- Email: [email protected]

Surfax

and Coatings

Technology

94-95

(1997)

138432

The influence of the discharge power on film composition obtained by reactive magnetron sputtering

Abstract The preparation of the binary nitrides or oxides with narrow homogeneity range at low substrate temperatures is considered. An attempt is made LOestablish the analytical dependence of film composition on both discharge power and ratio of argon and reactive gas pressures. It is shown that under the increase of the discharge power up to a certain limit, the increase of target sputtering rate is compensated by zhe proportional growth of the number of active nitrogen particles. Starting from some discharge power, the rate of arrival of these particles is steadily decreased and metal excess appears in the film. The metal/reactive gas ratio grows proportional to the discharge power at its high values. The aforementioned is illustrated by the data obtained for magnetron sputtering of aluminium target in the gas mixture Ar-N2. 0 1997 Elsevier Science S.A. Kcy~r*ovds:

Reactive magnetron sputterin g; Film composition:

Discharge power

1. Introduction The preparation of the nitride or oxide films by the method of reactive magnetron sputtering is carried out by

the sputtering of the correspondingmetal target in a mixture of argon and a reactive gas(nitrogen or oxygen). The addition of argon to the reactive gas is expedient for several reasons: (I) for the discharge stabilization, (2) for the increasingof the film depositionrate, and (3) for the obtaining of films with the necessarycomposition and structure. The essentialparameter regulating the composition of the obtained films is the ratio of the partial pressuresof the reactive gas and argon (PJPA~).The required film composition for the majority of oxides or nitrides is obtained at the parametervalue which exceedscertain thresholdone. If this excessis high enough, the film deposition rate is decreased considerably and in somecasesthe films becomelessdense due to the excessof the reactive gas.On the other hand, at a constant PAP*, value and if the increaseof dischargepower is higher than certain level, this leads to deficiency of the reactive gas component in the films. The influence of the parameterPJPAr on the films composition has been investigated in many studies.However, * Corresponding author. Tel,: +380 11 4113101; e-mail: [email protected]

0257~8972/97/$17.00 0 1997 Elsevier PII s0257-8972(97)00273-9

fax:+380

44 1342078;

Science S.A. All rights resewed

the influence of dischargepower on film composition has only been reported in small number of articles (seee.g. [l51). In these papersit was experimentally establishedthat substantialincreasein dischargepower leadsto decreaseof volatile component in the film. In ref. [3], with help of mass spectroscopyand optical emissionspectroscopy, the influence of discharge power on transition process when the target surFaceis changedfrom compoundto metal was studied in detail. This, in turn, leadsto drastic changeof sputtering rate and film composition. But the questionwhy the rate of metal component sputtering grows faster then active particle quantity in reactive gasat dischargepower increase remainsunclear and needsfurther elucidation. In the presentpaper, we make an attempt to establishthe analytical dependenceof the obtained film composition on theseparametersand on dischargepower in particular. To be specific, we shallexamine the metal-gas binary systems, where the reactive gas molecules are not chemisorbedon the sputtered metal surface at low temperatures. It is known that for the synthesisof the aforementioned compoundsat low temperaturesof substrate,the presenceof active particles in the reactive gas is necessary.Such particles may appearin the dischargeonly. First of all, we shall try to determinethe number of atoms of metal and the reactive gas (e.g. nitrogen), arriving at the substrate.Next, we shallanalyze how the compositionof the

formed films depends on the partial pressure of the nitrogen in the chamber and discharge power.

2. Results and discussion At the sputtering of the metal target in nitrogen/argon atmosphere, part of the nitrogen is bound into the compound in the growing film. This leads to a decrease of the nitrogen pressure in the chamber in comparison with that when the discharge is absent. However, only active nitrogen particles can be bound in the film. It was shown earlier that at low substrate temperatures, these active particles are mainly atomic nitrogen [6,7]. It is clear that the higher the discharge power, the higher the number of active nitrogen particles. At constant discharge voltage (v) the nitrogen pressure variation in the chamber will be proportional to the nitrogen ion current. dhz = - WJk&

(1)

where k is the coefficient of proportionality which depends on the discharge voltage. The total target current consists of currents of me argon and nitrogen ions I=I*,+I,,

(2)

At a pressure less than 0.66 Pa (5 x lo-.’ Torr) the atom mean free path is more than 1 cm. This is greater than cathodic space, where ions are accelerated to substrate. If we consider for simplicity that sputtering occurs~at sufficiently high vacuum (better than 0.66 Pa), then charge exchange effects can be neglected, and as it was shown in [8], following approximate equation is valid

4&ir = WN,

(3)

Differentiating it at PAr = co~sr and with respect to Eq. (1): one can obtain

?=A& 2 where A is constant. From the comparison of this expression with Eq. (5). it follows that: (1) k(U) is constant, and does not depend on U; (2) all suggestions and approximations used in Eqs. (l)-(3) are correct. Thus, the increase of bound nitrogen quantity in the film can be explained by the increase of the active nitrogen particles number, which is proportional to the argon ion current Iti. This, in our opinion. is the physical meaning of Eq. (1) and hence of Eq. (5). The sputtering of metal is necessary only for the formation of a clean metal surface, where the active nitrogen particles will be bound. This can be confirmed by experiments. when the metal is evaporated in the non-activated nitrogen. In this case the nitrogen pressure in the chamber is almost constant, if trivial, ‘walled-up’ trapped nitrogen molecules in the growing films are not taken into account. Thus, the quantity of active nitrogen, formed in the discharge and suitable for the nitride formation, is proportional to the variation of nitrogen pressure with the switching on the discharge. PNzok + PN20

j-J:+

pN, =

‘3 1 +k(U)+-

(4) Ar

where PN,o is the nitrogen pressure in the absence of discharge. Let us consider the physical consequences of this errpression. It follows from Eq. (4) that

(5) It was shown by Seidman [l] that at the magnetron sputtering of silicon in the argon-nitrogen atmosphere the following empirical relation holds in wide range of discharge voltage U variations:

(6)

Ar

This dependence is shown in Fig. 1. If the discharge power is low (low ZA,), then AP-I,,.. Let us define the rate of metal sputtering from the target. At small discharge voltage variations when_the-sputter yield depends on ion energy linearly, this rate is proportional to the discharge power SUZ, where I is discharge current,sputter yield of metal when discharge voltage equals one unit. Since the part of the target 0 is covered by the nitride film, then for the [email protected]=of m~etalo_ne-can~vrite s zzs,gy

The solution of this equation gives the nitrogen pressure in the chamber, that depends on the current of the argon ions

AK

-PN2=AP=

+ ( 1 - 0%1,

(7)

where Skit” is selective sputter yield of metal from nitride and &,?, is this yield from clean metal. Coefficrents $zY and Shlr depend by themselves on the rat-io of quantities of bombarding ions Ar and Nz. which varies with discharge power variation. But this dependence is much weaker than that of 0 on theseparameters.So, below for simplicity we shall consider S,“::” and S1ieasconstants. Let us consider the numb-f metalLandnitrogen atoms arriving to the substrateper unit time. This quantity is proportional to the sputtering rate in SIU for a metal, and to the pressuredecreaseAP (Eq. (6)) in the chamber after the dischargeis switched on for the nitrogen. Taking into account the aforementioned considerations, we obtain the following expressionfor the ratio of atomic fluxes of metal and nitrogen, arriving to the substrate Me N=

SIU BAP

(8)

430

V. Yu. Kulikotsky

PN20

_____

/ Sur&e

and Comings

-----_e-7--I

V IAr

Fig. 1. The nitrogen pressure decrease (LIP) as a function of the argon ions current (I,,) (solid curve) at the reactive magnetron sputtering of metal in the Ar-N2 gas mixture.

where AP is determined by Eq. (6) and B is somenormalization factor. For binary nitrides of type MeN with a narrow homogeneity range, the stoichiometric compound is formed when Me/N < 1. If Me/N > 1, there is an excess of metal atoms in the film. Eq. (8) is sufficient for the qualitative analysis of the discharge power influence on the films composition. It follows from Eq. (8) that at dischargepower (and consequentlylAr) growth, Me/N is growing too. However, one can attempt to analyze this relation in more detail, expressing I and S through I,,, and fixed parametersPAr and PNjI,. The resulting expression, though being complicated, allows us to extrapolate the behaviour of Me/N to the small and large dischargepowers. i.e. small and large I,. The number of reaction capablenitrogen particles grows and nitrogen pressurein the chamber decreasesalong with the nitrogen ions contribution to the total current, as the dischargepower increases.Substitution of Eq. (3) into Eq. (2) gives for the dischargecurrent

Technology

94-95

(1997)

428-432

In this casethe increaseof Z,, at constant dischargevoltage leads both to the proportional increase of sputtering rate due to the dischargecurrent increaseand to the growth of reaction capablenitrogen particles number. The ratio Me/N in the film remains almost constant. The growth rate of reaction capable nitrogen atoms number will decrease with further increase of IAr (i.e. of the discharge power) until all the nitrogen moleculesbecome plasma activated. The latter case is, of course, ideal and can never be realized. Further increase of the power would enlarge target sputtering rate rather than reaction-capable nitrogen particles number. So, under the increase of the discharge power up to a certain limit, the increaseof target sputtering rate is compensatedby the proportional growth of active nitrogen particles number. Starting from some discharge power, the arriving rate of theseparticles (~AP/cIZA,)is steadily slowing down. The number of nitrogen atoms arriving at the target per unit time becomessmallerthan the number of sputtered ones. The target surface would become more and more nitrogen depleted and the target nitride coverage B would decreasefrom its initial value 00by some A0. The latter is proportional to the relative decreasein obtained nitrogen quantity. Let us estimate it. One can seefrom Fig. 1 that at someargon ion current IAr, the number of reaction-active nitrogen particles is proportional to PA correspondingto the point A (Fig. 1). However, if the nitrogen partial pressuredecreasewere absentin the sputtering process,the partial current I,, and therefore, AP would grow proportionally to IAt. Then the number of active nitrogen atoms would be proportional to the value PB, correspondingto the point B. Pressuredifference betweenthese points PA - PB is proportional to the numberof active nitrogen atomsor ions which do not reach the target due to the pressuredecrease.The aforementionedconsiderationspermit us to write the following expression A0 PB-PA -...-=--00 Af’l~~~+o Substituting

(9)

and Now let us try to express0 and therefore S as a function of argon ion current Iti. For this let us consider Fig. 1. At small dischargepowers, i.e. at small I,,. the nitrogen pressure decreaseis, as follows from Eq. (6), proportional to I,, AP=

PN2kk PA,

Discharge current growth with I,, linearly too (seeEq. (9))

into this expression, we get A0 ~ -= ‘O

kk

l+k&

Since 0 = B0- Ad, then

where C = COKV, i.e. the ratio Me/N is proportional to I AP The qualitative dependence of Me/N on Z,, is depicted in Fig. 2. An increase of the nitrogen concentration in the gas mixture makes P/PN2, and S,, decrease. It leads to a decrease of the Me/N ratio at all discharge power values. In reality. for most of the current sources an increase of the discharge current is followed by small increase of the discharge voltage, i.e. U=/3I+U, where Uo is the minimum discharge voltagee,[email protected] is-~small coefficient. In this caSe at small discharge current values

Fig. 2. The dependence of the ratio of the number of arriving to the substrate atoms of metal and nitrogen Me/N on the argon ions current. Plot 1 corresponds to the lower PN?/P*~ value.

The Eq. (10) resembles very much Eq. (4). This is because the part of the target surface covered by nitride varies similarly to the nitrogen amount arriving at the target, which in turn is proportional to PNz. If Z,, + 0, then 0 + BO-and if ZAr + 03, then % -+ 0. The 80 value is a function of many parameters of the discharge (power, PN2/PAr ratio, r,,, number of nitrogen atoms and ions created in the discharge, the dissociation rate of N: on the target, etc). It cannot be described quantitatively in general, but its behaviour can be analysed qualitatively, and conclusions can be made about its influence on the composition of obtained films. Substitution of Eq. (7), Eq. (9) and Eq. (10) into Eq. (8) gives Me -= N

(so+$&,J

(;y;$;)u &k Ar

where P = P, + PNZ and ,SO= %,,S’:~& + (1 - %&Yhfr is the sputter yield at PN,, PAr fixed and small discharge power. At small ZAr

and therefore small power

so puo =~+EFQiyp

P I

The last formula does not change essentially the character of the dependence presented in Fig. 2, since /3 is usually small. Thus, in spite of the fact that IA, is almost unmeasurable, our approach permits to determine the character of the dependence of Me/N on discharge power at its small and large values. At large discharge current values

It is clear that the stoichiometric compound with formula MeN and a narrow range of homogeneity can be obtained for all values of Me/N < 1. If Me/N > 1. the film will contain metal excess. This excess will be realized when the discharge power is increased at constant value of PN,/PY The aforementioned can be illustrated by the data for magnetron sputtering of aluminium target in the gas mixture AI-N~ (see Table 1). The stoichiometric composition of AlN films was prepared at ratios of P,.+/P=OA and 0.5, and a discharge power of about 180 W. Further increase of power led to the appearance of excess Al in the film. The excess Al increases with increasing discharge power. In our case Table

where D = comt. Note that Me/N does not depend on the current I,,. At large U and I,, we obtain

/SOP

1

The dependence of the Al/N ratio in the films prepared by the reactive magnetron sputtering on the different discharge paramctcr> p,: I& *Ai Power, W Discharge voltage, V Al/W

0.4 180 360 1.00

0.5 171 110 1.02

0.5 230 390 1.28

0.5 280 585 1.61

0.6 290 570 1.43

0.7 308 580 1.15

0.8 300 570 0.97

432

1’. Yu. Kulikovsky

/ Su&ce

and Coatings

1.28-times power increase (from 230 to 280 W) led to 1.26 times increase of Me/N (from 1.28 to 1.61, see Table 1). This is in accordance with predictions mentioned above. At the constant discharge power this excess could be compensated by increasing the nitrogen partial pressure at PN2/P=0.8 only (see Table 1). This consideration points out the principal limitations for the sputtering rate by reactive magnetron sputtering for films of stoichiometric composition.

3. Conclusion The influence of discharge power on the film composition obtained by the reactive magnetron sputtering method has been investigated. It was shown that the film composition depends weakly on the discharge power at its low values.

Technology

94-95

(1997)

428-432

The ratio Me/N grows proportionally to the discharge power at its high levels.

References [l] L.A. Seidman, SOF. Electwn. ind., I32(4) (1984) 15, [2] L.A. Seidman, SOL’. Electron. Technick, 5(378,l (1985) 44 131 S. Berg, T. Larsson and H.-O. Blom, J. Vat. Sci. Techoi. A, 4(3) (1986) 594. [4] A.R. Nyaiesh, Thin So/id Films, 86(2/3) (1981) 267. [5] M.P. Delplancke, V. Vassileris and R. Winand, J. Vat. Sci. Technol. A, 1313) (1995) 1104. [6] A.N. Pilyankevich, V.Yu. Kulikovsky, L.R. Shaginyan and L.I. Novozhenyuk, Sov. Poverkhnost, 6 (1988) 68. [7] V.M. Vereshaka, V.Yu. Kulikovsky and L.R. Shaginyan, Sov. Poverkhnost, 3 (1992) 57. [S] A.N. Pilyankevich, V.Yu. Kulikovsky and L.R. Shaginyan, Sov. Poverkhnost, 12 (1991) 24.