^{1,a)}, C. H. Seager

^{1}, D. V. Lang

^{1}, P. J. Cooper

^{1}, E. Bielejec

^{1}and J. M. Campbell

^{1}

### Abstract

We have exposed silicon bipolar transistors to fast neutrons and characterized the properties of the resulting defects using capacitance-based spectroscopy of the -type collector. We have performed low-temperature electron capture measurement of the divacancy and vacancy-oxygen defects after the samples were annealed from 350–500 K. We show from a simple rate equation analysis that one can define an unambiguous test for cluster-induced reductions of defect level occupation due to slow capture. This allows easy identification of deep level transient spectroscopy(DLTS) levels where the capture is inhibited due to band bending. Our measurements show extremely long, temperature-dependent capture times for the doubly charged state of the divacancy. We have modeled the capture dynamics as a function of annealing using a simple electrostatic band-bending approach coupled with a realistic simulation of the cluster size and shape distribution as estimated from computer simulation of the damage cascades. We find that our simulation of neutron damage combined with electrostatic modeling of the capture data, with only a limited number of adjustable parameters, fits the measured data very well. Our annealing studies indicate, however, that *isolated* divacancies (those with visible DLTS signals from two acceptor states) comprise only about of the charge in the defect cluster.

We thank Andy Armstrong, Pat Griffin, Victor Harper-Slaboszewicz, Don King, Norman Kolb, J. Kyle McDonald, Normand Modine, Sam Myers, George Vizkelethy, Bill Wampler, and Alan Wright for stimulating discussions. Sandia National Laboratories is a multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Co., for the Department of Energy under Contract No. DE-AC04094AL85000.

I. INTRODUCTION

II. EXPERIMENTAL DETAILS

III. CAPTURE MEASUREMENTS

IV. MODELING OF CARRIER CAPTURE KINETICS IN DEFECT CLUSTERS

A. Equilibrium custer potentials

B. General properties of carrier capture kinetics in a cluster environment

C. Carrier capture kinetics for a simple divacancy cluster

1. Single divacancy defect in a cluster

2. Multiple defects in a cluster

D. Effect of cluster size distribution on carrier capture measurements

E. Carrier capture of VO centers

V. DISCUSSION

VI. CONCLUSIONS

### Key Topics

- Deep level transient spectroscopy
- 78.0
- Annealing
- 65.0
- Charged clusters
- 44.0
- Point defects
- 38.0
- Electrostatics
- 29.0

## Figures

DLTS of the base collector junction of a transistor using a rate window of , filling pulse width 1 and 10 ms period after neutron damage and 30 m anneals from 350 to 500 K at 0 V bias. Prior to recording the data the junctions were subjected to forward bias for 20 m at 300 K.

DLTS of the base collector junction of a transistor using a rate window of , filling pulse width 1 and 10 ms period after neutron damage and 30 m anneals from 350 to 500 K at 0 V bias. Prior to recording the data the junctions were subjected to forward bias for 20 m at 300 K.

DLTS signal vs filling pulse width measured at the DLTS peak at several temperatures following annealing at 350 K at 0 V. Prior to recording the data the junctions were subjected to forward bias for 20 m at 300 K. The DLTS emission times, (inverse of the rate window) are shown by vertical lines.

DLTS signal vs filling pulse width measured at the DLTS peak at several temperatures following annealing at 350 K at 0 V. Prior to recording the data the junctions were subjected to forward bias for 20 m at 300 K. The DLTS emission times, (inverse of the rate window) are shown by vertical lines.

Data taken in the same manner as those of Fig. 2, but after the transistor was annealed at 500 K for 30 m at 0 V.

Data taken in the same manner as those of Fig. 2, but after the transistor was annealed at 500 K for 30 m at 0 V.

DLTS peaks measured at several rate windows after 20 m injection at 300 K. Filling pulses have been chosen to ensure the DLTS signal is saturated. The decline in peak height as the peak temperature increases indicating partial filling of the center.

DLTS peaks measured at several rate windows after 20 m injection at 300 K. Filling pulses have been chosen to ensure the DLTS signal is saturated. The decline in peak height as the peak temperature increases indicating partial filling of the center.

Evidence for nonuniform defect creation from MARLOWE calculations of neutron damage before vacancy-interstitial recombination. The recombination is expected to reduce the defect numbers by as much as a factor of 10. Approximately 18 000 vacancy-interstitial pairs in are shown along with a single vacancy cluster in .

Evidence for nonuniform defect creation from MARLOWE calculations of neutron damage before vacancy-interstitial recombination. The recombination is expected to reduce the defect numbers by as much as a factor of 10. Approximately 18 000 vacancy-interstitial pairs in are shown along with a single vacancy cluster in .

Dashed curve is the solution to the electrostatic cluster potential in the negatively charged region occupied by the defect states [Eq. (3)], and the solid curve is the solution in the region of positively charged shallow donor atoms [Eq. (2)]. Defect states lie inside . The negative-going singularity of Eq. (3) at has been omitted from the plot.

Dashed curve is the solution to the electrostatic cluster potential in the negatively charged region occupied by the defect states [Eq. (3)], and the solid curve is the solution in the region of positively charged shallow donor atoms [Eq. (2)]. Defect states lie inside . The negative-going singularity of Eq. (3) at has been omitted from the plot.

This figure illustrates changes in the DLTS signal as the temperature and barrier height are changed. The self-consistent electrostatic barrier height is found by equating the temperature-dependent charge density [Eq. (5) shown for 112 and 154 K] to the barrier height obtained from the electrostatic solution [Eq. (4), thin solid line]. In this illustrative calculation we only consider cluster charge from the defect. For the partially filled example shown, the observed DLTS signal is temperature dependent, with the magnitude of decreasing as the temperature increases. Note also that if the defect density decreases and the is partially filled, the magnitude of the observed DLTS signal will increase dramatically.

This figure illustrates changes in the DLTS signal as the temperature and barrier height are changed. The self-consistent electrostatic barrier height is found by equating the temperature-dependent charge density [Eq. (5) shown for 112 and 154 K] to the barrier height obtained from the electrostatic solution [Eq. (4), thin solid line]. In this illustrative calculation we only consider cluster charge from the defect. For the partially filled example shown, the observed DLTS signal is temperature dependent, with the magnitude of decreasing as the temperature increases. Note also that if the defect density decreases and the is partially filled, the magnitude of the observed DLTS signal will increase dramatically.

Fractional occupancy of the shallow divacancy state computed from Eq. (7) for several defect densities leading to (singly occupied) barrier heights of (a) 80 meV and (b) 50 meV. For comparison, calculated fractional occupancy for the case of a constant capture rate is shown. The electron emission time, , is shown for each temperature.

Fractional occupancy of the shallow divacancy state computed from Eq. (7) for several defect densities leading to (singly occupied) barrier heights of (a) 80 meV and (b) 50 meV. For comparison, calculated fractional occupancy for the case of a constant capture rate is shown. The electron emission time, , is shown for each temperature.

Numerically computed fits of Eq. (7) (lines) compared to capture data for a neutron-irradiated sample forward bias annealed at 300 K. The measured electron emission time, , is shown for each temperature. The fits were calculated using single barrier heights (when is singly filled) at 112 and 154 K of 60 and 50 meV, which resulted in equilibrium filling fractions of 0.64 and 0.88, respectively. The value of was at 112 K.

Numerically computed fits of Eq. (7) (lines) compared to capture data for a neutron-irradiated sample forward bias annealed at 300 K. The measured electron emission time, , is shown for each temperature. The fits were calculated using single barrier heights (when is singly filled) at 112 and 154 K of 60 and 50 meV, which resulted in equilibrium filling fractions of 0.64 and 0.88, respectively. The value of was at 112 K.

Mean cluster size following neutron irradiation as a function of number of defects in a cluster as simulated from a binary-collision approximation (MARLOWE).

Mean cluster size following neutron irradiation as a function of number of defects in a cluster as simulated from a binary-collision approximation (MARLOWE).

Barrier height for the clusters plotted in Fig. 10 as a function of the number of defects in each cluster. Numbers above each histogram bar are a weighting function that consists of the probability of finding a cluster of a particular size times the average number of defects in that cluster.

Barrier height for the clusters plotted in Fig. 10 as a function of the number of defects in each cluster. Numbers above each histogram bar are a weighting function that consists of the probability of finding a cluster of a particular size times the average number of defects in that cluster.

Capture data measured at two temperatures following anneals at 350 and 500 K. Vertical lines are the measured electron emission times at the two measurement temperatures, 112 and 154 K. The curved lines are fits to the data using the barrier height distribution shown in Fig. 11.

Capture data measured at two temperatures following anneals at 350 and 500 K. Vertical lines are the measured electron emission times at the two measurement temperatures, 112 and 154 K. The curved lines are fits to the data using the barrier height distribution shown in Fig. 11.

Numerically computed fits of Eq. (7) compared to capture data for a neutron-irradiated sample forward bias annealed at 300 K. Both the data (points) and the fits (lines) are normalized to the long pulse width equilibrium values. The measured electron emission time, , is shown for each temperature. For the calculated curves the barrier heights (when VO is neutral) at 79 and 112 K are 25 and 28 meV, respectively. The equilibrium filling fractions are 1.00 for both measurement temperatures. The value of was at 79 K, chosen to match that of Hallen (Ref. 7).

Numerically computed fits of Eq. (7) compared to capture data for a neutron-irradiated sample forward bias annealed at 300 K. Both the data (points) and the fits (lines) are normalized to the long pulse width equilibrium values. The measured electron emission time, , is shown for each temperature. For the calculated curves the barrier heights (when VO is neutral) at 79 and 112 K are 25 and 28 meV, respectively. The equilibrium filling fractions are 1.00 for both measurement temperatures. The value of was at 79 K, chosen to match that of Hallen (Ref. 7).

Article metrics loading...

Full text loading...

Commenting has been disabled for this content