научная статья по теме CATHODOAND PHOTOLUMINESCENCE INCREASE IN AMORPHOUS HAFNIUM OXIDE UNDER ANNEALING IN OXYGEN Физика

Текст научной статьи на тему «CATHODOAND PHOTOLUMINESCENCE INCREASE IN AMORPHOUS HAFNIUM OXIDE UNDER ANNEALING IN OXYGEN»

CATHODO- AND PHOTO LUMINESCENCE INCREASE IN AMORPHOUS HAFNIUM OXIDE UNDER ANNEALING IN OXYGEN

E. V. Ivanovo,"*, M. V. Zamoryanskaya". V. A. Pustovarovb, V. Sh. Alievc, V. A. Gritsenko€, A. P. Yelisseyevd

Iofje Physical Technical Institute 194021, Saint Petersburg, Russia

b Ural State Technical University 620002, Yekaterinburg, Russia

'"Institute of Semiconductor Physics, Siberian Branch of Russian Academy of Sciences

630090, Novosibirsk, Russia

d Institute of Geology and Mineralogy Siberian Branch of Russian Academy of Sciences

630090, Novosibirsk, Russia

Received July 14, 2014

Cathodo- and photoluminescence of amorphous nonstoichiometric films of hafnium oxide are studied with the aim to verify the hypothesis that oxygen vacancies are responsible for the luminescence. To produce oxygen vacancies, hafnium oxide was enriched in surplus metal during synthesis. To reduce the oxygen concentration, the film was annealed in oxygen. A qualitative control of the oxygen concentration was carried out by the refractive index. In the initial, almost stoichiometric films we observed a 2.7-eV band in cathodoluminescence. Annealing in oxygen results in a considerable increase in its intensity, as well as in the appearance of new bands at 1.87, 2.14, 3.40, and 3.6 eV. The observed emission bands are supposed to be due to single oxygen vacancies and polyvacancies in hafnium oxide. The luminescence increase under annealing in an oxygen atmosphere may be a result of the emission quenching effect.

DOI: 10.7868/S0044451015040151

1. INTRODUCTION

A11 increase in the information volume and operation speed of silicon-based devices, and the attainment of a terabyte scale are underlain by scaling effects and a decrease in the channel length in metal oxide semiconductor (MOS) devices. That decrease is accompanied with a decrease in the thickness of the gate silicon oxide. Silicon oxide was used as a gate dielectric during four decades. The thickness of the SiO-2 layer is 1.2 11111 while the designed value is 60 11111. A further decrease in the SiO-2 thickness is unacceptable because of strong tunnel currents of leakage, which lead to the heating of silicon devices and to a decrease in their reliability. A general approach to solving this problem is to replace the gate SiO-2 with alternative or high-/,-

E-mail: ivanova'fl'mail.ioffr.ru

dielectrics with a high dielectric permeability fl 3]. Alternative dielectrics have a permittivity k = 15 25, to be compared with k = 3.9 for SÍO2. High k values allow increasing the physical thickness of the dielectric up to 5 11111, which allows suppressing the leakage current.

The most promising materials that are currently introduced instead of SÍO2 are dielectrics based 011 lir().,..\;/ (k « 15), 11IHK).,, (k « 15), and HID, (k = = 25). The dielectrics based 011 hafnium oxide have higher thermodynamic stability on a boundary with silicon, higher permittivity, and higher barriers for holes and electrons fl 3]. The unresolved problem is high enough leakage currents due to traps and the capture of charge carriers on these traps. Therefore, studying the origin (atomic and electronic structure) of defects that operate as traps in hafnium oxide is very important.

Experimental [4 6] and theoretical [7 9] investigations of hafnium oxide show that oxygen vacancy is one

of the main point defects. Oxygon vacancies wore established to produce absorption in the spectral range

4.4 5.3 oV and to be responsible for electrical conductivity of HfO-2 [4]. The aim of this paper is the experimental study of luminescence at excitation by an electron beam (cathodoluminosconco) and by synchrotron radiation (photoluminosconco) in order to determine the physical nature of the luminescence center. For this, we studied a set of nonstoichiomotric Hf03. (x < 2) samples enriched in surplus hafnium (in the oxygen vacancies). Two ways to affect the oxygen vacancy concentration were used: the enrichment in metal during synthesis and annealing in a reducing atmosphere (vacuum), and a decrease in the concentration of surplus hafnium (oxygen vacancy) by annealing samples in oxygen.

2. EXPERIMENTAL TECHNIQUE

The Hf03. (x < 2) films were produced using ion beam sputtering deposition (IBSD) [10]. Silicon plates Si(100) with the resistance 4.5fi-cm, which had been subjected to a deep cleaning by the RCA Co technique [11] were used as substrates. Before mounting into the vacuum chamber, the substrates were treated in a HF solution to remove the natural oxide.

A silicon substrate was placed near the target from metallic hafnium (Williams Inc., Hf content more than 99.9%). The target was sputtered by a beam of Ar+ ions, and simultaneously we delivered high-purity oxygen (more than 99.999%) into the area near the target and substrate. A beam of sputtered particles fell on the substrate surface, thus forming an Hf03. film. The films were produced at room temperature. The substrate heating by hot particles from the target did not exceed 70 °C.

The beam of Ar+ ions for sputtering of the target material was formed by a the Kaufmann-typo source [12]. The energy of Ar+ ions was 1.2 koV, while the density of ionic current on the Hf target was

1.5 niA/'cm2. The thickness and rate of layer deposition were controlled by a quartz sensor (TM-400, Max-tec. Inc), located near the substrate. The composition (.r-paramotor) of the film was defined by partial pressure of oxygen using a gas flow controller. For our experiments, we grew two sets of Hf03. samples at partial oxygen pressures 9 • 10-3 and 2 • 10-3 Pa. In such conditions, we produced the samples of an almost stoichiometric composition (x « 2) and nonstoichiomotric samples (x < 2).

Part of the stoichiometric and nonstoichiomotric films wore annealed in a vacuum with the residual pres-

sure in the chamber less than 10-4 Pa at 600 °C during 1 h. The other part of the films was annealed in a flow reactor in the atmosphere of pure oxygen at T = 600 °C and P = 1 atm during 1 h.

Ellipsometric measurements on our films were carried out with an LEF-3M ellipsometer operating at A = 632.8 mil and the incidence angle 70°C. The thickness and refractive index were calculated in the frames of the one-layer reflection system Si-substrato Hf03.-film atmosphere.

The conditions of film producing and annealing, as well as ellipsometric data (thickness and refractive indices) for the films are given in Table 1. Enrichment of hafnium oxide in surplus metal during synthesis and annealing or at vacuum annealing leads to an increase in the refractive index. Annealing in oxygen is accompanied by a decrease in the concentration of surplus hafnium (oxygen vacancies), and the refractive index decreases. A similar effect is observed when silicon nitride SiN3. is enriched in silicon [13,14].

Photoluniincsccncc (PL) and PL excitation spectra were measured at 7.5 Iv in both stationary and time-resolved regimes using synchrotron radiation on a SUPERLUM station of the DESYLAB laboratory (Hamburg, Germany) [15]. The time delay St relative to the excitation pulse and the time window length At were chosen taking the luminescence kinetics into account. Two time windows were used: with Sti = 2.7 lis, Ati = 11.8 lis for the fast component, and with St-2 = 60 lis, ¿li2 = 92 lis for the slow one. The PL excitation spectra were measured at the excitation energy Ecx = 4 40 eV and normalized to an equal number of incident photons using sodium salicylate, whose quantum efficiency does not depend on the photon energy at hv > 3.7 eV.

Cathodoluminosconco spectra of Hf03. samples were taken with an electron probe microanalyzcr "Canie-bax". This microanalyzcr is equipped by four X-ray spectrometers for quant it ivo X-ray microanalysis and additionally by two optical spectrometers for cathodo-luminesccncc (CL) measurements. CL spectra were recorded in the range 1.5 3.8 oV at 300 Iv. The sample was irradiated by a beam of 5 oV electrons at a 0.2 /mi depth of electron penetration and a current density of 1.2 A/cm2.

3. CATHODOLUMINESCENCE SPECTRA

In Fig. 1, CL spectra for a hafnium oxide film of the composition close to the stoichiometric one (sani-

Table 1. Conditions of the preparation and the ellipsometric data (thickness d and refractive index /?) for the hafnium oxide films (Ad and An are the respective errors for d and /?)

Sample number Conditions of the sample preparation Ellipsometric data, A = 632.8 rim, Ad ~ 1 À, An ~ 0.001*

22 The initial HfO-2 film after annealing in vacuum (P < 10 1 Pa) at 600 C 1 h d = 757.6 A n = 1.986

23 The initial HfC>2 film after annealing in oxygen ambient (/',.>, ~ 10' Pa) at 600 (' 1 h d = 767.2 À n = 1.955

41 The initial II!'().,. (>• < 2) film grown at /><>, = 2 • 10 Pa. Nonst ehiomet ric composit ion d = 1054.2 À n = 2.010

44 The initial Hf03. (x < 2) film after annealing in vacuum (P < 10 1 Pa) at 600 ( ' 1 h d= 1038.7 À n = 2.013

42 The initial Hf03. (x < 2) film after annealing in oxygen ambient (Pq2 < 105 Pa) at 600°C/'l h d= 1069.8 À n = 2.010

band omission (Fig. 2). It is natural to suppose that the 2.7-eV PL band is due to oxygen vacancies in hafnium oxide. This conclusion is confirmed by the PL decrease after further annealing the film in oxygen (see Fig. 2). Annealing in oxygen also results in a decrease in the refractive index to n = 1.955 (sample No. 23).

A further confirmation that the 2.7-eV luminescence band is due to oxygen vacancies was obtained in the experiments with nonstoichiomotric hafnium oxide samples enriched in surplus metal during synthesis (see Table 1). In Fig. 3, we show the PL spectrum of the original nonstoichiomotric film (sample No. 41). Such a film has the refractive index n = 2.010 to be compared with n = 1.976 for the stoichiometric sample No. 24. The Hf03. film No. 41 with n = 2.010 demonstrates PL emission in the 2.7-eV band. The intensity of this emission is about 600 pulses/'soc. Annealing in vacuum leads to an increase in surplus hafni

Для дальнейшего прочтения статьи необходимо приобрести полный текст. Статьи высылаются в формате PDF на указанную при оплате почту. Время доставки составляет менее 10 минут. Стоимость одной статьи — 150 рублей.

Показать целиком