Volume 6 Issue 1 - October 10, 2008
Off-State Avalanche Breakdown Induced On-Resistance Degradation in Lateral DMOS Transistors
Jone F. Chen* and J. R. Lee

Institute of Microelectronics, Department of Electrical Engineering

IEEE Electron Device Letters, Vol. 28, No. 11, pp. 1033-1035, November 2007

Font Normal   Font Enlarge
Recently, lateral DMOS (LDMOS) transistors are widely used in medium-voltage smart-power applications because they are easily integrated into the mature standard CMOS process. When the LDMOS device is used in switching applications, the device experiences avalanche breakdown during on-state to off-state transient because of unclamped inductive load. The maximum electric field of the LDMOS device in avalanche occurs at the Si/SiO2 interface near the drain side. Even if a rugged device survives breakdown initially, repeatedly operating under avalanche may gradually wear out the device because of high electric field near the drain. To evaluate this reliability concern, this work presents the on-resistance (Ron) degradation of LDMOS transistors caused by avalanche breakdown.

Devices investigated in this study are n-type LDMOS transistors fabricated by a 0.35 μm CMOS compatible technology. The cross section of the device, potential contour, impact ionization contour, and electric field distribution biased under avalanche breakdown is shown in Fig. 1. The channel region (Lch) and accumulation region (Lacc) are also indicated in the figure. The device has the following characteristics: the typical operating drain voltage (Vds) and gate voltage (Vgs) is 12 V, off-state breakdown voltage is 19.3 V, and Ron is 7 mΩ-mm2. To evaluate the damage created during fast on-off transient, constant-current pulse stressing is applied to the drain and grounding the source, gate, and p-channel terminals. The stressing is performed at room temperature with various current levels ranging from 1μa to 1 mA. The duration of each pulse is 0.1 s and the number of pulses is 100. During the stressing, Ron is monitored periodically. In addition, charge pumping current (ICP) is measured to extract stress-induced interface state density (ΔNit) and oxide trap density (ΔNot). The pulse in charge pumping measurement is applied to the gate while the drain and p-channel terminals are grounded. The source terminal is floating because the damage near the source side is negligible. The amplitude of the pulse is fixed at 10 V and the base voltage (Vbase) sweeps from -8 V to 0 V under a frequency of 500 kHz. Process (TSUPREM4) and device (Medici) simulations are also performed to investigate the degradation mechanism under avalanche breakdown condition.
Fig. 1. The device cross section and TCAD simulation results on potential contour (line), impact ionization contour (field), and electric field (vector) of our LDMOS device under avalanche breakdown condition.

The relationship between Ron degradation and the number of current pulse under various current levels is shown in Fig. 2 and two observations are found. First, higher current level produces larger Ron degradation at the beginning but eventually Ron degradation saturates at roughly 14%. Second, the magnitude of Ron degradation is closely related to the product of current level and the number of pulses. For instance, the Ron degradation under 10 stressing for 10 pulses is roughly equal to the Ron degradation under 100 stressing for 1 pulse. Such a result reveals that avalanche-breakdown-induced damage is determined by the total charge flowing through the drain. To identify the mechanism of breakdown-induced Ron degradation, charge pumping measurement and TCAD simulation are performed. Fig. 3 shows the ICP data of the device before and after 10μa current pulse stressing. Based on TCAD simulation results of flat-band voltage and threshold voltage along Si/SiO2 interface, it can be found that ΔNit located in the Lch, Lacc, and spacer region can be sensed when Vbase = -8 V. At Vbase = -4 V, however, only ΔNit created in the Lch and Lacc region can be probed. According to ICP data in Fig. 3, little increase of ICP (ΔICP) at Vbase = -4 V but significant ΔICP at Vbase = -8 V indicates that most of the ΔNit appears under the spacer region. To confirm the location of ΔNit, lateral distribution of ΔNit is extracted from ΔICP data and the result is depicted in the inset of Fig. 3. It shows that ΔNit is distributed under the spacer region centered at 70 nm away from the poly-gate edge and the width of distribution is less than 20 nm.
Fig. 3. ICP data of the device before and after 10μa current stressing for 100 pulses. ΔNit distribution is shown in the inset. The origin of X-axis in the inset is the poly-gate edge.
Fig. 2. Ron degradation as a function of the number of current pulse under various current levels.

As the number of stress pulse increases, in addition to ΔNit, a left shift in the ICP vs. Vbase characteristics indicates that positive oxide-trapped charges (ΔNot) are also created. The density of oxide-trapped charges can be extracted and the peak value of ΔNit distribution and ΔNot distribution as a function of the number of current pulse is drawn in Fig. 4 and two phenomena are evident. First, ΔNit is roughly 10 times greater than ΔNot, suggesting that the main mechanism responsible for Ron degradation is interface state generation. Second, both ΔNit and ΔNot saturate as the number of current pulse increases. This trend is identical to the saturation of Ron degradation as seen in Fig. 2. The results in Figs. 1 – 4 suggest the mechanisms responsible for breakdown-induced Ron degradation are as follows. Avalanche breakdown results in significant impact ionization near the drain. The electrons generated by impact ionization flow to the drain. Most of the holes created by impact ionization go to p-channel. However, some of the holes can inject into the gate oxide because the direction of the electric field near the drain is from drain toward the gate as in Fig. 1. Such hole injection leads to interface state generation and positive oxide-trapped charges. Although positive ΔNot can attract electrons under the spacer and reduce series resistance, ΔNit is much greater and leading to Ron degradation. Since ΔNit is distributed in a narrow region, the available interface trap sites are limited, leading to the saturation of ΔNit. Positive ΔNot also has the tendency to saturate as in Fig. 4. The saturation of ΔNit and ΔNot is responsible for the saturation of breakdown-induced Ron degradation.
Fig. 4. The peak value of ΔNit and ΔNot as a function of the number of current pulse.

In conclusion, Ron degradation in LDMOS transistors operated under avalanche breakdown has been discussed. Interface state generation is the main degradation mechanism. TCAD simulation suggests that the driving force of damage is breakdown-induced hole injection near the drain. The degradation of Ron saturates because of the saturation in interface state generation. Since the device experiences avalanche breakdown during unclamped inductive switching, breakdown-induced Ron degradation should be taken into consideration during product design and manufacturing.
< previousnext >
Copyright National Cheng Kung University