CORRESPONDENCE

Comments on "Complementary MOS-Bipolar Transistor Structure"'

There is a substantial difference between the lateral n-p-n described in the above paper' and the lateral p-n-p used in con- ventional integrated circuits that has appar- ently been ignored by the authors. Conven- tional lateral p-n-p's have an n+ diffusion underneath the emitter that suppresses in- jection from the emitter into the substrate. There is no equivalent diffusion in the lat- eral n-p-n shown in the paper. Therefore, the n-p-n is represented by the four-termi- nal structure shown in Fig. 1, where Q1 is the lateral n-p-n and Q s is a parasitic, vertical n-p-n. Under normal conditions the collector current of the parasitic transistor will be many times larger than the collector current of the lateral transistor. This se- verely limits the usefulness of the device in practical designs.

V (SUBSTRATE) +

Fig. 1. Equivalent circuit of a lateral n-p-n showing the parasitic, vertical transistor.

Conceptually it is possible to put a p+ diffusion underneath the lateral structure to solve this problem. However, this could well give rise to fabrication problems; there was no mention in the paper that it was or could be done.

ROBERT J. WIDLAR National Semiconductor Corp. Santa Clara, Calif. 95051

Manuscript received January 28. 1970. 1 H. C. Lin, J . C. Ho. R. R. Iyer. and K. Knong,

I E E E Trans. Elecfro!z Deuicep. vol. ED-16. DD. 945- 951. November 1969.

Further Comments on "Comple- mentary MOS-Bipolar Transistor Structureyf1

I t is true that the current collected by the substrate is many times larger than that collected by the lateral transistor, as clearly shown in Table I of the above-mentioned paper.' However, this substrate current does not affect the characteristics of the lateral transistor as evidenced in Fig. 7. In switch-

Manuscript received March 31, 1970. 1 I. C. Lin, J. C. Ho, R. R. Iyer, and K. Kmong.

I E E E Trans . Electron Devireo, vol. ED-16, pp. 945- 951, November 1969.

ing circuits for which the majority of CMOS devices are intended, the large substrate cur- rent merely means larger current spikes dur- ing the transient period.

H. C. LIN Advanced Technology Labs. IVestinghouse Electric Corp. Baltimore, Md. 21203

Millimeter-Wave Mode Conversion by a Solid-state Magnetoplasma

Abstract-A magnetoplasma such as that formed in n-type indium antimonide c a n e5ciently transduce energy from one waveguide mode to another. This correspon- dence reports observations of magneto- plasma mode conversion from the TEolD mode to the TMolo mode at 70 GHz. The maximum conversion efficiency is calculated to be 95 percent under conditions of the present experiment.

Solid-state magnetoplasma devices have recently been proposed as phase shifters, isolators, filters, and circulators for use a t millimeter and submillimeter wavelengths [I]-[3]. IYe herewith report observations of phenomena indicating that magnetoplasmas can also be utilized to efficiently transduce millimeter and submillimeter energy from one waveguide mode to another. Present microwave devices which accomplish this function are mechanical in nature and are thus subject to wavelength limitations im- posed by dimensional tolerances. Besides circumventing such limitations, magneto- plasma mode-conversion devices have the additional advantage of being magnetically controllable. In the present correspondence, we consider one-port conversion from the circular TEol" mode to the circular TMolo mode. The basic principle is, however, also applicable to two-port conversion and to several other choices of waveguide geometry and mode-pair.

Consider a circular waveguide containing an axially magnetized disk of n-type indium antimonide (InSb) placed next to a short circuit. An incident TEolO-mode wave excites axially symmetric "magnetoplasma modes" in the InSb-filled section which, in turn, excite a variety of reflected axially symmet- ric normal modes in the empty waveguide. Of these, if the waveguide radius is properly chosen, only the TEolOand TMol" modes will propagate. Thus, the sum of energies in the returning TEol" and TLIol" modes, plus the energy lost in the sample, must equal the energy contained in the incident TEo1' wave.

Factors affecting the proportion of re- turning TEoIO- and TMolO-mode energy can be best understood by considering the char- acter of the axially symmetric "magneto- plasma modes." Recent theoretical work [4] has shown that transverse E-fields of these

Manuscript received January 21, 1970. This work was supported by the National Science Foundation and by the Air Force Office of Scientific Research.

Fig. 1. lransverse E-field configuration of either the "TEo~"-limit" or "TMolO-limit" mode of the mag- netoplasma-filled waveguide as a function 8 = (ot -82).

( w t ) . 0 \

( w t l = n /

- 2 r -Jn - w -2 0 (&Zl 2 2

Fig, 2. Standi,pg-wave pattern produced by a single "twist mode incident upon a short circuit a t 5 -0.

modes consist of circular and radial func- tions propagating together while being separated by 90 degrees in both time and space phase. Field patterns of either of the two lowest order modes ("TEol"-limit" or "TMol"-limit") as function 0 = (ut-@) are shown in Fig. 1. One sees that transverse E-fields spiral between a circular configura- tion and a radial configuration as the mode propagates. Because of this unique wave motion, we refer to the axially symmetric "magnetoplasma modes" as "twist" modes

Fig. 2 shows the standing-wave pattern produced by a single "twist" mode incident on a short circuit at z = O . At every cross- sectional plane, transverse E-fields alter- nately assume circular and radial configura-

[5 I.