Solid-State Electronics xxx (2016) xxx–xxx

Contents lists available at ScienceDirect

Solid-State Electronics

journal homepage: www.elsevier .com/locate /sse

Improvement of the positive bias stability of a-IGZO TFTs by the HCNtreatment

http://dx.doi.org/10.1016/j.sse.2016.09.0100038-1101/� 2016 Published by Elsevier Ltd.

⇑ Corresponding author.E-mail address: duck@hanyang.ac.kr (D.-K. Choi).

Please cite this article in press as: Kim M-H et al. Improvement of the positive bias stability of a-IGZO TFTs by the HCN treatment. Solid State E(2016), http://dx.doi.org/10.1016/j.sse.2016.09.010

Myeong-Ho Kim a, Myung-Jea Choi a, Katsuya Kimura b, Hikaru Kobayashi b, Duck-Kyun Choi a,⇑aDepartment of Materials Science and Engineering, Hanyang University, Seoul 133-791, Republic of Koreab Institute of Scientific and Industrial Research, Osaka University, CREST, Japan Science and Technology Corporation, 8-1, Mihogaoka, Osaka 567-0047, Ibaraki, Japan

a r t i c l e i n f o

Article history:Available online xxxx

The review of this paper was arranged by Dr.Y. Kuk

Keywords:HCN treatmentAmorphous IGZOPBS instability

a b s t r a c t

In recent years, many researchers have attempted to improve the bias stability of amorphous indium gal-lium zinc oxide (a-IGZO) thin-film transistors (TFTs). In this study, the hydrogen cyanide (HCN) treatmentwas carried out to improve the positive bias stability of bottom-gate a-IGZO TFTs. The HCN treatment wasperformed using a 0.1 M HCN solution with a pH of 10 at room temperature. Before applying the positivebias stress, there were no differences in the major electrical properties, including the saturation mobility(lsat), threshold voltage (Vth), and subthreshold swing (S/S), between HCN-treated and non-HCN-treateddevices. However, after applying the positive bias stress, the HCN-treated device showed superior biasstability compared to the non-HCN-treated device. This difference is associated with the passivation ofthe defect states and the surface of the back-channel layer of the HCN-treated device by cyanide ions.

� 2016 Published by Elsevier Ltd.

1. Introduction

Recently, oxide thin-film transistors (TFTs) have attracted muchattention for applications in active-matrix liquid-crystal displays(AMLCDs) and active-matrix organic light-emitting diode displays(AMOLEDs) due to their various competitive advantages. In partic-ular, amorphous indium gallium zinc oxide (a-IGZO) TFTs have ahigh mobility (>20 cm2/V) compared to conventional amorphoussilicon TFTs (�0.5 cm2/V). As such, a-IGZO TFTs represent promis-ing switching devices for flat panel displays that possess ultra-high-definition (resolution: 2000 � 4000), large size (>50 in.), andhigh frame rate operation (>240 Hz) [1,2]. In addition, a-IGZO TFTshave a higher transparency over the entire visible light range, ascompared to silicon-based TFTs, because of their wide band gap.They also possess better uniformity than polysilicon TFTs becauseof their amorphous structure [3,4]. However, bias stability issues ofa-IGZO TFTs under gate bias stress still need to be resolved in flatpanel displays such as AMLCDs and AMOLEDs. According to manystudies investigating the stability of a-IGZO TFTs, the bias stabilityis very sensitive to oxygen vacancies in a-IGZO films as well as toambient oxygen molecules [5–7]. For these reasons, passivatingthe oxygen vacancy defect states and the a-IGZO back-channellayer is necessary in order to maintain good bias stability undergate bias stress.

Many studies have been made in attempts to improve the sta-bility of oxide TFTs by reducing the number of oxygen vacancydefect states in oxide semiconductors. The approaches used inthese studies typically utilize high pressure oxygen annealing orthe addition of carrier suppressors, such as gallium (Ga), hafnium(Hf), and zirconium (Zr), to the oxide semiconductor. Carrier sup-pressors can prevent the formation of oxygen vacancies becausethey have high ionic field strengths, which allow oxygen to bebound more tightly. However, these methods result in degradationof the electrical characteristics, including the mobility and thresh-old voltage (Vth), of oxide TFTs [8–10]. Additionally, there havebeen many studies investigating ways to protect the back-channel layer of a-IGZO TFTs; this is typically done by employinga passivation layer. However, passivation layers, such as SiO2 andSiNx, lead to Vth shifts in a-IGZO TFTs because hydrogen radicalsin the process chamber can diffuse into the a-IGZO bulk and actas donors during deposition [11,12].

Removing metal species by treating semiconductors withhydrogen cyanide (HCN) solutions has been found to be an effec-tive method to improve the electrical characteristics and bias sta-bility of semiconductor devices. To remove metal species on Si, forexample, the HCN treatment is performed using extremely dilute(3 ppm) HCN solutions at room temperature; this results in thecomplete removal (below �3 � 109 atoms/cm2) of metal contami-nants [13–15]. The removed metal species are present in the formof metal-cyano-complex ions, which are extremely stable in HCNsolutions, allowing them to avoid re-adsorption [15]. In addition,

lectron

2 M.-H. Kim et al. / Solid-State Electronics xxx (2016) xxx–xxx

cyanide ions (CN�) in HCN solutions selectively react with defectstates, such as Si and Cu dangling bonds, by forming Si-CN[13,16,17] and Cu-CN [18,19] bonds, respectively. This passivatesthe defect states. However, HCN treatments have yet to beattempted on oxide semiconductor-based devices.

In this study, we performed HCN treatments on a-IGZO TFTs toimprove their bias stability under a positive gate bias stress (PBS).To confirm the bias stability improvement of the HCN-treated a-IGZO TFTs, the standard device electrical properties were evalu-ated. Under bias conditions, we focused on the Vth shift of a-IGZO TFTs; this shift is closely related to the bias stability of a-IGZO TFTs. In order to understand the effect of the HCN treatmenton the bias stability of a-IGZO TFTs, the chemical states were alsoevaluated through X-ray photoelectron spectroscopy (XPS). Finally,to observe variations of the CN� ions in the a-IGZO bulk, we carriedout time-of-flight secondary ion mass spectroscopy (TOF-SIMS)depth profile analysis before and after the HCN treatment.

2. Experimental

Schematic diagrams of the device cross-sections for the HCNtreatment and the final structure of the device are shown inFig. 1(a) and (b), respectively. A 50-nm-thick Mo layer was depos-ited as the gate electrode on a glass substrate via DC sputtering atroom temperature. This was then patterned using a conventionalphotolithography process. Then, a 100-nm-thick SiNx gate dielec-tric was deposited on the substrate by plasma-enhanced chemicalvapor deposition at 250 �C. a-IGZO was selected as the active mate-rial and a 50-nm-thick a-IGZO filmwas deposited using RF sputter-ing with a power of 40W at room temperature. During deposition,we used an Ar:O2 ratio of 7:3, and the pressure was maintained at5 mTorr. After defining the active region, one group of TFT deviceswas fabricated as a reference by following the typical processingsequence (device A). The other group of samples was also treatedwith an HCN solution, as shown in Fig. 1(a) (device B). The HCNtreatment was performed using a 0.1 M HCN solution with a pHof 10 at room temperature for 10 min. Next, a 50-nm-thick Mo filmwas deposited via DC sputtering at room temperature. The sourceand drain Mo electrodes were patterned using the lift-off method.Finally, the devices were subjected to thermal annealing at 300 �Cfor 1 h in N2 ambient. The electrical characteristics were measuredwith current–voltage (I–V) measurements using an Agilent E5270Bparameter analyzer, and the chemical compositions of the oxidethin films were analyzed using XPS (Theta probe base system,Thermo Fisher Scientific Co.). The binding energy of the C1s peak,caused by contaminant hydrocarbon species on the specimen sur-faces, was taken as a reference for the binding energies (284.5 eV).To confirm variations of the various ionic species in the a-IGZO thin

Fig. 1. Schematic diagrams of a-IGZO TFT cross-sections: (a) HCN treatment using a 0.1 Mwith source/drain electrodes. The channel width (W) and length (L) were 50 lm and 80

Please cite this article in press as: Kim M-H et al. Improvement of the positiv(2016), http://dx.doi.org/10.1016/j.sse.2016.09.010

film, we conducted time-of-flight secondary ion mass spectroscopy(TOF-SIMS) depth profile analysis.

3. Results and discussion

Several important parameters of pristine devices A and B, suchas the saturation mobility (lsat), threshold voltage (Vth), and sub-threshold swing (S/S), were measured at room temperature(298 K). Device A exhibited a lsat of 16.9 cm2/Vs, Vth of 5.3 V, andSS value of 0.45 V/dec. Device B exhibited a lsat of 16.9 cm2/Vs,Vth of 5.7 V, and SS value of 0.42 V/dec. Both devices possessed verysimilar electrical characteristics. Such results indicate that the HCNtreatment did not affect the device parameters of pristine devices.Next, we investigated how the HCN treatment influenced the biasstability. Fig. 2(a) and (b) shows the transfer characteristics of non-HCN-treated (device A) and HCN-treated (device B) a-IGZO TFTs asa function of the PBS time. Devices A and B were stressed under thefollowing conditions: the Vg was set to +20 V at room temperatureand the maximum stress duration was 2000 s. Table 1 shows asummary of the changes in the lsat , Vth, and SS value accordingto the PBS time in devices A and B. During the bias stability test,neither device showed noticeable changes in the lsat or SS value.However, device A suffered from a positive Vth shift of �3.7 V afterapplying a PBS for 2000 s, whereas device B showed no significantchange in the Vth; a shift of only �0.2 V was observed under thesame PBS conditions. In other words, unlike other treatment meth-ods (e.g., high oxygen pressure annealing or the addition of carriersuppressors), the HCN treatment improves the bias stability whilealso retaining the desired electrical characteristics. Additionally,the lack of change in the SS value indicates that extra defect statesat shallow levels (i.e., the subgap density of states in the energyregion that are close to the conduction band) were not createdby the HCN treatment; shallow defect states are known to affectthe SS value [6]. In general, the positive Vth shift of oxide TFTs isattributed to either (i) electron trapping [8,9,20] or (ii) adsorptionof oxygen on the back-channel layer [5,12,21].

The electron trapping model can be further classified dependingon the electron trapping sites, such as the active/gate insulatorinterface or oxygen vacancy defects, in our devices. We can excludethe electron trapping model at the active/gate insulator interfacebecause the HCN treatment is carried out on the back-channellayer of a-IGZO TFTs; therefore, it is unlikely to affect the active/-gate insulator interface in our bottom-gate a-IGZO TFTs. However,the electron trapping model should be considered for the oxygenvacancy defect states because the bulk of the active layer can beinfluenced by the HCN treatment. According to Seo et al., the pres-ence of oxygen vacancy defect states leads to poor bias stabilityunder PBS and positive bias thermal stress (PBTS) conditions

HCN solution with a pH of 10 at room temperature for 10 min and (b) a-IGZO TFTlm, respectively.

e bias stability of a-IGZO TFTs by the HCN treatment. Solid State Electron

Fig. 2. Evolution of the transfer characteristics for the (a) non-HCN-treated a-IGZO TFT (device A) and (b) HCN-treated a-IGZO TFT (device B) as a function of the applied PBStime (W/L = 50/80 lm). A gate bias stress was applied under +20 V PBS conditions. The HCN-treated a-IGZO TFT shows good bias stability compared to the non-HCN-treateddevice under PBS conditions.

Table 1Comparison of the electrical properties of devices A and B as a function of the gate bias stress time.

Device Gate bias stress time (s) Mobility (cm2/Vs) Threshold voltage (V) Subthreshold swing (V/dec)

A 0 16.9 5.3 0.452000 16.7 9.0 0.43

B 0 16.9 5.7 0.422000 16.8 5.9 0.42

M.-H. Kim et al. / Solid-State Electronics xxx (2016) xxx–xxx 3

because the defect states act as electron trap sites in TFTs [22]. Forthis reason, the bias stability can be improved by passivating oxy-gen vacancy defect states in the active layer of oxide TFTs underPBS conditions.

Fig. 3 shows the O1s XPS spectra of the (a) non-HCN-treated a-IGZO thin film and (b) HCN-treated a-IGZO thin film. The O1s spec-tra were deconvoluted into three different peaks: (i) a strong peakat a lower energy, which represents a metal-oxide peak in theoxide lattice without an oxygen vacancy (529.8 eV); (ii) a metal-oxide peak in the oxide lattice with an oxygen vacancy(531.3 eV); and (iii) a metal-hydroxide peak (532.2 eV) [23]. Itcan be clearly seen that the peak related to the metal-oxide withan oxygen vacancy decreased after the HCN treatment. Presum-ably, this change is associated with CN� ion penetration. Penetrat-

Fig. 3. O1s XPS spectra of the (a) non-HCN-treated a-IGZO thin film and (b) HCN-trearesolved: (i) M-O state in the oxide lattice without an oxygen vacancy (529.8 eV), (ii) M-(532.2 eV).

Please cite this article in press as: Kim M-H et al. Improvement of the positiv(2016), http://dx.doi.org/10.1016/j.sse.2016.09.010

ing CN� ions selectively react with oxygen vacancies. As a result,the unsaturated metal-oxide bonds that originate from oxygenvacancies are present in the form of metal-cyano bonds, whichremain in the a-IGZO thin film [13,16,17]. This indicates that theunsaturated metal-oxide bonds near the surface of the a-IGZO thinfilm are passivated by CN� ions.

Additionally, we carried out time-of-flight secondary ion massspectroscopy (TOF-SIMS) depth profile analysis in order to observevariations in the various ionic species in a-IGZO thin films as afunction of the HCN treatment. For the TOF-SIMS measurements,a 25 keV Bi1+ ion beam was used as the analysis beam and a3 keV Cs+ ion beam was used as the sputtering beam. Fig. 4 showsthe TOF-SIMS depth profile analyses of Ga-O, In-O, Zn-O, SiO2, CN,and Si ions in non-HCN-treated and HCN-treated a-IGZO thin films.

ted a-IGZO thin film. In the O1s XPS spectra, three chemical binding states wereO state in the oxide lattice with an oxygen vacancy (531.3 eV), and (iii) M-OH state

e bias stability of a-IGZO TFTs by the HCN treatment. Solid State Electron

Fig. 4. TOF-SIMS depth profile analysis of the (a) non-HCN-treated a-IGZO thin film and (b) HCN-treated a-IGZO thin film. After the HCN treatment, the intensity of CN ionswas significantly increased in the a-IGZO region.

4 M.-H. Kim et al. / Solid-State Electronics xxx (2016) xxx–xxx

Although the Ga-O, In-O, Zn-O, SiO2, and Si ions in the a-IGZO thinfilms did not change noticeably, the intensity of the CN ions wassignificantly increased in the a-IGZO region after the HCN treat-ment. This suggests that the HCN treatment affects the bulk ofthe a-IGZO thin film. In other words, the unsaturated metal-oxide bonds on the surface and in the bulk of the a-IGZO thin filmwere passivated by CN� ions after the HCN treatment. It can beinferred that the presence of passivated, unsaturated metal-oxidebonds is one of the reasons for the improved bias stability of deviceB under PBS conditions.

The oxygen adsorption model can also be used to explain thepositive Vth shift under PBS conditions. Fig. 5(a) shows the generaladsorption process of oxygen molecules on the back-channel layerin the non-HCN-treated device (device A). In general, the interac-tion between the back-channel layer of a-IGZO TFTs and ambientoxygen is critical in determining the Vth instability because ambi-ent oxygen molecules can capture electrons from the surface of theback-channel layer by the following chemical reaction [5,12,21]:

O2 ðgasÞ þ e� ¼ 2O� ðsolidÞ: ð1ÞAccording to Le Chatelier’s principle, high ambient oxygen pres-

sure at a fixed temperature causes an increase of the O�solid. Simi-

larly, when the PBS is applied in a gate electrode, the interactionbetween the back-channel layer and oxygen molecules can becomemore activated, which leads to the poor bias stability of a-IGZOTFTs [5,12,21]. For this reason, many researchers have formed pas-sivation layers, such as SiO2 and SiNx, to protect the back-channellayer of a-IGZO TFTs and prevent oxygen adsorption [12,24]. How-ever, the HCN-treated device (device B) shows good bias stabilityunder PBS conditions (Fig. 2(b)). This can be explained by the factthat the absorbed CN� ions on the back-channel layer can passivate

Fig. 5. Process showing the adsorption of O2 molecules on the back-channel layer of (a) ddevice B, the back-channel layer of the device is passivated by metal-cyano bonds. As a

Please cite this article in press as: Kim M-H et al. Improvement of the positiv(2016), http://dx.doi.org/10.1016/j.sse.2016.09.010

the back-channel layer and block the adsorption of oxygen mole-cules that induce the Vth shift (Fig. 5(b)). Therefore, applying theHCN treatment on the back-channel layer of a-IGZO TFTs canenhance the bias stability of a-IGZO TFTs under PBS conditionswithout necessitating the need for a passivation layer. This alsodemonstrates that the HCN treatment plays a significant role inimproving the bias stability of TFTs under PBS conditions.

4. Conclusion

In this study, we investigated how the HCN treatment affectedthe instability of a-IGZO TFTs under PBS conditions. The HCN treat-ment caused no noticeable change in the SS value or saturationmobility. However, the non-HCN-treated device suffered from apositive Vth shift of �3.7 V after applying the PBS for 2000 s. Alter-natively, the HCN-treated device exhibited superior stability (4Vth

of �0.2 V) under the same conditions. In other words, the HCNtreatment can improve the bias stability of a-IGZO TFTs withoutdegrading their electrical characteristics. XPS and TOF-SIMS depthprofile analysis suggest that unsaturated metal-oxide bonds in thea-IGZO thin film are passivated by forming metal-cyano bonds.Such passivation leads to a significant reduction in the Vth shiftin HCN-treated a-IGZO TFTs under PBS conditions. Moreover, it isbelieved that the absorbed CN� ions can prevent the adsorptionof oxygen molecules at the surface, which enables the productionof passivation layer-free a-IGZO TFTs.

Acknowledgements

This research was supported by the Basic Science Research Pro-gram through the National Research Foundation of Korea (NRF)

evice A (non-HCN-treated) and (b) device B (HCN-treated) under PBS conditions. Inresult, the bias stability of device B is improved under PBS conditions.

e bias stability of a-IGZO TFTs by the HCN treatment. Solid State Electron

M.-H. Kim et al. / Solid-State Electronics xxx (2016) xxx–xxx 5

funded by the Ministry of Science, ICT & Future Planning(2013R1A1A2064715).

References

[1] Park JS. J Electroceram 2012;28:74.[2] Fortunato E, Barquinha P, Martins R. Adv Mater 2012;24:2945.[3] Kamiya T, Nomura K, Hosono H. Sci Technol Adv Mater 2010;11:044305.[4] Park JS, Maeng WJ, Kim HS, Park JS. Thin Solid Films 2012;520:1679.[5] Chen YC, Chang TC, Li HW, Chen SC, Lu J, Chung WF, et al. Appl Phys Lett

2010;96:26104.[6] Nomura K, Kamiya T, Hirano M, Hosono H. Appl Phys Lett 2009;95:013502.[7] Lee KH, Ok KC, Kim H, Park JS. Ceram Int 2014;40:3215.[8] Chiang HQ, McFarlane BR, Hong D, Presley RE, Wager JF. J Non-Cryst Solids

2008;354:2826.[9] Chen WT, Lo SY, Kao SC, Zan HW, Tsai CC, Lin JH, et al. IEEE Electr Dev L

2011;32:1552.[10] Kwon JY, Jeong JK. Semicond Sci Technol 2015;30:024002.[11] Liu SE, Yu MJ, Lin CY, Ho GT, Cheng CC, Lai CM, et al. IEEE Electr Dev L

2011;32:161.

Please cite this article in press as: Kim M-H et al. Improvement of the positiv(2016), http://dx.doi.org/10.1016/j.sse.2016.09.010

[12] Jeong JK, Yang HW, Jeong JH, Mo YG, Kim HD. Appl Phys Lett 2008;93:123508.[13] Takahashi M, Liu YL, Fujiwara N, Iwasa H, Kobayashi H. Solid State Commun

2006;137:263.[14] Fujiwara N, Liu YL, Takahashi M, Kobayashi H. J Electrochem Soc 2006;1534:

G394.[15] Fujiwara Y, Kobayashi Y, Kita K, Kakehashi R, Noro M, Katayama J, et al. J

Electrochem Soc 2008;155:D377.[16] Kobayashi H, Asano A, Asada S, Kubota T, Yamashita Y, Yoneda K, et al. J Appl

Phys 1998;83:2098.[17] Maida O, Asano A, Takahashi M, Iwasa H, Kobayashi H. Surf Sci 2003;542:244.[18] Okamoto Y, Ishizuka S, Kato S, Sakurai T, Fujiwara N, Kobayashi H, et al. Appl

Phys Lett 2003;82:1060.[19] Ishizuka S, Kato S, Okamoto Y, Sakurai T, Akimoto K, Fujiwara N, et al. Appl Surf

Sci 2003;216:94.[20] Hoshino K, Hong D, Chiang HQ, Wager JF. IEEE T Electron Dev 2009;56:1365.[21] Jeong JK. Semicond Sci Technol 2011;26:034008.[22] Seo JS, Jeon JH, Hwang YH, Park H, Ryu M, Park SHK, et al. Sci Reports

2013;3:2085.[23] Lee KW, Kim KM, Heo KY, Park SK, Lee SK, Kim HJ. Curr Appl Phys 2011;11:280.[24] Nomura K, Kamiya T, Hosono H. Thin Solid Films 2012;520:3778.

e bias stability of a-IGZO TFTs by the HCN treatment. Solid State Electron