† Corresponding author. E-mail:
The influence of characteristics’ measurement sequence on total ionizing dose effect in partially-depleted SOI nMOSFET is comprehensively studied. We find that measuring the front-gate curves has no influence on total ionizing dose effect. However, the back-gate curves’ measurement has a great influence on total ionizing dose effect due to high electric field in the buried oxide during measuring. In this paper, we analyze their mechanisms and we find that there are three kinds of electrons tunneling mechanisms at the bottom corner of the shallow trench isolation and in the buried oxide during the back-gate curves’ measurement, which are: Fowler–Nordheim tunneling, trap-assisted tunneling, and charge-assisted tunneling. The tunneling electrons neutralize the radiation-induced positive trapped charges, which weakens the total ionizing dose effect. As the total ionizing dose level increases, the charge-assisted tunneling is enhanced by the radiation-induced positive trapped charges. Hence, the influence of the back-gate curves’ measurement is enhanced as the total ionizing dose level increases. Different irradiation biases are compared with each other. An appropriate measurement sequence and voltage bias are proposed to eliminate the influence of measurement.
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
Figure
Figure
Figure
We also extract the maximum of body current from Fig.
From the analysis of the front-gate body current, we can obtain the similar conclusions to those presented in the Subsection
Figure
The drain leakage currents under ON bias and PG bias are also studied as shown in Fig.
Figure
The threshold voltages of the back-gate main transistor are also extracted as shown in Fig.
From the previous sections, we find the back-gate curves’ measurement has big influence on TID effect. In this subsection, we compare the influences of the back-gate curves’ measurement under ON bias and PG bias.
As for the front-gate parameters, there is no obvious difference between the first test and the second test, hence we only choose the front-gate parameters from the first test. We calculate the change rate of the front-gate threshold voltage (Vth, fg), the maximum of the body current (Ib, max), and the drain saturation current (Idsat) by using the following equation:
Figure
Figure
As for the back-gate parameters, the influence of the back-gate curves’ measurement is reflected in the difference between the first test and the second test. Hence, the change rate of the back-gate parameter Vth, bg is calculated from
What is more, from Figs.
According to previous studies,[1,6,16] the irradiation under ON bias induces more significant hump and higher drain leakage current, whereas the irradiation under PG bias induces the larger shift of threshold voltage, drain saturation current, and body current. What is very interesting is that the measurement-sequence-induced change of hump and drain leakage current under ON bias are larger than those under PG bias. The measurement-sequence-induced change rate of threshold voltage, drain saturation current, and body current under PG bias are larger than those under ON bias. Hence, we draw the following conclusions from these phenomena: the back-gate curves’ measurement weakens the TID effect; the more significant the TID effect, the more significant the weakening will be.
From Subsections
As shown in Table
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
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.
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.
[1] | |
[2] | |
[3] | |
[4] | |
[5] | |
[6] | |
[7] | |
[8] | |
[9] | |
[10] | |
[11] | |
[12] | |
[13] | |
[14] | |
[15] | |
[16] | |
[17] | |
[18] | |
[19] | |
[20] | |
[21] | |
[22] | |
[23] | |
[24] | |
[25] | |
[26] | |
[27] |