标签: SiGe

In recent weeks, much attention has been devoted to silicon-germanium (SiGe) heterojunction technology as the next step in building silicon-based ICs down below 10 nanometers. It is not the first time that the semiconductor industry has turned to this compound heterojunction semiconductor for help. However, this time most of the attention has centered around IBM's continued support for SiGe in its foundries.   In June IBM ramped up its foundry business with new SiGe and silicon-on-insulator processes targeted mostly at wireless radio-frequency (RF) circuits. Traditionally fabricated with more exotic heterojunction technologies such as gallium arsenide, RF circuits in mobile devices are being fabricated instead with SiGe because of it’s speed advantages, performance/power tradeoffs, and its compatibility with silicon.   And this month (July 5), IBM announced the first 7nm (nanometer) node test silicon ICs with functioning transistors, developed in collaboration with GlobalFoundries and Samsung at Suny Polytechnic Institute. This breakthrough may change the way companies build integrated circuits, especially as scaling gets close to 1 nm and below.   Michael Liehr of the SUNY College of Nanoscale Science and Engineering, left, and Bala Haranand of IBM examine a wafer comprised of the new SiGe 7nm-based integrated circuits. (Source/Credit: Darryl Bautista/IBM)   The team took advantage of the fact that SiGe increases the frequency of a transistor’s short-circuit current gain, thus creating the conditions that accelerate the movement of electrons across a transistor structure. In a conventional silicon transistor, higher doping would lower the current gain and allow leakage back to the collector. In a SiGe transistor, the bandgap potentials maximize the current gain and minimize leakage, making possible much higher gate frequencies at lower power.   In addition, germanium makes the silicon act more like a conductive metal structure and less like an insulator. The base is more conductive and so the more the base resistance of a transistor structure decreases, less noise it produces versus pure-silicon devices, especially important in the sub-10 nanometer range where noise is an ongoing problem.   Germanium, from which the first solid state semiconductor transistor was created, is now integral to the latest generation of nanometer-sized transistors used in silicon-based ICs. (Source: Bell Labs )   SiGe has always been the back-burner IC technology the semiconductor industry turned to for very specific problems. For example, analog IC designers have used variations of bipolar SiGe (silicon germanium) and BiCMOS (bipolar-CMOS) SiGe to get the performance they needed. And in some special cases germanium-strained-silicon CMOS creates fast PMOS (positive channel MOS) transistors that allow for fast, complementary low-leakage digital design in mission critical designs.   But beyond these specialized niches, the semiconductor industry has several times turned to SiGe as a way to increase the amount of digital information a logic circuit can process or a memory device can hold without necessarily reducing transistor size. Because SiGe is a heterojunction structure, theoretically it can generate more than one signal threshold level reliably: not just on or off, but several other intermediate signal levels as well.   Unlike homojunction-based digital circuits, which have only two threshold voltages that can be detected reliably (on or off), SiGe transistors can generate three or four such signals naturally. Several times since the early 1990s, companies and organizations such as Bell Labs, IBM, Intel, Signetics, National Semiconductor, and even Motorola were looking at SiGe and its multithreshold capabilities to create circuits that operated on other forms of logic other than binary. In one paper I remember reading that Motorola's Research Labs had come up with a SiGe CMOS transistor design that could reliably produce three threshold levels. They had also borrowed some techniques from analog circuit design that allowed them to detect and manipulate those signals and create some primitive trinary logic functions as well as a trinary logic memory device.   Theoretically, because SiGe could be used to build devices that move beyond simple 0/1, on-off based binary logic, ICs built using SiGe can reliably generate multiple signal levels that are easily discriminated. These ICs could be used to build 3-base, 4-base, and higher logic functions, effectively increasing a device's information density without further shrinking the transistor structure. Also, the interconnects could be reduced if signals in the circuit assume four or more levels rather than only two. In random access memory, storing two or three values instead of one doubles or triples the information density of the memory without increasing die size.   Each of these forays into multi-threshold, multi-valued logic came at a time when the semiconductor industry was up against the wall on how much further they could push lithography technologies to product affordable integrated circuits. Unfortunately, most of those walls were economic, while advances in photolithography allowed the industry to push transistor sizes down to the nanometer level.   But now the semiconductor industry is up against the wall of physical limits, as well as horrendously expensive financial ones. At IBM's 7nm node, we are not that far away from sub-nanometer devices in the picometer range.   Whether the multi-threshold capabilities of SiGe are ever used, SiGe has a number of other features that may extend silicon-based circuits’ lifetime a bit further. This fact gives the history of semiconductors in today's electronic devices a certain symmetry. If you remember, it was germanium, not silicon that started the semiconductor revolution in 1947. That was when researchers at Bell Laboratories used it to create the first solid-state device, a point-contact transistor. And the first mass consumer electronics devices, transistor-based portable radios, ran on germanium diodes.

硅是上帝送给人类的礼物。电路板中绝大多数器件都采用体硅CMOS工艺（硅的原材料是沙子）制造，但有一个部分却难以实现，那就是射频前端。目前射频前端 主要采用GaAs或SiGe工艺制造，但由于材料的稀缺性和工艺的复杂性，射频前端芯片（RFeIC）良率不高，成本太贵。这阻碍了物联网（IoT）传感 器节点（单价应低于1美元）的普及。 RFaxis是一家专注于射频前端设计的公司，该公司解决了以标准CMOS工艺生产RFeIC的 难题。RFaxis公司产品营销和客户应用工程总监虞强博士表示，以往有很多关于用CMOS技术设计制造射频前端的论文，但是真正实现起来却有很大难 度：1.CMOS的击穿电压较低，难以做到大功率；2.截止频率低即惰性强，要提高功率需选用较厚材料，然而却会增加惰性；3.电导率低，难以把直流转换 成射频信号，这又牵涉到转换效率（功耗）和增益问题。 RFaxis全球销售副总裁Raymond Biagan介绍道，RFaxis生产的RFeIC是高集成度的单芯片、单裸片解决方案。与传统采用分立器件或前端模块（FEM）所开发的RF前端相 比，RFaxis CMOS工艺器件不但大幅降低了成本，系统复杂度及噪声也同时减少。 射频前端的主要功能是连接收发器和天 线，用以增大输出功率，提高接收灵敏度，增加传输速率和距离。传统的RFIC由独立的PA（功率放大器）、LNA（低噪声放大器）、开关（SW）和分立器 件所组成，采用GaAs或SiC工艺将其结合在一个模块里。另外，其集成度远低于纯CMOS工艺产品。 图1：RFaxis用CMOS工艺将PA、LNA和SW集成在单个裸片上。 GaAs或SiC工艺产品无法做到高集成度，通常在一个芯片内包含两个以上裸片，然后通过引线键合连接在一起。而RFaxis的芯片则是完全由一个单裸片所组成，而且这样性能也得到更加优化。 图2：RFaxis纯CMOS工艺单芯片、单裸片RFeIC引领下一代射频前端发展。 RFaxis是唯一一家以纯CMOS工艺生产RFeIC的公司，其产品与竞争对手相比性能相当。而且因为是采用纯CMOS工艺生产，所以无论是工艺还是交货周期都优于GaAs工艺。 在整个半导体行业中，GaAs及SiGe相对于CMOS来讲只是非常小的一个部分。GaAs及SiGe是非常稀缺的资源，无线通信的发展受限于这些资源的稀缺性。这就是为什么最近4G LTE市场爆发，但是竞争对手却无法及时交货的原因。 传 统GaAs工艺的晶圆采用6英寸工艺，其成本远大于1000美元，而RFaxis采用0.18μm的8英寸晶圆的成本却远小于1000美元。这也就是说 RFaxis的晶圆比竞争对手的面积更大，但是价格却更便宜，因此RFaxis单颗裸片成本远小于GaAs工艺。另外，GaAs是特殊工艺，其良率远低于 CMOS工艺。 图3：高通Xb143参考设计用RFX8050/8051取代了Skyworks和RFMD产品。 “我们的产品的优势是：1.成本更低；2.集成度更高；3.超越竞争对手的性能；4.更好的ESD保护性能，因为CMOS工艺的电导性更好，工作温度也更高；5.外部电路更简单，更易于开发，系统成本也更低。”Biagan最后总结道。