Determination of interface trap capture cross sections using three-level charge pumping

A modified three-voltage-level charge pumping (CP) technique is described for measuring interface trap parameters in MOSFETs. Charge pumping (CP) is a technique for studying traps at the Si-SiO/sub 2/ interface in MOS transistors. In the CP technique, a pulse is applied to the gate of the MOSFET which alternately fills the traps with electrons and holes, thereby causing a recombination current I/sub cp/ to flow in the substrate. With this technique, interface trap capture cross sections for both electrons and holes may be determined as a function of trap energy in a single device. It is demonstrated that a modified three-level charge pumping method may be used to determine not only interface trap densities but also to capture cross sections as a function of trap energy. The trap parameters are obtained for both electrons and holes using a single MOSFET.

was supported by the Office of Naval Research.

20375.
1 k - (a) The three-level CP pulse. The bias levels above VACC and below VINV are used as in standard charge pumping [l]. Using three-level CP, the duration ( t p ) and voltage ( V , ) of the third level are varied to determine trap parameters. (b) Diagram of interface trap occupancy after filling with electrons. At long times r e , the traps are filled to levels E,, or Et2 depending on bias levels Vel and Vez. dependent of energy. We emphasize that these capture cross sections are not obtained using previous three-level CP meth- The test devices are p-channel MOSFET's with 10-pm gate length, 100-pm width, and a 48-nm gate oxide. The threelevel pulse used here is identical to that of Tseng but with both V , (as in [2]) and t , (as in [3]) variable. The pulse is applied to the MOSFET gate and I,, measured at the substrate [l]. A 100-Hz pulse with 50-ns rise and fall times was used to explore a wide range of trap time constants.
Experimental I,, data obtained as a function of t, for different values of Ve are shown in Fig. 2. At small t,, I,, decreases approximately as In (te). This behavior is expected since [l], [2] ods [2]-[6].
where k is Boltzman's constant and T i s the absolute temperature. If a is independent of E t , this reduces to 1 dI,, As discussed above, when te is long, the traps above the Fermi level determined by V , emit their electrons. Therefore, the traps reach equilibrium, and I,, saturates, at long times, as observed in Fig. 2. That a clean saturation characteristic is indeed obtained indicates that the trap levels are associated with a single emission time and thus each is characterized by a single value of U . Dit may then be determined using Tseng's method [2] from the variation of saturated I,, with V , using (1).
U.S. Government work not protected by U.S. copyright  [l], while the triangles are obtained using our improved three-level CP technique. The disagreement observed for electron traps above midgap occurs because U, is not energy independent as assumed in [l]. equal to q / k T (dashed line in Fig. 3). This difference may be caused by experimental error (although the apparent error is larger than we can account for), or it may be due to the energy dependence of 0 , . ue is found to be surprisingly large (~1 x cm2), and it decreases towards the conduction-band edge. This result does not agree with repofid here have b e n obtained on a single set of devices and may not be representative of "typical" devices.
In Fig. 3, we observe that the experimentally determined energy at which electron and hole emission times are equal occurs well below is considerably larger than u h . The offset is calculated from

( k T / 2 q ) In (uh/ue) [ l o ] . Using near-midgap U values, an
Our improved technique is based on the essential fact that the time at which saturation first occurs is given by t t = r e , where This offset is due to the fact that _ _ (4) offset of -0.10 eV is obtained, in excellent agreement with Fig. 3. and Ujh is the electron thermal velocity, ni is the intrinsic carrier density, and 4, is measured with respect to midgap. T, (for energies not too close to midgap) is obtained by graphical extrapolation as shown in Fig. 2. Values for 4, are determined within f0.05 V from V , by standard methods [ 2 ] .
Electron and hole emission times 7, are shown in Fig. 3 as a function of 4,. Using these data and (4), U , and uh may readily be determined as a function of energy (Fig.   4). The hole emission time depends exponentially on 4, with a slope of g / k T , in agreement with (3). uh is small (xl x cm2) and independent of energy, in reasonable agreement with previous work using ac conductance on capacitors ( u h = 2-4 x cm2 for (100) Fig. 4. Using the three-level CP technique with (1) (triangles), trap densities are found to be small (6-7 x lo9 traps/cm2.eV), nearly independent of energy, and about the same for electrons and holes. The average Dit values obtained from the slopes of I,, versus ln(t,) using (3) are shown by the dashed lines in Fig. 4. Agreement between the two methods is excellent for the hole traps below midgap, but differ significantly above midgap. This discrepancy occurs because is not independent of energy as assumed in the derivation of (3). If the actual energy dependence of U, is taken into account using (2), then Dit is about 30% smaller and is in much better agreement with the data. This comparison shows that average Dit values obtained by the standard CP approach [l] may have large errors when U varies rapidly with energy.
In summary, we have shown that a modified three-level charge pumping method may be used to determine not only interface trap densities as in [2] but also capture cross sections as a function of trap energy. The trap parameters are obtained for both electrons and holes using a single MOS-FET. This is a significant advantage over the conventional ac conductance technique [7], which is limited to one-half of the bandgap. The values obtained for (Th are in good agreement with previous results. However, values obtained for U, have a surprisingly large magnitude and strong energy dependence. Clearly, further experiments, including direct comparisons to ac conductance as in [9], are warranted.