- Email: [email protected]
Correlation of charge trapping and electroluminescence in highly efficient Si-based light emitters
Available online at www.sciencedirect.com
Physica E 16 (2003) 499 – 504 www.elsevier.com/locate/physe
Correlation of charge trapping and electroluminescence in highly e%cient Si-based light emitters T. Gebela; b;∗ , L. Rebohleb; c , J. Suna , W. Skorupaa; b , A.N. Nazarovd , I. Osiyukd a Forschungszentrum
Rossendorf, Bautzner Landstrasse 128, P.O. Box 510119, 01314 Dresden, Germany b Nanoparc GmbH, Dresden-Rossendorf, Germany c Vienna Technical University, Vienna, Austria d Institute of Semiconductor Physics, NASU Kiev, Ukraine
Abstract In this paper we report on recent results on charge trapping and electroluminescence (EL) from Ge rich SiO2 layers. Thermally grown 80 nm thick SiO2 layers were implanted with Ge ions at energies of 30 –50 keV to peak concentrations of 1–6 at%. Subsequently rapid thermal annealing was performed at 1000◦ C for 6, 30 and 150 s under a nitrogen atmosphere in order to form luminescence centers. A combination of capacitance–voltage (CV ) and current–voltage (IV ) methods was used for the investigation of the trapping properties. It was found that at electric :elds ¡ 8 MV=cm electron trapping dominates while at higher electric :elds which are typically required for the EL operation of the devices positive charge trapping occurs. It is assumed, that the trapping sites which are responsible for the trapping of the positive charge are in strong relation to the defects causing the luminescence. ? 2002 Elsevier Science B.V. All rights reserved. PACS: 73.40.Qv; 72.20.Jv; 72.10.Fk; 78.60.Fi; 78.67.Bf Keywords: Nanocluster rich oxides; Ion implantation; Electroluminescence; Charge trapping
1. Introduction Optical data transfer is the most promising tool to overcome the bottleneck of interchip and intrachip communication in future computer systems [1]. Since silicon is the key material for today’s microelectronic devices one is strongly interested in the integration of devices for the emission, modulation and detection of optical signals on one and the same chip, using the current Si-technology. Furthermore, the increasing market for integrated optical sensor systems and ∗ Corresponding author. Tel.: +49-351-2603036; fax: +49-3512603411. E-mail address: [email protected] (T. Gebel).
multi-functional microsystems (e.g. lab-on-chip systems in biotechnology) also causes a strong demand on miniaturized and integrated optical systems based on silicon technology. A promising method for the formation of Si-based light emitting structures is ion beam synthesis. Electroluminescence (EL) from Si implanted oxide layers was observed in the red [2] and blue wavelength range [3,4] and from Ge implanted SiO2 layers EL was found in the red/infrared and in the blue/violet spectral region [4–8]. In our previous work we demonstrated strong violet EL with a power e%ciency of 0.5% from 130 nm thick SiO2 layers implanted with Ge [9] and demonstrated the prototype of a silicon based optocoupler [10]. We found evidence for an oxygen de:ciency center (ODC) as the
1386-9477/03/$ - see front matter ? 2002 Elsevier Science B.V. All rights reserved. PII: S 1 3 8 6 - 9 4 7 7 ( 0 2 ) 0 0 6 3 6 - 7
500
T. Gebel et al. / Physica E 16 (2003) 499 – 504
luminescence center responsible for the PL excited around 5 eV. Possible ODCs are the neutral oxygen vacancy (≡ Si–Si ≡, ≡ Si–Ge or ≡ Ge–Ge ≡) and the twofold coordinated Si or Ge atom. In this work the EL and the charge trapping properties of Ge implanted 80 nm thick SiO2 layers are investigated to get a better understanding about the injection and charge transport mechanism in this system.
2. Experimental Thermally grown 80 nm thick SiO2 :lms on (1 0 0) n-type Si wafers were implanted with Ge+ ions at energies of 30 –50 keV to Kuences of 1:8 × 1015 – 1:0×1016 cm−2 . According to TRIM-calculations [11] the Kuences were chosen in such a way, that the Ge pro:le shows a maximum concentration of 1, 3 and 6 at%. The as-implanted devices were divided into 3 sets and rapid thermal annealing (RTA) at 1000◦ C was performed for 6, 30 and 150 s. MOS capacitors for electrical measurements were fabricated using 300 nm thick sputtered layers of Al for both, the gate electrode (area 1 × 10−3 cm2 ) and the back contact. For EL investigations on top of the oxide layer a 80 nm thick indium–tin-oxide (ITO) layer was deposited using a sputtering process. The size of the EL-devices was 0:5 mm in diameter in a periodic pattern of 2 mm pitch. Current–voltage (IV )-measurements were performed with a Keithley 237 source measuring unit. A Keithley 590 CV -meter was used for high-frequency capacitance–voltage (HF-CV )-investigations. High:eld electron injection from the silicon substrate into the oxide was performed at constant current (Jinj = 2 × 10−5 A=cm2 ) at room temperature. Charge trapping has been studied by combined measurements of the change of the voltage V (t) which was applied to the MOS structures at constant current regime [12,13] and the shift of Kat-band voltage (OVFB ) of high-frequency (1 MHz) capacitance–voltage (C–V ) characteristics performed at de:nite intervals [12]. Additionally trapping characteristics were investigated by a method using a combination of IV and CV -measurements. Voltage ramps were driven up to a certain electric :eld and then a CV scan was recorded in order to determine the trapped charge.
Fig. 1. Schematic band structure for injection from the Si-substrate. The V (t) method is not sensitive to the region close to the injecting interface.
3. Results and discussion When using the V (t) and the CV method one has to consider the diPerent oxide regions the two methods are sensitive to. The V (t) method is insensitive to the region close to the injecting interface, since charges in this position do not lead to strong band bending. This means that for the case of injection from the Si substrate the charges trapped at the Si=SiO2 interface cannot be detected by this method (Fig. 1). Contrary to that CV investigations give information over the whole layer. It has to be mentioned that charges in the volume give a weaker contribution to the shift of the CV curve than charges trapped at the Si=SiO2 interface. High-:eld electron injection from the Si substrate was performed in a constant-current regime by applying a positive voltage to the gate contact. The used current density of 2 × 10−5 A=cm2 corresponds to the typical EL operation regime of such devices. The amount of injected charge is 1:25 × 1014 e=cm2 =s. The initial CV characteristics is shifted towards positive voltages due to the trapping of electrons in the oxide up to an injected charge of 2:9 × 1016 e=cm2 . Then the curve shifts back towards negative voltages. This means that positive charges are generated. In addition,
T. Gebel et al. / Physica E 16 (2003) 499 – 504
Fig. 2. Trapped charge after constant current stress for oxides implanted with 50 keV Ge (3%), treated with RTA at 1000◦ C for 150 s. Injection from the Si substrate was carried out by applying a positive voltage to the gate electrode. The charging was performed at a constant current density of 2 × 10−5 A=cm2 which corresponds to the typical EL operation regime of such devices The CV and the V (t) method were used. The + or − sign is related to the type of trapped charge.
the slope of the CV -characteristic decreases indicating that the density of interface states at the Si=SiO2 interface is increased. With the V (t) method, only negative charge trapping was detected. The results of the two methods are shown in Fig. 2. The observed ePects imply the following: the amount of the trapped charge derived from CV measurements is related to the total net charge, which consists of positive and negative charges in the oxide. Since the V (t) method shows only electron trapping and is not sensitive to the interface region in the case of injection from the Si substrate, there is a strong indication for the trapping of positive charge at the interface. Therefore, by using the two methods, the separation of the diPerent kinds of charge can be carried out—at least for the case of injection from the Si substrate. In Fig. 2 the diPerence of the calculated charges from the CV and the V (t) method is given as Qpos and represents positive charges. So the following scenario occurs during the constant current stress: Electrons are trapped in the volume of the SiO2 layer, but with increasing stress time also positive charges
501
are trapped near the SiO2 =Si interface region which leads to a decrease of the trapped net charge. The inKuence of the electric :eld on the chargetrapping is of great interest for the EL operation regime. The MOS-devices were stressed using well de:ned voltage ramps, starting from 0 V up to a :eld Estress and then a CV scan was performed. The unimplanted layer shows no electron trapping but the trapping of positive charges which increases for higher Estress (Fig. 3). For implanted layers at moderate electric :elds (5 –7 MV=cm) electron trapping is observed. However, in the high :eld region the trapping characteristics is changed. A back-shift of the CV curve is observed which indicates additional trapping of positive charge. For the explanation of this behavior several aspects have to be taken into consideration. Electrical stress at high electric :elds is known to cause additional defects in the oxide. Hot electrons which are present in the SiO2 conduction band at high electric :elds (E ¿ 7 MV=cm) [14,15] may perform impact ionization. Like in the case of Ge implantation, if traps are already initially present, trap-assisted impact ionization can also occur. Electron–hole pairs are formed, and because of the high kinetic energy of the impacting electrons bonds in the SiO2 structure may be broken. The holes which are produced during the impact events may be trapped, e.g. in neutral oxygen vacancies (NOV) [16]. A comparison of the trapping behavior for diPerent implantation energies and doses is given in Fig. 3. The charge is related to trapped electrons, so negative values mean positive trapped charges. The Ge implanted layers show the trapping of charges, with a maximum at a Ge concentration of 3% (Fig. 3a). For a concentration of 6% Ge no further increase of the trapped charge can be seen. This might be caused by the possible transformation of “trapping” defects into cluster-like structures. These nanoclusters are still a source for defects, e.g. at the surface or in the vicinity of the cluster, but the total number of defects might be decreased. Fig. 3b shows the trapped charge as a function of the implantation energy. The strong increase for higher energies indicates the importance of the trap position relative to the Si=SiO2 interface, which is in good correlation to our results from Rutherford-Backscattering Spectrometry (RBS) were a clear increase of Ge at the Si=SiO2 interface was observed (to be published elsewhere). Since microstructural and optical
502
T. Gebel et al. / Physica E 16 (2003) 499 – 504
Fig. 3. Trapped charge after diPerent voltage ramps. The sweeps were performed from 0 up to the electric :eld Estress . The value of Qtrap is related to the number of trapped electrons per cm2 .
Fig. 4. Structure of a neutral oxygen vacancy (NOV) (after [16]).
investigations did not give a strong indication for cluster related quantum ePects, other processes must cause the trapping of charge. The results of our investigations showed that the implanted Ge which is present in the oxide layer is responsible for the trapping of charge. But how do the trapping centers look like and how can the trapping mechanisms be understood? During the process of ion implantation and the following annealing steps E’ centers and ODCs are formed. The basic structure of such an oxygen de:ciency center, here in the special case of the NOV, is sketched in Fig. 4. After hole trapping, the Si–Si bond is broken and thus an E’ center is formed. Hori [16] describes this as an irreversible process being a strong precursor for beginning device breakdown. However, in Ref. [17] the electron trapping of E’ cen-
ters is described as a process leading to the formation of a NOV which means that the process could also be reversible. Since the devices are operated at very high electric :elds close to breakdown, the device degradation and the stability become the major issues for the EL-device performance. IV measurements have been carried out in combination with EL investigations. Since the spectra do not change for diPerent excitation :elds in the case of Ge implanted samples [8], the EL intensity can be characterized by the height of the main emission peak at 390 nm. Fig. 5a shows the EL intensity in arbitrary units as a function of the applied electric :eld for Ge+ implanted 80 nm thick SiO2 layers. The background was estimated to be 5 a:u. The plots of the three diPerent annealing procedures are shown. The onset of the EL appears to be nearly independent of both the implantation energy and the annealing conditions. This implies that the electric :eld has to reach a critical level before the onset of the EL occurs. It has to be mentioned that the background signal and the noise which was at a constant level up to the observed onset do not allow a clear statement for lower electric :elds with our equipment. But obviously after passing the onset level hot electrons with energies su%cient for impact excitation are present. In the high :eld regime (HFR) with E ¿ 9:5 MV=cm the curves show diPerences. With
T. Gebel et al. / Physica E 16 (2003) 499 – 504
503
Fig. 5. Combined IV and EL measurements on 80 nm thick SiO2 layers implanted with 40 keV Ge+ ions. Panel (a) shows the EL intensity as a function of the applied electric :eld, (b) on the injection current density. In panel (c) the EL intensity divided by the current is plotted in arbitrary units as a measure of EL e%ciency.
increasing annealing time higher EL intensities at the same electric :eld are achieved. This can be explained by the IV characteristics, which showed a shift of the IV curves towards lower electrical :elds with increasing annealing time. This means that more electrons are injected leading to a higher EL intensity. In Fig. 5b the EL intensity is plotted as a function of the current density. Regarding the annealing times, no remarkable diPerences can be observed between 6 and 30 s RTA treated samples. However, the 150 s RTA treatment leads to a decrease of the EL intensity and higher currents have to be applied. This is even better visible in Fig. 5c, where the ratio of the EL intensity divided by the current density is plotted versus the electric :eld. After a maximum peak a decrease of the EL intensity/current density is observed. This might be due to the trapping of the positive charges which occur at such high electric :elds and which leads to a reduction of the amount of luminescence centers regarding the proposed mechanism (Fig. 4).
4. Conclusion The investigation of charge trapping in Ge-rich oxide layers showed electron trapping at electric :elds up to 8 MV=cm. At higher electric :elds positive charge trapping, preferably in the SiO2 =Si interface region, was observed. The charge trapping is assumed to be caused by defects (NOV)—the same centers which cause the EL. Further investigations, e.g. electron spin
resonance (ESR) measurements are necessary for better a understanding of the trapping process. Acknowledgements This work has been supported by the German Bundesministerium fUur Bildung und Forschung (BMBF), contract WTR UKR 01/054.
References [1] Mel Ari Optoelectronic Road Map, http://www.cordis.le/ esprit/src/melop-rm.htm. [2] L.S. Liao, X.M. Bao, N.S. Li, X.Q. Zheng, N.B. Min, Solid State Commun. 97 (12) (1996) 1039. [3] P. KnYapek, B. Rezek, D. Muller, J.J. Grob, R. LYevy, K. LuterovYa, J. Kocka, I. Pelant, Phys. Stat. Sol. A 167 (1998) R5. [4] L. Rebohle, J. von Borany, R.A. Yankov, W. Skorupa, I.E.Tyschenko, H. FrUob, K. Leo, Appl. Phys. Lett. 71 (1997) 2809. [5] K.V. Shcheglov, C.M. Yang, K.J. Vahala, H.A. Atwater, Appl. Phys. Lett. 66 (1995) 745. [6] J.Y. Zhang, Y.H. Ye, X.L. Tan, Appl. Phys. Lett. 74 (17) (1999) 2459. [7] W.S. Lee, J.Y. Jeong, H.B. Kim, K.H. Chae, C.N. Whang, S. Im, J.H. Song, Mater. Sci. Eng. B 69 –70 (2000) 474. [8] L. Rebohle, J. von Borany, H. FrUob, W. Skorupa, Appl. Phys. B 71 (2000) 131. [9] T. Gebel, L. Rebohle, J. Zhao, D. Borchert, H. FrUob, J. von Borany, W. Skorupa, Mat. Res. Soc. Symp. Proc. 638 (2001) F18.1.1 (invited).
504
T. Gebel et al. / Physica E 16 (2003) 499 – 504
[10] T. Gebel, W. Skorupa, J. von Borany, L. Rebohle, D. Borchert, W. Fahrner, European Patent Application EP 1 132 975 A1 (2001). [11] J.F. Ziegler (Ed.), www.srim.org. [12] M.V. Fischetti, J. Appl. Phys. 57 (1985) 2860. [13] P. Samanta, C.K. Sarkur, Solid-State Electron. 41 (1997) 459. [14] M.V. Fischetti, D.J. DiMaria, S.D. Brorson, T.N. Theis, J.R. Kirtley, Phys. Rev. B 31 (1985) 8124.
[15] D. Arnold, E. Cartier, D.J. DiMaria, Phys. Rev. B 49 (1994) 10 278. [16] T. Hori, Gate Dielectrics and MOS ULSIs, in: Springer Series in Electronics and Photonics, Vol. 34, Springer-Verlag, Berlin, 1997. [17] A. Kalnitsky, J.P. Ellul, E.H. Poindexter, P.J. Caplan, R.A.S. Lux, A.R. Boothroyd, J. Appl. Phys. 67 (1990) 7359.
Physica E 16 (2003) 499 – 504 www.elsevier.com/locate/physe
Correlation of charge trapping and electroluminescence in highly e%cient Si-based light emitters T. Gebela; b;∗ , L. Rebohleb; c , J. Suna , W. Skorupaa; b , A.N. Nazarovd , I. Osiyukd a Forschungszentrum
Rossendorf, Bautzner Landstrasse 128, P.O. Box 510119, 01314 Dresden, Germany b Nanoparc GmbH, Dresden-Rossendorf, Germany c Vienna Technical University, Vienna, Austria d Institute of Semiconductor Physics, NASU Kiev, Ukraine
Abstract In this paper we report on recent results on charge trapping and electroluminescence (EL) from Ge rich SiO2 layers. Thermally grown 80 nm thick SiO2 layers were implanted with Ge ions at energies of 30 –50 keV to peak concentrations of 1–6 at%. Subsequently rapid thermal annealing was performed at 1000◦ C for 6, 30 and 150 s under a nitrogen atmosphere in order to form luminescence centers. A combination of capacitance–voltage (CV ) and current–voltage (IV ) methods was used for the investigation of the trapping properties. It was found that at electric :elds ¡ 8 MV=cm electron trapping dominates while at higher electric :elds which are typically required for the EL operation of the devices positive charge trapping occurs. It is assumed, that the trapping sites which are responsible for the trapping of the positive charge are in strong relation to the defects causing the luminescence. ? 2002 Elsevier Science B.V. All rights reserved. PACS: 73.40.Qv; 72.20.Jv; 72.10.Fk; 78.60.Fi; 78.67.Bf Keywords: Nanocluster rich oxides; Ion implantation; Electroluminescence; Charge trapping
1. Introduction Optical data transfer is the most promising tool to overcome the bottleneck of interchip and intrachip communication in future computer systems [1]. Since silicon is the key material for today’s microelectronic devices one is strongly interested in the integration of devices for the emission, modulation and detection of optical signals on one and the same chip, using the current Si-technology. Furthermore, the increasing market for integrated optical sensor systems and ∗ Corresponding author. Tel.: +49-351-2603036; fax: +49-3512603411. E-mail address: [email protected] (T. Gebel).
multi-functional microsystems (e.g. lab-on-chip systems in biotechnology) also causes a strong demand on miniaturized and integrated optical systems based on silicon technology. A promising method for the formation of Si-based light emitting structures is ion beam synthesis. Electroluminescence (EL) from Si implanted oxide layers was observed in the red [2] and blue wavelength range [3,4] and from Ge implanted SiO2 layers EL was found in the red/infrared and in the blue/violet spectral region [4–8]. In our previous work we demonstrated strong violet EL with a power e%ciency of 0.5% from 130 nm thick SiO2 layers implanted with Ge [9] and demonstrated the prototype of a silicon based optocoupler [10]. We found evidence for an oxygen de:ciency center (ODC) as the
1386-9477/03/$ - see front matter ? 2002 Elsevier Science B.V. All rights reserved. PII: S 1 3 8 6 - 9 4 7 7 ( 0 2 ) 0 0 6 3 6 - 7
500
T. Gebel et al. / Physica E 16 (2003) 499 – 504
luminescence center responsible for the PL excited around 5 eV. Possible ODCs are the neutral oxygen vacancy (≡ Si–Si ≡, ≡ Si–Ge or ≡ Ge–Ge ≡) and the twofold coordinated Si or Ge atom. In this work the EL and the charge trapping properties of Ge implanted 80 nm thick SiO2 layers are investigated to get a better understanding about the injection and charge transport mechanism in this system.
2. Experimental Thermally grown 80 nm thick SiO2 :lms on (1 0 0) n-type Si wafers were implanted with Ge+ ions at energies of 30 –50 keV to Kuences of 1:8 × 1015 – 1:0×1016 cm−2 . According to TRIM-calculations [11] the Kuences were chosen in such a way, that the Ge pro:le shows a maximum concentration of 1, 3 and 6 at%. The as-implanted devices were divided into 3 sets and rapid thermal annealing (RTA) at 1000◦ C was performed for 6, 30 and 150 s. MOS capacitors for electrical measurements were fabricated using 300 nm thick sputtered layers of Al for both, the gate electrode (area 1 × 10−3 cm2 ) and the back contact. For EL investigations on top of the oxide layer a 80 nm thick indium–tin-oxide (ITO) layer was deposited using a sputtering process. The size of the EL-devices was 0:5 mm in diameter in a periodic pattern of 2 mm pitch. Current–voltage (IV )-measurements were performed with a Keithley 237 source measuring unit. A Keithley 590 CV -meter was used for high-frequency capacitance–voltage (HF-CV )-investigations. High:eld electron injection from the silicon substrate into the oxide was performed at constant current (Jinj = 2 × 10−5 A=cm2 ) at room temperature. Charge trapping has been studied by combined measurements of the change of the voltage V (t) which was applied to the MOS structures at constant current regime [12,13] and the shift of Kat-band voltage (OVFB ) of high-frequency (1 MHz) capacitance–voltage (C–V ) characteristics performed at de:nite intervals [12]. Additionally trapping characteristics were investigated by a method using a combination of IV and CV -measurements. Voltage ramps were driven up to a certain electric :eld and then a CV scan was recorded in order to determine the trapped charge.
Fig. 1. Schematic band structure for injection from the Si-substrate. The V (t) method is not sensitive to the region close to the injecting interface.
3. Results and discussion When using the V (t) and the CV method one has to consider the diPerent oxide regions the two methods are sensitive to. The V (t) method is insensitive to the region close to the injecting interface, since charges in this position do not lead to strong band bending. This means that for the case of injection from the Si substrate the charges trapped at the Si=SiO2 interface cannot be detected by this method (Fig. 1). Contrary to that CV investigations give information over the whole layer. It has to be mentioned that charges in the volume give a weaker contribution to the shift of the CV curve than charges trapped at the Si=SiO2 interface. High-:eld electron injection from the Si substrate was performed in a constant-current regime by applying a positive voltage to the gate contact. The used current density of 2 × 10−5 A=cm2 corresponds to the typical EL operation regime of such devices. The amount of injected charge is 1:25 × 1014 e=cm2 =s. The initial CV characteristics is shifted towards positive voltages due to the trapping of electrons in the oxide up to an injected charge of 2:9 × 1016 e=cm2 . Then the curve shifts back towards negative voltages. This means that positive charges are generated. In addition,
T. Gebel et al. / Physica E 16 (2003) 499 – 504
Fig. 2. Trapped charge after constant current stress for oxides implanted with 50 keV Ge (3%), treated with RTA at 1000◦ C for 150 s. Injection from the Si substrate was carried out by applying a positive voltage to the gate electrode. The charging was performed at a constant current density of 2 × 10−5 A=cm2 which corresponds to the typical EL operation regime of such devices The CV and the V (t) method were used. The + or − sign is related to the type of trapped charge.
the slope of the CV -characteristic decreases indicating that the density of interface states at the Si=SiO2 interface is increased. With the V (t) method, only negative charge trapping was detected. The results of the two methods are shown in Fig. 2. The observed ePects imply the following: the amount of the trapped charge derived from CV measurements is related to the total net charge, which consists of positive and negative charges in the oxide. Since the V (t) method shows only electron trapping and is not sensitive to the interface region in the case of injection from the Si substrate, there is a strong indication for the trapping of positive charge at the interface. Therefore, by using the two methods, the separation of the diPerent kinds of charge can be carried out—at least for the case of injection from the Si substrate. In Fig. 2 the diPerence of the calculated charges from the CV and the V (t) method is given as Qpos and represents positive charges. So the following scenario occurs during the constant current stress: Electrons are trapped in the volume of the SiO2 layer, but with increasing stress time also positive charges
501
are trapped near the SiO2 =Si interface region which leads to a decrease of the trapped net charge. The inKuence of the electric :eld on the chargetrapping is of great interest for the EL operation regime. The MOS-devices were stressed using well de:ned voltage ramps, starting from 0 V up to a :eld Estress and then a CV scan was performed. The unimplanted layer shows no electron trapping but the trapping of positive charges which increases for higher Estress (Fig. 3). For implanted layers at moderate electric :elds (5 –7 MV=cm) electron trapping is observed. However, in the high :eld region the trapping characteristics is changed. A back-shift of the CV curve is observed which indicates additional trapping of positive charge. For the explanation of this behavior several aspects have to be taken into consideration. Electrical stress at high electric :elds is known to cause additional defects in the oxide. Hot electrons which are present in the SiO2 conduction band at high electric :elds (E ¿ 7 MV=cm) [14,15] may perform impact ionization. Like in the case of Ge implantation, if traps are already initially present, trap-assisted impact ionization can also occur. Electron–hole pairs are formed, and because of the high kinetic energy of the impacting electrons bonds in the SiO2 structure may be broken. The holes which are produced during the impact events may be trapped, e.g. in neutral oxygen vacancies (NOV) [16]. A comparison of the trapping behavior for diPerent implantation energies and doses is given in Fig. 3. The charge is related to trapped electrons, so negative values mean positive trapped charges. The Ge implanted layers show the trapping of charges, with a maximum at a Ge concentration of 3% (Fig. 3a). For a concentration of 6% Ge no further increase of the trapped charge can be seen. This might be caused by the possible transformation of “trapping” defects into cluster-like structures. These nanoclusters are still a source for defects, e.g. at the surface or in the vicinity of the cluster, but the total number of defects might be decreased. Fig. 3b shows the trapped charge as a function of the implantation energy. The strong increase for higher energies indicates the importance of the trap position relative to the Si=SiO2 interface, which is in good correlation to our results from Rutherford-Backscattering Spectrometry (RBS) were a clear increase of Ge at the Si=SiO2 interface was observed (to be published elsewhere). Since microstructural and optical
502
T. Gebel et al. / Physica E 16 (2003) 499 – 504
Fig. 3. Trapped charge after diPerent voltage ramps. The sweeps were performed from 0 up to the electric :eld Estress . The value of Qtrap is related to the number of trapped electrons per cm2 .
Fig. 4. Structure of a neutral oxygen vacancy (NOV) (after [16]).
investigations did not give a strong indication for cluster related quantum ePects, other processes must cause the trapping of charge. The results of our investigations showed that the implanted Ge which is present in the oxide layer is responsible for the trapping of charge. But how do the trapping centers look like and how can the trapping mechanisms be understood? During the process of ion implantation and the following annealing steps E’ centers and ODCs are formed. The basic structure of such an oxygen de:ciency center, here in the special case of the NOV, is sketched in Fig. 4. After hole trapping, the Si–Si bond is broken and thus an E’ center is formed. Hori [16] describes this as an irreversible process being a strong precursor for beginning device breakdown. However, in Ref. [17] the electron trapping of E’ cen-
ters is described as a process leading to the formation of a NOV which means that the process could also be reversible. Since the devices are operated at very high electric :elds close to breakdown, the device degradation and the stability become the major issues for the EL-device performance. IV measurements have been carried out in combination with EL investigations. Since the spectra do not change for diPerent excitation :elds in the case of Ge implanted samples [8], the EL intensity can be characterized by the height of the main emission peak at 390 nm. Fig. 5a shows the EL intensity in arbitrary units as a function of the applied electric :eld for Ge+ implanted 80 nm thick SiO2 layers. The background was estimated to be 5 a:u. The plots of the three diPerent annealing procedures are shown. The onset of the EL appears to be nearly independent of both the implantation energy and the annealing conditions. This implies that the electric :eld has to reach a critical level before the onset of the EL occurs. It has to be mentioned that the background signal and the noise which was at a constant level up to the observed onset do not allow a clear statement for lower electric :elds with our equipment. But obviously after passing the onset level hot electrons with energies su%cient for impact excitation are present. In the high :eld regime (HFR) with E ¿ 9:5 MV=cm the curves show diPerences. With
T. Gebel et al. / Physica E 16 (2003) 499 – 504
503
Fig. 5. Combined IV and EL measurements on 80 nm thick SiO2 layers implanted with 40 keV Ge+ ions. Panel (a) shows the EL intensity as a function of the applied electric :eld, (b) on the injection current density. In panel (c) the EL intensity divided by the current is plotted in arbitrary units as a measure of EL e%ciency.
increasing annealing time higher EL intensities at the same electric :eld are achieved. This can be explained by the IV characteristics, which showed a shift of the IV curves towards lower electrical :elds with increasing annealing time. This means that more electrons are injected leading to a higher EL intensity. In Fig. 5b the EL intensity is plotted as a function of the current density. Regarding the annealing times, no remarkable diPerences can be observed between 6 and 30 s RTA treated samples. However, the 150 s RTA treatment leads to a decrease of the EL intensity and higher currents have to be applied. This is even better visible in Fig. 5c, where the ratio of the EL intensity divided by the current density is plotted versus the electric :eld. After a maximum peak a decrease of the EL intensity/current density is observed. This might be due to the trapping of the positive charges which occur at such high electric :elds and which leads to a reduction of the amount of luminescence centers regarding the proposed mechanism (Fig. 4).
4. Conclusion The investigation of charge trapping in Ge-rich oxide layers showed electron trapping at electric :elds up to 8 MV=cm. At higher electric :elds positive charge trapping, preferably in the SiO2 =Si interface region, was observed. The charge trapping is assumed to be caused by defects (NOV)—the same centers which cause the EL. Further investigations, e.g. electron spin
resonance (ESR) measurements are necessary for better a understanding of the trapping process. Acknowledgements This work has been supported by the German Bundesministerium fUur Bildung und Forschung (BMBF), contract WTR UKR 01/054.
References [1] Mel Ari Optoelectronic Road Map, http://www.cordis.le/ esprit/src/melop-rm.htm. [2] L.S. Liao, X.M. Bao, N.S. Li, X.Q. Zheng, N.B. Min, Solid State Commun. 97 (12) (1996) 1039. [3] P. KnYapek, B. Rezek, D. Muller, J.J. Grob, R. LYevy, K. LuterovYa, J. Kocka, I. Pelant, Phys. Stat. Sol. A 167 (1998) R5. [4] L. Rebohle, J. von Borany, R.A. Yankov, W. Skorupa, I.E.Tyschenko, H. FrUob, K. Leo, Appl. Phys. Lett. 71 (1997) 2809. [5] K.V. Shcheglov, C.M. Yang, K.J. Vahala, H.A. Atwater, Appl. Phys. Lett. 66 (1995) 745. [6] J.Y. Zhang, Y.H. Ye, X.L. Tan, Appl. Phys. Lett. 74 (17) (1999) 2459. [7] W.S. Lee, J.Y. Jeong, H.B. Kim, K.H. Chae, C.N. Whang, S. Im, J.H. Song, Mater. Sci. Eng. B 69 –70 (2000) 474. [8] L. Rebohle, J. von Borany, H. FrUob, W. Skorupa, Appl. Phys. B 71 (2000) 131. [9] T. Gebel, L. Rebohle, J. Zhao, D. Borchert, H. FrUob, J. von Borany, W. Skorupa, Mat. Res. Soc. Symp. Proc. 638 (2001) F18.1.1 (invited).
504
T. Gebel et al. / Physica E 16 (2003) 499 – 504
[10] T. Gebel, W. Skorupa, J. von Borany, L. Rebohle, D. Borchert, W. Fahrner, European Patent Application EP 1 132 975 A1 (2001). [11] J.F. Ziegler (Ed.), www.srim.org. [12] M.V. Fischetti, J. Appl. Phys. 57 (1985) 2860. [13] P. Samanta, C.K. Sarkur, Solid-State Electron. 41 (1997) 459. [14] M.V. Fischetti, D.J. DiMaria, S.D. Brorson, T.N. Theis, J.R. Kirtley, Phys. Rev. B 31 (1985) 8124.
[15] D. Arnold, E. Cartier, D.J. DiMaria, Phys. Rev. B 49 (1994) 10 278. [16] T. Hori, Gate Dielectrics and MOS ULSIs, in: Springer Series in Electronics and Photonics, Vol. 34, Springer-Verlag, Berlin, 1997. [17] A. Kalnitsky, J.P. Ellul, E.H. Poindexter, P.J. Caplan, R.A.S. Lux, A.R. Boothroyd, J. Appl. Phys. 67 (1990) 7359.