3. III-V compound semiconductor material systems

This chapter offers a short overview of the employed material systems and of the epitaxy techniques used for the layer structure growth. Furthermore, the Gunn diode layer structures considered in this dissertation are described. In one particular case, the resonant tunneling injector Gunn diode, simulation results support the design choice.

3.1 Group-III/As and group-III/N material systems

Figure 3.1: Bandgaps of the most important elemental and binary cubic semiconductors versus their lattice constant at 300$^\circ$K. The right-hand scale gives the light wavelength $\lambda$, corresponding to the band gap energy [IL03,Riz01].

3.1.1 GaAs, AlAs

Since the early days of semiconductor technology, the investigation on new phenomena in solid state physics lead to new devices and ICs. A number of different materials have been studied. Among all the new materials, the one that has received most attention is gallium arsenide, together with some of its closely related binary, ternary and quaternary compounds. After silicon, gallium arsenide is now regarded as the second most important electronic material.

Goldschmidt was the first who created GaAs in 1929. He found that it has a zincblende lattice with a FCC symmetry [Gol29]. Only in 1952, in fact, GaAs has been identified as a semiconductor by Welker (Siemens)^3.1. The nature of the bond between gallium and arsine is predominantly covalent. The first device exploiting the direct band-gap of GaAs dates from 1962, when Hall et al. at GE [HFKC62] and Redhiker at al. at MIT Lincoln Laboratory [QRK⁺62], independently, obtained the first semiconductor laser. In the same year, Gunn (IBM) discovered the transferred electron effect and developed the first solid state microwave oscillator [Gun63].

Figure 3.2: Bandgap energies of $Al_xGa_{1-x}As$ in different points of the Brillouin zone plotted versus the Al concentration [Kel96,Mal02]

Figure 3.1 shows the energy gaps and the lattice constants of of the most important elemental and binary cubic semiconductors. The connecting lines consider the case of ternary compounds, composed of various ratios of the corresponding binary materials. The compound $Al_xGa_{1-x}As$ maintains nearly the same lattice constant with the change of the Al mole fraction. This property and the related high quality heteroepitaxy have opened new possibilities for advanced devices like double heterostructure lasers, high electron mobility transistors and heterostructure bipolar transistors; to describe the large possibilities offered by the epitaxial growth using $Al_xGa_{1-x}As$, and in general III/V compounds, a new expression has been coined: bandgap engineering.

In Fig. 3.2, the bandgap energy of the $Al_xGa_{1-x}As$ material system is plotted versus the Al concentration in different points of the Brillouin zone. At about 45% Al, the transition between direct and indirect bandgap can be observed.

3.1.2 GaN, AlN

Figure 3.3: Bandgaps of III-N semiconductors with wurtzite ( $\alpha - phase$) and zincoblende ( $\beta - phase$) structure versus their lattice parameter at 300$^\circ$K. The right-hand scale gives the light wavelength $\lambda$, corresponding to the band gap energy [Riz01].

As in the case of arsenides, between group III - elements and nitrogen, the nature of the bond is mainly covalent. However, for the nitrides, the large difference in the electronegativities causes a strong ionic component, which means very high bonding energies (AlN 11.5, GaN 8.9 and InN $7.7 \thickspace eV/atom$) and consequently excellent thermal and chemical stability. In contrast to GaAs, the thermodynamically stable phase of these materials is the hexagonal wurtzite structure, $\alpha - phase$. Beside to $\alpha - phase$, a metastable $\beta - phase$ with zincblende structure exists if very thin layers of GaN and InN are grown on cubic substrates like GaAs or silicon. The nitride materials with wurtzite structure form an alloy system (InGaN, AlGaN, InAlN), whose direct bandgaps range from 0.7 to $6.2 \thickspace eV$ (Fig. 3.3). These give to group III-nitride unique optical properties, making them suitable for a large spectrum of optoelectronic applications. Another peculiarity of group III nitrides, in comparison with arsenides, is the existence of strong spontaneous and piezoelectric polarization fields. This property leads to an additional carrier accumulation at the strained interfaces $Al_xGa_{1-x}N/GaN$ in 2DEG structures, enhancing the electron concentrations in GaN HEMTs.

3.2 GaAs, AlAs grown by MBE

Molecular beam epitaxy (MBE) is a physical deposition method for growing thin epitaxial structures made of semiconductors, metals or insulators. The word epitaxy stems from the Greek words $\epsilon \pi \iota$ (epi=on) and $\tau \alpha \xi \iota \zeta$ (taxis=order). The etymology of the term already suggests the growth of thin films in an ordered way on a substrate or on a previously deposited layer. Solid source materials contained in effusion cells evaporate after the electrical heating. Under Ultra High Vacuum (UHV) conditions, a molecular (or atomic) beam builds from the evaporated materials. The epitaxial film grows from the molecular beams impinging on the substrate. The beams from different sources (cells), which have a conic distribution directed to the heated substrate, can be independently controlled by means of shutters. In the transition from the gas-phase to the solid-phase, the reciprocal interaction between the molecules causes the crystal formation and the growth takes place. Under optimal conditions, the growth rates, which are typically in the order of 1 monolayer per second (ML/s), are low enough to allow surface migration of the impinging species and smooth surfaces. The homogeneity of the deposited layers along the wafer can be improved adjusting the growth rate and the substrate holder rotation.

Figure 3.4: Schematic view of a MBE system from [IL03]. (Top cross-view)

The GaAs Gunn diode structures considered in this thesis have been grown by molecular-beam epitaxy (MBE) on 2-inch semi-insulating GaAs substrates in a Varian ModGen II MBE system. The principle scheme of the MBE growth chamber is shown in Fig. 3.4. The substrate holder can be heated up to 900 $\thinspace$  $^\circ$C . The source materials Ga, As, In, Al, Si (for n-doping) and Be (for p-doping) are placed in the effusion cells. The growth rate can be measured by a RHEED^3.2 system and is controlled by tuning the cells temperature. In order to monitor the doping levels, calibration samples are periodically grown and characterized by Hall and CV measurements (sections 4.3 and 4.4).

3.3 GaAs Gunn diode structures

3.3.1 Graded Gap injector GaAs Gunn diodes

In this section, the layer structures of the GaAs Gunn diodes with graded gap injectors (GGI) are presented. The starting point for the chosen layer-system are the structures suggested by Hutchinson at al. [HSCK96] and by Couch et al. [CBK⁺88]. The optimization work, performed at the institute by Stock [Sto03], provided further helpful indications for the layer-system chioice.

Figure 3.5: Typical layer sequence of the GaAs Gunn diode with a graded gap injector.

A typical layer sequence of a GaAs Gunn diode with a graded gap injector is sketched in Fig. 3.5. It consists mainly of an undoped AlGaAs graded barrier structure followed by a $\delta -$doping and a thick low doped GaAs active region. The grading is linear, starting from 1.7% up to the maximum Al value. The role of the two $5\hspace{1mm} nm$ GaAs spacers is to avoid doping diffusion in the graded barrier.

In this work, different structures have been considered:

• One with a $1.6\hspace{1mm} \mu m$ active region without spacers. The absence of spacers makes the graded gap barrier not effective [Sto03](W16016).
• Two with $1.5 \hspace{1mm} \mu m$ and $1.8 \hspace{1mm} \mu m$ active regions and with a $50 \hspace{1mm} nm$ long, 32% Al graded barrier each (W18006 and W18021 respectively).
• Four with $1.6\hspace{1mm} \mu m$ active regions and with $65\hspace{1mm} nm$ long, 28%, 30%, 32% and 34%, respectively, Al graded barriers (W18041, W18040, W18038, W18039).

The first structure (W16016) is mainly used as a reference. Even if the electrical measurements show no hint of a hot electron injector, W16016 allows a comparison with the other structures. Wafers W18006 and W18021, which have the same full-working graded gap injector, were grown to demonstrate the influence of the active region length on the diode high frequency behaviour. Finally, in wafers from W18038 to W18041, the role of the maximum Al content in the injector has been examined.

3.3.2 Resonant tunneling injector GaAs Gunn diodes

The basic principles of the resonant tunneling injector (RTI) have been already discussed in section 2.2.3. The design of the AlAs double barrier RTI started from numerical simulations. The RTI is a completely new structure and no reference could be found in the literature.

Figure 3.6: Conduction band profile and local density of states plot for a Resonant Tunnelling Injector

The simulation of the resonant tunneling injector was done with the software package Wingreen [IM]. This is based on a self-consistent real-time Green's function approach [Ind99]. A simulation example is shown in Fig. 3.6. The conduction band and the local density of states of a RTI are presented for an applied bias voltage. The chosen bias voltage corresponds to a current density in the range 23-27 $kA/cm^2$. The layer structure has to be so that the following condition is satisfied: the first transmission energy level for the given current density range has to match the energy difference between the L- and the $\Gamma$-valley ( $\Delta E \thickspace \approx \thickspace 310 \thickspace meV$ for GaAs). In order to make the RTI competitive, a further specification has been defined: the voltage drop on the RTI for the working current conditions, must be much lower than the one on the GGI.

Figure 3.7: Typical layer sequence of the GaAs Gunn diode with a resonant tunneling injector.

In Fig. 3.7, a typical layer sequence of the RTI GaAs Gunn diode is presented. The structure is very similar to that of GGI GaAs Gunn diode: the active region, the spacers and the $\delta$-doping layers have not been changed. The injector itself is undoped and consists in a sequence of AlAs/GaAs/AsAs ( $6\thickspace ML %\footnote{ML: abbreviation for monolayer.}$/ $17\thickspace ML$/ $6\thickspace ML$, W18069). A second wafer has been grown decreasing the AlAs thickness from 6 to $5\thickspace ML$ (W18069). The High Resolution Transmission Electron Microscopy (HRTEM) image illustrates a resonant tunneling double barrier with $6\thickspace ML$ of AlAs. The sample has been grown at $580 \thickspace ^\circ$C, an optimal temperature for well defined GaAs/AlAs interfaces [Lan99].

Figure 3.8: HRTEM image of a resonant tunneling double barrier AlAs/GaAs/AlAs ( $6\thickspace ML$/ $17\thickspace ML$/ $6\thickspace ML$). It can be noticed that the interfaces are well-defined and the crystal quality is very good [Lan99].

3.4 GaN grown by MOVPE

In this section, the growth of GaN by Metal Organic Vapor Phase Epitaxy (MOVPE) is shortly explained. To avoid confusion, it should be noted that the epitaxial technique MOVPE is also known by the acronyms MOCVD (metalorganic chemical vapor deposition), OMVPE and OMCVD [Mor99].

The MOVPE is characterized by large-area growth capability, high surface mobility of the precursor gaseous molecules, good layer uniformity and precise control of the epitaxial deposition. These reasons together with the higher growth rate, made the MOVPE to be the favourite growth method for industrial purposes. By this technique, the gas phase growth precursors are transported by a carrier gas to a heated substrate, where the precursors are pyrolysed and the nitride film is deposited. The diffusion of the active materials to the substrate are favored by the depletion at the surface and the consequent concentration gradient of these materials in the gas phase, due to their consumption during the growth. The obtained byproducts are then transported out from the reactor together with the unused reactants. As group III sources, trimethylgallium or triethylgallium (-indium,-alluminium) are used, whereas the common nitrogen source is ammonia ($NH_3$). The high thermal stability of $NH_3$, although still low compared to $N_2$, is one reason to use high substrate temperatures, typically more than 550$^\circ$C for InN and above 900$^\circ$C for GaN and AlN. The high growth temperature and thus the high nitrogen vapor pressure lead to the problem of nitrogen loss from the nitride film and to carbon contamination from the decomposition of the organic radical during metalorganic pyrolysis. The loss of nitrogen is usually alleviated by the use of high V/III gas ratios during the deposition. The extreme requirements for the nitride growth have led also to the development of new MOVPE reactor designs.

The GaN Gunn diode structures, considered in this thesis, have been grown by MOVPE in an Aixtron AIX200 reactor on two inch $Al_2 O_3$ substrates. Unlike the MBE system, RHEED is not suitable for in-situ monitoring of the growth rate in the high-pressure environment like MOVPE. RHEED requires a high electron mean free path, which can be achieved only in ultra high vacuum conditions. The growth rate is therefore determined by normalized reflectometry. A more accurate description of the MOVPE epitaxy and the experimental setup can be found in [Kal03].

3.5 GaN Gunn diode structures

During this work, four GaN wafers have been considered: G102, G394, G582, G695. The first 3 wafers are uniformly doped; they have been used to optimize the ECR-RIE process and to produce GaN nanocolumns; the GaN layers are about 4, 2.2, and $4 \thickspace \mu m$ thick, respectively.

Figure 3.9: Layer sequence of the GaN Gunn diode.

The layer sequence of a GaN Gunn diode is sketched in Fig. 3.9. It is a typical Gunn diode without injector and $\delta$-doping. It consists of a high doped $800\thickspace nm$ thick contact layer, followed by a $1600\thickspace nm$ low doped active region and by another high doped $200\thickspace nm$ thick contact layer. The typical surface morphology for the grown GaN Gunn wafer is shown in Fig. 3.10. The surface roughness for the $2 \times 2 \mu m^2$ AFM scanned area is about $0.3\thickspace nm$. This is a good result for GaN.

Figure 3.10: AFM image of the surface morphology for GaN Gunn structure (wafer G695).