Technical Insight
MBE looks to new applications
Sales of reactors have suffered from the contraction in MBE's traditional markets, but there are new applications for MBE to keep researchers busy, reports Richard Campion.
The 12th European MBE conference took place in the picturesque village of Bad Hofgastein high in the Austrian Alps on February 16-19. With MBE s traditional application sectors suffering an economic downturn, it was interesting to hear discussions about the technique s application in the emerging III-nitride microelectronics field, and how MBE is making itself useful in the CMOS industry, potentially a big market for MBE system manufacturers.
The first session was opened by N Grandjean of CRHEA-CNRS Valbonne, France, who discussed the progress in MBE growth of GaN and compared the use of N2-plasma with ammonia (NH3). The two methods require different approaches; with N2-plasma the best layers are grown under gallium-rich conditions, whereas nitrogen-rich conditions are best for NH3. A comparison was made of growth onto sapphire, silicon and single-crystal GaN substrates. This showed that the roughening of the surface of MBE-grown GaN is due to kinetic phenomena, not edge dislocations as is often thought. Finally, the effect of NH3 on the surface free energy of strained GaN (on AlN) was shown to be a useful tool to control (reversibly) the 2D-3D transition utilized in the formation of quantum dots.
F Semond, also from CNRS, focused on the growth of GaN on silicon and SiC substrates using NH3. He showed that the introduction of a thick AlN interlayer on top of GaN nucleated on a thin AlN layer allows the relaxation of strain. This increases the critical thickness for cracking of a subsequent layer to several microns. Data for AlGaN/GaN HEMTs grown in this way on SiC were presented, and included mobilities of 10,000 cm2/Vs (at 10 K) and electron sheet densities of 1013/ cm-2.
G Kobulmuller of Infineon Technologies discussed experiments that closely monitor the growth rate of GaN in the early stages of nucleation by line-of-sight mass spectrometry of the Ga desorption rate from the surface. A drop in desorption rate is associated with the opening of the nitrogen shutter, and thus the growth rate can be deduced. This study, combined with AFM and RHEED data, shows that growth starts instantly on sapphire but in two stages on an AlN buffer. The detail of this two-stage process is dependent on both temperature and III/V ratio. The change of the growth rate from being linear to exponential occurs at 2.5 monolayers, which coincides with 2D-3D transition to Stranski-Krastanow growth (figure 1).
J Schmitz of the Fraunhofer Institute, Freiburg, Germany, talked in detail about the wide variety of heterostructures afforded by the type-I and type-II band alignments in the (AlGaIn)(AsSb) system. This allows the formation of structures with effective bandgaps between 20 m2/Vs and carrier densities of 4-5 x 1015/m2). He reported a novel low temperature/low growth rate method for a 15 nm AlAs QW with AlGaAs spacers above and below, which separate the QW from Si-doped regions above and below the spacer layers. The low growth temperature prevents Si migration into the AlAs well. Hall data showing the integer Hall effect and FQHE were presented, demonstrating the high quality of the samples.
Aggressive scaling in CMOS devices is leading to ever-thinner gate oxides. As electron tunneling increases exponentially with decreasing oxide thickness, the CMOS industry is now looking for new materials with higher dielectric constants (K) to replace SiO2. Single-crystal metal oxides are likely high-K candidates, and MBE is being explored as the deposition method.
Hans-Jorg Osten of the University of Hanover, Germany, presented a road map for the future development of CMOS devices with particular focus on the equivalent oxide thickness (EOT) of the gate insulator layer. Because of the strength of the Si-O bonds, the matching of the oxygen sublattice with the silicon lattice is of critical importance. After consideration of several candidate metal oxides, work involving praseodymium oxide (as Pr2O3 in the bixbyte structure) was described. Growth on Si (001) leads to orthogonal (110) domains and overgrowth with Si has not yet been achieved. However, Osten s group has obtained perfect oxide growth on Si (111) and overgrown the oxide layers with Si. The layers showed extremely good properties with K of about 30, compared with 20 for SiO2, low leakage, high breakdown voltage and, importantly, high thermal stability.
J Fompeyrine of IBM s Zurich labs continued Osten s theme with a discussion of other candidate materials. The work focused on (La,Zr) oxides using sequential deposition approach to tune the lattice match of the oxide layer to the silicon (Compound Semiconductor September 2001). The most successful single-phase layers have been grown on Si(111), and show promising properties.
There is much interest in using strained Si channels for future generations of CMOS devices (Compound Semiconductor September 2002). S Mantle of the Jülich Research Center, Germany, reported a novel method of obtaining relaxed SiGe on which to grow strained Si. This starts with the growth of strained SiGe on Si(001), followed by ion implantation of He+ ions at 18 keV into the Si substrate to create a defect-laden layer just below the SiGe. A carefully controlled anneal causes the formation of "bubbles", creating dislocation loops that move to the interface and yield strain relieving misfit dislocations. The degree of relaxation depends on the He dose.
B Voigtländer and J Myslivecek of the Jülich Research Center gave a joint talk to introduce the session on Si growth kinetics. They described equipment that can take real-time STM images of the evolving growth surface (figure 3). The audience was treated to astounding "movies" showing details of the growth process on the atomic level. Examples of these can be seen at www.fz-juelich.de/video/voigtlaender. They then addressed the problem of step bunching in the growth of Si on vicinal Si (001) and introduced the results of kinetic Monte Carlo simulations. These showed that the diffusion anisotropy inherent on the observed 2 x 1 reconstructed surface gives rise to a sticking anisotropy and an effective downward mass current, which leads to instability and step bunching. The model suggests that this behavior will only occur in a certain window of growth temperatures.
S Andrieu of CNRS, France, introduced the meeting to the field of MBE of metals. One main advantage of MBE for metal film deposition is that, due to the relatively low surface diffusion energy, growth can take place at or near room temperature. This allows for the alloying of normally non-miscible metals such as AgCu and AuNi. Also, due to the elasticity of metals, thin highly strained metastable structures can be grown, including face-centered cubic Co or Fe, and body-centered cubic Ni. Thus there are many possible substrate to epilayer relationships.
The discussion then moved to the growth of magnetic tunnel junctions, which consist of two magnetic layers separated by a thin (~1 nm) oxide layer. In such a device the electrons tunnel through the barrier while preserving their spin state. In an ideal device where magnetic polarization is 100%, if the magnetization directions of the two layers are parallel then current will pass; if they are anti-parallel then current will not pass. In real devices using Fe, Co, Ni and associated compounds, the polarization of the electrodes is not 100%, but the search is on for new materials (e.g. Fe3O4, CrO2 and LaSrMnO3), and MBE is at the forefront of this work.
In a related area to the session on metals, L Daweritz of the Paul Drude Institute gave an introduction to the field of ferromagnetic semiconductors and their possible use in spintronic devices. Compound semiconductors are ideal candidates for spintronic devices because ferromagnetic metals can be epitaxially deposited on their surfaces, opening another application for MBE. Daweritz discussed two possible routes to forming the magnetic semiconductors. First, there are dilute magnetic semiconductor (DMS) materials (e.g. GaMnAs) which are intrinsically ferromagnetic, and second, there are materials where nanoparticles of magnetic material are embedded in a semiconductor matrix. For the nanoparticle materials, the importance of obtaining the correct nanocrystal size was stressed, since crystals that are too small will behave superparamagnetically. Careful annealing of GaMnAs grown with a high Mn concentration gives rise to Mn-rich hexagonal crystallites of an appropriate size, which are ferromagnetic at room temperature. Work on the overgrowth of GaAs with a MnAs layer was also discussed, and again the importance of obtaining the correct phase was stressed.
Richard Campion discussed the work at Nottingham University, UK, on the DMS material GaMnAs. Currently the DMS materials are only ferromagnetic at temperatures well below room temperature, but the maximum Curie temperature of this material system has risen from 100 K in 2001 to a current value of 160 K. Campion s talk concentrated on the mapping of the low-temperature growth-phase diagram and the annealing procedures that have led to this improvement in Curie temperature. The detrimental role of interstitial Mn and arsenic antesite defects was discussed, along with strategies to reduce these. Data presented showed that the increase in Curie temperatures can be directly related to increased hole densities, which is consistent with theoretical predictions.
The first session was opened by N Grandjean of CRHEA-CNRS Valbonne, France, who discussed the progress in MBE growth of GaN and compared the use of N2-plasma with ammonia (NH3). The two methods require different approaches; with N2-plasma the best layers are grown under gallium-rich conditions, whereas nitrogen-rich conditions are best for NH3. A comparison was made of growth onto sapphire, silicon and single-crystal GaN substrates. This showed that the roughening of the surface of MBE-grown GaN is due to kinetic phenomena, not edge dislocations as is often thought. Finally, the effect of NH3 on the surface free energy of strained GaN (on AlN) was shown to be a useful tool to control (reversibly) the 2D-3D transition utilized in the formation of quantum dots.
F Semond, also from CNRS, focused on the growth of GaN on silicon and SiC substrates using NH3. He showed that the introduction of a thick AlN interlayer on top of GaN nucleated on a thin AlN layer allows the relaxation of strain. This increases the critical thickness for cracking of a subsequent layer to several microns. Data for AlGaN/GaN HEMTs grown in this way on SiC were presented, and included mobilities of 10,000 cm2/Vs (at 10 K) and electron sheet densities of 1013/ cm-2.
G Kobulmuller of Infineon Technologies discussed experiments that closely monitor the growth rate of GaN in the early stages of nucleation by line-of-sight mass spectrometry of the Ga desorption rate from the surface. A drop in desorption rate is associated with the opening of the nitrogen shutter, and thus the growth rate can be deduced. This study, combined with AFM and RHEED data, shows that growth starts instantly on sapphire but in two stages on an AlN buffer. The detail of this two-stage process is dependent on both temperature and III/V ratio. The change of the growth rate from being linear to exponential occurs at 2.5 monolayers, which coincides with 2D-3D transition to Stranski-Krastanow growth (figure 1).
J Schmitz of the Fraunhofer Institute, Freiburg, Germany, talked in detail about the wide variety of heterostructures afforded by the type-I and type-II band alignments in the (AlGaIn)(AsSb) system. This allows the formation of structures with effective bandgaps between 20 m2/Vs and carrier densities of 4-5 x 1015/m2). He reported a novel low temperature/low growth rate method for a 15 nm AlAs QW with AlGaAs spacers above and below, which separate the QW from Si-doped regions above and below the spacer layers. The low growth temperature prevents Si migration into the AlAs well. Hall data showing the integer Hall effect and FQHE were presented, demonstrating the high quality of the samples.
Aggressive scaling in CMOS devices is leading to ever-thinner gate oxides. As electron tunneling increases exponentially with decreasing oxide thickness, the CMOS industry is now looking for new materials with higher dielectric constants (K) to replace SiO2. Single-crystal metal oxides are likely high-K candidates, and MBE is being explored as the deposition method.
Hans-Jorg Osten of the University of Hanover, Germany, presented a road map for the future development of CMOS devices with particular focus on the equivalent oxide thickness (EOT) of the gate insulator layer. Because of the strength of the Si-O bonds, the matching of the oxygen sublattice with the silicon lattice is of critical importance. After consideration of several candidate metal oxides, work involving praseodymium oxide (as Pr2O3 in the bixbyte structure) was described. Growth on Si (001) leads to orthogonal (110) domains and overgrowth with Si has not yet been achieved. However, Osten s group has obtained perfect oxide growth on Si (111) and overgrown the oxide layers with Si. The layers showed extremely good properties with K of about 30, compared with 20 for SiO2, low leakage, high breakdown voltage and, importantly, high thermal stability.
J Fompeyrine of IBM s Zurich labs continued Osten s theme with a discussion of other candidate materials. The work focused on (La,Zr) oxides using sequential deposition approach to tune the lattice match of the oxide layer to the silicon (Compound Semiconductor September 2001). The most successful single-phase layers have been grown on Si(111), and show promising properties.
There is much interest in using strained Si channels for future generations of CMOS devices (Compound Semiconductor September 2002). S Mantle of the Jülich Research Center, Germany, reported a novel method of obtaining relaxed SiGe on which to grow strained Si. This starts with the growth of strained SiGe on Si(001), followed by ion implantation of He+ ions at 18 keV into the Si substrate to create a defect-laden layer just below the SiGe. A carefully controlled anneal causes the formation of "bubbles", creating dislocation loops that move to the interface and yield strain relieving misfit dislocations. The degree of relaxation depends on the He dose.
B Voigtländer and J Myslivecek of the Jülich Research Center gave a joint talk to introduce the session on Si growth kinetics. They described equipment that can take real-time STM images of the evolving growth surface (figure 3). The audience was treated to astounding "movies" showing details of the growth process on the atomic level. Examples of these can be seen at www.fz-juelich.de/video/voigtlaender. They then addressed the problem of step bunching in the growth of Si on vicinal Si (001) and introduced the results of kinetic Monte Carlo simulations. These showed that the diffusion anisotropy inherent on the observed 2 x 1 reconstructed surface gives rise to a sticking anisotropy and an effective downward mass current, which leads to instability and step bunching. The model suggests that this behavior will only occur in a certain window of growth temperatures.
S Andrieu of CNRS, France, introduced the meeting to the field of MBE of metals. One main advantage of MBE for metal film deposition is that, due to the relatively low surface diffusion energy, growth can take place at or near room temperature. This allows for the alloying of normally non-miscible metals such as AgCu and AuNi. Also, due to the elasticity of metals, thin highly strained metastable structures can be grown, including face-centered cubic Co or Fe, and body-centered cubic Ni. Thus there are many possible substrate to epilayer relationships.
The discussion then moved to the growth of magnetic tunnel junctions, which consist of two magnetic layers separated by a thin (~1 nm) oxide layer. In such a device the electrons tunnel through the barrier while preserving their spin state. In an ideal device where magnetic polarization is 100%, if the magnetization directions of the two layers are parallel then current will pass; if they are anti-parallel then current will not pass. In real devices using Fe, Co, Ni and associated compounds, the polarization of the electrodes is not 100%, but the search is on for new materials (e.g. Fe3O4, CrO2 and LaSrMnO3), and MBE is at the forefront of this work.
In a related area to the session on metals, L Daweritz of the Paul Drude Institute gave an introduction to the field of ferromagnetic semiconductors and their possible use in spintronic devices. Compound semiconductors are ideal candidates for spintronic devices because ferromagnetic metals can be epitaxially deposited on their surfaces, opening another application for MBE. Daweritz discussed two possible routes to forming the magnetic semiconductors. First, there are dilute magnetic semiconductor (DMS) materials (e.g. GaMnAs) which are intrinsically ferromagnetic, and second, there are materials where nanoparticles of magnetic material are embedded in a semiconductor matrix. For the nanoparticle materials, the importance of obtaining the correct nanocrystal size was stressed, since crystals that are too small will behave superparamagnetically. Careful annealing of GaMnAs grown with a high Mn concentration gives rise to Mn-rich hexagonal crystallites of an appropriate size, which are ferromagnetic at room temperature. Work on the overgrowth of GaAs with a MnAs layer was also discussed, and again the importance of obtaining the correct phase was stressed.
Richard Campion discussed the work at Nottingham University, UK, on the DMS material GaMnAs. Currently the DMS materials are only ferromagnetic at temperatures well below room temperature, but the maximum Curie temperature of this material system has risen from 100 K in 2001 to a current value of 160 K. Campion s talk concentrated on the mapping of the low-temperature growth-phase diagram and the annealing procedures that have led to this improvement in Curie temperature. The detrimental role of interstitial Mn and arsenic antesite defects was discussed, along with strategies to reduce these. Data presented showed that the increase in Curie temperatures can be directly related to increased hole densities, which is consistent with theoretical predictions.
