CHARACTERIZATION OF DEFECT TRAPS IN SiO THIN FILMS

Department of Crystall#te Materials Science, Graduate School of Engineer#tg, Nagoya University, Furo-cho, Chikusa-ku, Nagoya-city, 464-8603, Japan; bSemiconductor Physics, Department of Fundamental Researches, Perpignan University, 52 avenue de Villeneuve, 66860 Perpignan Cedex, France; CMOPS-CLOES-SUPELEC, Metz University, 2 rue Edouard Belin, 57070 Metz, France; Center for Cooperative Research on Advanced Science and Technology, Nagoya University, Furocho, Chikusa-ku, Nagoya-city, 464-8603, .Japan


INTRODUCTION
Under normal operating condition, interface states and oxide defects are generated in the oxide SiO2 film and at the oxide-silicon interface of a metal-oxide-silicon (MOS) structure of microelectronic devices.
High electric fields, which appear in very thin isolative oxide layers, are Final proof reading by Max Blanco: < mblanco@arrow.utias.utoronto.ca> known [1][2][3] to be the cause of this generation which induces de- gradations of the devices properties.
In this work we are interested in the study of slow traps, which are responsible for long time electrical instabilities of MOS devices [4].This paper introduces a new measurement technique to characterize these defects.The method is based on capacitance-voltage C(V) [5][6][7][8] mea- surements at low temperature and it uses thermal energy to generate an inversion regime during the hysteresis cycle.It is applied on p-type poly-Si gate-MOS capacitors with thin (17.8 nm thickness) oxide layer to separate the effects of the different types of charges and traps.

THEORETICAL APPROACH
The C(V) characteristic is measured along a cycle which is described by varying the applied bias across the MOS capacitor from -5V to +5 V and back to -5 V.An hysteresis effect is observed.The method is based on the observation of the modification of the hysteresis cycle induced by a de- gradation process, which consist in a Fowler-Nordheim electron injec- tion from the gate, made under a constant voltage.
The C(V) characteristics measurements were carried out at tem- peratures below 100 K.In the case of p-type MOS devices firstly, the C(V) characteristics were measured from a negative bias to a positive bias.In this measurement, a deep depletion situation is observed at positive biases due to the low-temperature measurement.At the maximum positive bias, the sample temperature was raised to 300 K and kept for one hour, (heating and cooling cycle).Due to this high- temperature process, electrons are thermally excited and are accumu- lated in the inversion layer.Moreover, the excited electrons are cap- tured in the traps in the oxide and at the interface.All slow-state traps are also occupied in electrons and the charging process obeys Jonsher's law [9].Once the traps have been occupied by electrons, the sample was cooled down below 100 K again, and then the C-V measurement was performed from the positive bias to the negative bias.During this process, the electrons trapped by slow states in the oxide were not observed to be emitted from the states.Therefore, a complete hyster- esis can be obtained.We can separate slow-state traps from fast-state traps because the electrons are quickly released from the fast states.
A complete description of one hysteresis cycle needs a global ap- proach with the parameters: temperature, voltage and time.From the differences between the C(V) hysteresis obtained after degradation and the initial hysteresis after the heating process described above was ap- plied to the sample, three voltage shifts A Vii may be determined from curves displacements ('ii' will be 'ss' for slow states, 'fs' for fast states, 'of' for oxide fixed charges or 'rag' for migration species), together with the evolution in time of the voltage shift A V,,. Differences are observed between charging and discharging processes in the relaxation of traps.
From these voltage shifts, we can obtain the values Nii of each trap density: / Vo.Nil (1) qeox Heree,,.,. is the oxide thickness and eox is the dielectric permittivity of the oxide.In the case of slow-state traps, Jonsher's law [9,10] is written as: (2) Nd is the equilibrium density of defect sites.In this equation, a slow- state trap is taken as an oxide trap.The model considers the slow-state trap as a vacancy.This model can be applied successfully to the E' center case [11][12][13][14].AV, is then the gate voltage shift in the C-V characteristics due to the slow-state traps, Co., the maximum capaci- tance, q charge of the electron, the time, tc the time needed to dis- charge half of the slow-state traps and is a coefficient depending on tc.Experimental results for t and will be given.Our assertion, here, is to give the saturation term AV,(c)--qN/Cox.The general evo- lution Eq. ( 2) is also developed in [9].A V,, represents a part of A relative to the variation of the effective slow state traps, where A V,,,g is the voltage shift due to defects in the mid-gap and is indeed a true measure of the effective net oxide.

EXPERIMENTS AND DISCUSSION
Experiments were performed with sample #1" a MOS capacitors fab- ricated on a p-type Si (100) substrate with a boron concentration of 2 X 1017 cm -3 using LOCOS isolation (Local Oxidation of Silicon).The oxide film was grown in a wet environment at 850C and its thickness was 17.8 nm.The sample was subjected to boron implantation through the gate oxide under 40 keV to adjust the transistor threshold voltage.The polycristalline silicon gate area was 3.82 x 10 --4 cm2.
For this new High/Low Temperature C(V) (HLTCV displacement method), thermal energy is used to generate inversion regime and also to study slow-state traps more thoroughly.As far as formation of the inversion layer is concerned, both light illumination /13/and thermal heating have physically comparable effects.The heating treatment results in generation of extra oxide charges which effects can be se- parated.These effects are investigated in this present paper.
Figure shows C(V) characteristics measured at a temperature of 100 K for sample #1 after a voltage stress under an applied field of 10MV/cm, where the density of injected electrons was Ni,j 5 1017 cm-2.A good hysteresis is obtained for the separation of traps.The C-V curve from negative bias to positive bias has a feature of deep depletion condition due to the low measurement temperature.At 15V, the sample was warmed up to 308 K for 15 min and thus the capacitance increases by the formation of the inversion layer.Figure 2 displays the capacitance change at +5 V during warming up to 308 K, as a function of warming-up time.This result which would seem to show that the capacitance saturates above 500 s, which means that a 10 min warming-up is sufficient to form the inversion layer.
In Figure 1, at around 0V, the reverse C(V) curve has a plateau originating from the fast-state traps.Under the depletion region, the reverse curve runs parallel with the forward curve, which clearly in- dicates the existence of the slow-state traps.In this manner various traps effects are observed as induced voltage shifts.The shift between the forward and the reverse curve at the plateau region is A Vs, + A V./:, and at the depletion region A Vs. is determined.Accordingly, we can distinguish the slow-state traps from the fast-state traps and obtain here, N.,., 5.4 10 I cm-2, Nf 6 10 li cm-2.
The voltage shift between the non-stressed and the reverse char- acteristics, A Vou, corresponds to the density of oxide fixed charge No./ We found Nof= 1.2 10 l cm-2.Figure 3 shows these contribu- 00 I 41001 o00l 8001 'I0001 'I01 LIfo0 t(s) Vg /olts) FIGURE 3 Zoom of C-V characteristics.Superposition and expansion of curves obtained for the stressed sample (Fig. 1) and for a non-stressed sample.
tions due to slow-state traps and fixed oxide charges by superposition of both hysteresis cycles.It is a zoom of the two C(V) curves to show the separation of charge and to calculate densities.The total density of oxide trapped charges, Not, is Not-Nss +Nof, Not--6.8x l011 cm-2.
The total measurement process takes about two hours.It was found that, for an injected electrons density of N,,j 5 x 1017 cm -2 and for an applied electric field of E= 10MV/cm, the relaxation time t, is roughly one hour.The parameter 0 was evaluated, for an applied bias of Vg-5 V, to be 0.14 [9].

CONCLUSION
We have established a new C(V) method in order to separate slow- state traps from fast-state traps.Various oxide defects can also be separated (as migration ions at high temperature in inversion).
FIGUREHysteresis after a constant voltage stress that injected Ni, 5 x 1017 cm -2 at inversion conditions: 308 K, one hour.

FIGURE 2
FIGURE 2 Capacitance increase at the inversion region by warming up at room temperature.