Characterization of deep-level defects in GaNAs/GaAs heterostructures grown by APMOVPE

Fig. 2 presents low-temperature DLTS spectra for all the studied GaNAs/GaAs samples. In this range and for the selected measurement conditions, a broad and positive DLTS-line signal is observed. In our system, the positive signal indicates that the peaks are correlated with thermal emission of majority carriers from deep levels. It is generally accepted that the majority peaks (traps) are those for which the capacitance transients are increasing function of time, whereas minority peaks manifest themselves by decreasing capacitance transient [12]. According to this statement, in the case of DLTS, a positive signal corresponds to majority traps while a negative one to minority traps.

First of all, we should answer the question whether the holes or electrons are the majority carriers observed in the DLTS measurements of our samples. It is generally known that while the un-doped GaAs is usually n-type, a dilute GaNAs can be p-type or n-type, depending on the growth conditions and the type of elemental sources used in different growth techniques. The results presented in the literature indicate that as-grown, unintentionally doped GaNAs grown by MOVPE [13, 14] or MBE [15] is usually p-type, however for example a CBE growth can result in the n-type GaNAs material [16]. It has been speculated that this p-type background doping is a result of unintentional carbon contaminations [1, 17, 18]. Further confusing matter is the fact that annealing can strongly change both the carrier concentration and type in nitrogen diluted III-V materials. These changes are probably related to the hydrogen passivation of background carbon acceptors or even their dissociation from carbon upon annealing, but the existence of other types of defects that can contribute to the background carrier concentration cannot be excluded [1, 17, 19]. In the present studies, the TBHy is used as a source of nitrogen in the growth of dilute GaNAs and it could be also the additional source of carbon contaminations, causing p-type doping in the epitaxial layer. Nevertheless, although the undoped epitaxial layers were grown on the Si-doped n+-GaAs substrate, we observed only a majority carriers signal in DLTS spectra shown in Fig. 2, thus the layers are in fact slightly n-type. If a p-n junction existed, both the positive and negative signal (dependent on the bias conditions) would be observed in DLTS measurements. It should be noticed yet that after annealing of similar GaNAs/GaAs samples we actually observe the DLTS signal coming from both hole and electron traps [20]. It has been concluded that the existence of the acceptor-like defects in MOVPE-grown samples could be related to activation of carbon impurities or reduction of hydrogen impurities density after annealing. These results strongly suggest that post-growth annealing clearly influences the electrical properties of the GaNAs layers grown at low temperatures (566 °C in our case) by means of APMOVPE technique. Both the electron and hole traps were also observed in undoped GaNAs/GaAs triple quantum well structures after annealing [21]. Finally, in the present study we correlated the positive DLTS signal with the existence of only electron traps in the as-grown and undoped GaNAs/GaAs heterostructures grown by APMOVPE.

Furthermore, a broad DLTS-line suggests that deep levels, giving rise to this signal, can be actually a continuous distribution of deep level states or groups of closely spaced discrete energy levels within the band gap. The positive DLTS signal observed for all studied samples in the 100 K to 300 K temperature range is in fact a sum of at least three overlapping peaks originating from three different electron traps, labeled by us tentatively as E1, E2 and E3. In order to resolve the closely spaced defect levels, high-resolution Laplace DLTS (LDLTS) was applied. The LDLTS technique is a powerful tool for thermal emission studies, because of its significantly improved spectral resolution compared to a conventional DLTS technique [22]. In case of ideal conditions, Laplace DLTS can resolve defect levels with emission rates differing by a factor of 2. Exemplary Laplace DLTS spectra, shown in Fig. 3 for one of the samples (1.6 % of N), clearly demonstrate a high resolution of this technique, making possible to resolve all the peaks detected by conventional DLTS. In Fig. 3, one can also observe a distinct evolution of particular peak positions on increasing temperature of the measurement, i.e. the peaks E1 and E2 move towards higher emission rates, if the temperature increases. Moreover, when the temperature reaches 210 K, another peak E3 reveals, while peak E1 disappears from the LDLTS spectrum.

The activation energy (E_a) and apparent capture cross-section (σ_a) of the traps can be extracted from the slope and intersection of the Arrhenius plot of ln(e_n/T²) vs. 1000/T, respectively, where e_n is the carrier emission rate from deep level at temperature T. It follows from the detailed balance equation of the emission rate according to the formula [12]: en=σaνthNCexp⁡[−Ea/kBT]$${e_n}={\sigma _a}{\nu _{th}}{N_C}\exp [-E_a/{k_B}T]$$(1)

where ν_th is a thermal velocity of electrons, N_C is an effective density of states in the conduction band and k_B is the Boltzmann’s constant. In case of electron traps, activation energy E_a determines a deep energy level position (E_T) in the band gap in relation to the conduction band edge (E_C), according to the relationship E_a = E_C – E_T. Exemplary Arrhenius plots of the three traps observed in low temperatures for the sample with 1.6 % of N, are shown in Fig. 4. For comparison, the results obtained by conventional DLTS (crossed points) as well as high-resolution Laplace DLTS (open points) analysis were presented. For conventional analysis, the Arrhenius plots were obtained by measuring a shift in the DLTS temperature peak position as a function of emission rate, whereas in high-resolution analysis, the emission rate peak position was measured at various fixed temperatures. Furthermore, the deep-level trap concentration (N_T) was determined using the formula [12]: NT = 2NDΔC/C0 $${N_T}{\text{ = 2}}{N_D}\Delta C/{C_{\text{0}}}$$ (2)

where C₀ is a capacitance of the Schottky diode at a fixed reverse bias and N_D is a net doping concentration of shallow levels, obtained from C-V measurements. Calculated values of activation energy and capture cross section (in these studies assumed to be temperature independent) of electron traps E1, E2 and E3 observed in a low-temperature DLTS spectrum are collected in Table 1. Moreover, concentrations of traps, as extracted from DLTS-peak amplitudes for long enough filling pulses equal to 1 ms (ensuring that all the traps are really filled up, i.e. DLTS-peaks intensities are saturated), are collected in Table 2.

According to the data presented in Table 1, the deep-level position of the trap E1 varies significantly with varying nitrogen concentration in the range of 0.15 eV to 0.24 eV, suggesting its attribution to the nitrogen-related defect in GaNAs. It is generally known, according to the BAC model [10], that the parameters of the N-related traps in dilute-nitrides are strongly dependent on the N content, due to the band gap reduction with incorporation of the N atoms into the crystal. They may also vary depending on the sample quality, growth method, and post-growth processing. According to theoretical predictions, the N complexes, such as nitrogen split interstitials N–As and N–N on single As sites are the dominant defects in GaNAs [23]. Experimental data confirm that the as-grown GaNAs layers contain a significant concentration of nitrogen interstitials [7, 24]. It is also known that formation of the (N-As)_As complexes leads to compressive strain in the epilayer while the (N–N)_As complexes induce less tensile strain compared to the substitutional N_As atom. Formation of the N complexes is mainly responsible for the observed significant deviation of the lattice parameter in GaNAs from the Vegard’s law [25] and a low luminescence efficiency of as-grown (Ga, In)(N, As) materials [7, 24]. Therefore, the E1 trap with the activation energy of 0.15 to 0.24 eV was tentatively associated with one of such N complexes. Similar electron traps with comparable activation energies (depending on the N content) have been already observed in GaNAs grown by MBE [26, 27], CBE [16] as well as MOVPE [28].

On the contrary, the electron traps E2 and E3 maintain almost the same activation energies in all studied samples ranging from 0.30 eV to 0.32 eV and 0.48 eV to 0.49 eV, respectively. The energy positions of these traps are very close to two commonly observed deep levels in a bulk and epitaxial GaAs, as well as GaAs-based structures grown by MOVPE, called EL6 and EL3, according to the Martin’s et al. [29] classification scheme. These defects were previously observed also in the GaNAs and GaInNAs heterostructures [30]. The macroscopic structure of the EL3 is usually determined as the off-centre substitutional oxygen on the arsenic site (O_As) [31, 32] or it is related to the As vacancy (V_As) [33, 34]. On the other hand, EL6 is commonly associated with native complex defects involving arsenic antisite (As_Ga) and electron-ically shallower center, as gallium or arsenic vacancy (V_Ga, V_As) [33–35]. These defects are generally regarded by some authors as the possible candidates for major recombination centers in GaAs-based alloys [33, 36].

A high-temperature DLTS spectrum consists of a weak positive peak E4, emerging near 370 K for all studied samples, as it is shown in Fig. 5. The energy position of the trap E4 is located at about 0.81 eV below the conduction band for all studied samples. Furthermore, the density of the trap, calculated from the peak amplitude was also very similar and equal to about 2.0 × 10¹³ cm⁻³ to 2.6 × 10¹³ cm⁻³ (Table 2). High-resolution Laplace DLTS spectra, measured at 370 K (i.e. near the DLTS-peak maximum), are presented in the inset of Fig. 5. As it can be seen from this inset, the DLTS peak corresponds to the sharp Laplace DLTS peak, suggesting a mono-exponential emission process expected from a well-defined single energy level related to a point defect.

The corresponding Arrhenius plots obtained from conventional DLTS (crossed points) as well as Laplace DLTS (open points) measurements are shown in Fig. 6. The calculated trap parameters are collected in Table 3.

Similarly to the traps E2 and E3, also the trap E4 manifests the parameters comparable with another native point defect, commonly observed in GaAs and GaAs-based semiconductors, called EL2 [29]. The EL2 defect has been the most important and extensively studied deep-level defect in GaAs in the last two decades. It is generally considered that EL2 level is related to the isolated arsenic antisite defect (As_Ga) or defect complex involving As_Ga with interstitials and/or vacancies (As_i, V_As, V_Ga) [34, 37] and its formation is favorable under the As-rich conditions. The EL2 defect is also commonly observed in GaNAs and GaInNAs layers grown by different techniques [6, 14, 27, 30].

In order to correlate the revealed deep levels with electrical and optical properties of the GaNAs/GaAs structures, more detailed studies are required. In the near future, we are planning to perform successive electrical and optical measurements on as-grown as well as annealed samples, which should provide sufficient information about a real nature of the deep-level defects observed in our samples and their possible influence on the parameters of future GaNAs-based devices obtained by means of APMOVPE technique.