- Email: [email protected]
Plasma etching of high-k and metal gate materials
ARTICLE IN PRESS
Vacuum 80 (2006) 761–767 www.elsevier.com/locate/vacuum
Plasma etching of high-k and metal gate materials Keisuke Nakamura, Tomohiro Kitagawa1, Kazushi Osari, Kazuo Takahashi, Kouichi Ono Department of Aeronautics and Astronautics, Graduate School of Engineering, Kyoto University, Yoshida-Honmachi, Sakyo-ku, Kyoto 606-8501, Japan
Abstract Etching characteristics of high-k dielectric materials (HfO2) and metal electrode materials (Pt, TaN) have been studied in high-density chlorine-containing plasmas at pressures around 10 mTorr. The etching of HfO2 was performed in BCl3 without rf biasing, giving an etch rate of about 5 nm/min with a high selectivity of 410 over Si and SiO2. The etching of Pt and TaN was performed in Ar/O2 with high rf biasing and in Ar/Cl2 with low rf biasing, respectively, giving a Pt etch rate of about several tens nm/min and a TaN etch rate of about 200 nm/min with a high selectivity of 48 over HfO2 and SiO2. The etched profiles were outwardly tapered for Pt, owing to the redeposition of etch or sputter products on feature sidewalls, while the TaN profiles were almost anisotropic, probably owing to the ionenhanced etching that occurred. r 2005 Elsevier Ltd. All rights reserved. Keywords: Plasma etching; Chlorine-containing plasma; High-k dielectrics; Metal electrode materials; HfO2; Pt; TaN
1. Introduction As integrated circuit device dimensions continue to be scaled down, the gate width of advanced microelectronic devices is projected to be scaled down to much less than 100 nm, and the thickness of gate oxides is also reduced down to 2 nm or less for conventional SiO2. Correspondingly, the technological challenge continues for growing ultrathin SiO2 films of high quality, to maintain the gate capacitance without increasing the gate leakage current and reducing the oxide reliability; however, the ultimate solution would rely on high dielectric constant (k) materials. Recent efforts have been made to replace SiO2 with silicon-oxynitrides of slightly high k, and nowadays, new high-k (420) dielectrics or metal oxides such as Al2O3, HfO2, ZrO2, and their silicates and aluminates are being developed to replace SiO2 [1,2]. In practice, the metal oxides give the required specific gate capacitance at a considerably larger thickness as compared to SiO2. In integrating these materials into device fabrication, an understanding of the etching characteristics of high-k Corresponding author. Tel.: +81 75 753 5793; fax: +81 75 753 5980.
E-mail address: [email protected] (K. Ono). Present address: NTT communications Corporation, Hitotsubashi, Chiyoda-ku, Tokyo 100-8128, Japan. 1
0042-207X/$ - see front matter r 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.vacuum.2005.11.017
materials is strongly required for their removal prior to forming the source and drain contacts. Moreover, for gate stacks with high-k dielectrics, gate electrodes of conventional polycrystalline silicon (poly-Si) tends to cause some problem of the depletion layer present in doped poly-Si gate materials. Thus, for lower equivalent oxide thicknesses, recent efforts have also been made to replace poly-Si gate electrodes with metal gates of WN, TiN, TaN, Ru/RuO2, Pt, and/or Ir [1,2]. For gate etch processes, a precise control of the profile and critical dimension of gate electrodes and a high selectivity to gate oxides have historically been the two most important issues to be addressed. Plasma etching of high-k oxides and metal electrodes has been studied in the application of ferroelectric materials, buffer layers, and capacitor dielectrics to memory devices. However, only a few studies have recently been concerned with the plasma etching of high-k dielectrics [3–13] and metal electrodes [11,13] for the application to high-k gate stacks. Pelhos et al. studied the etching of high-k gate dielectric Zr1xAlxOy thin films in helical resonator plasmas with Cl2/BCl3 [3]. Sha et al. investigated the etching of ZrO2 in electron–cyclotron–resonance (ECR) plasmas with Cl2/Ar [4] and BCl3/Cl2 [6], and also the etching of HfO2 in ECR BCl3/Cl2 plasmas [7,8]. Norasetthekul et al. reported on the etching of HfO2 in inductively
ARTICLE IN PRESS 762
K. Nakamura et al. / Vacuum 80 (2006) 761–767
coupled plasmas (ICP) with Cl2/Ar, SF6/Ar, and CH4/H2/ Ar [5]. Moreover, Maeda et al. reported on the HfO2 etching in ICP CF4 and Cl2/HBr/O2 plasmas [9]. Chen et al. investigated the etching of HfO-based high-k films in ICP plasmas with Cl2/HBr and CF4/CHF3 [10]. Emphasis of most of these studies has been placed on etch chemistries giving an etch selectivity of 41 over the underlying Si [3–9] and SiO2 [9]. More recently, Kota et al. have reported on the etching of metal/high-k stacks (TiN/TaN/HfO2) in ICP plasmas with BCl3-based chemistries [11]. It is further noted that the etching of high-k dielectrics with low ion energies and/or less ions is also indispensable in mass production, for chamber cleaning of the chemical vapor deposition (CVD) and atomic layer deposition (ALD) apparatuses to prepare high-k thin films. We have investigated the high-density plasma etching of HfO2 with attention being focused on etch chemistries and plasma conditions to achieve a much higher selectivity of b1 over Si and SiO2. We found the highly selective etching of HfO2 over Si in ICP fluorocarbon plasmas [12]; the selectivity was 45 in C4F8/Ar plasmas and the selectivity was further increased to b10 in C4F8/Ar/H2 plasmas, where carbon and/or carbonaceous species would work as surface inhibitors on Si, as in the situation of highly selective etching of SiO2 over Si with fluorocarbon plasmas. This paper presents a highly selective etching of HfO2 over Si and SiO2 in ECR chlorine-containing plasmas without rf biasing [13]. The paper also presents the etching of Pt and TaN in ECR chlorine-containing plasmas [13], with emphasis on the etch anisotropy and selectivity of metal electrodes over HfO2 and SiO2 (over the underlying high-k and the overlying mask materials). 2. Experiment Fig. 1 shows the experimental setup. The plasma reactor consisted of a grounded plasma and specimen chamber made of stainless steel. A set of solenoid coils were placed around the former, providing divergent magnetic fields in the specimen chamber. The discharge was established by 2.45-GHz right-hand circularly polarized microwaves of TE11 mode injected through an entrance quartz window at the top of the plasma chamber, and the 875-G ECR resonance was located near the end of the plasma chamber. The specimen chamber, 36 cm in diameter and 40 cm in length, had an end port for pumping and several side ports for diagnostics, containing a 16-cm-diam floating electrode or substrate holder 20 cm downstream (BE200 G) from the ECR resonance region. The electrode was capacitively coupled to a 13.56-MHz rf power supply through an impedance matching network for additional rf biasing. Feedstock gases employed were Cl2, BCl3., O2, and their mixtures with Ar, being introduced into the reactor evacuated to a base pressure o106 Torr, through a set of small holes at side walls of the reactor. The experiments were performed at incident microwave powers of nominally PMW ¼ 400–600 W, rf bias powers of nominally
Microwave (2.45 GHz) Solenoid Coils
Quartz Window
Gas Wafer Stage
OES Langmuir Probe
To Pump
Matching Box RF (13.56 MHz)
Fig. 1. Experimental setup for ECR plasma etching.
Prf ¼ 0–150 W, and a gas flow rate of 40 sccm in total; the reactor gas pressure was varied from P0 ¼ 5 to 20 mTorr (0.665–2.66 Pa). The etched depth was measured by stylus profilometry; in some cases, the film thickness of dielectrics was measured by optical interferometry, and the film thickness of metals was obtained from the sheet resistance measured by the four-point probe method. The etch rates were calculated by dividing the etched depth or thickness by the etching time of typically 5 min. Moreover, the chemical composition on wafer surfaces was examined with X-ray photoelectron spectroscopy (XPS), and the etched profile was with scanning electron microscopy (SEM). In addition, optical emission spectroscopy (OES) and Langmuir probe measurement were employed to characterize the plasma around the wafer position during processing. Samples for etching were 50-nm-thick HfO2 films prepared by CVD on Si substrates without after annealing, and 200-nm-thick Pt and TaN films prepared by reactive sputtering on thermally oxidized Si substrates. Separate Si and thermal SiO2 substrates were employed for reference of etching. To examine the etched profiles, patterned samples were also employed with a 200-nm-thick hard mask (SiO2 mask) pattern feature of lines and spaces, where the nominal pattern width was in the range 0.410 mm. It is
ARTICLE IN PRESS K. Nakamura et al. / Vacuum 80 (2006) 761–767
noted here that the Pt and TaN samples had 10-nm-thick Ti films deposited as an adhesion layer between the metal and SiO2. The samples were cleaved into 2 cm2 pieces and attached on a 6-in.-diam Si wafer, which was then mechanically clamped into place on the wafer stage water-cooled at 20 1C; backside helium was injected for increased wafer thermal contact. No significant effects of reaction products from the Si wafer underlying would be considered on the etching characteristics concerned, because no significant peaks of Si were identified in the XPS spectrum of etched HfO2 sample surfaces.
surfaces of HfO2, Si, and SiO2, and that at higher P0X15 mTorr, the etching of HfO2 did not occur. Langmuir probe measurements indicated that at P0 ¼ 10 mTorr with PMW ¼ 600 W, the difference between the plasma and floating potentials was typically V p V f 10 V, increasing with decreasing P0 as also shown in Fig. 2(a). The significantly reduced etch rates at high pressures were attributable to the reduced plasma production efficiency threat, which resulted in the reduced plasma density and thus in the reduced concentration of reactive neutrals BCl and Cl, as inferred from Langmuir probe and OES measurements. In addition, as the microwave power was increased from PMW ¼ 400 to 600 W at P0 ¼ 10 mTorr, the HfO2 etch rate tended to increase, where the etch rates of Si and SiO2 remained almost unchanged. Fig. 2(c) shows the etch rates of HfO2, Si, and SiO2 in BCl3 without rf biasing, as a function of pressure P0 at PMW ¼ 600 W, recalculated from the samples of Fig. 2(a) after removal of the deposited films on surfaces by using the solvent of alcohol. It should be noted that at low pressures P0p6 mTorr, the etching was found to occur on HfO2 surfaces, while the etching was insignificant on Si and SiO2 surfaces. This etch rate behavior of HfO2 was similar to that of the optical emission intensity of BCl bands
3. Results and discussion 3.1. Etching of high-k dielectrics Figs. 2(a) and (b) show the etch rates of HfO2, Si, and SiO2 in BCl3 without rf biasing (Prf ¼ 0), as a function of pressure P0 at PMW ¼ 600 W and as a function of microwave power PMW at P0 ¼ 10 mTorr, respectively. The etching of HfO2 films occurred at pressures between P0 ¼ 8 and 12 mTorr. The etch rate of HfO2 was about 5 nm/min in maximum at P0 ¼ 10 mTorr, where the etch selectivity was 410 over Si and SiO2. It should be noted that at lower P0p6 mTorr, the deposition occurred on all 10
8 Microwave power 600 W RF 0 W HfO2 Etch
0
Depo -5 Vp - Vf
SiO2 Si
10
Vp -Vf (V)
20
-10
Gas Pressure 10 mTorr RF 0 W
6 Etch Rate (nm/min)
5 Etch Rate (nm/min)
763
HfO2 4
2 Etch
Si SiO2
0 Depo
BCl3 plasma
BCl3 plasma -15
0
5
(a)
10 15 Pressure (mTorr)
20
-2 350
0
(b)
10
HfO2
Depo
HfO2 SiO2 Si
5
-100
SiO2
-200 Si -300
0
0
5
Etch
0 Etch Depth (nm)
Etch Rate (nm/min)
Microwave 600 W RF 0 W BCl3
Microwave 600 W RF 0 W BCl3 5 mTorr
-400
(c)
600
100
15
-5
500 550 400 450 Microwave Power (W)
10 15 Pressure (mTorr)
-500
20
(d)
0
1
2
3 4 Time (min)
5
6
Fig. 2. Etch rates of HfO2, Si, and SiO2 in ECR BCl3 plasmas without rf biasing (Prf ¼ 0), (a) as a function of pressure P0 at PMW ¼ 600 W and (b) as a function of microwave power PMW at P0 ¼ 10 mTorr. The gas flow rate was 40 sccm. Also shown are (c) the etch rates recalculated from the samples of (a) after removal of the deposited films on surfaces, and (d) the etched depth or thickness measured as a function of time at P0 ¼ 5 mTorr and PMW ¼ 600 W.
ARTICLE IN PRESS 764
K. Nakamura et al. / Vacuum 80 (2006) 761–767
(272 nm), implying that BCl is a primary etchant for HfO2. Fig. 2(d) shows the etched depth or thickness of HfO2, Si, and SiO2, measured as a function of time in BCl3 without rf biasing at P0 ¼ 5 mTorr and PMW ¼ 600 W. It should be noted that after an induction period of o0.5 min when no etching occurred on all sample surfaces, the etching of HfO2 was found to occur during approximately o0.5 min, and then the deposition started to occur on HfO2; in contrast, on Si and SiO2 surfaces, only the deposition occurred after the induction period. The instantaneous etch rate of HfO2 prior to the deposition was an order of magnitude higher than the time-averaged etch rate as in Fig. 2(c), implying that the HfO2 etch rate would be significantly enhanced if the deposition can be suppressed under some operating conditions. In addition, the induction period was not observed at higher P0 ¼ 10 mTorr, where the etching of HfO2 occurred at an almost constant rate after the start of etching (tX0); thus, the induction period observed at P0 ¼ 5 mTorr may be attributed in part to the competition between the significant deposition and the etching or removal of deposited materials on surfaces. Fig. 3 shows SEM micrographs of the HfO2, Si, and SiO2 surfaces etched in (or exposed to) ECR BCl3 plasmas during 5 min, without rf biasing at PMW ¼ 600 W and P0 ¼ 5 and 10 mTorr. At P0 ¼ 5 mTorr, all surfaces of
HfO2, Si, and SiO2 were found to be rough, or the surface morphologies were poor, probably owing to the films or materials deposited. In contrast, all the surfaces were smooth at P0 ¼ 10 mTorr, on which the etching occurred. Fig. 4(a) shows the XPS spectrum of Hf 4f7/2 and Hf 4f5/2, obtained from HfO2 sample surfaces under three different conditions: (i) HfO2 surfaces before plasma exposure; and HfO2 surfaces exposed to BCl3 plasmas during 5 min without rf biasing at PMW ¼ 600 W and P0 ¼ (ii) 10 mTorr (etched) and (iii) 5 mTorr (deposited). Note that the conditions (ii) and (iii) correspond to those of Fig. 3. Also shown is the spectrum obtained from (iv) Si surfaces exposed to BCl3 plasmas under the same operating conditions of (iii). The similar spectrum of Hf 4f was observed from (i) virgin and (ii) etched HfO2 surfaces, while the spectrum intensity was relatively small from (iii) HfO2 sample surfaces obtained under deposition conditions, implying that the surfaces were covered with thin films deposited or material layers containing no Hf species. Fig. 4(b) shows the XPS spectrum of Cl 2p and B 1s, obtained from HfO2, and Si surfaces under the same conditions of Fig. 4(a). No significant spectrum was observed from (i) virgin and (ii) etched HfO2 surfaces. In contrast, the similar spectrum of Cl 2p and B 1s was observed from (iii) HfO2 and (iv) Si sample surfaces obtained under deposition conditions, implying some thick layer of BClx species or films deposited on the surfaces of interest. Moreover, Fig. 4(c) shows the XPS spectrum of O 1s from HfO2 and Si surfaces under the same conditions of Figs. 4(a) and (b). The similar O 1s spectrum was observed from (i) virgin and (ii) etched HfO2 surfaces, being attributable probably to Hf–O bonding of HfO2. In contrast, the O 1s spectrum on (iii) HfO2 and (iv) Si sample surfaces under deposition conditions exhibited some chemical shift attributable possibly to B–O bonding, which may be caused by the exposure of samples to atmosphere containing O2 after the experiment. 3.2. Etching of metals
Fig. 3. SEM micrographs of the HfO2, Si, and SiO2 surfaces etched in (or exposed to) ECR BCl3 plasmas without rf biasing during 5 min at PMW ¼ 600 W and P0 ¼ 5 and 10 mTorr.
Figs. 5(a) and (b) show the etch rates of Pt, TaN, HfO2, SiO2, and Si as a function of Cl2 concentration ratio in Ar/ Cl2 and as a function of O2 concentration ratio in Ar/O2, respectively, at a microwave power of PMW ¼ 600 W, an rf bias power of Prf ¼ 150 W, and a pressure of P0 ¼ 10 mTorr. Under these conditions, the peak-to-peak amplitude of the rf bias voltage was measured to be in the range Vpp ¼ 440–530 V, decreasing with increasing the concentration ratio. The etch rate of Pt in Ar/Cl2 was about several tens nm/min, being almost independent of Cl2 concentration. The selectivity of Pt was in the range 0.6–2.0 over HfO2 and in the range 1.0–2.0 over SiO2, indicating that highly selective Pt etching over HfO2 and SiO2 is difficult in Ar/Cl2. Note that under these conditions, the TaN etch rate exceeded 4300 nm/min when adding a small amount of Cl2, probably owing to the formation of highly volatile products TaClx.
ARTICLE IN PRESS K. Nakamura et al. / Vacuum 80 (2006) 761–767 (i) HfO2 (before plasma exposure) (ii) HfO2 (BCl3, 10 mTorr, etched) (iii) HfO2 (BCl3, 5 mTorr, deposited) (iv) Si (BCl3, 5 mTorr, deposited)
765
350 Microwave 600 W, RF 150 W TaN
300
(x 0.5)
16000 17.93 eV Hf 4f7/2 (HfO2)
14000
15.98 eV Hf 4f5/2 (Hf)
12000 XPS Intensity
Etch Rate (nm/min)
16.7 eV Hf 4f7/2 (HfO2)
10000 8000 (i)
6000
(ii)
0
Si 200 HfO2 100
50
4000 2000
250
Pt SiO2
(iii)
0
(iv )
25
20 15 Binding Energy (eV)
(a)
10
0
25
(a)
Ar/Cl2 10 mTorr
50 75 [Cl2]/([Cl2]+[Ar]) (%)
100
80 Microwave 600 W, RF 150 W
8000
Ar/O2 10 mTorr Cl 2p 199.4 eV
60 Etch Rate (nm/min)
6000 XPS Intensity
(i ) B 1s 190.0 eV
(ii)
4000
HfO2
40
SiO2 Si 20
(iii)
2000
Pt
TaN (i v )
0
0 210
200 190 Binding Energy (eV)
(b)
180
0 (b)
O 1s (HfO2) 530.4 eV
25000
XPS Intensity
15000 (i ) 10000
(ii) (iii )
5000 (i v) 0 545 (c)
540
535 530 525 Binding Energy (eV)
20 30 [O2]/([O2]+[Ar]) (%)
40
50
Fig. 5. Etch rates of Pt, TaN, HfO2, SiO2, and Si (a) as a function of Cl2 concentration ratio in Ar/Cl2 and (b) as a function of O2 concentration ratio in Ar/O2, at a microwave power PMW ¼ 600 W, an rf bias power Prf ¼ 150 W, and a pressure P0 ¼ 10 mTorr. The gas flow rate was 40 sccm in total.
O 1s (B2O3, SiO2) 533.0 eV
20000
10
520
515
Fig. 4. XPS spectrum of (a) Hf 4f7/2 and Hf 4f5/2, (b) Cl 2p and B 1s, and (c) O 1s, obtained from HfO2 and Si sample surfaces under three different conditions: (i) HfO2 surfaces before plasma exposure; and HfO2 surfaces exposed to BCl3 plasmas during 5 min without rf biasing at PMW ¼ 600 W and P0 ¼ (ii) 10 mTorr (etched) and (iii) 5 mTorr (deposited). Also shown is the spectrum obtained from (iv) Si surfaces exposed to BCl3 plasmas under the same operating conditions of (iii).
In contrast, by adding a small amount of O2 to Ar, the etch rates of TaN, HfO2, SiO2, and Si significantly decreased, while the etch rate of Pt decreased slightly, resulting in a high etch selectivity of48 over HfO2 and SiO2 with 420% O2 added. These results of the etching behavior of Pt is similar to previous studies for the application of high-k metal oxides to capacitor dielectrics in memory devices [14–16]; in practice, the etching of Pt relies primarily on physical sputtering through energetic ion bombardment on surfaces, because Pt is a difficult material for etching, owing to its non-volatile halogen compounds. Fig. 6(a) shows the etch rates of Pt, TaN, HfO2, SiO2, and Si as a function of Cl2 concentration ratio in Ar/Cl2, at
ARTICLE IN PRESS K. Nakamura et al. / Vacuum 80 (2006) 761–767
766
250 Microwave 600 W RF 30 W
200 Etch Rate (nm/min)
Ar/Cl2 10 mTorr 150
TaN
100 Si 50 HfO2 SiO2
0
Pt 0
25
(a)
50 75 [Cl2]/([Cl2]+[Ar]) (%)
100
140 Microwave power 600 W
120
Ar/Cl2=20/20 sccm 10 mTorr
100 Etch Rate (nm/min)
TaN
a high etch selectivity of48 was achieved over HfO2 and SiO2. Fig. 6(b) shows the etch rates of TaN and Pt as a function of rf bias power Prf at 50% Cl2 added in Ar/Cl2, otherwise under the same operating condition of Fig. 6(a). The threshold of Prf or the incident ion energy required was found to be significantly lower for TaN than for Pt; in practice, the etching of TaN may occur at Prf ¼ 0 or without rf biasing after some breakthrough at the beginning of etching. These results imply that the etching of TaN relies primarily on ion-assisted etching, with the chemical nature being high as compared even to Si. Fig. 7 shows SEM micrographs of a nominally 0.4-mm line feature of Pt and TaN, approximately just etched at 20% O2 in Ar/O2 with Prf ¼ 150 W for Pt and at 75% Cl2 in Ar/Cl2 with Prf ¼ 30 W for TaN; otherwise, the operating conditions were the same for both with PMW ¼ 600 W and P0 ¼ 10 mTorr. The etched profile of Pt was outwardly tapered, owing to the redeposition of etch or sputter products on feature sidewalls [14–16]. In contrast, the etched profile of TaN was found to be almost anisotropic, probably owing to the ion-enhanced etching that occurred.
80 60 40 Pt
20 0
0 (b)
10
20 30 40 RF Bias Power (W)
50
60
Fig. 6. (a) Etch rates of Pt, TaN, HfO2, SiO2, and Si as a function of Cl2 concentration ratio in Ar/Cl2, at PMW ¼ 600 W, Prf ¼ 30 W, and P0 ¼ 10 mTorr. The gas flow rate was 40 sccm in total. Also shown are (b) the etch rates of Pt and TaN as a function of Prf at 50% Cl2 added in Ar/Cl2, otherwise under the same operating condition of (a).
a microwave power of PMW ¼ 600 W, an rf bias power of Prf ¼ 30 W, and a pressure of P0 ¼ 10 mTorr. Under these conditions, the peak-to-peak amplitude of the rf bias voltage was measured to be in the range Vpp ¼ 100–160 V, decreasing with increasing the concentration ratio. Note that under these low rf bias power conditions, the etching of Pt did not occur, owing to lowered energy of the ion bombardment; moreover, without rf biasing (Prf ¼ 0), the etching did not occur for all the films of interest. The etch rate of TaN increased significantly with increasing Cl2 concentration, being about 200 nm/min in pure Cl2, where
Fig. 7. SEM micrographs of a 0.4-mm line feature of Pt and TaN, approximately just etched at 20% O2 in Ar/O2 with Prf ¼ 150 W for Pt and at 75% Cl2 in Ar/Cl2 with Prf ¼ 30 W for TaN. Otherwise, the operating conditions were the same for both with PMW ¼ 600 W and P0 ¼ 10 mTorr.
ARTICLE IN PRESS K. Nakamura et al. / Vacuum 80 (2006) 761–767
4. Conclusions Etching characteristics of high-k dielectric materials (HfO2) and metal electrode materials (Pt, TaN) have been studied using high-density ECR chlorine-containing plasmas. Attention was focused on etch chemistries and plasma conditions to achieve a high etch selectivity of b1 for HfO2 over the underlying Si and SiO2; regarding Pt and TaN, the emphasis was placed on the etch anisotropy and selectivity over HfO2 and SiO2 (over the underlying high-k and the overlying mask materials). The etching of HfO2 was performed in BCl3 without rf biasing, giving an etch rate of about 5 nm/min with a high selectivity of 410 over Si and SiO2. The etching of Pt and TaN was performed in Ar/O2 with high rf biasing and in Ar/Cl2 with low rf biasing, respectively, giving a Pt etch rate of about several tens nm/min and a TaN etch rate of about 200 nm/min with a high selectivity of 48 over HfO2 and SiO2. The etched profiles were outwardly tapered for Pt, owing to the redeposition of etch or sputter products on feature sidewalls, while the TaN profiles were almost anisotropic, probably owing to the ion-enhanced etching that occurred. Further investigations are now in progress for the etching of stacks of metal/high-k, along with a better understanding of the physics and chemistry underlying the phenomena observed. Acknowledgments This work was supported by NEDO/MIRAI project and by Taiyo Nippon Sanso Corporation. The work was also
767
supported in part by a Grant-in-Aid for Scientific Research from the Japan Society for the Promotion of Science. References [1] Hirose M. Oyo Buturi 2002;71:1091–101 [in Japanese]. [2] Toriumi A, Horikawa T, Nabatame T. Nikkei Microelectron 2002; 163–170 [in Japanese]. [3] Pelhos K, Donnelly VM, Kornbilt A, Green ML, Van Dover RB, Manchanda L, et al. J Vac Sci Technol A 2001;19:1361–6. [4] Sha L, Cho BO, Chang JP. J Vac Sci Technol A 2002;20:1525–31. [5] Norasetthekul S, Park PY, Baik KH, Lee KP, Shin JH, Jeong BS, et al. Appl Surf Sci 2002;187:75–81. [6] Sha L, Chang JP. J Vac Sci Technol A 2003;21:1915–22. [7] Sha L, Puthenkovilakan R, Lin YS, Chang JP. J Vac Sci Technol B 2003;21:2420–7. [8] Sha L, Chang JP. J Vac Sci Technol A 2004;22:88–95. [9] Maeda T, Ito H, Mitsuhashi R, Horiuchi A, Kawahara T, Muto A, et al. Jpn J Appl Phys 2004;43:1864–8. [10] Chen J, Yoo WJ, Tan ZY, Wang Y, Chan DSH. J Vac Sci Technol A 2004;22:1552–8. [11] Kota GP, Ramalingam S, Lee S, Coenegrachts B, Lee C. In: Proceedings of the fourth international symposium on dry process. Tokyo: IEEJ; 2004. p. 133–8. [12] Takahashi K, Ono K. In: Proceedings of the fourth international symposium on dry process. Tokyo: IEEJ; 2004. paper 369–74 (also submitted to J Vac Sci Technol). [13] Nakamura K, Kitagawa T, Osari K, Takahashi K, Ono K. Presented at the JSAP spring meeting. Tokyo: JSAP; 2005. p. 30a-G-5 [in Japanese]. [14] Shibano T, Nakamura K, Oomori T. J Vac Sci Technol A 1998;16:502–8. [15] Ono K, Horikawa T, Shibano T, Mikami N, Kuroiwa T, Kawahara T, et al. Technical digests of the international electron devices meeting. Piscataway, NJ: IEEE; 1998. p. 803–6. [16] Shibano K, Nakamura K, Takenaga T, Ono K. J Vac Sci Technol A 1999;17:799–804.
Vacuum 80 (2006) 761–767 www.elsevier.com/locate/vacuum
Plasma etching of high-k and metal gate materials Keisuke Nakamura, Tomohiro Kitagawa1, Kazushi Osari, Kazuo Takahashi, Kouichi Ono Department of Aeronautics and Astronautics, Graduate School of Engineering, Kyoto University, Yoshida-Honmachi, Sakyo-ku, Kyoto 606-8501, Japan
Abstract Etching characteristics of high-k dielectric materials (HfO2) and metal electrode materials (Pt, TaN) have been studied in high-density chlorine-containing plasmas at pressures around 10 mTorr. The etching of HfO2 was performed in BCl3 without rf biasing, giving an etch rate of about 5 nm/min with a high selectivity of 410 over Si and SiO2. The etching of Pt and TaN was performed in Ar/O2 with high rf biasing and in Ar/Cl2 with low rf biasing, respectively, giving a Pt etch rate of about several tens nm/min and a TaN etch rate of about 200 nm/min with a high selectivity of 48 over HfO2 and SiO2. The etched profiles were outwardly tapered for Pt, owing to the redeposition of etch or sputter products on feature sidewalls, while the TaN profiles were almost anisotropic, probably owing to the ionenhanced etching that occurred. r 2005 Elsevier Ltd. All rights reserved. Keywords: Plasma etching; Chlorine-containing plasma; High-k dielectrics; Metal electrode materials; HfO2; Pt; TaN
1. Introduction As integrated circuit device dimensions continue to be scaled down, the gate width of advanced microelectronic devices is projected to be scaled down to much less than 100 nm, and the thickness of gate oxides is also reduced down to 2 nm or less for conventional SiO2. Correspondingly, the technological challenge continues for growing ultrathin SiO2 films of high quality, to maintain the gate capacitance without increasing the gate leakage current and reducing the oxide reliability; however, the ultimate solution would rely on high dielectric constant (k) materials. Recent efforts have been made to replace SiO2 with silicon-oxynitrides of slightly high k, and nowadays, new high-k (420) dielectrics or metal oxides such as Al2O3, HfO2, ZrO2, and their silicates and aluminates are being developed to replace SiO2 [1,2]. In practice, the metal oxides give the required specific gate capacitance at a considerably larger thickness as compared to SiO2. In integrating these materials into device fabrication, an understanding of the etching characteristics of high-k Corresponding author. Tel.: +81 75 753 5793; fax: +81 75 753 5980.
E-mail address: [email protected] (K. Ono). Present address: NTT communications Corporation, Hitotsubashi, Chiyoda-ku, Tokyo 100-8128, Japan. 1
0042-207X/$ - see front matter r 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.vacuum.2005.11.017
materials is strongly required for their removal prior to forming the source and drain contacts. Moreover, for gate stacks with high-k dielectrics, gate electrodes of conventional polycrystalline silicon (poly-Si) tends to cause some problem of the depletion layer present in doped poly-Si gate materials. Thus, for lower equivalent oxide thicknesses, recent efforts have also been made to replace poly-Si gate electrodes with metal gates of WN, TiN, TaN, Ru/RuO2, Pt, and/or Ir [1,2]. For gate etch processes, a precise control of the profile and critical dimension of gate electrodes and a high selectivity to gate oxides have historically been the two most important issues to be addressed. Plasma etching of high-k oxides and metal electrodes has been studied in the application of ferroelectric materials, buffer layers, and capacitor dielectrics to memory devices. However, only a few studies have recently been concerned with the plasma etching of high-k dielectrics [3–13] and metal electrodes [11,13] for the application to high-k gate stacks. Pelhos et al. studied the etching of high-k gate dielectric Zr1xAlxOy thin films in helical resonator plasmas with Cl2/BCl3 [3]. Sha et al. investigated the etching of ZrO2 in electron–cyclotron–resonance (ECR) plasmas with Cl2/Ar [4] and BCl3/Cl2 [6], and also the etching of HfO2 in ECR BCl3/Cl2 plasmas [7,8]. Norasetthekul et al. reported on the etching of HfO2 in inductively
ARTICLE IN PRESS 762
K. Nakamura et al. / Vacuum 80 (2006) 761–767
coupled plasmas (ICP) with Cl2/Ar, SF6/Ar, and CH4/H2/ Ar [5]. Moreover, Maeda et al. reported on the HfO2 etching in ICP CF4 and Cl2/HBr/O2 plasmas [9]. Chen et al. investigated the etching of HfO-based high-k films in ICP plasmas with Cl2/HBr and CF4/CHF3 [10]. Emphasis of most of these studies has been placed on etch chemistries giving an etch selectivity of 41 over the underlying Si [3–9] and SiO2 [9]. More recently, Kota et al. have reported on the etching of metal/high-k stacks (TiN/TaN/HfO2) in ICP plasmas with BCl3-based chemistries [11]. It is further noted that the etching of high-k dielectrics with low ion energies and/or less ions is also indispensable in mass production, for chamber cleaning of the chemical vapor deposition (CVD) and atomic layer deposition (ALD) apparatuses to prepare high-k thin films. We have investigated the high-density plasma etching of HfO2 with attention being focused on etch chemistries and plasma conditions to achieve a much higher selectivity of b1 over Si and SiO2. We found the highly selective etching of HfO2 over Si in ICP fluorocarbon plasmas [12]; the selectivity was 45 in C4F8/Ar plasmas and the selectivity was further increased to b10 in C4F8/Ar/H2 plasmas, where carbon and/or carbonaceous species would work as surface inhibitors on Si, as in the situation of highly selective etching of SiO2 over Si with fluorocarbon plasmas. This paper presents a highly selective etching of HfO2 over Si and SiO2 in ECR chlorine-containing plasmas without rf biasing [13]. The paper also presents the etching of Pt and TaN in ECR chlorine-containing plasmas [13], with emphasis on the etch anisotropy and selectivity of metal electrodes over HfO2 and SiO2 (over the underlying high-k and the overlying mask materials). 2. Experiment Fig. 1 shows the experimental setup. The plasma reactor consisted of a grounded plasma and specimen chamber made of stainless steel. A set of solenoid coils were placed around the former, providing divergent magnetic fields in the specimen chamber. The discharge was established by 2.45-GHz right-hand circularly polarized microwaves of TE11 mode injected through an entrance quartz window at the top of the plasma chamber, and the 875-G ECR resonance was located near the end of the plasma chamber. The specimen chamber, 36 cm in diameter and 40 cm in length, had an end port for pumping and several side ports for diagnostics, containing a 16-cm-diam floating electrode or substrate holder 20 cm downstream (BE200 G) from the ECR resonance region. The electrode was capacitively coupled to a 13.56-MHz rf power supply through an impedance matching network for additional rf biasing. Feedstock gases employed were Cl2, BCl3., O2, and their mixtures with Ar, being introduced into the reactor evacuated to a base pressure o106 Torr, through a set of small holes at side walls of the reactor. The experiments were performed at incident microwave powers of nominally PMW ¼ 400–600 W, rf bias powers of nominally
Microwave (2.45 GHz) Solenoid Coils
Quartz Window
Gas Wafer Stage
OES Langmuir Probe
To Pump
Matching Box RF (13.56 MHz)
Fig. 1. Experimental setup for ECR plasma etching.
Prf ¼ 0–150 W, and a gas flow rate of 40 sccm in total; the reactor gas pressure was varied from P0 ¼ 5 to 20 mTorr (0.665–2.66 Pa). The etched depth was measured by stylus profilometry; in some cases, the film thickness of dielectrics was measured by optical interferometry, and the film thickness of metals was obtained from the sheet resistance measured by the four-point probe method. The etch rates were calculated by dividing the etched depth or thickness by the etching time of typically 5 min. Moreover, the chemical composition on wafer surfaces was examined with X-ray photoelectron spectroscopy (XPS), and the etched profile was with scanning electron microscopy (SEM). In addition, optical emission spectroscopy (OES) and Langmuir probe measurement were employed to characterize the plasma around the wafer position during processing. Samples for etching were 50-nm-thick HfO2 films prepared by CVD on Si substrates without after annealing, and 200-nm-thick Pt and TaN films prepared by reactive sputtering on thermally oxidized Si substrates. Separate Si and thermal SiO2 substrates were employed for reference of etching. To examine the etched profiles, patterned samples were also employed with a 200-nm-thick hard mask (SiO2 mask) pattern feature of lines and spaces, where the nominal pattern width was in the range 0.410 mm. It is
ARTICLE IN PRESS K. Nakamura et al. / Vacuum 80 (2006) 761–767
noted here that the Pt and TaN samples had 10-nm-thick Ti films deposited as an adhesion layer between the metal and SiO2. The samples were cleaved into 2 cm2 pieces and attached on a 6-in.-diam Si wafer, which was then mechanically clamped into place on the wafer stage water-cooled at 20 1C; backside helium was injected for increased wafer thermal contact. No significant effects of reaction products from the Si wafer underlying would be considered on the etching characteristics concerned, because no significant peaks of Si were identified in the XPS spectrum of etched HfO2 sample surfaces.
surfaces of HfO2, Si, and SiO2, and that at higher P0X15 mTorr, the etching of HfO2 did not occur. Langmuir probe measurements indicated that at P0 ¼ 10 mTorr with PMW ¼ 600 W, the difference between the plasma and floating potentials was typically V p V f 10 V, increasing with decreasing P0 as also shown in Fig. 2(a). The significantly reduced etch rates at high pressures were attributable to the reduced plasma production efficiency threat, which resulted in the reduced plasma density and thus in the reduced concentration of reactive neutrals BCl and Cl, as inferred from Langmuir probe and OES measurements. In addition, as the microwave power was increased from PMW ¼ 400 to 600 W at P0 ¼ 10 mTorr, the HfO2 etch rate tended to increase, where the etch rates of Si and SiO2 remained almost unchanged. Fig. 2(c) shows the etch rates of HfO2, Si, and SiO2 in BCl3 without rf biasing, as a function of pressure P0 at PMW ¼ 600 W, recalculated from the samples of Fig. 2(a) after removal of the deposited films on surfaces by using the solvent of alcohol. It should be noted that at low pressures P0p6 mTorr, the etching was found to occur on HfO2 surfaces, while the etching was insignificant on Si and SiO2 surfaces. This etch rate behavior of HfO2 was similar to that of the optical emission intensity of BCl bands
3. Results and discussion 3.1. Etching of high-k dielectrics Figs. 2(a) and (b) show the etch rates of HfO2, Si, and SiO2 in BCl3 without rf biasing (Prf ¼ 0), as a function of pressure P0 at PMW ¼ 600 W and as a function of microwave power PMW at P0 ¼ 10 mTorr, respectively. The etching of HfO2 films occurred at pressures between P0 ¼ 8 and 12 mTorr. The etch rate of HfO2 was about 5 nm/min in maximum at P0 ¼ 10 mTorr, where the etch selectivity was 410 over Si and SiO2. It should be noted that at lower P0p6 mTorr, the deposition occurred on all 10
8 Microwave power 600 W RF 0 W HfO2 Etch
0
Depo -5 Vp - Vf
SiO2 Si
10
Vp -Vf (V)
20
-10
Gas Pressure 10 mTorr RF 0 W
6 Etch Rate (nm/min)
5 Etch Rate (nm/min)
763
HfO2 4
2 Etch
Si SiO2
0 Depo
BCl3 plasma
BCl3 plasma -15
0
5
(a)
10 15 Pressure (mTorr)
20
-2 350
0
(b)
10
HfO2
Depo
HfO2 SiO2 Si
5
-100
SiO2
-200 Si -300
0
0
5
Etch
0 Etch Depth (nm)
Etch Rate (nm/min)
Microwave 600 W RF 0 W BCl3
Microwave 600 W RF 0 W BCl3 5 mTorr
-400
(c)
600
100
15
-5
500 550 400 450 Microwave Power (W)
10 15 Pressure (mTorr)
-500
20
(d)
0
1
2
3 4 Time (min)
5
6
Fig. 2. Etch rates of HfO2, Si, and SiO2 in ECR BCl3 plasmas without rf biasing (Prf ¼ 0), (a) as a function of pressure P0 at PMW ¼ 600 W and (b) as a function of microwave power PMW at P0 ¼ 10 mTorr. The gas flow rate was 40 sccm. Also shown are (c) the etch rates recalculated from the samples of (a) after removal of the deposited films on surfaces, and (d) the etched depth or thickness measured as a function of time at P0 ¼ 5 mTorr and PMW ¼ 600 W.
ARTICLE IN PRESS 764
K. Nakamura et al. / Vacuum 80 (2006) 761–767
(272 nm), implying that BCl is a primary etchant for HfO2. Fig. 2(d) shows the etched depth or thickness of HfO2, Si, and SiO2, measured as a function of time in BCl3 without rf biasing at P0 ¼ 5 mTorr and PMW ¼ 600 W. It should be noted that after an induction period of o0.5 min when no etching occurred on all sample surfaces, the etching of HfO2 was found to occur during approximately o0.5 min, and then the deposition started to occur on HfO2; in contrast, on Si and SiO2 surfaces, only the deposition occurred after the induction period. The instantaneous etch rate of HfO2 prior to the deposition was an order of magnitude higher than the time-averaged etch rate as in Fig. 2(c), implying that the HfO2 etch rate would be significantly enhanced if the deposition can be suppressed under some operating conditions. In addition, the induction period was not observed at higher P0 ¼ 10 mTorr, where the etching of HfO2 occurred at an almost constant rate after the start of etching (tX0); thus, the induction period observed at P0 ¼ 5 mTorr may be attributed in part to the competition between the significant deposition and the etching or removal of deposited materials on surfaces. Fig. 3 shows SEM micrographs of the HfO2, Si, and SiO2 surfaces etched in (or exposed to) ECR BCl3 plasmas during 5 min, without rf biasing at PMW ¼ 600 W and P0 ¼ 5 and 10 mTorr. At P0 ¼ 5 mTorr, all surfaces of
HfO2, Si, and SiO2 were found to be rough, or the surface morphologies were poor, probably owing to the films or materials deposited. In contrast, all the surfaces were smooth at P0 ¼ 10 mTorr, on which the etching occurred. Fig. 4(a) shows the XPS spectrum of Hf 4f7/2 and Hf 4f5/2, obtained from HfO2 sample surfaces under three different conditions: (i) HfO2 surfaces before plasma exposure; and HfO2 surfaces exposed to BCl3 plasmas during 5 min without rf biasing at PMW ¼ 600 W and P0 ¼ (ii) 10 mTorr (etched) and (iii) 5 mTorr (deposited). Note that the conditions (ii) and (iii) correspond to those of Fig. 3. Also shown is the spectrum obtained from (iv) Si surfaces exposed to BCl3 plasmas under the same operating conditions of (iii). The similar spectrum of Hf 4f was observed from (i) virgin and (ii) etched HfO2 surfaces, while the spectrum intensity was relatively small from (iii) HfO2 sample surfaces obtained under deposition conditions, implying that the surfaces were covered with thin films deposited or material layers containing no Hf species. Fig. 4(b) shows the XPS spectrum of Cl 2p and B 1s, obtained from HfO2, and Si surfaces under the same conditions of Fig. 4(a). No significant spectrum was observed from (i) virgin and (ii) etched HfO2 surfaces. In contrast, the similar spectrum of Cl 2p and B 1s was observed from (iii) HfO2 and (iv) Si sample surfaces obtained under deposition conditions, implying some thick layer of BClx species or films deposited on the surfaces of interest. Moreover, Fig. 4(c) shows the XPS spectrum of O 1s from HfO2 and Si surfaces under the same conditions of Figs. 4(a) and (b). The similar O 1s spectrum was observed from (i) virgin and (ii) etched HfO2 surfaces, being attributable probably to Hf–O bonding of HfO2. In contrast, the O 1s spectrum on (iii) HfO2 and (iv) Si sample surfaces under deposition conditions exhibited some chemical shift attributable possibly to B–O bonding, which may be caused by the exposure of samples to atmosphere containing O2 after the experiment. 3.2. Etching of metals
Fig. 3. SEM micrographs of the HfO2, Si, and SiO2 surfaces etched in (or exposed to) ECR BCl3 plasmas without rf biasing during 5 min at PMW ¼ 600 W and P0 ¼ 5 and 10 mTorr.
Figs. 5(a) and (b) show the etch rates of Pt, TaN, HfO2, SiO2, and Si as a function of Cl2 concentration ratio in Ar/ Cl2 and as a function of O2 concentration ratio in Ar/O2, respectively, at a microwave power of PMW ¼ 600 W, an rf bias power of Prf ¼ 150 W, and a pressure of P0 ¼ 10 mTorr. Under these conditions, the peak-to-peak amplitude of the rf bias voltage was measured to be in the range Vpp ¼ 440–530 V, decreasing with increasing the concentration ratio. The etch rate of Pt in Ar/Cl2 was about several tens nm/min, being almost independent of Cl2 concentration. The selectivity of Pt was in the range 0.6–2.0 over HfO2 and in the range 1.0–2.0 over SiO2, indicating that highly selective Pt etching over HfO2 and SiO2 is difficult in Ar/Cl2. Note that under these conditions, the TaN etch rate exceeded 4300 nm/min when adding a small amount of Cl2, probably owing to the formation of highly volatile products TaClx.
ARTICLE IN PRESS K. Nakamura et al. / Vacuum 80 (2006) 761–767 (i) HfO2 (before plasma exposure) (ii) HfO2 (BCl3, 10 mTorr, etched) (iii) HfO2 (BCl3, 5 mTorr, deposited) (iv) Si (BCl3, 5 mTorr, deposited)
765
350 Microwave 600 W, RF 150 W TaN
300
(x 0.5)
16000 17.93 eV Hf 4f7/2 (HfO2)
14000
15.98 eV Hf 4f5/2 (Hf)
12000 XPS Intensity
Etch Rate (nm/min)
16.7 eV Hf 4f7/2 (HfO2)
10000 8000 (i)
6000
(ii)
0
Si 200 HfO2 100
50
4000 2000
250
Pt SiO2
(iii)
0
(iv )
25
20 15 Binding Energy (eV)
(a)
10
0
25
(a)
Ar/Cl2 10 mTorr
50 75 [Cl2]/([Cl2]+[Ar]) (%)
100
80 Microwave 600 W, RF 150 W
8000
Ar/O2 10 mTorr Cl 2p 199.4 eV
60 Etch Rate (nm/min)
6000 XPS Intensity
(i ) B 1s 190.0 eV
(ii)
4000
HfO2
40
SiO2 Si 20
(iii)
2000
Pt
TaN (i v )
0
0 210
200 190 Binding Energy (eV)
(b)
180
0 (b)
O 1s (HfO2) 530.4 eV
25000
XPS Intensity
15000 (i ) 10000
(ii) (iii )
5000 (i v) 0 545 (c)
540
535 530 525 Binding Energy (eV)
20 30 [O2]/([O2]+[Ar]) (%)
40
50
Fig. 5. Etch rates of Pt, TaN, HfO2, SiO2, and Si (a) as a function of Cl2 concentration ratio in Ar/Cl2 and (b) as a function of O2 concentration ratio in Ar/O2, at a microwave power PMW ¼ 600 W, an rf bias power Prf ¼ 150 W, and a pressure P0 ¼ 10 mTorr. The gas flow rate was 40 sccm in total.
O 1s (B2O3, SiO2) 533.0 eV
20000
10
520
515
Fig. 4. XPS spectrum of (a) Hf 4f7/2 and Hf 4f5/2, (b) Cl 2p and B 1s, and (c) O 1s, obtained from HfO2 and Si sample surfaces under three different conditions: (i) HfO2 surfaces before plasma exposure; and HfO2 surfaces exposed to BCl3 plasmas during 5 min without rf biasing at PMW ¼ 600 W and P0 ¼ (ii) 10 mTorr (etched) and (iii) 5 mTorr (deposited). Also shown is the spectrum obtained from (iv) Si surfaces exposed to BCl3 plasmas under the same operating conditions of (iii).
In contrast, by adding a small amount of O2 to Ar, the etch rates of TaN, HfO2, SiO2, and Si significantly decreased, while the etch rate of Pt decreased slightly, resulting in a high etch selectivity of48 over HfO2 and SiO2 with 420% O2 added. These results of the etching behavior of Pt is similar to previous studies for the application of high-k metal oxides to capacitor dielectrics in memory devices [14–16]; in practice, the etching of Pt relies primarily on physical sputtering through energetic ion bombardment on surfaces, because Pt is a difficult material for etching, owing to its non-volatile halogen compounds. Fig. 6(a) shows the etch rates of Pt, TaN, HfO2, SiO2, and Si as a function of Cl2 concentration ratio in Ar/Cl2, at
ARTICLE IN PRESS K. Nakamura et al. / Vacuum 80 (2006) 761–767
766
250 Microwave 600 W RF 30 W
200 Etch Rate (nm/min)
Ar/Cl2 10 mTorr 150
TaN
100 Si 50 HfO2 SiO2
0
Pt 0
25
(a)
50 75 [Cl2]/([Cl2]+[Ar]) (%)
100
140 Microwave power 600 W
120
Ar/Cl2=20/20 sccm 10 mTorr
100 Etch Rate (nm/min)
TaN
a high etch selectivity of48 was achieved over HfO2 and SiO2. Fig. 6(b) shows the etch rates of TaN and Pt as a function of rf bias power Prf at 50% Cl2 added in Ar/Cl2, otherwise under the same operating condition of Fig. 6(a). The threshold of Prf or the incident ion energy required was found to be significantly lower for TaN than for Pt; in practice, the etching of TaN may occur at Prf ¼ 0 or without rf biasing after some breakthrough at the beginning of etching. These results imply that the etching of TaN relies primarily on ion-assisted etching, with the chemical nature being high as compared even to Si. Fig. 7 shows SEM micrographs of a nominally 0.4-mm line feature of Pt and TaN, approximately just etched at 20% O2 in Ar/O2 with Prf ¼ 150 W for Pt and at 75% Cl2 in Ar/Cl2 with Prf ¼ 30 W for TaN; otherwise, the operating conditions were the same for both with PMW ¼ 600 W and P0 ¼ 10 mTorr. The etched profile of Pt was outwardly tapered, owing to the redeposition of etch or sputter products on feature sidewalls [14–16]. In contrast, the etched profile of TaN was found to be almost anisotropic, probably owing to the ion-enhanced etching that occurred.
80 60 40 Pt
20 0
0 (b)
10
20 30 40 RF Bias Power (W)
50
60
Fig. 6. (a) Etch rates of Pt, TaN, HfO2, SiO2, and Si as a function of Cl2 concentration ratio in Ar/Cl2, at PMW ¼ 600 W, Prf ¼ 30 W, and P0 ¼ 10 mTorr. The gas flow rate was 40 sccm in total. Also shown are (b) the etch rates of Pt and TaN as a function of Prf at 50% Cl2 added in Ar/Cl2, otherwise under the same operating condition of (a).
a microwave power of PMW ¼ 600 W, an rf bias power of Prf ¼ 30 W, and a pressure of P0 ¼ 10 mTorr. Under these conditions, the peak-to-peak amplitude of the rf bias voltage was measured to be in the range Vpp ¼ 100–160 V, decreasing with increasing the concentration ratio. Note that under these low rf bias power conditions, the etching of Pt did not occur, owing to lowered energy of the ion bombardment; moreover, without rf biasing (Prf ¼ 0), the etching did not occur for all the films of interest. The etch rate of TaN increased significantly with increasing Cl2 concentration, being about 200 nm/min in pure Cl2, where
Fig. 7. SEM micrographs of a 0.4-mm line feature of Pt and TaN, approximately just etched at 20% O2 in Ar/O2 with Prf ¼ 150 W for Pt and at 75% Cl2 in Ar/Cl2 with Prf ¼ 30 W for TaN. Otherwise, the operating conditions were the same for both with PMW ¼ 600 W and P0 ¼ 10 mTorr.
ARTICLE IN PRESS K. Nakamura et al. / Vacuum 80 (2006) 761–767
4. Conclusions Etching characteristics of high-k dielectric materials (HfO2) and metal electrode materials (Pt, TaN) have been studied using high-density ECR chlorine-containing plasmas. Attention was focused on etch chemistries and plasma conditions to achieve a high etch selectivity of b1 for HfO2 over the underlying Si and SiO2; regarding Pt and TaN, the emphasis was placed on the etch anisotropy and selectivity over HfO2 and SiO2 (over the underlying high-k and the overlying mask materials). The etching of HfO2 was performed in BCl3 without rf biasing, giving an etch rate of about 5 nm/min with a high selectivity of 410 over Si and SiO2. The etching of Pt and TaN was performed in Ar/O2 with high rf biasing and in Ar/Cl2 with low rf biasing, respectively, giving a Pt etch rate of about several tens nm/min and a TaN etch rate of about 200 nm/min with a high selectivity of 48 over HfO2 and SiO2. The etched profiles were outwardly tapered for Pt, owing to the redeposition of etch or sputter products on feature sidewalls, while the TaN profiles were almost anisotropic, probably owing to the ion-enhanced etching that occurred. Further investigations are now in progress for the etching of stacks of metal/high-k, along with a better understanding of the physics and chemistry underlying the phenomena observed. Acknowledgments This work was supported by NEDO/MIRAI project and by Taiyo Nippon Sanso Corporation. The work was also
767
supported in part by a Grant-in-Aid for Scientific Research from the Japan Society for the Promotion of Science. References [1] Hirose M. Oyo Buturi 2002;71:1091–101 [in Japanese]. [2] Toriumi A, Horikawa T, Nabatame T. Nikkei Microelectron 2002; 163–170 [in Japanese]. [3] Pelhos K, Donnelly VM, Kornbilt A, Green ML, Van Dover RB, Manchanda L, et al. J Vac Sci Technol A 2001;19:1361–6. [4] Sha L, Cho BO, Chang JP. J Vac Sci Technol A 2002;20:1525–31. [5] Norasetthekul S, Park PY, Baik KH, Lee KP, Shin JH, Jeong BS, et al. Appl Surf Sci 2002;187:75–81. [6] Sha L, Chang JP. J Vac Sci Technol A 2003;21:1915–22. [7] Sha L, Puthenkovilakan R, Lin YS, Chang JP. J Vac Sci Technol B 2003;21:2420–7. [8] Sha L, Chang JP. J Vac Sci Technol A 2004;22:88–95. [9] Maeda T, Ito H, Mitsuhashi R, Horiuchi A, Kawahara T, Muto A, et al. Jpn J Appl Phys 2004;43:1864–8. [10] Chen J, Yoo WJ, Tan ZY, Wang Y, Chan DSH. J Vac Sci Technol A 2004;22:1552–8. [11] Kota GP, Ramalingam S, Lee S, Coenegrachts B, Lee C. In: Proceedings of the fourth international symposium on dry process. Tokyo: IEEJ; 2004. p. 133–8. [12] Takahashi K, Ono K. In: Proceedings of the fourth international symposium on dry process. Tokyo: IEEJ; 2004. paper 369–74 (also submitted to J Vac Sci Technol). [13] Nakamura K, Kitagawa T, Osari K, Takahashi K, Ono K. Presented at the JSAP spring meeting. Tokyo: JSAP; 2005. p. 30a-G-5 [in Japanese]. [14] Shibano T, Nakamura K, Oomori T. J Vac Sci Technol A 1998;16:502–8. [15] Ono K, Horikawa T, Shibano T, Mikami N, Kuroiwa T, Kawahara T, et al. Technical digests of the international electron devices meeting. Piscataway, NJ: IEEE; 1998. p. 803–6. [16] Shibano K, Nakamura K, Takenaga T, Ono K. J Vac Sci Technol A 1999;17:799–804.