No abstract

In this paper a 0.25 pm SO1 CMOS technology is described. It uses imdepleted SO1 devices with nominal channel length of 0.15 pm, minimum channel length in the 0.1 pm range, supply voltage of 1 .S V, local interconnect, 6 levels of metal, artd same ground rules as the comparable bulk 0.25 pm CMOS. Key technology elements considered include device, penbrmance, reliability, ESD, and circuit functionality. Using this SO1 CMOS, a 4 Mb SRAM is demonstrated. This is the highest performance 0.25 pm CMOS technology reported to date. JNTRODUCTIONDespite many decades of work and promise of higher performance, SO1 CMOS technology has not been able to become the substra1.e for mainstream CMOS. The most important obstacle to SO1 becoming mainstream has been the floating-body ef'fects, and in particular the low nFET breakdown, which in turn severely limits high voltage screening of IC's. As; the CMOS technology moves into the 0.25 pm region, the supply voltage is reduced to 1.8 V.The breakdown of SO1 nFETs with LE^ in the 0.1 pm range is about 3 V. Lowered supply voltages, for the first time, have opened an opportunity for SO1 CMOS to become a mainstream technology. In developing the 0.25 pm SO1 CMOS technology, the goal has been to optimize the technology for maximum performance over the bulk, while keeping the standby current the same as a 0.25 bulk CMOS technology under the worst temperature condition and minimum channel length. In this work, the supply voltage chosen was based on achieving the highest possible performance. Devcloping a low power (voltage) technology was not part of this effort, although a 0.25 p m CMOS technology ht3s excellent low power potential.In addition to lhe usual technology issues addressed in a bulk CMOS, there are unique issues related to the floating-body effects (in particular pass gate leakage and history dependence of the delay) that should be addressed when developing an SO1 CMOS. These effects become more important as technology is applied to circuits which have tight timings or are sensitive to node leakage. In this work floating-body effects have been minimized to a great extent. The technology was applied to a 4 Mb SRAM, modified to work on SOI. Fast access time, as well as a wide operating voltage range is demonstrated. TECHNOLOGY and DEVICE RESULTSThis technology uses shallow trench isolation (!$TI), 40 A gate oxide, local interconnect, and the 6-lever1 A1 interconnect from 0.25 pm bulk CMOS [2-31. Figure 1 is an SEM of finished cross section (only the first 3 levels of BEOL shown for clarity). It has the same basic ground rules as a 0.25 pm technology. Only small density improvement over a 0.25 p m bulk technology is expected in a SO1 CMOS: Latch-up is of little concern in 0.25 pm 1.8V technologies, well contact in bulk technologies takes up little area, and most circuits are BEOL pitch limited. Use of STI eliminates any improvement that one might get in isolation in SOI. Non-depleted device was chosen for this technology [ 11. The effective channel length, LEFF, of this Figure 1-Cros...

No abstract

Doped-channel i-In0.52Al0.48As/n+-In0.53Ga0.47 As/i-In0.52Al0.48As heterojunction metal-semiconductor field-effect transistors lattice matched to the InP substrates with gate lengths in the submicron range have been fabricated and characterized. The dc and microwave performance of the devices are presented in this paper. Drain current anomalities, or the kink effects, were observed at room temperatures as well as at 77 K in the dc measurements. The kinks are associated with the deep-level electron trapping, and are not present at microwave frequencies. The dc and microwave bias points for achieving maximum transconductance are different. Normal microwave characteristics are exhibited. A compression of transconductance gm is observed in the dc measurement, while such gm degradation phenomenon does not appear at microwave frequencies. At 10 GHz, an extrinsic gm of 507 mS/mm, a current-gain-cutoff frequency ft of 49.5 GHz, and a power-gain-cutoff frequency (fmax) of 70.5 GHz were observed for a 0.25-μm-gate device. For a 0.3-μm-gate device, a gm of 545 mS/mm, an ft of 42 GHz, and an fmax of 114 GHz were measured. Output conductance as low as 7.7 mS/mm was measured. A voltage gain (gm/gds) as high as 64 was observed. The voltage gain for measured devices is well above 20 for a wide range of bias conditions. Fabricated devices show their potential for high-frequency operations. Further investigation of the origin of traps should result in improved low-frequency device characteristics.

No abstract