depicts the transmission spectra (1,300 to 1,330 nm) of the QD waveguide for the three types of condition measured. Note that the insertion loss of the devices improved significantly in the annealed waveguides. There was also a significant blueshift in the transmission spectrum of the 600A waveguide as compared to the AG waveguide due to the blueshift of the transition energy of QDs which is in accordance with [14
Transmission spectra of AG, 600A, and 750A with respective single-mode shapes measured at 1,310 nm. Inset shows the FP spectrum of 750A, which was used to calculate the propagation loss.
A good indication of the single-mode propagation obtained was by observing the single-mode Fabry-Perot (FP) spectrum as shown in the inset of Figure
. The calculated propagation loss based on the respective FP spectra was 4.0 dB/cm for AG, 3.7 dB/cm for 600A, and 3.0 dB/cm for 750A at the wavelength range of 1,308 to 1,315 nm. The improvement in the propagation loss indicated the diffusion of the QD layers and an unintentional passivation of the device. When measuring the propagation loss, shorter waveguides from the batch of devices were cleaved instead of using actual DUTs. This was because longer devices will give much finer mode spacing, and this would result in less accurate data.
Besides the improvement in the propagation loss, a significant change to the DUTs after annealing was that it became impervious to wavelength change, hence making the DUTs less sensitive to wavelength variation. As shown in Figure
, when the range of transmission intensities of the AG and 600A DUTs in 1,308 to 1,315 nm were compared, an approximately 50% lesser transmission difference was observed on the latter device, i.e., the range was smaller. For example, at −4 V, the range of DC transmission was approximately 8 dB for the AG DUTs as compared to approximately 2 dB and approximately 0.5d B for DUTs 600A and 750A, respectively.
DC transmission curves of AG, 600A, and 750A for 1,308 to 1,315 nm, in 1-nm steps. Notice that the ‘width’ of the wavelength band decreases (hence less sensitive to wavelength change) with increasing annealing temperature.
The applied reverse bias voltage for the measurement of the DC optical transmission of the DUTs was capped at 7.0 V. This corresponded to the electric fields of 0 to 150 kV/cm. This was because it was too power intensive to drive an EAM at higher voltage. However, it is worth highlighting that the largest leakage current in the measured range was still less than −0.5 μA, suggesting that the breakdown voltage of the QD device was in excess of −7 V. For the as-grown DUT, we had previously reported an extinction ratio of up to 13 dB at a reverse bias of 10 V and approximately 10 dB of ON/OFF ratio for 8 V [6
]. The DC measurement observed indicated that at the length of 1.6 mm, the absorption of the QD-EAM began to saturate at a reverse bias voltage of 6 V and above. Note that due to the observed suppression of absorption at low reverse bias (<2 V), a higher bias voltage was required for the as-grown device [2
]. Nevertheless, since the optical power capability of conventional EAM is normally limited by the piling up of photogenerated holes as a result of heavier effective mass as compared to that of electron, a larger bias voltage would be beneficial to the power handling capability [15
]. This is because the field screening effect due to the trapped holes inside the confinement region will be reduced at higher electric field [16
]. In the case of the annealed samples, the intermixing lowered the field screening effect at lower electric field. Therefore, 600A demonstrated a reduced built-in potential which was in accordance with the interdiffusion induced [17
]. However, the maximum extinction ratio achieved was reduced to 7 dB. The extinction ratio of 750A was further reduced to <3 dB. Hence, although interdiffusion enhances the QD Stark shifts and greatly reduced the built-in dipole moment, at a RTA range which is too high, it reduces the modulation range at higher voltage. The increased transfer curve gradient of 750A followed by weakened modulation at higher voltage could be due to the thermally induced bandgap shrinkage [18
] due to the increased transmitted output light in 750A when compared to AG or 600A. The extinction ratio and propagation loss comparisons of all three DUTs are presented in Figure
to further illustrate the effects of annealing on these two parameters.
Extinction ratio (top) and propagation loss (bottom) of AG, 600A, and 750A.
Due to the low transmitted intensity of the as-grown DUTs and limitation of the photodetector's sensitivity, only the experimental results of the annealed DUTs were obtained. Figure
shows the small-signal intensity modulation of the annealed DUTs measured at 1,310 nm. A significant advantage of intermixing was the reduced DC reverse bias (driving voltage) needed for the small-signal intensity modulation. A similarly structured QD EAM has been reported to demonstrate a small-signal modulation bandwidth of 2 GHz at a reverse bias of 4 V [1
]. For the 600A device, the reverse bias introduced was as low as 0.5 V, and for 750A, no reverse bias was applied. The elimination of a need for DC reverse bias due to the change in the transmission curve brought on by intermixing will also reduce the complexity of the modulator setup and is a promising indication of selective annealing for on-chip integration.
Intensity modulation response of 600A and 750A. Note that the reverse bias voltages are 0.5 and 0 V, respectively.
Note that although the DC extinction ratio of 600A (750A) was reduced to less than 70% (30%) of its original modulation ability, RF measurement on the devices was still possible due to lower propagation loss after annealing. The 3-dB bandwidth of both 600A and 750A is approximately 1.6 GHz. Noting that these are preliminary RF results, similar frequency responses of approximately 1.6 GHz for both 600A and 750A might be due to the non-optimized WG structures and RF matching. That is, the obtained RF performance was limited by the device design and not by the QD materials. Therefore, we believe that an improvement in the high-speed performance will be expected following the optimization of QD waveguide design and improved RF matching.
The realization of RF measurement on the processed (annealed) lumped-element QD-EAM confirms the prospect of QD epiwafer in monolithic integration for future references. By applying low-cost intermixing, such integration will have low insertion loss and polarization-independent properties [14
]. This is because the integrated devices would actually be made from the same epilayers unlike other types of integration. Therefore, the EAMs would naturally be tuned to the same polarization as that of the emitted radiation from the corresponding QD lasers, and improved extinction ratio may even be observed due to the improved absorption strength of the same platform that integrated devices share.