Q24 ECS Journal of Solid State Science and Technology, 3 (2) Q24-Q27 (2014) 2162-8769/2014/3(2)/Q24/4/$31.00©The Electrochemical Society
Effect of Mg Doping on the Electrical Characteristics of High
Performance IGZO Thin Film Transistors
Hung-Chi Wu,aTung-Sheng Liu,aand Chao-Hsin Chiena,b,z
aDepartment of Electronics Engineering and Institute of Electronics, National Chiao Tung University, Hsinchu, Taiwan
bNational Nano Device Laboratories, Hsinchu, Taiwan
In this work, we report that the electrical characteristics of Indium-Gallium-Zinc-Oxide Thin Film Transistors (IGZO-TFTs) can be improved by Mg doping in the IGZO layer. Several composite IGZO/Mg-doped IGZO channel structures were employed. With appropriate Mg-doped IGZO thickness, the electrical properties such as mobility, subthreshold swing (S.S.) and on/off ratio can be significantly improved by Mg doping. Mg doping in IGZO layer can stabilize device performance whatever the device is stressed under positive gate bias or negative gate bias with illumination. By X-ray photoelectron (XPS) analysis, we observe the presence of Mg can change the oxygen bonding states and influence the properties of IGZO layer. The atomic percentage of Mg in the IGZO layer is about 6.53%. As a result, Mg doping is a useful method to enhance the performance and stability of IGZO-TFTs. © 2013 The Electrochemical Society. [DOI:10.1149/2.012402jss] All rights reserved.
Manuscript submitted August 8, 2013; revised manuscript received November 19, 2013. Published December 4, 2013.
IGZO thin film transistors have attracted industrial and commercial interests in recent years and now there have already been new prod-ucts using IGZO. It has the advantages including low temperature processing, large area fabrication and being compatible with flexi-ble electronics.1Its wide applications to flexible electronics, TFTs, electronic papers and sensors make amorphous oxide semiconduc-tors (AOS) have the possibility to be the prevailing material of next generation TFT technology.2 Even though there are many kinds of AOS materials, IGZO is one of the most promising candidates. The TFTs with IGZO channel layer usually exhibit superior electrical per-formance such as high channel mobility, a good on/off ratio, a small S.S. and an acceptable threshold voltage (VT).3These features make IGZO be more attractive than other materials such as amorphous silicon, poly-silicon and organic conjuagated polymers for the TFT technology.4
Recently, many scientists have been focusing on understanding and improving the properties of IGZO layer by many different ways. Some focused on the IGZO composition adjustment; they tried to find the optimized electrical performance by understanding what kinds of function these atoms have in the oxide semiconductor.5,6 Some groups focused on varying oxygen flow during deposition because the properties of IGZO are known to be very sensitive to the deposition condition.7 Also, Kim et al. tried to replace the metal atom in the IGZO layer by other kind of metal element, for example, Hf-In-Zn-O (HIZHf-In-Zn-O).8 The purpose of all the researches is to make the AOS material have better electrical performance and more stable reliability such as bias, illumination and environment-stabilities.
In this work, we systematically investigate the effect of Mg doping on the electrical properties of IGZO-TFTs. We find the TFTs with composite IGZO/ Mg-doped IGZO channel structure possess better electrical performance than that of the conventional IGZO-TFTs in the aspects of on/off ratio and S.S.. By XPS analysis, we observe the presence of Mg can change the oxygen bonding states and influence the properties of IGZO layer. As a result, the electrical performance of the IGZO-TFTs can be improved by using the composite IGZO/Mg-doped IGZO channel.
In this study, we used bottom-gate top-contact device structure. After standard RCA cleaning, we used an n++Si wafer as the starting substrate and its back side was coated by Al as the gate electrode. A 200 nm SiOxlayer was formed by plasma enhanced chemical vapor deposition system and its capacitance was 1.38× 10−8F/cm2. Then, we defined the channel region by sputter and lithography. The device channel width and length were 200 and 50μm, respectively. In this step, we employed two different recipes to deposit the channel layer. One was Mg and IGZO co-sputter (M-I) and the other was IGZO (I) sputter only. In other words, we varied the channel structure. The DC power for Mg and IGZO targets were 50 W and 70 W respectively.
zE-mail:[email protected]
The Ar/O2 ratio was 24/6 SCCM (SCCM denotes cubic centimeter per minute). Keeping the thickness of active layer to be c.a. 30 nm, the different combinations were 300 Å I (Sample A), 10 Å M-I+290 Å I(Sample B), 50 Å M-I+250 Å I (Sample C), and 300 Å M-I (Sample D), respectively. After the channel deposition, we defined source and drain electrodes region by lithography and deposited 50 nm Ti film by E-gun; the device channel width and length were 200 and 50μm, respectively. All samples were annealed at 300◦C with N2flow for 30 minutes. The transistor performances were measured by HP4156 semiconductor parameter analyzer, the drain voltage was set to 5 V and the gate voltage was swept from−20 V to 20 V. We also measured its output characteristics with the drain voltages from 0 V to 20 V and the gate voltage variation between 0 V to 20 V in steps of 5 V. The thin film properties were measured by XPS.
Figure 1shows the transfer characteristics of our IGZO TFTs. TableIsummarize the electrical parameters for all the samples. All parameters are obtained based on the conventional transistor current equation. We can see sample A in which the channel was composed of pure IGZO depicts good electrical performance. The carrier mo-bility is of 12.29 (cm2/V sec), the S.S. value is of 461 (mV/decade), the on/off ratio is of 1.62× 108and the V
T is of−2.1 V. All these parameters of sample A indicate that the performance of our control IGZO TFTs is comparable with those reported in the other works.9,10 It can be found the electrical performances of sample B are better than sample A, it means Mg doping can improve the IGZO-TFT perfor-mances. We also measured the Hall mobility and carrier concentration. The Hall mobility for IGZO and Mg-doped IGZO is 15.37 and 26.13 (cm2/V sec) and the carrier concentration is 1.43× 1018and 9.18× 1017(/cm3), respectively. Mg doping can reduce the carrier concentra-tion because of the reducconcentra-tion in oxygen vacancy density, but increase the Hall mobility resulting from the bonding state change in the IGZO layer. The properties of the IGZO-TFTs are mainly determined by the ZnO system11so that we can explain what we have observed in the Mg-doped IGZO-TFTs according to the results in the ZnO system.
-20 -10 0 10 20 10-13 10-12 10-11 1010-10 -9 10-8 10-7 10-6 10-5
Drain current (A)
Gate voltage (V)
Sample A Sample B Sample C Sample D
Figure 1. Transfer characteristics ID−VGof TFTs with different thickness of
Mg-doping. The channel length and width of devices are 50 and 200μm. ) unless CC License in place (see abstract). ecsdl.org/site/terms_use
address. Redistribution subject to ECS terms of use (see 140.113.38.11
ECS Journal of Solid State Science and Technology, 3 (2) Q24-Q27 (2014) Q25
Table I. Comparison of the various parameters for the sample A, B, C and D.
Mobility S.S. On/Off
Sample VT(V) (cm2/ V sec) (mV/decade) ratio
Sample A −2.1 12.29 461 1.62× 108
Sample B −1.3 16.81 347 4.29× 108
Sample C 0.2 17.36 331 5.96× 108
Sample D 4.9 4.97 355 1.09× 108
Remashan et al.12reported the Mg doped ZnO as a buffer layer can improve the electrical performances of the ZnO-TFTs; they attributed to the resultant grain size with the addition of Mg doping. The larger the grain size, the lower defect density in the IGZO is. This is con-sistent with our measuring results. Besides, the Mg2+substitution in the Zn2+site does not lead to significant lattice distortion due to their similar ionic radii.13Therefore, Mg atoms can effectively be doped in the IGZO system and improve the electrical performances. Clearly, samples B and C have higher on/off ratio, lower VTnear 0 V, larger mobility and smaller S.S. value with the comparison of sample A. These results imply that Mg doping can change the bulk property of IGZO, which is closely related to the bonding states in IGZO; it does not degrade the property of the interface between the IGZO channel and the SiOxdielectric layer. However, with increasing the thickness of Mg-doped IGZO layer to 30 nm (Sample D), we can observe the degraded electrical performance, such as obvious on current reduction and moving away of VTfrom 0 V. These results are similar to those in the paper reported by Remashan et al.12and might be attributed to an increased effective mass of the electrons.14Hence, Mg doping in the IGZO layer can improve the transistor performance effectively and be able to adjust VTtoward the positive direction. Besides, the thickness of Mg-doped IGZO layer shall be optimized.
Figure2shows the output characteristics of all samples. It should be noted that the drain current scale of sample D is different from those of the other samples. It can be observed the currents of samples A, B and C saturate in high drain voltages under each gate voltage. But, it is different for sample D; its current does not saturate in high drain voltages, especially under high gate voltages. We also checked the contact resistance by transmission line method (TLM).15The contact resistances for sample A, B, C and D are 598, 538, 497 and 3317 (-cm), respectively. These values, especially for sample D, mean that the Mg doped IGZO layer does not form a good contact with the Ti electrodes. We think there might be the formation of oxide at the interfacial layer between the Mg-doped IGZO and Ti. The oxide resulting from the Ti deposition certainly degrades the conductivity and the addition of Mg in IGZO even makes this situation worse. In
0 5 10 15 20 0 20 40 60 80 D ra in cu rren t (µ A) Drain voltage (V) 0V 5V 10V 15V 20V (a) 0 5 10 15 20 0 2 4 6 8 10 D ra in cu rren t (µ A) Drain voltage (V) 0V 5V 10V 15V 20V (d) 0 5 10 15 20 0 20 40 60 80 Drain voltage (V) D ra in cu rren t (µ A) 0V 5V 10V 15V 20V (c) 0 5 10 15 20 0 20 40 60 80 D ra in cu rren t (µ A) 0V 5V 10V 15V 20V (b) Drain voltage (V)
Figure 2. ID−VDoutput characteristics of sample A, B, C and D.
1.5 2.0 2.5 3.0 3.5 4.0 0 200 400 600 800 1000 1200 1400 (a E ) 1/ 2 (c m -1 eV ) 1/ 2
Photon energy (eV) Mg-IGZO IGZO
3.24 eV
3.49 eV
Figure 3. Extrapolated bandgap values for IGZO and Mg-doped IGZO films
by the transmittance spectra.
other words, we can say that the potential barrier between Ti and the Mg-doped IGZO layer is higher than that between Ti and the undoped IGZO, and then increases the difficulty of carrier injection from the Ti electrode into the Mg-doped IGZO layer. This might be ascribed to the grain size resulting from Mg doping. Fang et al. have ever reported that the grain size of ZnO is highly dependent on the ratio of Mg doping.16 The defect density is closely related with the grain size; the Mg-doped IGZO layer in this work is supposed to have small grains and thus high defect density. We think they do not influence the electrical performance during device operation because the impact of defects can be significantly suppressed by the large amount of accumulation carrier. However, the defects can affect the carrier injection from the electrode to the active layer. Hence, the contact between Ti and the Mg-doped IGZO layer is deteriorated. Therefore, the effect of Mg doping can increase the electrical performance of IGZO-TFTs and improve the interface between the IGZO and SiOx layer. However, Mg-IGZO layer cannot form good contact with Ti. The situation at the interface between the Mg-doped IGZO and Ti is quite complicated so that it should be further investigated.
Figure3shows the extrapolated bandgap values for the undoped IGZO and the Mg-doped IGZO films by the transmittance spectra. The bandgap of IGZO is increased from 3.24 to 3.49 eV with the addition of Mg during sputter process. There might be the formation of MgOx during the Mg-doped IGZO deposition by the co-sputter process. MgOx is not conductive and belongs to the wide bandgap materials. The bandgap of MgOxis 7.8 eV which is higher than that of IGZO.17 The bandgap of IGZO can be widened by the existence of MgOx. The formation of MgOxcan suppress the oxygen vacancy generation and change the oxygen bonding states in IGZO, which is beneficial in stabilizing the IGZO-TFTs during operation. The ratio of MgOxshall not be too high to degrade the property of IGZO thin film.18Therefore, the effect of Mg doping on the IGZO can improve the electrical performance of the IGZO-TFTs.
Figure4shows the transfer curve shift of all samples as a function of gate bias stress duration. The stress voltage was 20 V. We can clearly observe that there is no obvious S.S. change after stressing; implying there is no new traps created at the interface during stress. The VTshift for samples A, B, C and D is 4.6, 3.4, 1.2 and 1.3 (V), respectively. The suppression of VTshift by Mg doping is remarkable when the thickness of M-I layer is above 50 Å. Mg doping can help in forming MgOx in IGZO and in turn suppress the oxygen vacancy generation. The oxygen vacancies are thought to be able to give the electrons during stress; the VTshift is caused by the electron trapping inside the gate dielectric and the interface between the channel and the gate insulator. Since MgOxhas very strong Mg-O bond, the electrons thus are suppressed during operation. Hence, Mg doping in IGZO can improve the reliability of the IGZO-TFTs against the positive bias stress. The reduction in oxygen vacancy density by adding Mg into IGZO will be addressed in the following statement.
Figure5shows the transfer curves of all samples as a function of gate voltage for various stress durations under white light illumination. The stress voltage was−20 V. The VTshift for samples A, B, C and D is−6.1, −3.4, −2.8 and −1.2 (V), respectively. The tail curve just as when the device is turned on can be observed in sample A after
) unless CC License in place (see abstract). ecsdl.org/site/terms_use
address. Redistribution subject to ECS terms of use (see 140.113.38.11
Q26 ECS Journal of Solid State Science and Technology, 3 (2) Q24-Q27 (2014) -20 -10 0 10 20 10-12 10-11 1010-10 -9 10-8 10-7 10-6 10-5 10-4 D ra in cu rren t ( A ) Gate voltage (V) 0sec 100sec 1000sec 2000sec 3600sec -20 -10 0 10 20 10-12 10-11 1010-10 -9 10-8 10-7 10-6 10-5 10-4 D ra in cu rren t ( A ) Gate voltage (V) 0sec 100sec 1000sec 2000sec 3600sec -20 -10 0 10 20 10-12 10-11 1010-10 -9 10-8 10-7 10-6 10-5 10-4 D ra in cu rren t ( A ) Gate voltage (V) 0sec 100sec 1000sec 2000sec 3600sec (d) (c) (b) -20 -10 0 10 20 10-12 10-11 10-10 10-9 10-8 10-7 10-6 10-5 D ra in cu rren t ( A ) Gate voltage (V) 0sec 100sec 1000sec 2000sec 3600sec (a)
Figure 4. Transfer curves of the IGZO-TFTs as a function of stress time for
samples (a) A, (b) B, (c) C and (d) D. Stress was performed with a gate bias of 20 V.
a certain time stress under illumination. However, this abnormality can be suppressed when Mg doping is added in the IGZO layer. The mechanism of degradation is caused by the electron-hole pair generation under illumination. The shorter the wavelength, the severer the degradation is.19The negative gate bias can lead to hole trapping inside the gate dielectric and the interface between the channel and the gate insulator. Since Mg doping in IGZO can lead to the formation of MgOx, the bandgap of IGZO thus increases. Higher bandgap can reduce the electron-hole pair generation under the illumination with the shorter light wavelength; the stability of the IGZO-TFTs for the negative bias stress under illumination can then be improved. As a result, we speculate that the MgOx formation might be main factor in stabilizing the reliability of the IGZO-TFTs under negative bias stressing with illumination after the addition of Mg atoms.
Figure6illustrates the O1sXPS narrow scan spectra of the IGZO thin film with and without Mg doping. It should be noted that all data have been calibrated with respect to the carbon signal. There is no obvious shape change for In, Ga and Zn elements (data not shown) implying that Mg elements do not interact with other metal elements in the IGZO layer. Besides, the atomic percentage of Mg in the IGZO layer is about 6.53%, which means Mg is still the minor element in the IGZO layer. Therefore, we assume the structure of IGZO with minor Mg doping will not change significantly and remains similar to that
-20 -10 0 10 20 10-12 10-11 10-10 10-9 10-8 10-7 10-6 10-5 D ra in cu rren t ( A ) Gate voltage (V) Sample D 0sec 100sec 1000sec 2000sec 3600sec -20 -10 0 10 20 10-12 10-11 10-10 10-9 10-8 10-7 10-6 10-5 D rai n c u rr ent ( A ) Gate voltage (V) Sample C 0sec 100sec 1000sec 2000sec 3600sec (d) (c) (b) -20 -10 0 10 20 10-12 10-11 10-10 10-9 10-8 10-7 10-6 10-5 Sample A 0sec 100sec 1000sec 2000sec 3600sec D ra in cu rren t ( A ) Gate voltage (V) -20 -10 0 10 20 10-12 10-11 10-10 10-9 10-8 10-7 10-6 10-5 D ra in cu rren t ( A ) Gate voltage (V) 0sec 100sec 1000sec 2000sec 3600sec Sample B (a)
Figure 5. Transfer curves of the IGZO-TFTs as a function of stress time
with light illumination for samples (a) A, (b) B, (c) C and (d) D. Stress was performed with a gate bias of−20 V.
538 536 534 532 530 528 526 524 OI OII OIII In ten si ty (a .u .)
Binding energy (eV) 538 536 534 532 530 528 526 524 In te n sit y ( a .u .)
Binding energy (eV) OI
OII
OIII
(b) (a)
Figure 6. O 1s XPS narrow scan spectra on the IGZO thin film surface
with-and without Mg doping with-and its bonding states analysis.
of undoped IGZO. The O1ssignal has an obvious shape change and becomes broader after the addition of Mg. MgOx is not conductive and its presence can widen the energy bandgap of IGZO.20This result means that the bonding states in the IGZO layer have been changed with the Mg addition. We also deconvoluted the O spectra to inspect the change of bonding states, as shown in Fig.6. The O 1s peak can be fitted by three nearly Gaussian distributions, respectively, centered at OI(529.96± 0.1 eV), OII(531.55± 0.1 eV), and OIII(532.7± 0.1 eV). OIis related to the O2−ion in the lattice surrounded by the Zn, Ga and In atoms in the IGZO compound system. OIIis associated with O2− ions that are in oxygen deficient regions within the matrix of IGZO. OIIIis usually attributed to the presence of loosely bound oxygen on the surface of the film termed the specific chemisorbed oxygen.21,22 It can be found there is an increase in OI and OII, and the ratio of OIIIis reduced with Mg doping. The obvious increase in the amount of OIIshows the oxygen vacancy can be reduced with Mg doping. Moreover, the bonding energy of MgO (393.7 kJ/mole) is higher than ZnO (284.1 kJ/mol) at 298 K.23It is hard to break the MgO bond as compared to the ZnO bond. Thus, we think the increase in OIIratio, the MgOxformation in IGZO and the high bonding energy of MgO bond result in good stability of the IGZO-TFTs during operation. This indicates that Mg doping can enhance the ratio of O2−ion in the lattice surrounded by the Zn, Ga and In atoms in the IGZO compound system, and increase the ratio of oxygen deficient regions within the matrix of IGZO. However, the ratios of indium, gallium and zinc oxides cannot be convincingly identified at this stage; further investigation is under its way for clarifying the ratios of various oxides. But, we still believe, according to our experimental data, the presence of Mg can change the intrinsic properties of the IGZO thin film and in turn influence the electrical characteristics of IGZO-TFTs.
In this study, the effects of Mg doping on the electrical properties of IGZO-TFTs have been investigated. The TFTs with composite IGZO/Mg-doped IGZO channel depict better electrical performance than the IGZO-TFTs without Mg doping in terms of on/off ratio and S.S.. We measured the output characteristics and also analyzed the contact resistance by TLM. The mobility degradation of Mg-doped IGZO-TFTs is due to the poorer contact resistance between the Mg doped IGZO layer and the Ti electrode, which is supposed to come from the smaller grain size in the Mg-doped IGZO. From the bandgap measurement by the optical experiment, the bandgap can be tailored by the ratio of Mg doping; however, it should be optimized. Mg doping in IGZO can suppress the effect of positive and negative biases effectively. By XPS analysis, it can be observed the composition in the IGZO layer does change with Mg doping. The presence of Mg can change the oxygen bonding states and influence the properties of IGZO layer. Consequently, Mg doping can improve the electrical performance of the IGZO-TFTs and the optimization such as Mg concentration and the thickness of Mg-doped IGZO layer would be studied in the future.
References
1. T. Kamiya, K. Nomura, and H. Hosono,Sci. Technol. Adv. Mater., 11, 044305 (2010). 2. H. Yabuta, M. Sano, K. Abe, T. Aiba, T. Den, H. Kumomi, K. Nomura, T. Kamiya,
and H. Hosono,Appl. Phys. Lett., 89, 112123 (2006).
) unless CC License in place (see abstract). ecsdl.org/site/terms_use
address. Redistribution subject to ECS terms of use (see 140.113.38.11
ECS Journal of Solid State Science and Technology, 3 (2) Q24-Q27 (2014) Q27
3. K. Nomura, H. Ohta, A. Takagi, T. Kamiya, M. Hirano, and H. Hosono,Nature, 432, 488 (2004).
4. S. Yang, C.-S. Hwang, J.-I. Lee, S.-M. Yoon, M.-K. Ryu, K.-I. Cho, S.-H. K. Park, S.-H. Kim, C.-E. Park, and J. Jang,Appl. Phys. Lett., 98, 103515 (2011). 5. H. Hosono,J. Non-Cryst. Solids, 352, 851 (2006).
6. K. Nomura, A. Takagi, T. Kamiya, H. Ohta, M. Hirano, and H. Hosono,J. J. Appl. Phys., 45, 4303 (2006).
7. Y.-S. Lee, W.-J. Chen, J.-S. Huang, and S.-C. Wu,Thin Solid Films, 520, 6942 (2012).
8. C.-J. Kim, S. Kim, J.-H. Lee, J.-S. Park, S. Kim, J. Park, E. Lee, J. Lee, Y. Park, J. H. Kim, S. T. Shin, and U.-I. Chung,Appl. Phys. Lett., 95, 252103 (2009). 9. H.-S. Uhm, S.-H. Lee, W. Kim, and J.-S. Park,IEEE Electron Device Lett., 33, 543
(2012).
10. L.-F. Teng, P.-T. Liu, Y.-J. Lo, and Y.-J. Lee,Appl. Phys. Lett., 101, 132901 (2012). 11. J.-S. Park, J. K. Jeong,Y.-G. Mo, H. D. Kim, and C.-J. Kim,Appl. Phys. Lett., 93,
033513 (2008).
12. K. Remashan, Y. S. Choi, S. J. Park, and J. H. Janga,J. Electrochem. Soc., 157(12), H1121 (2010).
13. C.-J. Ku, Z. Duan, P. I. Reyes, Y. Lu, Y. Xu, C.-L. Hsueh, and E. Garfunkel,Appl. Phys. Lett., 98, 123511 (2011).
14. A. Ohtomo, S. Takagi, K. Tamura, T. Makino, Y. Segawa, H. Koinuma, and M. Kawasaki,Jpn. J. Appl. Phys., 45, L694 (2006).
15. C. R. Kagan and P. Andry, Thin-Film Transistors, Dekker, New York, 2003. 16. D. Fang, C. Li, N. Wang, P. Li, and P. Yao,Cryst. Res. Technol., 48(5), 265 (2013). 17. D. M. Roessler and W. C. Walker,Phys. Rev., 159, 733 (1967).
18. B.-Y. Su a, S.-Y. Chu, Y.-D. Juang, and S.-Y. Liu, Journal of Alloys and Compounds, 580, (2013).
19. C. S. Chuang, T. C. Fung, B. G. Mullins, K. Nomura, T. Kamiya, H. P. D. Shieh, H. Hosono, and J. Kanicki,SID Int. Symp. Digest Tech. Papers, 39, 1215 (2008). 20. P. Bhattacharya, R. R. Das, and R. S. Katiyar,Thin Solid Films, 447, 564 (2004). 21. B. D. Ahn, J. H. Lim, M.-H. Cho, J.-S. Park, and K.-B. Chung,J. Phys. D: Appl.
Phys., 45, 415307 (2012).
22. T. T. Trinh, V. D. Nguyen, K. Ryu, K. Jang, W. Lee, S. Baek, J. Raja, and J. Yi, Semicond. Sci. Technol., 26, 085012 (2011).
23. Y. Yao and T. Xie, The Handbook of Physics and Chemistry (CRC Press, Boca Raton, FL, Vol. 111), (1985).
) unless CC License in place (see abstract). ecsdl.org/site/terms_use
address. Redistribution subject to ECS terms of use (see 140.113.38.11
