2535

IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. ED-34, NO. 12, DECEMBER 1987

Common Origin for Electron and Hole Traps in MQS Devices MOHAMMAD ASLAM,

Abstrucf-Experimentalevidence is provided to showthatmany electron and hole traps found in ultraclean and annealed SiO, layers are related to intrinsic oxygen deficient defects. These trapping sites are found to play a dominantrole inlow-field oxide breakdown, radiation sensitivity, and interface state generation in MOS devices. Thg saturation of SiOz with oxygenleads to the elimination of a large number of these traps and to the stabilization of SiOz layers for use in submicrometer devices.

I. INTRODUCTION LTRA-LARGE-SCALE integration (ULSI) requires feature sizes well below 1 pm. Theoperation of these submicrometer structures at existing supply voltages and their manufacturing processes are connected with several problems that ultimately deteriorate MOSFET characteristics: threshold shifts, transconductance, avalanche noise. breakdown, and conductance ( f The instabilities appearing during device operation at existing supply voltages may be divided into two groups: those due to high fields present in the oxide and those caused by the existence of high fields in the channel of a MOSFET. The high fields present in Si02 can encourage electron injection from the gate electrode or from the Si into Si02 through Fowler-Nordheim tunneling [l]. If, however, the field exceeds the breakdown strength of the oxide, complete device failure may take place. The existence of high fields in the channel ultimately leads to injection of hot electrons and holes into S O 2 . The injection process is enhanced near the drain edgewhere the channel field attains its highest value, causing the generation of electron-hole pairs through the avalanche process. Depending upon the transverse oxide field, the injection of electrons and/or holes may result. An additional problem arises through radiation damage due to ionizing radiation present in ULSI processing technologies and device operation in a radiation environment. Charge injection into the oxide by a Fowler-Nordheim [l] or avalanche process [2] causes two series instability problems for the MOSFET: trapping of some of the injected charge by the oxide traps [3] and the generation of new traps at the Si-Si02 interface [4], [5]. As a result of

U

-')

Manuscript received March 28, 1987; revised June 9, 1987. Most of the experimental work reported in this paper was performed at the Technical University of Aachen, Aachen, Germany, and was supportedby the DAAD. The author is with the Department of Electrical and Computer Engineering, Wayne State University, Detroit, MI 48202. IEEE Log Number 8716783.

MEMBER, IEEE

intensive investigations directed toward 'Na traps [6]-[E] and H20-related trapping centers 121, [3], [7], [9], it is now possible to eliminate these traps in thermally grown Si02 through ultraclean technology and post-oxidation anneal (POA). POA, however, produces both shallow electron traps (detectable at temperatures below 150 K) and hole traps [lo]-[13] with capture cross sections in the range of 10-'5-10-16 and 10-13-10-14cm2, respectively. The nature of these centers, as well as their relationship to submicrometer device instabilities, is not completely understood. In the present paper, new experimental data are presented to show a possible relationship between electron and hole traps found in annealed SiOz layers. The results lead to the conclusion that an oxygen deficiency in Si02 is responsible for electron and hole traps, the generation of new interface traps, low-field oxide breakdown, and radiation damage in MOS systems that have been treated at higher temperatures. The saturation of SiO, with oxygen leads to a high-quality oxide and the Si-Si02 interface acceptable for use in ULSI.

11. SAMPLEPREPARATION The MOS capacitors used in this study were fabricated from p- and n-type (100) silicon wafers with resistivities of 0.1 and 0.2 Q cm, respectively. They were oxidized and annealed at 1000 or 1050°C in a double-walled resistance-heated quartz furnace tube using very dry 02,N2, and Ar ( H20 content I 1 ppm). All high-temperature steps were terminated by pulling the samples into the cold zone of the furnace and allowing them to cool for 10 min in a nitrogen atmosphere. O2 annealing was performed immediately after POA simply by closing the valve for N2 while simultaneously opening the O2valve. The samples were eventually pulled from the hot zone and cooled in a Nz atmosphere. After the final high-temperature step, an aluminum layer 300 nm thick was e-beam evaporated; photolithography was or3.85 X used to define dots of areas equal to 1 X cm2. Finally, the samples were sintered at 450°C for 30 min in N2 to remove e-beam damage. For Si-gate samples an n-doped poly-Si layer 800 nm thick was deposited at 650°C. After deposition, the samples were annealed at 1000°C inN2 for 20 min and subsequently covered with a layer of Al. Electrodes were defined using chemical and plasma etching.

-

0018-9383/87/1200-2535$01.OO 0 1987 IEEE

2536

IEEE TRAWACTIONS ON ELECTRON DEVICES, VOL. ED-34, NO. 12, VECEMBER 1987

111. MEASURING TECHNIQUES Electrons or holes were injected from the Si into the Si02 by an avalanche injection technique [2]. A 50-kIH:r sawtooth voltage was used to generate the hot carriers ill Si that were injected into the Si02 at 100 K in order to fill the shallow electron traps, at 400 K for the deep electro:] traps, and at 300 K for the hole traps. Theavalanche CUIrent was kept constant by adjusting the sawtooth voltage. The trapped charge was monitored periodically by a highfrequency C-Vmeasurement of the flat-band voltage shilt AVFBas a function of time. A microprocessor system was employed for recording the data, which were finally tram-. ferred to a microcomputer (HP9845B) for evaluation. The effective densities NeEand capture cross sections c d the traps were evaluated assuming a first-order capture process, no detrapping, and the presence of one or mole centers of discrete cross section [12]. The densities of interface states were measured using a quasi-static C-V technique [14] with the help of an automatic setup interfaced to the microcomputer. This setup, equipped with a multifunction picoammeter (HP4140A), was also used to study the breakdown behavior for which a staircase voltage was used. IV. RESULTSAND DISCUSSIONS A . N2 Anneal The measurements of AVFBas a function of time were evaluated to obtain NeEand u for various electron and hole traps in SiO,. The results are summarized in Fig. 1, whfxe NeEis plotted as a function of POA time Tan,. In accordance with earlier reports in the literature, it was found cm2 are eliminated that electron traps with u = through POA [151. This center is related to the presence of H 2 0 in SiO, [9] and, therefore, its elimination is a consequence of H 2 0 outdiffusion during POA. However, an cm2 is found toptrelectron trapping site with u = sist even for longer POA times, although its density decreases sharply at in the beginning of the annealing process. This seems to suggest that there are two centers vvith the same value, but only one of these is eliminated through POA. The other center, which appears to be Imaffected by extended POA, could have been generated by POA at the beginning of the annealing process. POA generates shallow electron traps with u = 10- 19and hole traps with u = 10-'4-10-13 cm', in agreement with earlier reports [IO]-[12]. Our results (Fig. 1) demonstrate, in addition, that the effective densities of generated hole and shallow electron traps are comparable; their generation kinetics are also similar. Studies regarding the effect of the annealing ambien Ion traps revealed that the hole and shallow electron traps are also generated in an Ar atmosphere [16]. This rules out the possible role of N2 in the trapgeneration process. The high-temperature treatment, however, could cause some structural changes in Si02. To discuss the nature of these changes, a knowledge of the spatial distribution of the generated centers is required. Our earlier results reverled

HOLETRAPS 15

300°K

SHALLOW ELECTRONTRAPS

100°K

-0 L

z DEEPELECTRONTRAPS

DEEP ELECTRON TRAPS

t ann (rninl

Fig. 1. Effect of POA at 1000°C in N, on NeEforelectron and hole traps in SiO,;Do, = 50 nm.

a nonuniform distribution, with the centroid lying about 10 nm away from the Si02 interface [ 161, [ 171. These findings appear to indicate that the electron and hole traps present in ultraclean and annealed Si02 layers are due to intrinsic defects. In another experiment, the Si02 was sealed off by depositing a layer of poly-Si on it before the N2 annealing step. Here it was found that the same hole and shallow electron traps, with the above mentioned u values, are generated [13]. In addition, the effective density of the 10-'9-cm2 deep electron trap showed an increase with annealing time. Samples with poly-Si as the gate material are known to haveextremely low densities of H20-related traps. It is plausible, therefore, to assume that the lO-I9cm2 trap is the same center that appeared in the Al-gate samples (Fig. 1). These findings lend further support to the idea that thetrapping centers present in annealed Si02 films are intrinsic.

B. 0, Anneal To look into thenature of these centers, we considered the possibility of reversing the generation process. For this purpose the Si02 layer was given a short oxygen exposure, the so-called O2 anneal, immediately after a prolonged POA treatment. The effect of the O2 anneal on various trapping centers, as shown in Fig. 2, is surprising. It is seen that the electron and hole traps, generated through POA, can be eliminated through an O2 anneal. The effective densities of the shallow electron traps with CT = 10-17-10-19cm2, not shown in Fig. 2, are also diminished by the oxygen exposure [I 81. The O2 anneal has no appreciable effect on the H20-related traps [ 131. The optimal time for O2 anneal is found to be about 30 s for a 50-nm oxide at 1000°C. The change in oxide thickness after O2 anneal was found to be 5 0.2 nm. C. Suggested Model These findings suggest that the electron and hole trapping sites present in annealed Si02 layers are related to an O-deficiency in S O 2 . This implies that the formation of

2537

MOS DEVICES

ASLAM: COMMON IN ORIGIN FOR TRAPS

the injected charge that is captured by electron and hole traps. In fact, the interface state generation process can be enhanced through POA and reversed by O2 anneal, as shown in Fig. 3. This implies that intrinsic defects like 03=Si are also responsible for the instabilities related to interface state generation.

r SHALLOW ELECTRON TRAPS

0

1

-

L I

I

DEEP ELECTRON TRAP I

I

I

I

25

I

I

I

L

1

u =1019cm2 I

I

I

50

I

I

)

75

t ann (sec)

Fig. 2. Reduction of trap densities through O2 anneal at 1000°C,Do, 50 nm: POA 1 hr.

=

an 0-deficiency entails a redox reaction between the silicon substrate (in the case of a poly-Si gate, also between the electrode) and Si02 03=Si-O-Si~03

S 03=Si

-

*

+0

Si=03

E. Radiation Hardness Fig. 4 shows electron trapping behavior of two X-ray irradiated samples with and without a second PMA after irradiation. The radiation hardness of Si02 improves through O2 anneal for both of the samples. This indicates that the intrinsic defects are responsible for radiation damage in Si02. 0-deficient Si02becomes more sensitive to radiation since it contains large concentrations of electron and hole traps. The hole traps, due to their higher capture probabilities, are dominant in localizing holes generated in SiOz through irradiation. The localized holes can act as electron traps (Fig. 4(b)).

The removal of the traps through an O2 anneal would mean a reversal of this reaction. The defect 03=Si % E O 3 F. Low-Field Breakdown may undergo an internal electron transfer resulting in a To study whether the intrinsic defects also affect the dipole-like center. It could then be responsible for both breakdown behavior of Si02, two samples were given a electron and hole traps. Alternatively, or additionally, one prolonged POA treatment in order to produce a larger might think of single 03=Si centers, obtained upon concentration of intrinsic traps. Then, one of them was fragmentation through difision at high temperatures of exposed to O2 for 30 s at 1050°C. The I-Vcharacteristics Si=03 structure. This diffusion process of the samples are shown in Fig. 5 . the 03=Si would be promoted by the strain near the Si-Si02 interThe low-field breakdown peaks completely disappear in cen- the sample that was treated with 02.Both Al-gate and face, leading to a largerconcentration of the 0 3Si ~ ter in this region. This centercould conceivably also form poly-Si-gate samples show this behavior. At first sight, upon release of H 2 0 through high-temperature annealing this finding appears to contradict a report appearing in the literature [20] onthe improved breakdown strength of 03=Si-OH + HO-Si=O3 pyrolytically deposited Si-rich Si02 layers. In contrast to 03=Si-0 Si=03 H20 this, but supporting our results, the beneficial effect of a The fragmentation of 03=Si-0 Si-03 through dif- high-temperature O2 treatment on the breakdown strength of Si02 has recently been reported [21]. These findings fusion also produces the 03=Si-0 * center. The possible 0-deficient defects in Si02, which could suggest that the low-field breakdown phenomenon is rebe generated through POA and removed through O2 an- lated to intrinsic traps. neal, are 03=Si Si=03 and 03=Si - . Because of the limited number of possible intrinsic defects and large G. Submicrometer Technology number of trapping sites found in Si02, it is assumed that The rapid oxygen anneal is useful in suppressing inthe same defect may be responsible for more than one stabilities related to the high-temperature processing betrapping center. Experimental evidence to this effect for fore the deposition of a polysilicon layer on the oxide. the case of electron traps is provided by the transfer of The intrinsic defects generated through the high-temperelectrons, initially captured into shallow levels, to deeper ature processing performed after thedeposition of a polyelectron traps [ 191. silicon layer cannot be affected by the oxygen anneal. However, if the post-deposition processing time is limited D. Interface Traps to a few tens of seconds, the generation of electron and An indication that a similar relationship exists between hole traps may be suppressed. A study [22], using conelectron and hole traps has recently been provided by our ventional processing for oxide growth and rapid thermal results regarding the generation of interface states via processing (RTP) for post-deposition treatment, showed electron and hole injection [5]. It was found that the dis- that for an RTP of 10 s at 1000°C the densities of gentribution of generated interface states has a remarkable erated electron and hole traps were below 10" cmP2. similarity for electron and hole injection into annealed However, for longer RTP times (60 s ) the trap densities Si02. The densities of interface states, generated through increase to the upper 10" cm-2 range. Thus, RTP seems electron and hole injection, depend upon that fraction of to be a very promising technique for post-deposition pro-

-

-

-

-

*

+

-

- -

--

2538

IEEE TIMNSACTIONS ON ELECTRON DEVICES, VOL. ED-34, NO. 12, DECEMBER 1987

I

nealed Si02. A deep electron trap with u = lo-’’ cm2 is also found. Since all these centers are removed by a brief O2 anneal, they are apparently related to intrinsic defects such as 03=Si * and 03=Si S i s 0 3 . The intrinsic defects are responsible for instabilities related to charge trapping, interface state generation, radiation sensitivity, and low-field breakdown in ultraclean and annealed Si02 films. The saturation of Si02 with oxygen produces highly stable Si02 layers that can be used in VLSI and ULSI.

--

NO 0,ANNEAL

30 sec.

E-E,(eV)

:iF-

Fig. 3. Effect of O2 anneal on the generation of interface traps throuzh hole injection; Do, = 45 nm, Ninj= 4 X lo1’cm-’; POA 16 h at 1000°C.

REFERENCES

t

-E222150 > 5y:.-:::

NE:

1

>

NO

0,ANNEAL, 30 sec.

Q

[I] M. Lenzlinger and E. H. Snow, “Fowler-Nordeim tunneling into thermally grown SO2,” J. Appl. Phys., vol. 40, pp. 278-283, 1969.

O2 ANNEAL

O2 ANNEAL, 30 sec.

75

0

I-

0

1.0

2.0 16

0

-2

0.5

1.0

1.5

Nin, Hd6 cni3

Ninj 110 crn)

ACKNOWLEDGMENT The author is indebted to P. Balk and D. R. Young for very fruitful discussions at various stages of this work. A critical reading of the manuscript by M. P. Shaw is greatly acknowledged.

(a) (b) Fig. 4. Effect of O2 anneal on electron trapping (a) with and (b) withoJt PMA at 450°C in N2; dose is 10 Mrad; Do, = 50 nm; POA 1 h at 1000°C.

t sec.

E,,(Mvc~“)

Fig. 5. Effect of O2 anneal at 1050°C onlow-field breakdown in SO2; D,,, = 80 nm; POA 10 h in N2.

cessing in ULSI fabrication. Additionally, RTP can alsl:, be employed for growing thin oxide layers for use in ULSI devices. V. CONCLUSIONS Shallow electron traps ( u = 10-’6-10-’9 cm2) and hol,: traps ( u = 10-’3-10-14 cm2) with comparable densities and a similar spatial distribution are always present in an-

[2] E. H. Nicollian, A. Goetzburger, and C. N. Berglund, “Avalanche injection currents and charging phenomena in thermal Si02.” Appl. Phys. Lett., vol. 15, pp. 174-177, 1969. [3] D. J. DiMaria, “Defects and impurities in thermal SOz,’’ in Physics o f S i 0 , and its Interfaces, S. T. Pantilides, Ed. New York: Pergamon, 1978. [4] C. T. Sah, J. Y. Sun, and J. J. Tzou, “Generation-annealing kinetics and atomic models of a compensating donor in the surface space charge of oxidized silicon,” J . Appl. Phys., vol. 54, pp. 944-956, 1983. [5] L. Do Thanh, M. Aslam, and P. Balk, “Defect structure and generation of interface states in MOS structures,” Solid-state Electron., V O ~ .29, pp. 829-840, 1986. [6] R. Williams, “Photoemission of electrons from silicon into SO2,” Phys. Rev., vol. 140, pp. A569-575, 1965. [7] D. J. DiMaria, J. M. Aitken, and D. R. Young, “Electron trapping in aluminum-implanted Si02 films on Si,” J. Appl. Phys., vol. 46, pp. 1216-1222, 1975. [SI D. J. DiMaria, F. J. Feigl, and S. R. Butler, “Capture and emission of electrons at 2-4-eV-deep trap level in SiOz films,” Phys. Rev., V O ~ .Bl1, pp. 5023-5030, 1975. [9] A. Hartstein and D. R. Young, “Identification of electron traps in thermal Si02 films,” Appl. Phys. Lett., vol. 38, pp. 631-638, 1981. [IO] J. M. Aitken, D. R. Young, and K. Pan, “Electron trapping in electron-beam irradiated Si02,” J. Appl. Phys., vol. 49, pp, 3386-3391, 1979. [ l l ] T.H. Ning, “Thermal reemission of trapped electrons in SiOz,” J . A&. Phys., ~ 0 149, . pp. 5997-6003, 1978. [12] D. R. Young, E. A. Irene, D. J. DiMaria, R. F., Dekeersmaecker, and M. Z. Massoud, “Electron trapping in SiO, at 295 annd 77”K,” J . Appl. Phys., V O ~ .50, pp. 6366-6372, 1979. [13j M. Aslam, R. Singh, and P. Balk, “Nature of electron and hole traps in MOS systems with poly Si electrode,” Phys. Status Solidi, vol. 84, pp. 659-668, 1984. [14] C. N. Berglund, “Surface states at steam-grown Si02-Si interfaces,” IEEE Trans. Electron Devices, vol. ED-13, pp. 701-705, 1966. [15] D. R. Young, “Characterization of electron traps in Si02 as influenced by processing parameters,” J . Appl. Phys., vol. 52(b),pp. 4090-4094,1981. [16] M. Aslam, P. Balk, and D. R. Young, “High temperature annealing behavior of electron traps in thermal SiOz,” Solid-state Electron., V O ~ .27, pp. 709-719, 1984. [17] M. Aslam and P. Balk, “Processing dependence and structure of the hole traps in Si02,” in Insulating Films on Semiconductors, J. F. Venvay and D. R. Wolters, Eds. Amsterdam: North-Holland, 1983. [18] M. Aslam, to be published. [19] -, “Electron self-trapping in Si02,” J. Appl. Phys., vol. 62, no. 1, pp. 159-162, 1987. [20] S. K. Lai, Semiconductor Silicon 198I, H. R. Huff, R. J. Kriegler, and Y . Takeishi, Eds. Pennington, NJ: The Electrochem. SOC., 1981, p. 416.

ASLAM: COMMON ORIGIN FOR TRAPS IN MOS DEVICES

[21] S . S .Cohen, “Electrical properties of post-annealed thin SiO, films,” J . Electrochem. SOC.,vol. 130, pp. 929-932, 1983. [22] M. Aslam, P. Balk, G. K. McGinty, and K. H. Nicholas, to be published.

*

Mohammad Aslam (”87) was born in Gujrat, Pakistan. He received the M.Sc. degree in physics from Panjab University, Lahore, in 1969. He received the M.S. and Ph.D. degrees in microelectronics from the Technical University of Aachen, Aachen, Germany, in 1979 and 1983, respectively. He worked as a Lecturer during 1970-1974 at Panjab University and as an Assistant Professor during 1974-1975 at the Aeronautical Engineering College, Karachi. He held a DAAD fellow-

2539 ship from 1975 to 1983 and a post-doctoral appointment during the period of 1983-1984 at the Technical University of Aachen. Before joining Wayne State University as an Assistant Professor of Electrical and Computer Engineering in 1986, he was serving as a Squadron Leader in the Pakistan Air Force in Karachi. His currentinterests include theuse of new technologies in MOS processing, rapid thermal processing, ULSI instabilities related to the charge trapping centers in MOS, radiation hardness and breakdown behavior of S O z , and electron self-trapping in Si02. Dr. Aslam is a member of the American Physical Society and the American Vacuum Society.