Materials Science

Professors: B Monemar, head of division; E Janzén, adjunct prof, semi-conductor physics, (100%); L Lindström, adjunct prof, semiconductor physics, (20%); U Lindefelt, adjunct professor, semiconductor physics, (20%). Additional senior research staff: L I Johansson, docent, P O Holtz, docent, W M Chen, docent, H Weman, docent, P Bergman, Dr Techn, A Henry, PhD. In addition there are continuosly between 5 and 10 post doc visitors in the division. We also have 10 PhD students. Technical support: L Samuelsson, research technician, (PhD), (25%), O Kordina, research engineer (techn lic 50%), S Andersson, research engineer (civ ing, 100%), C Hallin, research engineer (civ ing, 100%), A Eklund, research engineer, lab manager, (50%), K Larsson, research engineer, (25%).

The budget for research in our division was about 9 MSEK during the year 930701-940630, excluding equipment grants. The major part of this budget comes from external sources, the faculty support is about 1.6 MSEK.

We would like to point out that a substantial part of the SiC research program is financed via the NUTEK/NFR Thin Film Consortium.

The research activities in the materials science division cover a broad spectrum from basic research (which dominates) to projects of an applied character, partly in direct cooperation with industry. There is a strong international cooperation within most research projects. The present research program can be divided into the following partly overlapping areas:

- studies of the electronic properties of semiconductor quantum structures, such as heterostructures, quantum wells, quantum wires and superlattices, with various spectroscopic techniques.

- studies of electronic properties of defects in semiconductors, including both bulk, surface and interface defects, with several techniques including laser spectroscopy and magnetic resonance.

- preparation and characterization of epitaxial films of semiconductors with molecular beam epitaxy (MBE), chemical vapor deposition (MOCVD) and liquid phase epitaxy (LPE).

- studies of bulk and surface bandstructure, geometrical surface structure and adsorption phenomena on clean surfaces from single crystals (semiconductors, metals, alloys) as well as from interfaces and polycrystalline materials, with photoemission (including synchrotron radiation), inverse photoemission, and other surface-sensitive techniques.

Below we shall present an account of the results obtained during the last year. We find it useful to divide the description under four major headlines, the three first ones describe the activities in semiconductor physics, and the fourth research in photoelectron spectroscopy:

I. III-V Semiconductors and Quantum Structures

II. Silicon Carbide

III. Silicon, Porous Silicon and SiGe Structures.

IV. Electronic and Surface Structure of Metals and Compounds.

I. III-V SEMICONDUCTORS AND QUANTUM STRUCTURES

I.1 . Optical properties of quantum wires and quantum dots

(H Weman, J Hammersberg, J P Bergman, C I Harris, K Swiatek)

I.1.1. (Al,Ga)As serpentine superlattice quantum wires

General optical properties. The optical properties of various types of quantum wires fabricated at the QUEST Center at University of California, Santa Barbara (M S Miller, J L Merz and P M Petroff) are studied by luminescence spectroscopy. In particular a detailed study has been made on the (Al,Ga)As quantum wires grown on vicinal (100) GaAs substrates by molecular beam epitaxy. From polarization-dependent photoluminescence (PL) of (Al,Ga)As serpentine superlattice (SSL) structures grown on vicinal GaAs substrates, an optical anisotropy has been found which confirms the 1D properties for the valence band. Conclusive evidence about the origin of these PL peaks and bands was obtained by PL excitation (PLE) and a complete analysis of the polarization, recombination lifetime, diamagnetic shift, thermal activation energy and pump power dependence.

We have measured PLE spectra and radiative lifetimes as a function of temperature. At low temperature we observe excitons to be strongly localized within potential fluctuations along the wires. The degree of localization decreases with increasing temperature, a localization depth of approximately 10 meV is estimated from the thermalization behavior. At higher temperatures, the excitons have sufficient thermal energy for motion along the wires. The radiative lifetime increases continuously with temperature, reaching 20 ns for the free carrier recombination at room temperature. The long recombination lifetime and high external PL efficiency are due to the lateral barriers between the wires in the SSL structure and could have important implications for their use in semiconductor lasers. In linear polarized PLE, we have been able to resolve an SSL-induced heavy-light hole splitting of the ground state exciton. The observed heavy-light hole splitting is of the order of 5 meV with an electron-heavy hole subband separation of 15-20 meV. We extract the lateral composition modulation between the barrier and the wire to be 15% from the observed heavy-light hole splitting, corresponding to a lateral potential-energy difference in the conduction band of about 120 meV.

Microwave modulation. We have studied the modulation of exciton PL in (Al,Ga)As SSLs, induced by external microwave and magnetic fields at low temperature. We observe that the inhomogeneously broadened PL band of excitons is split into high-energy positive- and low-energy negative- modulation components. Analysis of the spectral shape of modulated PL and its dependence on microwave power draw us to the following interpretation. Photo-excited free carriers are heated by microwave irradiation. They can activate (by impact) the excitons in the wire region from localized to more delocalized states (with higher energy). The localized excitons are "trapped" by short-range potential fluctuations in interfacial islands, whereas delocalized (activated) excitons are more mobile and can move along the wire axis. Thus, the microwave radiation can change the distribution of excitons in localized to more delocalized states in the quantum wire. From a comparison of the cyclotron resonance spectra we can also estimate in a qualitative way the relative degree of carrier confinement in the ordered (Al,Ga)As structures.

High magnetic fields. Low-temperature high magnetic field studies on quantum wire arrays have been performed in fields up to 24 Tesla (in collaboration with Dr M Potemski, High-Magnetic Field Lab, Grenoble). The PL and PLE from (Al,Ga)As SSLs in a magnetic field applied parallel and in two perpendicular directions to the wire was studied. In PL we found an anisotropy in the diamagnetic shift as well as in the linewidth dependence. The amount of the energy shift is critically dependent on the direction of the applied magnetic field. The energy shift for a band-to-band electron-heavy hole transition in the SSL structure has been calculated for the three perpendicular magnetic field orientations (Dr M Lazzouni and Prof L J Sham, UC San Diego). In the model the intermixing of the barrier and well material has been included and the best fit to the data is achieved with an Al composition in the barrier (wire) of 16 % (4 %).

The data give a clear indication that the ground state exciton wave packet is experiencing an anisotropic environment. The linewidth anisotropy dependence is interpreted as a result of how the shrinking exciton volume is probing the amount of potential fluctuations in the ordered AlGaAs structure. This indicates in particular that the lateral potential is not very uniform. In PLE we observe four subbands which all have a weak diamagnetic field dependence. When the magnetic field is applied perpendicular to the plane of the quantum wire superlattice we observe (spin split) Landau-level-like magneto exciton transitions from fields as low as 2 Tesla. We find that the exciton binding energy increases from 8 meV for a 2D reference to 11.5 meV and 13 meV for the 54 Å and 160 Å wide wire, respectively. The decreased binding energy for the narrowest wire is due to the strong lateral penetration of the exciton wavefunction into the barrier region and is in good agreement with recent calculations.

I.1.2. InGaAs/InP quantum wires

We have intensified our collaboration with NTT, Opto-electronics lab., Atsugi, Japan (Dr Tamamura and Notomi) to study the optical properties of lattice-matched InGaAs/InP and strained InAsP/InP quantum wires. Dr. Weman spent six months at NTT where he was involved in the fabrication, design and characterization of these wires. The ultra-narrow quantum wires are fabricated by electron-beam lithography, reverse-mesa wet etching, and an overgrowth achieving a lateral dimension of down to 100 Å. InGaAs/InP quantum wires have been studied with magneto-photoluminescence measurements. The results show a gradual transition of the quasi-one-dimensional sub-bands to quasi-two dimensional Landau levels with increasing magnetic field. This transition appears at fields higher than 10 T for the most narrow wires. This is an evidence for the strong lateral confinement in these wire structures. A pronounced effect of the lateral barriers was the quenching of classical Landau-orbits with cyclotron-diameter wider than the wire width. At higher magnetic fields the magneto-photoluminescence spectra split into clear discrete peaks for each sub-band. By varying the excitation density at 14 T we are able to observe some intriguing many-body effects. E.g. we see a reduced bandgap renormalization for the most narrow quantum wires. In the near future we are planning to study the optical properties of the strained (modulation doped and undoped) InAsP/InP quantum wires.

I.1.3. Selforganized InGaAs quantum discs

We are studying the optical properties of self-organized InGaAs quantum dots fabricated at the NTT Opto-electronics lab. in Atsugi, Japan (Dr Tamamura, Dr Nötzel and Dr Temmyo). The disks are formed during spontaneous reorganization of a sequence of AlGaAs and strained InGaAs epitaxial films grown on GaAs (311)B substrates by MOCVD. The size of the quantum dots are as small as 20 nm, and the PL is characterized by narrow linewidths and well resolved exciton resonances in excitation spectroscopy.

The magneto-luminescence properties of InGaAs quantum disks have been studied using the High-Magnetic Field Lab. in Grenoble. The circularly polarized excitation spectra have been measured for three different disk sizes in fields up to 20 Tesla. Near 10 Tesla we see several higher Landau levels in the disks with the number of levels depending on the size of the disk. Above 10 Tesla the magnetic confinement starts to dominate and there is not much difference in the Landau-level energy spectrum between a disk and a reference quantum well sample. There are however some interesting differences in the oscillator strength for the different polarization states. In addition we have investigated the high-excitation spectra using a pulsed Nd:YAG laser of one quantum disk (80 nm wide) in fields up to 28 Tesla.

Further experiments on the quantum disks are directed towards more systematic studies for different disc sizes under high excitation and to study the interaction between excitons in strongly 3D confined systems.

I.2. Optical properties of the GaAs/AlGaAs two-dimensional system

I.2.1. Undoped GaAs/AlGaAs quantum wells

I.2.1.1. Exciton dynamics

(C I Harris, J P Bergman, P O Holtz, Q X Zhao, and B Monemar)

The radiative emission in structures showing quantum size effects is well known to be dominated by exciton recombination. It is therefore important to have a clear understanding of the interaction of excitons with impurities and defects in the structure, in particular at the interface between layers. The predominant interaction mechanisms occur on a time scale from one picosecond up to one nanosecond, a range accessible to measurements using a streak camera. Using this equipment we have studied the detailed mechanisms determining the capture of a free exciton at impurities and at defects present at the interface between the quantum well and the barrier layers.

Localisation as a mechanism in narrow quantum wells has indeed been found to be a key to many of the radiative properties observed for these structures. Continuing work over the last few years has highlighted a number of fundamental mechanisms which depend on localisation. We have shown that for narrow quantum wells the capture of excitons at impurities becomes extremely inefficient and is thus of relative unimportance for the smallest quantum structures, this result is due to the importance of interface roughness in these systems. The localisation of a free exciton due to a non-smooth interface inhibits the motion of the free exciton, such that there is a reduced probability of being in the vicinity of an impurity and therefore able to be captured. The presence of this mechanism leads to a temperature dependence of exciton capture that is strongly anomalous in comparison to the recognised behaviour in bulk material. We have also recently demonstrated that this same mechanism is particularly important for lower dimensional systems (see section I.1.1) where the additional confinement of carriers severely limits impurity capture.

A further example we have studied is how the interaction of the Coulombic potential of a shallow impurity together with the localising potential due to interface roughness affects the total confinement of the exciton. The two potentials are essentially independent, however, the confinement of an exciton will be determined by interface roughness in the case that the localisation island, in which the impurity is located, defines a smaller radius than does the Bohr radius of the exciton at the impurity. For narrow quantum wells we have found a strong dispersion in the exciton to impurity binding energy as a result of this coupling of potentials.

Using a time resolved technique for studying the dynamics of exciton recombination we are now beginning to better understand the interactions that determine the final radiative recombination. From this basic level we can move on to understand how the mechanisms are perturbed under extreme conditions which for example mimic the parameters under which practical devices such as semiconductor lasers will operate.

I.2.1.2. Non-radiative recombination processes

(J P Bergman, C I Harris, P O Holtz, Q X Zhao, and B Monemar)

GaAs/AlGaAs is widely used for optoelectronic components as both emitters and detectors of light. We have previously shown that the recombination at low injection at room temperature in GaAs/AlGaAs quantum wells (QW's) is dominated by nonradiative recombination. This is evident from the observed decrease of the PL intensity and the decay time as a function of temperature. We have shown that this channel is most likely due to thermal escape of the carriers out of the well followed by a non-radiative recombination via deep defect levels in the AlGaAs barrier layer. We are now investigating QW's grown by different methods in order to determine whether this process is related to the growth method. We are also studying samples with different growth structures, like different numbers of multiple QW's or with additional barriers on each side of the QW, to study the changes of the decay time due to carrier escape.

At low temperatures the non radiative recombination processes are of minor importance and the recombination is dominated by radiative processes. We have, however, intentionally introduced deep level defects in undoped GaAs/AlGaAs QW's by electron irradiation. This is expected to produce deep defects in both well and barrier material. In the irradiated material we observe both a decrease of the total PL intensity and a decrease of the observed decay time, with increasing irradiation dose. At the highest irradiation level we observe decay times faster than 50 ps, which is comparable to the time resolution of our present system. In these samples we do not observe the temperature dependence of the decay time which is normal for a radiative recombination. It is obvious that the recombination is determined by a non radiative recombination process.

We have further observed that the non radiative recombination process can be saturated or deactivated by intense optical illumination. Using normal pulsed excitation intensity, we observe an increase of the low temperature decay time from 50 ps to several hundreds picosecond during a time period of about 10 minutes. The final value for the decay time is comparable to the decay time in the non irradiated samples. We have not yet determined the actual non radiative recombination process or what is required in order to activate this process after illumination.

I.2.1.3. Shake-up processes

(P O Holtz, Q X Zhao, and B Monemar)

Shake-up transitions involving QW hole subbands have been observed as satellites in selective photoluminescence spectra of undoped GaAs/AlGaAs QWs. These shake-up transitions are explained in terms of an interaction between localized excitons and valence-band hole states attached to the QW subbands, in which holes are shaken up from the n=1 heavy hole subband to higher subbands, either the n=1 light hole subband or the n=2 heavy hole subband. The required localization is due to the interface roughness; thus these new transitions are of intrinsic origin. From the observation of the intersubband shake-up processes we derive direct information about the hole inter-subband energies. Furthermore, the satellite intensity is strikingly enhanced in the presence of a magnetic field due to an increasing exciton localization related to the compression of its wave function in the field. The exciton wave function compression continues until its radius in the plane of the well is comparable with the radius of the "flat island" characterized by constant QW width. Accordingly, we can from the magnetic field dependence of the shake-up satellite intensity roughly estimate the size of the "flat islands" and consequently probe the interface roughness. For instance, for a 50 Å wide QW, the intensity enhancement of the satellite reaches its maximum at ~10T, which corresponds to a wave function radius of ~90+/-10 Å. Although the accuracy of the method used is limited, we derive unique information on the dimensions of the island size for characterizing the interface smoothness.

I.2.2. Doped GaAs/AlGaAs quantum wells

I.2.2.1. Acceptors in GaAs/AlGaAs QWs

(P O Holtz, Q X Zhao, A C Ferreira, J Dalfors, M Singh, B Sernelius and B Monemar)

The electronic structure of the confined acceptor. The confinement effects on the ground state as well as the excited states of the shallow acceptor in center-doped GaAs/AlGaAs quantum wells have been investigated via the neutral acceptor bound exciton (BE). Several novel excited states of the acceptor have been observed both via resonant Raman scattering and two-hole transition satellites of the BE recombination observed in selectively excited PL (SPL) and PLE spectroscopy. The observation of satellites in SPL spectra also opens the possibility to monitor the BE in PLE spectra. By detecting the THT satellite in PLE, the splitting of the acceptor 1S ground state into the hh-like acceptor 1S([[Gamma]]6) ground state and the lh-like 1S([[Gamma]]7) state has been spectrally resolved for the first time. The proposed interpretation of the acceptor ground state splitting has been confirmed by polarized PLE measurements.

Magneto-optical properties. The observed acceptor bound exciton (BE) peak is found to split into two components in emission in the presence of a magnetic field. In order to further proceed with the interpretation of the magneto optical results achieved, the electronic structures of the initial excited acceptor BE state and the final acceptor ground state have been treated separately. The g-values for the confined acceptor holes describing the linear magnetic field splitting of the acceptor 1S[[[Gamma]]6] ground state for varying QW width have been calculated. The excited acceptor states are computed within a four-band effective-mass theory, in which the valence-band mixing as well as the mismatch of the band parameters and the dielectric constants between well and barrier materials have been taken into account. The calculated g-values for the acceptor 1S[[[Gamma]]6] ground states, i.e. the final state in the PL emission, in a 150 Å wide QW is g3/2 = 0.61 and g1/2 = 0.35. For the initial state in PL emission, the acceptor BE state, the j-j coupling between the two bound holes and the single electron involved in the BE system gives rise to three BE states with J = 1/2, 3/2 and 5/2, of which the J = 5/2 BE state is concluded to be at lowest energy for shallow acceptor BEs in our QWs. Accordingly, the BE emission observed in PL at low temperatures originates from the initial J = 5/2 neutral acceptor BE state to the final neutral acceptor 1S[[[Gamma]]6] ground state. The derived Zeeman results agree well with the theoretically predicted exciton transitions in the presence of a magnetic field.

Furthermore, the energy separation between the acceptor mj = +/-3/2 1S[[[Gamma]]6] ground state and the excited mj = +/-3/2 2S[[[Gamma]]6] state as a function of applied magnetic field has been determined from selective photoluminescence and resonant Raman scattering measurements. These experimental results have also been compared with theoretical predictions for the confined acceptor states.

Doping level dependence. A systematic study of various optical properties on the acceptor doping up to 6x10¹⁸cm^-3has been performed. Several unexpected and interesting properties have been revealed. The excitons exhibit a red shift with increasing acceptor doping level. Two effects dominate the exciton position: The blue-shift due to band-filling (the Burstein-Moss shift) and the counteracting effect due to the bandgap renormalisation. For n-type GaAs, the Fermi-level shift dominates, which results in an exciton blue-shift, while the reverse situation applies to p-type GaAs. The experimentally observed red shift in our p-type quantum wells is in nice agreement with theoretical predictions if band-filling and many-body effects such as exchange and correlation effects are taken into account. Another interesting observation is the decreasing energy separation between the light hole and heavy hole excitons with increasing doping level. Also this at first sight surprising doping dependence is fully explained within the above described theoretical model. Furthermore, a novel bound exciton with a larger binding energy than the normal neutral acceptor bound exciton, appears at higher doping levels. This novel exciton is interpreted as the ionized acceptor bound exciton.

I.2.2.2. Donors in GaAs/AlGaAs QWs

(P O Holtz, Q X Zhao, and B Monemar)

Satellite spectra. The excited states of the confined Si donor have been monitored via two electron transitions (TETs) of the confined donor BE, observed in SPL and PLE spectra. The energy position of the observed TET satellite yields an accurate determination of the 1S - 2S transition energy for donors in QWs. Furthermore, the TET satellites in SPL spectra have been monitored also in the presence of a magnetic field. From such magneto-optical studies of TET satellites, the dependence of S-like excited states of the confined donor on the applied magnetic field has been determined for the first time. Previously, only transitions between the ground state and p-like excited donor states, observed in optical absorption, have been experimentally studied.

The D^- centers. The negatively charged donor is formed when a neutral donor traps an extra electron. There are accordingly two electrons bound to this D^- center, which form a singlet state at lowest energy and a higher triplet state. This D^- center has recently attracted a lot of attention. In addition to a number of theoretical predictions of the D^- center, the experimental observation of the D^- center by far infrared magneto-optical measurements has recently been reported. We have observed a novel feature in photoluminescence spectroscopy, which is interpreted as the D^- bound exciton. The binding energy of this exciton is dependent on the excitation intensity level, but measures to 1.3 meV for a 200 Å wide quantum well at low excitation intensity conditions. This D^- bound exciton also exhibits a striking dependence on temperature and applied magnetic field in accordance with the expectations.

I.2.3. The Transition from Quantum Well to Superlattice

(Q X Zhao, P O Holtz, C I Harris, and B Monemar)

This project has been performed in collaboration with E Veje at H C Ørsted Institute, Copenhagen.

The electronic structure. When the barrier widths in multiple quantum wells (MQWs) become narrow enough, neighbouring wells become coupled to each other. The MQWs are converted into a superlattice (SL). The electronic structure is strongly influenced by this coupling strength. With decreasing barrier thickness, Lb, the neighbouring wells become electronically more coupled. The electron and hole progressively lose their 2D character, and the structure is instead gradually transformed to a more three-dimensional (denoted as quasi-3D in the following discussion). We therefore have a unique system, which enables us to control the 2D character relative to the quasi-3D-like character by varying the barrier thickness Lb. A series of samples has been used to investigate both the free exciton (FE) and the Be-acceptor bound exciton (BE) in the transition region between the 2D and the quasi-3D system.

Free excitons. The FE binding energy is strikingly reduced with decreasing barrier thickness. For instance, the 1S-2S energy separation for the heavy hole state of the FE in a 100 Å wide QW decreases from about 8.5 meV to 3.3 meV as the barrier width is reduced from 150 to 10 Å. Our results show that the energy separations between the heavy hole and light hole states of the FE 1S states remain constant for barrier thicknesses down to about 40 Å, to decrease for further reduction of Lb.

The Be acceptors. A similar characterization as for the FE has been carried out for the Be-acceptor BE. From these measurements, it is found that a) the 1S-2S transition energy of the acceptor does not change as the barrier thickness is reduced down to 20 Å, b) the binding energy of the BE (relative the position of the FE) remains constant within the experimental accuracy as the barrier thickness is reduced, c) the decay time of the BE increases drastically with decreasing AlGaAs barrier width. We have concluded that the oscillator strength of the BE in a QW is related to the relative extension of the electron and hole wave functions, and does not necessarily correlate with the BE binding energy.

I.2.4 Coupled double GaAs/AlGaAs quantum wells

(Q X Zhao, B Monemar, P O Holtz)

This project has been undertaken in co-operation with the MBE group at NTH in Trondheim (B Fimland, T Westgaard, K Johannessen).

Electronic structure of excitons. If the barrier in a double quantum well structure is sufficiently narrow, there will be an interaction between the two wells, and we have a so called coupled double quantum well (CDQW) structure. The electronic states corresponding to a single quantum well split into symmetric and anti-symmetric states. The exciton states are formed from either symmetric or anti-symmetric hole and electron one-particle wave functions and are consequently denoted "symmetric" and "anti symmetric" excitons. An important feature in the experimental PLE spectra is a significant reduction of the oscillator strength for optical transitions related to anti symmetric heavy-hole and light-hole exciton states compared to the strength of the symmetric exciton states, when the symmetry splitting becomes comparable to the single quantum well exciton binding energies. This effect is absent in the simplest theory for CDQW's, but can be obtained within an extended effective-mass theory, in which the Coulomb interaction between electrons and holes, which gives rise to the mixing between the symmetric and anti symmetric exciton states, has been taken into account.

The binding energy of the free excitons. To investigate the exciton binding energies, combined unpolarized and polarized photoluminescence excitation measurements have been performed to measure the exciton 2S states. Discrete peaks, which are identified as the excited 2s state of the n=1 symmetric heavy hole exciton, are observed in GaAs/AlGaAs symmetric coupled double quantum well (SCDQW) samples. The 1S-2S energy splitting of the symmetric heave hole exciton in the SCDQW is significantly reduced in comparison with the uncoupled quantum wells. A simple model to calculate the binding energies of the 1S and 2S exciton states is developed. The predicted 1S-2S energy separation derived from such calculations are in very good agreement with the achieved experimental results.

Diamagnetic shift of the FE. The diamagnetic shift of the FE in SDCQW structures has been measured and compared with results from isolated QW. The results show that the diamagnetic shift of the FE in SDCQWs is larger than the corresponding diamagnetic shift in isolated QWs. The experimental results of the diamagnetic shift are also in excellent agreement with our theoretical predictions.

1.2.5. High Density Carrier Effects in Quantum Wells

(J P Bergman)

We have performed studies of the radiative recombination in GaAs/AlGaAs QW's during intense picosecond and femtosecond pulsed excitation. In specially designed GRINSCH - structures (GRaded INdex Separate Confinement Heterostructures ) it is easy to obtain high carrier densities and forming a two-component plasma of electrons and holes. At these carrier densities excitons do not longer exist due to screening and phase space filling effects and the electrons and holes in the well recombine through a band to band transition on a picosecond time scale.

With time resolved spectroscopy on this time scale we have studied the radiative recombination at different carrier densities and observed many-body effects such as bandgap renormalisation and large broadening of all transitions. In narrow QW's, with a well width smaller than 40 Å, we have observed a blue shift instead of the expected red shift due to the bandgap renormalisation. This is in disagreement with standard many body calculations in 2D systems, but recent theoretical studies, which treats the Hartree term on equal footing with the exchange and correlation terms in the calculation, support our experimental results.

This work is expected to continue and evolve due to the installation of new and more powerful pulsed laser systems in the femto- and pico-second time domain in our laboratories.

I.2.6. GaAs/AlGaAs/InGaAs Tunneling Structures

(Q X Zhao, T Lundström, P O Holtz, J P Bergman, and B Monemar)

The pioneering work on the double barrier resonant tunnelling structures (DBRTS) was made 20 years ago (Chang et al, Appl Phys Lett 24, 593 (1974)). There are a number of potential applications based on the tunnelling structures, e.g. the resonant tunnelling light-emitting diode. One important parameter is the on-off ratio for the QW light output when the device switches from resonance to off-resonance. Different ways to improve the optical on-off ratio have been proposed, for instance via the triple-barrier resonant tunnelling structures (TBRTS).

We have based our project on the GaAs/AlGaAs TBRTS structures, but modified this structure by introducing a narrow InGaAs well in the GaAs layer region to further improve the optical on-off ratio. Our preliminary data show that the narrow well has a strong influence on the electronic and optical properties of the TBRTS structures. The optical and electrical characterization will continue on such structures in order to optimise the conditions for the on-off ratio. We will in this project focus on the fundamental properties such as the effects of defects, shallow impurities and deep level centers in both the barrier and the well region on the carrier escape time and the recombination efficiency.

I.2.7. The GaAs/AlGaAs Modulationdoped Heterostructure

(T Lundström, Q X Zhao, P O Holtz, J P Bergman, C I Harris, and B Monemar)

Recombination processes. We have studied the radiative recombination processes in a series of modulation doped GaAs/AlGaAs single heterojunctions, with an active GaAs layer ranging from 400 Å to 1500 Å. In these structures a 2D electron-gas (2DEG) is formed and trapped in a notch potential near the GaAs/AlGaAs heterojunction due to charge transfer from the modulation doping in the AlGaAs barrier to the active GaAs layer. Radiative recombination processes involving this 2DEG have been studied by means of PL,SPL,PLE and time resolved PL. In all our samples a PL-band denoted H-band 1 (HB1) can be observed. This emission is due to radiative recombination between the 2D-electrons and free holes located in the active GaAs layer. The spectral position of HB1 is strongly dependent on the excitation conditions, the emission shifts towards higher recombination energies with increasing excitation intensities. This dependence is explained by a band bending model, where the equilibrium built in band bending in the active GaAs layer is decreased by creation of spatially separated photo excited electrons and holes. By applying an external electric field perpendicular to the layers we can control the band bending in the active region. The behaviour of the HB1 emission with varying applied electric fields confirms the proposed band bending model. The applied field also gives us the possibility to vary the separation between the Fermi level and the next unoccupied 2D electron subband. At sufficiently small separation, an efficient scattering path near k=0 is available for electrons at the Fermi edge, which makes it possible for 2D electrons with an energy close to the Fermi edge to radiatively recombine with free holes at k=0. This many-body excitonic transition, usually referred to as the Fermi edge singularity (FES) can be observed both in PL and PLE in our samples. In time resolved PL we observe a spectral shift of the HB1 emission towards lower recombination energies with increasing time after the pulsed photo-excitation, which also is consistent with the band bending model.

Theoretical model. The interpretation of the experimental data is supported by theoretical calculations of energy levels, wave-functions and band bending in the heterojunction. The theoretical model is formulated in the effective mass approximation, where a one-dimensional Schrödinger equation and Poissons equation are solved simultaneously and self-consistently by numerical methods. Simulations of the dependence of recombination energies on the excitation intensity shows exactly the same behaviour as the experimentally observed. We also plan to simulate the transient behaviour of the radiative recombinations in the modulation doped heterojunction following pulsed excitation and compare with time resolved PL measurements.

Intersubband relaxation at a single GaAs/AlGaAs heterojunction. The triangular-like confining potential at a single heterojunction implies that the wavefunction for progressively higher subbands is considerably more extended than for the lowest subband. As a result the recombination probability increases dramatically for higher subbands. The observed recombination from a single heterojunction therefore depends dramatically on the rate of inter-subband relaxation for photo-excited carriers. We have recently investigated the dependence of relaxation rate as a function of carrier density. By examining the low temperature time-resolved photoluminescence using a streak camera we have been able to demonstrate a dramatic decrease in relaxation rate as the Fermi-level energy increases beyond a threshold fraction of the intersubband spacing (EF/DEn,n+1 > 0.7). The slow relaxation rate implies that a large fraction of the total recombination occurs while carriers remain in the higher subbands. Thus there is a dramatic relative increase in the higher subband emission. In addition we are able to observe under transient conditions the formation of the Fermi-Edge-Singularity (FES). The small variation in Fermi-level relative to the higher subband, during recombination following pulsed excitation, passes through the threshold for FES formation. We are thus able to observe the full kinetics of the heterojunction recombination and to understand the dependence in terms of band bending and relative occupation of different subbands.

I.3. Optical characterization of the InGaAs/InP two-dimensional system

I.3.1 Intersubband Transitions in Strained InGaAs/InP Quantum Wells

(S A Stoklitsky, Q X Zhao, T Lundström, P O Holtz, and B Monemar)

Theoretical calculations. The infrared (IR) detector based on intersubband transitions in p-type quantum wells has been paid considerable attention in recent years. The major advantage of using p-type QW structures for the IR detector is the possibility to monitor infrared radiation at normal incidence. Owing to a strong mixing between the light hole and heavy hole subbands, optical transitions between the valence subbands are allowed also for normally incident light. We have theoretically investigated the lattice mismatch (strain) effects on the infrared absorption in p-type In1-xGaxAs/InP quantum wells for both tensile (x>0.47) and compressive (x<0.47) strains. The results show that normal incidence optical matrix elements substantially increase in the case of the compressive strain and decreases in the case of the tensile strain at a constant hole density. The peak of the normal incidence absorption in the compressively strained QW is shown to reach a considerable value of 5000-6000 cm^-1 for a sheet hole concentration of 10¹² cm^-2. On the other hand, for the z-polarization of the light, we found a substantial enhancement of the optical matrix elements in the case of a tensile strain. We intend to perform far-infrared measurements in order to compare with the calculated results. A prototype infrared detector will be fabricated at the optimised conditions according to our theoretical model.

I.3.2. The InGaAs/InP Heterostructure

(T Lundström, M Singh, Q X Zhao, P O Holtz, and B Monemar)

Fundamental parameters. The InGaAs/InP system has a bandgap energy which is favourable for optical communication device applications, and has accordingly attracted great interest in recent years. But many of the basic physical parameters such as effective masses, bandgap offset and g-values for both electrons and holes etc. are still not well documented. Therefore it is important both for fundamental physical reasons and application purposes, to study the InGaAs/InP system more in detail. We have chosen to concentrate on modulation doped InGaAs/InP heterostructures since they are very interesting as device structures and ideal for studying fundamental physical properties of two-dimensional (2D) carriers. The samples are grown at the Institute of Microelectronics in Kista (G Landgren), and measured by optical methods such as PL, Fourier-transform PL (FT-PL) and Fourier-transform PL excitation (FT-PLE). In PL-measurements we observe band to band transitions involving different subband levels for the 2D carriers. The energy separation between these levels can be used in theoretical band structure calculations from which physical parameters such as the bandgap offset can be deduced. Effective carrier masses can be extracted from magneto-optical PL-measurements, where an applied magnetic field splits the 2D levels into Landau levels. Studies of the effective masses as a function of internal strain, i.e. the In-composition in the InGaAs-layer are in progress.

Many-body effects. The many-body excitonic transition, usually referred to as the Fermi-edge singularity (FES), was also observed in the InGaAs/InP system. Hole localization due to strong potential fluctuations is claimed to be the mechanism for the required momentum conservation in this radiative transition. But in the case of GaAs/AlGaAs systems, where the potential fluctuations are much smaller, momentum conservation for the FES transition is usually explained in terms of a scattering process via the next unoccupied state. By applying an electric field across an InGaAs/InP structure we are able to control the energy separation between the Fermi-level and the next unoccupied state. This gives us a unique possibility to investigate the relative importance of the scattering process compared to the hole localization mechanism for the momentum conservation in the FES transition.

I.4. Low temperature III-V compounds

I.4.1 Low temperature InP

(W M Chen, E Sörman and B Monemar)

This project is a cooperation with Department of Materials Sciences, University of California at Berkeley, USA (E R Weber, P Dreszer, A Prasad, A Kurpiewski, W Walukiewicz).

There has been a continuous effort in searching for methods to produce undoped semi-insulating (SI) InP, as a desirable substrate material or for device isolation in InP-based optoelectronics. The SI behavior of GaAs grown at low temperature (LT), which is highly As rich, has inspired an effort to grow InP at low temperature under phosphorus over-pressure in the attempt to achieve undoped SI InP. Very unexpected and surprising results were observed, however, as as-grown LT-InP was shown to be highly n-type conductive.

In this work we show with correlated optical and electrical measurements that the PIn antisite is the prevailing defect in LT-InP. We show that the increase in free electron concentration measured by the Hall effect measurements correlates very well with the increase in the PIn⁺ concentration determined by magnetic circular dichroism of absorption (MCDA), with decreasing growth temperature. The results therefore give evidence that the dominant deep donor level at Ec + 0.12 eV in LT-InP corresponds to PIn (0/+). It is the auto-ionization of the PIn antisite via its first ionization stage which leads to the n-type conductivity in as-grown LT-InP. This work indicates that the introduction of PIn antisites in InP is not at all suitable to achieve SI material as suggested in the past.

I.4.2 Low temperature GaAs

(W M Chen)

This project is a cooperation with Department of Materials Sciences, University of California at Berkeley, USA (E R Weber, X Liu, A Prasad).

Highly As-rich LT-GaAs grown below 400deg. C has attracted great attention in recent years, because of their unique properties of record-high resistivity, extremely short carrier lifetime and reasonably good mobility. In as-grown LT-GaAs, the excess As exists in the form of As antisite (AsGa) defects in a large quantity (~ 10²⁰ cm^-3). The material is conductive due to carrier hopping between the defects. Upon annealing at ~ 600 deg.C, LT-GaAs becomes semi-insulating. The AsGa defect concentration decreases significantly, accompanied by the formation of As precipitates. There exist controversial models to account for the SI properties of the annealed LT-GaAs, as due to the residual AsGa defects or the As precipitates. If the AsGa defects pin the Fermi level close to midgap, the metal precipitates formed in such a GaAs matrix should have a near-flat-band condition. Hence the key question is whether the AsGa-defect concentration is high enough in the material.

In this work we determine the concentration of AsGa defects in neutral and positive charge states, AsGa⁰ and AsGa⁺, by near-infrared absorption and MCDA. We show that samples grown at ~ 200 deg.C contain a large amount of AsGa defects. The defect concentration decreases significantly for annealed materials, and for samples grown at higher temperatures. However, we find that a high concentration of the defects is always present in the LT-GaAs studied, with the concentration of AsGa⁰ larger than that of AsGa⁺ defects. This result indicates that the defects can account for the pinning of the Fermi energy, and consequently also the semi-insulating properties.

II. SILICON CARBIDE

II.1. CVD growth of Silicon Carbide

(O Kordina, C Hallin, R C Glass, C Hemmingsson, L-O. Björketun,
E Janzén)

Growth of epitaxial layers of SiC has been made by Chemical Vapour Deposition (CVD). The growth is performed in a hot-wall reactor by passing reactive gases (silane and propane or silane and methane) over a heated susceptor where a substrate of SiC or Si is placed. As carrier gas hydrogen is used. N- or p-type doping is achieved by small additions of nitrogen or trimethylaluminium.

The aim with the work has been to produce thick low-doped layers with long carrier lifetimes for bipolar high power devices. The work has been concentrated on the 6H-SiC polytype, however, a considerable amount of fundamental growth studies have been made on the 3C-SiC polytype which may be grown on Si substrates. Recently we have also began studies on the, perhaps, most promising polytype, 4H-SiC. The material we have produced on 6H-SiC is the, to date, purest ever reported. The residual doping concentration of n-type layers is in some cases below 10¹⁴ cm^-3. For uncompensated layers this value is slightly higher, 5 . 10¹⁴ cm^-3. The corresponding mobilities at 6 K is as high as 1.1 . 10⁵ cm²/Vs and the minority carrier lifetimes close to 0.5 us. A bipolar device structure has been made from 6H-SiC with a 45 um thick low-doped n-type base layer and a 1.5 um heavily p-type doped emitter layer. The device structure was processed by the Industrial Microelectronic Center (IMC) in Stockholm into mesa shaped diodes. The reverse blocking voltages obtained by these diodes were as high as 4.5 kV which is the, to date, highest reported value for any wide bandgap material.

The grown 4H-SiC material show even more promising results than the 6H-SiC. Low-doped 2 . 10¹⁴ cm^-3, uncompensated n-type layers have been produced with mobilities as high as 1.8 . 10⁵ cm²/Vs.

Several important studies will in the near future be carried out: The nitrogen and aluminium incorporation of the layers will be investigated, the growth conditions for obtaining high carrier lifetimes will be determined and new hot-wall reactor designs will be looked into.

II.2 Silicon Carbide Bulk Growth

(M Tuominen, R Yakimova, E Janzén)

An experimental set-up for large crystal sublimation growth of SiC has been designed and tested. This is a vertical system consisting of a graphite crucible assembly placed in a quartz tube and heated by RF-power. Graphite felt or graphite foam can be used as heat insulation. The growth process employs a SiC single crystal seed located at one end of a cylindrical growth cavity while at the other end the source material is placed. An axial temperature gradient in the cavity maintains the source material at its sublimation temperature (~2400deg.C) and the seed at a lower temperature (most commonly 2200deg.C). The growth takes place on the seed crystal in an argon atmosphere at 10 to 200 mbar gas pressure.The seeds used in all runs are Lely-grown platelets. The temperature is controlled by a two-wavelength pyrometer. Different RF-generators and different coils have been used to minimize reflected power losses and to increase the heating efficiency. Growth parameters like seed temperature, temperature difference between seed and source, source material, and gas pressure in the growth chamber have been varied in order to optimize the growth regimes. As a result monocrystalline SiC material has been obtained. Depending on the growth conditions the growth rate varies from tens to hundreds um/h. The crystals grown have been characterized by means of microscopy, X-ray diffraction (XRD) and PL techniques. The crystal structure quality is affected mainly by seed imperfections and vapor supersaturation while the doping level depends on the source material purity.

II.3 Dispersive photoluminescence characterisation of SiC

(A Henry, O Kordina, E Janzén)

The main part of the results obtained since the SiC project was started in our laboratory (about 3 years ago), is related to the CVD growth technique. Improvement of the quality of the material for three different polytypes (3C, 6H and 4H) was obtained and we are now able to grow epitaxial SiC film of state-of-the-art purity. The characterisation of the material using various techniques is used as feed back to the growth technique and is greatly responsible for the remarkably fast development of the research project. The photoluminescence technique is used as the first characterisation process of the grown material. We also characterise with the luminescence technique bulk or epilayer samples grown by other groups such as 3C and 6H SiC epilayers grown at IMC, 3C-SiC epitaxial films grown by reactive magnetron sputtering (RMS) by the thin film group at IFM, or bulk and epitaxial samples from external sources (from USA or Russia). All these studies allow us to compare the quality of the samples.

Generally in SiC the dominant PL behaviour is due to donor-acceptor pair emissions giving rise to broad PL bands. This fact is mainly due to the high concentration level of the doping impurities. However in good material it is possible to observe excitons bound to the shallow nitrogen donor(s), characteristic with sharp, near-bandgap and phonon related replica lines. In high purity material strong free-exciton (FE) phonon related lines can be identified. We did observe strong FE related lines, firstly in 6H-SiC and recently also in 4H-SiC, as never reported in the literature to our knowledge. We also observe in some of our 6H-SiC epilayers unresolved aluminium bound-excitons (BE), showing a compensation effect due to the acceptor unintentionally introduced during the growth.

As a first example, the intensity of impurity (bound exciton) and intrinsic (free-exciton) luminescence lines reflect the impurity doping concentration. We have presented a calibration procedure for the nitrogen impurity concentration in 6H SiC epilayers grown in our laboratory. The calibration is valid for a large range of doping from 10¹⁴ cm^-3 to 10¹⁷cm^-3. We also presented the possibility to distinguish between n-type and p-type low doped material by using simply the PL technique.

Characterisation of 6H-SiC epilayers grown at IMC (e.g. in a completely different CVD reactor) is also performed and shows the importance of the growth parameters. From these epilayers very strong hydrogen related lines could be identified, whereas they were never observed in epilayers grown in Linköping. More investigation about hydrogen in SiC is presently in progress.

In high quality 6H CVD epilayer grown on 6H SiC substrate we also observe the DI PL spectrum, which has earlier been observed in as-grown 3C SiC grown on Si substrates. The DI center, a radiation-induced defect, is one of the most studied defects in bulk SiC, both in 3C and 6H SiC polytypes. It is generally observed after implantation with heavy ions and subsequent annealing at temperatures between 1000deg.C and 1700deg.C, and gives rise to a well-resolved line structure in PL spectra. It has been speculated that it is due to a vacancy related defect and to have a Jahn-Teller effect in the PL spectrum. We have started to measure the decay time of the DI center from both 3C and 6H CVD layers. For comparison electron irradiated and annealed materials should also be investigated. In the 3C epilayer the DI spectrum is well separated from other luminescent spectra whereas in 6C SiC the DI spectrum is often superimposed over a broad PL band associated with a donor-acceptor pair emission. The decay time of this background must then be taken in consideration in the analysis. The high-temperature spectrum of the DI center in both polytypes is found to have decay times in the range of tens of microseconds. This confirms that the DI defect behaves as an isoelectronic center.

Investigation of 3C-SiC epitaxial films grown using the RMS method was also performed. The nitrogen BE lines could be observed together with other weak and broad PL lines associated to interface defects. This behaviour proved that 3C-SiC films with a crystalline quality comparable to the best CVD-grown 3C-SiC epilayers could be grown by the RMS technique.

II.4. Infrared FTPL characterization of CVD SiC

(M Singh, O Kordina, E Janzén)

Unintentional contamination with extremely low concentrations of transition metals in SiC is known to give rise to luminescence-active centres that emit in the infrared region. Vanadium is a well-known example of one such transition metal and gives rise to the 0.8 eV emission band. The positions of the zero-phonon lines are dependent on the various polytypes of SiC. The presence of the 6H, 4H and 15R polytypes can be determined from the 0.8 eV vanadium PL band. In addition to this band we also observed a weak band with zero-phonon lines around 1.17 eV. Consistent with the site symmetry, three lines could be observed in 6H material. However, an additional line could be observed in CVD material with significant 3C polytype inclusions. The intensity of this additional line which appears to increase systematically, relative to the 6H lines, as the 3C inclusions incease. We believe that this PL band is related to low level contamination with a transition metal and serves as a PL probe of the presence of 3C-SiC during the growth of 6H. The exact identity of this transition metal is unclear as yet but a similar transition has been observed in other wide-gap materials such as GaN doped with chromium, and is thought to be an internal ³T2(F)--³A2(F) transition in Cr⁴⁺.

II.5 Time Resolved Photoluminescence Measurements in SiC

(J P Bergman, C I Harris, E Janzén)

We have performed time resolved measurements of radiative recombination in different polytypes of SiC. To do this it is necessary to have a pulsed laser excitation source in the UV spectral region, to excite above the bandgap of SiC. This is made possible by up-converting the picosecond pulses from a modelocked IR dye laser by non linear optical crystals. The luminescence is detected by a photomultiplier tube and a standard time correlated photon counting system, with a time resolution in the picosecond time scale.

We have measured the decay time of the bound excitons (BE) at the neutral nitrogen donor in the 3C, 6H, 4H and 21R polytype of SiC. Nitrogen is the most common residual impurity in SiC, and nitrogen BE's are normally the dominating emission at low temperatures. The measured decay times are extremely fast, ranging from 1.5 ns for the S BE in SiC:6H to 160 ns for the BE in 3C SiC. We have found a strong correlation between the observed decay time and the binding energy of the corresponding donor. We conclude that the dominating recombination mechanism for the BE's is a non-radiative Auger process.

We have also studied the decay characteristics of the free exciton (FE) in both the 4H and 6H polytype of SiC. These are only present in low doped material and their presence influences the observed decay of the BE's. A study of the dynamical interaction between the FE's and BE's has been started. At low temperatures we are trying to observe the capture and creation of the BE from the FE state. The opposite process, i.e. the thermal ionisation of the BE into a FE, is studied at higher temperatures where the thermal energy is comparable to the binding energy of the exciton.

We have made extensive time resolved measurements of the optical emission at room temperature in 6H SiC. From this we have obtained the minority carrier lifetime in a series of samples with different doping concentrations. We have found a weak correlation with doping concentration with a maximum lifetime of 450 ns for a sample with Nd= 3 - 5 10¹⁴ cm-³. We have also found a reduction of both the observed lifetime and optical efficiency in samples with compensating doping. This study continues in order to completely understand the processes that limit the carrier lifetime, since long carrier lifetimes is of great importance for power device applications. A special interest is focused on the carrier dynamics during high injection, which previously has not been possible to obtain with our excitation sources.

II.6 Electrolytic Etching of SiC and Formation of Porous SiC

(C I Harris, A O Konstantinov, R C Glass, A Henry, B Monemar,
E Janzén)

The chemically inert nature of SiC makes it a very difficult material to etch in a controlled manner. In order to achieve a chemical reaction vigorous enough to remove significant quantities of material we optimise the physical conditions to drive the reaction. We have recently investigated the use of electrolytic etching of SiC whereby the rate of reaction is enhanced by the application of an electrical potential difference between the SiC substrate and the electrolyte. In addition photo-assisted electrochemical etching provides a further parameter to control the reaction. Our initial work in this area demonstrated that these techniques provide both a means for selective etching and also a valuable tool in the assessment and characterisation of structural defects in the material. For electrolytic etching in the dark the nominal etch rate of n-type 6H SiC is extremely low, this etch rate is however enhanced along defects such as screw dislocations. As a result the etch is able to highlight certain defects, thereby allowing a routine assessment of the structural quality. The use of additional photo-excitation further enhances the reaction such that substantial etch rates are achieved even in low defect areas. By careful optimisation of the parameters an etch can be achieved which yields both a high quality surface and sharp profiles defined by the photo-excitation.

More detailed recent investigation has shown that material produced under low etch-rate conditions, in particular for the non-cubic (a) polytypes, has a number of interesting and unique properties. Scanning electron microscopy (SEM) reveals that the etch process results in a porous surface layer resembling a complex sponge. Careful selection of the preparation conditions allows us to control the mean fibre size within a wide range (200-500nm). By controlling the size of the fibres which make up this 'sponge' we radically alter the electrical properties of the material. A small fibre size results in a very high resistivity material. A patent application in collaboration with ABB has indeed been made to use the formation of a high resistivity layer as a means of device isolation and passivation. The larger fibre size material has relatively lower resistivity but also shows interesting properties including high photo-sensitivity. Work is continuing to demonstrate the use of porous SiC both in photo-sensitive and chemical sensitive detectors. The mechanism for the unusual properties of the porous layer has been investigated using electrical (I-V, C-V) and optical (photoluminescence (PL), photo-conductivity) techniques. We have shown conclusively that the dominant process is one of Fermi-level pinning at surface states in the fibres produced by the etch. A post oxidation treatment passivates these surface states and results in a lower resistivity. In particular C-V measurements provide a nice illustration of the mechanism of depletion within thin fibres as the capacitance is modulated according to the active surface area.

PL measurements have demonstrated a strong room temperature blue-green luminescence resulting from the etched layers. This was originally interpreted in terms of a similar model to that proposed for porous silicon, although the efficiency was considerably lower. Further work highlighted the fact that the emission is independent of SiC polytype. We therefore interpret the luminescence as being from molecular-like states at the etched surface. These states, as yet unidentified, are almost certainly related to those responsible for Fermi-level pinning as the post oxidation process also results in a loss of luminescent emission.

Work on the use of electrochemical etching will in the future focus primarily on the use of the porous material for practical application in power device termination and sensors, but further study on the identification of the surface state mechanism will also continue.

II.7 Hydrogen passivation of SiC

(C I Harris, A O Konstantinov, E Janzén)

Preliminary experiments have recently been carried out to investigate the passivation of defects/ impurities in SiC using a dc hydrogen plasma technique. A dc plasma is used to prevent surface damage of the material which has been shown to be a problem in the treatment of semiconductors. Passivation of both n and p type material has been achieved using a potential difference applied to the sample to enhancethe hydrogen ion flux and to lower the surface barrier to diffusion. At present a relatively low passivation efficiency has been demonstrated from increased sample resistivity, however further work is underway to use higher substrate temperatures during passivation in order to achieve a higher rate of hydrogen diffusion.

II.8. ODMR characterization of SiC

(W M Chen, N T Son, E Sörman, B Monemar, E Janzén)

As the first goal of this project, we attempt to reveal and investigate the lifetime limiting defect in SiC by optical and magnetic resonance spectroscopies. The samples studied include 4H and 6H bulk crystals and epilayers grown by CVD, either n- or p-type doped. The carrier recombination processes are monitored here by PL emissions from the materials. A defect-related magnetic resonance signal at around g=2 is observed via these PL emissions, giving rise to the so-called optically detected magnetic resonance (ODMR) signal. The resonance line at X-band frequencies probably contains unresolved hyperfine structure. This ODMR signal corresponds to an enhancement in the intensity of the PL band peaking at about 1.7 eV and a decrease in the intensity of other PL emissions ranging from near band edge shallow bound excitons to near-infrared deep PL bands. This is interpreted as due to the competing process in carrier recombination: the magnetic resonance at the dominant recombination center promotes carrier recombination via this center (the PL band at 1.7 eV) and consequently reduces carrier recombination via other recombination channels. This dominant recombination center is shown to be a deep level defect, evident from the related deep PL emission and a photo-excitation spectrum of the ODMR signal. The latter reveals the photo-ionization of the center, which determines the energy level of the defect to be at about 1.1 eV below the bottom of the conduction band. The fact that this ODMR signal has been observed in both 6H and 4H SiC, regardless of their doping type and whether they are bulk or epilayers, indicates that this must be a common and basic defect in SiC. This defect provides a very efficient recombination channel, limiting the carrier lifetime.

II.9. ODCR studies of SiC

(W M Chen, N T Son, B Monemar, E Janzén)

We employ optically detected cyclotron resonance (ODCR) technique, for the first time, to determine electron effective masses and mobilities in 6H and 4H SiC. High purity undoped 6H and 4H layers grown by CVD, with residual n-type doping concentrations in the range of 10¹⁴ - 10¹⁵ cm^-3, were investigated.

In 6H SiC, the electron effective mass values were determined as m||* = (2.0+/-0.2)m0 and m^* = (0.42+/-0.02)m0. From the fit of the ODCR line shape, a remarkably high mobility at 6 K was deduced as u^ ~ 1.1x10⁵ cm²/Vs for electrons in the basal plane. In 4H SiC, the electron effective mass values were obtained as m||* = (0.29+/-0.02)m0 and m^^*= (0.42+/-0.01)m0. The electron mobility in the basal plane at 6 K was deduced as u^ ~ 1.8x10⁵ cm²/Vs. These results represent the first direct experimental determination of the electron effective masses in 6H and 4H SiC. These results are consistent with the assumption that the electron valleys in 6H and 4H SiC can be described as ellipsoids, which locate at the M point on the side planes of the Brillouin zone with the principal axis along the c-axis.

II.10 Structural analysis of SiC wafers

(M Tuominen, L -O Björketun, R Yakimova, R C Glass, E Janzén)

Synchrotron X-ray topography has been combined with high resolution X-ray diffraction (XRD) analysis and microscopy examination to study specific structural imperfections in commercially available 4H SiC wafers. Preferential chemical etching has been additionally applied to reveal strain associated defects. Several types of macro-defects have been observed: micropipes, basal plane tubes, dislocation networks comprising domain boundaries, strain and cracks. It has been concluded from the high resolution XRD measurements that mosaicity with domain misorientation is a dominant imperfection in the wafers. Synchrotron topographs have shown strain contrast around micropipes and little or no strain associated with cracks. It has been found out in the section topographs that misorientation occuring in the crystal is closely related to the micropipes.

A comparison has been made with the crystal structure quality of Lely grown platelets. The results obtained suggest that the growth direction and the growth process (growth rate and supersaturation) play an essential role in crystal defect formation. The investigation of structural defects in 4H SiC wafers has shown some similarity with the 6H material as far as micropipes and mosaicity are concerned. However it appears that the micropipe distribution in 4H material is somewhat different. A model for micropipe formation has been proposed. It seems to be more difficult to prevent crack formation in 4H wafers because of higher supersaturation required for 4H polytype formation and higher growth rate compared with the Lely process. The post growth processing of the wafers might also be critical for developing additional structural macro-defects.

TEM and XRD have been used for characterisation of 3C-SiC grown by CVD. The work carried out so far has been devoted to study the effect of growth parameters on film quality. Conventional X-ray diffraction was used for phase analysis and to study the Full Width Half Maximum (FWHM) of the SiC(200) peak to estimate film quality. In addition to conventional X-ray diffraction there were also X-ray texture measurements done to confirm that the films were single crystalline. Cross-sectional TEM has been used to study the quality of the interface, the top surface and the defect density of the grown films. The density of voids in the substrate as a function of the growth temperature has also been studied.

II.11. Physical models of SiC

(C Persson, A Konstantinov, U Lindefelt, E Janzén)

The development of semiconductor devices benefits strongly from computer simulation of device properties. Such simulation, however, requires knowledge of fundamental transport properties (e.g., electron- and hole mobilities as functions of temperature, doping etc., and impact ionization properties) of the material from which the device is made. One of the objectives of this project is to provide, not only quantitative information on material (SiC) characteristics associated with transport properties of importance for simulation, but also an understanding in terms of fundamental interaction processes.

To achieve this, the theoretical part of the project will focus on band structure calculations (based on the Schrödinger equation within the local density approximation) and calculations of transport properties (based on Boltzmann's transport equation) using the Monte Carlo method.The SiC-polytypes of main interest are the 3C, 2H, 4H, and 6H polytypes, containing 2, 4, 8, and 12 atoms per unit cell, respectively. The large number of atoms per unit cell for the 4H and 6H polytypes puts very demanding requirements on the numerical method used to solve the band structure problem within the self-consistent local density approximation. Within this project we benefit from a collaboration with Carl-Olof Almbladh, Dept. of Theoretical Physics, University of Lund, who has developed a band structure program based on the density functional LAPW (Linearized Augmentet Plane Wave) method. Fully self-consistent potentials and band structures have so far been calculated for the 3C, 2H, and 4H polytypes. From these results, important band structure parameters necessary for calculating transport parameters can be deduced.

Through good contacts with Dept. of Integrated Systems, ETH, Zürich, Switzerland, we now also have access to a Monte Carlo program for performing transport calculations (i.e., solving Boltzmann's transport equation). Some work on modifying this program to suit the SiC crystal structure has been performed.

The major part of an extension of the deformation potential acoustic phonon scattering model, to suit the SiC crystal structure and based on group theoretical considerations, has also been performed.

III. SILICON, POROUS SILICON AND SiGe STRUCTURES

III.1 Magnetic resonance for complex defects in silicon

(W M Chen, E Sörman, A Henry, M Singh, E Janzén and B Monemar)

The activities in this research area have continued during the year, in close cooperation with the group at Huygens Laboratory at the University of Leiden, Holland (A Frens, M T Bennebroek and J Schmidt).

We have completed the analysis of experimental results obtained on an S-Cu-related metastable complex defect in Si, by optical detection of magnetic resonance (ODMR) at X-band and K-band. Two photoluminescence (PL) emissions arising from the bound exciton (BE) recombination at the defect in two different configurations were monitored in the ODMR experiments. The spin-triplet nature of the lowest BE state for both BE's was confirmed. The symmetry of each configuration has been determined, to be monoclinic-I and triclinic, respectively. The unusually broad ODMR linewidth is argued to originate from unresolved hyperfine interaction with a copper atom involved in the defect, at which the primary bound particle (i.e. the hole) of the BE is highly localized. The configurational metastability has been demonstrated in the ODMR experiments.

We have carried out ODMR studies of a similar metastable selenium-related complex defect in silicon. This defect gives rise to two different deep photoluminescence spectra, originating from bound exciton recombination at two different configurations (referred to as A and B) of the same selenium-related centre. The lowest BE state in the A configuration is shown to be a spin triplet state, and the symmetry of the defect in this configuration is determined to be monoclinic I. In this symmetry one of the principal spin axis (z) is along the direction, while the other two principal spin axes (x and y) lie in the plane. The x-axis is shown to be 22deg. away from the [111] axis towards the [001] axis, and the y-axis is orthogonal to the x- and z-axis. The broad ODMR line width is attributed to an unresolved hyperfine interaction, most likely due to a copper atom in the centre. This is further supported by level anti-crossing effects observed in the experiments. The conversion between the two configurations, either by above bandgap light (SeA=>SeB) or by thermal annealing at temperatures > 60 K (SeB=>SeA), can be followed in PL and also in ODMR.

III.2. Direct determination of the electron-electron-hole (eeh) Auger threshold energy in silicon

(W M Chen, B Monemar and E Janzén)

This project is also part of a cooperation with the Leiden group (A Frens, M T Bennebroek and J Schmidt).

The electron-electron-hole (eeh) Auger threshold energy Ea is a fundamental property of the Auger recombination process, resulting from the requirement of both energy and momentum conservation and the band structure. In this work we have directly determined Ea in silicon, for the first time, by a novel experimental approach. The success of the new experimental approach is greatly attributed to our recent discovery of the excitonic Auger process as the dominant process inducing configurational changes of a metastable defect in silicon. For the particular defect studied, a S-Cu-related complex defect in silicon, the eeh Auger process was shown to govern the conversion of the defect from its stable configuration to the metastable configuration. By measuring the conversion rate as a function of the excitation photon energy, Ea can be determined directly. We have shown that the Ea value of 5 meV obtained for the defect-mediated eeh Auger process gives the upper limit for Ea in the case of the intrinsic eeh Auger process. This provides a direct experimental evidence that the eeh Auger threshold energy in silicon is indeed very small, and consequently this important Auger process must be dominated by the direct (phonon-less) mechanism. The results from this work are therefore expected to resolve the controversies on the subject, and lead to a better understanding of both the fundamental carrier recombination processes and the band structure in silicon.

III.3. Shallow excited states of deep luminescent centers in silicon studied by Fourier transform photoluminescence excitation spectroscopy

(M Singh, N T Son, W M Chen, B Monemar and E Janzén)

We employ the novel Fourier transform photoluminescence excitation spectroscopy (FTPLES) to investigate the shallow excited states of deep defects in Si. Excitation spectroscopy in the near infrared region has been hindered in the past by the lack of suitable tunable light sources which has meant that only a handful of studies in narrow spectral regions have been carried out. We show that FTPLES is a powerful technique for studying the excited states of deep defects. In Ag-doped silicon we observe the dominant emission band with zero-phonon lines (ZPL) around 780 meV. The FTPLE spectrum with a series of sharp lines in the 780-830 meV range can be attributed to 1s-ns transitions of a pseudo-donor electron. The FTPLE spectrum of the main PL band in Fe-doped Si with a ZPL at 735.1 meV, exhibits weak excited effective-mass-like states which indicates that the defect is hole-attractive with a bound electron in a delocalized effective mass orbit. We reveal the general features on the electronic structure of the excited states of deep BE systems in silicon doped with transition-metal impurities. These results elucidate the nature of excitons bound to deep isoelectronic impurities in semiconductors.

III.4. Efficient excitation transfer between deep defects in
silicon

(M Singh, W M Chen, N T Son and B Monemar)

Inter-defect excitation transfer is believed to be an important mechanism in semiconducting crystals. However, the role of excitation transfer processes in Si is not fully appreciated due to a general lack of experimental evidence. In this work we provide direct experimental evidence, by using a novel approach of the recently developed FTPLE spectroscopy, that one defect can be excited efficiently by a transfer process from an electronic excitation of another defect in Si. One such example is a deep iron-related isoelectronic defect in Si, which gives rise to a deep photoluminescence emission with a no-phonon line at 735 meV. In the FTPLE spectrum we observe the shallow electronic excited states in the 735-780 meV region, which provides evidence on the pseudo-donor nature of the defect. The superior resolution, sensitivity and the large spectral ranges that can be covered in Fourier spectroscopy allow us to obtain the entire excitation spectrum of the defect up to the bandgap. In addition to an efficient excitation at the fundamental band edge, two strong peaks are present in the excitation spectrum at 976.3 meV and 1115.6 meV. These two excitation processes are believed to originate from electronic excitation of other defects, since the energy of these excitations is higher than the ionization energy of the 735 meV defect. This clearly shows that the 735 meV defect can be efficiently excited through an excitation transfer process via these defects. From the width of the 976.3 meV excitation line, we estimate its lifetime, governed by the transfer process, to be less than 1 ps - a remarkably short lifetime for defect states in Si. Since the excitation transfer dominates, no radiative recombination of the electronic excitations at these defects could be observed experimentally.

III.5. Theoretical identification of new defects in P- doped electron-irradiated Czochralski silicon

(A B Van Oosten, W M Chen, J L Lindström and B Monemar)

Irradiation defects in oxygen and phosphorus rich silicon are identified by comparing ab initio theoretical predictions for a variety of possible complexes with experimental results. It is shown that infrared absorption and optically detected magnetic resonance spectra detect two different complexes with similar formation kinetics and cannot be accounted for by a single complex. Evidence for the existence of a (V-O-P) complex with a neutral as well as a negative charge state is presented and the complex is identified as the center responsible for the IR absorption bands. The ODMR spectrum is accounted for by the positively charged (V-P2) complex. It is shown that the negatively charged (V-P2) should also exist and can account for the NL1 electron paramagnetic resonance spectrum. (V-O-P) and (V-P2 ) are proposed to be formed through capture of the (V-P) complex (E center) by interstitial O and substitutional P, respectively, which explains why both complexes are formed simultaneously. We also confirm the identification of the 889 cm^-1 IR band with the (V-O2) complex, which is predominantly formed in the absence of P.

III.5. Porous Silicon

(C I Harris, M Syväjärvi, O Kordina, P Bergman, B Monemar)

The field of research surrounding porous silicon has expanded rapidly since the first demonstration of bright room temperature luminescence some three years ago. The small program at Linköping has focussed in the last year on two specific areas which are of interest both to the understanding of the luminescence mechanism itself and from the viewpoint of device application. ODMR has been applied to investigate the recombination in the visible band and also the less studied near infra-red emission. The technologically interesting blue emission from porous silicon has also been examined in detail using both time integrated and time resolved photoluminescence techniques. The blue emission is observed in as-etched samples prepared using a standard electrochemical method. The porous material giving the blue emission is unstable and degrades rapidly under UV photo-excitation. A corresponding increase in red intensity with decreasing blue intensity is observed, demonstrating a correlation between the two types of emission. Under ambient light conditions at room temperature the material is more stable and remains emissive for a number of days. The radiative decay timefor the blue emission is found to be extremely fast ([[tau]]=0.86 ns), and remarkably remains unchanged from room temperature down to 77K. This independence of decay time on temperature suggests a molecular origin for the luminescence rather than the more widely accepted quantum size picture.

III.6 Processinduced defects in silicon

III.6.1 Oxygen microclusters in silicon

(T Hallberg, L Lindström)

Clustering of oxygen is a well-known but not well understood phenomenon in heat treated Czochralski silicon. We have studied oxygen clustering in the temperature range 350-650 ^oC where oxygen related thermal donors are formed causing resistivity changes. We have developed a new method to study the clustering process which can be an important tool to better understand the huge amount of data published since the 1950:s. Our results show that oxygen atoms are clustering in different structures which can be correlated with different electrically active centers. The thermal donor formation process is found to be governed both by ordinary oxygen diffusion and by structural transformation processes of pre-existing clusters.

III.6.2 Characterization of silicon exposed to reactive ion etching (RIE)

(L Lindström, A Henry, G S Oehrlein, B Monemar)

Photoluminescence and thermal wave modulated reflectance investigations have been performed on silicon wafers exposed to different plasmas in a reactor equipped with an electron cyclotron resonance (ECR) source. The work has been performed in collaboration with G S Oehrlein at SUNY Albany, USA. Various etch conditions have been investigated and compared to results from a standard type of RIE reactor.

III.6.3 Other defects in silicon

(T Hallberg, L Lindström, B G Svensson, A Henry, B Monemar)

An effort to characterize a group of defects in n-type silicon (P-, As-,
Sb-, and Bi-doped) induced by electron irradiation has continued using FTIR spectroscopy and DLTS. In addition the effect of hydrogenation of Hg-implanted silicon has been studied and found to give rise to new photoluminescence lines.

III. 7. Characterisation of Si-SiGe structures

(A Henry, B Monemar, W X Ni and G V Hansson)

The new solid-source molecular beam epitaxy (MBE) system has produced various Si-Ge structures (single epilayers, single quantum wells, multi quantum wells or superlattices) which were investigated by the photoluminescence (PL) technique. Until now no luminescence was detected from the superlattice structures. However epilayers and quantum wells structures show good PL properties.

Epilayers. The PL spectra of the SiGe alloy epilayers grown on silicon substrates are characteristic of near bandedge excitons bound to shallow impurities such as phosphorus or boron. Our PL data showing the energy position of the PL lines as the function of the Ge concentration (in the range 8 to 17%), are comparable to some previous results from the literature. These PL lines have a full width at half maximum (FWHM) relatively narrow of about 2.7 meV. However inhomogeneity in the Ge concentration for some of the samples was revealed by the observation of the broadening of the PL lines or net superposition of different lines related to different Ge concentrations. The SiGe luminescence is observed at low temperature and is rapidly thermally quenched when the experimental temperature is increased. When the boron concentration introduced during the growth is increased to around 3.7x10¹⁸ cm^-3 a weak luminescence feature can indicate a band-gap narrowing of about 54 meV with a filling energy of 50 meV.

Single quantum wells. From single quantum wells (SQW) very intense and narrow (FWHM ~ 2.7 meV) PL lines can be observed indicating the good quality of the structures. The PL is detected up to high temperatures, comparatively to the case of alloy epilayers, such as 77K, depending of the Ge concentration and well width. However no structure of the PL lines could be observed as expected if bound exciton and free exciton PL lines could be resolved. The excitation power dependence of the luminescence at low temperature shows a broadening of the line width together with a slight shift to the high energies. These both behaviours (temperature and excitation power dependence) are characteristics of localised excitons in SQW where the potential fluctuations are relatively smooth. To check the homogeneity of the doping in the case of a Sb [[partialdiff]]-doped SQW, a large wafer (3") was used as the substrate and that without rotation during the growth. A variation of the energy position of the PL was found to be only 3 meV, indicating a small variation in the doping concentration. A Comparative study between an undoped SQW and a modulated structure shows as expected a decrease in intensity of the PL together with a shift to higher energy.

Multiple quantum wells. Multiple quantum wells (MQW) were also investigated. However their PL intensity is less efficient than in the case of the SQW, showing the importance of non-radiative channels existing in the Si barriers. This behaviour is confirmed by using excitation below the Si band gap during the PL experiments instead of the usual visible excitation.

IV. ELECTRONIC STRUCTURE AND SURFACE STRUC-TURE STUDIES OF METALS AND COMPOUNDS

(L I Johansson and H Johansson).

Experimental studies aimed to increase the understanding of the surface properties of selected metals and metallic compounds represent our main research efforts. Synchrotron radiation at the MAX-laboratory in Lund has been utilized to a large extent but also inhouse photoemission and inverse photoemission equipment. Instrument development work on an inhouse LEED I-V instrument and on a spherical grating monochromator for beamline 33 at the MAX laboratory have been continued during this year. Below short summaries of different projects are given.

IV 1. Transition metal carbides and nitrides

High resolution core level photoemission studies of some surfaces of carbides and nitrides have been continued for the purpose of making a systematic study of the occurence of surface shifts in the metal and nonmetal levels of these materials. Results on TiC(100), TiN(100) and WC(0001) have been published during this year while the data collected on ZrN(100), NbC(100) and NbN(100) not yet have been summarized in reports. The investigation of WC(0001) revealed presence of surface shifts in both the metal and nonmetal level for the first time for these materials. In the earlier studies presence of surface shifts in the nonmetal levels only have been possible to identify unambigously. The plan is to utilize these findings in future studies of selective surface reactions.

These core level results have together with earlier results from studies of the band structure and the surface structure of these materials been summarized in a review for Surface Science Reports.

IV2. Transition metal silicides

High resolution core level photoemission investigations of the (001) and (110) surfaces of MoSi2 and of WSi2(110) have been reported during this year. Surface shifted Si 2p components were identified on all surfaces, indicating Si termination. A surface shifted W 4f level could also be identified on WSi2 while no shifted Mo 4p level could be identified on MoSi2. Initial oxidation studies revealed a fairly rapid Si oxidation for both silicides. No Mo oxidation could not be detected on the MoSi2 surfaces but a weak W oxide signal was revealed on WSi2 at the large exposures. Investigations of FeSi2 silicide surfaces have been started but not completed.

IV 3. Beryllium metal

Be metal appears to be unique since we could identify three surface shifted Be 1s components on the close-packed Be(0001) surface. These components were interpreted to originate from the three outermost layers, two subsurface Be 1s components were thus identified and the first layer showed the largest surface shift observed so far on any close-packed metal surface. These experimental results initiated a theoretical study (Aldén et al PRL 71, 2457 (1993)) which predicted negative shifts for the three outermost atomic layers with magnitudes close to the observed values. The calculated results showed that what makes Be anomalous compared to other metals is the absence of a p-core, against which the screening orbitals would have to orthogonalize.

A photoemission study of the Be(1010) surface gave a further example of the anomalous surface related core level shifts of Be metal. Three surface shifted Be 1s components were unambigously identified also on this surface. From considerations of extracted surface to bulk intensity ratios these components were assigned to emission from the first, second and third plus fourth atomic layer. Surprisingly, the surface shifts of -700, -500 and -220 meV determined for this surface were found to be smaller in magnitude than the shifts of -825, -570 and -265 meV previously determined for the close-packed Be(0001) surface. This finding of smaller surface core level shifts on a more open surface was most unusual, since the close-packed surface have shown the smallest shift for all other metals.

The electron mean free path in Be metal was determined from the surface to bulk intensity ratios extracted from these two surfaces. The values determined were found to show a good overall agreement and the mean free path was found to exhibit a pronounced minimum at an electron kinetic energy around 25 eV. The experimental results were compared with the mean free path calculated using an approach originally applied by Quinn. The experimental and theoretical values agreed well at higher energies but at lower energies the experimental values were found to drop significantly below the theoretical values and locate the minimum at a markedly smaller kinetic energy.

IV 4. Instrument development

Development work on the LEED I-V instrument, equipped with a video camera and computer system, for surface crystallography studies has been completed during this year.

The spherical grating monochromator designed for beamline 33 at the MAX laboratory has been built and delivered and is presently under installation. Much work remains to be done, however, before the monochromator is characterized and operational.

IFM Activity Report 1994: Materials Science

General Information

Scientific Areas

Education

Materials Science

III.5. Theoretical identification of new defects in P- doped electron-irradiated Czochralski silicon