Manuscript Received
5-31-89

Laser-Induced Changes in the Electrical Performance of

Silicon MOS Device Structures

Chen-Zhi Zhang, Steve E. Watkins, Rodger M. Walser and Michael F. Becker
Center for Material Science and Engineering and
Department of Electrical and Computer Engineering
The University of Texas at Austin
Austin, TX 78712

Laser-induced morphological damage and electrical changes in the basic silicon MOS structure have been investigated using Q-switched Nd:YAG laser irradiation at 1064 nm. The threshold values for the onset of morphological damage to poly-silicon and Al layers have been obtained. It was found that the two polysilicon films have different laser-induced damage thresholds. Both 1-on-1 and N-on-1 electrical damage behavior were studied for sheet resistors and MOS capacitors fabricated from the poly-silicon films. Severe electrical parameter changes were observed at fluences below the onset of visible morphological damage. Poly-silicon gate MOSFETs have been characterized in terms of leakage current, transconductance change, and threshold voltage shift. The experiments indicated that changes in these parameters after laser irradiation at low fluence were probably due to changes in poly-silicon gate morphology. Drain/source edge junctions which were not covered by the gate were also very sensitive to the effects of transient laser heating. Laserinduced defects are thought to be introduced into this region.

Key words: laser damage, MOS structure, electrical degradation, morphological damage.

1. Introduction

It is well known that the silicon MOS structure is a basic unit in MOS integrated circuits and CCD arrays, which are widely used to perform logic, memory, and imaging functions [1]. Many special applications such as communication, military and space systems require these devices to be operated in laser environments, where it becomes important to understand their laser damage processes [2]. A series of tests have been conducted to investigate both the morphological and electrical effects of pulsed laser irradiation on poly-silicon, aluminum film, and MOS structures. We have also correlated surface damage behavior with the electrical degradation of MOS devices at different laser fluence levels.

The test samples were poly-silicon gate, n-channel and buried-channel MOSFETs, as shown in figure 1. For imaging use, poly-silicon is chosen as the material for the gate of most devices since it is semi-transparent for visible light. The samples were irradiated with 1064 nm, 10 ns pulsed laser radiation from a Nd:YAG laser. The low frequency performance of MOSFETs for both analog and digital applications is characterized in terms of leakage current, transconductance, and threshold voltage both before and after laser irradiation.

In this paper, we first describe the devices under test and the Nd:YAG laser system in some detail, then report the results of experiments on laser-induced morphological changes to poly-silicon and Al films. Next we describe experiments on laser-induced changes to the electrical properties of poly-silicon sheet resistors and of MOSFETs which have various ratios of channel length to width. The experimental relation of morphological damage to electrical degradation at various laser fluences is given. Finally, a concluding summary of laser damage to MOS structures is given.

2. Experiments

The samples had been fabricated on 100 mm diameter Si wafers by standard MOS-LSI process which include LOCOS (Local Oxidation of Silicon) field oxidation, poly-silicon deposition and Al lines for electrical connection. The devices were buried-channel and n-channel TDI CCD arrays of 2048x96 elements. Each wafer contained additional test and characterization devices including:

(1) resistor bars of poly-silicon, 960 x 60 μm2 and 960 x 10 μm2 in size,

(2) serpentines of poly-silicon and n+ wells which have square resistance of 70 ohm/sq (10 mm width),

(3) poly-silicon capacitors, 350 x 350 μm2 and 500 x 620 μm2 in size, and

(4) n-channel and buried-channel MOSFETs which have varying ratios of channel width to length, i.e., W/L = 54...3/6, W/L 36/36...3 (see fig. 2). Some of the MOSFETs were shielded by Al and polyimide layers.

=

Figure 3 shows the laser test system. The laser source for these experiment was a Q-switched Nd:YAG laser operating at 1064 nm with 10 ns pulses (FWHM) and with a 10 Hz repetition rate. The incident energy on the samples was controlled by an attenuator consisting of a rotating half-wave plate followed by a fixed, thin-film polarizer and

monitored by a reference energy meter. The standard deviation of the pulse-to-pulse energies was two percent or less. The spatial profile of the Gaussian beam was measured by the scanning slit method and the FWHM was determined. The 1/e2 beam-spot radius wo was obtained using the equation 2wo = FWHM (In2/2)1/2. The beam was focused to an approximately 300 μm spot radius. The peak-on-axis irradiation fluences were calculated using F = 2Е/лw。2, where E is the pulse energy. The reference energy meter was calibrated and the spot size was determined during each test session.

In the damage testing process, the samples were positioned using a He-Ne beam and a 20x alignment microscope. The computer recorded the energy of each incident pulse, controlled the shutter which blocked the beam after the desired number of pulses, and calculated the peak-on-axis fluence. The alignment microscope was used to check for gross surface damage. Details of the morphological damage were observed subsequently with a Nomarski microscope and a SEM. Morphological damage is defined as any visible change induced by laser irradiation using the 50 - 500x Nomarski optical microscope.

Before irradiation and at each of the intermediate levels of fluence, the following measurements were performed on the MOSFET samples using an HP4145B Semiconductor Parameter Analyzer:

(1) Breakdown voltage of the source and drain well junctions at various values of Vg.

(2) Output characteristics, i.e., Id vs. Vd curves for various values of Vg.

(3) Transfer curves, Id vs. Vg at several values of Vsub.

(4) Transconductance curves, Gm vs. Vg at several values of Vsub.

3. Results and Discussion

The results showing the onset of laser-induced morphological damage at several major stages of successive laser damage to the poly-silicon and aluminum resistor and capacitor patterns are listed in tables I and II. Laser damage threshold values are defined as the fluence at which a color change (surface height change) can be detected by Nomarski microscopy for poly-silicon films and at which surface roughness occurs for Al films. A few general observations can be made. First, poly-1 (poly-silicon layer 1) films had a lower damage threshold than poly-2 films for all geometries. Second, the damage threshold for aluminum lines and plates fell generally close to that for poly-1. Third, the N=10 pulse thresholds are all about 20-40% lower than the corresponding one shot thresholds for the same test structure. Finally, all of the capacitor plate structures had higher damage thresholds than the line (or serpentine) structures. The latter is presumably due to the more effective thermal diffusion in the 2-D plates as compared to the 1-D linear structures, thus allowing the linear structures to reach melting temperature with less fluence. Figures 4 (a) - (d) display typical near-threshold damage morphologies for the poly-silicon capacitor plates.

Another planar structure tested was shielding layers of aluminum and polyimide. In some cases, active devices were covered by this optical shielding layer pair. In other cases, three layer pair structures were damage tested with multiple laser pulses. We found that the aluminum-polyimide layer-pairs made good ablating shields for high fluence pulses. Each layer pair provided good shielding for 1 shot, up to fluences of about 5 J/cm2. MOSFET devices showed no significant change in static electrical performance even after the top shielding layer had been stripped away by laser irradiation. Multiple layer pairs gave corresponding protection from multiple pulses. Only one layer pair was removed per pulse so that the three layer-pair structures gave effective protection from three relatively high fluence pulses. The related damage morphologies are shown in figures 4 (e) and (f). The aluminum is just beginning to melt in figure 4 (e) while the layer pair has been stripped from above a set of interconnect lines in figure 4 (f).

Laser-induced electrical parameter changes were examined for both the linear resistor structures described above and for MOSFET devices. Changes in sheet resistance were measured as a function of laser fluence for the poly-silicon films and the n+ well resistors, and the results are shown in figures 5 (a) and (b). For poly-silicon, the sheet resistance decreased slightly at very low fluences, starting at about 0.2 J/cm2, and then decreased significantly at about 0.4 J/cm2. All of these changes occurred below the visible damage onset fluence at around 1.0 J/cm2. This decrease may be due to grain growth by fast laser heating or due to the creation of electrically active impurities or defects. Successive laser pulses continued this trend, as shown in figure 5 (c). The sheet resistance for the n+ wells only decreased by about 0.2%, and this small change may not be experimentally significant. The relatively small resistivity changes observed in the poly-silicon as well as the n+ wells would have a negligible effect on the electrical performance of a MOSFET if this were the only change taking place. In experiments on MOSFETs, other changes appear to take place at fluences above 0.2 J/cm2, and they are described in the next section.

Electrical measurements on MOSFETs included leakage current (both source/drain to substrate and source to drain), transconductance, and threshold voltage. Leakage current as a function of laser fluence is shown in figure 6. Leakage from the drain/source wells to the substrate seems to dominate drain to source leakage, as shown in figure 6 (a). The intersection of the surface and the edge of the source or drain depletion region is both sensitive to defects and exposed to the full laser fluence. Laser induced defects at this point can significantly increase the surface leakage current. Close inspection of the data reveals a slight improvement in the leakage current at fluences around 0.4 J/cm2, well below the morphological damage threshold. This leakage reduction may be related to modification of the SiO2/Si

interface states at the junction intersection with the surface. The subsequent rapid increase in the substrate leakage with increasing fluence could easily result in device failure. This surface-junction intersection can be protected by a polysilicon layer, as in the CCD active area or for a closed gate design FET.

The onset fluence for drain to source leakage was about 0.4 J/cm2 as shown in figure 6 (b). This leakage appeared to be insensitive to the gate width or length as all samples showed similar behavior. The threshold fluence for changes in electrical properties was always much lower than the onset fluence for morphological damage to the same device. Leakage current was observed to increase as fluence increased until the poly-silicon gate was open circuited. During this process, the poly-silicon gates were observed to successively suffer color change (poly-1 only), edge erosion, melting or shrinkage, and breaking, as summarized in tables I and II. The edge erosion can be observed in figure 2 in the entire top row of MOSFETs. In addition, this figure shows that the shortest gate device (uppermost right) has punctured the cap oxide over the gate, presumably by melting and boiling of the poly-silicon below.

Degradation of the transconductance and threshold voltage shifts were observed at generally the same fluence levels as laser-induced leakage, as shown in figures 7 and 8. The measured changes in these parameters were greater for larger ratios of W/L, that is, for shorter channel lengths. The larger area gates suffered less laser-induced change in transconductance and threshold voltage. In addition, the polarity of AVth behaved irregularly. In some samples it was positive, but for others it was negative. We deduce that the laser pulses have the less effect on the interface states behind the gate area, due to strong optical absorption in the poly-silicon gate, than they would in an uncovered region. That leaves changes in the poly-silicon gates as being largely responsible for the transconductance reduction and change in threshold voltage in MOSFETs. In practice, a reduction in Gm causes more serious degradation in analog electrical performance than in digital performance. The laser-induced changes in Ids and Vth below the onset for visible damage may well be considered as acceptable for digital applications such as CCD arrays and their associated logic circuitry.

4. Conclusions

Experiments have been conducted to investigate both the morphological and electrical effects of pulsed laser irradiation on poly-silicon films and interconnects, aluminum films, and MOS transistors. In most cases, as the laser fluence increased from around 0.7 J/cm2 to 1.1 J.cm2, morphological damage appeared in the form of color change (thickness change), gate edge erosion or damage pits, melting or shrinkage, and breaking of the poly-silicon gates and aluminum lines. For shorter gates and other narrow structures, laser damage thresholds at the lower end of this range were observed. Also the poly-2 layer had a significantly higher damage threshold than either the poly-1 or the aluminum interconnects. This is attributed to processing differences which also resulted in poly-2 having a 40% higher sheet resistance than poly-1. The sheet resistance of the poly-silicon layers was found to decrease at fluences below the onset of visible damage. For poly-silicon interconnects, the sheet resistance decreased significantly above 0.4 J/cm2 up to the melting and open circuiting of the interconnect at 1.1 to 1.3 J/cm2. No corresponding change was observed in the n+ well resistors fabricated in the base wafer material. Grain growth and defect activation are the suggested mechanisms for this resistivity decrease.

Changes in several electrical parameters of MOSFETs were detected at fluences below the onset of visible morphological damage. The drain/source edge junction regions not covered by the gate, the laser induced leakage current to the substrate increased rapidly above 0.4 J/cm2. The surface termination of these p-n junctions was particularly vulnerable to laser-induced damage resulting from surface defects produced during the laser heating transient. Degradation of the electrical performance of MOSFETs also became significant at 0.4 J/cm2. Increased leakage, Ids, decreased transconductance, Gm, and threshold voltage shifts, AVth, were all observed at fluences below the onset of visible damage. Electrical degradation increased as the poly-silicon melted, and eventually, the poly-silicon connections were open circuited rendering the device non-functioning. At fluences below those that caused catastrophic failure, the electrical parameter changes are probably not severe enough to render the MOS device useless except in critical applications.

This work was supported in part by Acurex Corporation /Aerotherm Division.

5. References

[1] S. M. Sze, Physicals of Semiconductor Devices, 2nd Ed., 1981.

[2] M.F. Becker, C.-Z. Zhang, S.E. Watkins, R.M. Walser: "Laser-Induced Damage to Silicon CCD Imaging Sensors", Proceedings of SPIE 1105, (1989).

[3] H.-Y. Tsoi, J.P. Ellul, M.I. King, J.J. White, and W.C.Bradley, "A Deep-Depletion CCD Imager for Soft XRay, Visible, and Near-infrared Sensing," IEEE Trans. on Elect. Dev. ED-32, 1525 (1985).

[4] J.P.Ellul, H.-Y. Tsoi, J.J. White, and W.C. Bradley," State-of-the Art Imaging Arrays and Their Applications," Proceedings of SPIE 501, 117 (1984).