1. Introduction

Silicon photonics is a rapidly-developing field of technology that is expected to enable the fabrication and integration of optical components with CMOS technology. Various silicon photonic devices such as waveguides, modulators, and detectors have been successfully demonstrated [1–8]. Efficient silicon light emitters and lasers are highly desirable but also especially challenging due to the indirect band gap of silicon. Various approaches to overcome this limitation have been investigated. Among these are erbium-doped silicon devices [9,10], the all-silicon Raman-conversion laser [11,12], quantum-confined structures such as silicon nanocrystals [13,14], dislocation and point defect engineered devices [15,16], and hybrid technology for wafer-scale integration of silicon waveguides with InP (or InAlGaAs)-based lasers [17].

Recently, stimulated emission and optical gain, characteristics of lasing, at 1.28μm have been observed in periodic nanopatterned crystalline silicon under optical excitation at cryogenic temperatures [18]. The source of this emission is attributed to the bi-stable carbon-substitutional carbon-interstitial (CsCi) complex known as the G center. Evidences from absorption and emission spectroscopy and transmission electron microscopy suggest that the G centers are incorporated, in the nanopatterning process, into the side-wall of the nano-pores [19]. The structural, thermal, vibronic, optical and electronic properties of the individual G center itself in irradiated silicon have been extensively investigated [20–28].

Substitutional carbon atoms are necessary for G center formation and exist naturally in silicon wafers at concentrations between 10¹⁵ and 10¹⁷ atoms/cm³ depending on the crystal growth technique. A G center is created when a mobile interstitial carbon atom (Ci) binds with a substitutional carbon atom (Cs). The process is understood to proceed as follows: silicon interstitials (Si_i) are created as the result of a lattice damage event. A mobile Si_i can then migrate to a Cs. The Cs then gives its lattice site to the Si_i in a process known as the Watkins exchange mechanism. The resulting mobile Ci can then migrate until it binds to another Cs forming a CsCi pair, known as the G center. Thus, G-center creation depends both on the density of Cs present in the lattice [29] and on the density of displacement-introduced Si_i’s.

Early studies were interested in investigating the effects of radiation damage, as such in the past studies G centers were created via electron, ion or gamma ray bombardment. These bombardment techniques necessarily inflict damage on the entire lattice, and thereby increase both the electronic and optical losses without intentionally increasing the capture of carriers by the G centers. A weak G line has also been observed to result from reactive ion etching (RIE) [30]. The creation of G centers by RIE was shown to affect only a thin layer at the surface of the Si crystal. Previously, electroluminescence from G centers was demonstrated in an electron-bombarded silicon p-n junction at cryogenic temperatures [31].

The advantage of G-center introduction via nanopatterning for the purpose of fabricating light-emitting devices is that the damage to the lattice is contained in thin shells at the walls of etched pores [18] while leaving the majority of the lattice in its pristine form and thereby minimizing the optical and electronic recombination losses. It was not clear, however, if G centers created this way could be activated electrically.

In order to further explore the possibilities of this system, we investigate in this paper G center electroluminescence in a nanopatterned p-n junction based on silicon pre-implanted with carbon followed by solid-phase epitaxial regrowth.

2. Fabrication of carbon rich silicon (Si-C) using solid phase epitaxial regrowth

In [18] electronics grade silicon on insulator (SOI) with [Cs]~2.5×10¹⁶ cm^-3 was used. In order to increase the G-center originated luminescence of this system it is necessary to increase the amount of substitutional carbon [30].

Carbon doping of silicon using conventional ion implantation is not feasible due to its low solid solubility—on the order of 10¹⁷ cm^-3. Instead, a combination of a two-step implantation and solid phase epitaxial regrowth is employed. This method takes advantage of the increased carbon solubility at the interface between crystalline and amorphous silicon [32]. The top layer of the silicon crystal is ‘pre-amorphized’ by silicon ion implantation prior to the carbon ion implantation. The crystal is then annealed in nitrogen or argon atmosphere to induce solid-phase epitaxial regrowth and thereby re-crystalization. Si_1-xC_x alloy with [Cs] of up to 7×10²⁰ cm^-3 was achieved using this technique [33].

In this work, boron doped p-type silicon with a resisitivity of 0.2–0.4Ωcm and [Cs]~2×10¹⁶ cm^-3 was used as a substrate for carbon enrichment. The silicon was first amorphized by implantation of 200keV Si⁺ ions at a dose of 2×10¹⁵ cm^-2, followed by implantation of 65keV C+ ions at a dose of 1×10¹⁴ cm^-2. The sample was then annealed at 675 °C for 10s to induce re-crystallization. The implanted carbon formed a 250nm think layer with a mean concentration of 10¹⁹ cm^-3.

Substitutional carbon in silicon is usually detected by its IR absorption at 607 cm^-1 [33]. Due the ultra-small thickness of the Si-C layer, the expected absorbance was less than 0.1% making the detection of this absorbance line below the sensitivity limit of available FTIR instrumentation.

3. Enhanced G center photoluminescence of nano-patterned Si-C

The periodically nanopatterned Si-C structure was fabricated using the highly-uniform self-organized anodized aluminum oxide (AAO) nanopore membrane as an etch mask [18, 34]. The free standing AAO membrane was lifted from an aqueous solution and placed on top of the Si-C and subsequently etched in an RIE machine. The RIE conditions were: Cl₂, 30sccm; BCl₃, 5sccm; 50mT; 100W; 4 minutes. The result was a hexagonal mesh of nanopores with 50nm diameter, 100nm pitch, and a depth of about 200nm. The fabrication process and an SEM micrograph of the nanopatterned Si-C are shown in Fig. 1.

 

Fig. 1. (a). Schematic of the SPE growth process. A pristine Si lattice is amorphized by bombarding with Si ions followed by a carbon implantation. The crystal is then re-grown using the undamaged bottom layer as a seed layer. b. A SEM image of an AAO membrane atop a SOI wafer. c. A SEM image of an array of nanopores in the silicon wafer resulting from RIE through the AAO etch mask. The AAO has been removed after the etching.

Download Full Size | PDF

Fig. 2 shows the photoluminescence spectra at 25K of the nanopatterned carbon-rich silicon compared to that of nanopatterned silicon without carbon enrichment. The 514nm line of an argon ion laser was used for excitation. The excitation power was 200mW and the beam spot size was 3mm in diameter. The emitted photoluminescence was collected using a 3 inch, F/4 concave mirror and focused onto a spectrometer. An InGaAs photodiode array was used to measure the intensity of the PL.

In the nanopatterned silicon control sample without being subjected to implantation, the G line is present but its intensity is lower than that the signature phonon-assisted band-edge PL at 1130nm. In the implanted and nanopatterned Si-C the intensity of the band-edge is reduced while the G line increases by a factor of 20.

While the concentration of the implanted carbon atom was about 500 times more than the native carbon concentration present in the silicon, the increase in G line luminescence was not as many times. One likely explanation is that only a small fraction of the implanted carbon was incorporated as Cs during re-crystallization. Another possibility is that only a limited number of silicon interstitials was created during the nanopatterning, and consequently the number of G centers was also limited. It is interesting to note in this context that we have also experimented with nanopatterning using fluorine based RIE chemistries such as CF₄, CBrF₃, and SF₆, and while nanopores were obtained no G line was observed. We hypothesize that even with chlorine based RIE the number of G centers created is limited by the number of Si_i introduced during RIE.

 

Fig. 2. Photoluminescence spectra of nanopatterned silicon and Si-C at 25K excited with 514nm line of argon ion laser. Bandedge PL at 1130nm and G-line at 1278nm are observed.

Download Full Size | PDF

4. G center electroluminescence in a nanopatterned Si-C p-n junction

The process of carbon enrichment described in the preceding section laid the foundation for advancing to the next goal—electroluminescence —and was used in the fabrication of a nanopatterned p-n junction. The same p-type substrate was used. Again, the silicon was first amorphized by implantation of 200keV Si⁺ ions at a dose of 2×10¹⁵ cm^-2, followed by implantation of 65keV C+ ions at a dose of 1×10¹⁴ cm^-2, and implantation of 80keV P⁺ ion at a dose of 10¹³ cm^-2. The sample was then annealed at 675c for 10s to induce re-crystallization. This implantation process was designed to produce a junction at 200nm depth with N_a=5×10¹⁶ cm^-3, and N_d=10¹⁸ cm^-3, and also a carbon rich layer at 150–250nm to coincide with the junction.

Metallization of this device was particularly challenging, because the G centers are known to be annealed away at temperatures above 200c. Therefore, metallization had to be done before nanopatterning. Furthermore, since the nanopatterning was done using chlorine based RIE, metals such as aluminum or gold could not be used for the top n-type contact because they are readily etched by the chlorine plasma. Instead nickel was used. A nickel layer 100nm thick was evaporated onto the surface of the device and then patterned using photolithography and liftoff to produce strips of nickel 40μm wide and 200μm apart. For the p-type contact a 1μm thick layer of aluminum was evaporated on the bottom surface of the p-type substrate. The metal was annealed for 10min at 450c in a forming gas atmosphere. Finally, an AAO was placed on the device as described previously, and pores were etched using the same RIE conditions described in section 3. The schematic sketch of the device shown in Fig. 3(a) illustrates the relationship between the contacts, the junction, and the nanopatterned (i.e. activated G-center region). An SEM image of the nanopatterned region next to one of the nickel strips is given in Fig. 3(b), and the measured I-V curve of the device at room temperature is given in Fig. 3(c).

 

Fig. 3. (a). Schematic drawing of the G center LED. (b). A SEM image of the nanoparrened region near one of the nickel strips. c. IV curve at room temperature.

Download Full Size | PDF

The device was mounted on a cold finger cryostat. Due to the limited heat dissipation capacity of the setup, temperatures below 60K could not be achieved. Fig. 4 shows the electroluminescence from the device at 60K with a current of 50mA. A prominent emission peak is observed at the 1278nm G-line. The broad and continuous ‘G-band’ spanning over 1250nm to 1350nm and centered at 1300nm is also clearly seen [19].

 

Fig. 4. Electroluminescence spectrum of G-center LED at 60K and a current of 50mA.

Download Full Size | PDF

5. Discussion and conclusion

The mechanism of electroluminescence is presented in Fig. 5. The G center is bi-stable and can alternate between two atomic configurations ‘A’ and ‘B’ [19]. Both configurations can trap an electron from the conduction band in an ‘acceptor state’ (which is normally unpopulated), or trap a hole from the valence band in a ‘donor state’ (which is normally populated with two electrons). Below 50K, the optically active ‘B’ configuration is dominant. The G center can be excited optically, promoting an electron from the donor level to the acceptor level. In forward bias, electrons and holes are simultaneously injected across the junction. The G center can trap either carrier type. In order for a photon to be emitted, the G center must trap both an electron and a hole thus becoming ‘excited’. The trapped electron and hole can recombine directly producing a phononless emission at 1278nm (0.97eV), or indirectly producing an emission band extending from 1250nm to 1350nm.

 

Fig. 5. Electroluminescence mechanism of G center showing electrons and holes trapped in the acceptor and donor level, respectively, and the emission of a photon.

Download Full Size | PDF

Two important processes limiting the emission of photons from G centers are that G centers can emit captured electrons (holes) back to the conduction (valence) band and that G centers can convert from the B configuration to the A configuration which is not optically active. As the temperature increases, these processes become more efficient, and the luminescence decreases and finally vanishes at around 80K.

The marked increase in the G line shown in section 3 is undoubtedly due to an increase in the concentration of Cs in the lattice prior to the etching of pores. Yet the nanopatterning step is also crucial in creating G centers as it allows for a large surface area with which the etchants can interact with the crystal to create silicon self-interstitials which lead to G centers while maintaining the crystallinity, band structure and carrier lifetime of the crystal away from the etched surfaces of the pores. In this way, the unetched regions of the crystal provide the pathway to bring in the carriers to the G centers embedded in the pore walls. This is assisted by the short distances between the pores (on the order of 10nm) compared with the diffusion lengths of carriers in Si (on the order of microns).

The results show that the process of carbon implantation followed by solid-phase epitaxial recrystallization and then nano-patterning creation and activation of localized G centers via nanopatterning can indeed give rise to enhanced photoluminescence and electroluminescence, as desired.

While this work demonstrates that light emission by electrical pumping is possible in this system, the system is sufficiently complex and should be further investigated toward the realization of an electrically pumped silicon laser. The relatively weak G-line electroluminescence may be due to poor lateral current spread from the nickel strips to the nanopatterned area leading to a majority of vertical current flow which bypasses the G-center rich zones. Better and more efficient current supply to the G-center zones of the device with minimized shunt current should be an effective measure to explore in order to increase the electroluminescence.