Charge Storage Characteristics of Mo Nanocrystal Memory Influenced by Ammonia Plasma Treatment

(1)

Charge Storage Characteristics of Mo Nanocrystal Memory

Influenced by Ammonia Plasma Treatment

Chao-Cheng Lin,aTing-Chang Chang,b,

*

,zChun-Hao Tu,aWei-Ren Chen,a

Chih-Wei Hu,a Simon M. Sze,a Tseung-Yuen Tseng,a Sheng-Chi Chen,c and

Jian-Yang Linc a

Institute of Electronics, National Chiao Tung University, Hsin-Chu 300, Taiwan b

Department of Physics, Center for Nanoscience and Nanotechnology, National Sun Yat-Sen University, Kaohsiung 804, Taiwan

c

Graduate School of Opto-Electronic Engineering, National Yunlin University of Science and Technology, Yunlin 64002, Taiwan

Mo nanocrystal memory was fabricated through annealing the oxygen-incorporated Mo and Si layers. We then investigated the influence an of ammonia plasma treatment on the nonvolatile memory characteristics of a charge storage layer composed of Mo nanocrystal memory embedded in SiOx. Transmission electron microscopy revealed the nanostructure of the charge storage layer, and X-ray photoelectron spectra analyses revealed that nitrogen was incorporated into the charge storage layer. Electric analyses indicated that the memory window reduced, and both retention and endurance improved after the treatment. The reduction in the memory window was attributed to the decrease in charge trapping centers in the surrounding oxide after the treatment. The improvement of retention was interpreted in terms of the nitrogen passivation of traps in the oxide around the Mo nanocrystals. The robust endurance characteristic was attributed to the improvement of the quality of the surrounding oxide by nitrogen passivation.

Manuscript submitted February 27, 2009; revised manuscript received May 8, 2009. Published July 10, 2009.

Nonvolatile memory composed of a floating structure plays an important role in portable devices such as cell phones, notebook computers, and digital cameras. However, conventional floating structure confronts a bottleneck on the reduction in tunnel oxide in the memory structure because carriers stored in an electrical con-tinuously conducting polycrystalline silicon floating gate totally lose if one defect chain exists in the tunnel oxide.1,2The nanocrystals embedded in dielectric as charge storage centers in the floating gate structure has attracted much attention for next-generation nonvola-tile memory because carriers stored in electrically isolated nanocrys-tals can suppress carrier loss.2Several methods have been investi-gated to fabricate nanocrystals such as ion implantation,3,4 sputtering,5,6and oxidation.7,8It is expected that those processes can induce defects or traps in the dielectric around nanocrystals during the fabrication process. The induced deficiency in the surrounding dielectric can lead to stored charges leaking out of the nanocrystals through trap-assisted tunneling.9Therefore, the quality of the sur-rounding oxide, which electrically isolates the nanocrystals, is an important issue in research on nanocrystal memory. Two possible ways to gain good reliability are improving the surrounding oxide of the nanocrystal by using a hydrogen treatment with a high pressure at high temperature共 ⬎ 900°C兲 and nitrogen annealing at high tem-perature共 ⬎ 1000°C兲. However, the Si–H bond is weak and easily to be broken during endurance tests. For the nitrogen annealing, a high temperature process共 ⬎ 1000°C兲 is required to dissociate the nitrogen to strengthen the surrounding oxide layer.10,11 Therefore, the process is not appropriate for application in semiconductor in-dustry and for nanocrystal memory in terms of thermal budget.

In this study, Mo nanocrystal memory was fabricated through annealing the oxygen-incorporated Mo silicide layer.12We investi-gated the effect of ammonia共NH₃兲 plasma treatment on memory characteristics of the Mo nanocrystal embedded in SiOx as the

charge storage layer. The plasma-enhanced process is widely used in the semiconductor industry for its benefit with regard to small ther-mal budget, which is important for integrated circuit technology because the thermal budget may lead to the redistribution of source/ drain dopant or density and size of the nanocrystal. Furthermore, the NH3plasma treatment technique has been investigated to improve

the quality of gate dielectric.13,14Our experimental results show that the memory characteristics including memory window, retention, and endurance were influenced by the incorporation of nitrogen into the surrounding oxide.

Experimental

The experimental process flow and memory structure is shown in Fig. 1. The memory cells were fabricated on 6 in. p-type Si sub-strate. After the substrate was cleaned by a standard RCA process, a 5 nm thick dry oxide layer was grown at 950°C on the substrate in a horizontal furnace. An 8 nm thick Mo silicate layer was then de-posited on the oxide layer using cosputtering Mo and Si in Ar 共24 sccm兲/O 共2 sccm兲 ambience. 30 nm thick Si oxide was depos-ited on the Mo silicate layer as the blocking oxide by plasma-enhanced chemical vapor deposition共PECVD兲 at 300°C. The ther-mal annealing process was performed in N2ambience at 900°C for

60 s to form Mo nanocrystals embedded in SiO_x. Cells were subse-quently treated with NH3plasma in the PECVD chamber for 30 min

with an NH3gas flow rate of 20 sccm and a chamber pressure of

67 Torr at a power of 50 W. For the electrical characteristic mea-surement, 500 nm thick Al was thermally evaporated with a shadow mask on the blocking oxide to form the metal-oxide-semiconductor 共MOS兲 structure. The nanostructure of Mo nanocrystals was

ana-*Electrochemical Society Active Member.

z

E-mail: [email protected] Figure 1. Process flow and memory structure of this work. Journal of The Electrochemical Society, 156共9兲 H716-H719 共2009兲

0013-4651/2009/156共9兲/H716/4/$25.00 © The Electrochemical Society H716

) 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

(2)

lyzed through a transmission electron microscope. The chemical bonding and composition of the charge storage layer were analyzed with X-ray photoemission spectroscopy共XPS兲. Electrical character-istics were measured using a Keithley 4200 and HP 4284 precision LCR meter with a frequency of 1 MHz.

Results and Discussion

Figure 2a and b shows the transmission electron microscopy 共TEM兲 for the charge storage layer with and without the plasma treatment. The average size and density of both samples are esti-mated at about 4–5 nm and 1012_cm−2_{. A density as high as}

1012_cm−2_{is a basic requirement for nanocrystal memory to prevent}

the variation in memory characteristics between cells during reduc-tion in the nonvolatile memory structure. In this case, however, the high density results in the thickness of the surrounding oxide, which electrically isolates nanocrystals, being only 4–5 nm, as shown in Fig. 2. Therefore, if traps exist in the oxide, this can result in a serious tunnel process of the stored carriers between nanocrystals 共trap-assisted tunnel mode兲.9

In our previous study, the oxide surrounding the Mo nanocrystal was deficient共oxygen deficiency兲.12Figure3shows XPS Mo 3p and N 1s spectra of the charge storage layer共a兲 without and 共b兲 with the plasma treatment, both performed by using a monochromatic Al K␣ 共1486.6 eV兲 X-ray. An additional XPS peak 共397.8 eV兲 appears in Mo 3p and N 1s spectra of the charge storage layer after the plasma treatment due to the formation of O–Si–N bonds. This suggests that the incomplete bonds in SiOxbonded with nitrogen after the plasma

treatment, as shown in Fig.3.15

Figure 4a and b shows the C-V curves of the MOS structure embedded with the Mo nanocrystal for the sample with and without the plasma treatment, respectively. At the smaller sweeping voltage of 2 V, there is a negligible memory window in Fig. 4a and b corresponding to the quasi-neutral state共i.e., no charge is stored in the charge storage layer under this sweeping range兲. At the larger sweeping voltages共−11 to 9 V and vice versa兲, there are counter-clockwise memory hystereses, as evident in Fig.4aandb. The

coun-terclockwise hystereses are due to carrier transport through tunnel oxide between the charge storage layer and the Si substrate. The memory windows were reduced 共0.6 V reduction兲 for the sample with the plasma treatment. The reduction in the memory window after the treatment was related to the nitrogen passivation in the charge storage layer. It has been suggested that the traps in the oxide around nanocrystals can capture carriers and contribute to the memory window.12According to XPS results, nitrogen was incorpo-rated into the oxide around the Mo nanocrystals after the treatment. The incorporated nitrogen can passivate the traps in the oxide, which reduce the charge storage centers in the surrounding oxide, leading to the smaller memory window.

Figure5is a comparison of retention behavior for the samples with and without the plasma treatment, respectively. Retention was measured at a stress voltage of 10 V on the Al gate electrode for 5 s. The memory window was obtained by comparing the C-V curves between the charged state and the quasi-neutral state. When carriers are stored in the nanocrystals, the stored carriers raise the potential energy of the nanocrystals and increase escape probability of the stored carriers. Furthermore, carriers trapped in the shallow traps are unstable and can easily leak back to the silicon substrate. Figure5 shows that for retention time before 102_{s, the charge loss rate is}

significant, becoming stable over a longer retention time. This result is inconsistent with partial carriers trapping in the shallow trap states of the SiOxmatrix around the nanocrystals. Because the decline rate

of the flatband voltage after 102_{s retention time is stable, the}

reduc-tion rate of the flatband voltage has an exponential dependence on retention time according to the one-dimensional direct tunneling model.16Therefore we can extrapolate the retention characteristics to 10 years with a slope of the stable range

Figure 2. Plane-view TEM image of the sample共a兲 without and 共b兲 with the plasma treatment.

Figure 3. Mo 3p and N 1s core-level spectra of the charge storage layer composed of Mo nanocrystals embedded in SiOx共a兲 without and 共b兲 with the plasma treatment.

Figure 4. C-V curves of the MOS structure without and with the plasma treatment.

Figure 5. The retention behavior of MOS structures with and without the plasma treatment.

H717

Journal of The Electrochemical Society, 156共9兲 H716-H719 共2009兲 H717

(3)

m = VFB2− VFB1 log t₂− log t₁

where VFB1 and VFB2 are the flatband voltage shifts at retention

times of 102_s _共t

1兲 and 104s 共t2兲, respectively. A 1.4 V memory

window共charge remaining at 68%兲 can be obtained after 10 years by analyzing the extrapolation value of retention data. In contrast, the charge loss rate of the sample after the treatment is slower than that without the treatment, with the memory window remaining at 86%. The superior retention of the sample with the treatment can be explained by the nitrogen passivation of the traps in the oxide around Mo nanocrystals, schematically expressed in Fig.6. When carriers are stored in nanocrystals after programming, if there is a defect chain in the tunnel oxide, the nanocrystal that aligns with the defect chain is discharged immediately. As shown in Fig.6afor the sample without the plasma treatment, the carriers stored in the neighbor nanocrystals can escape to the discharged nanocrystal by assistance of traps共trap-assisted tunneling兲 in the surrounding oxide, which results in a faster loss rate of the stored carriers. For the sample with the plasma treatment, the traps in the surrounding oxide were reduced by nitrogen passivation. The trap-assisted lateral mi-gration of the stored carriers was suppressed, which improves the retention of the memory structure, as shown in Fig.6b.

Figure7ais the comparison of the endurance characteristics of the samples with and without the plasma treatment under pulse con-ditions of V_G= ⫾ 15 V for 1 ms, respectively. For the sample without the plasma treatment, the⌬VFB共the difference between

pro-gramming and erase states兲 was reduced significantly, remaining 51% after 106 _{program/erase cycles. However, the plasma-treated}

sample exhibits a robust endurance characteristic 共⌬V_FB of 89% after 106_{program/erase cycles}_{兲. Previously the ⌬V}

FBreduction

dur-ing such an endurance test was due to the degradation of the gate oxide.13Carriers tunneling from the substrate into the charge storage layer during the endurance test can release energy to destroy the

surrounding dielectric共Fig.7b兲, which results in the reduction in the ⌬VFB. For the sample without the plasma treatment, there are

sev-eral weak or dangling bonds in the surrounding oxide induced dur-ing the formation of nanocrystals. The bonds can easily break durdur-ing the programming/erase cycles, leading to a faster degradation rate of the surrounding oxide. After the plasma treatment, because the traps were reduced, and the surrounding oxide was strengthened by the nitrogen incorporation, the generation rate of traps reduced, result-ing in the better endurance characteristic.

TableIcompares the memory characteristics of our experimental result with those of Si nanocrystal memory in recent research. The NH₃plasma-treated Mo nanocrystal memory has a large memory window and good retention, and is suitable for nonvolatile memory application.17-19

Conclusion

The nonvolatile memory characteristics of the Mo nanocrystals were influenced by ammonia plasma treatment. The C-V hysteresis reduced from 3.6 to 3.0 V due to a reduction in trapping centers in the surrounding oxide. The retention characteristic improved from 68 to 86% remaining charges after the plasma treatment because the carriers’ lateral migration between nanocrystals was suppressed through the nitrogen passivation of the traps in the surrounding ox-ide. The incorporation of nitrogen into the charge storage layer through the NH3 plasma treatment can strengthen the endurance

characteristic of Mo nanocrystal memory. These results suggest the promising usage of NH3plasma treatment to passivate the

surround-ing oxide of nanocrystals in the next generation of floatsurround-ing gate flash memory devices.

Acknowledgments

This work was performed at National Science Council Core Fa-cilities Laboratory for Nano-Science and Nano-Technology in Kaohsiung-Pingtung area and was supported by the National Sci-ence Council of the Republic of China under contract no. NSC-97-3114-M-110-001 and no. NSC 97-2112-M-110-009-MY3.

National Sun Yat-Sen University assisted in meeting the publication costs of this article.

References

1. D. Kahng and S. M. Sze, Bell Syst. Tech. J., 46, 1288共1967兲.

2. S. Tiwari, F. Rana, K. Chan, H. Hanafi, W. Chan, and D. Buchanan, Tech. Dig. -Int. Electron Devices Meet., 1995, 521.

3. Y. C. King, T.-J. King, and C. Hu, IEEE Trans. Electron Devices, 48, 696共2001兲. 4. C. Y. Ng, T. P. Chen, L. Ding, M. Yang, J. I. Wong, P. Zhao, X. H. Yang, K. Y. Liu,

M. S. Tse, A. D. Trigg, et al., IEEE Trans. Electron Devices, 53, 730共2006兲. 5. V. Ho, L. W. Teo, W. K. Choi, W. K. Chim, M. S. Tay, D. A. Antoniadis, E. A.

Fitzgerald, A. Y. Du, C. H. Tung, R. Liu, et al., Appl. Phys. Lett., 83, 3558共2003兲. 6. K. I. Han, Y. M. Park, S. Kim, S. H. Choi, K. J. Kim, I. H. Park, and B. G. Park,

IEEE Trans. Electron Devices, 54, 359共2007兲.

7. L. W. Teo, W. K. Choi, W. K. Chim, V. Ho, C. M. Moey, M. S. Tay, C. L. Heng, Y. Lei, D. A. Antoniadis, and E. A. Fitzgerald, Appl. Phys. Lett., 81, 3639共2002兲. 8. J. K. Kim, H. J. Cheong, Y. Kim, J. Yi, H. J. Bark, S. H. Bang, and J. H. Cho, Appl.

Phys. Lett., 82, 2527共2003兲.

9. M. Houssa, M. Tuominen, M. Naili, V. Afanas’ev, A. Stesmans, S. Haukka, and M. M. Heyns, J. Appl. Phys., 87, 8615共2000兲.

10. G. Taraschi, S. Saini, W. W. Fan, and L. C. Kimerling, J. Appl. Phys., 93, 9988 共2003兲.

11. C. H. Tu, T. C. Chang, P. T. Liu, H. C. Liu, S. M. Sze, and C. Y. Chang, Appl. Phys. Lett., 89, 162105共2006兲.

Figure 6.共Color online兲 Schematic explanation of the retention behavior for the sample共a兲 before and 共b兲 after the plasma treatment.

Figure 7. 共a兲 Endurance characteristic of the MOS structures without and with the plasma treatment and共b兲 band diagram of carrier transport during the electron tunnel.

Table I. Comparison of the memory characteristics of our experi-ment result with those of Si nanocrystal memory in recent research. Reference Memory window 共V兲 Retention共%兲 This study 3 86 Laha et al.16 1.5 50 Kuan et al.17 1 ⬍50 Porti et al.18 1.4 30

H718 Journal of The Electrochemical Society, 156共9兲 H716-H719 共2009兲

H718

(4)

12. C. C. Lin, T. C. Chang, C. H. Tu, W. R. Chen, C. W. Hu, S. M. Sze, T. Y. Tseng, S. C. Chen, and J. Y. Lin, Appl. Phys. Lett., 93, 222101共2008兲.

13. C. Busseret, A. Souifi, T. Baron, S. Monfray, N. Buffet, E. Gautier, and M. N. Semeria, Mater. Sci. Eng., C, 19, 237共2002兲.

14. P. Pavan, R. Bez, P. Olivo, and E. Zanoni, Proc. IEEE, 85, 1248共1997兲. 15. Z. H. Lu, S. P. Tay, R. Cao, and P. Pianetta, Appl. Phys. Lett., 67, 2836共1995兲. 16. W. Guan, S. Long, M. Liu, Q. Liu, Y. Hu, Z. Li, and R. Jia, Solid-State Electron.,

51, 806共2007兲.

17. A. Laha, D. Kühne, E. Bugiel, A. Fissel, and H. J. Osten, Semicond. Sci. Technol., 23, 085015共2008兲.

18. Y. Kuang, Y. Li, D. Wu, Z. Yu, R. Tang, and R. Huang, in IEEE Nanoelectronics Conference, IEEE, p. 24共2008兲.

19. M. Porti, M. Avidano, M. Nafría, X. Aymerich, J. Carreras, O. Jambois, and B. Garrido, J. Appl. Phys., 101, 064509共2007兲.

H719

Journal of The Electrochemical Society, 156共9兲 H716-H719 共2009兲 H719