博文

[笔记] A Brief History of the MOS transistor, MOS晶体管简史

已有 741 次阅读 2025-5-7 22:33 |个人分类:集成电路（资料）|系统分类:科研笔记

[笔记] A Brief History of the MOS transistor, MOS晶体管简史

Steven Leibson Leibson85R-2005-mid.jpg

图1 Steve Leibson

https://theamphour.com/wp-content/uploads/2012/06/Leibson85R-2005-mid.jpg

https://theamphour.com/the-amp-hour-99-impavid-ideopraxist-insider/

Part 1: Early Visionaries

https://www.eejournal.com/article/a-brief-history-of-the-mos-transistor-part-1-early-visionaries/

In 1939, Shockley wrote: “It has today occurred to me that an amplifier using semiconductors rather than vacuum is in principle possible.” Shockley appears to have been unaware of the earlier patents granted to Lilienfeld and Heil.

1939年，肖克利写道：“今天我突然想到，原则上使用半导体而不是真空的放大器是可能的。”肖克利似乎不知道Lilienfeld和Heil之前获得的专利。

When Bell Labs announced the junction transistor in 1951, it quickly became the dominant transistor in the industry because it was superior in every way to the original point-contact device. However, it was still a bipolar transistor and not a MOSFET.

当贝尔实验室于1951年宣布推出结型晶体管时，它迅速成为行业中的主导晶体管，因为它在各个方面都优于最初的点接触器件。然而，它仍然是双极晶体管，而不是MOSFET。

Significantly, the Bell Labs practice was to remove the silicon dioxide layer after diffusion because it was considered “dirty,” as in “laced with contaminants.” Two months after the Frosch and Derick article appeared, Jean Hoerni at Fairchild Semiconductor realized that the silicon dioxide layer was important for several reasons and that it should be left in place. A pure silicon dioxide layer became an integral part of Hoerni’s planar manufacturing process and would be a key element of IC manufacturing.

值得注意的是，贝尔实验室的做法是在扩散后去除二氧化硅层，因为它被认为是“脏的”，就像“夹杂着污染物”一样。在Frosch和Derick的文章发表两个月后，Fairchild Semiconductor的Jean Hoerni意识到二氧化硅层的重要性有几个原因，应该留在原地。纯二氧化硅层成为Hoerni平面制造工艺不可或缺的一部分，并将成为IC制造的关键要素。

Atalla further refined this discovery into a more formal silicon dioxide passivation technology that permitted the silicon to be doped far more precisely in specific locations when coupled with newly developed photolithographic and etching techniques. Using this technology, Atalla and Kahng managed to build a working MOSFET by the beginning of 1960, three decades after Lilienfeld first conceived of the device. Although the device worked, after a fashion, this first MOSFET had several problems. Notably, it was 100 times slower than contemporary bipolar transistors, mainly because of its relatively large channel length of 20 µm.

Atalla进一步将这一发现改进为一种更正式的二氧化硅钝化技术，当与新开发的光刻和蚀刻技术结合时，该技术允许硅在特定位置更精确地掺杂。使用这项技术，Atalla和Kahong在1960年初成功构建了一个工作的MOSFET，这是Lilienfeld首次构思该器件的三十年后。尽管该器件工作正常，但在某种程度上，第一个MOSFET存在几个问题。值得注意的是，它比当代双极晶体管慢100倍，主要是因为其20µm的相对较大的沟道长度。

https://www.eejournal.com/article/a-brief-history-of-the-mos-transistor-part-1-early-visionaries/

Part 2: Fairchild – The Big Engine that Couldn’t

https://www.eejournal.com/article/a-brief-history-of-the-mos-transistor-part-2-fairchild-the-big-engine-that-couldnt/

Shockley’s autocratic management style and ego fractured his team, causing a showdown on May 29, 1957. The research team’s demand was for the “Shockley Problem” to be solved. It wasn’t, and eight members of Shockley’s research team – which became known as the Traitorous Eight – left in September 1957. That core group cut a deal with Sherman Fairchild and founded Fairchild Semiconductor on October 1, 1957.

肖克利的专制管理风格和自我意识使他的团队四分五裂，导致1957年5月29日的摊牌。研究小组的要求是解决“肖克利问题”。事实并非如此，肖克利研究小组的八名成员——后来被称为叛徒八人组（八叛逆）——于1957年9月离开。该核心小组与Sherman Fairchild达成协议，于1957年10月1日成立了Fairchild Semiconductor。

Bell Labs thought this oxide was too dirty to leave in place but Hoerni realized that a sufficiently clean insulating layer would prevent contamination from dust, dirt, and water. With few changes, Hoerni filed for a patent on the planar process on January 14, 1959.

The second important step towards the semiconductor breakthrough that was needed to allow MOSFETs to achieve their destiny occurred on January 23, 1959. That was the day that Fairchild Semiconductor’s founder Robert Noyce wrote down the ideas for a monolithic integrated circuit in his lab notebook. He’d been spurred to think of ways to use Hoerni’s planar process to make more than discrete transistors. He realized that the silicon dioxide layer was a perfect insulator and allowed metal interconnect to be deposited on top to complete the connections among multiple devices on the IC. With that flash of insight, Noyce changed the electronics industry forever and transformed soldering and wiring into a high-tech printing process.

贝尔实验室认为这种氧化物太脏了，不能留在原地，但霍尼意识到，一个足够干净的绝缘层可以防止灰尘、污垢和水的污染。几乎没有什么变化，霍尼于1959年1月14日申请了平面工艺的专利。

1959年1月23日，MOSFET实现其命运所需的半导体突破的第二个重要步骤发生了。就在那天，飞兆半导体的创始人罗伯特·诺伊斯在他的实验室笔记本上写下了单片集成电路的想法。他一直在思考如何利用霍尼的平面工艺制造出不仅仅是分立晶体管。他意识到二氧化硅层是一种完美的绝缘体，允许在顶部沉积金属互连，以完成IC上多个器件之间的连接。凭借这一瞬间的洞察力，诺伊斯永远改变了电子行业，将焊接和布线转变为高科技印刷工艺。

In February, 1963, Wanlass and his manager C.T. Sah presented a paper at ISSCC that revealed that Wanlass had conceived of circuits that combined p- and n-channel MOSFETs on the same IC. He’d invented CMOS as a mere byproduct of his work.

1963年2月，Wanlass和他的经理C.T.Sah在ISSCC上发表了一篇论文，揭示了Wanlass已经构思出将p沟道和n沟道MOSFET组合在同一IC上的电路。他发明CMOS只是他工作的副产品。

https://www.eejournal.com/article/a-brief-history-of-the-mos-transistor-part-2-fairchild-the-big-engine-that-couldnt/

Part 3: Frank Wanlass – MOS Evangelist, Inventor of CMOS

https://www.eejournal.com/article/a-brief-history-of-the-mos-transistor-part-3-frank-wanlass-mos-evangelist-inventor-of-cmos/

It’s hardly surprising that semiconductor companies were reluctant to invest much energy into MOSFET development in the early 1960s. Early MOSFETs were 100 times slower than bipolar transistors, and they were considered unstable, for good reason: their electrical characteristics drifted badly and unpredictably with time and temperature. A lot of research and development work would be needed to transform MOSFETs into reliable electronic components.

毫不奇怪，在20世纪60年代初，半导体公司不愿意在MOSFET的开发上投入太多精力。早期的MOSFET比双极晶体管慢100倍，并且它们被认为是不稳定的，这是有充分理由的：它们的电特性随着时间和温度的变化而严重漂移，且不可预测。将MOSFET转化为可靠的电子元件需要大量的研究和开发工作。

Along the way, Wanlass and his manager C.T. Sah patented the idea for CMOS circuits, which combine p- and n-channel MOSFETs on one silicon die. CMOS is the foundational transistor technology for nearly every IC now manufactured. (Note: Sah is often listed as the sole inventor of CMOS, but his name is on the patent because he was Wanlass’s manager, and it was customary to list the manager along with the inventor on the patent application.)

在此过程中，Wanlass和他的经理C.T.Sah为CMOS电路的想法申请了专利，该电路将p沟道和n沟道MOSFET结合在一个硅管芯上。CMOS是目前制造的几乎所有IC的基础晶体管技术。（注：Sah通常被列为CMOS的唯一发明人，但他的名字出现在专利上，因为他是Wanlass的经理，通常在专利申请中会将经理和发明人一起列出。）

In an interview, Wanlass said, “I was driving down the 101 freeway one day, on a Sunday, and it hit me. It’s got to be sodium, and I had sent out some aluminum wire to be spectroscopic analyzed. I knew from my college work, some of my thesis work, that sodium… I knew for a fact that sodium, with just a little help from temperature and voltage, would diffuse right through quartz. It had a very high diffusion quotient. I knew that. That was from experiments in college.”

在一次采访中，Wanlass说：“有一天，在一个星期天，我在101高速公路上开车，它撞到了我。它一定是钠，我送了一些铝线进行光谱分析。我从我的大学工作，我的一些论文工作中知道钠……我知道一个事实，即钠在温度和电压的帮助下，会直接通过石英扩散。它有很高的扩散系数。我知道。这是从大学的实验中知道的。”

Eventually, the industry would understand that sodium ion contamination causes drift and kills p-channel MOSFETs over time, while making it impossible to build working n-channel MOSFETs. The aluminum was being contaminated with sodium when it was drawn through wire dies to make the aluminum wire that served as the feed stock for the vapor deposition. The wire dies were lubricated with sodium. Electron-beam vaporization employs a shutter mechanism that shields the wafers from the crucible of molten aluminum until the aluminum reaches its evaporation temperature. Sodium has a much lower boiling point than aluminum, so the sodium boiled off and dissipated before the shutter in the evaporation chamber opened and exposed the wafer to the aluminum vapor.

最终，业界会明白，钠离子污染会导致漂移，并随着时间的推移杀死p沟道MOSFET，同时使其无法构建工作的n沟道MOSFET。当铝被拉过线材模具以制造用作气相沉积原料的铝线时，铝被钠污染了。金属丝模具用钠润滑。电子束蒸发采用一种快门机制，将晶片与熔融铝的坩埚隔开，直到铝达到其蒸发温度。钠的沸点比铝低得多，因此钠在蒸发室中的闸门打开并将晶片暴露于铝蒸气之前沸腾并消散。

https://www.eejournal.com/article/a-brief-history-of-the-mos-transistor-part-3-frank-wanlass-mos-evangelist-inventor-of-cmos/

Part 4: IBM Research, Persistence, and the Technology No One Wanted

https://www.eejournal.com/article/a-brief-history-of-the-mos-transistor-part-4-ibm-research-persistence-and-the-technology-no-one-wanted/

IBM’s bipolar transistor problem was one of reliability. Some of the transistors in its SLT devices would fail due to electrical leakage, and CD’s work on the problem uncovered the culprit and the solution. The culprit was sodium, and the solution was phosphorus. By 1964, IBM had discovered that adding trace amounts of phosphorus to the silicon dioxide layer in the planar process, the phosphorus bonds with the sodium and prevents it from descending into and poisoning the silicon below. A further development determined that applying a negative voltage to the silicon substrate further alleviated the problem of sodium contamination by increasing the MOSFET’s threshold voltages.

IBM的双极晶体管问题是可靠性问题。其SLT器件中的一些晶体管会因漏电而发生故障，CD在这个问题上的工作揭示了罪魁祸首和解决方案。罪魁祸首是钠，溶液是磷。到1964年，IBM发现，在平面工艺中向二氧化硅层中添加微量磷，磷会与钠结合，防止其落入下方的硅中并中毒。进一步的发展确定，通过增加MOSFET的阈值电压，向硅基板施加负电压进一步缓解了钠污染的问题。

IBM’s systems development group in Poughkeepsie then identified the perfect vehicle for MOS ICs: memory. The competing memory technology, magnetic cores, was incredibly slow. Magnetic memory’s access and cycle times were on the order of microseconds. MOS memory ICs would be faster than that. In addition, the use of MOS memory ICs could significantly cut memory’s cost per bit while speeding access times.

IBM位于波基普西的系统开发团队随后确定了MOS IC的完美载体：内存。与之竞争的内存技术，磁芯，速度慢得令人难以置信。磁存储器的访问和循环时间大约为微秒。MOS存储器IC将比这更快。此外，使用MOS存储IC可以显著降低存储器的每比特成本，同时加快访问时间。

Currently, IBM Research has developed a 2nm MOS process technology and is in partnership with Rapidus in Japan to commercialize the process. The history of IBM Research’s involvement with MOSFETs is nearly synonymous with the MOSFET’s history, and IBM Research’s work was clearly instrumental in finding a home for MOSFETs, the technology that no one wanted.

目前，IBM Research已经开发出一种2nm MOS工艺技术，并与日本Rapidus合作将该工艺商业化。IBM Research参与MOSFET的历史几乎与MOSFET的历史同义，IBM Research的工作显然有助于为MOSFET找到一个家，这是一种没有人想要的技术。

https://www.eejournal.com/article/a-brief-history-of-the-mos-transistor-part-4-ibm-research-persistence-and-the-technology-no-one-wanted/

Part 5: RCA – The Persistent CMOS Contrarian

https://www.eejournal.com/article/a-brief-history-of-the-mos-transistor-part-5-rca-the-persistent-cmos-contrarian/

RCA Labs also started developing “integrated semiconductor devices,” ICs, during this period. It appears that RCA Labs was well on its way to developing ICs before Texas Instruments (TI) and Fairchild but apparently failed to build fully functional devices. RCA Labs then took a wrong turn and started looking at Shockley’s unipolar transistor as a candidate for making ICs. That work turned out to be a dead end.

在此期间，RCA实验室也开始开发“集成半导体器件”IC。看起来，RCA实验室在德州仪器（TI）和飞兆半导体（Fairchild）之前就已经在开发IC，但显然未能制造出功能齐全的设备。RCA实验室随后误入歧途，开始将肖克利的单极晶体管视为制造IC的候选者。那项工作最终被证明是一条死胡同。

Throughout the 1970s, RCA was the only major semiconductor company to stick with CMOS, or COSMOS, because CMOS circuits were still slow when compared with both bipolar and NMOS ICs. Most projects called for speed. By the middle of the 1970s, PMOS devices had fallen by the wayside and NMOS was the new MOS king. Ultimately, however, RCA would prove to have been right all along, because CMOS got fast.

在整个20世纪70年代，RCA是唯一一家坚持使用CMOS或COSMOS的大型半导体公司，因为与双极型和NMOS IC相比，CMOS电路仍然很慢。大多数项目都要求速度。到20世纪70年代中期，PMOS器件已经半途而废，NMOS成为新的MOS之王。然而，最终，RCA将被证明一直是正确的，因为CMOS发展得很快。

Hitachi’s Central Research Laboratories made a breakthrough when Yoshio Sakai and Toshiaki Masuhara developed twin-well CMOS, which cured CMOS of its speed handicap. Makimoto and his group adopted the twin-well CMOS process and took aim at Intel’s fastest static RAM (SRAM): the NMOS, 4-kbit 2147 SRAM. Intel’s 2147 memory had a minimum access time of 55 nsec and consumed 115 milliamps.

当酒井义雄和Toshiaki Masuhara开发出双井CMOS时，日立中央研究实验室取得了突破，解决了CMOS的速度障碍。Makimoto和他的团队采用了双井CMOS工艺，并瞄准了英特尔最快的静态RAM（SRAM）：NMOS，4-kbit 2147 SRAM。英特尔的2147存储器的最小访问时间为55纳秒，消耗115毫安。

These two Hitachi CMOS SRAMs were big sellers and they electrified Hitachi’s competitors in the semiconductor industry by proving that low-power CMOS ICs could also be fast. Within a few years, thanks to RCA’s persistence and Hitachi’s innovative process, NMOS joined PMOS on the ever-growing heap of obsolete semiconductor technologies, and CMOS became the king of the hill.

这两款日立CMOS SRAM是畅销产品，它们通过证明低功耗CMOS IC也可以很快，使日立在半导体行业的竞争对手感到振奋。几年内，由于RCA的坚持和日立的创新工艺，NMOS加入了PMOS不断增长的过时半导体技术堆，CMOS成为了王者。

https://www.eejournal.com/article/a-brief-history-of-the-mos-transistor-part-5-rca-the-persistent-cmos-contrarian/

参考资料：

[1] Steven Leibson. 2023-04-03, A Brief History of the MOS transistor, Part 1: Early Visionaries

https://www.eejournal.com/article/a-brief-history-of-the-mos-transistor-part-1-early-visionaries/

[2] Steven Leibson, 2023-04-05, A Brief History of the MOS transistor, Part 2: Fairchild – The Big Engine that Couldn’t

https://www.eejournal.com/article/a-brief-history-of-the-mos-transistor-part-2-fairchild-the-big-engine-that-couldnt/

[3] Steven Leibson, 2023-04-10, A Brief History of the MOS transistor, Part 3: Frank Wanlass – MOS Evangelist, Inventor of CMOS

https://www.eejournal.com/article/a-brief-history-of-the-mos-transistor-part-3-frank-wanlass-mos-evangelist-inventor-of-cmos/

[4] Steven Leibson, 2023-04-17, A Brief History of the MOS transistor, Part 4: IBM Research, Persistence, and the Technology No One Wanted

https://www.eejournal.com/article/a-brief-history-of-the-mos-transistor-part-4-ibm-research-persistence-and-the-technology-no-one-wanted/

[5] Steven Leibson, 2023-04-24, A Brief History of the MOS transistor, Part 5: RCA – The Persistent CMOS Contrarian