1. Introduction

There has been a growing demand for an efficient Si-based near-infrared emitter in the field of silicon photonics. As compared to the mature band-gap engineering, manipulation of such optically active lattice imperfections as point defects has only recently entered into the focus of interest since it turned out to be a viable technique for providing room-temperature-emitting Si [1–4]. Recently, Cloutier et al. have even reported light amplification and stimulated emission of the sub-bandgap emission centered at 1278 nm (970 meV) that are associated with the carbon-related G-center in a periodic nano-patterned crystalline Si [5]. These are expected to serve as the stepping stone in demonstrating the capability of emissive defects that, if engineered properly, allow Si-based light-emitting diodes (LEDs) [6,7] and eventually Si-based lasers.

So far, a rich variety of emissive defects in Si have been found, and some of them are already well documented. The {311} rod-like defects (RLDs) represent a well-known but little explored example of these. They exhibit a sharp photoluminescence (PL) feature often referred to as E-line, which is centered at 1377 nm (900 meV) at cryogenic temperature. Its spectrum gradually broadens and shifts to lower energies with increasing temperature, while it is still observable near room temperature unlike many of defect-related emissions in Si [8–10]. Nonetheless, no attempt has ever been reported to observe electroluminescence (EL) of the E-line. This may have to do with how the RLDs are introduced to Si. It has been pointed out that Si interstitial pairs, regardless of preparation methods, play a vital role in creating the {311} RLDs. In fact, methods vary but they are more or less dependent on the bombardment of charged particles, including ion implantation [9] and electron beam irradiation [11], except for a few examples based on strain relaxation in the structures containing Si during prolonged annealing [8,10,12,13]. As excess damage often gives rise to nonradiative recombination centers, the luminescence efficiency is at stake, which is a concern. In this context, thin film growth technology is instrumental in that it provides a possible solution to this by offering control over the creation and elimination of defects.

In this letter, we demonstrate EL sharply peaked at 1377 nm due to the {311} RLDs in a Si layer with embedded InSb quantum dots (QDs). A thermal budget was given to the sample chip in order to relax the built-in strain. Passing a current through the chip caused the E-line to develop out of the otherwise broad background of EL, which allows the operation of such an electric-field-active narrow-line LED up to 150 K.

2. Experimental

InSb-QD-embedded Si was grown by solid-source molecular beam epitaxy (MBE) on a p-type CZ-Si(001), 5-10 Ωcm. Si was deposited using an electron-beam evaporator and effusion cells were used for the evaporation of In and Sb at a rate of approximately 0.1 nm/sec. The detailed growth conditions are found elsewhere [14]. Briefly, the growth temperature was 300 °C, and in order of growth there are a 50-nm Si buffer layer, five bilayers of InSb QDs, and a 100-nm thick Si cap. Spontaneous n-doping of the grown layer occurred due to the incorporation of Sb atoms that are surface segregating during growth and/or remaining in the vacuum. This indicates that a p-n junction is easy to be formed if we start with a p-type Si substrate. The post-growth annealing was performed ex situ in a furnace flooded with nitrogen so that RLDs are launched into Si as the thermal strain relief proceeds. Preliminary cross-section transmission electron microscopy study shows that strain contrast has changed significantly following the annealing, which is in qualitative agreement with the previous reports [8,10,12]. PL and EL were measured by lock-in detection using a liquid-nitrogen cooled Ge pin photodiode. The excitation source of PL was the second harmonic of a continuous-wave diode-pumped Nd:YAG laser (532 nm). The maximum excitation power was 15 mW over a spot size of approximately 1 mm in diameter. An LED structure was created by vacuum deposition of Al on both chip faces with the rear electrode being silver epoxied to a Cu heat sink. A dc source was used and the edge-emitting EL was collected by using lens optics. Figure 1(a) shows the schematic sample cross-section while Fig. 1(b) shows typical current-voltage (I-V) traces recorded at 9-K and room temperature, showing standard rectifying characteristics. The fairly large turn-on voltage at 9 K, ≈6V, as compared to that at room temperature, ≈3 V, is due presumably to the carrier freeze-out effect at very low temperature.

 

Fig. 1 (a). Schematic LED sample structure. The active region contains an MBE-grown InSb-QDs-embedded-Si. {311} RLDs were introduced ex situ by post-growth annealing. (b) Typical I-V characteristics of LED at 9-K and room temperature.

Download Full Size | PDF

3. Results and discussion

Figure 2 compares the 9-K PL spectra of the InSb-QDs-embedded-Si ((a) as-grown and (b) annealed at 600 °C for 30 min.) on a normalized scale. The spectrum of the as-grown sample is dominated by a broad background emission due to the InSb QDs, over the wavelengths, 1075 −1700 nm [14]. The doublet centered at 1112 nm (1114 meV) and the peak located at 1180 nm (1050 meV) are presumably due to point defects, which tend to be diminished after annealing and therefore are not of interest here. The phonon replicas of Si are not observed although the pump light with a penetration length ~1 μm largely excites the substrate side of Si. This is indicative of carrier funneling effects and/or a fairly large cross-section of the relevant electronic states other than the Si band-edge. After annealing at 600 °C, however, a small signature of the E-line (indicated by the arrows) begins to be visible at 1377 nm (900 meV) riding on the broad EL band that has undergone a three-fold increase in terms of intensity as compared to the as-grown spectrum. Note further that the spectral dip around 1385 nm (895 meV) that is consistently observed for all the spectra throughout this work is an artifact attributed to O-H absorption.

 

Fig. 2 Normalized 9-K PL spectra of InSb-QDs-embedded-Si LED: (a) as-grown, (b) after receiving post-growth annealing 600 °C for 30 min. Arrow indicates the developing E-line due to {311} RLDs. The spectra have been shifted vertically for clarity.

Download Full Size | PDF

Interestingly, as visible in Fig. 3(a) , the E-line was found to grow rapidly as the electric field was applied (electric-field-active EL), and the resultant spectrum was such that the broadband emission was nearly totally suppressed, leaving the predominant E-line feature. Such a spectral dominance switch upon current injection was continuous as a function of bias voltage as shown in Fig. 3(b). The broad background is seen at a very low injection current, 0.06 mA, while it tends to be quenched above 10 mA. This is more clearly represented by the suppression ratio, defined as the intensity of the E-line divided by that of the broad band, as a function of current as shown by the filled triangles in Fig. 3(c). Also plotted in Fig. 3(c) is the E-line intensity. Interestingly, the E-line evolves in an unusual nonmonotonic way, which seems to be connected with the spectral dominance switch. The underlying physics of the electric-field-active EL clearly merits further study.

 

Fig. 3 (a). Comparison of PL and EL spectra taken at 9 K from InSb-QDs-embedded-Si LED containing {311} RLDs. A broad background visible for PL is quenched in the case of EL with only the sharp E-line dominating at 10 mA. (b) Synopsis of EL spectra as a function of forward injection current. Note the spectral changes including a monotonic shift of the peaks with increasing current. (c) E-line intensity versus injected current at 9-K, and the ratio of the intensities of the E-line and the broadband emission due to InSb QDs. (d) EL peak energies as a function of current. The peak energies have been determined by Lorentzian fitting.

Download Full Size | PDF

With increasing current, the peak shifts to lower energies while the single 900-meV peak splits into a doublet, and a single peak of increased width follows. (Note that the O-H absorption dip does not fully account for the appearance of the doublet feature discussed here.) The peak energies are plotted as a function of current in Fig. 3(d), which suggests a dominance switch of two relevant electronic states with increasing current. A previous work has pointed out that there are at least two distinct but correlated classes of defects contributing to the PL features near 900 meV: line interstitial defects and planar defects [10]. In fact, the spectral features reported previously are very similar to what is observed in the present work. In the light of technology, such characteristics are potentially useful for application: as inferred from Fig. 3(d), the repeatable spectral variation upon changing current suggests emission wavelength tuning by controlling current.

The EL survived up to 150 K as visible in Fig. 4 . Importantly, an anomalous temperature variation of EL intensity is evident as captured in the inset. Up to 40 K, the E-line feature steadily decreases in intensity with increasing temperature, i.e., standard thermal roll-off. However, the trend is reversed above 80 K, and the EL intensity begins to increase and at 150 K it is even larger than that at 9 K. Considering the total absence of the Si-band-edge related emission at 80 K and 150 K, the restoration of the EL intensity at elevated temperatures indicates the presence of a shallow trap accommodating a large fraction of injected carrier while releasing them at increased temperature. Further study is now in progress.

 

Fig. 4 Temperature dependent EL spectra of InSb-QDs-embedded-Si LED containing {311} RLDs at a constant forward bias current of 20 mA. Inset shows the integrated EL intensity versus reciprocal temperature. Solid lines are to guide the eye.

Download Full Size | PDF

The apparent loss of spectra at temperatures higher than 150 K seems to be consistent with the spectral changes shown in Fig. 3. The EL intensity at increased current levels tends to drop starting at 80 mA and a sharp decline follows. Such a tendency naturally invokes an argument based on the local lattice heating, which should in part explain the absence of spectra above 150 K.

It is worth noting that the present work was the first to demonstrate the introduction of RLDs to Si by means of epitaxy in combination with post-growth treatment. As a matter of fact, there have been no successful reports of introducing the {311} RLDs directly into Si without recourse to irradiation of highly energetic particles. Strain relaxation of the embedded InSb QDs that occurred during annealing might have played a key role in giving rise to the emission pertaining to the {311} RLDs. These points have to be justified in the future by performing thorough morphological characterization, including the TEM observations mentioned in the preceding section, in combination with dedicated EL measurement.

4. Conclusions

In conclusion, an attempt was made to develop a new class Si-based LED containing the narrow-line, electric-field-active emissive {311} RLDs by first growing InSb-QDs-embedded-Si followed by strain relaxation of the grown layers during post-growth annealing. LED was operated up to 150 K, accompanied by repeatable spectral changes with varying temperature and current. Engineering of appropriate defects in Si is expected to function as the touchstone in developing a viable Si-based light emitters including LEDs.