Telegraph Noise as A Probe of Microscopic Hydrogen Motion in Amorphous Silicon
- PDF / 439,826 Bytes
- 6 Pages / 414.72 x 648 pts Page_size
- 15 Downloads / 155 Views
401 Mat. Res. Soc. Symp. Proc. Vol. 377 ©1995 Materials Research Society
probe coplanar configuration and the signal is amplified using an Ithaco 564 current preamplifier and then recorded on a Tektronix 2232 100 MHz digital storage oscilloscope. The conductance fluctuation time traces are stored, and are then fast Fourier transformed (FFT) to obtain the power spectra for the current noise. Comparison of four probe (van der Pauw geometry) and two probe measurements confirms that the noise data is not influenced by contact effects. Figure 1 shows a time record of the n-type a-Si:H film's conductance at 400 K when random telegraph switching noise is observed. RTSN in a-Si:H is observed intermittently and can persist for several hours or only appear once among many non-switching time traces. Multiple state switching is occasionally observed, but the two-state telegraph noise is far more common. This irregular behavior and the large fractional resistance change in figure 1 suggest that the conventional source of RTSN, single electron trapping, is not a plausible explanation. If charge trapping were the cause of the RTSN, then in the macroscopic a-Si:H samples this would imply that 105 charge carriers would have to simultaneously be trapped to result in a fractional resistance change on the order of 1%. Trapping due to single electron hopping at room temperature would yield switching events on the order of microseconds, not the characteristic times that are on the order of milliseconds as shown in figure 1. In addition there was no distinction between the average times spent in the up and down state at any temperature, which is in direct contrast with the known temperature dependence of average capture and emission times associated with changes in electronic occupancies of defects. Previous studies of l/f noise in a-Si:H have found that the current fluctuations are strongly non-Gaussian [8], exhibiting large correlations of the noise power between neighboring frequency octaves. The correlations of the noise power for both the RTSN and the non-switching data are quantified by considering 100 FFTs generated from the individual current time traces and calculating the correlation coefficient using the expression [9] Pij
=
NI-l(Qin - )(Qjn - )/(ij n
where Qin is the power spectra summed over an octave frequency interval i,is the average noise power per octave for that frequency range, ai is the corresponding standard deviation and the sum extends over all N = 100 power spectra. For noise statistics originating from a large number of statistically independent fluctuation processes the correlation coefficients are Gaussian, reflected in Pij being approximately zero for all different octaves. Deviations from Gaussian results indicate that the noise may be dominated by a few events within the system, or equivalently strongly correlated interactions are present [9]. As shown in figure 2 the correlation coefficients for both the RTSN and non-RTSN are strongly non-Gaussian at T = 400 K. The theoretical prediction for correlati
Data Loading...
