Charging Effect of Aluminum Nitride Thin Films Containing Al ... - NTU

Copyright © 2010 American Scientific Publishers All rights reserved Printed in the United States of America

Journal of Nanoscience and Nanotechnology Vol. 10, 599–603, 2010

Charging Effect of Aluminum Nitride Thin Films Containing Al Nanocrystals Y. Liu1 ∗ , T. P. Chen2 , L. Ding2 , J. I. Wong2 , M. Yang2 , Z. Liu2 , Y. B. Li3 , and S. Zhang3 1

State Key Laboratory of Electronic Thin Films and Integrated Devices, University of Electronic Science and Technology of China, Chengdu, Sichuan, 610054, P. R. China 2 School of Electrical and Electronic Engineering, Nanyang Technological University, 639798, Singapore 3 School of Mechanical and Aerospace Engineering, Nanyang Technological University, 639798, Singapore

In this work, the Al-rich AlN thin film is deposited on Si substrate by radio frequency (RF) sputtering to form a metal-insulator-semiconductor (MIS) structure. Al nanocrystals (nc-Al) are formed and embedded in the AlN thin film. Charge trapping/detrapping in the nc-Al leads to a shift in the flatDelivered by Ingenta to: band voltage (VFB ) of the MIS structure. The charge storage ability of the AlN thin films containing Nanyang Technological University Al nanocrystals provides the possibility of memory applications. On the other hand, charge trapping IP : 155.69.4.4 in nc-Al reduces the current conduction because of the breaking Mon, 01 Mar 2010 06:24:01 of some tunneling paths due to Coulomb blockade effect and the current conduction evolves with a trend towards one-dimensional transport.

Keywords: Al Nanocrystal, Charging Effect.

Aluminum nitride (AlN) thin film has attracted much research attention because it is a promising material for applications in surface acoustic wave (SAW) devices and light-emission devices.1–3 In addition, AlN also has an application potential in the area of solar cell synthesis.4 Many techniques can be used to synthesize AlN thin films such as reactive evaporation, magnetron sputtering, pulsed laser deposition, and hybrid techniques such as plasma- or ion-assisted depositions.5–7 With its high thermal conductivity,8 reasonable thermal match to semiconductors (such as Si, GaAs and GaN) and small lattice mismatch8 9 and wide band gap (6.2 eV),10 AlN thin film could be used as a gate dielectric for field-effect transistors. In this work, the Al-rich AlN thin film is deposited on Si substrate by radio frequency (RF) sputtering to form a metal-insulator-semiconductor (MIS) structure. Al nanocrystals (nc-Al) are formed and embedded in the AlN thin film. Charge trapping/detrapping in the nc-Al leads to a shift in the flat-band voltage (VFB ) of the MIS structure. The situation is similar to the memory effect observed for SiO2 thin films containing Si nanocrystals (nc-Si), which has already been used for nonvolatile memory applications.11 12 The charge storage ability of the AlN thin films containing Al nanocrystals provides the possibility of memory ∗

Author to whom correspondence should be addressed.

J. Nanosci. Nanotechnol. 2010, Vol. 10, No. 1

applications with low cost. On the other hand, the charging effect on the current conduction in AlN films has been studied also. It is found that charging in the nc-Al leads to a reduction in the current conduction, and the system evolves with a trend towards one-dimensional transport, which has been defined in Ref. [13]. The phenomena can be explained by a model of charge transport in nc-Al arrays.

2. EXPERIMENTAL DETAILS A series of Al-rich AlN films were deposited on n or p-type, 100 oriented Si wafers. Depositions were carried out by RF magnetron sputtering of a pure Al target in a gas mixture of argon and nitrogen. The purities of the argon and nitrogen gas are more than 99.99 in percentage. The flow rates of Ar and N2 were varied to obtain different Al to N ratios. A 200 nm aluminum layer was then deposited to form the gate electrode. The wafer backside was coated with a layer of aluminum with a thickness of about 500 nm after removing the initial oxide. Finally, an alloying process was conducted at 425 C in N2 ambient to form ohmic contacts. X-ray photoemission spectroscopy (XPS) analysis was performed using a Kratos AXIS spectrometer with monochromatic Al K (1486.71 eV) X-ray radiation. After the correction with the appropriate relative sensitivity factors, the Al and N atomic ratios were determined from the calculation of the ratio of IAl /IN where IAl is the peak area of the Al 2p peak

1533-4880/2010/10/599/005

doi:10.1166/jnn.2010.1590

599

RESEARCH ARTICLE

1. INTRODUCTION

Charging Effect of AlN Thin Films Containing Al Nanocrystals

Liu et al.

and IN is the peak area of the N 1s peak. As the ratio is larger than 1, ranging from 1.2 to 3.5, the as-fabricated films are rich of aluminum. Depth XPS analysis was also carried out. There is no obvious difference in the Al:N ratio at various depths. Transmission Electron Microscopy (TEM) image shows that excess Al atoms form nano-size crystals in the AlN matrix as illustrated in the inset in Figure 1(a), which is also presented in Refs. [14, 15]. In air

environment, the experiment of charging/discharging the Al nanocrystals and the current–voltage (I–V ) measurements were carried out with a Keithley 4200 semiconductor characterization system, and capacitance–voltage (C–V ) measurements were carried out with a HP4284A LCR meter at the frequency of 1 MHz.

3. RESULTS AND DISCUSSION

Figure 1(a) shows the C–V characteristics obtained by sweeping the voltage from −8 V to 2 V first and then from (a) 2 V back to −8 V. A flat-band voltage shift of ∼11 V 80 is observed from Figure 1(a). The large flat-band voltage shift indicates significant charge trapping in the Al-rich 60 AlN film. The charge trapping is concerned with the Al nanocrystals embedded in the AlN thin films, and the situ40 ation here is very similar to that of Si-rich SiO2 (i.e., SiOx , x < 2) thin films where charge trapping in the Si nanonc-Al 20 crystals to: leads to a flat-band voltage shift.11 12 Both posiDelivered by Ingenta tive and negative charge trapping in the Al nanocrystals Nanyang Technological University 0 are possible, depending on the gate voltage applied. As IP : 155.69.4.4 shown in Figure Mon, 01 Mar 2010 06:24:01 1(b), the application of 30 V to the gate for 0.1 s shifts the C–V characteristic to the positive with –8 –6 –4 –2 0 2 a positive flat-band voltage shift of ∼18 V (i.e., VFB ≈ Voltage (V) +18 V) indicating a negative charge trapping in the Al 120 (b) nanocrystals. In contrast, as shown in Figure 1(c), the 100 application of −23 V for 0.1 s leads to a negative flatband voltage shift of about 1.3 V (i.e., VFB ≈ −13 V), 80 indicating positive charge trapping in the Al nanocrystals. These results show that the charge trapping depends on 60 the voltage applied. The capability of charge storage in the 40 Al nanocrystals and the large effect on the flat-band voltage shift provide the possibility of the application of the 20 Al-rich AlN thin films in memory devices. Initial After 30 V for 0.1 s 0 Figure 2(a) shows flat-band voltage shift as a function of writing time at various voltages. The flat-band voltage shift is increased with the writing time. The application of –8 –6 –4 –2 0 2 a higher voltage also leads to a higher flat-band voltage Voltage (V) shift. This is because under a positive gate voltage, elec120 trons are injected from the substrate and then trapped in (c) 100 nc-Al. A longer writing time or a higher voltage causes more electron trapping in nc-Al, and thus a larger flat-band 80 voltage shift is observed. For a charging duration that is long enough, for example 100 seconds at 20 V, the flat60 band voltage shift tends to saturate, because most of nc-Al 40 are charged up in this situation. This is not shown in the figure, because a writing duration of 100 s is too long for 20 memory applications. On the other hand, there is no shift Initial After –23 V for 0.1 s 0 for t ≤ 10−4 s, which means that only a voltage application longer than 10−4 s can write the device. Figure 2(b) shows the retention study of the MIS device. Initially, an appli–8 –6 –4 –2 0 2 Voltage (V) cation of +25 V for 0.1 s was used to write the device. As can be seen in the figure, the flat-band voltage shift Fig. 1. (a) Double-sweep C–V characteristics; (b) shift in C–V charis reduced from ∼16 V to ∼10 V after 24 hours, and it acteristic after the application of +30 V for 0.1 s; and (c) shift in C–V is predicted that the flat-band voltage shift will be further characteristic after the application of −23 V for 0.1 s. The inset in (a) shows nc-Al embedded in the AlN matrix. reduced to ∼05 V after one year. Capacitance (pF)

Capacitance (pF)

RESEARCH ARTICLE

Capacitance (pF)

100

600

J. Nanosci. Nanotechnol. 10, 599–603, 2010

Liu et al.

Charging Effect of AlN Thin Films Containing Al Nanocrystals 10–6

2.5 2.0

10–8

1.5 1.0

10–9 10–10

0.5

10–11

0.0

10–12

10–7

10–6

Virgin After +15 V for 100 s Power law fitting

10–7

25 V 23 V 20 V 18 V

Current (A)

Flat-band voltage shift (V)

(a) 3.0

10–5

10–4

10–3

10–2

10–1

100

10–13

101

1

2

Writing time (s) (b) 1.8

5

1.4

One year

Delivered by Nanyang Technological University IP : 155.69.4.4 The value (=1.93) for the virgin situation exceeds Mon, 01 Mar 2010 the 06:24:01 two-dimensional transport limitation (i.e., = 5/3).

1.2 1.0 0.8 0.6 0.4 0.2 0.0 100

101

102

103

104

105

106

107

108

Fig. 2. (a) VFB as a function of writing time at various charging voltages; and (b) flat-band voltage shift as a function of waiting time. For the experiment shown in (b), the device was charged at +25 V for 0.1 s.

On the other hand, charge trapping on the current conduction is also studied. Figure 3 shows the power law fitting on the I–V characteristics for the virgin situation (i.e., no charge trapping in nc-Al) and the one after +15 V for 100 s. The power-law behavior could be explained by a model similar to the one of collective charge transport in arrays of normal-metal quantum dots.13 The current through an array of metallic quantum dots separated by tunnel barriers is shown to follow the power-law relationship13 16 I = I0 V − Vth

1.8

(1)

where is the scaling exponent, and Vth is the threshold voltage. The scaling exponent = 1 and 5/3 for the one- and two-dimensional arrays of quantum dots, respectively.13 Equation (1) can be used to fit our experimental result of the system of nc-Al embedded in AlN matrix, and the fitting can yield the values of the factors including , Vth and I0 of Eq. (1). Figure 3 shows such a fitting to the I–V characteristics before and after the application of +15 V for 100 s. is found to be larger than 5/3 for the virgin situation, but it reduces to the range of 1–5/3 after charging by +15 V for 100 s. J. Nanosci. Nanotechnol. 10, 599–603, 2010

2.0

1.6

1.4

1.2

1.0 0

20

40

60

80

100

120

Charging time (s) Fig. 4. as a function of charging time.

601

RESEARCH ARTICLE

Waiting time (s)

This indicates that the current conduction of the system is beyond the regime of two-dimensional transport. We believe some three-dimensional transport exists in the system in addition to the one- and two-dimensional transport for the virgin situation.13 15 After charging by +15 V for 100 s, the is reduced to 1.13 falling in the range of 1–5/3, indicating that the current conduction of the system becomes a quasi two-dimensional transport (i.e., in between one- and two-dimensional transport). The situation could be like that there are two-dimensional tunneling paths in addition to one-dimensional tunneling paths. For both cases, the threshold voltage is found to be approximately zero at room temperature. As can be observed from Figure 4, the charge trapping leads to a reduction in . decreases from 1.93 for charging time t = 0 (i.e., the virgin case) to 1.13

ζ

Flat-band voltage shift (V)

4

Fig. 3. I–V characteristics before and after application of +15 V for 100 s. The lines shows the fittings based on Eq. (1). The fittings yield I0 = 311 × 10−10 A and = 193 for the situation before the voltage application and I0 = 600 × 10−11 and = 113 for the situation after the Ingenta to: voltage application.

Previously charged by +25 V for 0.1 s

1.6

3

Voltage (V)


Liu et al.

for t = 100 s. It means that it starts from a value bigger than the upper limit ( = 5/3) and drops in the range between the upper limit and the lower limit ( = 1). This indicates that the current conduction of the system evolves from a situation beyond two-dimensional transport (some three-dimensional transport in the system) to a quasi two-dimensional transport. The evolution has a trend towards the one-dimension transport when the charge trapping is increased.13 15 This evolution is due to the breaking of some tunneling paths as a result of charge trapping in the nc-Al, as discussed below. Under the influence of a voltage applied to the gate, electrons can tunnel into the nc-Al from the gate (or substrate) and tunnel out from the nc-Al towards the substrate (or the gate) depending on the polarity of the voltage. Electron tunneling can also take place between adjacent uncharged nanocrystals, and many such nanocrystals form conduction paths connecting the Si substrate to the metal gate as shown in Figure 5(a). For the virginby Delivered (a)

Metal gate

case, in addition to the one or two-dimensional transport, there may be some electron tunneling occurs in the third dimension, which makes beyond 5/3. The scenario of the influence of charge trapping in nc-Al on the tunneling paths is shown in Figure 5(b). Electron trapping in one nc-Al will affect the tunneling of other electrons into this nc-Al, and the relevant tunneling paths could be broken due to the Coulomb blockade effect. Therefore, as a result of the charging in nc-Al, with the breaking of some tunneling paths in the other two directions, the current conduction of the system will evolve towards the one-dimension transport. As the electric field of the applied voltage is along the vertical direction, the charges trapped in the paths along the vertical direction have a higher probability to tunnel out from the nanocrystals than those trapped in the paths along the lateral directions. Therefore, the charging in the paths along the lateral directions is more significant than in the paths along the vertical direction. The charging Ingenta to: in the lateral directions causes the breaking of some tunneling paths in the lateral directions, leading Nanyang Technological University to the evolution of the current conduction of the system IP : 155.69.4.4 towards the one-dimension transport. Mon, 01 Mar 2010 06:24:01

RESEARCH ARTICLE

4. CONCLUSIONS In conclusion, Al-rich AlN thin films deposited by RF sputtering of Al target in an argon and N2 gas mixture exhibit a memory effect. The memory effect is due to the charge trapping in the nc-Al embedded in the AlN matrix. The AlN thin films containing nc-Al provide the possibility of memory applications with low cost. On the other hand, the I–V characteristic of the system follows a power law, i.e., I = I0 V . With charge trapping in nc-Al, decreases towards = 1 of one-dimensional arrays, showing that the current conduction evolves with a trend towards the one-dimensional transport due to the breaking of some tunneling paths in the other two directions as a result of the charge trapping in nc-Al.

Si substrate Metal gate

(b)

Acknowledgments: This work has been financially supported by Singapore Millennium Foundation and by the Academic Research Fund from Ministry of Education, Singapore, under project No. ARC 1/04.

References and Notes Si substrate Uncharged nc-Al Tunneling path

Charged nc-Al Blocked path

Fig. 5. (a) Schematic illustration of the formation of tunneling paths between adjacent uncharged nc-Al embedded in the AlN matrix; and (b) illustration of the influence of charge trapping in some nc-Al on the tunneling paths.

602

1. Y. Taniyasu, M. Kasu, and T. Makimoto, Nature 441, 325 (2006). 2. G. F. Iriarte, J. Appl. Phys. 93, 9604 (2003). 3. V. Mortet, O. Elmazria, M. Nesladek, M. B. Assouar, G. Vanhoyland, J. D’Haen, M. D’Olieslaeger, and P. Alnot, Appl. Phys. Lett. 81, 1720 (2002). 4. S. Zhao and E. Wäckelgård, Sol. Energy Mater. Sol. Cells 90, 1861 (2006). 5. R. P. Netterfield, K.-H. Müller, D. R. McKenzy, M. J. Goonan, and P. J. Martin, J. Appl. Phys. 63, 760 (1988). 6. C. Chu, P. P. Ong, H. F. Chen, and H. H. Teo, Appl. Surf. Sci. 137, 91 (1999).


Liu et al.


7. K. Jagannadham, A. K. Sharma, Q. Wei, R. Kalyanraman, and J. Narayan, J. Vac. Sci. Technol. A 16, 2804 (1998). 8. E. S. Dettmer, B. M. Romenesko, H. K. Charles, Jr., B. Carkhuff, and D. J. Merrill, IEEE Trans. Components, Hybrids, Manufac. Technol. 12, 543 (1989). 9. I. C. Oliveira, K. G. Grigorov, H. S. Maciel, M. Massi, and C. Otani, Vaccum 75, 331 (2004). 10. S. Strite and H. Morkoç, J. Vac. Sci. Technol. B 10, 1237 (1992). 11. C. Y. Ng, T. P. Chen, L. Ding, and S. Fung, IEEE Electron Device Letters 27, 231 (2006).

12. C. Y. Ng, T. P. Chen, M. Yang, J. B. Yang, L. Ding, C. M. Li, A. Du, and A. Trigg, IEEE Trans. Electron Devices 53, 663 (2006). 13. A. A. Middleton and N. S. Wingreen, Phys. Rev. Lett. 71, 3198 (1993). 14. Y. Liu, T. P. Chen, P. Zhao, S. Zhang, S. Fung, and Y. Q. Fu, Appl. Phys. Lett. 87, 033112 (2005). 15. Y. Liu, T. P. Chen, H. W. Lau, J. I. Wong, L. Ding, S. Zhang, and S. Fung, Appl. Phys. Lett. 89, 123101 (2006). 16. H. E. Romero and M. Drndic, Phys. Rev. Lett. 95, 156801 (2005).

Received: 11 December 2006. Accepted: 10 April 2007.

Delivered by Ingenta to: Nanyang Technological University IP : 155.69.4.4 Mon, 01 Mar 2010 06:24:01

RESEARCH ARTICLE


603