dielectric films deposited on silicon substrates are reported. In the growth of ... and field strength (in excess of 6 x 106 V cm- t) with low ionic content (typically.
Thin Solid Films, 37 (1976) 453-460 © Elsevier Sequoia S.A., Lausanne--Printed in Switzerland
453
ELECTROCHEMICAL PROPERTIES OF DIELECTRIC FILMS OF A L U M I N I U M OXIDE DEPOSITED ON SILICON
SAMPURAN -SINGH Department of Electrical Engineering, The University, Southampton ( Gt. Britain) K. V. ANAND Electronics Laboratory, University of Kent, Canterbury ( GL Britain) (Received March 23, 1976; accepted April 14, 1976)
Chemical, physical and some electrical properties of aluminium oxide dielectric films deposited on silicon substrates are reported. In the growth of these films we utilized the vapour phase reaction of the hydrolysis of aluminium trichloride (AICI3) in an r.f. heated horizontal reactor tube to give films of good thickness uniformity and run-to-run consistency. The activation energies associated with the constituent components of this complex chemical reaction were determined. It is shown that the films exhibit high dielectric constant (7.7 + 0.2) and field strength (in excess of 6 x 106 V cm- t) with low ionic content (typically about 101° cm-2), making them an attractive contender for passivation and for use as the gate dielectric of metal-insulator-semiconductor (MIS) devices. 1. INTRODUCTION The use of aluminium oxide films as gate dielectrics for MIS technology has recently received much attention 1. Their potential use can be envisaged both in replacing the conventional silicon dioxide (SiO2) insulator or in creating a double-layer structure in combination with an ultra-thin SiO 2 film to produce a non-volatile field effect memory device 2. We report here some characteristic properties such as typical density, refractive index, dielectric constant, ionic content etc. together with the etch rate and the IR spectrum as a function of deposition conditions of the aluminium oxide films. There are various methods available for the preparation of aluminium oxide. These include techniques such as the decomposition of organic oxyaluminium compounds 3, reactive sputtering 4, glow discharge s, anodization 6 and hydrolysis 7 of A1Cl 3. Films reported in this paper were grown by the last of the abovementioned procedures. 2. EXPERIMENTAL DETAILS OF THE DEPOSITION SYSTEM
2.1. The deposition system A diagrammatic representation of the hardware associated with the deposition system is shown in Fig. 1. Details of the design criteria have been discussed elsewhere z .
454
s. SINGH, K. V. ANAND
Reaction
R.f.coit
churner NNN 0 0/0 O f
/ 0
0
0
Silic°nslice I
)Exhaust
MSMS
0
Aluminium pellets Exhaust +x
Fig. 1. Aschematicdiagramofthealumina(Al203)filmdepositionsystem:/Xneedlevalve; x Whitey valve; [] two-stagegas regulator; MS molecular sieve drying column; F flowmeter;NRV non-return valve; LPR low pressure regulator. 2.2. Chemical reactions Deposition of aluminium oxide by the hydrolysis of A1C13 involves two serial reactions. These are as follows: (i) the formation of water vapour in the gas phase by the reaction between hydrogen and carbon dioxide, i.e. H 2 + C O 2 --~ H 2 0 + CO and (ii) the deposition of aluminium oxide (A1203) by the reaction between A1C13 and the formed water vapour, i.e. 2A1Cl3 + 3H20-~A1203 + 6HC1 As shown in Fig. l, gaseous A1Cl s was formed in the chlorinator by passing chlorine over heated (at 4 3 0 _ 10 °C) aluminium pellets. 2.3. Optimization of gas flow rates The physical quality, i.e. uniformity, run-to-run consistency and clarity, of the deposited films was studied as a function of the flow rates of various gases. It was found that high flow rates (in excess of 25 cm 3 m i n - 1) of chlorine caused etch pits in the silicon substrate slice as shown in Fig. 2(a). This arises from the fact that, at high flow rates, some of the chlorine passes unreacted through the chlorinator and subsequently attacks the silicon according to the reaction 2C12 + Si ~ SIC14 It was also found that for low flow rates of hydrogen and carbon dioxide, i.e. less than about 150 cm 3 min -1, the resulting film was " g r a i n y " in texture when
ELECTROCHEMICAL PROPERTIES OF ALUMINIUM OXIDE FILMS
w
455
o
(a)
(b) Fig. 2. (a) A microphotograph of the silicon substrate showing circular etch pits; (b) a microphotograph of the silicon substrate showing the "grainy" texture.
viewed under an optical microscope (see Fig. 2(b)). This was due to the resulting insufficiency of water vapour to carry the hydrolysis to completion. Hence it is likely that under these conditions a chlorine-bearing oxygen-deficient alumina layer of the form A12ClxO3_x would be formed s. In order to obtain films free of etch pits and grains, the optimum flow conditions given in Table I were arrived at after extensive tests. These flow conditions were used to deposit all films reported in this paper. At a deposition temperature of 900 °C, the uniformity of the film within the same slice was better than 40 A and run-to-run consistency was usually better than 100 A.
456
S. SINGH, K. V. ANAND
TABLE I
OPTIMUM FLOW CONDITIONS Gas
Flow rate (cm 3 r n i n - l )
Chlorine Argon carrier through chlorinator Hydrogen Carbon dioxide Main argon carrie1
16 42 700 470 870
3. RESULTS AND DISCUSSIONS
3.1. Some properties of the films Table II summarizes (for details see ref. 2) some of the experimentally determined properties of aluminium oxide layers grown at 900 +_l0 °C. T A B L E Ii
SOME MEASUREDPROPERTIESOF ALUMIN1UM OXIDE FILMS DEPOSITEDAT 900 °C Property
Value
Density Refractive index Dielectric constant Breakdown field
3.8 + 0 . 4 g c m -3 1.75+0d 7.7+0.2 (6.0+0.3) × 106 V c m - t for positive bias (6.8_+0.2) × 106 V cm -1 for negative bias 101° cm -2 typical
Mobile ionic charge
For the breakdown field a positive bias represents the condition when the metal plate of the MIS capacitor was positive with respect to the silicon substrate. It has been shown 9 that the most probable reason for the polarity dependence of the breakdown field is the existence of positive surface states at the aluminasilicon interface. Their presence effectively shields the semiconductor in the negative bias condition, so that for a given applied voltage of either polarity the magnitude of the actual field in the dielectric is lower for negative bias than it is for positive bias, thus resulting in a higher breakdown voltage (and apparent field) for the former. The mobile ionic charge was measured by using the wellknown temperature bias technique 1°. The value was found to be low in comparison with those quoted for thermally grown silicon dioxide which confirms that aluminium oxide has a dense structure.
3.2. Colour chart for the thickness of aluminium oxide films The colour chart given in Table III was established by depositing films of different thickness at 900 °C. In the range of about 600 /~ to 1250 /!~ the film thickness was measured by ellipsometry xl with high accuracy (i.e. _ 1 0 /~). Outside this range the film thickness was calculated by measuring the capacitance of the metal-aluminium oxide-silicon structure under the accumulation condition.
ELECTROCHEMICAL PROPERTIES OF ALUMINIUM OXIDE FILMS
457
TABLE III COLOUR CHART FOR ALUMINIUM OXIDE; THE FILMS WERE VIEWED UNDER VERTICAL FLUORESCENT LIGHTING
Thickness range (A,)
Colour
400-550 550-700 700 -850 850-900 900-950 950-1050 1050-1150
Light brown or tan Golden brown Brown -violet Blue-violet Deep blue Light blue Metallic blue Light yellow Bright yellow Orange Red -violet Blue-green
1150-1300
1300-1550 1550-1700 1700-1900 1900-2150
3.3. Temperature dependence of deposition rate of aluminium oxide A logarithmic plot of the deposition rate versus 1/Twhere Tis the deposition temperature (in degrees absolute) resulted in a straight line as shown in Fig. 3. A value of 0.57_ 0.01 eV for the activation energy associated with the deposition reaction was calculated from this plot. Two reactions take place during the deposition of aluminium oxide: firstly between hydrogen and carbon dioxide to produce water vapour and subsequently between this water vapour and aluminium trichloride. The activation energy of 0.57 eV represents the sum of the activation energies associated with each of the two reactions. In order to isolate the energies of the individual components, the reaction was restricted to the formation of water vapour only by eliminating A1CI 3 from Temperature (°C) 1100 1050 1000 950 900 850 i , i i i =
800 i
I000
Alumina ..~vation
energy = 0.5?-* 0.01 eV
o
