真乄科技業的頂尖投資團隊

美國麻省理工學院(MIT)的研究人員最近發明了一種用於大規模生產昂貴電路的新型「複製/貼上」(copy/paste)方法。這項技術是在製造上覆石墨稀的「供體」(donor)晶圓後，採用「沉積－剝離」(複製/貼上)的方式，從而降低電路與下層晶圓的成本。

該技術能有助於製造商以較簡單且低成本的方式，結合矽(Si)以及諸如用於電晶體管通道的砷化鎵(GaAs)等昂貴材料。

麻省理工學院教授Jeehwan Kim表示，「我們利用了石墨烯的強度及其潤滑的特質，而非其電氣性能，因而可製造昂貴材料的極薄電路，從而大幅降低使用特殊材料的總成本。」

例如，目前，英特爾(Intel)、IBM等許多製造商正致力於矽晶圓上生長GaAs電晶體通道，但由於矽與GaAs之間的晶格無法匹配，終究無法達到高品質的結果。然而，據Kim表示，採用遠端外延的方式，能夠分別從石墨烯頂部供體晶圓剝離的GaAs薄層上製造這些GaAs通道，並接合於矽晶圓上，而不至於造成任何晶格不匹配的問題。

研究人員利用僅沉積單原子層厚的石墨烯單層方式，形成了供體晶圓。由於石墨烯接合至供體晶圓頂部且「可滑動」，因而可在此堆疊頂部製造昂貴的材料薄層，而又不至於黏在該供體晶圓上。Kim的研究團隊再從供體晶圓上簡單地剝離昂貴的頂部半導體層，然後移植到低廉的基板(例如矽基板)上。這種低成本的基板甚至可用晶片中的源極與汲極製圖，然後進行蝕刻而將GaAs通道添加到其他的CMOS電路。由於不存在晶格不匹配的問題，所取得具有GaAs通道的矽晶電晶體將超越今所用的任何矽通道電晶體。

研究人員在石墨烯上生長發光二極體(LED)，剝離後再放置於另一基板上 （來源：MIT）

由於石墨烯比鋼強600倍，而且能與下方供體晶圓緊密地接合，使得這一過程得以不斷地重複，如同用於大規模生產橡膠車輪的鋼模具，有助於石墨烯大量生產結合特殊材料的CMOS晶圓。這些特殊材料本身還可用於LED、太陽能電池、高功率電晶體或其他元件。研究團隊宣稱，單供體晶圓採用這種「複製-貼上」的方式目前還沒有次數的限制。

該研究團隊已經實驗證實了採用特殊材料降低成本的可行性，包括GaAs、磷化銦(InP)和磷化鎵(GaP)等特殊材料，他們還將特殊材料的主動層移植到低成本的軟性基板上，成功地製造出LED。

從矽晶圓上剝離鎳薄膜，展現使用2D材料轉換晶圓製程的概念 （來源：Jose-Luis Olivares / MIT）

接下來，研究人員計劃嘗試用於服裝和其他穿戴式材料上的技術，並試圖堆疊多層以創造更複雜的元件。

LAKE WALES, Fla. -- A new copy/paste method for mass production of expensive circuits has just been invented at the Massachusetts Institute of Technology (MIT). After the fabrication of a "donor" wafer topped by graphene, the technique uses a deposit-and-peel-off (copy/paste) method that reduces the cost of the circuitry and makes the cost of the underlying wafer relatively insignificant. The technique could encourage manufacturers to combine silicon (Si) with expensive materials, like gallium arsenide (GaAs) for transistor channels, very easily and inexpensively.

(left to right): Post doc Kyusang Lee, professor Jeehwan Kim (sitting), and graduate students Samuel Cruz and Yunjo Kim. 
(Source: Jose-Luis Olivares/MIT)

"We are taking advantage of the strength and slippery qualities of graphene, rather than its electrical properties, in order to fabricate extremely thin circuits of expensive materials, thereby greatly reducing the overall cost of using exotic materials," professor Jeehwan Kim told EE Times in an exclusive interview before the public announcement of his "remote epitaxy" technology.

Today many semiconductor researchers are struggling with, for instance, growing GaAs transistor channels on silicon wafers for transistor channels, ending up with low quality results because of the lattice mismatch between silicon and GaAs, according to Kim. However, using remote epitaxy, those GaAs channels can be separately fabricated on a thin layer of GaAs peeled off the graphene-topped donor wafer and bonded to the silicon wafer without any problems with lattice mismatch.

The researchers fashioned the donor wafers by merely depositing a single-atom thick monolayer of graphene. Because graphene bonded to the donor wafer, but is “slippery” on top, a thin layer of an expensive material can be fabricated atop the stack which does not stick to the donor wafer. Kim's group members then simply peel the expensive top semiconducting layer from the donor wafer and transfer it to an inexpensive substrate, such as Si. The substrate can even be patterned already with the source and drain in silicon, then etch to add the GaAs channels to its otherwise CMOS circuitry. And with no lattice mismatch problems, the resulting silicon transistors with GaAs channels will outperform any silicon channel transistors used today.

Light emitting diodes (LEDs) being grown on graphene and then peeled off for placement on another substrate. 
(Source: MIT)

Because graphene is 600 times stronger than steel and bonds strongly to the underlying donor wafer, the process can be repeated over-and-over, like the steel mold used to mass produce rubber car tires, giving graphene a leg-up on the mass production of CMOS wafers combined with exotic materials. The exotic materials can also be used by themselves for LEDs, solar cells, high-power transistors or other devices. The team claims not to have yet found a limit on how many copy-and-paste operations can be performed by a single donor wafer.

Kim's team perfected the method at MIT’s Research Laboratory of Electronics with the help of doctoral candidate Yunjo Kim, graduate students Samuel Cruz, Babatunde Alawonde, Chris Heidelberger, Yi Song, and Kuan Qiao, postdoctoral researchers Kyusang Lee, Shinhyun Choi, and Wei Kong, visiting research scholar Chanyeol Choi, MIT Professor of Materials Science and Engineering Eugene Fitzgerald, professor of electrical engineering and computer science Jing Kong, and assistant professor of mechanical engineering Alexie Kolpak, along with contributors from other universities including Jared Johnson and Jinwoo Hwang from Ohio State University, and Ibraheem Almansouri of Masdar Institute of Science and Technology.

The team has proven the feasibility experimentally of lowering the cost of using exotic materials, including GaAs, indium phosphide, and gallium phosphide. They have also transferred the exotic active layers onto cheap flexible substrates, successfully fabricating light-emitting diodes there.

Next they plan to try the technique on clothing and other wearable materials, as well as attempt to stack multiple layers to create more complicated devices.

Funding was provided by the One to One Joint Research Project between the MI/MIT Cooperative Program and LG electronics R&D center.

— R. Colin Johnson, Advanced Technology Editor

編譯：Susan Hong

(參考原文：New Method Cuts Cost of GaAs Circuits，by R. Colin Johnson)