Barrier capability of TaNx films deposited by different nitrogen flow rate against Cu diffusion in Cu/TaNx/n(+)-p junction diodes

Barrier capability of TaN

x

®lms deposited by di€erent nitrogen

¯ow rate against Cu di€usion in Cu/TaN

x

/n

‡

±p junction diodes

Wen Luh Yang

, Wen-Fa Wu

, Don-Gey Liu

, Chi-Chang Wu

,

Keng Liang Ou

a_{Institute and Department of Electrical Engineering, Feng Chia University, Taichung 407, Taiwan, ROC} b_{National Nano Device Laboratories, Hsinchu 300, Taiwan, ROC}

c_{Institute and Department of Mechanical Engineering, National Chiao-Tung University, Hsinchu 300, Taiwan, ROC}

Received 14 July 2000; received in revised form 24 October 2000; accepted 26 October 2000

Abstract

This paper investigates the barrier capability of tantalum nitride (TaNx) layers against Cu di€usion. The TaNxlayers

were reactively sputtered in contact holes to a thickness of 50 nm by using a di€erent nitrogen ¯ow rate. Results indicate that the TaNxlayers fail to be a di€usion barrier due to a relative high resistivity for nitrogen ¯ow ratios exceeding 10%.

In addition, we found that the phase of a-Ta(±N) functions as an e€ective barrier against Cu di€usion and that Cu/ TaN(3±5%)/n‡_{±p junction diodes are able to sustain a 30 min furnace anneal up to 500°C without causing degradation}

of the electrical characteristics. The high-temperature failure of barrier capability for the TaNxlayers is due to

inter-di€usion of Cu and Si across the TaNx®lm structure to form Cu3Si. The surface roughness and the ®lm structure of

TaNxlayers determine the ability of Cu and Si interdi€usion. Ó 2001 Elsevier Science Ltd. All rights reserved.

Keywords: Di€usion barrier; Copper; Tantalum; Cu3Si

1. Introduction

A high performance interconnection network on a chip is becoming increasingly important for ultralarge-scale integration (ULSI) of Si integrated circuits. Con-tinued shrinking of devices has led to a discrepancy between the device and interconnect performance. The obvious advantages for using Cu as interconnect mate-rial to substitute aluminum alloys are related to im-proving the operation speed and the reliability of ULSI circuits [1,2]. Compared with aluminum alloys, Cu provides a lower bulk resistivity, higher electro-migra-tion and stress-migraelectro-migra-tion resistance, higher melting point, and lower reactivity with commonly used di€u-sion barrier materials [3,4]. Unfortunately, Cu has drawbacks that retard its widespread application, such

as problems of dry etching, poor adhesion to oxide and other dielectric materials, easy oxidation in air, and fast di€usion in Si and oxide even at room temperature, re-sulting in degradation of the device characteristics [5,6]. However, if an appropriate di€usion barrier between Cu and its underlying layers is provided, Cu will satisfy the needs of future integrated circuits.

Recently, Cu has been used for global (long distance) interconnect. It is known that the variation of Cu sheet resistance with anneal temperature provides a good measure of barrier capability for the Cu/barrier/dielec-tric system. Furthermore, with continued shrinking of the devices, the local (short distance) interconnects will also change to a Cu metallization system to match the improving performance of ULSI circuits. For the local interconnect, Cu is directly connected to the source/ drain area of MOSFETs and the variation of Cu sheet resistance with anneal temperature is no longer a good criterion for evaluating the barrier capability of the Cu/ barrier/Si system. Although much research has been

*_{Corresponding author. Fax: +886-4-4516842.}

E-mail address: [email protected] (W.L. Yang).

0038-1101/01/$ - see front matter Ó 2001 Elsevier Science Ltd. All rights reserved. PII: S0038-1101(00)00228-8

devoted to evaluate the barrier capability by the mea-surement of Cu sheet resistance, few literature is avail-able on the measurement of junction leakage of Cu contacted p±n junctions [7,8]. Hence the objective of this research is to examine the reliability of Cu metallization for local interconnect.

Many di€usion barriers against Cu di€usion have been examined extensively in recent years [9±14]. Among them, tantalum nitride (TaNx) is by far the most

com-mon di€usion barrier for Cu because of the absence of any compounds between Cu and Ta, and also Cu and N [15±17]. In addition, the reaction temperature of possible silicide formation at the barrier/Si interface can be raised to a higher value, as compared with Ta/Si, by adjoining Ta atoms to Si as a compound in the form of a nitride [18]. Several researchers have investigated that the phase of TaNx ®lms sequentially formed by sputtering Ta

under increasing amounts of nitrogen partial ¯ow in-cludes tetragonal metastable phase Ta (b-Ta), nitrogen-incorporated cubic Ta (a-Ta(±N)), hexagonal Ta2N and

NaCl-type TaN [19,20] However, in most of the studies the barrier capability of TaNx®lms has been determined

by material analysis. In this paper, we investigate the barrier capability of TaNx ®lms deposited at various

nitrogen ¯ow ratios during reactive sputtering by elec-trical properties and material characteristics. Mean-while, we compare the variation of junction leakage with the variation of Cu sheet resistance for Cu/TaNx/p±n

junctions as the anneal temperature is changed. Our results may help to clarify whether the Cu sheet resis-tance or the junction leakage is a suitable measure of the di€usion capability of TaNx ®lms against Cu di€usion

for the Cu/barrier/Si system. 2. Experimental procedure

The barrier capability of TaNx ®lms against Cu

dif-fusion was investigated using a structure of Cu/TaNx/

n‡_{±p junction diodes. The key feature of this experiment}

is the di€erent nitrogen ¯ow rate during reactive sput-tering of TaNx formation. First, p-type (1 0 0)-oriented

Si wafers with a resistivity of 6±9 X cm were used in this study. After standard RCA cleaning, the wafers were administered the LOCOS process to de®ne active re-gions. The n‡_{±p junctions were formed by As}‡

im-plantation at 60 keV with a dose of 5  1015_cmÿ2

followed by the rapid thermal annealing (RTA) process at 1050°C for 30 s in N2 ambient. After the contact

windows were cleaned by dipping sample in HF, a re-actively sputtered TaNx ®lm of 50 nm thick was

de-posited onto the active regions with di€erent nitrogen ¯ow ratio. In this paper, nitrogen ¯ow ratio is de®ned as a ratio of N2partial ¯ow to total gas ¯ow …N2‡ Ar† and

the deposited TaNx ®lm is denoted as TaN (%). For

example, TaN (5%) is a TaNx ®lm reactively sputtered

by 5% nitrogen ¯ow ratio. Then Cu ®lm with a thickness of 300 nm was deposited subsequently in the same sputtering system without breaking vacuum. During the sputtering, gas pressure was maintained at 6 mTorr with a power selected at 500 and 1500 W for TaNx and Cu,

respectively. Finally, Cu and TaNxlayers were patterned

by dilute HNO3 and Cl2 plasma respectively for the

formation of Cu/TaNx/n‡±p junctions.

To investigate the barrier capability of TaNx ®lms

against Cu di€usion, the devices were thermally an-nealed at a temperature ranging from 400°C to 600°C for 30 min in a vacuum of 10ÿ3 _{Torr. For electrical}

analysis, the leakage current of the diodes was mea-sured by a HP4145B semiconductor parameter analyzer at a reverse bias of ÿ5 V. The sheet resistance (Rs) of

the Cu ®lms was determined by the 4-point probe mea-surement. In addition, X-ray di€raction (XRD), sec-ondary ion mass spectroscopy (SIMS), and atomicforce microscope (AFM) were used for material analysis.

3. Results and discussion

Fig. 1 shows the resistivity of the reactively sputtered TaNx®lms as a function of nitrogen ¯ow ratio. In this

®gure, pure b-Ta with a resistivity of 197 lX cm is ob-served without any nitrogen ¯ow. As seen in this ®gure, the resistivity of TaNx ®lms initially decreases with

in-creasing nitrogen ¯ow ratio and reaches a minimum value of 159 lX cm for 5% nitrogen ¯ow ratio. This is due to the change of phase from b-Ta to a-Ta(±N) and b-Ta exhibits a higher resistivity than nitrogen incor-porated a-Ta [21]. On the contrary, the resistivity of TaNx®lms increases slightly between nitrogen ¯ow ratio

of 5% and 10% and then increases dramatically for ni-trogen ¯ow ratio exceeding 10%. Fig. 2 shows the XRD spectra of TaNx®lms deposited at di€erent nitrogen ¯ow

ratios on the Si substrates. It is known that the crys-tallographic structure of TaNx®lm is a€ected by

nitro-gen ¯ow ratio during the reactive sputtering [22,23]. As seen in Fig. 2, di€raction patterns taken from the Ta ®lm can be indexed to a b-Ta (tetragonal) structure. At the nitrogen ¯ow ratio of 3%, the XRD spectrum shows a a-Ta(±N)(bcc) preferred orientation. Although the a-Ta (±N) peak is not prominent, it is still observed for the ®lms at nitrogen ¯ow ratios of 5%, 7%, and 10%. This indicates that the ®lms were mainly composed of amorphous-like materials. When nitrogen ¯ow ratio increases to 15%, NaCl-type TaN(1 1 1) and weak amorphous Ta2N(1 0 1) peaks were observed. As

nitro-gen ¯ow ratio is further raised up to 25%, the NaCl-type TaN peak predominates and becomes broader in the XRD spectra for the ®lms with the nitrogen ¯ow ra-tio higher than 25%. Since the resistivity of TaN(1 1 1) and Ta2N are larger than that of a-Ta(±N), hence the

resistivity of TaNx increased dramatically for nitrogen

¯ow ratios exceeding 10% as shown in Fig. 1. It should be noticed that although current ¯ows in the Cu/TaNx/

dielectric system is horizontal type and the current ¯ow path in TaNx is in parallel with the Cu path, while for

the Cu/TaNx/Si system, the current path is a vertical

type and TaNxresistance is in series to Cu (see the inset

of Fig. 1). Because the series resistance of TaNxlayer in

the Cu/TaNx/junction diodes depends on both the

re-sistivity and ®lm thickness of TaNx. In this paper, the

resistivity of TaNxis thought to be a relative high value

for the nitrogen ¯ow ratio exceeding 10% (with a TaN thickness of 50 nm). Decreasing the barrier thickness may lead to a higher acceptable range of resistivity while the thinner barrier will result in the poor barrier capa-bility. Therefore, when TaNx is used as a di€usion

bar-rier between Cu and Si, the thickness and resistivity of TaNx ®lm should be formed as small as possible. The

results shown in Figs. 1 and 2 suggest that NaCl-type TaN(1 1 1) or Ta2N(1 0 1) (nitrogen ¯ow ratio

exceeds 10%) is not suitable as a di€usion barrier for the Cu/TaNx/Si system due to the relative high

resistivity.

The variation of Cu sheet resistance as a function of the annealing temperature is commonly used to examine the capability of di€usion barrier against Cu di€u-sion. The di€erence of sheet resistance between the an-nealed and as-deposited samples, normalized to the sheet resistance of as-deposited samples, is called the variation percentage of sheet resistance …DRs=Rs %† and

is de®ned as follows:

DRs

Rs % ˆ

Rs;after annealÿ Rs;as-deposited

Rs;as-deposited  100%

It is well known that Cu di€uses fast in Si and forms Cu±Si compounds at a temperature as low as 200°C. The formation of Cu±Si compounds results in the sheet resistance of Cu/Si increase. Fig. 3 illustrates the varia-tion percentage of sheet resistance vs. annealing tem-perature for the Cu/TaNx/Si samples with nitrogen ¯ow

ratio ranging from 0% to 10%. In this ®gure, the Cu/ TaNx/Si samples remain stable in the measurement of

sheet resistance following anneal at temperature up to 650°C (DRs=Rs % slightly decrease with increasing

tem-perature due to the defect healing by thermal annealing). However, drastic increases in sheet resistance are found after annealing above 700°C. The drastic increase in sheet resistance is attributed to the formation of Cu3Si

precipitates from the XRD measurement (to be shown later in Fig. 4). On the other hand, as nitrogen ¯ow ratio exceeds 15%, there was the appearance of peeling which happened for Cu ®lms deposited on the TaNx®lms after

600°C annealing. Fig. 4 shows the XRD spectra for Cu/ Ta/Si and Cu/TaN(5%)/Si samples subjected to anneal at various temperatures. The di€raction patterns reveal that the two structures remain unchanged after anneal at temperature up to 600°C, while di€erent sets of peaks belonging to Cu3Si, Ta5Si3, and TaSi2 are found after

700°C annealing. The high-resistivity Cu3Si

forma-tion and related Cu decrease resulted in the drastic in-creases of sheet resistance as shown in Fig. 3. As seen in Fig. 4(b), peak intensity of Cu3Si for the Cu/TaN(5%)/Si

Nitrogen Flow Ratio (%)

0 5 10 15 20 25 30 35 40

Res

is

tivity

(

µΩ

-cm)

Fig. 1. Resistivities of the reactively sputtered TaNx®lms as a function of nitrogen ¯ow ratio. The inset is the cross-sectional view of

Fig. 2. XRD spectra for TaNxdeposited at various nitrogen ¯ow ratios.

sample is larger than that observed for the Cu/Ta/Si sample shown in Fig. 4(a). Hence, the fact that variation percentage of sheet resistance increases as nitrogen ¯ow ratio increases when annealed at 700°C (see Fig. 3) may be due to the di€erent amount of Cu3Si formation. On

the other hand, it had been reported that Cu(1 1 1) provides higher electro-migration resistance than that of Cu(2 0 0) [24]. In our experiment, the ratio of Cu(1 1 1) to Cu(2 0 0) for Cu ®lm deposited on the TaN(5%) layer was 315.72 (23050/73) while the ratio of Cu(1 1 1) to

Fig. 5. Leakage current densities of the Cu/TaNx/n‡±p junction diodes vs. nitrogen ¯ow ratio at various annealing temperatures.

Leakage current densities of the as-deposited sample (without any heat treatment) are also included for comparison.

Cu(2 0 0) on the Ta layer was 6.18 (1780/288). These results imply that TaN (5%) is a suitable di€usion bar-rier for the Cu metallization system from the views of resistivity, expected electro-migration resistance, and thermal stability. For further analyzing the thermal stability, SEM images were used to examine the surface morphologies of Cu ®lm after thermal treatments. The Cu ®lms remained stable on di€erent di€usion barrier (TaNxwith di€erent nitrogen ¯ow ratios) until anneal at

temperature up to 650°C. This is consistent with the results of sheet resistance and XRD measurements. Moreover, an increase in annealing temperature led to a change in color of Cu surface and a production of pre-cipitates. For example, the smaller Cu grain of about 1.5 lm in diameter under 650°C annealing and Cu±Si compound of bigger grain of about 6 lm in diameter

after 700°C annealing were observed for the Cu/Ta/Si sample. From the XRD spectra shown in Fig. 4, it is believed that the precipitates are Cu3Si phase.

Although the results of sheet resistance, XRD, and SEM measurements show that the Cu/TaNx/Si samples

remain stable as annealed up to 650°C, the junction characteristics of Cu/TaNx/p±n junction are necessary

for evaluating the barrier capability of TaNxagainst Cu

di€usion. Fig. 5 illustrates the reverse-biased current densities of the Cu/TaNx/n‡±p junction diodes with

di€erent nitrogen ¯ow ratios under di€erent annealing temperatures. In this measurement, the leakage current densities were obtained from an average value of 25 samples and the diode area was 1000  1000 lm2_{. For}

the diodes without any heat treatment (as-deposited), the leakage current densities remain stable (below 10

nA/cm2_{) as nitrogen ¯ow ratio is increased.}

Neverthe-less, the diode leakage increases with increasing the an-nealing temperature and most of diodes are degraded after annealing at 600°C. As seen in Fig. 5, the leakage current densities initially decrease, reaching a valley of minimum at 3% (for 400°C and 500°C annealing) or 5% (for 600°C annealing) nitrogen ¯ow ratio, and then in-crease with increasing the nitrogen ¯ow ratio. In other words, for the TaN (3%) and TaN (5%) di€usion bar-riers, the diodes endure thermal annealing at tempera-ture up to 500°C. For the TaNx®lms with nitrogen ¯ow

ratios exceeding 10%, the devices all failed (de®ned by a criteria of 10ÿ6 _A/cm2 _{leakage current density) after}

400°C annealing. According to the XRD results as shown in Fig. 2, our data provide the evidence that the a-Ta(±N) thin ®lm is a suitable barrier against Cu dif-fusion and the barrier capability of a-Ta(±N) is more e€ective than the b-Ta, Ta2N, and TaN ®lms. Moreover,

the barrier capability of TaN(1 1 1) (15±25% of nitrogen ¯ow ratio) is even inferior to the b-Ta(0 0 2) thin ®lm.

SIMS and AFM measurements were used to inves-tigate the failure mechanism of the fabricated diodes. As shown in Fig. 6(a), an interdi€usion of Cu and Si across

the barrier ®lm is found for the Cu/TaN(5%)/Si sample annealed at 500°C. After 500°C annealing, although the small amount of Cu atoms di€used into Si substrate does not a€ect the Cu sheet resistance signi®cantly (see Fig. 3), the junction leakage measurement shows a two order of magnitude increase in the leakage current measurement (from 10ÿ9_{to 10}ÿ7_A/cm2_{, as seen in Fig.}

5) for the Cu/TaN(5%)/n‡_{±p junction diodes. Therefore,}

it implies that the sheet resistance measurement is only valid for the global interconnections and the junction leakage evaluation is a suitable method for local inter-connections. As seen in Fig. 6(b), the large amount of interdi€usion between Cu and Si after 700°C annealing are observed by SIMS measurement, indicating that the reaction of Cu with Si results in the drastic increase of Cu sheet resistance. In addition, surface roughness of the TaNx ®lms was examined by AFM on unpatterned

samples. Fig. 7 shows the AFM images of TaNx ®lms

deposited on Si substrate with the nitrogen ¯ow ratio of 5%, 15%, and 25%, respectively. A fairly smooth surface with a root-mean-square (RMS) value of 0.135 nm is obtained for the TaN (5%) sample as shown in Fig. 7(a). Furthermore, increasing the nitrogen ¯ow ratio led to

deposition of relatively rough surface of TaNx ®lms.

RMS of 0.286 and 0.328 nm were obtained for nitrogen ¯ow ratios of 15% and 25%, respectively, as shown in Fig. 7(b) and (c). Since the larger RMS of TaNx ®lms

corresponds to a rougher interface and may have led to the poor barrier capability against Cu di€usion, these AFM results support the electrical measurements that junction leakage of the diodes increases with nitrogen ¯ow ratio when samples were thermally annealed. In addition to the surface roughness, the barrier capability of TaNx ®lms may be a€ected by the bulk structure of

TaNx®lms. With increasing the nitrogen ¯ow ratio, it is

known that the evolution of structure of the TaNx®lms

follows the zone model to progressively change from voided columnar (Ta), through ®brous of reduced grains (a-Ta(±N)), featureless structure (Ta2N), and ®nally to

columnar structure (TaN) [23]. The quasi-amorphous structure of a-Ta(±N) ®lms lengthen the di€usion path of Cu to react with Si, hence the TaN(3±5%) ®lms provided a better barrier capability against Cu di€usion. 4. Conclusion

The barrier capability of TaNx layers against Cu

di€usion by di€erent nitrogen ¯ow ratio was investi-gated. We found that the resistivity of TaNx ®lms

in-creases drastically for nitrogen ¯ow ratios exceeding 10% and the high-resistivity TaNx®lms are not suitable

for the use of Cu/TaNx/Si system. When Cu/TaNx/n‡±p

junction diodes were annealed up to 600°C, TaN-(0±10%) functions as an e€ective barrier against Cu di€usion for the sheet resistance measurement, while most of samples failed in the junction leakage evalua-tion. Results suggest that the junction leakage is more suitable for evaluating the barrier capability of TaNx

®lms on the Cu/TaNx/n‡±p junction diodes than the

variation of sheet resistance. In this paper, Cu/ TaN(3±5%)/n‡_{±p junction diodes are able to retain their}

integrity in electrical characteristics up to 500°C an-nealing. The high-temperature failure of barrier capa-bility for the TaNx ®lms is presumably due to the

interdi€usion of Cu and Si, forming Cu±Si related pre-cipitates, and the interdi€usion may be enhanced by the microstructure and resulting roughness of TaNxsurface.

Acknowledgements

This work was supported by the National Science Council (ROC) under the contract NSC-89-2215-E-035-013 and supported in part by Feng Chia University (FCU-RD-88-01). The authors would also like to thank the National Nano Device Laboratory for their techni-cal support.

References

[1] Murarka SP. Multilevel interconnections for ULSI and GSI era. Mater Sci Eng 1997;R19:87±151.

[2] Moslehi MM, Lino AP, Omsted T. Ultralarge scale integrated metallization and interconnects. J Vac Sci Technol 1999;A17(4):1893±7.

[3] Murarka SP, Hymes SW. Copper metallization for ULSI and beyond. Crit Rev Solid State Mater Sci 1995;20(2):87± 124.

[4] Chuang JC, Chen MC. Formation of Cr±O and Cr±N±O ®lms serving as Cu oxidation resistant layers and their N2

pre-sintering e€ect. Thin Solid Films 1998;335:146±52. [5] Chuang JC, Chen MC. E€ects of thermal N2annealing on

passivation capability of sputtered Ta(±N) layers against Cu oxidation. J Electrochem Soc 1998;145:4029±35. [6] Kaloyeros AE, Chen X, Stark T, Kumar K, Seo SC,

Peterson GG, Frisch HL, Arkles B, Sullivan J. Tantalum nitride ®lms grown by inorganic low temperature thermal chemical vapor deposition. J Electrochem Soc 1999; 146(1):170±6.

[7] Wang MT, Lin YC, Chen MC. Barrier properties of very thin Ta and TaN layers against copper di€usion. J Elec-trochem Soc 1998;145(7):2538±45.

[8] Kolawa E, Chen JS, Reid JS, Pokela PJ, Nicolet MA. Tantalum-based di€usion barriers in Si/Cu VLSI metalli-zation. J Appl Phys 1991;70(3):1369±73.

[9] Ono H, Nakano T, Ohta T. Di€usion barrier e€ects of transition metals for Cu/M/Si multilayers (M ˆ Cr, Ti, Nb, Mo, Ta, W). Appl Phys Lett 1994;64(12):1511±3. [10] Olowolafe JO, Mogab CJ, Gregory RB, Kottke M.

Interdi€usions in Cu/reactive-ion-sputtered TiN, Cu/chem-ical-vapor-deposited TiN, Cu/TaN, and TaN/Cu/TaN thin-®lm structures: low temperature di€usion analyses. J Appl Phys 1992;72(9):4099±103.

[11] Chuang JC, Tu SL, Chen MC. Sputtered Cr and reactively sputtered CrNx serving as barrier layers against copper

di€usion. J Electrochem Soc 1998;145(12):4290±6. [12] Chuang JC, Tu SL, Chen MC. Sputter-deposited Mo and

reactively sputter-deposited Mo±N ®lms as barrier layers against Cu di€usion. Thin Solid Films 1999;346:299±306. [13] Kwak MY, Shin DH, Kang TW, Kim KN. Characteristics

of TiN barrier layer against Cu di€usion. Thin Solid Films 1999;339:290±3.

[14] Wang MT, Lin YC, Lee JY, Wang CC, Chen MC. Thin-®lm properties and barrier e€ectiveness of chemically vapor deposited amorphous WSix®lm. J Electrochem Soc 1998;

145(12):4206±11.

[15] Oku T, Kawakami E, Uekubo M, Takahiro K, Yamaguchi S, Murakami M. Di€usion barrier property of TaN between Si and Cu. Appl Surf Sci 1996;99:265±72. [16] Cho SL, Kim KB, Min SH, Shin HK, Kim SD. Di€usion

barrier properties of metalloroganic chemical vapor de-posited tantalum nitride ®lms against Cu metallization. J Electrochem Soc 1999;146(10):3724±30.

[17] Lin JC, Lee C. Grain boundary di€usion of copper in tantalum nitride thin ®lms. J Electrochem Soc 1999;146(9): 3466±71.

[18] Takeyama M, Noya A, Sase T, Ohta A. Properties of TaNx

®lms as di€usion barriers in the thermally stable Cu/Si contact systems. J Vac Sci Technol 1996;B14(2):674±8.

[19] Stavrev M, Fischer D, Wenzel C, Drescher K, Mattern N. Crystallographic and morphological characterization of reactively sputtered Ta, Ta±N, and Ta±N±O thin ®lms. Thin Solid Films 1997;307:79±88.

[20] Min KH, Chun KC, Kim KB. Comparative of study tantalum and nitrides (Ta2N and TaN) as a di€usion barrier

for Cu metallization. J Vac Sci Technol 1996;B14(5):3263±9. [21] Chen GS, Chen ST, Yang TJ. Di€usion barrier properties of single- and multilayered quasi-amorphous tantalum nitride thin ®lms against copper penetration. J Appl Phys 2000;87(12).

[22] Stavrev M, Fischer D, Preu A, Wenzel C, Mattern N. Study of nanocrystalline Ta(N,O) di€usion barriers for use in Cu metallization. Microelectron Eng 1997;33:269± 75.

[23] Chen GS, Lee PY, Chen ST. Phase formation behavior and di€usion barrier property of reactively sputtered tantalum-based thin ®lms used in semiconductor metallization. Thin Solid Films 1999;353:264±73.

[24] Abe K, Harada Y, Onoda H. Study of crystal orientation in Cu ®lm on TiN layered structures. J Vac Sci Technol 1999;B17(4):1464±9.