2. Gamma-ray irradiation effects on SiC MOSFETs

Figure 1 shows the change in the subthreshold region of drain current (*I*D) – gate voltage (*V*Gdz 1.2!/z c/10\$.!/\$+( z 1.2!/dz "+.z \*w\$\*\*!(z Iw%z
/z 5z #))w.5z %.. %¥ tion. The bias of 12 V was applied to drain (*V*D) during measurements. The gate oxide of the MOSFETs was formed using pyrogenic oxidation (H2:O2 = 1:1) at 1100°C. The mark "+" on the each line indicates the value of *V*T. As shown in the figure, the value of *V*T shifts to the negative voltage side, and also the *I*D-*V*<sup>G</sup> curves stretches after irradiation. This suggests that charge in oxide and interface traps are generated by gamma-ray irradiation.

Event Transient (SET) arises as a serious issue for analog electronics and digital logic cells. \*z#!\*!.(\_z0\$!z/z%\*z\*(+#z!(!0.+\*%/z.!z.!"!..! z0+z/z/\_z\* z0\$+/!z%\*z %#%0(z+)¥ binatorial logic are referred to as DSETs. In contrast, the Single Event Latch-up (SEL), the Single Event Burnout (SEB), and the Single Event Gate Rupture (SEGR) in power electronic

Electron-hole pairs are induced in insulator layers of Metal-Insulator-Semiconductor (MIS) structure devices, such as Metal-Oxide-Semiconductor (MOS) devices by irradiation, and as a result, charge trapped in insulator (oxide) and/or traps near the interface between oxide and semiconductor (interface traps) are generated. Since such charge trapped in insulator and interface traps act give harmful influence to transport properties of semiconductors, the electrical characteristics of MIS devices are degraded by their generation [9, 10f^z+.z!4)¥ ple, the shift of threshold voltage (*V* T) and the decrease in the channel mobility (µchdz.!z+¥ served in MOS field effect transistors (FETs). This radiation effect is called the TID effect, \* z%\*z#!\*!.(\_z0\$!z2(1!z+"z0\$!z z!""!0/z#. 1((5z%\*.!/!/z3%0\$z%\*.!/%\*#z +/!z+"z. %¥ 0%+\*/z !1/!z 0\$!z )+1\*0z +"z . %0%+\*w%\* 1! z \$.#!z %\*z %\*/1(0+.z \* z %\*0!."!z 0.,/z %\*¥

When energetic particles are irradiated into semiconductor crystals, atoms at lattice sites are scattered into non-lattice sites (knock-on effects). As a result, vacancies and interstitials are created in semiconductor crystals. This is the origin of the DDD effect. However, in reality, 0\$!z/0.101.!z+"z.!/% 1(z !"!0/z%/z\*+0z/+z/%),(!z\* zz3% !z2.%!05z+"z !"!0/z/1\$z/z %2¥ \*%!/\_z2\*5z(1/0!./\_z\* z2\*5w%),1.%05z+),(!4!/z!4%/0/z%\*z.5/0(/z!1/!z#!\*!.¥ ated vacancies and interstitials thermally diffuse and finally they become stable defects. In general, such defects act as scattering/recombination centers to free carriers, and as a result, the electrical characteristics of semiconductors devices are degraded. In the case of the DDD !""!0\_z/%)%(.z0+z0\$!z z!""!0\_z0\$!z !#. 0%+\*z+"z0\$!z\$.0!.%/0%/z+"z/!)%+\* 10+.z !2%¥ ces becomes larger with increasing fluence of radiation. The degradation of the electrical performance of solar cells installed in space satellites is known as one of the examples of the

 \*z0\$%/z\$,0!.\_z0\$!z!""!0/z+"z. %0%+\*z+\*z0\$!z!(!0.%(z\$.0!.%/0%/z+"z%z !2%!/z.!z !¥

Figure 1 shows the change in the subthreshold region of drain current (*I*D) – gate voltage (*V*Gdz 1.2!/z c/10\$.!/\$+( z 1.2!/dz "+.z \*w\$\*\*!(z Iw%z
/z 5z #))w.5z %.. %¥ tion. The bias of 12 V was applied to drain (*V*D) during measurements. The gate oxide of the MOSFETs was formed using pyrogenic oxidation (H2:O2 = 1:1) at 1100°C. The mark "+" on the each line indicates the value of *V*T. As shown in the figure, the value of *V*T shifts to the negative voltage side, and also the *I*D-*V*<sup>G</sup> curves stretches after irradiation. This suggests that

scribed from the point of view of the TID effect and the SEEs.

2. Gamma-ray irradiation effects on SiC MOSFETs

charge in oxide and interface traps are generated by gamma-ray irradiation.

devices are known as the hard errors.

380 Physics and Technology of Silicon Carbide Devices

creases with increasing dose.

DDD effect [11-14].

Figure 1. Change in the subthreshold region of *I*<sup>D</sup> –*V*G curves (subthreshold curves) for n-channel 6H-SiC MOSFETs by gamma-ray irradiation. The bias of 12 V was applied to drain during measurements. The "+" mark on the each line indicates the value of *V*T.

+. %\*#z0+z
3\$+.0!.z\* z%\*+'1.zeLf\_z0\$!z !\*/%05z+"z\$.#!z0.,,! z%\*z#0!z+4% !zc*N*OX) \* z%\*0!."!z0.,/zc*N*ITdz#!\*!.0! z5z%.. %0%+\*z\*z!z!/0%)0! z".+)z0\$!z/\$%"0z+"z/10\$.!/¥ hold curves using a following analysis. Since charge trapped in gate oxide does not respond to bias applied to gate, the entire subthreshold curve is simply shifted by the generation of \$.#!z0.,,! z%\*z#0!z+4% !^z\*z0\$!z+0\$!.z\$\* \_z/%\*!z0\$!z\$.#!z/00!z+"z%\*0!."!z0.,/z !¥ pends on Fermi level (thus, the value of the bias applied to gate oxide), the subthereshold curve is stretched by the generation of interface traps. This behavior can be expressed as

$$
\Delta V\_{\rm T} = \Delta V\_{\rm OX} + \Delta V\_{\rm IT} \tag{1}
$$

where *V*T\_z*V*OX and *V*IT are the shift of the threshold voltage by irradiation, the voltage shifts due to the generation of oxide-trapped charge and interface traps, respectively. Also, since the charge state of interface traps is assumed to be neutral at midgap state, at which Fermi level corresponds to the intrinsic Fermi level, the shift of the midgap voltage (*V*MID) due to irradiation is caused by oxide-trapped charge. Thus,

$$
\Delta V\_{\text{MID}} = \Delta V\_{\text{OX}} \tag{2}
$$

Since the subthreshold curve between *V*MID and *V*<sup>T</sup> is stretched by the generation of interface 0.,/\_z*V*IT is determined as

$$
\Delta V\_{\text{IT}} = \left(V\_{\text{T}} - V\_{\text{MD}}\right)\_{\text{post}} - \left(V\_{\text{T}} - V\_{\text{MD}}\right)\_{\text{pre}} \tag{3}
$$

where "post" and "pre" mean after and before irradiation, respectively.

In order to obtain the value of *V*MID, firstly, the drain current corresponding to the midgap condition (*I*MID) is estimated. In the subthreshold region, *I*D is expressed as the formula [15]

$$I\_{\rm D} = 2^{1/2} \mu \text{(W/L)} (\text{qN}\_{\rm A} L\_{\rm B} / \beta) \left( n\_{\rm i} / N\_{\rm A} \right)^2 \exp(\beta \phi\_{\rm s}) (\beta \phi\_{\rm s})^{-1/2} \tag{4}$$

MOSFETs, the positive charge is trapped in gate oxide by gamma-ray irradiation and the trapped charge increases with increasing absorbed dose. Since the shift of *V*OX for the Pyro SiC MOSFETs is larger than that for the Dry SiC MOSFETs, the value of trapped charge for the Pyro SiC MOSFETs is larger than that for the Dry SiC MOSFETs. On the other hand, *V*OX for the H2 SiC MOSFETs shows complicated behaviors although the shift is very slight even after 530 kGy irradiation. Thus, firstly the *V*OX shifts to the negative voltage side at +/!/z!(+3zGCz'5^z+3!2!.\_z 0\$!z2(1!z/\$+3/zz,+/%0%2!z2+(0#!z/\$%"0z.+1\* zICz'5z(¥ though a negative shift appears at 180 kGy. Then, finally, the shift becomes positive again

Radiation Response of Silicon Carbide Diodes and Transistors

http://dx.doi.org/10.5772/51371

383

Figure 2. a*V* <sup>T</sup>9K9>MF;LAGFG>9:KGJ:=<<GK=>GJF;@9FF=D\$/A)+/"!0K0@=LJA9F?D=K

sent results obtained from MOSFETs of which gate oxide was fabricated by dry (Dry) and pyrogenic (Pyro) oxidations at 1100°C and pryogenic oxidation followed by hydrogen annealing at 700°C for 30 min at a pressure of 20 Torr (H2),

These behaviors indicate that both positively and negatively charges are generated in gate oxide for the H2 SiC MOSFETs by gamma-ray irradiation. It was reported from the change %\*z,%0\*!zxz2+(0#!z\$.0!.%/0%/z+"zIw%z
z,%0+./z 1!z0+z#))w.5z%.. %¥ tion that negative and positive trapped charges were generated near SiO2/SiC interface and in oxide at 40 nm from the interface, respectively [18]. Although the mechanism of H2w\*\*!¥ (%\*#z!""!0z+\*z0\$!z#0!z+4% !z\* z0\$!z%\*0!."!z!03!!\*z+4% !z\* z%z\$/z\*+0z5!0z!!\*z(.%¥ fied, since the initial value of *V*T decreased by H2 annealing [19], the large shift of *V*T to the positive voltage side and the unique behavior of *V*OX might occurs due to the reduction of H2w\*\*!(%\*#z!""!0/z5z#))w.5z%.. %0%+\*^z(/+\_z%0z/\$+1( z!z\*+0%! z 0\$0zz,.0z+"z%\*¥ terface traps might be detected as oxide-trapped-charge in this analysis since interface traps in the middle region of the band gap of 6H–SiC have extremely long charge release times at RT, and they act just as charge trapped in oxide [20]. In contrast to *V*OX, the values of *V*IT

;AJ;D=K9F<KIM9J=KJ=HJ=s

after irradiation at 530 kGy.

respectively.

where *N*A, *n*<sup>i</sup> \_z£s and *L*B are the acceptor (or donor) concentration in the channel region of a MOSFET, the intrinsic carrier concentration, band bending at the surface and the Debye length given by *L*BzRzc<sup>s</sup> uc*qN*A))1/2, respectively. Here, is equal to *q*/*k*BT, where *q* and *k*<sup>B</sup> are the electron charge and the Boltzmann constant, respectively. At the midgap condition, £s is equal to (*k*BT/*q*)ln(*N*A/*n*<sup>i</sup> ). Thus, *I*MIDz\*z!z!/0%)0! z".+)z!-^zcGdz1/%\*#z£s for (*k*BT/*q*)ln(*N*A/*n*<sup>i</sup> ). Then, the value of *V*MID can be obtained from the value of *V*G at *I*MID on subthreshold curves. It should be mentioned that for the determination of *V*MID for SiC, it is necessary to linearly extrapolate the lower position of the subthreshold curve down to the lower part of the curve, since the value of *I*MID is of the order of 10-30 A.

\$!z2(1!z+"z*N*OXz\* z*N*IT is estimated from

$$
\Delta N\_{\rm OX} = \ \Delta V\_{\rm OX} \ \ C\_{\rm OX} / q \tag{5}
$$

$$
\Delta N\_{\text{IT}} = \ \Delta V\_{\text{IT}} \ \ C\_{\text{OX}} / q \tag{6}
$$

where *C*OX is equal to OX/*t*OX, and OX and *t*OX are the relative dielectric constant of SiO2 and the thickness of gate oxide, respectively.

Figure 2 (a) shows the *V*T as a function of absorbed dose for n-channel 6H-SiC MOSFETs. The triangles, circles and squares represent results obtained from MOSFETs of which gate oxide was fabricated by dry (Dry) and pyrogenic (Pyro) oxidations at 1100°C and pryogenic oxidation followed by hydrogen annealing at 700°C for 30 min at a pressure of 20 Torr (H2), respectively. For the details of the fabrication process of those MOSFETs, please see Ref. [16, 17]. For the Dry SiC MOSFETs, the *V*T slightly shifts to the positive voltage side above 50 '5z(0\$+1#\$z0\$!z2(1!z +!/z\*+0z\$\*#!z!(+3zFCz'5^z+.z0\$!z5.+z%z
/\_z0\$!z2(¥ ue of *V*<sup>T</sup> shifts to the negative voltage side, and the negative shift become smaller above 30 kGy. For the H2 SiC MOSFETs, the *V*T shows the negative shift around 20 kGy, however, the voltage shift terns to the positive above 30 kGy. Since the value of *V*T is affected by the #!\*!.0%+\*z+"z\$.#!z0.,,! z%\*z#0!z+4% !z\* z%\*0!."!z0.,/\_z"+.z1\* !./0\* %\*#z0\$!z!\$2¥ ior of *V*T, it is necessary to know the information on *V*OXz\* z*V*IT. Therefore, the value of *V*OX and *V*IT is derived from the subthreshold curves using Eqs. (1) – (4). The absorbed +/!z !,!\* !\*!z+"z*V*OXz\* z*V*IT is shown in Figs. 3 (a) and (b), respectively. The values of *V*OX for the Dry and the Pyro SiC MOSFETs show the negative voltage shift and the shift becomes large with absorbed dose. These results indicate that for the Dry and the Pyro SiC MOSFETs, the positive charge is trapped in gate oxide by gamma-ray irradiation and the trapped charge increases with increasing absorbed dose. Since the shift of *V*OX for the Pyro SiC MOSFETs is larger than that for the Dry SiC MOSFETs, the value of trapped charge for the Pyro SiC MOSFETs is larger than that for the Dry SiC MOSFETs. On the other hand, *V*OX for the H2 SiC MOSFETs shows complicated behaviors although the shift is very slight even after 530 kGy irradiation. Thus, firstly the *V*OX shifts to the negative voltage side at +/!/z!(+3zGCz'5^z+3!2!.\_z 0\$!z2(1!z/\$+3/zz,+/%0%2!z2+(0#!z/\$%"0z.+1\* zICz'5z(¥ though a negative shift appears at 180 kGy. Then, finally, the shift becomes positive again after irradiation at 530 kGy.

where "post" and "pre" mean after and before irradiation, respectively.

382 Physics and Technology of Silicon Carbide Devices

curve, since the value of *I*MID is of the order of 10-30 A.

\$!z2(1!z+"z*N*OXz\* z*N*IT is estimated from

the thickness of gate oxide, respectively.

where *N*A, *n*<sup>i</sup>

equal to (*k*BT/*q*)ln(*N*A/*n*<sup>i</sup>

( ) ( )

*NL n N*

In order to obtain the value of *V*MID, firstly, the drain current corresponding to the midgap condition (*I*MID) is estimated. In the subthreshold region, *I*D is expressed as the formula [15]

> 2 1/2 1/2 <sup>D</sup> AB i A s s *I* 2 W / L (q / ) / exp( )( )

MOSFET, the intrinsic carrier concentration, band bending at the surface and the Debye length given by *L*BzRzc<sup>s</sup> uc*qN*A))1/2, respectively. Here, is equal to *q*/*k*BT, where *q* and *k*<sup>B</sup> are the electron charge and the Boltzmann constant, respectively. At the midgap condition, £s is

Then, the value of *V*MID can be obtained from the value of *V*G at *I*MID on subthreshold curves. It should be mentioned that for the determination of *V*MID for SiC, it is necessary to linearly extrapolate the lower position of the subthreshold curve down to the lower part of the

where *C*OX is equal to OX/*t*OX, and OX and *t*OX are the relative dielectric constant of SiO2 and

Figure 2 (a) shows the *V*T as a function of absorbed dose for n-channel 6H-SiC MOSFETs. The triangles, circles and squares represent results obtained from MOSFETs of which gate oxide was fabricated by dry (Dry) and pyrogenic (Pyro) oxidations at 1100°C and pryogenic oxidation followed by hydrogen annealing at 700°C for 30 min at a pressure of 20 Torr (H2), respectively. For the details of the fabrication process of those MOSFETs, please see Ref. [16, 17]. For the Dry SiC MOSFETs, the *V*T slightly shifts to the positive voltage side above 50 '5z(0\$+1#\$z0\$!z2(1!z +!/z\*+0z\$\*#!z!(+3zFCz'5^z+.z0\$!z5.+z%z
/\_z0\$!z2(¥ ue of *V*<sup>T</sup> shifts to the negative voltage side, and the negative shift become smaller above 30 kGy. For the H2 SiC MOSFETs, the *V*T shows the negative shift around 20 kGy, however, the voltage shift terns to the positive above 30 kGy. Since the value of *V*T is affected by the #!\*!.0%+\*z+"z\$.#!z0.,,! z%\*z#0!z+4% !z\* z%\*0!."!z0.,/\_z"+.z1\* !./0\* %\*#z0\$!z!\$2¥ ior of *V*T, it is necessary to know the information on *V*OXz\* z*V*IT. Therefore, the value of *V*OX and *V*IT is derived from the subthreshold curves using Eqs. (1) – (4). The absorbed +/!z !,!\* !\*!z+"z*V*OXz\* z*V*IT is shown in Figs. 3 (a) and (b), respectively. The values of *V*OX for the Dry and the Pyro SiC MOSFETs show the negative voltage shift and the shift becomes large with absorbed dose. These results indicate that for the Dry and the Pyro SiC

 

= (4)

). Thus, *I*MIDz\*z!z!/0%)0! z".+)z!-^zcGdz1/%\*#z£s for (*k*BT/*q*)ln(*N*A/*n*<sup>i</sup>

OX OX OX *N VC q* / (5)

IT IT OX *N VC q* / (6)

).

\_z£s and *L*B are the acceptor (or donor) concentration in the channel region of a

 

Figure 2. a*V* <sup>T</sup>9K9>MF;LAGFG>9:KGJ:=<<GK=>GJF;@9FF=D\$/A)+/"!0K0@=LJA9F?D=K ;AJ;D=K9F<KIM9J=KJ=HJ=s sent results obtained from MOSFETs of which gate oxide was fabricated by dry (Dry) and pyrogenic (Pyro) oxidations at 1100°C and pryogenic oxidation followed by hydrogen annealing at 700°C for 30 min at a pressure of 20 Torr (H2), respectively.

These behaviors indicate that both positively and negatively charges are generated in gate oxide for the H2 SiC MOSFETs by gamma-ray irradiation. It was reported from the change %\*z,%0\*!zxz2+(0#!z\$.0!.%/0%/z+"zIw%z
z,%0+./z 1!z0+z#))w.5z%.. %¥ tion that negative and positive trapped charges were generated near SiO2/SiC interface and in oxide at 40 nm from the interface, respectively [18]. Although the mechanism of H2w\*\*!¥ (%\*#z!""!0z+\*z0\$!z#0!z+4% !z\* z0\$!z%\*0!."!z!03!!\*z+4% !z\* z%z\$/z\*+0z5!0z!!\*z(.%¥ fied, since the initial value of *V*T decreased by H2 annealing [19], the large shift of *V*T to the positive voltage side and the unique behavior of *V*OX might occurs due to the reduction of H2w\*\*!(%\*#z!""!0/z5z#))w.5z%.. %0%+\*^z(/+\_z%0z/\$+1( z!z\*+0%! z 0\$0zz,.0z+"z%\*¥ terface traps might be detected as oxide-trapped-charge in this analysis since interface traps in the middle region of the band gap of 6H–SiC have extremely long charge release times at RT, and they act just as charge trapped in oxide [20]. In contrast to *V*OX, the values of *V*IT for all SiC MOSFETs show the positive voltage side and their shifts become larger with increasing absorbed dose although the absolute values depend on the fabrication process of gate oxide, as shown in Fig. 3 (c).

Figure 3. (a) ΔVχ and (b) ΔVg as a function of absorbed dose for n-channel 6H-SiC MOSFETs. The triangles, circles and squares represent results obtained from MOSFETs of which gate oxide was fabricated by dry (Dry) and pyrogenic (Ryro) oxidations at 1100°C and pryogenic oxidation followed by hydrogen annealing at 700°C for 30 min at a pressure of 20 Torr (H2), respectively.

The values of △Nox and △Nr are estimated from Figs. 3 (b) and (c), respectively, using Eq. (5)/(6). Figures 4 (a) and (b) show ΔNοx and ΔΝη, respectively, for the Dry (triangles), the Pyro (circles) and the H2 (squares) SiC MOSFETs as a function of absorbed dose. For comparison, the reported results of Si MOSFETs are also plotted in the figures (upside-down triangles) [9]. The value of ΔΝοχ for the Dry SiC MOSFETs is slightly smaller than that of the Pyro SiC MOSFETs and both values increase with increasing absorbed dose with an exponent of 2/3. It is also found that Si MOSFETs show the 2/3 power-low dependence, although the value of ΔΝοχ for the Si MOSFETs is larger than that for the SiC MOSFETs [9]. On the other hand, the change in ΔΝοχ for the H2 MOSFETs due to irradiation show a different behavior from others, and the value is in order of 1011 /cm² even after irradiation at 530 kGy. These results indicate that the characteristics of gate oxide fabricated by H2-annealing differ from those by nonannealing. However, it should be noticed that AN & estimated in this analysis is a value subtracting a positive component from a negative component. Thus, if both positive and negative components are almost the same value, the net number of ΔΝοχ is small. Therefore, from this result, we cannot simply conclude that the quality of gate oxide fabricated by Hz-annealing is higher than that of gate oxide fabricated by non-annealing or not.

For ΔΝη, the H2 SiC MOSFETs have lower values than the other MOSFETs at absorbed doses above 30 kGy. The characteristics of SiC MOS devices were reported to be degraded by carbon related compounds remaining around the interface between SiO2 and SiC [21]. Since such compounds might also act as precursors of radiation-induced interface traps, it is assumed that H₂ annealing to gate oxide of SiC MOSFETs reduces residual compounds near the interface. For the absorbed dose dependence of ANT7, the H2 and the Dry SiC MOSFETs have the 2/3 power-low dependence although ANm for the Pyro SiC MOSFETs increases with increasing the absorbed dose with an exponent of approximately 3/2. The 2/3 powerlaw dependence is also reported in Si of which gate oxide was formed using dry oxidation [9]. The power-law dependence comes from the generation mechanism of interface traps, and the structural and/or electrical properties of the interface between SiO2 and SiC for the H2 and the Dry SiC MOSFETs are different from those for the Pyro SiC MOSFETs. Therefore, it is suggests that the characteristics of the interface between SiO2 and SiC formed by pyrogenic oxidation followed by H2-annealing are similar to those formed by dry oxidation.

Figure 4. a) Δήχ and (b) ΔΝη for Dry (triangles), Pyro (circles) and HJ (squares) SiC MOSFETs as a function of absorbed dose. For comparison, the reported results of Si MOSFETs are also plotted in the figures (upside-down triangles) [9].

The u., for Si MOSFETs is known to decrease with increasing absorbed dose [10]. In order to confirm this for SiC MOSFETs, µm for the H2 SiC MOSFETs were plotted as a function of absorbed dose (Fig. 5). For comparison, the result reported for Si MOSFETs are also plotted in the figure [9]. The µ.ã for the H2 SiC MOSFETs does not change up to 20 kGy and the value decreases with increasing absorbed dose above 60 kGy. Then, the value of um reduces to be 50 % of the initial value at 530 kGy. On the other hand, um the Si MOSFETs decreases with increasing absorbed dose and becomes 50 % of the initial value by irradiation at 10 kGy. Although the initial value of µa for Si MOSFETs (600 cm²/Vs) is much higher than the initial value of µ.¸¸ for the H2 SiC MOSFET (~ 50 cm²/Vs), the value for Si MOSFETs is assumed to be almost zero after irradiation at 100 kGy whereas the H2 SiC MOSFETs still keep 25 cm²/Vs of µm even after irradiation at 530 kGy. In addition, it is mentioned that the stability of their electrical performance against irradiation is also important for Rad-hard devices. Therefore, it can be concluded that SiC MOSFETs are quite tolerant against radiation in comparison with Si MOSFETs. For the degradation mechanism of µ.j. Ohshima et al. reported [17] that the relationship between the decrease of Hot and ΔNm for SiC MOSFETs was described by the same relationship reported for Si MOSFETs (μω = μο ((1 + αΔΝη)) [10], where μο and a are the initial value of the channel mobility and a constant (= 7.0±1.3×10+3), respectively. This suggests that um for SiC MOSFETs as well as Si MOSFETs can be explained in term of carrier scattering in the channel region by interface traps generated by gamma-ray irradiation. Since interface traps located in the middle of the band gap behave just like charge trapped oxide for SiC, ΔΝη obtained in this analysis means the net density of interface traps which act as carrier scattering centers. It was reported that the channel mobility of 6H-SiC MOSFETs is affected by acceptor-like traps existing near the conduction band edge [22]. Although the relationship between interface traps induced by irradiation and intrinsic interface traps has not yet been clarified in the case of SiC MOS devices, it is assumed that the radiation resistance of SiC MOSFETs might be improved by the reduction of initial interface traps generated near the conduction band edge.

Figure 5. นู, for H2 SiC MOSFETs as a function of absorbed dose. For comparison, the result reported for Si MOSFETs are also plotted in the figure [9]. The value of the channel mobility is normalized by the initial value.

Next, the effects of the surface morphology on um of SiC MOSFETs irradiated with gammarays will be discussed. In this study, MOSFETs were fabricated on n-type 6H-SiC epitaxial layers using the same fabrication process except the procedures of high temperature annealing after implantation [23]. Thus, although all samples were annealed at 1650°C for 3 min in an Ar atmosphere, the surface of one series of samples was covered with carbon films (Ccoating) during the annealing to avoid the degradation of the surface morphology [24], and the other series of samples were annealed without the carbon coating (non-coating). After the annealing, the carbon films were removed by the oxidation at 800°C tor 30 min in Q2 gas. Gate oxide of both series of the MOSFETs were formed by pyrogenic oxidation (H2O2 = 1:1) at 1100°C for 30 min. For the details of the fabrication process, please see Ref. [23]. The initial values of µ.g for C-coating and non-coating SiC MOSFETs are 41 and 44 cm²/Vs, respective

ly. For the surface morphology, the values of root mean square (RMS) for the C-coating and non-coating SiC are obtained to be 0.67 and 1.36 nm, respectively, from AFM measurements, whereas the RMS was 0.25 nm before annealing.

Figures 6 (a) and (b) show µa and ANm respectively, for C-coating (squares) and non-coating (circles) SiC MOSFETs as a function of absorbed dose. As shown in the figure, no significant decrease or slight increase in um is observed for the C-coating SiC MOSFETs. The value of ANm for the C-coating SiC MOSFETs is estimated to be less than 4x1011 /cm², and no significant increase in △Nr is observed up to 3 MGy. In contrast, μ.α. for the non-coating SiC MOSFETs decreases with increasing absorbed dose above 2 MGy. In the absorbed region that µ¡¡ decreases, ΔNm increases with increasing absorbed dose, and the value becomes of the order of 1012 /cm² by irradiation above 2 MGy. As above-mentioned, μω is degraded by the generation of interface traps. Therefore, the decrease in μ.g. for the non-coating SIC MOS-FETs can be interpreted in terms of the generation of interface traps. Also, it was reported by Kimoto [24] the channel mobility can be affected by the surface roughness. So, the higher radiation resistance obtained for the C-coating SiC MOSFETs compared to non-coating ones is caused by the less surface roughness.

Figure 6. a) μ., and (b) ΔΝη for C-coating (squares) and non-coating (circles) SiC MOSFETs as a function of absorbed dose.
