Response to Response to the Comment on 'Performance of a Spin Based Insulated Gate Field Effect Transistor

Please download to get full document.

View again

of 8
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Document Description
Response to Response to the Comment on 'Performance of a Spin Based Insulated Gate Field Effect Transistor
Document Share
Document Tags
Document Transcript
  Response to “Response to Comment on ‘Performance of a Spin Based Insulated Gate Field Effect Transistor’” [cond-mat/0607432] S. Bandyopadhyay Department of Electrical and Computer Engineering, Virginia Commonwealth University, Richmond, VA 23284 M. Cahay Department of Electrical and Computer Engineering and Computer Science, University of Cincinnati, Cincinnati, OH 45221 ABSTRACT We show that the arguments in the posting cond-mat/0607432 by Flatté and Hall are flawed and untenable. Their spin based transistor cannot work as claimed because of fundamental scientific  barriers, which cannot be overcome now, or ever. Their device is not likely to work as a transistor at room temperature, let alone outperform the traditional MOSFET, as claimed. 1  Flatté and Hall have proposed a spin-based field effect transistor [1] that works as follows:  Nearly 100% spin polarized electrons are injected from the source region into the channel. If the gate voltage is low, the spin orbit interaction in the channel is weak so that the electrons do not flip spin and arrive at the drain with their spin polarizations mostly intact. The drain is an ideal half-metallic ferromagnet magnetized anti-parallel  to the source. It blocks these electrons from transmitting and the current is nearly zero. When the gate voltage is turned up, the spin orbit interaction in the channel increases in strength and carriers flip their spin before they reach the drain. Therefore, many carriers arriving at the drain have their spins aligned along the direction of the drain’s magnetization. These are transmitted, resulting in a large drain current. Thus, by changing the gate voltage, one can modulate the source-to-drain current and realize transistor action. Flatté and Hall claim that this transistor will work at room temperature, have an on-to-off current ratio of 10 5 , and a lower threshold voltage than a MOSFET, resulting in lower energy dissipation during switching. In [2], we questioned these claims. We believe that the Flatté-Hall device cannot even work as a  proper transistor at room temperature, let alone outperform the traditional CMOS device. Flatté and Hall have recently posted a response to our objections [3]. Here we show that their response contains flawed arguments. As a result, our objections stand. First, let us reiterate our basic objection to the Flatté-Hall device. These authors claim (see [3]) that the on-to-off current ratio in their device is 10 5  at room temperature. We hold that this is impossible and a more realistic ratio is 10, if even that much. Such a low ratio makes the device useless in all mainstream applications. We elucidate this below. 2   Unless the source that injects spin-polarized electrons into the channel of the Flatté-Hall device is an ideal, 100% spin-polarized, half metallic ferromagnet, and unless there is no loss of spin  polarization at the source-channel interface, the spin injection efficiency is less than 100%. Let us assume that it is η  . Then, min majtotal  II  I  η  −=  (1) where  I  maj  is the current due to majority spins,  I  min  is the current due to minority spins and  I  total  =  I  maj + I  min . Therefore,  I  min =  [(1-  η  )/2]  I  total . Assuming (generously) that the drain is a 100% efficient spin selective detector,  I  min will be the leakage current (  I  off  ) that flows through the transistor when it is “off”. When the transistor is “on”, the current that flows is due to spin randomization in the channel (spin flipping). If randomization is complete, then the spin  polarization of the current arriving at the drain is zero. Therefore, at best 50% of the arriving carriers will have their spins aligned along the drain. As a result, the maximum possible value of the on-current is  I  on  = (1/2)  I  total . Consequently, the maximum on-to-off current ratio is 11 onoff   I  I  η  =−  (2) In order for this to equal 10 5 , as claimed by Flatté and Hall, η    ≥  99.999%. Thus, even 99% spin injection efficiency is not enough! The injection efficiency at the source contact has to reach 99.999%. If we consider the fact that the drain cannot be a perfect spin detector either, then the required spin injection efficiency at the source end is even higher. Flatté and Hall do not dispute any of this, but claim in their response [3] that even though such high injection efficiency is not currently achievable, it may become achievable by 2018, because (in their view) there is no 3  fundamental barrier to achieving 100% injection efficiency. Here, we show that there are    fundamental barriers  to ~100% spin injection efficiency, particularly at room temperature. Flatté and Hall have proposed two routes to achieving ~100% spin injection efficiency: (1) the use of 100% spin-polarized half metals as spin injectors, and (2) the use of spin selective  barriers. Unfortunately, there can be no  half metals with 100% spin polarization at any temperature above absolute zero. Ref. [4] has shown that all half metals lose their high degree of spin polarization at temperature T > 0 K because of magnons and phonons. Even at T = 0 K, there are no ideal half metals with 100% spin polarization because of surfaces and inhomogeneities [4]. Therefore, half metals will not achieve ~100% spin injection efficiency, ever  , even at 0 K, let alone room temperature. Consequently, half metals are not a viable route. Spin selective barriers can at best transmit one kind of spin at one specific injection energy. The  best spin selective barriers use resonant tunneling. At 0 K, the transmission energy bandwidth can approach zero, but at any non-zero temperature, thermal broadening of the carrier energy will ensure that the spin injection efficiency is far less than 100%. Therefore, this route will not work either at room temperature, not now and not ever. In summary, there are no known methods to approach 100% spin injection efficiency because there are no half metals at T > 0 K, and there are no perfect spin selective barriers at T > 0 K. Thus, there are fundamental barriers  to achieving ~100 % spin injection efficiency at T > 0 K. There may be other ways of achieving high spin injection efficiency in addition to the two that Flatté and Hall mention (see, for example, [5]), but they too do not work at T > 0 K. Thus, unless 4  revolutionary new ideas for spin injection are found, there is no hope of achieving 100% spin injection efficiency at room temperature either now, or in 2018, or 3018. Simply stated, there is no strategy in sight that can produce, even theoretically, 100% spin injection efficiency at room temperature. Flatté and Hall’s optimism (that somehow ~100% spin injection at room temperature will be possible by 2018) has no scientific rationale. The largest electrical spin injection efficiency demonstrated from a permanent ferromagnet so far is ~70% [6]. We showed in [2] that this results in a current on-to-off ratio of 3.3, not 10 5 . Even if 90% spin injection efficiency is achieved at room temperature by 2018, the resulting current on-to-off ratio will be only 10. That is not sufficient for any mainstream application. Therefore, we  believe that the Flatté-Hall device is not a viable transistor. Today’s CMOS transistors have a current on-to-off ratio exceeding 10 5 , which is unlikely to be ever possible for the Flatté -Hall device. This alone is sufficient to make this device non-competitive with the CMOS device and most other transistor devices currently extant. In [2], we provided a direct comparison between the Flatté-Hall device and a MOSFET. We showed that if the same structure  is used as either a traditional CMOS transistor, or the Flatté-Hall spin transistor, the CMOS version can always have a lower threshold voltage (and therefore win) by lowering the carrier concentration sufficiently 1 . In [3], Flatté and Hall challenge this 1  Flatté and Hall also ignored the voltage drop across the gate insulator in their srcinal calculation. They now address this issue in ref. [3]. They first state that the voltage drop across the oxide in CMOS is 70% of the total. Then they state: “if the same oxide is used for the spin-FET as planned for CMOS, the presence of the oxide changes the total spin-FET capacitance by only ~ 10%, and the voltage drop across the oxide is ~ 10% of the total voltage drop, changing the threshold voltage by ~10%.” We ask: what has capacitance got to do with the voltage drop across the oxide? The total voltage drop is the sum of the drop across the oxide and the transverse drop across the channel plus substrate. This is a simple voltage divider circuit. The drop across the oxide is 5
Similar documents
View more...
Search Related
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks