Technical Insight
InP battles to beat the slump in optical communications
While new processing technologies ensure that InP microelectronics will keep getting faster, new compounds are displacing InP as the material of choice for some long-wavelength lasers, writes Jon Newey.
This year s Indium Phosphide and Related Materials (IPRM) conference took place on May 12-16 in Santa Barbara, CA. The Santa Barbara area has a distinguished history in InP and other compound semiconductors, with a number of universities and companies active in the field located nearby.
It has been a tough couple of years for those involved in InP, with the collapse of optical communications infrastructure spending meaning reduced demand for long-wavelength emitters and a limited uptake of InP microelectronics. In the current cost-conscious climate, it is often argued that InP is too expensive and that its use should be avoided unless absolutely necessary. Even InP s monolithic integration capabilities are being pushed to one side in favor of using InP with other materials in hybrid solutions. Given this situation, it was perhaps not surprising that there was considerable emphasis at the meeting on the related materials aspects, where alternatives to InP were discussed. The sessions on lasers included much on dilute nitrides and in the microelectronics area there was considerable discussion on metamorphic HEMTs. Both of these technologies rely on GaAs instead of InP substrates. However, right from the opening session it was clear that InP has some powerful backers, and that there are applications where nothing else will do the job well enough (figure 1).
The opening session of three plenary talks gave an overview of a range of subjects, including InP opto- and microelectronics for the recovering telecom industry and InP HBT mixed-signal circuit technology. Henning Riechert of Infineon discussed the status of 1300 nm VCSELs and the various approaches being taken to fabricate them. He concentrated mainly on devices with GaInAsN active regions, as this is the route Infineon has chosen and refined for its recently announced volume production of 1300 nm VCSELs.
InP HBTsHistorically it has often been a military need that has pushed compound semiconductor technology to new levels of performance, and InP HBTs are now benefiting from the attention of a DARPA program called Technology for Frequency Agile Digitally Synthesized Transmitters (TFAST). The program manager, John Zolper, described the challenging scaling and integration issues that InP HBT technology faces in its quest to stay ahead of aggressively scaled SiGe bipolar technology. InP HBT technology has relied heavily on vertical scaling and design to push its performance along, but this is no longer enough. SiGe HBTs with 0.18 µm emitters have demonstrated a peak ft of 350 GHz with a current density of 20 mA/µm2, although BVCEO was just 1.4 V. This performance comes from processing technologies that reduce the extrinsic base-collector capacitance and extrinsic base and emitter resistance. TFAST calls for super-scaled InP HBT technology that will keep InP ahead of SiGe in speed and power by using many of the scaling approaches employed for SiGe HBTs. InP DHBTs have many attractive features that will help meet military requirements for mixed-signal circuits such as DACs and ADCs. Devices are required that can directly generate millimeter-wave frequency signals over a large range of signal levels. The desired outcomes of the TFAST program are summarized in Table 1.
Yingda Dong from the University of California, Santa Barbara (UCSB) reported on the selective area growth (SAG) of InP through narrow openings in an SiO2 mask on InP, and also showed the DC characteristics of HBTs fabricated using the SAG material as the extrinsic base region. By fabricating SiO2 stripe masks arranged radially on an InP substrate with various widths for the SAG openings, it was possible to establish the degree of InP lateral overgrowth onto the SiO2 as a function of the width of the opening and its angle of deviation from the (110) direction (figure 2). Dong described a device structure where the base mesa is etched down to the subcollector, leaving a narrow base shoulder. SiO2 is deposited to surround the collector pedestal, defining an intrinsic base-collector junction area slightly larger than the emitter-base junction, thus minimizing base-collector capacitance. The 0.6 µm base shoulder provides the opening for extrinsic base selective growth and overgrowth onto the SiO2 (figure 3).
The DC performance of the prototype regrown devices showed a DC gain of 15 at a VBC of 1.5 V, and the ideality factor for the base current was 1.93. This high factor was attributed to excess base current resulting from surface damage prior to the regrowth, and interface degradation during the regrowth. Work is now under way to improve this, and Dong hopes to be able to report RF characteristics once further efforts are made to reduce process-related damage and improve the quality of the selectively grown material.
R Kopf of Bell Labs discussed the planar processing of InP DHBTs by ion implantation as a way of overcoming the heat dissipation problems that may arise in the large-scale integration of mesa HBTs. InP has a higher thermal conductivity than InGaAs, and using an ion-implanted InP subcollector reduces thermal impedance by a factor of 10 for high current density operation. By selective implantation the extrinsic base-collector capacitance can be eliminated.
It has been a tough couple of years for those involved in InP, with the collapse of optical communications infrastructure spending meaning reduced demand for long-wavelength emitters and a limited uptake of InP microelectronics. In the current cost-conscious climate, it is often argued that InP is too expensive and that its use should be avoided unless absolutely necessary. Even InP s monolithic integration capabilities are being pushed to one side in favor of using InP with other materials in hybrid solutions. Given this situation, it was perhaps not surprising that there was considerable emphasis at the meeting on the related materials aspects, where alternatives to InP were discussed. The sessions on lasers included much on dilute nitrides and in the microelectronics area there was considerable discussion on metamorphic HEMTs. Both of these technologies rely on GaAs instead of InP substrates. However, right from the opening session it was clear that InP has some powerful backers, and that there are applications where nothing else will do the job well enough (figure 1).
The opening session of three plenary talks gave an overview of a range of subjects, including InP opto- and microelectronics for the recovering telecom industry and InP HBT mixed-signal circuit technology. Henning Riechert of Infineon discussed the status of 1300 nm VCSELs and the various approaches being taken to fabricate them. He concentrated mainly on devices with GaInAsN active regions, as this is the route Infineon has chosen and refined for its recently announced volume production of 1300 nm VCSELs.
InP HBTsHistorically it has often been a military need that has pushed compound semiconductor technology to new levels of performance, and InP HBTs are now benefiting from the attention of a DARPA program called Technology for Frequency Agile Digitally Synthesized Transmitters (TFAST). The program manager, John Zolper, described the challenging scaling and integration issues that InP HBT technology faces in its quest to stay ahead of aggressively scaled SiGe bipolar technology. InP HBT technology has relied heavily on vertical scaling and design to push its performance along, but this is no longer enough. SiGe HBTs with 0.18 µm emitters have demonstrated a peak ft of 350 GHz with a current density of 20 mA/µm2, although BVCEO was just 1.4 V. This performance comes from processing technologies that reduce the extrinsic base-collector capacitance and extrinsic base and emitter resistance. TFAST calls for super-scaled InP HBT technology that will keep InP ahead of SiGe in speed and power by using many of the scaling approaches employed for SiGe HBTs. InP DHBTs have many attractive features that will help meet military requirements for mixed-signal circuits such as DACs and ADCs. Devices are required that can directly generate millimeter-wave frequency signals over a large range of signal levels. The desired outcomes of the TFAST program are summarized in Table 1.
Yingda Dong from the University of California, Santa Barbara (UCSB) reported on the selective area growth (SAG) of InP through narrow openings in an SiO2 mask on InP, and also showed the DC characteristics of HBTs fabricated using the SAG material as the extrinsic base region. By fabricating SiO2 stripe masks arranged radially on an InP substrate with various widths for the SAG openings, it was possible to establish the degree of InP lateral overgrowth onto the SiO2 as a function of the width of the opening and its angle of deviation from the (110) direction (figure 2). Dong described a device structure where the base mesa is etched down to the subcollector, leaving a narrow base shoulder. SiO2 is deposited to surround the collector pedestal, defining an intrinsic base-collector junction area slightly larger than the emitter-base junction, thus minimizing base-collector capacitance. The 0.6 µm base shoulder provides the opening for extrinsic base selective growth and overgrowth onto the SiO2 (figure 3).
The DC performance of the prototype regrown devices showed a DC gain of 15 at a VBC of 1.5 V, and the ideality factor for the base current was 1.93. This high factor was attributed to excess base current resulting from surface damage prior to the regrowth, and interface degradation during the regrowth. Work is now under way to improve this, and Dong hopes to be able to report RF characteristics once further efforts are made to reduce process-related damage and improve the quality of the selectively grown material.
R Kopf of Bell Labs discussed the planar processing of InP DHBTs by ion implantation as a way of overcoming the heat dissipation problems that may arise in the large-scale integration of mesa HBTs. InP has a higher thermal conductivity than InGaAs, and using an ion-implanted InP subcollector reduces thermal impedance by a factor of 10 for high current density operation. By selective implantation the extrinsic base-collector capacitance can be eliminated.
