05 Oct 2007
Overcoming silicon's intrinsically poor light emission is a crucial problem that many research groups are tackling. Graham Turnbull and Ifor Samuel describe how their simple hybrid silicon-polymer laser could pave the way towards optical chip-to-chip interconnects.
Organic semiconductors are attractive materials for wavelength-tunable visible solid-state lasers. The simple processing and excellent optical properties of these light-emitting plastics makes them suitable for integration with other optoelectronic technologies.
In the Organic Semiconductor Centre at the University of St Andrews, UK, we have been developing compact polymer lasers and optical amplifiers. We are also exploring hybrid organic–inorganic lasers that could provide a new route to chip-to-chip optical interconnects. This work aims to address silicon's long-standing limitation for optoelectronic applications, that it is a very inefficient light emitter. Such novel sources, which can be fabricated simply using standard CMOS processing, offer a cost-effective new way of implementing optical chip-to-chip interconnects and addressing a current bottleneck in high-speed computing.
Organic semiconductor lasers
Organic semiconductors are efficient, visible light emitters that can be processed simply from solution or by thermal evaporation. Since the discovery of electroluminescence in organic semiconductors, there has been considerable progress in developing organic light-emitting diodes for displays and applications including transistors, solar cells and lasers.
These remarkable materials come in three sub-groups: small molecules, long-chain conjugated polymers and highly branched dendrimers. Small molecules are usually thermally evaporated to form amorphous films or grown as molecular crystals. The polymers and dendrimers may be deposited from solution; essentially they are semiconductors that can be printed to assemble optoelectronic devices.
As laser gain media, conjugated polymers have many distinctive features. Firstly, they exhibit very strong absorption bands, with peak extinction coefficients of approximately 105 cm–1. Large optical gain cross-sections (10–15 cm2) also give rise to substantial gain in very compact devices. They typically have emission bandwidths of >100 nm and changes in chemical structure can tune the emission from 400 to 700 nm (figure 1). Optical gain is possible throughout the visible spectrum making conjugated polymers well suited to applications such as tunable lasers or broadband amplifiers.
Unlike conventional laser dyes, polymers show little emission quenching with increased concentration. Indeed, they can exhibit fluorescence quantum efficiencies of 50% and higher in the undiluted solid state. Polymers naturally form four-level energy systems in which exciton migration along disordered polymer chains, plus vibrational and structural relaxations, substantially separate the absorption and emission bands. This allows simple optically side-pumped geometries to be used for lasers where the peak absorption can be 105 cm–1, and the residual absorption at the laser's emission wavelength is typically <30 cm–1.
The polymer's capacity for simple processing from solution is also relevant and many different microresonator configurations have been demonstrated, including microcavities, microrings, distributed feedback and photonic crystal lasers. To illustrate this processing simplicity, microring whispering-gallery-mode lasers can be made by dip coating an optical fibre in a polymer solution. Wavelength-scale structures for active photonic crystals can readily be moulded into polymer films using soft lithographic techniques.
Recent advances in materials and resonators have led to very low threshold pulsed operation enabling us, and also groups from Braunschweig and Karlsruhe, to report organic semiconductor lasers pumped by an InGaN diode laser. In terms of applications, there has been an exciting recent demonstration by the MIT group of polymer lasers intrinsically working as sensors for ultralow vapour concentrations of explosives.
Together with partners at Imperial College within the UK Ultrafast Photonics Collaboration, we have demonstrated high-gain polymer optical amplifiers and ultrafast sources and modulators for datacomms applications. Our review of organic semiconductor lasers gives a detailed overview of these and other recent advances (Chemical Reviews 107 1272).
Adding light emission to silicon
Silicon is the dominant material in modern electronics and the mass fabrication of wavelength-scale structures is extremely well developed. However, there is a growing need for high-bandwidth interconnects on silicon chips, which may be addressed in the future by integrating optoelectronic components directly onto the silicon processors.
For optoelectronics, silicon has a significant limitation: it is an indirect bandgap semiconductor and an intrinsically poor light emitter. To overcome this poor optical emission efficiency, researchers have pursued approaches including optically excited Raman and nanopatterned silicon lasers; quantum confinement in silicon nanocrystals; the use of rare-earth dopants and bonding of III-V semiconductor lasers to silicon chips.
Our alternative approach combines microstructured silicon resonators with light-emitting organic semiconductors. Despite the strong absorption of silicon at visible wavelengths, under optical pumping, our hybrid silicon-polymer structure works as a red surface-emitting laser (Applied Physics Letters 91 051124).
Integrating a solution-processed polymer adds only a simple supplementary fabrication step, compatible with standard CMOS processing, and does not require careful electrical or optical coupling to the silicon wafer. Crucially, our approach also adds visible light emission to the silicon at wavelengths easily detected by conventional silicon photodetectors.
There are two significant challenges to overcome when integrating organic semiconductors with silicon. The first is the substantial absorption of silicon at the visible wavelengths of the polymer photoluminescence, and the second is silicon's high refractive index, which complicates the way that the laser emission is confined within a polymer waveguide.
Our hybrid device uses a distributed Bragg reflector resonator design and is based on a silicon-on-insulator (SOI) substrate (figure 2). Two periodically microstructured silicon segments act as mirrors, while the silicon epilayer between these is removed, exposing the buried SiO2 layer. A thin film of the organic amplifying medium covers the whole structure.
Due to the higher refractive index of the polymer compared with SiO2, a polymer waveguide is formed between the two silicon mirrors with a confinement factor limited only by the thickness of the polymer layer. The area between the two periodically microstructured silicon reflectors also reduces the absorption of laser light by the silicon.
To fabricate our devices, we used reactive ion etching to reduce the 220 nm-thick silicon epilayer of the SOI substrate to 30 nm. We then defined the Bragg mirrors using electron-beam lithography and reactive ion etching, which in turn exposes the underlying oxide layer.
The silicon microstructured mirrors comprise linear gratings with a period of 360 nm. For an operating wavelength of around 630 nm, the stop-band of the Bragg mirrors is due to second-order diffraction and provides both in-plane feedback and a surface-emitted output coupling of the laser light. The mirror spacing is 50 µm and the entire structure is coated with a thin film of the prototypical conjugated polymer MEH-PPV, to form a polymer waveguide that supports a single transverse mode.
To characterize the laser, we used 1.2 ns pulses at 5 kHz from a frequency-doubled microchip laser to optically excite the hybrid structure at room temperature. The spectral characteristics of the polymer laser were then measured using a fibre-coupled CCD spectrometer.
Surface-emission spectra from our hybrid lasers, below and above lasing threshold, are shown in figure 3. Below threshold, spontaneous emission from the red-light-emitting polymer couples to several optical modes with wavelengths within the stop-band of the Bragg mirrors. These modes can be seen as a series of narrow peaks.
For excitation energies above 45 nJ, the intensity of the lowest wavelength modes increases at a faster rate, corresponding to the onset of lasing. At higher pumping levels (not shown) these dominate the emission spectrum. Not all of the longitudinal modes reach threshold, which we attribute to a variation in surface-emission losses across the in-plane stop-band of the mirrors. Our lasers are tunable over 20 nm, with typical pulse energies in the picojoule range (Applied Physics Letters 91 051124).
The threshold gain for lasing was calculated to be 370 cm–1, indicating that the in-plane mirror reflectivities are <20%. The reflectivity is quite low due to the surface-emission and silicon absorption from the Bragg mirrors, and can be increased with optimized grating design. The optical gain in the silicon-polymer laser is comparable with values achieved with silicon nano-crystals and nanopatterned crystalline silicon (the latter at cryogenic temperatures) and very large compared with silicon-Raman lasers. The high gain surmounts the absorption loss and allows very compact lasers to be fabricated – an important factor for high-density integration.
Our present lasers are optically pumped but we envisage using the underlying silicon chip to modulate the emission either via charge injection or an applied electric field to encode information. For interconnects, a move to higher repetition rates is necessary. Semiconducting polymer lasers operating at MHz frequencies have recently been reported by the Braunschweig group, suggesting that triplet accumulation is less significant than in conventional dye lasers.
One key problem to be solved in the longer term is achieving electrically pumped lasing. Polymer LEDs have been demonstrated with p-doped silicon as a hole injection layer. However, charge-induced absorption and field quenching are currently recognized as substantial obstacles to injection lasing in organic semiconductors. Electrical excitation is also a problem for lasers based on silicon nanocrystals and rare-earth dopants. While electroluminescence is possible, an electrically pumped silicon-based laser remains a major outstanding goal.
• This article originally appeared in the October 2007 issue of Optics & Laser Europe magazine.