Electron spin resonance of separation by implanted oxygen oxides: Evidence for structural change and a deep electron trap

Separation by implantation of oxygen (SIMOX) is the leading technology for silicon-on-insulator ( SOI), a device fabrication method with great promise for use in satellites and ultralarge scale integration. SIMOX buried oxides contain trapping centers that may play a significant role in the operation of devices utilizing this technology. We study trapping centers in SIMOX buried oxides with a combination of electrical measurements and electron spin resonance (ESR) .I Combining ESR and capacitance vs voltage (CV) measurements, we recently found that very high densities ( 10i8/cm3) of paramagnetic (ESR active) point defects called Ev centers are generated in SIMOX buried oxides exposed to vacuum ultraviolet (WV) irradiation (AC//~ 210.2 eV).25 This indicated the presence of a very high density of E precursors in the buried oxides. The E’ center is an unpaired spin on a silicon bonded to three oxygens; the E’ ESR spectrum has a zero crossing g-2.0005. We searched for other ESR spectra in the vicinity of g=2.000 but have not yet been able to detect other signals including the “amorphous silicon” signal reported by Stessmans, Revesz, and Devine.” The creation of high densities of E centers was accompanied by virtually no net space charge in the buried oxide.39J7 This absence of net space charge in the presence of a large E density suggests two possibilities:3-5Y7 ( 1) that the E’ centers are neutral, or (2) that the E’ centers are positively charged (E’ centers are the dominant deep hole trap in thermal oxides)’ and compensated by negatively charged centers.3-5’7 To test these possibilities and determine whether or not SIMOX E centers are electrically active, we performed a series of charge injection experiments.2-5*7 Injection of electrons into VUV illuminated oxides reduced E’ amplitude; injection of holes into the oxides increased E” amplitude.4’5*7 Both of these results demonstrate that SIMOX E’ centers are electrically active and that at least some of them are positively charged when paramagnetic. However, in these experiments we consistently observed an E’ density greater than total charge density.4*5.7 The fact that the trapped charge density is considerably lower than the presumably positive E’ center density leads one to suspect some form of electrically compensating mechanism to account for the discrepancy. In this letter, we determine more directly the effects of electron and hole injection on the buried oxide and provide very strong evidence for compensating positive and negative charge. We also provide evidence for structural changes and the creation of a deep electron trap as a result of VUV illumination. The samples used in this study include P( 100) 405 nm single implant and N( 100) 385 nm multiple implant SIMOX oxide samples. Both single and multiple implant samples received a 5 h anneal in 99.5% argon and 0.5% oxygen at 1315 “C. All samples received a total oxygen dose of 1.8 X 10”/cm3 at an energy of 200 keV, a current of 34 mA, and an implant temperature of 640 “C. A residual oxide and the top layer of silicon were removed by subsequent etches in HF and then KOH at room temperature. The behaviors of the multiple and single implant oxides were qualitatively the same though not identical. We made X-band ESR measurements at room temperature using a TE,,, “double” resonant cavity and a “weakpitch” spin standard. Relative spin-concentration measurements are accurate to f 10% with an absolute accuracy of a factor of two. High frequency CV measurements were taken at room temperature using a mercury probe. Net oxide space charge density was determined from CV curve shifts. (Etchback experiments indicate charge trapping throughout the oxides). E’ centers were generated by exposing (bare) buried oxides to VUV light from a 50 W deuterium lamp in a vacuum. In some cases, a filter passing only 10.2 eV photons was used; in these cases the oxides were illuminated briefly under positive bias. Biasing was performed by depositing low-energy ions created by corona discharge’ onto the samples. Surface potentials were measured with a Kelvin probe electrostatic voltmeter. [Most of these 10.2 eV photons are absorbed in the top 100 A of the oxide where they create electron hole pairs.‘oP” The positive bias drives holes across the oxide (hole injection) while sweeping electrons out.] In other cases, the oxides were VUV illuminated unbiased without the filter (he/A< 10.2 eV) for an extended period ( -40 h). Exposing the samples to this “extended” VUV illumination generates extremely high densities of paramagnetic E’ centers ( 10*8/cm3). Ultraviolet illumination (UV) from a sub-SiOl band gap (he/A. = 5.5 eV) mercury-xenon lamp was also used in combination with positive corona bias. The UV illumination results in the internal photoemission of electrons from