Web http://www.jcse.org/

Volume 3 Paper 13


Qualitative/Quantitative Model to Study the Deposition of Aluminium Nitride in Intensified Plasma Assisted Processing

E. De Silva and 1W. Ahmed
Faculty of Research, Knightsbridge University, Postboks 13, 8981 Spentrup DENMARK, E-mail:-
1Department of Chemistry and Materials, Manchester Metropolitan University, P.O.Box 34, Oxford Road, Manchester M9 8DN. 

Abstract

This work describes a theoretical model for the mechanism involved in a triode glow discharge employed to grow aluminium nitride films. A series of systematic studies were carried out at varying degrees of cathode voltage, cathode current density, pressure range and ion collector voltage values over a fixed period of time. All films developed were characterised using Scanning Electron Microscopy, Electron Dispersion X- ray Analysis, X- ray Diffraction and micro hardness tests. This project has collated a basic theoretical model that can be applied to produce films of different materials.

Key Words: Glow discharge, Triode, Aluminium nitride

Introduction

Aluminium nitride (AlN) is a binary III-V compounds with a wurtzite-type structure. The electron to atom ratio is 4:1. The structure shows an irregular distribution of tetrahedral bonds with a c/a axis ratio of AlN being 1.6. [] It has an outstanding combination of properties. They are high thermal conductivity (320 W/mK), excellent physical and chemical stability at high temperatures (melting point is about 2400 °C), optical transparency in the visible and near infrared (IR), piezoelectric and good insulating properties (the band gap is approximately 6.3eV ) and high hardness (HV 1400). Aluminium nitride is mainly used as a high strength special material to harden the surface and in the microelectronics industry as a basic material acting for passivation of semiconductor surface, transmission of surface acoustic waves etc. []

The use of an intensified glow discharge in a triode system to enhance the plasma to deposit aluminium nitride was initiated in 1991. [] This technique known as Intensified Plasma Assisted Processing (IPAP), is an improvement from traditional glow discharge techniques. Plasma intensification is produced by thermionic emission. Meletis [] carried out the first High Intensified Plasma Assisted Processing (IPAP) work to develop coatings on aluminium. Their initial work proved the system to be advantageous over conventional nitriding using a diode. They carried out the work in two ranges of the plasma energy spectrum. The first HEPAP work at Manchester Metropolitan University was started in 1996 [] to extend the operating parameters using a balanced RF system to the substrate for the deposition of AlN. This paper describes the details of the optimum conditions achieved and the quantitative/qualitative model used in IPAP.

Experimental

The IPAP system was built around a water cooled non-magnetic stainless steel vessel ( Figure 1). The negative emission source with filament and a DC powered positive plate were positioned facing one another across the inside of the vessel and were controlled separately. The RF powered substrate holder was connected to a matching network that enabled the process operation in the automatic mode. The system was continually pumped and included a heating and pre-mixing nitrogen-argon container. A plasma was sustained in the system across the pressure range 1.33 x 103 Pa to 1.33 x 10 -1 Pa. The substrate used was 1.5 cm thick and had a diameter of 5 cm. It was air craft quality HE30 grade aluminium (Al 98% by wt.)

Initially the system was evacuated to 1.33 x 10 -4 Pa and flushed several times with argon. In order to remove the Al2O3 layer formed on the surface due to oxidation the samples were sputter cleaned with the substrate voltage at 2000V with argon for 10 minutes. The argon gas was phased out gradually whilst simultaneously introducing nitrogen. All samples after each experiment were removed from the chamber and characterised. Several sets of experiments were done using unpolished and polished aluminium grade HE30. In all cases the experiments were carried out by varying each parameter over 7 stages. There were four variable parameters (viz: cathode voltage, nitrogen gas pressure, cathode current density and ion collector voltage) and initially one parameter was changed whilst the other three parameters were kept constant. Next, two parameters were changed while the other two were kept constant. Then three parameters were changed whilst one parameter was kept constant and finally all parameters were changed to obtain the optimum working range which gave a thicker deposition. These values were compared with predicted optimum parameters to establish a possible correlation for the future.

�Range of parameter values used in this study

Parameter Range
Cathode Current Density       (0.25 � 3.5) mA cm-2
Ion Collector voltage (20 � 100) V
Nitrogen Pressure 1.33 x (103 -10 �1) Pa
Bias Voltage (100 � 3000) V

Scanning Electon Microscopy (SEM) and Electron Dispersive X-ray Analysis (EDS), X-ray diffraction (XRD) and Auger Electron Spectroscopy (AES) and surface hardness were performed on the coatings deposited. The quality of the layer produced was affected by the cleanliness of the surface to be coated . The physical nature of the surface to be treated determined the pre-treatments required to get the thicker deposition. At this stage the surface was carefully inspected for signs of contamination or damage. Any undesirable matter was removed from the surface prior to treatment. This involved breaking of the bonds between the contaminants and the substrate surface either by physical or chemical methods. The polishing therefore was done until a fine mirror-like surface was obtained. After the polishing stage, the sample was washed with acetone to obtain a clean surface.

Results

The layers deposited were characterised for hardness, structure and composition. The appearances of these samples were light black due to compositional difference and the surface layers were highly adherent and dense. In conventional nitriding at temperatures lower than 400 °C the adhesion is very poor and improves with increasing temperature. During HEPAP a meta -stable Al3N is thought to form due to the energetic ion bombardment. It is believed that this aspect plays an important role in promoting adhesion by reducing compressive thermal stress arising due to differences in the thermal expansion coefficients between the Al3N coating and the Al substrate.

The microstructure of the IPAP films grown on to Al substrates revealed interesting features. The films appeared to be growing in a vertical direction as is evident by the columnar structure shown in figure 2. Following initial deposition, the rate of growth in the vertical direction appeared to be much greater than the growth rate in the horizontal direction. It is also evident that outgrowth occurs from the substrate surface as aluminium atoms are removed due to sputtering caused by high energy bombardment of the surface. The Al3N phase seemed to be an important feature in this columnar mode of growth in this deposition method.

For engineering applications the micro-hardness is an important property that determine the in-service performance. Figure 3 shows the surface micro-hardness of IPAP coatings compared to conventional plasma coatings. It is clearly evident that the IPAP process produces coatings which are three times harder than plasma nitride coating. In general, a greater resistance to wear can be expected with an increase in micro-hardness which ultimately means longer in-service of the component.

The composition of the AlN coatings was analysed using a SEM and EDS. Peaks showing the presence of Al ,N, Si/W, Ti, Fe and Ar were evident. Signals from the base material and the filament also showed up in the spectrum. Initial examination of the texture revealed the presence of cubic Al3N phases at orientations of (200),(311),(400) (see figure 5).

Discussion

1)     In order to explain the mechanism of HEPAP we attempted to develop an acceptable theoretical model. In this the major issue was the energy distribution of ions and neutrals in the dark space. Davis and Vanderslice [] and Rickards [] have done several studies on the distribution and their effect

Work done by Davis and Vanderslise [] suggested that the distribution of energies to be dependent of cathode fall length and the mean free path for exchange collision.�

In the primary ionic state for triode systems the relationship between the mean free path, λ, and charge collision is given by

���λ = 1/ nσ���� ---------------- ( 1 )

� σ� = collision cross section

�n = number density of the gas

In their analysis they made four assumptions.

The modification of the above model by Rickards [] gave a more general form which enabled field distributions other than the linear case to be considered.

The following equation defines the number of ions per energy interval as:-

dN/dE = (No/m) (L/λ) (1-E) (1/m) -1 exp{ - L/λ + L/λ (1-E) 1/m } ------- ( 2 )

E= ion energy relative to the maximum (this lies between 0 and 1.)

L = Cathode Fall Length

No = number of ions entering the cathode fall region from the negative glow.

λ = Mean free path for charge-exchange collision

When m = 2 the equation is equivalent to the Davis and Vanderslice model [].

According to the work by Matthew et al [] when m equals to 4/3 the value corresponds to the more realistic space charge limited (free-fall) case, the value of m chosen has minimal effect on the distribution shape. The value of L/λ influences the distribution shape. A reduction of L/λ produces an energy spectrum in which a greater proportion of the ions have energies nearer the maximum. Similarly, by decreasing the value of L/λ the energies of ions could be controlled near to the minimum level. This control helps to change the film formation and growth process.

Chapman [] has shown that L can be calculated, knowing the current density, by using the free-fall version of the Child-Langmuir equation.

L = (4Eo/9J)1/2 (2q/M)1/4 V3/4� ------------------------- (3)

J= Cathode current density

Q= Ionic charge

M = Ion mass

V= Bias Cathode Voltage

Eo = Permitivity of free space

This above equation has been derived for situations where the cathode is bombarded on one face only, which is flat and infinitely large. The ions originate from the negative glow with an initial energy of zero. The work done by Matthew et al [] to test the validity of this equation showed differences when compared with the results of other researchers []. Although all data showed good precision the accuracy was not comparable. In addition they have also commented about the possible inaccuracy (about 10%) due to secondary electron current density which was ignored in those experiments.

Where different gas pressures have been used e.g. higher pressures, the mobility limiting effects could be expected to become important. An alternative equation developed by Grass-Marti et al [] have taken into account of intermediate situations between the free-fall and mobility limited regimes. Yet, Matthews et al [] have stated that this equation did not prove their model to have any greater applicability than the free-fall Child �Langmuir equation. Their work further proved that Davis and Vanderslice assumption of sigma is independent of energy to be untrue. Sheldon�s [] predictive equation for sigma and Hegerberg and co-workers [] data have been proved to be more reliable. In this project the HEPAP process was studied as a total system with no controlled ionic states. Hence we have attempted to consider all possible occurrence during the HEPAP subject to following assumptions.

Although previous work showed HEPAP to yield better results than conventional systems under low temperature and low pressures, this did not produce the best results with high precision. We attempted to run the HEPAP process within the theoretical model given below. This included several assumptions and extensions to already suggested mechanisms. []

The assumptions involved in the theoretical model

1) We have considered Rickards equation where m= 4 / 3 to have a minimum effect on the distribution shape. []

2) L/λ is considered to have a critical influence on the distribution. []

3) Mobility limiting factor to be minimal in HEPAP process where the pressure range was less than 100 mPa. []

4) The effect of the secondary electrons to the ion current density to be 10% of the total as reported in previous studies. []

5) Neutral energy distributions to be displaced to lower energies than that for ion energy distributions. []

6) There is no interaction between Ar and N2 gases. []

7) Ionisation efficiency to be more effective as an indicator or ions arriving at the cathode. []

8) The total Et = Eletn + Ehen + Eions []

Et = total energy

Eletn = Low Energy Thermal Neutrals

Ehen = High Energy Neutrals

Eions� = Energetic Ions

9) The N2 gas could exist in any of the following forms given below.

Activity of total Nitrogen = Activity (N2 +N. + N+ +N2+)

Yet, N2+ state to be more dominant ionic species [] and the dissociative charge transfer to be less likely because of the short L.

10) The electronic configurations of aluminium and nitrogen are as follows:-

Al = Atomic number 13 :- 1s2 2s22p6 3s2sp1

N = Atomic number 7 :- 1s2 2s22p3��

In order to form aluminium nitride, the aluminium and nitrogen reaction has to cross the energy barrier. According to the results of previous work [] the two possible forms of aluminum nitride coatings recorded have been AlN and Al3N.

11) The diffusing flux for ith the layer of Al3N from the sample face to follow Fick�s first law []:-

Ji = Di . (ΔCi/Li)

Di = Nitrogen diffusivity of the layer

ΔCi = Concentration drop across the layer

L= Thickness of the layer

12) Overall pressure to represent the pressure at the cathode

In our work where a triode as opposed to a traditional diode was used, the following was clearly highlighted.

The IPAP has become a triode by the simple fact that the system includes a thermionic negative element. Hence, this being the only change, it is correct to assume that the results obtained have been influenced by this factor alone. In extending our understanding of the system the obvious mechanism could be described as follows:-

The total energy at each stage was calculated by combining equations 2 and 3.

Bearing the above factors in mind, one can extend the above equations, to receive the average ion energy as follows:-

Eia = Eit/N0 = Vc [ 2 (λ /L) � 2(λ /L) + 2(λ /L)e-L/λ ]

Where m= 2, N0 = Total ions and neutrals and Eia = average ion energy, Eit = Total ions energy, Nhen = number of high energy neutrals

The neutrals in the system can be either high energy or low energy. The high energy neutrals gain energy by collision with ions.

Therefore Nhen = N0 (L/λ)

For High energy neutrals this could be written as :-

Eahen = Ethen/ Nn

Then

Eahen = Ethen / N0 (L/λ) = Vc (λ /L) [ 1-2 (λ /L) + 2(λ /L) - 2(λ /L)e L/λ ]

Eahen = average high energy neutrals, Ethen = total high energy neutrals, Vc = cathode voltage

Hence the dependency of L and λ is apparent and thus demands the correct setting up of the parameters in HEPAP.

The best results were obtained when the system was running at the following parameters:-

Parameter�������������������

Value

Cathode Current Density     

2.5 mA cm-2

Ion Collector voltage

100V

Nitrogen Pressure

7 Pa

Cathode Voltage

2500 V

For the energies below 880eV a thinner nitride diffusion layer has been observed. [] Hence the total energy had to be higher than this level. The calculations were made to ensure that the total ion energy (without the neutral energy) was higher by increasing the cathode voltage and cathode current density.�

If the cathode voltage is kept constant, then for the same gaseous ions,

L = k. (1/ cathode current density)1/2 --------------------- ( 4 )

Although our experiments extended to higher cathode current densities and cathode voltages that gave higher energy values, we did not see any major improvements. One reason for this deviation could be the assumptions made in our theoretical model. Nevertheless, it shows that continuous bombardment on the substrate surface over a longer period at high cathode voltages seemed to have given the back scattering effect thereby reducing the thickness of the layers. There seemed to be an optimum level above which any increase will not show any further improvement. This has been confirmed by other studies. []

Another reason will be the total number of bombarding particles. If the majority is thermal neutrals then they will dilute the effect. This will need further investigations using the ion efficiency.

Nevertheless, the calculations of energies will enable to predict the parameters near to the optimum levels.

The constant values used in our calculations were [], []:-

Constant����������������������������������������������

Value

Ionic charge

1.6 x 10-19 C

M�����������������������������������������������

4.66 X 10-26 kg for N2

E0 = Permittivity of free space

8.85 x 10-12 F m-1

σ

2.5 x 10-15 cm2 for N2

Our experiments finally confirmed the following facts:-

We suggest any of the following two possibilities for the mechanism responsible for the formations of AlN.

The first possibility is the formation of aluminium nitride in the plasma. This happens when active nitrogen ions react with sputtered aluminium ions. Initially they will be high in nitrogen (AlN) and upon condensation on to the surface of the cathode they decomposes and release nitrogen. This nitrogen may diffuse in to the substrate. Thus the composition becomes Al3N type.

The second possibility is that ionic bombardment induces vacancies and vacancy clusters and therefore increases nitrogen diffusion. This could generate a coating with a thickness of few atomic layers and some may have defects. These defects will also promote the diffusion of nitrogen.

                                                       

Suggested energy barriers involved in AlN formation

Hence our explanation is that there are two possible energy barriers ( vide: above graphs) and depending on the energy available the reaction path will vary. In normal conditions where sufficient extra energy is provided AlN could be the predictive state. In an environment such as IPAP, Al3N metastable would be more desirable since the temperature will be comparatively lower and the continuous bombardment support the nitrogen diffusion inwardly.

Conclusions

Our experiments have shown that IPAP system to be a feasible, low temperature deposition method for surface coatings. A triode created by thermionic emitter showed a significant improvement in deposition. Any comparisons with conventional systems will further established the efficient and effective process development of IPAP over other established methods. When aluminium substrates were exposed to IPAP conditions, they showed a satisfactory growth of aluminium nitride with an improved hardness. Although initial sputtering was useful in the removing oxide layer the results were not satisfactory. One of the major aspects in aluminium nitride coating was the cleaning. Our experiments showed, a significant improvement was observed by the use of polished aluminium substrates. The formed nitride layer did not show any further growth after reaching an optimum level. The ability to control power independently for the emitter, the specimen and the positive electrode facilitated the independent control of voltage and cathode current density. The entire process was successfully completed at temperature less than 400 °C. Therefore, HEPAP would definitely be an attractive process for coatings of temperature sensitive materials. The theoretical model developed will assist to calculate the ion energies, which in turn will give an indication of optimum parameters that, could be used. The model will become more reliable as and when the assumptions are confirmed to reflect ideal situations. This model could be now used for coatings of other material using IPAP.

One of the draw back of IPAP was the difficulty of controlling the thickness of the coating. The fact that the nitride layers did not grow further than a certain limit due to shielding effect of previous layers which in turn made it difficult to control the thickness. The author suggest the supply of an extra source of substrate material viz ; another aluminium disc, the substrate to be connected to a rotating device or to supply an additional source of the substrate material in powder form that will enable to enter the cathode fall region.

Figures

Figure 1. Schematic Diagram of IPAP

Figure 2. SEM of AlN

Figure 3. Surface Microhardness

Figure 4. EDS profile

Figure 5. X-Ray Diffractions


References

M. Morita, S. Isogai, K. Tsubouchi and N. Mikoshiba, Applied Physics Letters,38(1)1981, 50

J. S. Tjorp, D. Evans, M. Al-naief and M. Akhataruzzaman, Journal of Materials Science,25, 1990, 4965.

E. I. Meletis and S. Yan, J.Vac.Sci.Technol.A9(4),Jul/Aug 1991, 2279-2284

D. Tither, W. Ahmed, E. de Silva, D. Whittaker, P. Skeldon, G. Thompson and X. Zhou, Product Finishing, Vol.50, No.1,Jan 1997, 12-15.

W. D. Davies and T.A. Vanderslice, Physics Review,131, 1963, 219.

J. Rickards, vacuum, 34, 1984, 559

A. Matthews and D. G. Teer, Thin Solids Films, 80,1981, 41-48

B. N. Chapman, Glow Discharge Processes, 1980

B. Rauscgenbach, A. Kolitsch and E. Richeter, Thin Solids Films, 109,1983,37-45

A. Gras-Marti, I. Abril and J .A. Valles-Abarca, Thin Solids Films, 124, 1985, 59-65

J. W. Sheldon, Phys. Rev. Lett.,8 , 1962, 64A

R. Hegerberg, M. T. Elford and H. R. Skullerud, J. Phys. B., 15 , 1982 , 797

K. S. Fancey and A. Matthews, Surface and Coatings Technology, 33, 1987, 17-29

A. A. Adjaottor, E. Ma, and E. I. Meletis, Surface and Coatings Technology 89, 1997, 197-203

H. Valizadeh, Ph.D. thesis, Salford University, 1982

A. Brokman and F. R. Tuler, J. Appl. Phys. 52, 1, Jan 1981, 468-471