• MBE-GROWTH OF STRAIN ENGINEERED GAN THIN FILMS UTILIZING A SURFACTANT
• ABSTRACT
• INTRODUCTION
• EXPERIMENTS
• RESULTS AND DISCUSSION
• CONCLUSIONS
• ACKNOWLEDGMENTS
• REFERENCES

MBE-GROWTH OF STRAIN ENGINEERED GAN THIN FILMS UTILIZING A SURFACTANT

ABSTRACT

GaN films were grown on sapphire substrates at temperatures below 725 ^oC utilizing a Constricted Glow Discharge plasma source. A three dimensional growth mode is observed at such low growth temperatures resulting in films that are composed of individual but oriented grains. The strain that originates from the growth on the lattice mismatched substrate with a different thermal expansion coefficient is utilized to influence the thin film growth. The strain can be largely altered by the growth of suitable buffer layers. Thereby, optical and structural film properties can be engineered. It is argued that the surface diffusion of Ga ad-atoms is affected by engineering the strain. Alternatively, surface diffusion can be influenced by surfactants. It is demonstrated that the use of bismuth as a surfactant allows to modify the surface morphology of the GaN films that reflects the size of the grains in the films. The results suggest that a substantial increase of the oriented grain sizes in the films is possible while maintaining a low growth temperature.

INTRODUCTION

Group-III nitrides are the most promising materials for optoelectronic light sources [1, 2] that can emit light from the ultraviolet to the visible spectral range and as well for high frequency power devices [3]. Usually, GaN films are grown at temperatures that are low compared with its melting point (T_growth < 0.5 T_m). In contrast to MOCVD, a MBE growth process can exploit deviations from thermodynamic equilibrium to a larger extend. This may help to increase doping levels or to grow high quality AlN/InN/GaN quantum well structures. However, due to the lower growth temperature (typical: MBE growth at 1000 K; MOCVD growth at 1300 K) the films grown by MBE usually exhibit a three dimensional growth mode with grain sizes that are limited by the growth temperature and by strain. This determines the structural quality of the thin films [4].

Recent progress in understanding GaN thin film growth revealed that strain is one of the key issues that determines the growth and physical properties of the thin films [4-8]. Strain can largely be altered by the growth of a buffer layer [6]. In fact, the active utilization of the buffer- layer growth-temperature [7], of its thickness and of the III/V flux ratio [8] open the possibility to strain engineer desired film properties such as the surface morphology, the size of the grains in the films [4], or the optical film properties [9]. It was argued that the surface diffusion of the Ga ad-atoms and the stoichiomety of the GaN films are affected by strain [4, 6]. Nevertheless, GaN thin film growth at MBE growth temperatures remained three dimensional and there are several options that can be explored to improve the structural film quality by enhancing the surface diffusion length of the Ga ad-atoms during growth and, thereby, the size of the grains in the films.

Obviously, the MBE growth temperatures can be increased. This is possible because nitrogen sources with large fluxes are now available such as our Constricted Glow Discharge (CGD) plasma source (demonstrated growth rate: 1.4 µm/h) [10]. Using several of such sources allows to overcome the film decomposition rate even at growth temperatures above 800 ^oC. However, this approach puts limits on the exploration of the low temperature growth. It is for this reason that we choose to perform experiments with surfactants in order to enhance surface diffusion while maintaining a growth temperature of 725 ^oC.

This contribution focuses on the utilization of bismuth (Bi) as a surfactant for the growth of GaN by MBE. Bi seems to be a promising choice for at least two reasons. First, Bi is isoelectronic with N but of extremely different size (covalent radii: N=0.07 nm, Bi=0.15 nm). Thus, it is unlikely that large amounts of Bi will be incorporated into the growing film. On the surface, however, Bi will substitute for N and, thereby, alter the bonding to the diffusing Ga ad-atoms. Second, GaN was grown from Bi solutions.

EXPERIMENTS

GaN layers are grown using a rebuilt Riber 1000 MBE system. Knudsen cells are used to evaporate pure Ga (99.9999%), Mg (99.99%), and Bi (99.9999%) while the activated nitrogen is produced by the CGD plasma source with pure nitrogen gas (99.9999%) along with a Millipore nitrogen purifier. Some details of the source design are given elsewhere [10]. A dc voltage generates a glow discharge that is constricted to an area in the plasma chamber close to the gas exit. It is the pressure difference between the plasma chamber and the MBE growth chamber that extracts the activated nitrogen species with energies less than 5 eV. A liquid nitrogen cryoshroud is used during growth to obtain a base pressure in the chamber of ~ 2 x 10^-10 Torr. A thin titanium (Ti) layer on the back of the 10 x 11 mm² c-plane sapphire substrate absorbs the heat radiated from the tungsten (W) filament heater. The temperature of the substrate is monitored with a pyrometer.

The substrates are degreased by boiling in acetone and ethyl alcohol for 5 minutes each and blown dry with nitrogen. After degassing in the load lock for 30 min at 500 ^oC, they are transferred into the growth chamber. The substrates are then heated up to 700 ^oC for thermal desorption of surface contaminants. At this temperature, they are exposed to activated nitrogen for 10 minutes. Subsequently, a thin low temperature GaN buffer layer (~ 250 ) is deposited on the substrate. Its particular thickness was determined and optimized by intentionally engineering the strain [7, 8]. Finally, the main epitaxial layer is grown on the buffer layer during 4 hours. Typical grown conditions are: Ga source temperature: 1210 K; nitrogen flow rate: 5 - 80 sccm; buffer-layer growth-temperature: 773 K; main-layer growth-temperature: 1000 K. The Bi source temperature was varied in the range of 250 to 550 ^oC; some among these samples were in addition Mg doped. During the growth, the nitrogen partial-pressure in the chamber is the range of 10^-5 to 10^-2 Torr.

RESULTS AND DISCUSSION

In figure 1 we present atomic-force-microscopy (AFM) images of nominally undoped samples grown without Bi and with different amounts of Bi (T_Bi = 350 to 550 ^oC); each depicted area is 2 x 2 µm². We assume that the observed features on the surface reflect the size of the grains in the films. The sample in Fig. 1a) that is grown without any bismuth exhibits small grains with irregular boundaries. From previous investigations it is known that these grains can be disconnected [4]. Therefore, a drastically reduced lateral Hall mobility is observed. A small amount of Bi makes the grains coalesce (Fig. 1 b) and the Hall mobility increases from 6 to 73 cm²/Vs (cf. Table 1). The background n-type carrier concentration was unchanged and exceeded 10¹⁸ cm^-3. This rather high n-type carrier concentration is probably caused by oxygen contamination of the growth chamber which was opened before these runs. A further increase of the Bi temperature to 450 ^oC (Fig. 1 c) and to 550 ^oC (Fig. 1 d), respectively, results in an enlargement of the grains. However, a poorer grain coalescence is obtained and the Hall mobility decreases.

Magnesium (Mg) doped GaN thin films exhibit grain sizes that compare well with those of the unintentionally doped n-type films of figure 1a). Unexpectedly, the size of the surface features increase drastically to about 10 µm if the Bi surfactant is used (figure 2; T_Bi = 350 ^oC and T_Mg = 280 ^oC). Similar effects are observed on films that are grown with different Bi source temperatures. In these particular runs the background impurity concentration is lower and the intrinsic n-doping did not exceed 10¹⁷ cm^-3. Thus, an influence of the impurity background concentration on the enhanced surface diffusion cannot be excluded.

Our previous experiments suggested that the feature sizes that can be observed on the films relate to the size of oriented grains which form the GaN thin films. This grain size limitation was attributed to a temperature and strain dependence of the Ga surface diffusion coefficient [4]. Following these arguments, the Bi surfactant seems to alter the Ga surface diffusion coefficient on the GaN (0001) faces, too. If we assume that surface diffusion only occurs during the growth of a double layer of Ga and N (thickness: 0.26 nm), a surface diffusion length x ~ (D_o)^-1/2 can be estimated where D is the surface diffusion coefficient and _o is the time that is required to grow the double layer. Since, both, the growth rate and the grain size can be extracted from the experiments, we can estimate Ga surface diffusion coefficients. They are shown in figure 3. Open circles are taken from reference 4 and they depict the temperature and stress dependence of the surface diffusion coefficient. In addition, we present the value obtained from the sample shown in the microphotography of Fig. 3 (solid circle). It is seen that the utilization of Bi as a surfactant leads to a further increase of the diffusion coefficient. We estimate a surface diffusion coefficient that we would have expected for a growth temperature of 985 ^oC even though the sample is grown at 725 ^oC. Thus, the use of a bismuth surfactant is beneficial in several respects. First, it allows to extend the growth of GaN thin films with a desired grain size to lower growth temperatures. This is important if deviations from a thermodynamic equilibrium should be exploited. Second the use of a surfactant provides an independent way of tuning the growth to values that result in a film with coalesced grains.

At present, details of the growth mechanism using surfactants are not well understood. We expect that the presence of impurities in the growth chamber must affect the experimental results. Also, a dependence of the feature size on the film thickness is subject of current investigations. Nevertheless, we give experimental evidence on an impact of Bi on the thin film growth if this semi-metal is used as a surfactant.

Fig. 1: Atomic Force Micrographs of MBE-grown GaN samples with different amount of Bi as surfactant: a) no Bi, b) TBi = 350 oC, c) TBi = 450 oC, and d) TBi = 550 oC, respectively. The size of the depicted areas is 2 x 2 m2.

Table 1: Carrier mobility of GaN samples grown with and without surfactant in dependence on the bismuth temperature.

TBi (oC)	-	350	450	550
µ (cm2/Vs)	6	73	10	3

Fig. 3:

Surface image of a MBE GaN sample grown with TBi = 350 oC and TMg = 280 oC taken by optical microscopy.

Fig. 4:

Surface diffusion coefficients of the sample grown with Bi (solid circle) in comparison with samples grown without any surfactant (open circles) in dependence on the reciprocal temperature.

CONCLUSIONS