INVESTIGATION OF TRANSFERRED-ELECTRON AMPLIFIER DIODES WITH A DOPING NOTCH *)

Transcription

1 R948 Philips Res. Repts 3, , 976 INVESTIGATION OF TRANSFERRED-ELECTRON AMPLIFIER DIODES WITH A DOPING NOTCH *) by J. MAGARSHACK, P. HARROP and A. RABIER Abstract A computer model has been developed which simulates both small- and large-signal performances oftransferred- Electron Amplifier (TEA) diodes stabilised by a doping notch. Experimental confirmation of the simulation has been sought by means of devices which have been fabricated to include doping notches of different depths. Measurements have been taken with the device biased in both polarities so that the doping notch appears at either the cathode or the anode ofthe diode, and a comparison is made between the results. Experimental results are reported concerning the small-signal impedance, noise figure, large-signal impedance, maximum added power, efficiency and intermodulation-product performance in both polarities. Indication is given of the large-signal stability and bandwidth of negative resistance encountered in these devices. The results, both of the computer simulation and measurements, suggest that, for cathode-notch devices, more efficient amplifier diodes can be produced by using a shallow notch structure and that this configuration is not optimum for lowest noise figure. Furthermore, a useful intermodulation performance is achieved by using an anode-notch configuration. Introduction Reflection amplifiers using Gunn diodes are finding applications in both narrowband systems at frequencies which are as yet prohibitive to transistor amplifiers (> 2 GHz) and as wideband amplifiers capable of producing ~edium power (> 00 mw at db gain compression).t is the need for amplifiers of the latter category that has formed the basis for the present study. It has previously been established -4) that the form of the diode doping profile plays a role of fundamental importance in determining the characteristics of this type of amplifier. In particular, it has been shown 4) that relatively low noise figures (0.9 db at 2 GHz) can be obtained by tailoring the profile so that a uniform field is established throughout the active layer. This profile is characterized by a doping notch near the cathode. The purpose of this article is to report results of a theoretical and experimental study of Gunn diodes which have doping notches. The influence of the notch, its depths and position at either the anode or the cathode, on *) This work has been supported by Centre National d'etudes et Télécommunications (CNET LANNION).

2 258 J. MAGARSHACK, P. HARROP AND A. RABIER parameters such as maximum added power, efficiency, noise figure and intermodulation have been investigated. The article will be presented in three sections, the first of which will describe the computer model that has been used to simulate diode behaviour under both small- and large-signal conditions, thus enabling the assessment of the efficiencies of diodes of different doping profiles. The second section will include a description of experimental results, at small and large signals, measured with the device biased such that the notch appears either at the cathode or at the anode. These results include measurements of impedance, maximum added power, efficiency and noise figure. The final section will present conclusions which can be drawn from the work and will indicate the type of doping profile which is considered necessary for a specific application whether this be primarily a question of lowest possible noise figure, highest possible efficiency or lowest intermodulation products.. The computer simulation This section will first give a brief account of the classical technique of simulation of this type of active device before analysing the computer results which have been obtained with the doping notch at the cathode and finally with the notch at the anode... Definition and description of the model The method is based on the unidimensional numerical solution of Poisson's equation and basic continuity incorporating appropriate values of vee) and D(E) 5). The model of the device is placed in a resistive circuit and is subjected to a d.c. bias and an alternating signal whose amplitude and frequency are variables. The d.c. bias may either be a constant-voltage source, which is the case in the large signal simulation, or a constant-current source, which is the case in the calculation of the stationary electric field. The two fundamental equations are presented in the following form. be - q = - [n - no(x)], bx e () where no(x) = the donor concentration which corresponds to the doping profile at a position x into the sample, n = electron concentration. be e - = J(E) - bn q n vee) + q D(E) -, bt bx (2)

3 INVESTIGATION OF TEA DIODES WITH A DOPING NOTCH 259 where J(E) = current density in the sample, vee)= electron velocity, D(E) = diffusion coefficient. The analytical representation of vee)which has been adopted 6) is!-loe + vo(e/eo)4 v(e) (E/Eo)4 ' (3) where Eo = threshold field. There remains some doubt as to the exact dependence on electric field of the diffusion coefficient of electrons in GaAs 5.7-). The analytical expression used in this study is based on the calculations of Hammer and Vinter 0) and is the following: Do + Ds[(E - Ep)/Ee]4 D(E) = --+-[-(E--Ep )-/E- -]4-. ' (4) e where Do = 30 cmz/s; D, = 25 cmz/s; Ep = 0 kv/cm; Ee = 5.78 kv/cm. The doping profile is introduced into the calculations through where N F = doping concentration in the :flat region, N; = doping concentration in the notch, Cl = width of notch, Kç, Kz = coefficients which describe the form of the notch, i.e. the slopes of doping in and out of the notch. The equations (3), (4) and (5) are substituted in eqs () and (2) which are solved using the numerical technique of finite differences. Thus information is obtained concerning the static-field distribution, small-signal impedance as a function of frequency and large-signal impedance evolution with increasing applied r.f. power at discrete frequencies of interest. The data obtained in this manner have led to the estimation of the level of third-order intermodulation products since it is possible to deduce the coeffi-, cients of an nth-degree polynomial of the dependence of r.f. current on r.f. voltage. If n then the insertion into this expression of two equal-amplitude voltage waves of frequencies Wl and Wz (5 MHz apart) permits the deduction of the ratio

4 260 J. MAGARSHACK, P. HARROP AND A. RABIER of power in the third-order intermodulation frequencies (2fl - f2 and 2f2 - fl) to that in the fundamental frequencies (fl and f2). So that P3 (~a3 V 3 )2 Pr = (2 al V + a3 V3)2. (6) The knowledge of al and a3 thus enables the ratio P 3 /Pr to be calculated for different values of incident power..2. Simulation of cathode-notch diodes. Influence of ratio of the doping levels in the flat to notch regions (r) It is useful to define a parameter, r, which is the ratio of the doping concentration in the fiat region to that in the notch. The performance ofthree differentdoping profiles (fig. la) has been examined. These three profiles have different static electric-field distributions as shown in fig. lb. Profile () (r = 2.42) presents an electric-field distribution which rises rapidly to a field well in excess of the threshold field at the edge of the notch 0 5 ç ~--:;----, I.38 '0 5 cm- 3 /_. /:;.=:-..::.:-:::':' cm cm-3 =.~2-=--:'::_-:- 3 Fig. la. Doping profiles used in simulations: profile, r = 2.42, profile 2, r =.88. profile 3. r = L(pm) Fig. lb. Computed electric-fields configurations in the diodes for each of the three profiles.

5 INVESTIGATION OF TEA DIODES WITH A DOPING NOTCH 26 after which it decreases and is below threshold at the anode. Profile (2) (r =.88) generates an electric field which rises to kv/cm at the edge of the notch and then remains approximately constant throughout the active layer. The electric field produced by profile (3) (r =.47) rises to 6 kv/cm at the edge of the notch and continues to rise steeply until it has a value of 25 kv/cm at the anode. ~n order that a realistic appraisal of their amplifier efficienciescould be made, the diodes doping concentrations were adjusted so that each sample dissipated the same d.c. power and therefore each sample has the same operating temperature (200 C). Furthermore, the small-signal impedances of the three devices from 7 to GHz, which ate presented in fig. 2a, are seen to be very similar. The large-signal impedances, at 7, 9 and GHz, along with the equivalent circuit used in the calculations, are presented on fig. 2b for each profile. The parameter on the curves is added power which is marked at maximum gain r=2.4.2 r= T r = T.4.7 Fig. 2a. Computed small-signal impedance as a function of frequency for each of the three profiles. é~ L 6coswt ar =2.4.2 o r= T.88 I>.r= T.47 Fig. 2b. Computed large-signal impedance, at 7, 9 and GHz, along with the equivalent circuit used.

6 262 J. MAGARSHACK, P. HARROP AND A. RABIER and at its maximum value. It is clear from the figure that there is generally an expansion of gain into a 50 n load, which occurs before the diode begins to saturate. However, there are several observations which distinguish between the performances of the three simulated devices. Profile () produces the smallest values of maximum added power, and therefore efficiency, throughout the frequency range 7- GHz. It does, however, present no gain expansion at 7 and 9 GHz although at GHz there is gain expansion. Profile (2) produces intermediate values of maximum added power and the evolution of its large-signal impedances takes the form of wide loops corresponding to significant gain expansion into a 50.Q load. Profile (3) generates the largest values of maximum added power and the large-signal impedance loops are generally more closed. Table I presents the maximum added power at 7, 9 and GHz obtained from the three simulated devices. TABLE I frequency r =.47 r =.88 r = GHz 29.3 mw 6 mw 3.2mW 9 GHz 3.7 mw 8.4 mw 7.5mW GHz 93.4 mw mw It is evident from table I that the structure which is capable of being the most efficient is that which has the shallowest notch. In addition, it is interesting to note that the form of the impedance loop presented by the device with smallest r is also that which is most conducive to linear medium power and wideband amplifier design. A circuit could be designed for this device which would present, throughout the band 7- GHz, an impedance which would minimize the effect of gain expansion whilst still taking advantage of the higher added power..3. Simulation of anode-notch diodes The majority of the results which have been obtained in this study have been measured on devices biased so that the notch appeared at the cathode. In the course of the measurements, it became apparent that the device could operate equally well and in certain cases with more interesting results, under reversed polarity, i.e. with the notch at the anode. It was therefore interesting to try to simulate this mode of operation with the existing computer model. This is easily effected by replacing x by (L - x) in expression (5) so that

7 INVESTIGATION OF TEA DIODES WITH A DOPING NOTCH 263 [ ( (L - X»)2] N F - N, [ (L-X- (})] n(x) = Ne exp - Kl + N; + 2 v + tanh K 2 (7) Figure 3 represents the result of this attempt applied specifically to the case of one of the epitaxial layers which was realized (If 823). The figure includes both the small-signal impedance (5.5-8 GHz) and the large-signal impedance evolution at 9 GHz. The static electric-field distribution for this configuration is such that a highfield region exists at the anode with the field exceeding its threshold value approximately halfway along the sample. The large-signal results indicate a much reduced gain-expansion phenomenon and a value of maximum added power which is comparable with the cathodenotch configuration. The field distribution of the anode-notch device is very similar to that of a device with a doping profile which steadily decreases from the cathode to the anode as has been previously reported 2) to give very stable diodes. The evolution of the impedance of such a device as the signal drive increases can be compared qualitatively to the corresponding behaviour of a cathode-notch(constant Jl823 D(E) MIRCEA V=6.4V Fig. 3. Computed small-signal impedance of anode-notch diode and large-signal evolution (added power as parameter) at 9 GHz.

8 264 J. MAGARSHACK, P. HARROP AND A. RABIER field) device by considering the variation of the electric field as it is modulated by the signal. For the uniform-field device (notch at the cathode) as the signal increases, it drives the field into the high negative differential mobility region at each point in the sample. The negative resistance has a tendency to increase therefore and as the drift velocity increases, the reactance changes as well. For the anode high-field device (notch at anode, or sloping down profile) the signal drives the field into the negative-mobility region only in a small part of the sample. In the region near the cathode the signal drive produces a more positive resistance, so that the total resistance of the sample saturates towards a smaller value of the negative resistance. 2. Experimental results Devices have been fabricated with a variety of different doping profiles whose r values range from The diodes were mounted in S4 packages and characterized in 50 Q coaxial support which was water-cooled. The first part of this section will discuss results taken with the notch at the cathode and the second part will describe a similar set of measurements taken from the same devices biased so that the notch appears at the anode. TABLE Principle parameters of the epitaxies epitaxy length of doping cone. doping cone, number active layer in that region in notch r ((LID) (0 4 cm=") (0 4 cm=")

9 INVESTIGATION OF TEA DIODES WITH A DOPING NOTCH Diodes with notch at the cathode Extensive small-signal impedance measurements have been performed on diodes of each of fourteen epitaxies with different doping profiles. It was ob-. served that there was a small dispersion of results within anyone epitaxy but the results which will be presented in table II are considered typical of the epitaxy which they represent. Table II presents the principal parameters of the epitaxies, An example of the measured device impedance is shown in fig. 4. The device is taken from the epitaxy V 823. The negative of the device impedance (i.e. -Zd) is plotted on a positive Smith chart. The device is biased with.5 volts and exhibits a negative resistance from GHz. The evolution of this devices negative resistance at 7, 9 and GHz as a function of applied bias is presented in fig. 5 which.also includes the development of the device impedance as the incident power is increased, for two different bias voltages (9.8 and.5 volts). The manner in which the device impedance changes with increasing incident power is clearly a function of applied bias and takes the form predicted by the computer simulation. Complementary small-signal measurements ~ich show the dependence of bandwidth of negative resistance on applied bias are presented in fig. 6b which shows the small-signal gain (into a 50 Q load) plotted against frequency. These results indicate the tendency for both the negative resistance of the diode and V=.55V I=4.00mA V823/Ta diode T Fig. 4. Measured small-signal impedance as a function of frequency of a diode from the epitaxy V 823.

10 266 J. MAGARSHACK, P. HARROP AND A. RABIER Fig. 5. Variation of small-signal impedance as a function of voltage at 7, 9 and GHz as well as large-signal measurements, for two bias values, at GHz. 25~~ r~.!!! '" o c: 0 - ~----~----~~------~----~o 3 - frequency (GHz) Fig. 6. (a) Noise-figure measurements against frequency for different bias voltages (V 823). (b) Corresponding measurements of small-signal gain against frequency for different bias voltages (V 823).

11 INVESTIGATION OF TEA DIODES WITH A DOPING NOTCH 267 the value of the gain to decrease as the bias is increased. In addition there is a shift in the frequency of maximum gain towards lower frequencies as the bias is increased. Measurements of noise figure which correspond to the measured values of gain mentioned above are presented as fig. 6a. The noise figure exhibits a minimum for each value of applied bias. This minimum occurs at lower frequencies as the bias is increased. For frequencies less than 0 GHz increasing bias is accompanied by decreasing noise figure in accordance with an existing theoretical model 3). The minimum measured value is 0.9 db at 8 GHz for an applied bias of 7.0volts and the noise figure measured over a 4 GHz bandwidth from 7- GHz is 5.0 db ± db for an applied bias of.55 volts. Measurements of noise figure have been carried out on all the epitaxies at 9 GHz under an.5 volts bias and the results are plotted in fig. 7b. This graph gives an indication of the dependence of noise figure on r and consequently on the electric-field profile. The noise figure decreases from 2 db to 4 db as the value of I' increases from.4 to 3.4, indicating that smaller noise figures ensue from structures in which the field profile is most nearly flat from cathode to anode 4). The device with smallest r value (I' =.2), corresponding to the epitaxy V 836, gives a smaller noise figure as a result of its lower flat-region doping concentration rather than as a result of an effect related to r. The large-signal performance of all the epitaxies has been measured, in the same 50 n coaxial mount, in order to compare the results with the computer s x x '-- -_ x t- ~-.!!! -- c: x - :g -- 3 xx -..._ x- --- lu I x x 2 23".:.s 'J 2-E 0 0, r{=n /Nv} Fig. 7. (a) Variation of amplifier efficiency, at 9 GHz, as a function of r. (b) Variation of noise figure (measured at 9 GHz) as a function of r. QJ :t7 IU 9.~ c: 7-5 t

12 268 J. MAGARSHACK. P. HARROP AND A. RABIER Fig. 8. Large-signal measurements comparing two diodes of different r values with added power as parameter. Diode (a) (V 836) has an r value of.2 whilst diode (b) (V 823) has an r value of 3.5. predictions. A comparison has therefore been made of the large-signal impedance loops of two diodes which have r values at the extremes of the epitaxies which have been prepared. The result is shown in fig. 8. In general, this loop phenomenon represents an increase of the diodes negative resistance as the input 'power is increased which occurs before the device saturates. This is probably due to r.f. excursions about the d.c. bias point which tend to increase the average negative differential mobility. The d.c. power supplied to the two diodes whose characteristics are shown in fig. 8 is about 3.7 watts in each case. Diode (a) corresponds to a diode from the series V 836 and diode (b) to a diode from the series V 823. It is apparent that the impedance loops of diode (a) are more closed than those presented by diode (b) which is in broad agreement with the results of the computer model. The parameters which are marked in the curves correspond to the values of added power at the maximum gain and at the maximum value of added power. The values of maximum added power for the two diodes at 9 GHz (midband) are very similar which appears to contradiet the predictions of the model, but it is necessary to point out that diode (a) has a flat-region doping profile (5.5 X 0 4 cm=") which is significantly lower than that of diode (b) (7.5x 0 4 cm=") and would therefore be expected to be less efficient. It may be concluded, therefore, that despite its lower doping concentration diode

13 INVESTIGATION OF TEA DIODES WITH A DOPING NOTCH 269 (a) is as efficient as diode (b) due to its shallower notch, which agrees with the general trend of the computer predictions. With the exception of diode (a) all the devices have flat-region doping concentrations of about 8 X 0 4 cm- 3 and so a comparison of their efficienciesis reasonably valid. This comparison forms the basis of fig. 7a where efficiency at 9 GHz is plotted against r. The amplifier efficiency is defined as P aj max 00% V IPol max 0' where Pal rnax = maximum value of added power, V = d.c. voltage across the diode, lpol max = d.c. current at maximum added power. The efficiency in amplification is seen to increase from 2.5 % to 3.7% as the value of r decreases from 3.4 to.5. There is therefore good agreement between the trends predicted by the computer and the measurements. It should be pointed out that not all devices exhibited well-behaved saturation characteristics and that some devices were subject to an abrupt change of impedance from small-signal to a point corresponding to a higher gain for the same value of incident power. These devices were generally among those of highest efficiency. This phenomenon manifests itself as a "jump" in the amplifier transfer characteristics and can be attributed to the form of the impedance loop 4). Some large-signal parameters of other epitaxies are summarised in table Ill, in which it can be seen that the epitaxy V 830 produced 22.6 mw of maximum added power at 9 GHz with an efficiencyof 3.7%. The intermodulation performances of the devices have been measured by injecting two equal-amplitude signals displaced in frequency by 5 MHz. The power in each of the frequencies f' f2 at the input and at each of the frequencies f' f2' 2f - f2 and 2f2 - f at the output was measured on a spectrum analyser. Figure 9 shows intermodulation measurements on a diode from the epitaxy V 823 at 9.5 GHz. It is interesting to note that the onset of gain expansion occurs at the same value of incident power which produces a nonlinearity on the third-order intermodulation curve. The large-signal behaviour of this device was simulated and the third-order intermodulation product (LM.P.) was calculated as described in the previous section. The results of this calculation are shown in fig. 9, from which it can be seen that the model predicts a larger expansion of gain than measured in practice but that there is excellent agreement between the calculated value of I.M.P. intercept point of 5.8 dbm and measured value of 4.3 dbm. Measurements of LM.P. have been performed on this device, under identical bias conditions, in two circuits of different characteristic impedances namely

14 TABLE III tv Cl Comparative performances of diodes with different notch depths diode N F bias conditions bandwidth noise figure max. efficiency intermodula- noise in oscillaserial no. NN addedpower ti on product tion 0 khz r=- (at 9 GHz) (at 9 GHz) (at 9 GHz) intercept point from carrier (GHz) (db) (mw) co (dbm) (Hzrms/lOOHz) V ~ 385 ma ~.55 V ma / / 3.5.~ ~.55 V ma.54 V ma.55 V ~ 320 ma ~- 6 V ma (i.e. notch at anode) ~ I '" ~ ~~»- ~ ~ e?> ~ til ~

15 INVESTIGATION OF TEA DIODES WITH A DOPING NOTCH T~ 30 Pa 0 a sa a F}n(d8m) Fig. 9. Comparison between experimental and theoretically deduced curves of linearity and 3rd-order I.M.P. at 9.5 GHz. Power values correspond to the power measured at each of the frequencies >2, 2/2 - and 2/ - 2' Here 2 - = 5 MHz. 50 Q and 30 Q. The results are shown in fig. 0. The degree of gain expansion measured into 30 Q is very much larger than that measured into 50 Q and the corresponding departure from linearity on the third-order I.M.P. is more pronounced for the 30 Q measurements. Furthermore the I.M.P. intercept point for the 30 Q case is +0.0 dbm compared with dbm for the 50 Q measurements which confirms that it would be prudent to use a low-gain final stage in a chain of reflection amplifiers in order to improve the overall I.M.P. performance. Table III summarises the I.M.P. performance of some other epitaxies. The best performance was offered by the series V 836 which produces an I.M.P. intercept point, without gain optimization, of dbm for a small-signal gain of 9.5 db into a 50 Q load.

16 272 J. MAGARSHACK. P. HARROP AND A. RABIER 20 Tr823-4 Po (dbm) t 0 o V=5V I=320mA +4.3dBm ,[ o l7n(dbm). Fig. 0. Linearity and 3rd-order I.M.P. measurements (diode V 823) in circuits of 50 and 30!l Diode with notch at the anode The following measurements were performed on the same devices biased such that the notch appeared at the anode. The results of small-signal impedance measurements are shown in fig. for bias conditions of 6 volts and 430 ma. This device (V 823) exhibits negative resistance, under these d.c, conditions, from 6. to 2.8 GHz. Additional data on the dependence of the bandwidth of negative resistance on bias conditions is plotted in fig. 2b which indicates that near-octave bandwidths of negative resistance can be obtained for bias voltages ranging from 2 volts to 8 volts. The centre frequency (approximately the frequency of maximum gain) occurs at lower frequencies as the bias is increased. (The opposite dependence was observed for cathode-notch diodes.) The noise figure (fig. 2a) is generally 3 to 4 db larger for this polarisation than for the cathode-notch case. However, a noise figure of 8.0 db ±.5 db can be obtained (V = 6.0 volts) which is associated with reasonable gain over the frequency range 7- GHz.

17 INVESTIGATION OF TEA DIODES WITH A DOPING NOTCH V=6V [;:432mA V823/3 diode 273 Fig.. Small-signal impedance measurements of a diode (V 823) biased so that the notch appears at the anode, as well as large-signal measurements at 7, 9 and GHz with added power as parameter. ëo ~ , ~ lij.~ oc: 3V 4.V 5V 6V ëo ~.s 5 g, 2.5! Fig. 2. (a) Noise-figure variation with frequency for different values of applied bias (diode V 823). Notch at the anode. (b) Corresponding values of small-signal gain.

18 274 J. MAGARSHACK, P. HARROP AND A. RABIER Before describing the large-signal results it is worth noting that the devices are not thermally optimized to operate in this sense and that the thermal resistance is larger when the notch is at the anode. Despite this fact, however, the results are very interesting. Large-signal impedance measurements are shown in fig. at 7, 9 and GHz where the parameter on the curves is maximum added power. It is clear by inspection of the figure that there is no loop phenomenon and therefore no gain expansion whatsoever under these conditions. The low efficiency of % which is measured at 9 GHz can be considered to be due to the low value of effective nz-product since the high-field region of this device is centred on the low-doped notch region. The disagreement which exists between the computer model and measurements may well have its origin in the non-equivalent thermal conditions. A comparison of the intermodulation performances has been made in figs 3 and 4 for a diode from the epitaxy V 823 biased with the notch at the cathode and the notch at the anode respectively. In order to effect a meaningful comparison, the small-signal gains have been chosen equal in the two cases. The I.M.P. intercept points are dbm and dbm respectively. The com- 20 'Jl823-t Po (dbm) to 0 -to V=tt.55V I=322mA I / +t4.5dbm-;f/ I I I I I la 20 --_ fin (dbm) Fig. 3. Linearity and 3rd-order I.M.P. (diode V 823) at 9 GHz. Notch at the cathode.

19 INVESTIGATION OF TEA DIODES WITH A DOPING NOTCH Jl823-T Po (dbm) i 0 0 V=-T6V I=-4.45mA -0-4) o TO 20 -F}n(dBm) Fig. 4. Linearity and 3rd-order I.M.P. (diode V 823) at 9 GHz. Notch at the anode. parison reveals that an anode-notch structure may prove interesting for certain applications where low intermodulation products are of prime importance. Figure 5 shows a similar characteristic for a diode from another epitaxy (V 836) which gives rise to an I.M.P. intercept point of dbm with an associated small-signal gain of 4 db. Conclusions It has been shown that a noise figure of 5 db ± db is possible in a 4 GHz bandwidth from 7- GHz and that a noise figure of db can be obtained in a narrow bandwidth. This performance was obtained from a device with a, deep cathode notch. The results on several epitaxies of different doping profiles show that shallow notch devices have larger noise figures as a result of an electric-field distribution which departs from the constant profile desirable for low noise figure. Large-signal impedance loops have been encountered for cathode-notch diodes, both theoretically and experimentally, which, for conditions of equal d.c. power, are more closed when the notch is shallow. This may be interesting in the design of broadband linear amplifiers. _ ~~_~...

20 276 J. MAGARSHACK. P. HARROP AND A. RABIER 20 Jl836/3 -Ia V=-6V I =-355mA +25dBm-// /,, //, /.t / , Ia o la 20 --"fin (dbm) Fig. IS. Linearity and 3rd-order I.M.P. at 9 GHz of V 836 diode. Notch at the anode. Experimental and theoretical evidence has been presented which indicates that shallow notch devices have higher efficiencies than deep notch devices. The required doping profile for low noise is thus incompatible with that required for high efficiency. Devices have been produced which are capable of producing more than 200 mw maximum added power. It has been demonstrated that reversing the polarity of a diode such that the notch appears at the anode, may result in improved intermodulation performance and under these, conditions an intermodulation intercept point of dbm has been obtained. This study has demonstrated the feasibility of wideband medium-power reflection amplifiers using Gunn diodes. It is reasonable to expect that devices with similar characteristics can be produced which would be used in amplifiers up to frequencies around 40 GHz. Laboratoires d' Electronique et de Physique appliquée Limeil-Brévannes, March 976

21 INVESTIGATION OF TEA DIODES WITH A DOPING NOTCH 277 REFERENCES ) R. CharIton, V. R. Freeman and G. S. Hobson, Electron. Lett, 7,575,97. 2) R. Spitalnik, M. P. Shaw, A. Rabier and J. Magarshack, Appl, Phys, Lett. 22, 62, ) R. CharIton and G. Hobson, IEEE Trans. ED-2, 652,974. 4) J. Magarshack, A. Rabier and R. Spitalnik, IEEE Trans. ED-2, 652, ) R. Spitalnik, IEEE Trans. ED-23, 58, ) P. N. Butcher, W. Fawcett and N. R. Ogg, Br. J. appl. Phys. 8, 755, ) J. A. Copeland and S. Knight, Semiconductors and semimetals 7A, R. K. Willardson and A. C. Beer (eds), Acad. Press, New York, 97, pp ) W. Fawcett and H. D. Rees, Phys. Lett. 29A, 578, ) J. G. Ruch and G. S. Kino, Phys. Rev. 74, 92, ) C. Hammar and B. Vinter, Electron. Lett. 9, 9, 973. ) B. Kramer and A. Mircea, Appl. Phys. Lett, 26, 623, ) J. Magarshack and A. Mircea, Proc. 8th MOGA Conference, 970, Amsterdam, pp ) J. E. Sitch and P. N. Robson, 4th European Microwave Conference, 974, Montreux paper B.42.