Kalna, K. and Asenov, A. (1) Multiple delta doping in aggressively scaled PHEMTs. In, Ryssel, H. and Wachutka, G. and Grunbacher, H., Eds. Solid-State Device Research Conference, 11-13 September 1, pages pp. 319-3, Nuremburg, Germany. http://eprints.gla.ac.uk/31/ Glasgow eprints Service http://eprints.gla.ac.uk
Abstract Multiple delta doping in aggressively scaled PHEMTs K. Kalna and A. Asenov Device Modelling Group, Dept. of Electronics & Elec. Eng. University of Glasgow, Glasgow G1 LT Scotland, United Kingdom E-mail:kalna@elec.gla.ac.uk The introduction of a second delta doping layer into the structure of aggressively scaled pseudomorphic high electron mobility transistors can dramatically enhance their performance. The results obtained from Monte Carlo device simulations show significant improvement in the device linearity when a second delta doping layer is placed below the channel. The positioning of the additional delta doping layer above the channel near the gate results in an increase in the transconductance of about %. In both cases the introduction of the second delta doping increases the current and hence the delivered power. 1. Introduction Scaling of pseudomorphic high electron mobility transistors (PHEMTs) with a low indium content channel into deep decanano dimensions can dramatically improve their performance (, and )[1]. However, a decrease in the carrier density observed in the channel, when the gate-tochannel separation is reduced in the scaling process, has a detrimental effect on the device current, and consequently, on the device linearity. The decrease in the current can be compensated with a second # % ) -. # % 5 % 9 % ) - % 5 =.. % ) = @. 9 ) D F = H % 5 # ).. @ % =. 9 ) @ = L N @ % 5 # 5 = ) ) @ F N H % 5 # ).. @ % =. 9 ) = N % 5 ) = @. 9 ) @ = L ) = % % 5 = % delta doping layer introduced into the device structure []. The concentration and the position of the additional doping layer have to be carefully chosen to provide the desired effect. Monte Carlo (MC) device simulations [3] have been employed to investigate the impact of the second delta doping layer on the performance of PHEMTs with gate lengths of 1, 7, 5, and 3 nm.. Double delta doped structures The PHEMT under investigation has a typical T-shaped gate and an indium content of [ \ ] in the InGaAs channel (see Fig. 1) [3].
¹ Conduction Band [ev]..... (a) x1 1 cm - 1x1 1 cm - x1 1 cm - 3x1 1 cm - (b) (c) -. Vertical Distance [nm] Second delta Second delta 5x1 1 x1 1 3x1 1 x1 1 1x1 1 Density [cm -3 ] ^ _ ` a b c d e g h c j k l m a j n _ k l o q l m r b k t u c q l m j q b b _ c b m c l x _ n y q x { a l j n _ k l x k { n h c m c r n h _ l n h c ~ d l ƒ g _ n h q x _ l ` u c m c u n q m k r _ l ` Š q Œ _ n h q l q m m _ n _ k l q u u q y c b o c u k n h c j h q l l c u Š o Œ k b q o k c n h c j h q l l c u Š j e The active doping concentration of the standard delta doping layer is assumed to be š -. In order to increase the carrier density in the channel, a second delta doping layer can be placed either below the channel [Fig. 1(a)] or above the original delta doping near the gate [Fig. 1(b)]. The conduction band profile of the original 1 nm single doped PHEMT [Fig. (a)] is compared with the profiles of the double doped structures when the second delta doping layer is placed below the channel [Fig. (b)] and above the channel [Fig. (c)]. Fig. also shows that the carrier densities in the double doped structures are much higher compared to the single doped one. š The carrier sheet density of œ œ - in the single doped 1 nm PHEMT increases to Ÿ š - in the double doped one if the concentration of the second doping is š -. This increase is even more dramatic for the 7 nm device where the carrier sheet density of š - rises to š - in the double doped device. Note that the carrier sheet density of the single doped structure calculated from the self-consistent Poisson-Schrödinger so- ² µ µ ³ ª ± ª ª«Intrinsic single doped Intrinsic double doped doped with external R Experimental data -.5 -. -1.5-1. -.5..5 1 ^ _ ` a b c º e» ½ À  j h q b q j n c b _ x n _ j x Š x y o k u x q l m n b q l x j k l m a j n q l j c x Š u _ l c x k { n h c ~ d l ` q n c u c l ` n h ƒ g x e ^ a u u j _ b j u c x q b c { k b n h c x _ l ` u c m k r c m m c _ j c q l m k r c l j _ b j u c x q b c { k b n h c m k a o u c m k r c m k l c e g h c x _ a u q n c m m q n q _ n h c É n c b l q u b c x _ x n q l j c x _ l j u a m c m ` _ c l o y k r c l x Ê a q b c x q b c j k r q b c m n k n h c c q x a b c c l n x o y j b k x x c x e lution is precisely calibrated against measured density in material grown at Glasgow and used in the fabrication of the referenced 1 nm PHEMT. The simulation study have been carried out using our finite element MC device simulator MC/HF [1, 3]. To carry out a reliable study, the whole simulator has been meticulously calibrated against experimental I ½ -V ½ characteristics obtained from the real 1 nm gate length PHEMT [1]. The calibration procedure requires to include the contact resistances into intrinsic characteristics simulated by the MC/HF in a postprocessing stage [5]. Figure 3 demonstrates that the drain current increases twice in the 1 nm gate length double doped PHEMT in comparison with the single doped. The double doped device consists the second delta doping layer with an active concentration of š - which is placed below the channel and separated by the nm GaAs spacer. Although the maximum transconductance of the sin-
1 1 Double, n=1x1 1 cm - Double, n=x1 1 cm - Double, n=3x1 1 cm - -.5 -. -1.5-1. -.5..5 Þ ß à á â ã ä æ ç è ê â ß è ë ß ì ç í î ï ð ì ñ ò â ò ì ê ã â ß ë ê ß ì ë ê ñ ã ü ý è õ à ò ê ã ø ã è à ê ñ ú ã ÿ ß ì ã ò ê ï í ï æ ñ ã ú á ö ø ã ú ã ú ñ ò ë ê ñ ã ë ã ì è ú ú ã ø ê ò ú ß è à ø ò ì ã ú ö ã ø ê ñ ã ì ñ ò è è ã ø æ 1 1 1 1 1 Double, n=1x1 1 cm - Double, n=x1 1 cm - Double, n=3x1 1 cm - -.5 -. -1.5-1. -.5..5 Þ ß à á â ã æ ç è ê â ß è ë ß ì ç í î ï ð ì ñ ò â ò ì ê ã â ß ë ê ß ì ë ê ñ ã ý è õ à ò ê ã ø ã è à ê ñ ë ò ê ï í ï # ß ê ñ ê ñ ã ë ã ì è ú ú ã ø ê ò ú ß è à ø ò ô ã â ø ò ì ã ú ö ã ø ê ñ ã ì ñ ò è è ã ø æ 1 1 1 1 Double, n=1x1 1 cm - Double, n=x1 1 cm - -.5 -. -1.5-1. -.5..5 Þ ß à á â ã æ ç è ê â ß è ë ß ì ç í î ï ð ì ñ ò â ò ì ê ã â ß ë ê ß ì ë ê ñ ã ú ã ÿ ß ì ã ß è Þ ß à æ ä ö á ê ê ñ ã ú á ö ø ã ú ã ú ñ ò ë ê ñ ã ë ã ì è ú ú ã ø ê ò ú ß è à ø ò ô ã â ø ò ì ã ú ò ö ÿ ã ê ñ ã â ß à ß è ò ø ú ã ø ê ò ú ß è à ø ò ô ã â æ 1 1 1 1 1 Double, n=1x1 1 cm - Double, n=x1 1 cm - -.5 -. -1.5-1. -.5..5 Þ ß à á â ã ü æ ç è ê â ß è ë ß ì ç í î ï ð ì ñ ò â ò ì ê ã â ß ë ê ß ì ë ê ñ ã ý è õ à ò ê ã ø ã è à ê ñ ú ã ÿ ß ì ã ò à ò ß è ò ê ï í ï ñ ã è ê ñ ã ë ã ì è ú ú ã ø ê ò ú ß è à ø ò ô ã â ß ë ø ò ì ã ú ò ö ÿ ã ê ñ ã â ß à ß è ò ø ú ã ø ê ò ú ß è à ø ò ô ã â æ 1 gle and the double doped devices are similar the double delta doping broadens and improves the device linearity. I í -V ð characteristics and transconductances at a fixed drain voltage of VareshowninFigs. and for the 7 nm and 5 nm gate length PHEMTs, respectively, with additional delta doping behind the channel. Figs. 5 and 7 provide analogous informations for 7 and 5 nm devices with the second delta doping above the channel. The current in the doub- le delta doped PHEMTs continuously increases with the increasing of doping concentration in the second delta layer. If the additional doping is placed below the channel, the transconductance peak broadens up resulting in the large improvement in linearity of the device, although its maximum values remain close to the maximum transconductance in the single doped structure. An effect of the second delta doping layer above the original delta doping is rather dif-
.. < :;,,, 997* 5- ) /1 )3., +,- ()* 1 1 (a) < :;,,, 997* 5- ) /1 )3., +,- ()* 1 3 1 3 Delta doping concentration [x 1 1 cm - ] 1 1 1 (b) = >? A C E G H I K L > N K P Q R C K U V X P U Z A X R K U X E V _ E C ` V A V R c E V E X P U Z Z E e R K Z P h > U? X P U X E U R C K R > P U Q P C K U > U R C > U V > X Z E _ > X E p K r K U Z s > R c E L R E C ` U K e C E V > V R K U X E V > U X e A Z E Z p z r H { } A K C E V K C E Q P C R c E U N Z > K N P U Z V K C E Q P C R c E U N X > C X e E V K C E Q P C R c E Œ U N K U Z R C > K U? e E V K C E Q P C R c E U N? K R E e E U? R c I H c E V E X ` P U Z Z E e R K Z P h > U? X P U X E U R C K R > P U E } A K e R P š E C P C E h C E V E U R V K V > U? e E Z P h E Z I H ferent. The increase in the current shown in Figs. 5 and 7 is not as large as in the devices with delta doping below the channel [Figs. and ] but the maxima of transconductances are much larger (about 5% for the 1 nm, % for the 7 nm, and 13% for the 5 nm gate length device) than the maximum transconductance of the single doped PHEMT. This improvement in the device performance is illustrated in Fig. for a set of the scaled devices. The most pronounced improvement in transconductance appears for the PHEMT with a gate length of 7 nm. The concentration used for the second delta doping does not affect the transconductance improvement except for the PHEMT with a gate length of 3 nm. When external resistances, which have been assumed to be the same as for the 1 nm single doped PHEMT, are included into calculations the magnitude of the transconductances reduce but the relative scale of improvement remains the same. 3. Conclusion The use of a second delta doping below the channel in the scaling of PHEMTs improves the device linearity. The second delta doping above the original one, and placed near to the gate, can improve the transconductance by about 5% in the 1 nm gate length PHEMTs to nearly % in the 7 nm ones. At the 5 nm and 3 nm gate lengths the increase in the transconductance is less pronounced due to the small distance between the additional and the original delta dopings but an improvement is still observable. In all double doped devices the current, and hence the available power from the PHEMT, increases substantially. [1] K. Kalna, S. Roy, A. Asenov, K. Elgaid, and I. Thayne, "RF analysis of aggressively scaled phemts", in Proceedings of ESS- DERC, ed. by W. A. Lane, G. M. Crean, F. A. McCabe and H. Grünbacher () 15-159. [] K. Y. Hur, K. T. Hetzler, R. A. McTaggart,D.W.Vye,P.J.Lemonias,andW.E. Hoke, "Ultralinear double pulse doped AlInAs/GaInAs/InP HEMTs", Electron. Lett. 3 (199) 151-151. [3] S. Babiker, A. Asenov, J. R. Barker, and S. P. Beaumont, "Finite element Monte Carlo simulation of recess gate compound FETs", Solid-State Electron. 39 (199) 9-35. [] Ch. Köpf, H. Kosina, and S. Selberherr, "Physical models for strained and relaxed GaInAs alloys: band structure and low-field transport", Solid-State Electron. 1 (1997) 1139-115. [5] S. Babiker, A. Asenov, N. Cameron, and S. P. Beaumont, "A simple approach to include external resistances in the Monte Carlo simulation of MESFETs and HEMTs", IEEE Trans. Electron Devices 3 (199) 3-3.