Figure 1. Probe station setup
BHATTACHARYA, Pradeep1, PHAN, Tri2, DABIPI, Ibibia3 & MAJLESEIN, Hamid4
124 J. B. Moore Hall, Electrical Engineering Department, Southern University, Baton Rouge, LA 70813, U.S.A.
1 bhattach@engr.subr.edu
2 trip@engr.subr.edu
3 dabipi@engr.subr.edu
4 hamid@engr.subr.edu
Abstract: The worldwide semiconductor industry is presently facing immensely difficult challenges to create specifications and solutions of process related problems as it moves into manufacturing ultra large scale integrated chips with feature size approaching 100 nm. Due to increasing integration density and the role of X-ray lithography in semiconductor research and processing, displayed by smaller microelectronic devices, it is felt students must be given practice to observe X-ray irradiation effects. The objective for the research is to test and observe the response of short channel IGFETs to various absorbed X-ray dosages. The new methods of improved lithography use high levels of absorbed X-ray energy in the oxide thus creating reliability problems. High energy X-rays(>>10 eV) create energetic particles whose side effects include the bombardment of the device interface, which further leads to the damage of the gate insulator. This leads to instability problems in the devices. In short channel devices, hot carrier charge trapping in the gate oxide occurs causing deterioration of the characteristics of IGFET. By using a wafer with chips having various short channel length devices, we performed threshold voltage studies. To obtain the results of the experiment, the electrical parameters of the short channel devices were measured in detail. By measuring the Current-Voltage(I - V) curves of the open or closed large area devices, it is possible to get the transconductance (gm) or gradient change of the short channel devices and shift of the threshold voltage from the larger devices. To perform these I-V measurements, instrument such as the HP4145B (Semiconductor Parameter Analyzer) and a manual probe station system are used. For example the average transconductance for the non-radiated device is 33.1E-6S/m and threshold voltage of 3.01V and the radiated side of the wafer a similar device has a gradient of 32.2E-6S/m V and threshold voltage of 0.926V. This shows how the exposure of the wafer to radiation does not change the transconductance of short channel devices significantly but the threshold voltages show dramatic change. However, nature of annealing also shows that there are portions of the short channel curve (device threshold voltage versus channel length curve) a reverse short channel effect is visible. It is seen in terms of a bump or a knee rather than smooth bend of the short channel curve to the anchor point of its origin. The curve should decrease in the shape of the knee at lower radiation doses, when studied for various radiation absorbed doses in the gate oxide but it does not. It shows a similar behavior when devices are annealed. This project is a constructive experience and electrical engineering elective (ELEN418) covering device physics.
Keywords: Practiced Based Engineering Education, laboratory experience
The objective of this research is to test and observe the response of short channel devices to various absorbed X-ray dosages. An important parameter in testing of these short channel devices in an enhancement mode MOS transistor is the threshold voltage1 VT which is the minimum gate voltage required to induce the channel. By obtaining the measurements of the (I - V) Current - Voltage curves of the open and closed large area devices, it is possible to get the transconductance (gm = ¶ I/¶ V) or gradient change of the short channel devices and shift of the threshold voltage from the larger devices.
The equation for the threshold voltage is given by:
VT = Φms - (Qi/Ci) - (Qd/Ci)+(2*ΦF)
where:
VT is the threshold or turn on voltage of the device
Φms is the metal-semiconductor (work function) potential difference (V)
Qi is the effective MOS interface charge per unit area (C/cm2)
Qd is the charge per unit area in the depletion region (C/cm2)
Ci is the insulator capacitance per unit area (F/cm2)
ΦF is a surface potential equal to (Ei - EF)/q(V)
Thus the voltage required to create strong inversion must be large enough to first achieve the flat band condition ( Φms and (Qi/Ci), then accommodate the charge in depletion region (Qd/Ci) and finally induce the inverted region (2*ΦF) ). This equation accounts for the dominant threshold voltage effects in typical MOS devices. It can be used for both n-type and p-type substrates if appropriate signs are included for each term.
Figure 1. Probe station setup
For the experimental procedure, a Hewlett Packard 4145B Semiconductor Parameter Analyzer is used. Before one receives the correct data of the measurement the Analyzer must be set to the proper channel division. A clear setup of the components for the I- V measurement is shown in Figure 1.
The experimental setup includes a (Micromanipulator) semi-automatic probe station, a semiconductor parametric analyzer, and optically processed wafers, which can be used to verify the final annealing method in terms of threshold voltage shift of IGFET devices. The probes are connected to the corresponding source, drain, and gate terminals of an IGFET on the processed wafer, and provide the contact for the I- V stimulus through a personality module.
Wafer #21 was the particular silicon wafer used in the research. There are an estimated one hundred and twenty chips on the wafer. The wafer has two sections of an open ended (W/L: 50:1000) and a closed device (W/L: 50:1000) located on the extreme right of each chip. On the left side of these lie a series of short-channel devices gradually decreasing in channel length from right to left.
These devices were n-channel enhancement mode devices on 0.1 ohm-cm Si (100) wafers. Each wafer is divided into four quadrants. Then (I - V) measurement method (in dark) was used by varying the gate from 0V to 10V on various control and damaged IGFETs to obtain threshold voltages near flatband conditions. This technique can be used in a simple manageable form by using a ramp up power supply for VG, a sensitive dc-supply for the drain bias and two simple analog low current integrator or current meters for measuring current sweep. These I_V values will have to be plotted using an X-Y recorder.
To prepare a processed wafer optimally one has to anneal the interface. A conversion of charge in terms of injection of hole and electron densities must be initiated. The devices on the non-irradiated part of the wafer will be injected by hot electrons to determine pre-irradiation fixed positive charge.
The annealing will take place in a convection type microwave with its ambient at temperatures that range typically from 400&grad; - 450&grad;F. The microwave system uses 3 GHz of microwave power in a standing wave pattern inside its heating enclosure that is irradiated on silicon wafers with oxide and metals over layers. The microwave is first allowed to heat to a required stipulated temperature is reached and then a wafer is heated for about twenty-five minutes in the gas ambient. The typical gas ambient was 90% Argon and 10% Oxygen. After that interval of time, new I - V measurements are taken on the annealed wafer using the same FETs as before.
Finally, the current - voltage characteristics (in dark) for both the annealed cases are compared and analyzed. These measurements, over the entire wafer, will give results on the gate oxide thickness across the wafer and its processing conditions. Shown in (Figure 2.) is actual measurement of an I - V curve of a radiated wafer after being exposed to about 6 MRads. Calculations of the absorbed dose is made using Beer -Lambert's Law and the corresponding X-ray mass-absorption coefficients and thickness of each layer. No partial back-reflection from any adjacent layer is included in producing the damage. The dose of absorbed energy in the oxide layer is considered as irradiated dose and not the total dose. However, as the X-ray beam of the
Center for Advanced Microstructures is not monochromatic and we get only total absorbed dose using a Silicon diode, we had to calibrate this for a single wavelength.
Figure 2. I-V Curve of an irradiated insulated gate field effect transistor
(of the same device)
Sample measurement of the irradiated left-hand side short-channel devices is shown below.
| GRAD | 1/GRAD | X-INT | Y-INT | |
| LINE 1 | 7.80E-06 |
1.28E+05 |
7.32E-01 |
-5.71E-04 |
Sample measurement of the non-radiated right hand side short-channel devices is shown below.
| GRAD | 1/GRAD | X-INT | Y-INT | |
| LINE 1 | 8.93E-06 | 1.12E+05 | 3.18E+00 | -2.84E-05 |
The threshold voltage for a short-channeled non-radiated sample device, which is 3.18 Volts, is higher than threshold voltage of the radiated sample device, which is 0.732Volts. This implies a shift to the left on the I - V graph. Shown in Figure 3. Normal un-irradiated and unannealed devices usually show a maximum of 10 mV of threshold voltage shift for fixed positive charge traps and a bit larger for neutral electron traps (60-120 mV) and approximately same is the shift for fixed negative charge as seen in fixed positive charge case. Statistically, we found a minor shift of threshold voltage2 due to geometry effects of 5.9% (between open and closed devices) before irradiation. Post irradiation shift was found to increase to 6.4%. The process of annealing or the treatment of heat to let the devices be able to heal themselves of the damage of the irradiation was performed subsequently. Using convection microwave source at 450°F for around twenty minutes to heat the wafer in an inert gas (Ar) does the annealing. The results showed that after annealing the decrease in threshold voltage was reduced to 2%. This was due to residual defect or fixed charges in annealed and irradiated devices. The same was found for change in transconductance measurement but the change was larger ~34%. This goes to suggest that transconductance recovery is a better tool to analyze device recovery.
Figure 3. Threshold Voltage versus Channel length plots for various doses
Examination of the data reveals a shift in the threshold voltage3 (x-intercepts). This shift is due to the irradiation process, and this sampling can be used to illustrate that all the results of the various devices will have the same results. The original graphs have their X-intercepts further to the right hand side, whereas the radiated graphs shift to the left. It can also be seen that the Y-intercept of the original graphs is further down whereas the radiated graphs have a shift further upwards. This would be seen in sub-threshold studies. It is also seen that (Fig.3) the knee of the short channel curves bulge up just before they dip to the origin showing that at shortest channel lengths they show a tendency of reverse short channel effect. This effect was seen increasing from the bunch of devices that were unirradiated (series 1) to series 2, which equated to an irradiation dose of 6 MRads. It further increased for the devices irradiated to 12 MRads of CAMD non-ionizing X-rays. More experiments are going to be performed, say up to a dose level of 100 MRads, in order to differentiate this effect from short channel effect.