Hydrogenation of polycrystalline silicon thin film transistors by plasma ...

42 1

IEEE ELECTRON DEVICE LETTERS. VOL. 16. NO. IO. OCTOBER 1995

Hydrogenation of Polycrystalline Silicon Thin Film Transistors bv Plasma Ion Imdantation 4

1

James D. Bernstein, Student Member, IEEE, Shu Qin, Member, IEEE, Chung Chan, Senior Member, IEEE, and Tsu-Jae King

Abstract-Both n- and p-channel polycrystalline silicon (polySi) thin Mm transistors (TFT’s) have been hydrogenated using the plasma ion implantation (PII) technique. Significant improvements in device characteristics have been obtained. Because PI1 is capable of greater dose rates than plasma immersion, it allows for significantly shorter process times than other methods investigated thus far.

I. INTRODUCTION

P

OLYCRYSTALLINE silicon (poly-Si) thin film transistors (TIT’S)have recently been under investigation due to their application to large area flat panel active matrix liquid crystal displays (AMLCD’s). Compared to the current amorphous silicon TFT technology, poly-Si TIT’Shave better device characteristics which will allow for higher quality flat panel displays (FPD’s). However, poly-Si TFT performance is limited by the trap states caused by grain boundary and intragranular defects inherent in polycrystalline structures, as well as the interface traps common to all types of devices, including amorphous and crystalline. Hydrogen passivation has been shown to be an effective method for the reduction of these trap states. One of the most thoroughly investigated hydrogenation methods so far [ 11-[3] consists of immersing the TFT’s in the plasma produced by an alternating current (AC) parallel-plate reactor. Another method [4]-[6] is by immersion in an electron cyclotron resonance (ECR) plasma. In both AC and ECR plasmas, the introduction of hydrogen into the device is believed to take place through plasma ion penetration and bulk diffusion. The rate of ion penetration is dependent on the plasma’s ion density. An AC hydrogen plasma typically has an ion density of ni z lo9 ~ m - while ~ . for an ECR hydrogen plasma, density can be as high as n 1 z lo1’ cmP3. During hydrogenation, the TFl- is usually kept near 300-350”C to enhance hydrogen diffusion. The relatively low ion densities characteristic of AC plasmas necessitate long processing times. While an ECR plasma has a higher ion density, the sheath potential limits the hydrogen ion current to the TFT’s, which also prolongs process time. Plasma ion implantation (PII) is a promising technique for materials [7] and semiconductor processing [8], [9]. Ion energies can range from 1-100 keV with average ion flux densities as high Manuscript received April 1 I , 1995; revised June 16, 1995. J. D. Bernstein, S . Qin, and C. Chan are with the Plasma Science and Microelectronics Research Laboratory, Department of Electrical and Computer Engineering, Northeastern University, Boston, MA 021 15 USA. T.-J. King is with the Xerox Palo Alto Research Center, Palo Alto, CA 94304 USA. IEEE Log Number 9414366.

as 10l6 cmP2 sec-’. This work represents the first application of PI1 to hydrogenation of poly-Si TFT’s. The PI1 process is performed by repetitively applying a large negative voltage pulse to a wafer immersed in a hydrogen plasma. Hydrogen ions are accelerated by the target potential and implanted into the sample. Although the primary mechanism for the introduction of hydrogen into the device is ion implantation, ion penetration also takes place between the pulses. Because ion implantation is capable of delivering higher dose rates than methods which rely on surface penetration and bulk diffusion, the PI1 method allows for shorter process times. 11. EXPERIMENTAL PROCEDURE

Undoped 100 nm-thick amorphous silicon films were deposited on 525 pm-thick fused silica substrates by lowpressure chemical vapor deposition (LPCVD) and crystallized at 600OC. After the definition of silicon island regions, a 85 nm-thick gate oxide was deposited by LPCVD and then annealed at 95OOC in an 0 2 ambient, resulting in a final gate-dielectric thickness of 100 nm. A 350 nm-thick polySi gate layer was then deposited and patterned. Self-aligned source/drain regions were formed by phosphorus or boron ion implantation for n-channel or p-channel devices, respectively. Afterwards, a 700 nm-thick passivating layer of LPCVD oxide was deposited and a dopant activation anneal was performed at 600°C. Device fabrication was completed with conventional contact hole formation and AlCu deposition and etch processes, and the TFT’s were sintered at 450°C for 30 minutes in forming gas. PI1 hydrogenation process conditions were as follows: base pressure lop6 Torr, working pressure 50 mTorr, microwave power 730 W, pulse voltage -6 kV, pulse repetition rate 4000 Hz, pulse width 5 psec, and sample temperature 35OOC. The PI1 system has been described in a previous paper [ 101. Conventional AC plasma immersion hydrogenation was performed in a parallel plate reactor at 35OOC with a H2 and Ar gas mixture [2]. TFT characteristics were measured throughout the experiment with a Hewlett Packard 5 155A Semiconductor Parameter Analyzer. 111. RESULTSAND DISCUSSION

With the aforementioned PI1 process conditions, improvements in n- and p-channel device characteristics saturate in 30 minutes, whereas up to 8 hours is required with conventional plasma immersion. Drain current ( I d , ) versus gate

0741-3 106/95$04..OO GI 1995 IEEE

IEEE ELECTRON DEVICE LETTERS, VOL. 16, NO. 10, OCTOBER 1995

422

pchannel 'I"

beburndm as 8-4~. -fabricated immersion

80411. ECR

30-m.b.

as

30-min.

fabricated

8-hr. immersion

8O-m.b.

PII

ECR

PI1

pdT(cm2Ns)

66.9

94.5

86.3

97.6

48.3

56.1

68.4

56.9

0

135.4

7.8

31.2

8.3

-109.1

-36.8

-15.8

-13.2

IL

J a l k

1 . 1 ~ 1 0 ~ 5 . 0 ~ 1 0 ~ l.0x107

4 . 6 ~ 1 0 ~ 9.9~10'

2.6~10~2.0~10~1.2~10~

vu3 09

5.4

2.4

1.9

2.3

-5.0

-3.5

-1.1

-3.6

s

0.88

0.3 1

0.40

0.36

1.17

0.61

0.25

0.54

(v/lw

voltage (V,,) curves for unhydmgenated, PII hydrogenated, and plasma immersion hydrogenated n- and p-channel devices with channel width-to-length ratios W/L = 50-pd50-pm are shown in Fig. l(a) and (b). The drain voltage (vds) is 10 V. The threshold voltage, mobility, leakage current, and subthreshold slope all undergo marked improvement during the PII process. Table I summarizes the TIT device parameters. Results for an ECR plasma immersion technique [ 5 ] , which required 80 minutes to achieve TIT performance saturation for our device structure and fabrication process, are also shown for comparison. The threshold voltage is defined at a fixed normalized drain current, I d = 10-7x ( W L ) A, to avoid the ambiguity from the nonlinearity of the I-V characteristics and to compare thresholds at the same degree of band bending [ll]. The mobility is calculated from the maximum transconductance at vds = 0.1 V. The on-off current ratio Ion/Ioff represents the ratio of drain current at V, = 20 V to the leakage current, I d s , m l n . The mobility and threshold voltage for a n-channel TIT are plotted against process time in Fig. 2(a) while the leakage current and inverse subthreshold slope are plotted against process time in Fig. 2(b). It is interesting to note that each parameter is affected at the onset of PII hydrogenation. This is in contrast to conventional plasma hydrogenation for which the mobility and leakage current experience almost no change for the first several hours of processing [2]. This phenomena in the conventional plasma immersion process has been attributed to the relatively small initial concentration of hydrogen which passivated only the midgap grain boundary defects that influence threshold voltage and subthreshhold slope. The tail states from strain bonds which affect mobility and leakage current are not passivated until the deep dangling bond states are filled. The different response to hydrogenation suggests that there may be a different passivation mechanism for the PII process. After hydrogen enters the device, whether it is by conventional parallel plate plasma immersion, ECR plasma immersion, or our low energy plasma ion implantation, it reaches the active channel by diffusion. A 10 keV H+ ion will penetrate Si02 to a depth of approximately 1500 A [12], so a -6 kV pulse applied to a collisional plasma with an ion energy distribution certainly cannot penetrate the 7000 A-thick Si02 cap, gate

7 Y

---

as fabricated

t-

z W

U U

3

0

3

n

GATE VOLTAGE (-V) (b)

Fig. 1. I-I/ characteristics before and after PII hydrogenation and conventional parallel plate plasma immersion hydrogenation for (a) n-channel and (b) p-channel poly-Si CMOS TITS.

oxide, and poly-Si gate. However, the PI1 implanted hydrogen will diffuse faster, due to the concentration dependence of diffusion. As large quantities of highly concentrated hydrogen diffuse through the device and encounter the channel defects, both deep and tail states will be passivated simultaneously if the hydrogen concentration is large enough compared to that of the coordination defects.

BERNSTEIN et al.: HYDROGENATION OF POLYCRYSTALLINE SILICON THIN FILM TRANSISTORS

, im

55,

423

limiting factor for PI1 hydrogenation assumes that diffusion from the dielectric cap into the channel and charging during the pulse will not be constraints. This was the case for our processing conditions (where the pulse generator output was the limiting factor), however further experiments should be conducted at higher dose rates. It should be noted that different fabrication methods and device structures make it difficult to compare these results with the results of other work. For example, devices that either lack [ 5 ] , or have a thinner [4], [6] passivation oxide layer would require shorter hydrogenation times.

IV. CONCLUSIONS Hydrogenation by PI1 has been found to dramatically improve poly-Si TFT characteristics. It is capable of much shorter process times than methods which rely on plasma immersion as the means for introduction of hydrogen into the device. We believe that the rate-limiting factor for PI1 is the maximum pulse repetition frequency.

REFERENCES

10

XI

33

40

PROCESS T W (minutes)

(b) Fig. 2. (a) Threshold voltage and carrier mobility and (b) inverse subthreshold slope and leakage current versus process time for a PI1 hydrogenated n-channel TFT.

The main advantage of PI1 is its capability to provide high dose rates. Ion penetration is not a constraint. The dose rate for the PI1 process can be calculated with a model that accounts for collisional plasmas and pulses with finite rise times [lo], [13]. The current dose rate for our experiment is 3.3 x cmP2 sec-’. A dose of 6 x 1017 cm-’ will saturate device parameters. The time to saturation was limited by the maximum ion density from our plasma source (ni z 2.5 x 10” crnp3) and the average power of our high voltage pulse generator. Very short processing times are possible. For example, increasing the pulse repetition rate to 15 kHz while holding other process conditions constant will decrease the required process time for device characteristic saturation to approximately 5 minutes (dose rate = 2.0 x 10’’ cm-’ sec-’). Note that there is a physical limitation on the maximum dose rate-and hence the minimum process time. This is due to ion depletion around the target [14]. For our conditions, the calculated maximum allowable pulse frequency is 104 kHz. The pulse width is intentionally kept short to reduce charge accumulation during the time of the pulse. This is necessary since the devices were fabricated on a quartz substrate, an insulator. Excessive charging can result in damage to the device. Citing the pulse repetition frequency as the rate-

[ I ] T. I. Kamins and P. J. Marcoux, “Hydrogenation of transistors fabricated in polycrystalline-silicon films,” IEEE Electron Device Lett., vol. I, no. 8, pp. 159-161, 1980. [2] I. Wu, T. Huang, W. B. Jackson, A. G. Lewis, and A. Chiang, “Passivation kinetics of two types of defects in polysilicon TFT by plasma hydrogenation,” IEEE Electron Device Lett., vol. 12, no. 4, pp. 181-183, 1991. [3] U. Mitra, B. Rossi, and B. Khan, “Mechanism of plasma hydrogenation of oolvsilicon thin film wnsistors,” J. Elecrrochem. Soc., vol. 138, no. I 1 ,* pp. 3420-3424, I99 1. 141 . _T. Takeshita, T. Unnagami, and 0. Kogure, “Study of ECR hydrogen plasma treatment on Gly-Si thin film Gansistors,”-Jpn. J. Appi. Phys., vol. 27, no. 11, pp. L2118-L2120, 1988. 151 R. A. Ditizio, G. Liu, S. J. Fonash, B. C. Hseih, and D. W. Greve, “Short time electron cyclotron resonance hydrogenation of polycrystalline silicon thin film transistor structures,” Appl. .. Phys. Lett., vol. 56, no. 123, pp. 1140-1142, 1990. 161 K. Baert. H. Murai, K. Kobayashi, H. Namizaki, and M. Nunoshita, “Hydrogen passivation of polysilicon thin-film transistors by electron cyclotron resonance plasma,” Jpn. J. Appl. Phys., vol. 32, pp. 2601-2606, 1993. [7] J. R. Conrad, J. Radtke, R. A. Dodd, and F. Worzala, “Plasma source ion implantation technique for surface modification of materials,” J. Appl. Phys., vol. 62, no. 11, pp. 45914596, 1987. [SI N. W. Cheung, “Plasma immersion ion implantation for ULSI processing,” Nucl. Instr. Meth., vol. B55, pp. 811-820, 1991. (91 S. Qin, N. McGruer, C. Chan, and K. Warner, “Plasma immersion ion implantation doping using a microwave multipolar bucket plasma,” IEEE Trans. Electron Devices, vol. 39, no. IO, pp. 2354-2358, 1992. [IO] S. Qin, C. Chan, N. McCruer, J. Browning, and K. Warner, “The response of a microwave multipolar bucket plasma to a high voltage pulse,” IEEE Trans. P l a s m Sei., vol. 19, no. 6, pp. 1272-1278, 1991. [ 1 I] A. G. Lewis, I. W. Wu, T. Y. Huang, M. Koyanagi, A. Chiang, and R. H. Bruce, “Small geometry effects in n- and p-channel polysilicon thin film transistors,” IEDM Tech. Dig., p. 260, 1988. [12] J. F. Ziegler, J. P. Biersack, and U. Littmark, The Stopping Range of Ions in Solids. New York: Pergamon Press, 1985. [I31 S. Qin and C. Chan, “The response of a microwave multipolar bucket plasma to a high voltage pulse with finite rise-time,” IEEE Trans. Plasma Sei., vol. 20, no. 5 , pp, 569-571, 1992. [ 141 B. P. Woods, “Displacement current and multiple pulse effects in plasma source ion implantation,” J. Appl. Phys., vol. 73, no. 10, pp. 47704778, 1993.

__