Charge Trapping and Transport Properties of SIMOX Buried Oxides with a Supplemental Oxygen Implant *

The radiation response characteristics of singleand multiple-implant SIMOX (separation by implantation of oxygen) buried oxide layers that had received a supplemental oxygen implant and anneal step were measured as a function of temperature and time after exposure to short radiation pulses. A fast capacitance-voltage technique was used for these measurements. The results indicate that, in comparison to standard SIMOX, the supplemental-implant SIMOX buried oxide shows hole motion through the oxide, greatly reduced bulk hole trapping, and little or no bulk shallow electron trapping. Substantial interfacial hole trapping was observed in these materials, as well as deep electron trapping in the single-implant material.


I. INTRODUCTION
Silicon-on-insulator (SOI) buried oxides prepared by the SIMOX (separation by implantation of oxygen) process have been shown to contain a high density of both deep hole traps and shallow electron traps, in contrast to standard gate oxides produced by direct free-surface thermal oxidation [1][2][3][4].The hole traps pose a problem for the radiation hardening of SIMOX layers, while the electron traps may contribute to oxide leakage current conduction.The major differences between thermal oxides and SIMOX oxides have been attributed to oxygen deficiency or excess silicon in the SIMOX buried oxide (BOX) layer.The dominant hole trap identified in the SIMOX BOX after exposure to ionizing radiation or a flux of holes is a form of E' center [4]; in the simplest model, the E' center is formed by interaction of a hole with an oxygen vacancy in the Si02 lattice [5].The excess silicon may result from reducing conditions generated in the BOX during the high-temperature (near 1300°C) anneal step common in SIMOX processing.During this anneal, the BOX is sealed off from a supply of additional oxygen by the top silicon layer [4].A logical approach to reducing the bulk hole trapping in the SIMOX material was proposed by Hughes, Revesz, and others [6]: to convert the SIMOX BOX into a more nearly stoichiometric Si02 layer (presumably, more like a "thermal" oxide) by adding additional oxygen to the BOX after completion of SIMOX processing.The additional oxygen is added by implantation and then the BOX is annealed to promote oxidation of the excess silicon at a temperature (near 1000°C) known to result in oxide layers with low trap densities.
In this work, we examine radiation-generated charge buildup in both single-and multiple-implant SIMOX buried oxides that have received an additional or supplemental oxygen implant and moderate-temperature anneal.We measured the time-dependent response of these materials to short-pulse radiation exposure as a function of buried oxides that have received an additional or supplemental oxygen implant and moderate-temperature anneal.We measured the time-dependent response of these materials to short-pulse radiation exposure as a function of temperature and applied bias, using a fast capacitance-voltage (C(V,)) technique, and analyzed the results using simple models to determine the primary response mechanisms and associated material parameters.The results indicate that the radiation response of the supplemental-oxygen-implant (SUPOX) SIMOX BOX differs substantially from that of the untreated or standard SIMOX.In contrast to standard SIMOX, SUPOX SIMOX shows greatly reduced trapping of radiation-generated holes in the interior (bulk) of the BOX and transport of holes through the BOX similar to that observed in thermal oxides.Also unlike standard SIMOX, and like thermal oxides, the single-implant material shows little or no shallow trapping and subsequent detrapping of electrons in the bulk of the BOX.However, unlike thermal oxides, at least the single-implant SUPOX material shows evidence for substantial long-term, or deep, trapping of radiation-generated electrons in the BOX.

SAMPLES AND EXPERIMENTAL TECHNIQUES
IBIS Technology Corporation supplied two types of supplemental-implant SIMOX in wafer form.The singleimplant SUPOX material was produced with a 1.8 x lOl*/cm2 200-keV oxygen implant and subsequent high-temperature anneal: the multiple-implant SUPOX material received three lower-dose oxygen implants with interspersed anneals.Follow-*This work was sponsored by the Defense Nuclear Agency.U.S. Government work not protected by U.S. copyright ing normal SIMOX processing, each wafer received a supplemental oxygen implant of 1 x lO17/cm2, followed by a 1-hour, 1000°C anneal.Nominal BOX thicknesses for the single-and multiple-implant wafers were 403.4 and 387.3 nm, respectively.BOX metal-oxide semiconductor (MOS) capacitors were fabricated on the SUPOX wafers at the Army Research Laboratory Semiconductur Engineering and Materials Technology Facility (ARL SEMT).These BOXCAPs consisted of n+-doped superficial-Si and polysilicon electrodes over the BOX layer on thep-type Si substrates.C(Vg) measurements on these units were sensitive to charge effects at the substrate/BOX (back) interface.
We irradiated the BOXCAPs with 4-p.s, 13-MeV electron pulses using the Armed Forces Radiobiology Research Institute (AFRRI) electron linear accelerator (LINAC).The samples were exposed under a selectable gate bias, Vg, in an evacuated sample holder with provisions for temperature control from 77 to 450 K and for both active and passive dosimetry (PIN diode, thermoluminescent dosimeters).To minimize perturbations of the internal electric field in the BOX by radiation-generated space charge, the BOX dose was kept below 30 krad(SiO2) and the magnitude of the electric field, E,,, applied across the BOX was kept relatively large (near 1 MV/cm for 40 V across 400 nm) [7].We recorded C(Vg) characteristics for the BOXCAPs before and from 0.2 ms to 800 s after a radiation pulse using a fast measurement system [8] by interrupting the DC gate bias, applying a 160-p.svoltage ramp, and simultaneously recording the sample capacitance.Midgap voltage shift as a function of time after irradiation, AV,,&), was obtained from the C(Vg) data.AVw is assumed b be equivalent to that component, AV, , , , of the radiation-induced shift in the C(Vg) characteristics caused by oxide trapped charge.The variability in the measured AVmg(') from curve to curve caused by pulse-to-pulse dose variahons, sample variations, and AVw measurement/extraction errors is estimated to be the greater of f 3 V or M.1 AVmg.Point-topoint error in AV, , &) within a given curve caused by AVmg measurement/extraction errors alone is estimated to be +1 V.

In. RESULTS
Figure 1 shows C(Vg) curves for a single-implant SUPOX sample irradiated to 26 krad(SiO2) at 194 K with Ve = 4 0 V.The dashed curve is the preirradiation characteristic; the other curves were taken at the indicated times following the radiation pulse.At 0.2 ms after irradiation, the sample shows a negative shift (AVmg = -15 V); this shift recovers with time past the preirradiation value and is positive at 800 s (AVmg = 8 V).
Figure 2 shows AV, &) for single-implant samples irradiated with Vg = 40 V ( E , near 1 MV/cm) (Fig. 2a) and Vg = 4 0 V (E, near-1 MV/cm) (Fig. 2b) at temperatures near 120, 195, and 405 K at an average dose of 26 krad(SiO2).For Vg = 40 V, the observed shifts are large and negative; for negative bias, the shifts are substantially smaller, and both negative and msition to a positive 8-V shift at late times (near 800 s).At 300 K, AVw is near 7 V and shows little change with time.
Overall, the data suggest a temperature-activated recovery process that is barely begun at 800 s at 120 K and is largely complete by 800 s at 300 K.A similar pattem is observable in the +4O-V data (Fig. 2a).In this case, AVv is near -56 V at early times at 120 and 193 K, a rapid transibon is made at 193 K to near -75 V, and AVw shows little change (slight positive shift to -71 v) from near -75 V at 300 K.At 405 K, AVw starts out near -7 1 V at 0.2 ms and then begins a rapid positive shift starting at about 0.01 s.A much weaker positive shift starting at this time is also detectable in the -40-V data at 405 K (Fig. 2b).
Figure 3 shows AV, , &) measured on triple-implant samples exposed to 14.5-krad(SiOz) pulses under 40-V (Fig. 3a) and -40-V (Fig. 3b) gate biases.For this material, AVmg is always negative for both positive and negative Vg, and shows relatively little change with time (note the expanded AVmg scales).Pulseto-pulse variability in the measured AVw (which primarily affects the positions of the AVw(t) curves on the voltage scale) makes these data more difficult to interpret.Nevertheless, within experimental error the general pattem of response seen in Figure 2 is discernible here: For Vs = 4 0 V, the early shift at 121 and 184 K is near -19 V; AVw makes a transition in the positive direction at 184 K to near -15 V and remains there at 294 K.For

A. Initial Charge Trapping
As noted in Section 111, for the single-implant samples irradiated with Vg = 40 V (Fig. 2a). the shifts at 120 K and at 193 K at early times are near -56 V.These shifts may be compared with AVO, which is the expected AVot for an irradiated oxide layer if all the radiation-generated holes are trapped or immobilized near their points of origin in a uniform distribution through the oxide and no electrons are trapped.These conditions are closely approached in a thermal oxide irradiated below 150 K [9]; at low temperatures, holes are essentially immobile in S i 4 [5].For an oxide irradiated with 1 ~~4 = 1 MV/cm, AVO = (-1.4k 0.2) x 10-8 dox2 D, where d,, is the oxide thickness in nanometers and D is the high-energy (WO or LINAC) dose in rads(Si0i) [9,10].For the present case, AVo = -59 4 9 V; therefore, within experimental error, the early/low-temperature 40-V shifts are close to AVO.On the other hand, the results for the single-implant samples irradiated under negative Vs (Fig. 2b) show AVmg near -15 V at 120 and 193 K at early times, and the shifts become positive at later times and higher temperatures.These results are clear evidence for substantial electron trapping in the single-implant SUPOX BOX.The apparent lack of significant effect of the electron trapping on the positive-Vg low-temperature AVmg suggests that any electron trapping is taking place near, or is at least strongly biased toward, the top interface and therefore has little effect on AVw.Reduced effect of electron trapping on AV under positive Vg would occur even if the eleccron traps were dstributed uniformly through the BOX (see discussion elsewhere

Earlybw-Temperature Recovery
Recent measurements using the 10-keV x-ray photoconduction technique show clearly that substantial hole motion takes place in single-implant SUPOX SIMOX BOX layers within (Slow transport of electrons in Si02 has not been observed.)For both the single-and triple-implant materials, the early/lowtemperature recovery process produces negative-going AVm&) under positive Vg and positive-going AVw(f) under negative Vg (Figs. 2 and 3).The sign of the change in AVw with time is no sure clue to the nature of the process: Transport of holes through the oxide can result in negative-going AVw if more than half of the holes are captured by traps at the sensitive SiOdSi interface (in the present case, the BOWsubstmte interface), and detrapping of electrons in the oxide bulk can cause positive-going AVmg if more than half of the detrapped electrons are retrapped at the sensitive interface.However, from the continuous-time randomwalk (CTRW) model for hole transport, when AVo,(r) resulting from hole transport is plotted as a function of log(r) (as has been done for our data), a characteristic S-shaped curve results.The shape of this curve is characterized by the parameter a, which determines the time dispersion, or spread, of the AVo&) recovery [ 11,5].Further, the CTRW model predicts specific behavior for AVot(f) at the extremes or asymptotes of the S-curve: i.e., at early times when AV,, begins to depart from its initial value and at late times as it approaches its final value.For transport under negative bias, the early and late stages of the transport are expected to be of the forms This model convolves the impulse response functions for hole transport, trapping, and detrapping with a radiation pulse to obtain the overall AV& of the system.(Interface trap effects may also be added.)For the 120 and 193 K cases, the hole transport was modeled with a= 0.32 and an initial shift AVmg(O+) = -56.5 V.The time for half of the transport to take place, rln, was adjusted so that a best fit was obtained to the data at each temperature.Fit values for fln were 0.4 s for 193 K and 2000 s for 120 K.The increasing negative shift with time in both cases implies that more than half of the holes transporting toward the BOWsubsuate interface under the positive gate bias become trapped at or near that interface.The fraction of holes captured at that interface,fh, was also adjusted for the best fit to the 193 K data (f,h = 0.67).
The greater "noise" and smaller range of recovery seen in the triple-implant results (Fig. 3) preclude fitting the CTRW model to these data.However, as noted in Section 111, the timedependent response is qualitatively similar to that of the singleimplant material.Overall, it is reasonable to conclude that hole transport is the dominant early/low-temperature recovery process in both SUPOX materials.

Late/High-Temperature Recovery
As noted in Section 111, a second recovery process that causes positive-going AVmg becomes important in both types of SUPOX material at high temperature (near 400 K) and later times (after 0.1 s).The recovery is most evident in the positive-Vg data for both materials (Figs.2a and 3a).The process that causes this recovery must be loss of holes from traps at or near the back interface, since the hole transport is already complete and electron motion under positive bias would be directed away from the back interface.The delayed onset of the second recovery and its appearance primarily at high temperature suggest that the process is thermal detrapping of the holes near the interface rather than their removal by a tunneling process, since tunneling is only weakly thermally activated and would be expected to begin concurrent with arrival of significant numbers of holes near the interface; i.e.. tunneling should overlap the last stages of the hole transport process.

C. Modeling of Trapped Charge Distributions
We employed a simple analytic model developed in previous work [ 151 to generate estimates of the approximate location and magnitude of charge trapping in the SUPOX BOX materials.The model assumes uniform distributions of electron and hole traps in the oxide layer that capture electrons and holes with mean free paths Se and Sh, respectively, and traps at the bottom (Siodsubstrate) interface that capture a specified fraction,feb of the electrons or holes, respectively, that reach that interface.(Trapping at the top interface is not considered here-recall that our measurements are not sensitive to trapping at that interface.)The shifts, AV+ and AV-, under positive and negative bias, respectively, are given by Table 1 shows the results of applying the model to selected AVms data for the SUPOX materials and also includes for comparison results from previous work for a standard SIMOX material [ll.The AV entries without parameter values (Se, Sh, f&,feb) are the measured values; the AV values with entries for the parameters are approximate fits to the measured values reached by trial-and-error adjustment of the parameters.Also shown are other results [2,13] for Ic,,JIco from x-ray photoconduction measurements performed with 1-s radiation pulses on the standard and single-implant SUPOX materials and the model predictions for these values.
= 1 corresponds to both holes and electrons moving completely through the BOX within the 1-s measurement time (see discussions elsewhere [2,13]).Results are fit for each material for early time at low temperature (before any hole transport), for 1 s at m m temperature (after most of the hole transport and/or shallow electron detrapping), and for late time at high temperature (into the late recovery stage).The model is crude-flat distributions of traps in the oxide bulk and &function distributions at the interface-and cannot be expected to produce an exact fit to the results.Nor are the fits obtained necessarily unique.The intent was to determine if the measurement results could be approximately reproduced by reasonable evolution of charge distributions in the BOX.We fit the standard SIMOX results by adjusting only Se and Sh (no interfacial trapping) as discussed elsewhere [ 11. S, is much smaller than the BOX thickness dox, reflecting the very efficient trapping of holes in the BOX bulk; Se is about 10 times larger.Both Sh and Se increase with time and temperature, reflecting the thermal detrapping of almost all the electrons and some detrapping of holes.
At low temperature, we fit the single-implant SUPOX results by assuming no hole motion and some bulk trapping of electrons (Se over six times doJ, together with some trapping of electrons at the back interface.At room temperature, Sh = loo0 nm andfhb = 0.64 indicate substantial hole motion and strong trapping of holes at the back interface, in agreement with our results in Section IV.At high temperature, bulk hole trap ping is eliminated (Sh 00).and the interfacial hole trapping is strongly reduced, reflecting the (probably thermal) hole anneal process.Some electrons are lost from the interface with increasing temperature, but significantly, the bulk electron trapping does not change up to late times at high temperature, indicating that some electrons are very deeply trapped in the single-implant material.
Fits to the triple-implant SUPOX data suggest that this material shows more bulk but less interfacial hole trapping than the single-implant material.At low temperature, the results indicate substantial electron trapping (Se comparable to do.J but by 1 s at room temperature these electrons are gone (Se + m).
This suggests that the triple-implant material has a significant density of shallow electron traps but no deep traps.

V. CONCLUSIONS
We have examined the time-dependent response to shortpulse irradiation of single-and triple-implant SIMOX buried oxides that had received a supplemental oxygen implant and anneal.The results show that, in comparison to standard (untreated) SIMOX BOX layers, both types of SUPOX materials show much less trapping of holes in the oxide bulk, as well   as evidence of hole transport through the oxide similar to that observed in thermally grown oxide layers.Like soft thermal oxides (e.g., unhardened field oxides), these materials also show evidence that a large fraction of the radiation-generated holes may become trapped at the SiOdSi interfaces and may cause back-channel voltage shifts in excess of those observed for 100-percent bulk hole trapping in standard SIMOX.Annealing of these trapped holes evidently takes place at high temperature by thermal detrapping.The electron-trapping properties of the single-and tripleimplant SUPOX materials differed from those of standard SIMOX and from each other.Unlike the standard SIMOX, the single-implant SUPOX showed substantial deep trapping of electrons in the BOX bulk and almost no shalhv trapping.The tripleimplant SUPOX material showed reduced shallow electron t r a p ping in comparison to standard SIMOX, and no deep trapping.
The nature of the electron traps in SIMOX is not clear at this time.The E' center-long considered to be the primary hole m@as also recently been identified as a candidate deep electron trapping site [16].It may also be a shallow electron trap.It is possible that substantial deep electron trapping in standard SIMOX is masked by the extensive deep hole trapping.DC conduction measurements on the SUPOX BOX materials [ 171 indicate that the triple-implant material h e lower conductivity (more like that of a thermal oxide) than does the single-implant material: this in turn suggests that more of the oxygendeficiency defects have been eliminated.If these defects include the deep electron traps, then the presence of these traps in the singleimplant material and their absence in the triple-implant material may be explained.However, this argument does not explain the apparent presence of shallow traps in the triple-implant material.
The reduction of the density of bulk hole traps in the SIMOX BOX brought about by the supplemental-oxygen processing opens the way for hardening of the BOX by techniques developed for reducing or compensating interfacial hole traps in conventional thermal oxides.The creation of deep electron traps in the single-implant supplemental-oxygen material may also be useful for radiation hardening of the buried oxide.
[l]).The early-time 121-and 184-K shifts for the triple-implant SUPOX samples are near -23 V for Vg = 40 V and near -19 V for Vg = -40 V (Fig.3).AVO for these samples exposed to 14.5-krad(Si02) pulses is -31 k 5 V.The reduction of the earlytime low-temperature shifts well below AVo for both negative and positive Vg suggests the presence of a substantial density of electrons trapped in the bulk of the triple-implant BOX at these times and temperatures.The spatial distribution of the trapped electrons in both the single-and triple-implant materials is examined further in Section IV C, Modeling of Trapped Charge Distributions."fgB.AVmg Recovery ProcessesStandard SIMOX BOX contains a high density of shallow electron traps.These traps capture a large fraction of the radiation-generated electrons and then release them by thermal depopulation in about 1 s at room temperature [l].This process is almost entirely responsible for the time dependence of AVot observed in standard SIMOX BOX structures at times less than a few thousand seconds following irradiation at or below mom temperature.At higher temperature and longer times, some positive shift in AVot is observed from thermal detrapping of holes [l].On the other hand, in thermal oxides, the major cause of time-varying AVot below mom temperature for times less than loo0 s after irradiation (earlyfiow-temperature AVot or AVms) is the slow, time-dispersive transport of radiationgenerated holes through the oxide [11,5].Following the hole bansport, additional AVot change is caused by removal of holes from traps primarily at the SiOdSi interfaces by thermal depopulation and tunneling processes [12,5].

1 s
after irradiation at m m temperature [131.(In contrast, no hole motion is observed in standard SIMOX materials 12,131.)Therefore, the candidate processes to describe the time dependence of earlyflow-temperature AVmg in the SUPOX BOX materials are hole transport and electron thermal detrapping.