Compared with bulk technology, silicon-on-insulator (SOI) technology has good immunity to single event effects (SEEs) and latchup for its fully dielectric isolation of shallow trench isolation (STI) and buried oxide (BOX) layer.^[1] Hence, it has been widely used in radiation hardening for many years. However, due to the additional buried oxide layer, the total ionizing dose (TID) effect in SOI technology is more complex than bulk technology.

In SOI technology, radiation-induced positive charges are trapped in gate oxide, STI, and BOX. In deep-submicron SOI technology, the thickness of gate oxide is below 10 nm, the radiation-induced positive trapped charges in the gate oxide is very small, which means that the effect from the trapped charges in the gate oxide is negligible.^[2] The radiation-induced positive trapped charges in the STI can reduce the threshold voltage of the STI sidewall parasitic transistor and induce the hump effect on the transfer characteristic curve of the front-gate transistor.^[3,4] The radiation-induced positive trapped charges in the BOX can deplete the body silicon film, and induce a lot of effects, such as body current lowering,^[5,6] coupling effect,^[7] drain-induced barrier lowering (DIBL),^[8,9] and so on. When the radiation-induced positive trapped charges in the BOX are enough and inverse the body silicon film, the device will suffer a large leakage current from the bottom body silicon film and fail to work.^[1,4]

In many studies of TID effect, the characteristics of the front-gate transistor and the characteristics of the back-gate transistor were both analyzed.^{[4–7,10–18]} Hence, both of them were all measured in experiment. In their papers, they only stated that the characteristics of the front-gate transistor and the characteristics of the back-gate transistor were measured before irradiation and after a certain total ionizing dose level. However, the measurement sequence was not stated clearly. Sometimes, we first measure the characteristics of the back-gate transistor for convenience. When we measure the characteristics of the back-gate transistor, the back gate is biased at a very high voltage (about 40 V–50 V). This high voltage may affect the device and influence the TID effect. Hence, if we first measure the characteristics of the back-gate transistor, then the characteristics of the front-gate transistor may be measured inaccurately. However, there is no paper which focuses on the influence of characteristic measurement sequence.

In this paper, we first study the influence of characteristics’ measurement sequence and the influence of double test on TID effect experimentally. The ON irradiation bias and Pass-Gate (PG) irradiation bias are compared with each either. An explanation to the influence is proposed. Finally, we give an appropriate measurement sequence and voltage bias in TID effect experiment to reduce the influence.

All of the devices in this paper were fabricated in 130 nm partially-deplete (PD) SOI CMOS process. The SOI wafer was 200 mm diameter UNIBOND® wafer from SOITEC corporation. The thickness of top silicon and BOX were 100 nm and 145 nm, respectively. All devices were isolated by STI structure, and TGate layout was used to contact the body silicon. Figure 1 shows the detail of the device structure. The body doping concentration was about 4.7 × 10¹⁷ cm⁻³. The input/output (I/O) NMOS devices with 6-nm gate oxide and 3.3-V operating voltage (VDD) were selected. All of the samples were 24-pin DIP ceramic packaged. And they were divided into two groups (group-front and group-back) for different measurement sequences. Radiation experiments were conducted in Xinjiang Technical Institute of Physics and Chemistry, Chinese Academy of Sciences. The ⁶⁰Co γ ray with a 100 rad(Si)/s dose rate was used as a radiation source. The ON bias (when gate is biased at 1.1 VDD, and others are grounded) and PG bias (when source and drain are biased at 1.1 VDD, others are grounded) were used during the radiation exposure. The typical characteristic curves as shown in Table 1 were measured by a Keithley 4200B parameter analyzer at room temperature before irradiation and after irradiation up to 50 krad(Si) and 100 krad(Si) respectively. As for the samples from group-front, the typical curves of the front-gate transistor were first measured, then the curves were measured the second time. When they were all finished, the typical curves of the back-gate transistor were measured the first time and the second time. As for the samples from group-back, the typical curves of the back-gate transistor were first measured, then the curves were measured the second time. When they were all finished, the typical curves of the front-gate transistor were measured the first time and the second time. What should be noted was that the second measurement was started after the first measurement has finished. To avoid experimental error and the difference among samples, we measured three samples from each group. The final characteristic curves were obtained by averaging the two samples that are most consistent with each other. All measures were taken within half an hour to avoid anneal effect.

Fig. 1.

	Figure Option ViewDownloadNew Window
	Fig. 1. (color online) Layout of PDSOI nMOSFETs, showing (a) top view, (b) side view, and (c) front view.

Table 1.

Table 1.

Table 1. Characteristic curves of PDSOI nMOSFETs and their voltage biases . Group Curve Front gate/V Drain/V Source/V Body/V Back gate/V Front gate curves Id–Vfg −0.5–3.3 0.1 0 0 0 Id–Vfg, Ib–Vfg −0.5–3.3 3.63 0 0 0 Id–Vd 3.3 0–3.3 0 0 0 Back-gate curves Id–Vbg 0 0.1 0 0 −5–45 Id–Vbg 0 3.3 0 0 −5–45

Table 1. Characteristic curves of PDSOI nMOSFETs and their voltage biases .

From Subsections 3.1–3.4, we can obtain the following results. (i) The back-gate curves’ measurement has big influence on the TID effect both under ON bias and under PG bias, whereas the front-gate curves’ measurement has a negligible influence on the TID effect both under ON bias and under PG bias. (ii) The double test of the back-gate curves’ measurement shows subdued TID effect, whereas the double test of the front-gate curves’ measurement displays no influence. From Subsection 3.5, we find that the back-gate curves’ measurement weakens the TID effect, and the more significant the TID effect, the more significant the weakening will be.

As shown in Table 1, the back gate is biased up to 45 V when we measure the back-gate curves. This large voltage induces very high electric field in the BOX. A TCAD simulation is performed to calculate the distribution of electric field during back-gate curves’ measurement. As shown in Fig. 13, when the back gate is biased up to 45 V, the electric field in the BOX rises up to 4 MV/cm, and the electric field at the bottom corner of STI reaches up to 6.5 MV/cm. According to the Nazarov et al’s study,^[20] when the electric field exceeds 5.8 MV/cm, electrons will be injected into BOX from the body silicon by Fowler–Nordheim (FN) tunneling in UNIBOND wafer. Hence, at the bottom corner of STI, the electrons can tunnel into oxide by FN tunneling mechanism during the back-gate curves’ measurement. The electrons of FN tunneling neutralizes the radiation-induced positive trapped charges, which induces the negative influence on the TID effect. Due to the low electric field in the BOX, there is no FN tunneling of the electrons in the BOX.

Fig. 13.

	Figure Option ViewDownloadNew Window
	Fig. 13. (color online) Distribution of electric field during back-gate curves’ measurement. Back gate is biased at 45 V.

However, there are anther two tunneling mechanisms called trap-assisted tunneling (TAT) and charge-assisted tunneling (CAT) which are easier to occur in SOI wafer. Figure 14 describes the two tunneling mechanisms of electrons and shows the energy band diagrams, respectively, at the bottom corner of STI and in the BOX.

Fig. 14.

	Figure Option ViewDownloadNew Window
	Fig. 14. (color online) Energy band diagrams of PDSOI nMOSFETs, respectively, (a) at the bottom corner of STI and (b) in the BOX. TAT is trap-assisted tunneling. CAT is positive charge-assisted tunneling. Yellow region is electron trap. Blue circles are positive charge centers.

Under the TAT mechanism, the electrons first tunnel into an intermediate trap (electron trap), they then tunnel out of the trap toward the anode.^[20–23] According to previous studies,^[23–25] there are many electron traps near the interface between BOX and body silicon. During the back-gate curves’ measurement, trap-assisted tunneling occurs both at the bottom corner of STI and in the BOX. The electrons from trap-assisted tunneling neutralize radiation-induced positive trapped charges, and thus weakening the TID effect. Due to the mechanism of electron traps assistant, the TAT is easier to occur than FN tunneling, which means that the TAT is more dominant than FN tunneling.

Under the CAT mechanism, the local barrier is lowered by positive charge centers.^[26,27] As shown in Fig. 14, the lower energy of the conduction band (E_c) is changed from the solid line to the dash line by the positive charge centers in the oxide. Hence, the tunneling of electrons is enhanced significantly. After irradiation, lots of positive trapped charges are induced in the STI and BOX. These positive trapped charges are the same as the positive charge centers, and thus inducing the charge-assisted tunneling. As the TID level increases, the positive trapped charges in the oxide increase, then the charge-assisted tunneling is enhanced more significantly, hence the back-gate curves’ measurement induces more negative influence on TID effect. With the assistance of mechanism of positive charge centers, the CAT is easier to occur than FN tunneling, which means that the CAT is also more dominant than FN tunneling. As the TID level increases, the positive charge centers increase significantly. This means that the CAT is more dominant than the TAT as the TID level increases, which can be proven by the bigger change of the characteristic curves as TID level increases.

Because there are higher electric fields and three kinds of tunneling mechanisms at the bottom corner of STI, the negative influence on TID effect at the bottom corner of STI is more significant. Before irradiation, FN tunneling and trap-assisted tunneling can also occur, hence the back-gate curves’ measurement can also influence the characteristics of device. However, compared with charge-assisted tunneling, FN tunneling and trap-assisted tunneling are unapparent. Hence, we do not find big changes of the characteristics of device before irradiation and we only observe a little change of hump and body current under ON bias.

In this paper, we investigate the influence of the characteristics’ measurement sequence on TID effect. We find that the back-gate curves’ measurement has big influence on TID effect, which weakens the radiation-induced hump, reduces the negative shift of threshold voltage, lowers the increment of the drain saturation current and leakage current, and subdues the radiation-induced body current lowering, We also find that the front-gate curves’ measurement shows negligible influence on TID effect.

The ON bias and PG bias are compared with each other. We find that the back-gate curves’ measurement weakens the TID effect: the more significant the TID effect, the more significant the weakening will be. Hence, the hump and drain leakage current are significantly influenced under the ON bias, whereas the threshold voltage, drain saturation current, and the body current are significantly influenced under PG bias.

In the TCAD simulation, we find that there is a high electric field at the bottom corner of STI during back-gate curves’ measurement. The electrons can tunnel into the bottom corner of STI by FN tunneling mechanism. The electric field in the BOX is not high enough. The electrons cannot tunnel into BOX by FN tunneling mechanism. However, there are electron traps and radiation-induced positive trapped charges in the oxide. The electrons can tunnel into the bottom corner of STI and BOX by trap-assisted tunneling and charge-assisted tunneling. The tunneling electrons neutralize the radiation-induced positive trapped charges, and thus weakening the TID effect. As the TID level increases, the radiation-induced positive trapped charges increase, thereby enhancing the charge-assisted tunneling and increasing the influence of back-gate curves’ measurement on TID effect. With the assistance of electron traps and positive charge centers, the TAT and CAT are more dominant than FN tunneling. As the TID level increases, the CAT is more dominant than the TAT.

To achieve accurate value of TID effect, we must first measure the front-gate curves, then measure the back-gate curves. However, a double test is not suggested because it induces a little influence on the TID effect. To weaken the influence of the back-gate curves’ measurement, the voltage biases can be appropriately lowered during measurement. Actually, we measure the transfer characteristic of the back-gate transistor to obtain the back-gate threshold voltage. The back-gate threshold voltage is below 15 V. There is no need to bias the back gate up to 45 V. If we reduce the voltage bias to 20 V, the electric field in the BOX will be lowered significantly. Hence, the influence from the back-gate curves’ measurement will be weakened.