Monolithic Silicon Photonic Integrated Circuits (PICs)

Technical Overview

Silicon photonics technology leverages the high-resolution lithography, large wafer size, and high manufacturing yield of the advanced silicon fabrication capabilities that have been developed for the CMOS electronics industry. These fabrication capabilities offer the potential of realizing PICs having greatly reduced size, weight, and power (SWaP), high complexity, and low cost. Silicon photonics is an emerging technology that has already been inserted into commercial digital optical communication products. The high level of integration afforded by silicon PICs may also be beneficial for microwave photonic (MWP) systems. Silicon photonics fabrication at Lincoln Laboratory makes use of the advanced process node (90-nm lithography, 200-mm-diameter wafers) available within the Microelectronics Laboratory cleanroom.

Program Goals

  • Identify and define critical communication, sensing, and signal processing system applications enabled by monolithic integration of silicon photonic components
  • Develop a toolkit of state-of-the-art silicon photonic devices that can be integrated together on a single platform. Devices included are
    • Low-loss waveguides
    • Wavelength filters
    • Optical modulators
    • Photodetectors
  • Design and fabricate silicon PICs and insert them into compelling system applications
  • Provide the government with access to silicon PIC design, fabrication, and packaging resources

Research Highlights

Lincoln Laboratory, in collaboration with researchers on the MIT campus, has been active in the field of silicon photonics for more than 10 years and has developed a toolbox of photonic components. This toolbox includes components such as low-loss waveguides, filters, modulators, and photodetectors. The toolbox has been developed for both analog and digital applications. The following describes some of the components that are available in the silicon PIC toolbox:

  • Waveguides are a fundamental building block for integrated photonic devices. Achieving low propagation loss in a waveguide is important for both maintaining signal strength during transmission through a PIC and achieving high performance in resonant structures such as ring resonator filters. In a typical single-mode silicon waveguide, propagation losses are limited to 1–2 dB/cm. Using techniques such as mode expansion, which was pioneered at Lincoln Laboratory, losses as low as 0.2 dB/cm can be achieved without necessarily sacrificing the performance or compactness of the devices.

    complementary filter banksFigure 1. Optical microscope image of one channel of two complementary filter banks. The two banks are labeled top and bottom. The dashed lines show the paths of the rings and waveguides, which are mostly hidden under the heaters and metal. The spiral-like patterns at each ring are the heaters.
  • Optical filters are used to separate different wavelengths or frequencies of light. In a ring-resonator filter, a resonant cavity is formed by a waveguide having a circular or ring path. At a resonant wavelength, the light couples through the ring to an adjacent waveguide. Off resonance, the light does not couple to the ring and continues along the input waveguide unimpeded. Figure 1 shows an optical microscope image of a ring-resonator filter. For telecom and datacom applications, such filters can be used to separate channels in a wavelength-division-multiplexed (WDM) system. In these applications, the channels are typically 25 to 100 GHz apart. Figure 2 shows the performance of two 16-channel filter banks with 80 GHz spacing. Note that the performance of the two filter banks is nearly identical, and only in few locations do the data not overlap. For RF applications, filters often require much narrower bandwidth on the order of 1 GHz.

    Drop-port transmission characteristics of a pair of 20-channel optical filter banks after tuningFigure 2. Drop-port transmission characteristics of a pair of 20-channel optical filter banks after tuning. The blue-green colors are the top bank, and the red-yellow colors are the bottom bank. Because the data from the two filter banks match well, the data from the top bank are largely hidden behind the data from the bottom bank.

    Additionally, RF applications require very high isolation between channels, often greater than 50 dB. Both of these demands make the fabrication of the filters much more challenging. Figure 3 shows how higher-order filters (the order is the number of coupled rings) can increase the extinction for filters with ~1.3 GHz of bandwidth.

    Comparison of narrow-bandwidth filters with increasing numbers of ringsFigure 3. Comparison of narrow-bandwidth filters with increasing numbers of rings. Increasing the number of rings, or the order of the filter, creates a steeper roll-off and increased extinction. These measurements were limited by the noise floor visible near 40 dB.
  • Optical modulators can be implemented in silicon by using electronic charge carriers (i.e., electrons and holes) to modify the refractive index of the silicon. One method of changing the carrier density is injecting carriers from a p-i-n diode. This type of device is somewhat slow, with time constants on the order of a nanosecond at best. However, these devices are sensitive (i.e., switching power ~1 mW) and can be compact. It is possible to achieve greater than a π phase shift in devices having length less than 0.25 mm. Another method for altering the carrier density in silicon is to use carrier depletion in a p-n junction. This effect is intrinsically very fast, typically only limited by the capacitance of the device. Speeds faster than 26 GHz have been achieved in modulators fabricated at Lincoln Laboratory. However, carrier depletion moves fewer carriers than carrier injection and therefore requires a much longer device.

  • Photodetectors convert optical power into electrical current. The bandgap of silicon is too large to absorb light at the wavelengths greater than ~1 μm. One method for obtaining absorption at longer wavelengths is to create defects in silicon that cause mid-gap states that can absorb light in the 1–2 μm wavelength region. Using this method, photodiodes having efficiencies near 25% (0.3 A/W) have been demonstrated. The absorption length in these devices is large, however, and devices a few millimeters long are necessary to absorb most of the light. Another method to obtain absorption in the 1–1.6 μm wavelength region on a silicon platform is to use germanium (Ge). Germanium can be epitaxially grown on silicon and has a bandgap that makes the absorption of wavelengths of 1.55 µm and shorter possible. Shown in Figure 4 is a cross section of a germanium photodiode fabricated at Lincoln Laboratory. This photodiode had an efficiency of 55% (0.7 A/W) and a low dark current (4 nA at 2V bias).

    Cross-sectional scanning-electron microscope (SEM)  image of a germanium P-I-N photo-diodeFigure 4. Cross-sectional scanning-electron-microscope (SEM) image of a germanium p-i-n photodiode.

The basic components necessary for many photonic systems have been demonstrated using Lincoln Laboratory's monolithic silicon PIC fabrication facility. We are currently working to integrate these components into useful subsystems and addressing issues such as feedback and control of large-scale PICs, electronic-photonic integration, electrical and optical power requirements and limitations, and packaging.

 

top of page