Photoreflectance spectroscopy of Si-doped InGaAsN thin films grown by molecular beam epitaxy

Several research groups have reported that the incorporation of a small fraction of N into InGaAs thin films on GaAs substrates has a strong effect on the optical and electronic properties due to the interaction between the conduction band and an impurity level introduced by the N atoms. In particular, the N incorporation causes a dramatic narrowing of the band gap. (Strain arising from the film/substrate lattice mismatch may also affect the properties of the film.) This phenomenon has made InGaAsN a material of interest for infrared optoelectronic devices such as 1.3µm lasers and 1.45µm light emitting diodes.

Photoreflectance (PR), a differential method that probes the modulation of the refractive index of the sample under laser photoexcitation, caused by the modulation of physical parameters such as the built-in surface electric field, is a widely used tool for studying the electronic structure of semiconductors. PR has several practical advantages: the technique is contactless, nondestructive, and relatively simple and low-cost to set up.

We present PR spectra for a series of Si-doped n-type GaAsN and InGaAsN alloys (with mole fraction, m(N), of N = 0.08% to 1.8%) grown on GaAs substrates by molecular beam epitaxy. Spectra were measured within the range 0.95 eV to 2.2 eV and analyzed by functional curve-fitting based on the Aspnes third derivative form. The temperature dependence of the PR, at temperatures from 25 K to 300 K, was examined for two samples, a lower-N sample (m(N)=0.2 %) and a higher-N (m(N)=1.8 %) sample in the InGaAsN series. Preliminary conclusions from the spectral measurements and curve-fitting results are as follows. (As an example, the spectrum of an InGaAsN film with m(N)=0.9 % is shown in Fig. 1. The fitted transitions are labeled by transition character, as discussed below, and photon energy.)

(1) At room temperature, all the samples show at least the following two transitions (which also occur in pure GaAs): the fundamental band gap or "E-" transition and the spin-orbit split peak or "E- +  {Delta}" transition. The higher N samples, the m(N)=0.34 % sample in the GaAsN series and the m(N)>0.9 % samples, in the InGaAsN series, also show an "E+" peak, which is ascribed to the splitting of the lowest conduction band by the interaction with the nitrogen impurity levels. The "E-" and "E- +  {Delta}" transitions shift down in energy with increasing N content, while the "E+" transition shifts up.

(2) All the transitions shift up in energy on going from lower Si (1x 10 17 cm-3) to higher Si (2 x 1017 cm-3 ) concentration. Also, the PR linewidths are larger in the higher Si than in the lower Si samples. The Si doping effect is likely caused by the dopant-induced carriers (electrons). One possible mechanism is the "carrier band-filling" or "Burstein-Moss" effect. The apparent strength of the doping effect, at relatively low Si concentration, is surprising; much smaller energy shifts have been observed in pure GaAs for similar dopant and carrier concentrations.

(3) Several samples show two near-band-edge peaks with a small energy splitting, with the lower peak having a n=2 critical point and the higher peak having a n=2.5 critical point. For example, the m(N)=0.08 % GaAsN sample shows transitions at 1.391 eV and 1.398 eV. This is consistent with a model that assumes both excitonic and free-to-free transitions occur at the fundamental band gap. According to this model, in the m(N)=0.08 % GaAsN sample the exciton binding energy is approximately 7 meV.

(4) Several samples show one or two additional peaks at higher energy than the fundamental band gap (i.e., higher than either the excitonic or free-to-free transition), but lower energy than the spin-orbit-split peak. These additional peaks may also arise from the dopant-induced carriers; the specific mechanism is being investigated.

(5) The temperature dependence of the electronic transition energies can be described with reasonable accuracy by either the Bose-Einstein equation or the Varshni equation. The fits of these two equations to the data are compared.