Self-assembled near-zero-thickness molecular layers as diffusion barriers for Cu metallization

Abstract

The present invention provides a diffusion barrier in an integrated circuit. The diffusion barrier comprises a self-assembled monolayer. The diffusion barrier is preferably less than 5 nm thick; more preferably it is less than 2 nm thick. The self-assembled monolayer typically contains an aromatic group at its terminus.

Description

STATEMENT AS TO RIGHTS TO INVENTIONS MADE UNDER FEDERALLY SPONSORED RESEARCH OR DEVELOPMENT

[0002] The US Government may have certain rights in this invention pursuant to National Science Foundation CAREER grant DMR-9984478.

REFERENCE TO A “SEQUENCE LISTING,” A TABLE, OR A COMPUTER PROGRAM LISTING APPENDIX SUBMITTED ON A COMPACT DISK

[0003] Not Applicable

BACKGROUND OF THE INVENTION

[0004] A. Field of the Invention

[0005] The present invention generally relates to integrated circuits. In particular, it relates to forming a diffusion barrier layer comprising a self-assembled monolayer in an integrated circuit.

[0006] B. Description of Related Art

[0007] Copper is becoming the metal of choice for forming conductive patterns in integrated circuits. There are, however, unresolved issues with its use. For instance, copper diffuses rapidly in silicon and silicon dioxide. The diffusion, over time, results in junction linkage, which decreases device efficiency.

[0008] To address the problem of copper diffusion, researchers have developed “diffusion barriers.” A diffusion barrier is part of the metallization scheme, comprising a layer of material formed between an overlying copper layer and an underlying silicon or silicon dioxide layer. The diffusion barrier serves to inhibit the diffusion of copper into the surrounding layer.

[0009] The use of amorphous alloys as diffusion layers has been discussed. Amorphous binary silicides, such as molybdenum-, tantalum and tungsten silicide and amorphous ternary alloys (e.g., Ti—Si—N) have been reported as diffusion barriers. The formation of these layers, however, uses sophisticated sputtering processes and results in the inclusion of substantial contaminants.

[0010] Diffusion barriers made of TiN, TiSiN and TiN/TiSiN have also been reported. The titanium was deposited by sputtering in the presence of ammonia, which was used as a nitridation agent. A two-step annealing process completed the layer formation.

[0011] While a number of diffusion barriers have been discussed in the art, improved diffusion barriers are always desirable, especially diffusion barriers that are formed in very thin layers.

SUMMARY OF THE INVENTION

[0012] The present invention provides a diffusion barrier in an integrated circuit. The diffusion barrier comprises a self-assembled monolayer. The diffusion barrier is preferably less than 5 nm thick; more preferably it is less than 2 mn thick. The self-assembled monolayer typically contains an aromatic group at its terminus.

BRIEF DESCRIPTION OF THE DRAWINGS

[0013] FIG. 1 shows representative C-V curves from a control sample (open legends) and a SAM1-coated MOS structure (filled legends) obtained prior to BTA, and at failure.

[0014] FIG. 2 shows (a) flat band voltage shift, AVFB and (b) leakage current density, leakage, plotted as a function of tBTA for control and SAMI -coated samples.

[0015] FIG. 3 shows a box plot of the failure times of MOS structures with the different SAMs at the Cu/SiO2 interface. The boxes edges represent the upper and lower quartile value, the error bars show the maximum and the minimum, and the central line indicates the median failure time.

DESCRIPTION OF THE SPECIFIC EMBODIMENTS

[0016] Devising ultra-thin barrier layers to prevent Cu diffusion into SiO2-based dielectrics is a major challenge that must be met to increase the speed, number density, and performance of microelectronics devices. Here, we demonstrate the use of near-zero-thickness (<2-nm-thick) self-assembled monolayers (SAMs) as barriers to Cu diffusion into SiO2. Cu/SiO2/Si(001-metal-oxide-semiconductor (MOS) capacitors, with and without SAMs at the Cu/SiO2 interface, were annealed at 200° C. in a 2 MV cm−1 electrical field. Capacitance-voltage (C-V) and current-voltage (I-V) measurements of MOS capacitors coated with SAMs having aromatic terminal groups consistently show as much as 5-orders-of-magnitude lower leakage currents and a factor-of-4 higher time to failure when compared with the corresponding values from uncoated interfaces. SAMs with short tail lengths or aliphatic terminal groups are ineffective in hindering Cu diffusion, indicating that the molecular length and chemical configuration are key factors determining the efficacy of SAMs as barriers. We propose that the steric hindrance offered by the terminal groups in the SAMs are responsible for the barrier properties.

[0017] Copper is the preferred metal for creating multilevel interconnect structures in ultra-large-scale-integrated (ULSI) circuits because of its high electrical conductivity and electromigration resistance. (S. P. Murarka, Mater. Sci. Eng. R19, 87-151 (1997).) One of the challenges in Cu metallization technology is to prevent the rapid diffusion of Cu into SiO2 under an electrical bias during device operation. This is because Cu incorporation degrades the dielectric properties of the oxide layer, causing leakage currents and leading to inferior device performance and failure. (J. D. McBrayer, R. M. Swanson, and T. W. Signmon, J. Electrochecm. Soc. 133, 1243 (1986).) Furthermore, the use of many low dielectric-constant materials such as fluorinated SiOx is critical for achieving greater device speeds. (W. Chang, S. M. Jang, C. H. Yu, S.C. Sun, and M. S. Liang, San Francisco, Calif., USA, 1999 (IEEE, Piscataway, NJ), p. 295, 131-3; G. S. Chen and S. T. Chen, J. Appl. Phys. 87, 8473 (2000).) It is thus crucial to devise solutions to prevent Cu diffusion across SiOx-based materials.

[0018] Several researchers have advocated the use of ≅10 to 30 -nm-thick diffusion barrier layers of Ti- or Ta-based compounds or Cu-based alloys to alleviate this problem. (C. Ahrens, D. Depta, F. Schitthelm, and S. Wilhelm, Appl. Surf. Sci. 91, 285-90 (1995); P. J. Ding, W. A. Lanford, S. Hymes, and S. P. Murarka, Appl. Phys. Lett. 64, 2897 (1994); and, P. J. Ding, W. A. Lanford, S. Hymes, and S. P. Murarka, Appl. Phys. 75, 3627 (1994).) While this approach has been successful thus far, new types of barriers are likely to be needed at device dimensions below 100 nm in contemporary device architectures, and very high-aspect ratio structures in 3D-integration of multiple-wafer devices. The barrier layer thickness should be kept below 5 nm for future devices to fully realize the advantage of high conductivity Cu. This is difficult to achieve by conventional physical and chemical vapor deposition methods without compromising either conformal coverage of high aspect ratio features and/or the barrier layer microstructure, both of which reduce the efficacy of the barrier. (A. Z. Moshfegh and O. Akhavan, Thin Solid Films 370, 10 (2000); A. Sekiguchi, J. Koike, and K. Maruyama, J. Japan Inst. Metals 64, 379 (2000).)

[0019] In this letter, we demonstrate for the first time that certain types of SAMs with typical lengths of <2 nm can serve as barriers for device applications. SAMs have near-zero thicknesses (NZT) and, by definition, will occupy an insignificant fraction of the total via/hole volume, thereby maximizing the room for filling in low-resistivity Cu. SAMs are expected to have good step coverage on high-aspect ratio features due to their high sticking probability with the substrate and low probability of depositing on themselves. Measurements also indicate that these monolayers may promote adhesion of Cu to dielectric surfaces. (L. Klapp, A. Krishnamoorthy, S. P. Murarka, and G. Ramanath, Unpublished (2000).) Moreover, the molecular dimensions of SAM also make them attractive for molecular electronics applications. (M. A. Reed and J. M. Tour, Scientific American 282, 86 (2000).)

[0020] Cu/SiO2/Si and Cu/SAM/SiOs/Si metal-oxide-semiconductor (MOS) structures were fabricated from p-and n-type device-quality Si(001) wafers capped with a 85-nm-thick dry-thermal oxide. The back-oxide was stripped using HF, and a 500-nm-thick Al back-contact was deposited by DC magnetron sputtering in an Ar plasma at 5 mTorr. The chamber base pressure was 9×10−7 Torr. The samples were annealed in a 2×10−7 Torr vacuum at 450° C. to ensure the formation of an ohmic contact. These substrates were successively rinsed in ultrasonic baths of xylene, acetone, isopropanol and deionized water, and dried with N2 to provide a clean surface for assembling the monolayers. The SAMs were deposited (see below) on one half of the wafer and the other half was used to make the control sample.

[0021] Organosilane compounds dissolved in toluene to obtain a 1% solution were used to form SAMs on SiO2 by the procedure described by Dressick et al. (W. J. Dressick, C. S. Dulcey, J. H. Georger, G. S. Calabrese, and J. M. Calvert, J. Electrochem. Soc. 141, 210 (1994).) The samples were then washed with toluene and baked for 4 minutes at 120° C. In all the SAMs, the trimethoxysilane group (Si with three —OCH3 groups) is tethered to the SiO2 substrate, while the fourth Si bond is attached to a tail consisting of aliphatic and/or aromatic groups (see Table 1). Finally, a 1000 -nm-thick Cu film was sputter-deposited in Ar at 5 mTorr through a shadow mask to form 1.2 -mm-dia gate contacts of the MOS capacitors.

[0022] Small samples with 3×3 dot arrays sliced from the wafers were used for thermal annealing at 200° C. and an electric field of 2 MV/cm in flowing N2— the treatment is referred to as bias thermal annealing (BTA). Capacitance-voltage (C-V) and current-voltage (I-V) characteristics were measured at 30-minute intervals using HP 4280 AC and HP4140 pA DC instruments, respectively, after rapidly cooling the samples to room temperature prior to each measurement. The C-V and I-V responses without any electrical bias were also measured before BTA. The measurements were continued until the sample failed; the failure criterion being defined as leakage current density jleakage>1000 nA cm−2 at −40 V gate voltage.

[0023] Several reports have shown that the shift in the flat-band voltage VFB and the increase in leakage current jleakage are characteristic signatures of Cu ion diffusion and incorporation into SiO2.2 Comparison of the C-V and I-V characteristics of our MOS structures with and without SAM at the Cu/SiO2. FIG. 1 shows a typical normalized-capacitance (C/Cmaximum) vs. gate voltage plots for a control sample and a SAM1-coated MOS structure prior to BTA (annealing time tBTA=0 min), and at failure-defined by jleakage>1000 nA cm−2. The flat-band voltage of the control sample VFBcontrol shifts to lower values with increasing tBTA, indicating Cu ion diffusion. The control samples failed at tBTA≅150 min, corresponding to a flat-band voltage shift of ΔVFBcontrol≅18 V. At the same tBTA, SAM1-coated samples showed ΔVFBSAM<1.5 V and jleakage<30 nA cm

[0024] 2 (see FIG. 2), and failed only at tBTA≅650 min—a four-fold increase in the time to failure.

[0025] FIG. 2 a shows ΔVFB plotted as a function of tBTA for control and SAM1-coated samples. ΔVFBcontrol increases with tBTA rapidly [d(ΔVFB)/dt˜0.11 V min−1], and continuously, all the way to failure. But, ΔVFBSAM remains relatively unchanged at ≅1.5 V with only a marginal increase of ≅0.0029 V cm−1 until failure, at which point ΔVFBSAM increases to ≅5V. The leakage current density jleakage also shows similar characteristics as ΔVFB (see FIG. 2b). In the control sample, jleakage continuously increases at a rapid rate, while in the SAM-coated sample a relatively constant jleakage value of ≅10-30 nA cm−2 persists right until failure, when it abruptly shoots up to values >105 nA cm−2. We note that at tBTA corresponding to the failure of the control sample, jleakage in SAM1-coated samples is more than four orders of magnitude smaller at ≅10 nA cm−2.

[0026] The above results clearly indicate that SAM1 effectively hinders Cu diffusion into SiO2. In the control sample Cu diffuses continuously during the BTA treatment and the continued Cu ion accumulation in the oxide layer causes monotonically increasing leakage currents, and finally failure. In the SAM1-coated samples, Cu diffusion is negligible for much longer times, and at failure, the leakage current increases catastrophically.

[0027] In order to obtain insights into the mechanism by which SAM1 prevents Cu diffusion into SiO2, we investigated the barrier properties SAMs with different molecular lengths and terminal groups (see Table 1). FIG. 3 is a box plot that summarizes the failure times of MOS structures with the different SAMs at the Cu/SiO2 interface. Our results indicate that both the molecular chain length and the terminal group are important factors that determine the barrier properties of SAMs. MOS capacitors with SAM1 and SAM2, both of which are terminated by aromatic rings—pyridyl and phenyl, respectively—show longer failure times. The slightly higher average failure time in samples with SAM1 is probably due to enhanced interaction of Cu with the N in the pyridyl ring. (S. Kotrly and L. Sucha, Handbook of Chemical Equilibria in Analytical Chemistry (Ellis Horwood Limited, Chichester, 1985).) The average failure times of MOS capacitors with SAM3, which has an aliphatic terminal group, is similar to that of the control samples. These results indicate that the aromatic terminal group plays an important role in preventing Cu diffusion into SiO2. MOS structures with SAM4—shortest length compared to the other SAMs, but has an aromatic terminal group—exhibit only a marginally greater failure time than the control sample, indicating that SAMs with longer chains are better barriers.

[0028] Based upon our results, the barrier properties of SAMs can be explained in terms of the size and configuration of the terminal group, and the molecular chain length. We propose that the larger volume (compared with, for example, aliphatic groups) occupied by the aromatic rings sterically hinder Cu diffusion between the molecules through the SAM layer. SAMs with long chain lengths will screen Cu atoms from the influence of the SiO2 substrate, thereby preventing ionization and consequent acceleration by the externally applied electric field. The Si—(OCH3)3 head group is unlikely to play any significant role in hampering Cu diffusion because the Si—O—Si linkages they form—to tether the SAMs to the substrate—are similar to those in SiO2. While the mechanism by which the SAMs eventually fail is not clear from our experiments, the probable cause is the development of defects such as pinholes at the interface. (E. E. Polymeropoulos, J. Appl. Phys. 48, 2404 (1977); D. Vuillaume, C. Boulas, J. Collet, J. V. Davidovits, and F. Rondelez, Appl. Phys. Lett. 69, 1646 (1996).)

[0029] In summary, we have shown that self-assembled monolayers of organosilane molecules can inhibit Cu diffusion into SiO2 during bias thermal annealing. The size and configuration of the terminal functional group, and the chain length of the SAMs, are two important factors that determine barrier properties. SAMs are attractive for Cu-based microelectronics circuits because of their near-zero-thickness, good conformality, and facile processing.

Molecule	Chemical formula	IUPAC Name
SAM1	1	3-[2-(trimethoxysilyl)ethyl]pyridine
SAM2	2	2-(trimethoxysilyl)ethyl benzene
SAM3	3	n-propyl trimethoxysilane
SAM4	4	Phenyl trimethoxysilane
1
2
CH3CH2CH2Si(OCH3)3
3
3

Claims

1. A diffusion barrier in an integrated circuit, wherein the diffusion barrier comprises a self-assembled monolayer.

2. The diffusion barrier according to claim 1, wherein the diffusion barrier is less than 5 nm thick.

3. The diffusion barrier according to claim 1, wherein the diffusion barrier is less than 2 nm thick.

4. The diffusion barrier according to claim 1, wherein the self-assembled monolayer contains an aromatic group at its terminus.

5. The diffusion barrier according to claim 1, wherein the diffusion barrier inhibits the diffusion of Cu into a silicon-based substrate.