EPR-study of nitrogen implanted silicon nitride
Introduction
Amorphous silicon nitride (Si3N4) and silicon oxide (SiO2) remain the two key dielectrics in modern semiconductor devices [1], [2]. It is known that Si3N4 possesses a high density of traps (N∼1021 cm−3) and has the property to localize injected charges (both electrons and holes) for a very long time (about 10 years) at room temperature (∼300 K). The typical density of charged traps, occupied by electrons (or holes) for Si3N4 is nt∼5×1018 cm−3 [3]. The effect of electron and hole localization in Si3N4 is widely used in electrically erasable read-only memory (EEPROM) silicon devices [4].
A lot of effort was made in understanding the origin (atomic and electronic structure) of charge traps in Si3N4. The most popular model of such a trap is an amphoteric threefold coordinated silicon atom Si with unpaired electron [5], [6], [7]. Here symbols (–) and () mean a normal chemical bond and an unpaired electron, respectively. According to this model, silicon nitride samples in their initial, non-charged state must be paramagnetic. However, in the previous experiments done on non-charged Si3N4 samples, no EPR signals were detected [8], [9]. It was shown theoretically that the creation of the pair of charged diamagnetic defects Si: (negatively charged) and Si (positively charged) via the reaction Si+Si→Si:+Si is energetically unfavorable [7], [10], [11]. In Ref. [12] the model of diamagnetic neutral Si–Si bond was proposed as the defect which is responsible for the localization of carriers in Si3N4. This model explains the absence of EPR signal in non-charged silicon nitride. Quantum chemical simulation predicts the appearance of EPR signal after the localization of electron (or hole) by Si–Si bond [13]. In contrast to this prediction, EPR experiments on corona-polarized Si3N4 films also show the absence of EPR [8], [9]. For the explanation of the latter experimental fact, spin coupling due to the resonant quantum tunneling of localized spins on occupied traps through unoccupied traps was proposed [12]. This model supposes a high density of neutral traps (∼1021 cm−3) in silicon nitride.
According to the aforementioned model, the spin coupling occurs due to the electron exchange through traps situated at typical distances of a∼N−1/3∼10 Å. This coupling is governed by the hopping frequency of a charge to a neighboring trap and it should decrease exponentially with the distance a between neighboring traps. Hence, in samples with lower concentration of traps, this coupling should decrease rapidly. Isolated localized charges, which cannot take part in the exchange because of long distances between neighboring traps, will occur and the appearance of an EPR signal belonging to non-coupled carriers is expected. The objective of the present study is the EPR investigation of localized electrons and holes in polarized Si3N4 samples in which the density of memory traps has been reduced by nitrogen implantation.
Section snippets
Experimental
According to the Mott octet rule 8-N, the interaction of nitrogen atoms with Si–Si bonds breaks those bonds followed by the creation of energetically much favorable Si–N bonds according to the reaction [14]This reaction was used in [14], [15] to explain the removal of hole traps in the process of the silicon oxide nitridation. In the present study the nitrogen implantation was used to remove part of the existing traps (Si–Si bonds) in silicon nitride, aiming to decrease the
Results
EPR spectra of initial (before the N-implantation) Si3N4 samples showed very weak EPR signal with g=2.0054 and ΔHpp=0.39 mT which was the same for both non-polarized sample and films containing 5×1012 electrons (holes) cm−2 (as estimated by C–V). The signal observed originates from triply coordinated silicon atoms with unpaired electron, Si3Si, and associates with mechanical and growth defects in amorphous silicon and on the surface of crystalline silicon [16]. No other EPR signals were detected
Discussion
Room temperature EPR signals from 1011–1012 localized charges have been successfully observed for various Si-based systems [8], [11], [13]. On the other hand, as it was mentioned above, the life-times of localized charges in Si3N4 are extremely long (∼108 s). Both these facts allowed us to rule out any other reasons for the non-detection of the EPR signals (like very short relaxation- and/or life-times, etc.) except for an antiferromagnetic coupling. Let us now investigate various physical
Acknowledgements
This work has been performed under INTAS grant 97-347.
References (21)
- et al.
Solid-State Electron.
(1990) - et al.
Appl. Surf. Sci.
(1989) - et al.
Microelectron. Reliab.
(1998) Structure and electronic properties of amorphous dielectrics in silicon MIS structures
Science
(1993)- et al.
IEEE Trans. Electron Devices
(1999) - et al.
Microelectronics (Sov)
(1987) - I. Fujiwara, H. Aozasa, A. Nakamura, Y. Komatsu, Y. Hayashi, Proc. IEDM 98, p....
- et al.
Phys. Rev. B
(2000) - et al.
J. Appl. Phys.
(1993) - V.A. Gritsenko, Yu.N. Morokov, Yu.N. Novikov, J.B. Xu,...