IEEE Copyright

Donald Sawdai, Xiangkun Zhang, Dimitris Pavlidis, and Pallab Bhattacharya
Department of Electrical Engineering and Computer Science
The University of Michigan
1301 Beal Avenue, Ann Arbor, MI 48109 USA

1. Introduction

InP-based HBTs have demonstrated excellent frequency performance and extremely good current-handling capability. Research on InP-based HBTs has almost exclusively focused on NPN HBTs since the electron velocity is several orders of magnitude higher than the hole velocity in InGaAs. PNP HBTs are also of interest, primarily for integration in circuits with NPN HBTs. Together, NPN and PNP HBTs can form a simple, efficient, and linear Class B power amplifier or output buffer. Such an AlGaAs/GaAs integrated NPN/PNP push-pull amplifier has demonstrated power performance at 2.5 GHz [3]. Multiple-stage amplifiers could be designed with alternating NPN and PNP stages, allowing for simple designs with a single power supply and direct coupling between stages, as demonstrated in a complementary AlGaAs/GaAs Gilbert cell [4]. PNP HBTs could also be used as active loads and current sources for NPN amplifier stages, which would provide higher gain per stage, reduced power consumption in the load, and reduced wafer area consumed by passive resistors. PNP and NPN HBTs could also be used together for compact, high-speed complementary digital circuits, such as I²L logic. Overall, the integration of PNP with NPN HBTs offers simpler circuits with reduced component count and reduced power consumption.

Due to their inherently lower speeds, PNP HBTs have not been thoroughly investigated. Since both types of HBTs are almost identical in theory, the design of PNP HBTs should be fairly similar to those of NPN HBTs. However, the lower hole mobility has several effects on the performance of PNP HBTs. The diffusion rate of minority carriers across the base is slower, which increases the base transit time (t _b) and reduces the gain by allowing more recombination in the neutral base. The emitter injection efficiency, which is proportional to m _p/m _n, is also much lower and results in reduced gain. The emitter and collector ohmic metals contact to p-type semiconductor, which typically have higher contact resistances than to n-type semiconductors. The holes do not experience velocity overshoot in the collector, which increases the collector transit time (t _C). However, the majority-carrier electrons in the base greatly reduce the base access resistance and the base contact resistance, which should reduce emitter current crowding and allow for thinner base layers. Overall, these effects result in reduced gain and slower performance for PNP HBTs when compared to NPN HBTs.

While not as impressive as their NPN counterparts, the current state-of-the-art frequency performance for PNP HBTs are sufficient for X-band and higher performance. The best published InAlAs/InGaAs PNP HBTs demonstrated b  = 170, f_T = 14 GHz, and f_max = 22 GHz [1], while the best published InP/InGaAs PNP HBTs demonstrated b  = 20, f_T = 11 GHz, and f_max = 25 GHz [5]. In the AlGaAs/GaAs material system, results have been presented with b  = 11, f_T = 33 GHz, and f_max = 66 GHz [6] and b  = 19, f_T = 37 GHz, and f_max = 30 GHz [7]. The apparent higher performance of the GaAs-based HBTs is mostly due to more aggressive base designs and higher attainable doping in the emitter cap and subcollector.

This paper first describes the drift-diffusion simulation of InAlAs/InGaAs PNP HBTs in order to determine designs for optimal DC and high-frequency performance. Then initial experimental results for InAlAs/InGaAs PNP HBTs on MBE-grown layers is presented. Finally, the results are discussed.

2. Simulation of PNP HBTs

The simulation of PNP HBTs was performed using a 2-dimensional drift-diffusion simulator which simultaneously solved Poisson’s equation, the carrier continuity equations for electrons and holes, and the energy balance equation for holes. Other physical effects that were simulated include a saturating field-veloctiy model, concentration-dependent low-field mobility, concentration-dependent Shockley-Read-Hall lifetime, Auger recombination, and thermionic emission and tunneling at the heterojunction. Material parameters for the simulations were obtained from current literature and are shown in Table 1. The HBTs were simulated on a 2D grid in order to include distributed base resistance effects on 2m m-wide emitter fingers; however, the extrinsic base-collector junction resulting from typical mesa-isolation processes was not included in the simulation. High-frequency performance was estimated by solving the drift-diffusion equations in the frequency domain for a sinusoidal excitation. Due to the lack of extrinsic junctions and surface recombination and to the disagreement of material parameters in current literature, the simulated values of currents, voltages, gain, and frequency performance for any specific HBT layer design should be taken as only approximate; however, the trends in these values between different HBT layer designs are accurate and indicate design trade-offs.

Parameter	InAlAs	InGaAs		Parameter	InAlAs	InGaAs
Eg (eV)	1.47	0.75		c (eV)	4.1	4.60
NC (cm-3)	5.1´ 1017	2.0´ 1017		NV (cm-3)	1.1´ 1019	6.1´ 1018
m n (cm2/Vs)				m p (cm2/Vs)
vnsat (cm/s)	8´ 106	7´ 106		vpsat (cm/s)	3´ 106	3´ 106
e r	12.70	13.88		BR (cm3/s)	1.0´ 10-11	2.0´ 10-10
CAn (cm6/s)	1´ 10-31	1.3´ 10-28		CAp (cm6/s)	1´ 10-31	1.25´ 10-29
t n,p SRH (s)
Table 1: Material parameters for drift diffusion simulations.

Layer	Type	Size (Å)	Doping (cm-3)
Emitter cap	p+ InGaAs	1000	1´ 1019
	p+ InAlAs	700	1´ 1019
Emitter	p InAlAs	1500	1.5´ 1017
Spacer	i InGaAs	100	-
Base	n+ InGaAs	500	5´ 1018
Collector	p- InGaAs	5000	5´ 1016
Subcollector	p+ InGaAs	2000	1´ 1019
Table 2: Basic layer structure for simulation of PNP HBTs

The basic layer structure for the simulations is shown in Table 2. The base layer is fairly thin (500 Å) for increased gain and reduced t _b, while the base doping is lower than for NPNs in order to increase hole mobility and reduce hole recombination in the neutral base. The energy band diagram and the hole concentration profile from a typical simulation of this structure are shown in Figure 1. Holes are injected from the wide-bandgap emitter to the narrow-bandgap base, where they diffuse across to the collector. Then the holes drift across the depleted collector to the heavily-doped collector contact region. The 100-Å spacer between the emitter and the base lowers the valence band spike by 0.21 eV at low currents in order to reduce the turn-on voltage. Since the heterojunction blocks reverse injection of electrons from the base to the emitter, over 99% of the total base current is due to recombination in the neutral base and in the emitter-base spacer, as can be seen in the hole concentration plot. The current gain is limited by Auger recombination, which typically generates over 60% of the base recombination current. Due to the lower hole velocity, base push-out occurs at lower collector current densities (~10⁴ A/cm²) than for NPN HBTs with a similar collector doping. The maximum current gain for the basic layer structure is around 30 at V_EC = 4 V, while the intrinsic values for f_T and f_max are 6.1 and 34 GHz, respectively. By estimating the additional R_B and C_BC due to the extrinsic base-collector junction area for a 2m m-wide base contact self-aligned on each side of a 2m m-wide emitter finger, the extrinsic f_max should be approximately 65% lower than the intrinsic f_max calculated here.

Figure 1: Simulation results for the layer structure in Table 2 at VEC = 4 V and JC = 7.56´ 103 A/cm2. The top figure shows the energy band diagram. The bottom figure shows the doping profile and the hole concentration. At this bias point, b  = 29.7.

 

The emitter-base junction design and the base layer thickness were investigated for effects on gain and high-frequency performance. Along with the basic structure with an abrupt heterojunction at the emitter-base spacer, HBTs were simulated with a compositionally-graded InAlAs® InGaAs spacer and without any spacer between the emitter and base layers. The basic layer structure was also simulated for base thicknesses from 300 Å to 600 Å. For a fixed base thickness of 500 Å, the abrupt spacer reduced the turn-on voltage by 0.10 V when compared to the design without a spacer due to the lowering of the valence band spike. However, the abrupt spacer also decreased the current gain by 16% (see Figure 2) due to hole recombination in the narrow-bandgap spacer region. When compared to the abrupt-spacer structure, the reduced charge storage in the spacer region caused f_T to increase by 35% and 45% for the no spacer and graded spacer designs, respectively. Reducing the base thickness from 600 Å to 300 Å tripled the gain of the HBT by reducing the recombination volume in the neutral base. Reducing the base thickness also significantly reduced the base transit time, which increased f_T by 33%.

Figure 2: Simulated effect of base thickness and emitter-base junction design on DC current gain.		Figure 3: Simulated effect of base doping on DC current gain. "Graded Doping" indicates one HBT with the base doping graded 5´ 1018 ® 1´ 1018 cm-3.

The effect of the base doping was investigated for two purposes: increasing the gain by using lighter doping, and decreasing the base transit time through a built-in electric field caused by a doping gradient in the base. For the latter case, the base doping was linearly graded 5´ 10¹⁸  ® 1´ 10¹⁸ cm^-3 from the emitter end to the collector end of the base. For uniform base dopings from 1´ 10¹⁹ cm^-3 down to 2´ 10¹⁸ cm^-3, the current gain increased dramatically by 19 times (see Figure 3), which strongly indicates the dominance of Auger recombination in the base. Over this doping range, Auger recombination was responsible for 40% to 75% of the total base current, where the increased Auger recombination is due to higher doping. The increase in gain and mobility caused an increase in f_T by 53% for the lowest-doped case. However, R_B increased significantly for lower base doping, which prevented improvement in the extrinsic f_max. The HBT with a doping gradient in the base demonstrated approximately the same gain and R_B as a uniformly-doped HBT at its average doping value, 3´ 10¹⁸ cm^-3. However, the drift field created by the graded base reduced t _b, resulting in approximately 9% increase in f_T.

Finally, HBTs were simulated with collector thickness from 3000 Å to 7000 Å, while the collector doping was fixed at 5´ 10¹⁶ cm^-3. The current gain did not vary significantly over this range of collector thicknesses. f_T decreased by 25% for the 3000-Å collector due to the decrease in the collector transit time. However, the accompanying increase in C_BC reduced the benefit to the extrinsic f_max.

3. Experimental Results

Two sets of PNP epilayers were grown on Fe-doped semi-insulating (001) InP by solid-source molecular beam epitaxy at The University of Michigan. The growth rate was 0.7 m m/h at 490º C, and an InGaAs/InAlAs superlattice was grown on the substrate to improve the material quality. The background doping for undoped InGaAs layers was below 5´ 10¹⁵ cm^-3, and undoped InAlAs layers were semi-insulating. Both layer structures were almost identical (see Table 3) except for the base doping: Wafer A had a uniform 5´ 10¹⁸ cm^-3 base doping, which resulted in a sheet resistance of 76 W /q and a contact resistivity to Ti/Pt/Au of 1.0´ 10^-6 W -cm². Wafer B had a linearly-graded base doping from 5´ 10¹⁸ cm^-3 at the emitter end to 1´ 10¹⁸ cm^-3 at the collector end, which resulted in a sheet resistance of 165 W /q and a contact resistivity to Ti/Pt/Au of 1.9´ 10^-6 W -cm². Both wafers were fabricated with the same process, which was based upon a high-performance NPN HBT process described elsewhere [8]. The Ti/Pt/Au base contacts were self-aligned to the Pt/Ti/Pt/Au emitter contacts, and the extrinsic base InGaAs was etched away using the base metal as a mask. Trenches etched under the emitter and base metals during mesa isolation were used to isolate the semiconductor junctions under the airbridge pads. Electroplated Au airbridges connected these airbridge pads to the interconnect metal for coplanar on-wafer probing. Nominal HBT emitter dimensions varied from 1-finger 1´ 10 m m² to 10-finger 5´ 40 m m², with typical values of 2´ 10 m m² and 5´ 10 m m² for high-frequency performance and high-power performance, respectively. Initial results from these wafers are presented below for a single-finger 5´ 10 m m² HBT.

Layer	Type	Size (Å)	Doping (cm-3)
Emitter cap	p+ InGaAs	2000	2´ 1019
	p+ InAlAs	700	1´ 1019
Emitter	p InAlAs	1500	8´ 1017
Spacer	i InGaAs	100	-
Base	n+ InGaAs	500	5´ 1018**
Collector	p- InGaAs	3000	3´ 1016
Subcollector	p+ InGaAs	5000	1´ 1019
Buffer	InGaAs/InAlAs superlattice	~1000	-
Substrate	Semi-insulating (001) InP

Table 3: Layer structure grown for Wafer A with uniformly-doped base. Layers for Wafer B were identical, except the base doping was linearly-graded 5´ 10¹⁸ ® 1´ 10¹⁸ cm^-3.

The forward I-V curve for a 5´ 10 m m² HBT from Wafer A is shown in Figure 4. The breakdown voltage BV_ECO (at 10 A/cm²) was 5.6 V, and the offset voltage was 0.20 V. From the Gummel plot, the ideality factors n_B and n_C were 1.60 and 1.00, respectively, and the maximum b was 12 at J_C = 34.8 kA/cm². As can be seen from the forward I-V curve, the gain increased significantly (b  > 20 and h_fe > 30) at higher V_EC. However, the gain compressed rapidly for J_C > 50 kA/cm² due to base push-out, which agrees with the maximum collector current density prior to base push-out from simulations using similar collector doping. Base push-out has a very significant effect on both the gain and the high-frequency performance of PNP HBTs due to the short diffusion length of the holes.



Figure 4: Forward I-V characteristics of 5´ 10 m m² HBT from Wafer A. I_B = 0.5 mA/step.

Microwave characteristics were measured on-wafer from 0.5 to 25.5 GHz using coplanar probes and an HP8510 network analyzer. |h₂₁|², G_max, and U were calculated from the S-parameters and extrapolated (when necessary) at 20 dB/decade to find f_T, f_max^Gmax, and f_max^U, respectively. For the 5´ 10 m m² HBT from Wafer A, the best frequency performance was f_T = 11 GHz, f_max^U = 27 GHz, and f_max^Gmax = 31 GHz at V_EC = 4.0 V and I_C = 11.69 mA. This is the highest reported f_max for any InP-based PNP HBT. The bias dependence of the frequency performance is shown in Figure 5. Note that the bias point for maximum frequency performance is very close to that of maximum gain, which indicates that hole transport rather than RC parasitics dominates the high-frequency performance. The effect of base push-out can be seen in the rapid decrease of f_T and f_max at higher collector currents.



Figure 5: Bias dependence of high-frequency performance for 5´ 10 m m² HBT from Wafer A.

The characteristics of the 5´ 10 m m² HBT on Wafer B were very similar to those on wafer A. The breakdown voltage BV_ECO (at 10 A/cm²) was 6.8 V, and the offset voltage was 0.20 V. From the Gummel plot, the ideality factors n_B and n_C were 1.65 and 1.03, respectively, and the maximum b was 4.2 at J_C = 14.2 kA/cm². The best frequency performance was f_T = 13 GHz, f_max^U = 23 GHz, and f_max^Gmax = 26 GHz at V_EC = 4.5 V and I_C = 6.47 mA. The forward I-V data exhibited a smaller Early voltage than Wafer A due to the lighter base doping at the collector edge. The gain characteristics were almost identical to Wafer A, although the peak gain occurred at lower current levels, perhaps either due to variations in geometries during fabrication or due to earlier base push-out from slight variations in the collector doping. Similarly, the maximum frequency performance also occurred at 45% lower collector current. The 15% increase in f_T is due to a 25% decrease in t _b caused by the drift electric field in the base. However, the doping gradient that created this drift field also significantly increased R_B, resulting in a 13% decrease in f_max.

4. Conclusion

According to 2-dimensional drift-diffusion simulations of InAlAs/InGaAs PNP HBTs, the most significant performance improvement can be obtained with a thin base (~300 Å) with a low base doping and an internal drift electric field, which results in 6 times more gain and 57% higher f_T than the conventional 500-Å base. A compositionally-graded emitter-base junction provides 19% more gain and is 45% faster than HBTs with a 100-Å spacer between the emitter and base, at the expense of 0.1-V higher turn-on voltage. Thin collectors (~3000-Å) increase f_T by 20% at the expense of higher C_BC and little improvement in f_max. In general, simulations indicate that the current gain is limited by Auger recombination in the neutral base, and the high-frequency performance is limited by the hole transit time.

Two wafers of InAlAs/InGaAs PNP HBTs were fabricated from MBE-grown epilayers. Wafer A, which had a 500-Å base doped uniformly at 5´ 10¹⁸ cm^-3, had a large-signal gain of 12, f_T = 11 GHz, and f_max = 31 GHz. This is the highest report f_max for any InP-based PNP HBT. The frequency performance was limited by the base transit time, which suggests that a thinner base or a base with a built-in electric field should be used. The gain and the frequency performance were both reduced at higher collector currents due to base push-out, indicating that an increased collector doping may allow for delayed base push-out and better performance at higher collector currents. Wafer B, which had a 500-Å base with a linearly-graded doping, gave lower base transit times and hence higher f_T than Wafer A, as predicted by the simulations. However, the lower average base doping increased the base resistance such that f_max was degraded by the R_BC_BC time constant.

This work is supported by ARO MURI (DAAH04-96-1-0001).

[1] S. Shi, K. P. Roenker, T. Kumar, M. M. Cahay, and W. Stanchina, "Simulation of PNP InAlAs/InGaAs Heterojunction Bipolar Transistors," IEEE Trans. Electron Dev., vol. 43, pp.1466-1467, 1996.

[2] D. Sunderland, and D. Dapkus, "Optimizing N-p-n and P-n-p Heterojunction Bipolar Transistors for Speed," IEEE Trans. Electron Dev., vol. 34, pp. 367-377, 1987.

[3] K. Kobayashi, D. Umemoto, J. Velebir Jr., A. Oki, ans D. Streit, "Integrated complementary HBT Microwave Push-Pull and Darlington Amplifiers with p-n-p Active Loads," IEEE J. Solid-State Circuits, vol. 28, pp. 1011-1016, 1993.

[4] D. Slater Jr., P. Enquist, F. Najjar, M Chen, and J. Hutchby, "Monolithic Integration of Complementary HBT’s by Selective MOVPE," IEEE Electron Dev. Lett., vol. 11, pp. 146-148, 1990.

[5] L. M. Lunardi, S. Chandrasekhar, and R. A. Hamm, "High-Speed, High-Current-Gain P-n-p InP/InGaAs Heterojunction Bipolar Transistors," IEEE Electron Dev. Lett., vol. 14, pp. 19-21, 1993.

[6] D. Slater, P. Enquist, J. Hutchby, A. Morris, and R. Trew, "Pnp HBT with 66 GHz f_max," IEEE Electron Dev. Lett., vol. 15, pp. 91-93, 1994.

[7] A. Kameyama, A. Massengale, C. Dai, and J. Harris, "Aluminum-Graded-Base PNp AlGaAs/GaAs Heterojunction Transistor with 37 GHz Cut-Off Frequency," IEICE Trans. Electron., vol. E79-C, pp. 518-523, April 1996.

[8] D. Sawdai, K. Hong, A. Samelis, and D. Pavlidis, "High Power Performance InP/InGaAs Single HBTs," 22^nd Int. Symp. Compound Semiconductors, pp. 621-626, 1995.