Spectral responses and the corresponding transition mechanisms

Figure 3.1(a) shows the PL spectra of the three samples measured at 10 K under a low excitation power of 1 mW. With increasing Sb compositions, a clear red shift in PL peaks is observed with its optimum optical properties, narrowest FWHM and the most intense PL intensity, in x=10 % sample. The phenomenon is attributed to the compressive strain relaxation together with the suppressed QD decomposition ability of GaAsSb CLs [40]. In this case, with improved carrier confinement and reduced QD size fluctuation, the PL intensity would increase and PL peak FWHM would decrease for x=10 % sample. As for x=20 % sample, significant PL intensity declines and broadened FWHM are observed, which

Fig. 3.1: (a) The 10 K PL spectra and (b) PL peaks as a function of cube roots of excitation powers P^1/3 for GaAsSb-capped samples with different Sb content x.

50

is attributed to the type-I to type-II hetero-interface transition with Sb composition exceeding 14 %. To confirm this attribution, power-dependent PL measurements are performed. As shown in Fig. 3.1(b), energy blue shift of Samples C shows linear dependences on the cube roots of excitation powers P^1/3, which are consistent with the expected type-II behaviors. On the contrary, the peak positions of x=0 and 10 % sample remain unchanged with increasing pumping powers. Therefore, with increasing Sb compositions in the GaAsSb CLs, the gradually relaxed compressive strain accumulated in the InAs QDs would lead to a band gap shrinkage and PL peak red shift. The FWHM reduction of x=10 % sample is attributed to the improved QD uniformity with the GaAsSb CLs. When Sb composition exceeds 14%, the significant PL intensity decline and FWHM broadenings are attributed to the type-I to type-II hetero-interface transition instead of optical characteristics degradation.

The normalized 10 K spectral responses of the three devices biased at 1.2 V are shown in Fig. 3.2. A significant spectral width reduction of x=10 and 20 % device compared with the x=0 % device is observed. The reference GaAs-capped QDIP (the x=0 % device) exhibits the broadest spectral width Δλ/λ of 0.30 at peak wavelength 6 μm, which is comparable with the typical reported values for such devices. A narrowest Δλ/λ of 0.06 at peak wavelength of 6.78 μm is obtained for the x=20 % device. One possible mechanism responsible for the spectral width narrowing is the improved InAs QD size uniformity. In this case, the narrower spectral widths of GaAsSb-capped InAs QDIPs is consistent with the narrower PL FWHM of the GaAsSb-capped sample with 10 % Sb composition. However, since the improvement in spectral widths is much more significant than that of PL FWHM narrowing, the other mechanism should also be involved for the spectral width narrowing of GaAsSb-capped QDIPs besides the improved QD size uniformity. The other phenomenon observed in the figure is a slight response peak red shift of the x=20 % device compared with the x=10 % device. The result suggests that the energy difference of the confinement states responsible for the intra-band transitions is reduced. For GaAsSb-capped InAs QD structures, larger VB

51

Fig. 3.2: Normalized 10 K spectral responses of GaAsSb-capped devices with different Sb content x at 1.2 V.

discontinuity than the conduction band is expected to fall on the interface. In this case, the longer detection wavelengths of GaAsSb-capped QDIPs with higher Sb compositions are directly attributed to the larger InAs QDs resulted from less significant QD decomposition with increasing Sb compositions. The results are different from the response peak blueshift with increasing In compositions of InGaAs-capped QDIPs since there is no additional confinement state appeared with the addition of the GaAsSb CL. Therefore, for InGaAs-capped InAs QDIPs, the possible elongated detection wavelength resulted from InGaAs QW state lowering would be cancelled out by the QD compressive strain relaxation with increasing In composition [80].

52

Fig. 3.3: 10 K PLE spectra of x=0 and 20 % sample and 10 K spectral responses of x=0 and 20 % device. The x-axis of the PLE spectra is re-adjusted by setting the QD first excited state as the “zero” energy. The insets depict the schematic conduction band structures and corresponding transitions response for the spectral responses.

To further investigate the transition mechanism of the QDIPs, the PLE measurements are performed on x=0 and 20 % sample. The 10 K PLE spectra of x=0 and 20

% sample with the PL peak energy as the detection energy are shown in Fig. 3.3. By setting the QD first excited state as the “zero” energy, the x-axis of the PLE spectrums are re-adjusted and compared to the spectrum responses of the x=0 and 20 % device operated at 1.2 V, which are also shown in Fig. 3.3. The peaks corresponding to the absorptions of QD bound states and the wetting layer (WL) are noted on the PLE spectrums. As shown in the

53

figure, for the x=0 % device, the spectral response is well agreed with the energy difference ΔE between the QD first excited and WL states. As discussed in previous literatures for traditional InAs QDIPs, the result suggests the broad response comes from the summation of transitions between QD bound and WL states [81,82]. For the x=20 % device, a dominant and a weaker response peaks with ΔE=184 meV and ΔE=367 meV are observed on the spectral response curve, respectively. Compared with the PLE spectrum of x=20 % sample, the high-energy response should be related to the transition between QD bound states and WL as the case of x=0 % device. The low-energy one should attributed to the transitions between QD bound states. The slight miss-alignment between the PLE and spectral response peaks may be resulted from the severe water absorption at ~ 6 μm (207 meV), which may hinder the actual response peaks. The results suggest that the dominant transition mechanism changes from QD excited to WL states to intra-band transition between QD bound states for x=0 and 20 % device, respectively. The phenomenon is attributed to the larger InAs QDs

resulted from the depression of QD decomposition for GaAsSb-capped InAs QD structures as the case of x=20 % device. In this case, along with the influence of strain-relaxation of GaAsSb CLs on InAs, the energy states in the InAs QDs would be lowered and energy difference between the confinement states would decrease. Therefore, for the x=20 % device, response peak blue shift corresponding to the QD excited to WL states would be observed, which would result in the dominant transition mechanism change to intra-band transition between QD bound states due to the inverse proportion of absorption coefficient over energy differences.

54

Fig. 3.4: The normalized responsivity ratios of x=0 and 20 % device at 1.2 V under different incident light polarizations.

It has been discussed in the previous publications that larger QDs would lead to lower normal incident absorption for QDIPs [83]. Therefore, to confirm the attribution of larger InAs QDs resulted from depression of QD decomposition with the GaAsSb CLs, the normalized peak responsivity ratios of x=0 and 20 % device at 1.2 V under different incident light polarizations are shown in Fig. 3.4. As shown in the figure, although only minor ~ 5 % normal incident absorption ratio difference is observed for the two devices, x=20 % device does exhibit lower incident absorption ratio compared with x=0 % device. The results are consistent with the previous discussions that by using GaAsSb CLs, QD decomposition will be depressed. In this case, larger InAs QDs are obtained for GaAsSb-capped QDIPs.

Therefore, the lowered QD bound states, reduced energy difference in-between the two states and the improved QD uniformity would together contribute to the longer wavelength absorption with narrow spectral widths of GaAsSb-capped QDIPs.

55

In summary, we investigate the effects of the GaAsSb CLs on the spectral responses of InAs/GaAs QDIPs. An extremely narrow spectral response of Δλ/λ=0.06 is observed for the device with 20 % Sb composition. The results suggest that the GaAsSb CLs would depress QD decomposition. In this case, larger InAs QDs with improved QD uniformity is obtained for GaAsSb-capped InAs QD structures, which result in energy level lowering and consequently, reduced energy difference in-between the states. Therefore, compared with standard GaAs-capped QDIPs, GaAsSb-capped QDIPs are of longer detection wavelength and narrower spectral width. The unique characteristic can be advantageous for selective detection at specific wavelengths by using QDIPs.

56