定义：基于硅芯片的光子学技术。

近年来在微电子学应用领域，基于硅芯片的平台技术占据了非常重要的作用。它如今是复杂的微处理器、大型存储电路和其它数字和模拟电子装置的基础。自从引入了绝缘硅片技术后，已经实现了将光子学功能集成到这种平台上，因此硅基光子集成回路成为可能。其中，不同的光学元件采用硅波导连接在一起。这种回路可用来实现电路板之间、电路板上的芯片间或者单个芯片间非常快的通信。这种通信链路的需求量很大，因为微处理器迅速的发展过程今后会受限于电子连接器有限的传输带宽。光数据传输能够实现更高的数据速率，同时也不存在电磁干扰的问题。这种技术在光纤通信的其它领域也非常重要，例如光纤入户。 
硅光子学也可以从光子学角度去看，后者是基于其它材料的。应用硅光子学器件（即使是电泵浦的硅激光器和硅放大器），可以实现更小的、更便宜的光子器件，能够使一些目前因为成本问题无法实现的应用成为可能。 

技术难点 
尽管硅光子学有很大的前景，但是该技术也面临很多挑战： 

• 由于硅具有非直接带隙，因此发光效率很低。尽管针对这一问题已经采用了很多解决方法，基于硅的激光器或放大器仍然无法与其它基于GaAs或者InP的激光器或放大器相媲美。 
• 硅的带隙也较大，因此无法探测通讯波段1500nm和1300nm附近的光。 
• 硅具有二阶非线性，因此无法制作电光调制器。 
• 芯片上激光光源的散热也是一个问题。 
• 光学连接器需要非常精准的对准，因此为了实现批量生产需要提高对准技术。 

也可以制作混合器件，即有三五族半导体材料（具有直接带隙和电光性质）制作的结构提供光子学功能，将其放置在包含体电子学元件的硅芯片上。其中一种技术采用外延层再生长过程，这种方法比较复杂并且会极大的降低产量。因此，混合器件价格比较昂贵，并且由于其工艺复杂而影响其发展。另一种方案是采用复杂的连接过程将包含波导的硅芯片与提供增益的铟磷芯片连接起来。当然，全硅方案更适合大范围的应用。 

研究现状 
下面简要描述硅光子学目前的研究现状： 

• 硅适宜于制作波导传输光。例如制作采用氧化物包层的条形波导，其传输损耗小于1dB/cm。硅的透明范围从约1100nm到远红外区域。它对模式有很强的限制，因此即使在很大弯曲时也不会有很大的弯曲损耗。它也可以利用非线性实现一些特定的功能，例如通过四波混频实现放大。制作纳米锥形结构可以实现与单模光纤进行高效的耦合，具有很大的有效模式面积。 
• 对于激光光源和放大器应用，由于硅具有间接带隙因此不能应用。有些在多孔硅和硅纳米颗粒方面的研究已经取得了一些进展，但是性能仍然不能与铟磷器件相比。但是，硅能够实现有效的拉曼放大，因为硅的拉曼增益系数很高，并且波导可以将模式限制在很小的范围内。尽管拉曼激光器或者放大器仍然需要光泵浦源，但是它能够得到更长的波长，甚至可以产生多个波长。另一个方法是在三五族半导体材料实现有源功能，然后与硅波导结构连接起来；硅波导的衰逝场足够强能够实现有效的放大。也可以利用直接在硅上生长锗制作单片结构，其中掺杂的锗作为激光器材料。 
• 采用马赫-曾德尔干涉仪和相位调制能够得到硅基光调制器：通过在电极注入载流子可以改变干涉仪一个干涉臂的折射率，然后转化成相位变化引起传输功率的变化。还可以采用微环谐振腔。这种器件可以达到几个千兆的传输带宽。在硅上采用外延层锗也可以实现非常紧凑和高效率的电吸收调制器。 
• 硅光电探测器只对波长小于1100nm的光敏感。在1500nm或1300nm附近通信波长的光电探测器可以采用SiGe材料实现。但是由于晶格失配会产生新的问题，产生晶格缺陷。 

在硅光子学发展成一门核心技术之前，还有很多工作要做，需要很大的资金投入。但是，其潜在优势也吸引了很多大公司，例如英特尔，认真寻求其发展。如果其获得成功，将会是非常有用的一门技术，会对光子学和微电子学应用带来极大的好处。

Definition: photonic technology based on silicon chips

For applications in microelectronics, an extremely powerful technology platform based on silicon chips has been developed in the recent decades. This is now the basis of complex microprocessors, large memory circuits, and other digital and analog electronics. With the introduction of the silicon-on-insulator technology [3] it has been demonstrated that photonic functions can be integrated into this technology platform, so that silicon-based photonic integrated circuits became possible. Here, different kinds of optical components can be connected with each other using silicon waveguides [1]. Such circuits could be used e.g. to establish very fast communication between circuit boards, between chips on a board, or even within single chips, e.g. connecting different cores of a microprocessor. There is a strong need for such fast communication links, because the rapid progress of microprocessors may soon be severely limited by the transmission bandwidth capabilities of electronic connections, made e.g. of copper. Optical data transmission allows for much higher data rates and would at the same time eliminate problems resulting from electromagnetic interference. The technology may also be useful for other areas of optical communications, such as fiber to the home.

Silicon photonics can also be considered from the viewpoint of photonics, which is so far normally based on other optical materials. (Fused silica = amorphous SiO₂ is common in photonics, e.g. in silica fibers, but much less elementary silicon.) The implementation of silicon-based photonic devices, maybe even electrically pumped silicon lasers and silicon amplifiers, could possibly lead to much smaller and much cheaper photonic devices, making accessible a range of applications which so far have been impossible already for reasons of too high cost.

It is clear that an enormous amount of work, corresponding to huge capital investments, is still required before silicon photonics can be established as a key technology. However, the potential merits motivate big players such as Intel to pursue this development seriously. If it is successful, it can lead to a very powerful technology with huge benefits for photonics and microelectronics and their applications.

Technological Challenges

Although the possible merits of silicon-based photonics are huge, there are also very substantial challenges for such a technology:

• Having an indirect band gap, silicon is a very inefficient light emitter. Although various tricks have been developed to get around this, the laser or amplifier performance of silicon-based devices cannot compete with that for other approaches, based on, e.g., gallium arsenide or indium phosphide.
• The band gap of silicon is also larger than desirable, making it impossible to detect light in the telecom spectral regions around 1.5 and 1.3 μm.
• Silicon has no χ⁽²⁾ nonlinearity, making it impossible to realize electro-optic modulators with this material.
• The heat dissipated by a laser source on a chip might well be more than is convenient.

It is possible to fabricate hybrid devices where the photonic functions are provided by structures made of III–V semiconductors (with a direct band gap of suitable width and electro-optic properties), such as indium phosphide, and these are placed on a silicon chip containing the bulk of the electronic components. One class of techniques is based on epitaxial regrowth procedures, which are complicated and often greatly reduce the yield. Another approach is to apply a sophisticated bonding process to combine a silicon chip containing waveguides with an indium phosphide chip providing the optical gain [17]. Here, precise alignment between the two chips is required.

Both technical approaches, leading to silicon hybrid devices, tend to be expensive and are strongly limited in complexity. Therefore, all-silicon solutions, arising from the “siliconization of photonics”, would be more suitable for widespread application. This is also tried for a range of mid-infrared applications such as biochemical sensors.

State of Research

The following paragraphs briefly describe the current state of research concerning basic building blocks of silicon photonics:

Guiding Light

For guiding light in waveguides, silicon is suitable [1, 2]. There are e.g. rib waveguides with oxide cladding, exhibiting propagation losses of well below 1 dB/cm. The transparency range of silicon extends from ≈ 1.1 μm to the far-infrared region. The tight mode confinement allows sharp bends without excessive bend losses. It also enables the use of nonlinearities for certain functions, e.g. amplification via four-wave mixing. Efficient coupling to single-mode fibers, having much larger effective mode areas, is possible with nanotapers [7].

Laser Sources and Amplifiers

For laser light sources and for amplifiers, the indirect band gap of silicon is hardly usable. Some progress has been achieved with porous silicon and with silicon nanoparticles in silica, but the performance achieved can not compete with that of e.g. indium-phosphide-based devices.

On the other hand, silicon allows for efficient Raman amplification, because the Raman gain coefficient of silicon is very high and the waveguides confine the mode to a very small area. Although a Raman laser [8, 19, 23] or amplifier [22] still requires an optical pump source, it can be useful for accessing longer wavelength regions, and possibly even to generate multiple wavelengths [23].

Another available amplification mechanism is parametric amplification based on the χ⁽³⁾ nonlinearity of silicon [18]. SOI channel waveguides provide good field confinement and thus substantial broadband parametric gain, at least when applying relatively intense (multi-watt) pump pulses.

Still another approach is to provide the active function in a III–V semiconductor material (see above), which is bonded to a silicon waveguide structure; the evanescent field of the silicon waveguide can then be strong enough for efficient amplification [17]. Also, it is possible to fabricate monolithic structures with germanium grown directly on silicon (Ge-on-Si technology), where doped germanium serves as the laser material [26].

If the realization of a laser source on a silicon chip is not feasible, one may use an external fiber-coupled laser diode and couple the light to the silicon chip with a single-mode fiber, using a nanotaper [7].

Modulation of Light

Silicon-based optical modulators can be realized with Mach–Zehnder interferometers and phase modulation via a change in carrier density [12]: injecting carriers with an electrode changes the refractive index in one arm of the interferometer, which translates the phase change into a change in power transmission.

Another possibility is to use a micro-ring resonator [20, 25]. Transmission bandwidths of multiple gigabits per second can be achieved with such devices.

Very compact and energy-efficient devices can also be realized as electroabsorption modulators made with epitaxial germanium on silicon [24].

Photodetection

A silicon photodetector (photodiode) is normally sensitive only for light with wavelengths below 1.1 μm, corresponding to the band gap. Photodetectors for telecommunication wavelengths around 1.5 or 1.3 μm are possible with silicon–germanium alloys (SiGe) [13]. Problems arise from the resulting lattice mismatch, which leads to crystal defects.

A new kind of silicon-based photodetector is the plasmonic internal-photoemission detector (PIPED). It is based on plasmonic effects at a metal–semiconductor–metal (MSM) interface, which forms an optical waveguide: light is absorbed in a metal barrier, generates surface plasmon polaritons (SPP); those generate “hot electrons” in the metal, which then can tunnel through the semiconductor barrier when a voltage is applied between the electrodes. Such photodetectors can be made very small and can have a very high detection bandwidth for light e.g. at 1550 nm wavelength [29].

Terahertz Silicon Photonics

Silicon technology can not only be used for working with light, but also with terahertz radiation. For example, one can realize plasmonic internal-photoemission detectors (PIPEDs) on the silicon photonics platform, which can be used as terahertz detectors [30].

Bibliography