- Email: [email protected]
Correlation between direct characteristic deficiencies and deep levels in 6H–SiC Schottky diodes
Materials Science and Engineering C 21 (2002) 283 – 286 www.elsevier.com/locate/msec
Correlation between direct characteristic deficiencies and deep levels in 6H–SiC Schottky diodes N. Sghaier *, A. Souifi, J.M. Bluet, G. Guillot Laboratoire de Physique de la Matie`re (UMR CNRS 5511), Institut National des Sciences Applique´es de Lyon, Baˆtiment Blaise Pascal, 7 Avenue Jean Capelle, 69621 Villeurbanne cedex, France
Abstract Excess current at low forward bias is observed for large area Ni Schottky diodes on n- and p-type 6H SiC. Random telegraph signal (RTS) measurements, carried out on these defective devices, show discrete time switching of the current. The traps signatures (Ea = 0.35 eV, r = 1.17 10 18 cm2 for the n-type, Ea = 0.43 eV, r = 1.8 10 20 cm2 for the p-type) extracted from DLTS measurement are very close to the one obtained from RTS. This strong correlation between the two different technique is attributed to the presence of an extended defect which presents different charge states (i.e. an extended defect decorated by punctual traps). This assumption is enforced by the DLTS measurements as a function of the filling time and as a function of the field. D 2002 Elsevier Science B.V. All rights reserved. Keywords: SiC; Schottky diodes; Barrier inhomogeneity; Random telegraph signal
1. Introduction Nowadays, a strong research activity is devoted to silicon carbide mainly for the development of high-power electronic devices. Indeed, because of its superior intrinsic electrical and physical properties (high electron drift velocity, high breakdown electric field, high thermal conductivity), SiC is a promising material to overpass the limit of the silicon technology in this field [1]. For instance, Schottky rectifiers are quite attractive for power conversion applications requiring high breakdown voltage ( > 600 V) and low on resistance. Nevertheless, even if promising demonstrators have been realized [2], the characteristics of large area diodes still show high reverse current density in comparison to the thermionic emission theory [3] and an excess current in the forward characteristic at low voltage [4,5]. These phenomena have been interpreted in terms of barrier height inhomogeneities attributed to material quality deficiencies. If the deleterious effect of micropipes (hollow core screw dislocation) on the electrical performance of SiC devices has been well established [6], many other defects can be responsible of the high power devices deficiencies [7]. This
*
Corresponding author. Fax: +33-4-72-43-85-31. E-mail address: [email protected] (N. Sghaier).
is why a detailed study of electrically active defects is necessary to understand the origin of nonideal current voltage characteristics. In this paper, we report on I –V – T analyses that reveal an excess current for low voltage. In the following sections, we focus on these defective diodes in order to analyze the physical origin of the excess current. Toward this end, random telegraph signal (RTS) and deep level transient spectroscopy (DLTS) have been performed.
2. Experimental results 2.1. Sample preparation The Ni/6H – SiC Schottky diodes studied in this work were realized at the North Carolina State University on nand p-type epilayers purchased from CREE Research. The layer thickness and doping are, respectively, 1.2 Am and Nd = 4.1 10 16 cm 3 for n-type, and 1.6 Am with Na = 2 1015 cm 3 for p-type. Back side ohmic contacts on n- and p-type was formed using, respectively, Ni and NiAl deposit with a subsequent annealing at 1000 jC in N2 ambient. The circular Ni/SiC Schottky contact were next formed by electron beam evaporation through a Mo shadow mask after a quick dip in HF in order to remove any oxide left on the SiC surface. The diode diameter is ranging from
0928-4931/02/$ - see front matter D 2002 Elsevier Science B.V. All rights reserved. PII: S 0 9 2 8 - 4 9 3 1 ( 0 2 ) 0 0 0 8 1 - 4
284
N. Sghaier et al. / Materials Science and Engineering C 21 (2002) 283–286
0.6 to 2.4 mm. C – V measurement have been performed at room temperature in order to control the doping profile of the different layers. The results revealed that the doping were homogenous in the layer depth and in agreement with nominal values. 2.2. I– V characteristics Direct and reverse I– V characteristics were next measured as a function of the temperature in the range 150 – 600 K. An important dispersion in the ideality factor values at 300 K and in the importance of the leakage current (direct or reverse) has been observed. These variations on a same sample are mainly dependent on the diodes diameter. From a qualitative point of view, just looking at the direct I –V characteristics shape, we can then classify the diodes in three categories. (1) For a few of the 0.6-mm diameter diodes (20%), either on n- or p-type, quasi-ideal direct characteristics are obtained with an ideality factor of 1.1 at 500 K typical of thermionic emission. Nevertheless, at room temperature and below, the ideality factor (n) is closer to 2. This value of n may be due to a generation/recombination component of the current which vanishes when rising the temperature. (2) For all the diodes with diameter greater than 1.2 mm (1.8 and 2.4 mm), a very important excess current for low forward voltage (over 1 AA at 0.1 V) in the direct mode of operation is observed above 280 K. For these diodes, a leakage component constant with the temperature (over 1 AA at 3 V) was also observed. From a plot of the current density versus the perimeter/surface ratio, we found that this excess current is mainly peripheric. (3) Finally, for the majority of the 0.6-mm diodes and some of the 1.2 mm diodes, either on n- or p-type, we observed two exponential regime or more in the direct characteristic and a thermal activation of the reverse current. An example of these characteristics of p-type is displayed in Fig. 1. For these diodes, the current is mainly in the volume (less than 15% of peripheric component). In the following, we will focus on this category of defective diodes. These characteristics with different exponential regimes have already been reported by different groups using different Schottky metalization [3 –7] and interpreted in terms of barrier height fluctuations. For the characteristics with two exponential regimes, a simple model considering two different diodes in parallel with different barrier heights has been proposed by Defives et al. [4]. These two barrier heights correspond with the real Schottky for the upper one high barrier height (HBH), and with a localized defective zone at the metal/SiC interface for the lower one low barrier height (LBH). Examples of the fit are displayed in solid lines on Fig. 1 for a p-type diode. The interesting parameter in this model is not such the different barrier heights or ideality factor but the ratio between the two current components which correspond to the ratio of the defective zone area with LBH reported to rest of the diode area with HBH. We used
Fig. 1. J – V characteristics as a function of temperature for a diode on ptype 6H – SiC showing barrier height inhomogeneity. The solid lines correspond to the fit.
this model to fit the direct I– V curves for all the diodes. We have extracted the area ratio between the defective zone corresponding to the LBH and the rest of the diode with HBH. For all the diodes we obtained, a ratio ranging from 10 5 to 10 6 at room temperature. This corresponds to a defective zone of 0.3– 3 Am2 for a 0.6-mm diameter diode. This value is consistent with the presence of extended defects (screw dislocations or micropipes) emerging at the surface. Indeed, we have observed many of these defects under the diode after metallization removing. In the following sections, we will focus on these defective diodes in order to determine precisely, the localization and the physical nature, of the defect responsible for the excess current. Toward this end we have first measured random time fluctuation of the current (RTS). 2.3. RTS measurements From the analysis of I – V– T results presented above, it is clear that a localized current path with smaller barrier height, corresponding to a defective area at the Ni/6H– SiC interface, is at the origin of the excess current. In these conditions, if a punctual trap is located at the neighborhood of the defective current path, the trapping/detrapping mechanism of an electron (a hole for p-type) will result in a discrete time fluctuation of the current. This phenomenon, called random telegraph signal (RTS), has been first observed in silicon diodes and attributed to the presence of dislocations decorated by metallic impurities [8]. In SiC Schottky diodes, RTS has been first observed by Ouisse et al. [9] for diodes showing an excess current. This RTS was shown to originate from the trapping/detrapping of an electron in the neighborhood of a localized current path. It was proposed that this localized current path can be due to the presence of extended defects like the micropipes. In our case, for all the diodes showing a two-barrier height behavior in the forward I –V characteristics, we have actually observed RTS both for n- and p-type samples. This
N. Sghaier et al. / Materials Science and Engineering C 21 (2002) 283–286
RTS was noticed only above 280 K (when an excess current is present) and was not observed for the quasi-ideal diodes (first category). For n-type, a combination of two to four RTS is frequently found at the same bias corresponding to different traps. In order to extract statistical parameters, we have selected diodes showing a single RTS over the temperature range. Examples of these single-time fluctuations of the current is given for the n- and p-type material in Fig. 2a and b, respectively. We clearly observe important current fluctuation with pulse time varying from 1 to 100 ms. According to the model of Hsu [10], the average pulse duration corresponding to high current value (sH) and low current value (sL) are, respectively, determined by the capture (sc) and emission time (se) of an electron trap. The energetic position of the deep trap is then given by: sc ET EF ¼ kT ln ð1Þ se Applying this model to n- and p-type material, we have extracted the defect signature corresponding to the preponderant current switching (i.e. the most frequently observed). The values of the activation energy and the cross-section are reported in Table 1. From these RTS measurements, we can conclude that individual defects channel the current through the layer. This result is consistent with the two-barrier height model used in Section 2.2. In order to obtain a fine
Fig. 2. (a) Random telegraph signal observed on an n-type sample at T = 330 K. (b) Random telegraph signal observed on a p-type sample at T = 280 K.
285
Table 1 Main traps signature for n- and p-type 6H – SiC obtained from RTS and DLTS measurements Ea RTS (eV)
r RTS (cm 2)
Ea DLTS (eV)
r DLTS (cm2)
NT (cm3)
n-type 0.33F0.03 8.7191019 0.35F0.03 1.21018 2.11013 p-type 0.43F0.03 1.7671020 0.43F0.05 7.7111020 1.61012 Ea: activation energy, r: capture cross-section and NT: traps concentration.
characterization of these defects (localization in the structure, electrical character, punctual or extended character), DLTS measurements were performed next. 2.4. DLTS measurements To confirm the results obtained from RTS, DLTS measurement has been carried out on the same diodes. Indeed, if punctual defects acting as trapping/detrapping centers located at the neighborhood of a defective current path are responsible of the RTS, we may also observe these deep defects in DLTS. Nevertheless, such a correlation between the two techniques has not been reported for silicon diodes. In our case, a DLTS signal is observed. For n-type sample, three different peaks appear with a predominant one. The shallower and deeper defects corresponding to the two other peaks were observed only for a few diodes and with lower concentration than for the principal defect. For p-type sample, a single peak is observed. The traps signature extracted from the DLTS data are reported in Table 1. In the case of p-type samples, no results in the literature have been reported yet about a deep defect with an energy of Ev + 0.43 eV. In the case of n-type samples, the obtained signature corresponds to the S center reported by Anikin et al. [11]. These punctual traps, observed in DLTS, can be either arbitrary distributed within the epitaxial layer or concentrated in the vicinity of an extended defect. In order to discriminate between these two possibilities, DLTS measurements as a function of filling time has been performed. Indeed, when a decorated extended defect is present, if the trap concentration around the extended defect is high enough, the interaction between individual levels results in a repulsive potential, which limits the subsequent capture. This phenomenon implies a logarithmic kinetic of the traps filling which has been reported for the first time by Omling et al. [12] in Si. In his case, this behavior was attributed to dislocation defects decorated by punctual traps. A typical result of the filling kinetic, for filling times ranging from 1 As to 100 ms, is given in Fig. 3. As shown by the linear fit on the figure, a logarithmic kinetic is observed. According to the model of Omling, we obtain a confirmation of the presence of extended defect decorated by punctual traps. Nevertheless, the model described above is valid only if the traps are charged after the capture. To check out this point, the variation of DLTS with the field has been measured.
286
N. Sghaier et al. / Materials Science and Engineering C 21 (2002) 283–286
for the time fluctuation of the current observed in RTS. These punctual traps are so gettered by an extended defect.
4. Conclusion
Fig. 3. DLTS peak amplitude versus filling time in logarithmic scale for peak 2 (n-type).
When rising the reverse bias of the pulse, an enlargement and a shift toward the low temperature side, of the DLTS peak, is noticed. This sensitivity to the electric filed is characteristic of occupied traps initially charged. Indeed, in this case, the electrostatic potential well of the defect, is deformed by the electric field. This results in a downshift of the activation energy. Two mechanisms can be responsible of this shift: Poole – Frenkel effect or phonon assisted tunneling. In our case, it is difficult to distinguish between the two, but the important point is that we are actually in presence of initially charged occupied traps.
3. Discussion In Table 1, the DLTS traps signature are reported and compared to the values obtained from RTS. We clearly obtain a very good correlation between the activation energies from the two measurement techniques, which, to the best of our knowledge has never been reported before for SiC. The discrepancy in the capture cross-section originates from two major differences in the two measurements techniques. The first one is the field conditions: a few volt in reverse mode is applied in DLTS and 0.5 V in direct mode was used for RTS measurement. The second one is the carrier dynamic: in DLTS measurement the trapped and emitted carrier are static, in the RTS measurement the carrier are mobile. From the correlation in the activation energies we can then conclude that the punctual defects observed in DLTS acting as trapping/detrapping centers are responsible
The excess current observed in the direct characteristics of Ni/6H– SiC (n- and p-types) Schottky diodes has been attributed to local defective zones under the metallization according to a model considering a low barrier height for the defective zone and a normal barrier height for the rest of the diode. The surface ratio obtained from the fits of about 10 6 is consistent with the presence of dislocations and micropipes in the epitaxial layer observed under the diodes after metallization removing. From RTS and DLTS measurements, deep defects with activation energy of 0.35 and 0.43 eV, respectively, for n- and p-type material has been observed. The strong correlation between the two techniques implies that the observed defects are extended defects surrounded by punctual traps. The presence of these traps from DLTS and RTS measurements is correlated with the observation of the two barrier height behavior in the I– V characteristics. Furthermore, in the case of n-type samples, photoluminescence mappings around the micropipes have shown a decrease of the photoluminescence intensity which can be explained by the gettering effect of the S center which is known as the principal lifetime killer center in 6H –SiC.
References [1] M. Bhatnagar, B.J. Baliga, IEEE Trans. Electron Devices 40 (1993) 645. [2] A. Itoh, T. Kimoto, H. Matsunami, IEEE Electron Devices Lett. 17 (1996) 139. [3] M. Bhatnagar, B.J. Baliga, H.R. Kirk, G.A. Rozgonyi, IEEE Trans. Electron Devices 43 (1996) 150. [4] D. Defives, O. Noblanc, C. Dua, C. Brylinski, M. Barthula, V. Aubray-Fortuna, F. Meyer, IEEE Trans. Electron Devices 46 (1999) 449. [5] U. Zimmermann, A. Hallen, A. Kostantinov, B. Breitholtz, MRS Proc. 512 (1998) 151. [6] P. Neudeck, J.A. Powell, IEEE Electron Devices Lett. 15 (1994) 63. [7] P. Neudeck, W. Huang, M. Dudley, IEEE Trans. Electron Devices 46 (1999) 478. [8] S.T. Hsu, Solid-State Electron. 14 (1971) 487. [9] T. Ouisse, E. Platel, T. Billon, A. Lahreche, Electron. Lett. 33 (22) (1997) 1907. [10] S.T. Hsu, IEEE Trans. Electron Devices 17 (1970) 496. [11] M.M. Anikin, A.S. Zubrilov, A.A. Lebedev, A.M. Strel’chuk, A.E. Cherenkov, Sov. Phys. Semicond. 25 (1991) 519. [12] P. Omling, E.R. Weber, L. Montelius, H. Alexander, J. MicheL, Phys. Rev., B 32 (1985) 6571.
Correlation between direct characteristic deficiencies and deep levels in 6H–SiC Schottky diodes N. Sghaier *, A. Souifi, J.M. Bluet, G. Guillot Laboratoire de Physique de la Matie`re (UMR CNRS 5511), Institut National des Sciences Applique´es de Lyon, Baˆtiment Blaise Pascal, 7 Avenue Jean Capelle, 69621 Villeurbanne cedex, France
Abstract Excess current at low forward bias is observed for large area Ni Schottky diodes on n- and p-type 6H SiC. Random telegraph signal (RTS) measurements, carried out on these defective devices, show discrete time switching of the current. The traps signatures (Ea = 0.35 eV, r = 1.17 10 18 cm2 for the n-type, Ea = 0.43 eV, r = 1.8 10 20 cm2 for the p-type) extracted from DLTS measurement are very close to the one obtained from RTS. This strong correlation between the two different technique is attributed to the presence of an extended defect which presents different charge states (i.e. an extended defect decorated by punctual traps). This assumption is enforced by the DLTS measurements as a function of the filling time and as a function of the field. D 2002 Elsevier Science B.V. All rights reserved. Keywords: SiC; Schottky diodes; Barrier inhomogeneity; Random telegraph signal
1. Introduction Nowadays, a strong research activity is devoted to silicon carbide mainly for the development of high-power electronic devices. Indeed, because of its superior intrinsic electrical and physical properties (high electron drift velocity, high breakdown electric field, high thermal conductivity), SiC is a promising material to overpass the limit of the silicon technology in this field [1]. For instance, Schottky rectifiers are quite attractive for power conversion applications requiring high breakdown voltage ( > 600 V) and low on resistance. Nevertheless, even if promising demonstrators have been realized [2], the characteristics of large area diodes still show high reverse current density in comparison to the thermionic emission theory [3] and an excess current in the forward characteristic at low voltage [4,5]. These phenomena have been interpreted in terms of barrier height inhomogeneities attributed to material quality deficiencies. If the deleterious effect of micropipes (hollow core screw dislocation) on the electrical performance of SiC devices has been well established [6], many other defects can be responsible of the high power devices deficiencies [7]. This
*
Corresponding author. Fax: +33-4-72-43-85-31. E-mail address: [email protected] (N. Sghaier).
is why a detailed study of electrically active defects is necessary to understand the origin of nonideal current voltage characteristics. In this paper, we report on I –V – T analyses that reveal an excess current for low voltage. In the following sections, we focus on these defective diodes in order to analyze the physical origin of the excess current. Toward this end, random telegraph signal (RTS) and deep level transient spectroscopy (DLTS) have been performed.
2. Experimental results 2.1. Sample preparation The Ni/6H – SiC Schottky diodes studied in this work were realized at the North Carolina State University on nand p-type epilayers purchased from CREE Research. The layer thickness and doping are, respectively, 1.2 Am and Nd = 4.1 10 16 cm 3 for n-type, and 1.6 Am with Na = 2 1015 cm 3 for p-type. Back side ohmic contacts on n- and p-type was formed using, respectively, Ni and NiAl deposit with a subsequent annealing at 1000 jC in N2 ambient. The circular Ni/SiC Schottky contact were next formed by electron beam evaporation through a Mo shadow mask after a quick dip in HF in order to remove any oxide left on the SiC surface. The diode diameter is ranging from
0928-4931/02/$ - see front matter D 2002 Elsevier Science B.V. All rights reserved. PII: S 0 9 2 8 - 4 9 3 1 ( 0 2 ) 0 0 0 8 1 - 4
284
N. Sghaier et al. / Materials Science and Engineering C 21 (2002) 283–286
0.6 to 2.4 mm. C – V measurement have been performed at room temperature in order to control the doping profile of the different layers. The results revealed that the doping were homogenous in the layer depth and in agreement with nominal values. 2.2. I– V characteristics Direct and reverse I– V characteristics were next measured as a function of the temperature in the range 150 – 600 K. An important dispersion in the ideality factor values at 300 K and in the importance of the leakage current (direct or reverse) has been observed. These variations on a same sample are mainly dependent on the diodes diameter. From a qualitative point of view, just looking at the direct I –V characteristics shape, we can then classify the diodes in three categories. (1) For a few of the 0.6-mm diameter diodes (20%), either on n- or p-type, quasi-ideal direct characteristics are obtained with an ideality factor of 1.1 at 500 K typical of thermionic emission. Nevertheless, at room temperature and below, the ideality factor (n) is closer to 2. This value of n may be due to a generation/recombination component of the current which vanishes when rising the temperature. (2) For all the diodes with diameter greater than 1.2 mm (1.8 and 2.4 mm), a very important excess current for low forward voltage (over 1 AA at 0.1 V) in the direct mode of operation is observed above 280 K. For these diodes, a leakage component constant with the temperature (over 1 AA at 3 V) was also observed. From a plot of the current density versus the perimeter/surface ratio, we found that this excess current is mainly peripheric. (3) Finally, for the majority of the 0.6-mm diodes and some of the 1.2 mm diodes, either on n- or p-type, we observed two exponential regime or more in the direct characteristic and a thermal activation of the reverse current. An example of these characteristics of p-type is displayed in Fig. 1. For these diodes, the current is mainly in the volume (less than 15% of peripheric component). In the following, we will focus on this category of defective diodes. These characteristics with different exponential regimes have already been reported by different groups using different Schottky metalization [3 –7] and interpreted in terms of barrier height fluctuations. For the characteristics with two exponential regimes, a simple model considering two different diodes in parallel with different barrier heights has been proposed by Defives et al. [4]. These two barrier heights correspond with the real Schottky for the upper one high barrier height (HBH), and with a localized defective zone at the metal/SiC interface for the lower one low barrier height (LBH). Examples of the fit are displayed in solid lines on Fig. 1 for a p-type diode. The interesting parameter in this model is not such the different barrier heights or ideality factor but the ratio between the two current components which correspond to the ratio of the defective zone area with LBH reported to rest of the diode area with HBH. We used
Fig. 1. J – V characteristics as a function of temperature for a diode on ptype 6H – SiC showing barrier height inhomogeneity. The solid lines correspond to the fit.
this model to fit the direct I– V curves for all the diodes. We have extracted the area ratio between the defective zone corresponding to the LBH and the rest of the diode with HBH. For all the diodes we obtained, a ratio ranging from 10 5 to 10 6 at room temperature. This corresponds to a defective zone of 0.3– 3 Am2 for a 0.6-mm diameter diode. This value is consistent with the presence of extended defects (screw dislocations or micropipes) emerging at the surface. Indeed, we have observed many of these defects under the diode after metallization removing. In the following sections, we will focus on these defective diodes in order to determine precisely, the localization and the physical nature, of the defect responsible for the excess current. Toward this end we have first measured random time fluctuation of the current (RTS). 2.3. RTS measurements From the analysis of I – V– T results presented above, it is clear that a localized current path with smaller barrier height, corresponding to a defective area at the Ni/6H– SiC interface, is at the origin of the excess current. In these conditions, if a punctual trap is located at the neighborhood of the defective current path, the trapping/detrapping mechanism of an electron (a hole for p-type) will result in a discrete time fluctuation of the current. This phenomenon, called random telegraph signal (RTS), has been first observed in silicon diodes and attributed to the presence of dislocations decorated by metallic impurities [8]. In SiC Schottky diodes, RTS has been first observed by Ouisse et al. [9] for diodes showing an excess current. This RTS was shown to originate from the trapping/detrapping of an electron in the neighborhood of a localized current path. It was proposed that this localized current path can be due to the presence of extended defects like the micropipes. In our case, for all the diodes showing a two-barrier height behavior in the forward I –V characteristics, we have actually observed RTS both for n- and p-type samples. This
N. Sghaier et al. / Materials Science and Engineering C 21 (2002) 283–286
RTS was noticed only above 280 K (when an excess current is present) and was not observed for the quasi-ideal diodes (first category). For n-type, a combination of two to four RTS is frequently found at the same bias corresponding to different traps. In order to extract statistical parameters, we have selected diodes showing a single RTS over the temperature range. Examples of these single-time fluctuations of the current is given for the n- and p-type material in Fig. 2a and b, respectively. We clearly observe important current fluctuation with pulse time varying from 1 to 100 ms. According to the model of Hsu [10], the average pulse duration corresponding to high current value (sH) and low current value (sL) are, respectively, determined by the capture (sc) and emission time (se) of an electron trap. The energetic position of the deep trap is then given by: sc ET EF ¼ kT ln ð1Þ se Applying this model to n- and p-type material, we have extracted the defect signature corresponding to the preponderant current switching (i.e. the most frequently observed). The values of the activation energy and the cross-section are reported in Table 1. From these RTS measurements, we can conclude that individual defects channel the current through the layer. This result is consistent with the two-barrier height model used in Section 2.2. In order to obtain a fine
Fig. 2. (a) Random telegraph signal observed on an n-type sample at T = 330 K. (b) Random telegraph signal observed on a p-type sample at T = 280 K.
285
Table 1 Main traps signature for n- and p-type 6H – SiC obtained from RTS and DLTS measurements Ea RTS (eV)
r RTS (cm 2)
Ea DLTS (eV)
r DLTS (cm2)
NT (cm3)
n-type 0.33F0.03 8.7191019 0.35F0.03 1.21018 2.11013 p-type 0.43F0.03 1.7671020 0.43F0.05 7.7111020 1.61012 Ea: activation energy, r: capture cross-section and NT: traps concentration.
characterization of these defects (localization in the structure, electrical character, punctual or extended character), DLTS measurements were performed next. 2.4. DLTS measurements To confirm the results obtained from RTS, DLTS measurement has been carried out on the same diodes. Indeed, if punctual defects acting as trapping/detrapping centers located at the neighborhood of a defective current path are responsible of the RTS, we may also observe these deep defects in DLTS. Nevertheless, such a correlation between the two techniques has not been reported for silicon diodes. In our case, a DLTS signal is observed. For n-type sample, three different peaks appear with a predominant one. The shallower and deeper defects corresponding to the two other peaks were observed only for a few diodes and with lower concentration than for the principal defect. For p-type sample, a single peak is observed. The traps signature extracted from the DLTS data are reported in Table 1. In the case of p-type samples, no results in the literature have been reported yet about a deep defect with an energy of Ev + 0.43 eV. In the case of n-type samples, the obtained signature corresponds to the S center reported by Anikin et al. [11]. These punctual traps, observed in DLTS, can be either arbitrary distributed within the epitaxial layer or concentrated in the vicinity of an extended defect. In order to discriminate between these two possibilities, DLTS measurements as a function of filling time has been performed. Indeed, when a decorated extended defect is present, if the trap concentration around the extended defect is high enough, the interaction between individual levels results in a repulsive potential, which limits the subsequent capture. This phenomenon implies a logarithmic kinetic of the traps filling which has been reported for the first time by Omling et al. [12] in Si. In his case, this behavior was attributed to dislocation defects decorated by punctual traps. A typical result of the filling kinetic, for filling times ranging from 1 As to 100 ms, is given in Fig. 3. As shown by the linear fit on the figure, a logarithmic kinetic is observed. According to the model of Omling, we obtain a confirmation of the presence of extended defect decorated by punctual traps. Nevertheless, the model described above is valid only if the traps are charged after the capture. To check out this point, the variation of DLTS with the field has been measured.
286
N. Sghaier et al. / Materials Science and Engineering C 21 (2002) 283–286
for the time fluctuation of the current observed in RTS. These punctual traps are so gettered by an extended defect.
4. Conclusion
Fig. 3. DLTS peak amplitude versus filling time in logarithmic scale for peak 2 (n-type).
When rising the reverse bias of the pulse, an enlargement and a shift toward the low temperature side, of the DLTS peak, is noticed. This sensitivity to the electric filed is characteristic of occupied traps initially charged. Indeed, in this case, the electrostatic potential well of the defect, is deformed by the electric field. This results in a downshift of the activation energy. Two mechanisms can be responsible of this shift: Poole – Frenkel effect or phonon assisted tunneling. In our case, it is difficult to distinguish between the two, but the important point is that we are actually in presence of initially charged occupied traps.
3. Discussion In Table 1, the DLTS traps signature are reported and compared to the values obtained from RTS. We clearly obtain a very good correlation between the activation energies from the two measurement techniques, which, to the best of our knowledge has never been reported before for SiC. The discrepancy in the capture cross-section originates from two major differences in the two measurements techniques. The first one is the field conditions: a few volt in reverse mode is applied in DLTS and 0.5 V in direct mode was used for RTS measurement. The second one is the carrier dynamic: in DLTS measurement the trapped and emitted carrier are static, in the RTS measurement the carrier are mobile. From the correlation in the activation energies we can then conclude that the punctual defects observed in DLTS acting as trapping/detrapping centers are responsible
The excess current observed in the direct characteristics of Ni/6H– SiC (n- and p-types) Schottky diodes has been attributed to local defective zones under the metallization according to a model considering a low barrier height for the defective zone and a normal barrier height for the rest of the diode. The surface ratio obtained from the fits of about 10 6 is consistent with the presence of dislocations and micropipes in the epitaxial layer observed under the diodes after metallization removing. From RTS and DLTS measurements, deep defects with activation energy of 0.35 and 0.43 eV, respectively, for n- and p-type material has been observed. The strong correlation between the two techniques implies that the observed defects are extended defects surrounded by punctual traps. The presence of these traps from DLTS and RTS measurements is correlated with the observation of the two barrier height behavior in the I– V characteristics. Furthermore, in the case of n-type samples, photoluminescence mappings around the micropipes have shown a decrease of the photoluminescence intensity which can be explained by the gettering effect of the S center which is known as the principal lifetime killer center in 6H –SiC.
References [1] M. Bhatnagar, B.J. Baliga, IEEE Trans. Electron Devices 40 (1993) 645. [2] A. Itoh, T. Kimoto, H. Matsunami, IEEE Electron Devices Lett. 17 (1996) 139. [3] M. Bhatnagar, B.J. Baliga, H.R. Kirk, G.A. Rozgonyi, IEEE Trans. Electron Devices 43 (1996) 150. [4] D. Defives, O. Noblanc, C. Dua, C. Brylinski, M. Barthula, V. Aubray-Fortuna, F. Meyer, IEEE Trans. Electron Devices 46 (1999) 449. [5] U. Zimmermann, A. Hallen, A. Kostantinov, B. Breitholtz, MRS Proc. 512 (1998) 151. [6] P. Neudeck, J.A. Powell, IEEE Electron Devices Lett. 15 (1994) 63. [7] P. Neudeck, W. Huang, M. Dudley, IEEE Trans. Electron Devices 46 (1999) 478. [8] S.T. Hsu, Solid-State Electron. 14 (1971) 487. [9] T. Ouisse, E. Platel, T. Billon, A. Lahreche, Electron. Lett. 33 (22) (1997) 1907. [10] S.T. Hsu, IEEE Trans. Electron Devices 17 (1970) 496. [11] M.M. Anikin, A.S. Zubrilov, A.A. Lebedev, A.M. Strel’chuk, A.E. Cherenkov, Sov. Phys. Semicond. 25 (1991) 519. [12] P. Omling, E.R. Weber, L. Montelius, H. Alexander, J. MicheL, Phys. Rev., B 32 (1985) 6571.