Study of thin hafnium oxides deposited by atomic layer deposition

Study of thin hafnium oxides deposited by atomic layer deposition

Nuclear Instruments and Methods in Physics Research B 219–220 (2004) 856–861 www.elsevier.com/locate/nimb Study of thin hafnium oxides deposited by a...

365KB Sizes 1 Downloads 14 Views

Nuclear Instruments and Methods in Physics Research B 219–220 (2004) 856–861 www.elsevier.com/locate/nimb

Study of thin hafnium oxides deposited by atomic layer deposition J.-J. Ganem a

a,*

, I. Trimaille a, I.C. Vickridge a, D. Blin b, F. Martin

c

Groupe de Physique des Solides, Universite Pierre et Marie Curie, Tour 23, 2, Pl. Jussieu, 75251 Paris cedex 05, France b ASM France, 1025 rue Henri Becquerel, 34036 Montpellier cedex 1, France c CEA-LETI-17, ave des Martyrs, 38054 Grenoble cedex 09, France

Abstract We have deposited thin films (3.5, 7.5 and 22 nm) by atomic layer deposition (ALD) using HfCl4 and H2 O precursors at 350 C. Growth, thermal annealing and thermal reoxidation of the thin hafnium oxide layers under controlled ultra-dry oxygen atmosphere were studied using ion beam techniques and isotopic tracing experiments. Secondary ion mass spectroscopy (SIMS) profiling shows that the composition of deposited films is homogeneous with depth and over a large area. RBS and NRA show that the films are under-stoichiometric in oxygen and contain trace chlorine contamination, more pronounced at the film–substrate interface. After oxidation for 20 min in 100 mbar O2 enriched to 99.9% in 18 O at 425 C, nuclear resonance depth-profiling using the 151 keV 18 O(p,a)15 N narrow resonance, reveals that the main process occurring is exchange between oxygen from the gas and oxygen from the film matrix. However, following a post deposition vacuum or inert gas anneal, the atomic exchange process during thermal reoxidation, in 18 O2 , is significantly inhibited and limited to the superficial region. We assume a link between this effect and the crystallization of the films previously reported.  2004 Elsevier B.V. All rights reserved. Keywords: High-k dielectrics; Hafnium oxide; HfO2 ; ALD; Oxidation;

1. Introduction Aggressive scaling down of CMOS devices imposes the use of alternatives to SiO2 as MOSFET gate dielectric material to avoid unacceptably high direct tunneling current and excessive boron penetration when the oxide thickness is below 1 nm [1,2]. One of the most promising dielectrics is hafnium oxide due to its high dielectric constant

*

Corresponding author. Tel.: +33-1-4427-4643; fax: +33-14354-2878. E-mail address: [email protected] (J.-J. Ganem).

18

O resonance; Stable isotopic tracing

(high-k) of about 20 [3] and its thermodynamic stability in comparison with silicon [4–6]. As microelectronic devices require very thin HfO2 films with accurate thickness control, atomic layer deposition (ALD) is considered as a potential breakthrough technology to deposit HfO2 layers with uniform thickness over a large surface area [7]. However, although the HfO2 is expected to be thermodynamically stable with silicon according to the Gibbs free energy under equilibrium condition [8,9], it is well known that HfO2 and ZrO2 are poor barriers to oxygen diffusion and silicides can be formed at the dielectric interface due to oxygen deficiencies [10,11]. Moreover, previous

0168-583X/$ - see front matter  2004 Elsevier B.V. All rights reserved. doi:10.1016/j.nimb.2004.01.176

J.-J. Ganem et al. / Nucl. Instr. and Meth. in Phys. Res. B 219–220 (2004) 856–861

studies have reported the appearance of crystalline phases in the growing film under certain experimental conditions [13,14]. Thus, the study of the microstructure of HfO2 is a key point for the performance of future devices since mass and electronic transport are possible along grain boundaries that can have deleterious effects on device performance. The aim of the work is to study the composition of thin HfO2 after deposition and its transformation during annealing and oxidation processes, with a particular focus on oxygen transport during the thermal oxidation.

857

chamber and a solid film is formed as a result of subsequent surface reactions. In order to avoid gas phase reactions caused by inter-mixing of the precursors, the reactor is purged with pure carrier gas after each precursor pulse. The deposition temperature was fixed at 350 C. Initial surface preparation was achieved via chemical means resulting in a 0.7 nm thick silicon oxide top layer with a good fraction of fully oxidized Si observed by XPS [12]. The 77, 190 and 490 cycles of sequential pulses were carried out to form 3.5, 7.5 and 22 nm HfO2 films respectively. Some samples were annealed in dry N2 atmosphere and/or oxidized in ultra-dry 99.9% 18 O enriched oxygen gas (18 O2 ) at 100 mbar in a static UHV Joule heated furnace. The anneals were performed at 425 C or 800 C for 15 min and the oxidations at 425 C for 20 min (see Table 1). Ion beam analyses were carried out in the 2.5 MV Van de Graaff accelerator of the GPS group (Paris). Rutherford backscattering spectroscopy (RBS) analyses were done using a 1500 keV 4 Heþ beam in a non-channelling geometry. Oxygen areal densities were determined by nuclear reaction analysis (NRA) using the 16 O(d,p)12 C and 18 O(p,a)15 N reactions induced by a deuteron beam of 830 keV and a proton beam of

2. Experimental procedures and analyses HfO2 films with a thickness of 3.5–22 nm were deposited in a Pulsar 2000 ALCVDe reactor from ASM Microchemistry Ltd. at the CEA LETI (Grenoble, France) on low doped, p-type, 8 in. oxidized Si (1 0 0) substrates. Gaseous hafnium tetra-chloride (HfCl4 ) and water vapour were used as precursors by volatilization at 196 C of solid HfCl4 and at 18 C of liquid H2 O and were carried into the reaction chamber in N2 carrier gas. The precursors are alternatively fed into the reaction

Table 1 Samples preparation Sample

Number of cycles

Thickness (nm)

R1 425 C

R2 800 C

18

P01 P01R1 P01R2 P01O P01R1O P01R2O P02 P03 P04 P02O P03O P04O P05 P06 P07 P05O P06O P07O

490 490 490 490 490 490 190 190 190 190 190 190 77 77 77 77 77 77

21.1 – – – – – 7.5 – – – – – 3.5 – – – – –

– p

– – p

– – – p p

– – p p – p – – p – – p – – p p

– – p – – p – – p – – p – – –

O2 oxidation 425 C

– – – p p p – – – p p p

The thicknesses were measured by ellipsometry just after the deposition process. R1 and R2 are 20 min N2 anneals at 425 and 800 C respectively. The 18 O2 oxidations were done under 100 mbar of gas.

858

J.-J. Ganem et al. / Nucl. Instr. and Meth. in Phys. Res. B 219–220 (2004) 856–861

Table 2 NRA and RBS results, the thickness films are calculated assuming an HfO2 density of 9.68 g/cm3 Sample

16 O (· 1015 at/cm2 )

18

O (· 1015 at/cm2 )

16 O + 18 O (· 1015 at/cm2 )

Hf (· 1015 at/cm2 )

Zr (· 1015 at/cm2 )

Cl (· 1015 at/cm2 )

O/Hf

Thickness (nm)

P01 P01R1 P01R2 P01O P01R1O P01R2O P02 P03 P04 P02O P03O P04O P05 P06 P07 P05O P06O P07O

115 116.5 115.5 116 115.5 112 44 44 47 38.5 44.5 43.5 20 19.6 22 12.5 14.2 20

0.2 0.2 0.2 3.6 3 3.7 – – – 8.4 3.7 2.6 – – – 8.1 7.9 1.3

115.2 116.7 115.7 119.6 118.5 115.7 44 44 47 46.9 48.2 46.1 20 21.1 22 20.6 22.1 21.3

58.5 – – – – – 21.7 20.5 21.7 – – – 9.3 9.4 9 – – –

0.6 – – – – – 0.26 0.25 0.26 – – – 0.1 0.1 0.07 – – –

3.8 – – – – – 2.1 2.3 2.2 – – – 0.8 0.7 0.3 – – –

1.92 – – – – – 1.89 2 2.02 – – – 1.83 2.02 2.1 – – –

21.1 – – – – – 7.8 7.4 7.8 – – – 3.3 3.4 3.2 – – –

The ratio O/Hf is calculated after subtracting 3 · 1015 O/cm2 corresponding to the SiO2 interfacial layer. Hf and O have been measured within a precision better than 2% while Zr and Cl within 30%.

730 keV, respectively. In front of the detector were placed a 19 lm and a 12 lm mylar foil, respectively to avoid the detection of retro-diffused particles. The detection angle was 150 with the respect to the incident beam. The results are summarized in Table 2. Oxygen 18 depth profiles in the samples were measured by nuclear reaction profiling using the 18 O(p;ac)15 N nuclear reaction induced by a proton beam around the 151 keV energy resonance. The samples were titled to 60 with respect in the incident beam direction in order to increase the depth resolution to about 0.7 nm. A 600 mm2 surface detector with a 3 lm mylar foil was used to stop the backscattered particles. SIMS measurements were performed using a Cameca IMS-5f magnetic sector instrument equipped to operate at low impact energy. Csþ primary ions were accelerated at 2.5 keV and the primary current was around 10 nA; the positive secondary ion extraction voltage was 1.5 keV giving an effective impact energy and incidence angle of 1 keV and 50, respectively. The primary beam was rastered over 200 lm · 200 lm for an erosion rate of approximately 0.05 nm/s. Vacuum in the sample chamber with caesium operating was

2 · 109 torr. The secondary ion signals MCsþ and MCsþ 2 were measured using an electron multiplier detector system with a post acceleration voltage. Measurements of Hf, H and Cl concentrations were determined by normalizing to OCsþ 2 and SiCsþ 2 signals in HfO2 and Si respectively, after calibrating with implanted samples.

3. Results and discussion 3.1. As deposited samples Typical RBS spectra are shown in Fig. 1(a), One can appreciate the very good sensibility of the technique for Hf detection even in the case of the thinner film where the areal density corresponds to a total Hf amount of 9.3 · 1015 Hf atoms/cm2 . The magnification of the spectrum, between the channels 250 and 450, reveals the presence of two contaminants as shown in Fig. 1(b) which corresponds to chlorine atoms and zirconium atoms. Zr concentration is found to be constant in all samples with the value of about 1% (Zr/Hf) which corresponds to the natural contamination of the

J.-J. Ganem et al. / Nucl. Instr. and Meth. in Phys. Res. B 219–220 (2004) 856–861 2500

100

P01 P02 P05

Hf

P01 80

1500

60

Counts

Counts

2000

1000

Cl

Zr

40

20

500

0

859

0

100

(a)

200

300

400

0 250

500

300

(b)

Channel Number

350

400

450

Channel Number

Fig. 1. RBS spectra on as deposited samples of three different thicknesses (a). In (b) is represented a magnification of the P01 sample spectrum which reveals chlorine and Zirconium contaminants.

1.0E+07

1.0E+22

Intensity (counts/s)

1.0E+06 16

1.0E+05

Hf

Si

Concentration (atoms/cm3)

28

O

1.0E+04 1.0E+03

18

O

1.0E+02

1.0E+21

H

1.0E+20

Cl

1.0E+19

1.0E+01 1.0E+00

1.0E+18

0

(a)

10

20

30

40

50

Thickness [nm]

0

(b)

10

20

30

40

50

Tickness [nm]

Fig. 2. MCsþ 2 SIMS profiles for the 21 nm sample. Oxygen and silicon in (a) and hafnium, hydrogen and chlorine in (b).

HfCl4 solid source of the ALD system. We also observe chlorine atoms with an areal density that increases with growing the film. This contamination is probably due to uncompleted chemical reactions during the process leaving Hf–Cl bonds in the dielectric. We expect this undesirable effect, also reported in previous studies, to be reduced by increasing the H2 O pulse time, which should more efficiently remove the chlorine through the formation of HCl molecules. As a consequence the O/ Hf ratio is below 2 due to incomplete replacement of Hf–Cl bonds by Hf–OH bonds during the H2 O pulse.

As expected, NRA analyses show that oxygen amount grows linearly with the number of ALD cycles indicating a steady state growing process. The SIMS profiles also reveal that the oxygen concentration is homogeneous in depth as shown in Fig. 2(a). The 18 O signal, 500 time less intense than the 16 O one, corresponds to the natural abundance of 18 O in oxygen (0.204%). The spectra reveal also that chlorine atoms preferentially concentrated in near the dielectric/silicon interfacial region as shown in Fig. 2(b). The apparent long tail of the Hf signal below the dielectric/silicon interface has been identified to be an artefact of the

860

J.-J. Ganem et al. / Nucl. Instr. and Meth. in Phys. Res. B 219–220 (2004) 856–861

SIMS technique most probably due to preferential ion intermixing mechanism, since by varying the effective sputtering energy the magnitude of this effect varies. The analysis of regions located at both interfaces (air/dielectric and dielectric/silicon) are also subject to caution as transitory sputtering effects occur. 3.2. N2 -anneals and oxidations in

18

O2

NRA results show that during the N2 annealing process a slight increase in the amount of oxygen occurs, more pronounced at 800 C and no losses of hafnium are observed. This probably indicates oxygen contamination from the annealing chamber. However, it is not clear where these additional oxygen atoms fix: they can diffuse through the hafnium oxide and contribute to grow the interfacial SiO2 layer, to fill oxygen vacancies in the dielectric bulk or a combination of the both. In order to clarify how oxygen behaves in the hafnium oxide films we have performed thermal oxidation using 18 O2 gas. This allows the discrimination between the initial natural oxide present in the hafnium oxide and the 18 O possibly incorporated during the oxidation. NRA results (Table 2) as well as SIMS spectra show clearly that the 18 O2 thermal oxidation is characterized by the incorporation of 18 O atoms in the films while 16 O atoms are lost. The most important mechanism is the exchange between the 18 O atom from the gas phase and the oxygen from hafnium oxide matrix.

However, a slight increase of total amount of oxygen (16 O + 18 O) after the oxidation treatment is also observed. The incorporation of 18 O depends strongly on the thermal history of the film as a post deposition anneal significantly reduces the oxygen incorporation and the higher the anneal temperature the higher is the resistance to oxidation as shown in Table 2. Moreover, the amount of 18 O present in the 22.1 nm film after oxidation is lower than that in the case of the thinner films. We suggest that during the thermal deposition the sample is also annealed at the temperature of the ALD chamber (350 C) for a longer time than for the thinner films since 490 cycles are needed to form the 22.1 nm sample while 190 and 77 cycles are required to form the 7.5 and 3.5 nm samples respectively. Thus, the thicker sample has been submitted to a 350 C thermal exposure of 25 and 32 min longer as compared to the 7.5 and 3.5 nm, respectively. Previous studies have shown that a thermal anneal is accompanied by the appearance of crystalline phases in the dielectric film and the authors suggest that the chlorine plays a key role in the crystallization [13,14]. This fact shows that the oxygen diffusion is not accelerated in contrast to the expected behaviour in a structure with grain boundaries and possible preferential diffusion pathways. The oxygen transport is then more likely due to initial presence of oxygen vacancies. SIMS profiles as well as 18 O nuclear resonance profiling using 18 O(p,ac)15 N at 151 keV show that most of 18 O atoms are fixed just below the sample 300

500

Sample surface

Sample surface

P05O P05R1O P05R2O

400

P02O

250

P03O P04O

Counts

Counts

200 300

200

150 100

100

0 150

(a)

50

151

152

153

154

Protons Energy [keV]

155

0 150

156

(b)

151

152

153

154

155

156

Protons Energy [keV]

Fig. 3. Excitation curves measured using the 151 keV 18 O(p;ac)15 N resonance on 3.5 nm (a) and 7.5 nm (b) HfO2 samples oxidized in 18 O2 atmosphere at 425 C just after: deposition (black circles), post-deposition N2 anneal at 425 C (open circles) and post-deposition N2 anneal at 800 C (open squares).

J.-J. Ganem et al. / Nucl. Instr. and Meth. in Phys. Res. B 219–220 (2004) 856–861

surface and homogeneously in the bulk of the dielectric film for the 21 and the 7.1 nm samples. In the case of the 3.5 nm film, only the NRP technique can give exploitable spectra thanks to the high depth resolution of the technique. In this case, for the as-deposited and the 425 C pre-annealed samples about 40% of the initial 16 O present in the films is exchanged with 18 O atoms during the thermal oxidation. However, after the 800 C anneal, the incorporation of 18 O in the film is significantly reduced as shown in Fig. 3 and the dominant effect is oxygen exchange between the gas phase and the oxygen from the dielectric in the superficial region.

4. Conclusion Thin HfO2 films deposited by atomic layer deposition on silicon substrate have been characterized by ion beam techniques. We have found that after deposition the films present chlorine contamination and a lack of oxygen. They are unstable toward thermal oxidation since a high oxygen transport and exchange mechanisms occur during the process. Oxygen diffusion can be significantly reduced after a thermal anneal in N2 atmosphere. However traces of water vapour can have effects on film properties since they are very reactive towards oxidation just after deposition. This work should be helpful in order to better control the HfO2 /Si system and to optimize process parameters with the aim to integrate hafnium oxide in the heart of future MOS devices.

861

Acknowledgements The authors want to thank Ph. Holliger for his precious help in the SIMS measurements. This work is supported by the Comite National de Formation en Microelectronique, CNFM (France).

References [1] D. Buchanan, IBM J. Res. Dev. 43 (1999) 245. [2] L. Feldman, E.P. Gusev, E. Garfunkel, in: E. Garfunkel, E.P. Gusev, A.Y. Vul (Eds.), Fundamental Aspects of Ultrathin Diectrics on Si-based Devices, Kluwer Academic, Dordrecht, 1998, p. 1. [3] J. Robertson, J. Vac. Sci. Technol. B 18 (2000) 1785. [4] H.J. Hubbard et al., J. Mater. Res. 11 (1996) 2747. [5] G.D. Wilk, R.M. Wallace, J.M. Anthony, J. Appl. Phys. 89 (2001) 5243. [6] L. Kang, B.H. Lee, W.J. Qi, Y. Jeon, R. Nieh, S. Gopalan, K. Onishi, J.C. Lee, IEEE Electron Dev. Lett. 21 (2000) 181. [7] M. Ritala, M. Leskela, Nanotechnology 10 (1999) 19. [8] O. Knache, O. Kubaschewshi, K. Hasselmann (Eds.), Thermochemical properties of Inorganic Substances, 2nd Ed., Springer, Berlin, 1991. [9] D.G. Schlom, J.H. Haeni, MRS Bull. 27 (2002) 198. [10] T.S. Jeon, D.L. Kwong, Appl. Phys. Lett. 78 (2001) 368. [11] K. Muraoka, Appl. Phys. Lett. 80 (2002) 4516. [12] O. Renault, D. Samour, D. Rouchon, Ph. Holliger, A.-M. Papon, D. Blin, S. Marton, Thin Solid Films 428 (2003) 190. [13] J. Aarik, A. Aidla, H. M€andar, V. Sammelselg, T. Uustare, Thin Solid Films 340 (1999) 110. [14] D. Blin, N. Rochat, G. Rolland, Ph. Holliger, F. Martin, J.-F. Damlencourt, T. Lardin, P. Besson, S. Haukka, M.-N. Semeria, ECS Spring Meeting, 2003.