DLTS Study of Oxide Traps Near the Si-SiO 2 Interface
- PDF / 368,806 Bytes
- 6 Pages / 420.48 x 639 pts Page_size
- 89 Downloads / 194 Views
Mat. Res. Soc. Symp. Proc. Vol. 54. 1986 Materials Research Society
588
Vp=Vpo
S
Vb=-lV
20
12-
20-
9
15
j P.le
Isec
z
._0 100K.oeV)
~
~
*~ 3
Vpo
Vb=-IV
______________50K
150K
5
0
_
Stp=20psec
(V)
Fig. 1. Evolution of DLTS response versus the pulse amplitude. The interface states density determined 1.6xlO'O cm- 22 eV7 11 at V.0 are 9.6xi099 cm- 2 eV(at 0.19 eV) 4.5x10 cm- eV-1 (at 0.3 eV) (at 0.4 eV) .Dotted lines are the saturation behaviors.
..
,..,
0 200K
10
05
FILLING PULSE
152K3V)10 10 1K
, "0
.01
1 PULSE WIDTH (msec)
10
Fig. 2. Evolution of the DLTS response versus the pulse width.
gives a description of experimental procedure and theoretical necessary to obtain a characterization of these oxide traps.
model
INFLUENCE OF OXIDE TRAPS The measurements are made on state-of-the art MOS capacitors which exhibit 1 2 a very low fast interface states density (3xlO9 cm- ev- at mid-gap). The thermal oxide of 1200 A thickness is grown at 950oC under wet atmosphere on oriented n-type 6-15 fl.cm Silicon substrate, followed by a N2 annealing. The ohmic contact on the backside faces is achieved by phosphorous implantation followed by annealing at 950°C for one hour. The influence of oxide traps on the DLTS response is shown on figure 1 and 2. The variations of the DLTS response versus the filling pulse amplitude (figure 1) is given at various temperatures which correspond to differents energy values in the Si gap. Though the pulse width (here 20t•s) is larger than the capture time constant of fast interface states, the DLTS response does not saturate when the filling pulse bias the MOS i.e. when the fast interface states are capacitor in accumulation, completely filled. The dotted lines in figure 1 are the usual saturation behaviors, and between Vp = 0 and a threshold value Vpo, the variations are due to incomplete filling of fast interface states. We have observed
589
the same unusual behavior with the variations of the pulse width t. (figure 2). We have kept Vp = Vp, in these measurements in order to minimize the influence of the filling pulse amplitude. For tv smaller than is performed due to limitation of our DLTS 20jis, no measurement apparaturs, and above 5004s the divergence is due to the behavior of the lock-in detection when the ratio tp/P of the pulse width over the period increases E83. CALCULATION OF THE DLTS RESPONSE WITH OXIDE TRAPS We have include in the classical DLTS calculation the possibility of electron tunneling between the oxide traps and the Si conduction band. We have taken into account the simple case of a square barrier, neglecting the slope of the S1o 2 conduction band which is less than few hundred mV with the electric fields applied on our samples. Such a value is negligible compared to the barrier height. We have used the classical tunneling emission rate [91, but the pre-exponential factor a is not well defined. A simple model using the effective mass approximation to calculate the tunneling transition probability is given in reference [101. Now, if we
Data Loading...
