow competitively. As a result of the different volume per formula unit between MnSb ðV MnSb ¼ 42:7 A˚ 3 Þ and InSb ðV InSb ¼ 68:0 A˚ 3 Þ, the total layer thickness is expected to increase linearly with the percentage of phase B present ðB% Þ. The value of B% is dependent on substrate temperature. Simulated total layer thicknesses L according to this model are shown by dashed lines in Fig. 8, calculated using the fits to B% data in Fig. 5. Even though the true (bulk) value of B% is likely to deviate somewhat from the value derived by plan view SEM, the simulation fits the J Sb=Mn ¼ 2 data well, exhibiting the correct onset and plateau in L. In addition, excess metallic species would be expected to appear as droplets on the surface and these were indeed identified. For the J Sb=Mn ¼ 6 samples, L exceeds that of the simple linear model at high temperatures (Fig. 8), indicating the Sb-limited growth is not occurring at this higher flux ratio. Under Sb-rich conditions, L becomes dependent upon f Mn and f In . While the Mn flux is constant, the effective In flux from the substrate depends on disruption of the InP lattice and migration of In atoms. While these processes are undoubtedly complex, it is likely that at least one step is thermally activated. If this is rate-limiting, the temperature dependence reduces to a single effective activation enthalpy. In this highly simplified approach, the layer thickness then takes the temperature-dependent form L / V MnSb f Mn þ V InSb AeDH eff =RT
(1)
in which A is a constant, DH eff is the effective enthalpy and the other symbols have their usual meaning. This equation has been fitted to the J Sb=Mn experimental data, and is shown in Fig. 8 by the solid line. While the fit is clearly superior to that of the Sb-poor model, it is by no means perfect. In fact, it is difficult to extract a reliable value for DH eff and values ranging from 40 to 70 kJ mol1 provide reasonable fits. We do not wish to speculate further on the kinetic processes involved, but it is clear that for J Sb=Mn ¼ 6 the total layer thickness is no longer Sb-limited. However, even when the total growth thickness is not limited by the Sb flux ðJ Sb=Mn ¼ 6Þ the MnSb has a highly Mn-rich surface layer (Fig. 6). This is not the case for the co-growing InSb phase, implying that Sb incorporation kinetics are rather different for MnSb and InSb. MnSb grown epitaxially on Si has shown a similar interfacial reaction [23], with the formation at the interface of MnSi in either epitaxial layer form or as islands depending on the substrate temperature. It was suggested that the formation of MnSi was thermodynamically due to the higher (more negative) enthalpy of formation of MnSi compared to MnSb [24] with the reaction kinetically limited by cross-diffusion of Mn and Si. An effective activation energy of 1.6 eV for cross-diffusion was derived [23]. In the present case, the enthalpies of formation of MnSb and InSb are rather similar, at 40 and 35 kJ mol1 , respectively, [24] with low miscibility. It is therefore not surprising that both phases are able to grow together.
MnSb clusters grown on sulfur terminated GaAs have been shown to generate a very large room temperature magnetoresistance effect [25,26]. In this respect, the highly granular MnSb/InSb films produced at growth temperatures above 300 C in the present study may be of interest. However, when smooth films with well-defined interface properties are required, such as for spin injection applications, it is clear that direct growth of MnSb on InP is not favorable due to the strong interfacial reactivity. 4. Conclusions The growth of MnSb on InP has been studied over a range of substrate temperatures and beam flux ratios. At 300 C, MnSb is grown with fairly smooth film morphology but only moderately planar interfaces. At higher temperatures the dominant phase formed becomes InSb. While both the MnSb and InSb are epitaxial, the film is granular with a very rough interface region. Disruption of the InP substrate liberates In which competes with the incoming Mn flux to form both InSb and MnSb. This disruption only occurs in the presence of an incoming Mn flux— simply annealing the substrate under Sb does not result in the formation of InSb except by uptake of residual surface In from the cleaning process. Depending on the Sb:Mn flux ratio, the total thickness of the InSb þ MnSb films was limited by either the temperature-dependent effective In flux (Sb-rich) or the available Sb (Sb-poor). Direct growth of MnSb on InP is not favorable for growth of sharp and well-defined spin injection interfaces. Acknowledgments We are grateful to S. York and R. Johnston for expert technical assistance. S.A.H. was supported by a Warwick Postgraduate Research Fellowship, and G.R.B. by the Royal Society (UK). References [1] T. Massalaski, H. Okamoto, P. Subramanian, L. Kacprzak, Binary Alloy Phase Diagrams, American Society for Metals, Metals Park, OH, 1990. [2] L. Da¨weritz, F. Schippan, A. Tampert, M. Ka¨stner, G. Behme, Z. Wang, M. Moreno, P. Schu¨tzendu¨ber, K. Ploog, J. Crystal Growth 227–228 (2001) 834. [3] M. Tanaka, Mater. Sci. Eng. B 31 (1995) 117. [4] M. Yokoyama, S. Ohya, M. Tanaka, Appl. Phys. Lett. 88 (2006) 012504. [5] K. Ono, T. Uragami, M. Mizuguchi, H. Fujioka, M. Oshima, M. Tanaka, H. Akinaga, J. Crystal Growth 209 (2000) 556. [6] J. Song, J. Lee, Y. Cui, J. Ketterson, S. Cho, Appl. Phys. Lett. 85 (2004) 4079. [7] R. Coehoorn, C. Haas, R. de Groot, Phys. Rev. B 31 (1985) 1980. [8] W. Pearson, A Handbook of Lattice Spacings and Structures of Metals and Alloys, Pergamon, London, 1958. [9] H. Tatsuoka, H. Kuwabara, M. Oshita, T. Nakamura, H. Fujiyasu, J. Crystal Growth 166 (1996) 754. [10] K. Sato, H. Ikekame, M. Akita, Y. Morishita, J. Magn. Magn. Mater. 177 (1998) 1379.
ARTICLE IN PRESS S.A. Hatfield, G.R. Bell / Journal of Crystal Growth 296 (2006) 165–173 [11] H. Ikekame, Y. Yanase, T. Ishibashi, T. Saito, Y. Morishita, K. Sato, J. Crystal Growth 173 (1997) 218. [12] H. Fujiyasu, T. Nakamura, M. Oshita, H. Kuwabara, H. Tatsuoka, Y. Nakanishi, Thin Solid Films 281–282 (1996) 499. [13] K. Ono, M. Shuzo, M. Oshima, H. Akinaga, Phys. Rev. B 64 (2003) 085328. [14] O. Rader, M. Lezaic, S. Blu¨gel, A. Fujimori, A. Kimure, N. Kamakura, A. Kakizaki, S. Miyanishi, H. Akinaga, New J. Phys. 7 (2005) 111. [15] M. Oshima, M. Shuzo, K. Ono, H. Fujioka, Y. Watanabe, S. Miyanishi, H. Akinaga, Appl. Surf. Sci. 130–132 (1998) 892. [16] P. Bach, A. Bader, C. Ru¨ster, C. Gould, C. Becker, G. Schmidt, W. Weigand, C. Kumpf, E. Umbach, R. Urban, G. Woltersdorf, B. Heinrich, Appl. Phys. Lett. 83 (2003) 521. [17] C. Kumpf, L.D. Marks, D. Ellis, D. Smilgies, E. Landemark, M. Nielsen, R. Feidenhans’l, O. Bunk, J.H. Zeysing, Y. Su, R. Johnson, Phys. Rev. Lett. 86 (2001) 3586.
173
[18] W. Schmidt, F. Bechstedt, Surf. Sci. 409 (1998) 474. [19] hhttp://www.nist.gov/srd/nist71.htmi, 2006. [20] Y. Sun, S. Cheng, G. Chen, R. Woo, R. Hicks, J. Appl. Phys. 897 (2005) 103512. [21] J.C. Ferrer, F. Peiro, A. Cornet, J.R. Morante, T. Uztmeier, G. Armelles, F. Briones, Appl. Phys. Lett. 69 (1996) 3887. [22] A. Godefroy, S. Ababou, B. Le´pine, A. Guivarc’h, G. Je´ze´quel, J. Crystal Growth 179 (1997) 349. [23] H. Tatsuoka, K. Isaji, K. Sugiura, H. Kuwabara, P.D. Brown, Y. Xin, C.J. Humphreys, J. Appl. Phys. 83 (1998) 5504. [24] I. Barin, Thermochemical Data of Pure Substances, VCH, Weinheim, Germany, 1993. [25] H. Akinaga, M. Mizuguchi, K. Ono, M. Oshima, Appl. Phys. Lett. 76 (2000) 357. [26] H. Akinaga, Surf. Sci. 514 (2002) 145.