Capacitively coupled RF-plasma reactor Background of the invention Customary, capacitively coupled RF-plasma reactors comprise first and second mutually spaced electrodes which concomit- antly confine the plasma reaction volume. Besides of the two electrodes, their electrode conductive surfaces, no further electroconductive parts act as a third electrode with an ex- ternally applied electrical potential into the plasma reac- tion volume. Therefore, typical capacitively coupled RF-plas- ma reactors are also said "diode type" reactors. Confining the plasma reaction volume between the two electrodes which are mutually driven at electric RF potential is normally achieved in providing a spacing at the border of the two electrodes which suffices for electric isolation in terms of DC, but which prevents spreading of the plasma discharge out- side the space confined by the two electrodes. This may be achieved e.g. by spacing the two electrodes at their periphe- ry by a distance smaller than dark space distance at the re- spective operating conditions.

Such capacitively coupled RF-plasma reactors have been used for many years in the field of plasma processing, especially for treating flat substrates such as silicon wafers.

The most common applications of such reactors are Plasma-En- hanced Chemical Vapour Deposition (PECVD) and Plasma Dry Etching. Plasma Dry Etching can be divided into Reactive Ion Etching (RIE), in which ion bombardment is used to promote an anisotropic etching, or in Plasma Etching (PE), in which ion bombardment is to be avoided.

Plasma cleaning, removal of a polymer resist (Ashing) or plasma induced surface activation of workpieces can also be performed in a capacitively coupled RF plasma reactor. Most such reactors are construed with planar electrodes.

In these applications the plasma is generated by a driving RF voltage, most commonly for industrial applications at a fre- quency of 13.56 MHz.

An RF-driven plasma develops a large self bias voltage bet- ween the plasma and the plasma reaction volume surrounding walls which are, as was stated, the electrode surfaces. This is due to the rectifying effect of the RF voltage across the plasma sheath by the plasma sheath according to the dark spa- ce adjacent to each of the electrode surfaces towards the plasma discharge or reaction volume. As a consequence, in a classical capacitively coupled RF reactor the ion bombardment of the workpiece or substrate is governed by the ratio bet- ween the two electrode surface areas which are in contact with the plasma discharge. Most plasma self bias, enhancing ion bombardment, arises adjacent to the electrode surface of the smaller electrode area, and the potential difference across the sheath or dark-space adjacent to the larger area electrode is accordingly smaller. This effect is conveniently exploited by plasma process designers to taylor the ion bom- bardment on the substrate. E.g. for RIE the workpiece or sub- strate is placed adjacent or on the smaller-area electrode coupled to the plasma which results in larger ion bombard- ment. Inversively, for PE the workpiece is placed adjacent or on the larger-area electrode where ion bombardment is smaller.

There is today a need for plasma processing of large surface workpieces as e.g. of large surface glass plates. E.g. the flat display industry is today considering manufacturing on such plates of about 1 m2. The type of processes to be per- formed to produce the pixel related electronic circuits on such glass plates are of the same nature as the processes used in micro-electronics, but have to be performed on signi- ficantly larger surface workpieces.

If a large surface substrate is to be processed by a capaci- tively coupled RF plasma reactor and at process conditions similar to those as used when treating silicon wafers, the plasma gap (the distance between the two electrodes, e.g. of a planar capacitive reactor) must be kept at the predetermin- ed value in order to achieve the number of collisions for exited species as desired. Thus, a typical value for plasma gap is between 1 and 10 cm (both limits included).

With the large surface workpieces and such plasma gaps the electrodes of the reactor take a very extended aspect ratio and the electrode surface area ratio becomes close to 1. This implies that in a capacitively coupled RF plasma reactor for large surface workpieces there is almost no control of ion bombardment. The ion bombardment occurs substantially equally at both electrode surfaces. Such a situation is strongly li- miting processing large-size workpieces and especially large- size substrates by known capacitively coupled RF plasma pro- cesses as they were developed for the micro-electronic indu- stry.

Summarv of the invention It is an object of the present invention to provide a capaci- tively coupled RF plasma reactor which allows treatment of large workpiece surfaces with an accurate control of ion bom- bardment onto the respective electrode surfaces and thus an adjacent workpiece, be it to a desired low or to a desired high level.

This is inventively achieved by the capacitively coupled RF- plasma reactor, comprising - a first and a second electrode arrangement mutually spaced and confining a plasma reaction volume, - at least one of the electrode arrangements comprising electrically mutually isolated sub-electrodes, - a first group of the sub-electrodes being commonly connected to a first electric input, - a second group of the sub-electrodes being commonly connected to a second electric input.

This concept of the novel capacitively coupled RF-plasma re- actor thus substitutes at least one of the two customary used reactor electrodes by an array of sub-electrodes which by means of their respective first and second electric inputs may be independently and thus differently electrically opera- ted, usually but not exclusively with RF voltages.

In a most preferred embodiment the sub-electrodes of one group and the sub-electrodes of the second group alternate periodically, thus e.g. one after the other as considered at least along one direction along the subdivided electrode ar- rangement. In the preferred mode of realisation where sub- electrodes of the at least two sub-electrode groups respec- tively connected to different electric inputs, especially RF inputs, are periodically alternating, the distance from one sub-electrode connected to one RF input to the next such sub- electrode connected to the same RF input and thus local peri- odicity of the sub-electrode pattern with respect to their electric feed should be of the order or less than the extent of the plasma gap between the two mutually facing electrode arrangements. This makes sure that the workpiece surface to be treated is subjected to an "averaged" effect of plasma discharge.

In a simplified consideration it may be said that with the inventive reactor the workpiece surface to be treated is ex- posed to a multitude of different sub-RF plasma discharges generated in local parallelism and operated at selectable and thus mutually different RF signal amplitudes and/or phasing and/or frequencies and/or voltage-shapes and which neverthe- less interact in the plasma reaction volume to result in the said "averaged" effect.

Short description of the fissures The present invention shall now be described by way of exam- ples and with reference to the accompanying drawings. They show: Fig. 1: Schematically a prior art capacitively coupled RF- plasma reactor, Fig. 2 to Fig. 5: four different possibilities of RF-feed to a reac- tor according to fig. 1 and of arranging a workpie- ce with respect to larger or smaller electrode, de- pending on the desired ion bombardment on the sur- face of the workpiece to be treated, Fig. 6: the dependency of percentage of RF voltage across the plasma sheath at one electrode from substrate size and depending of whether the substrate or workpiece is deposited on the smaller or on the larger area of the plasma reaction volume confining electrodes at a prior art reactor as shown in figs.

1 to 5, Fig. 7: in a schematic representation the generic electrode arrangement of a reactor according to the present invention, Fig. 8: the schematic representation of the generic elec- trode arrangement of a reactor according to the present invention for explaining the various possi- bilities of electrical feed to the different elec- trode arrangements, Fig. 9: departing from the approach as shown in fig. 7 or 8, schematically the electrode arrangement of the reactor according to the present invention with a preferred shape of sub-electrodes, Fig. 10: an enlarged view of the reactor according to fig.

9, Fig. 11: as a function of phase-shift between RF voltages applied to sub-electrode groups at the inventive reactor and as a function of sub-electrode shape the resulting calculated percentage of RF voltage at the plasma sheath of the substrate carrying electrode arrangement, Fig. 12: the percentage of RF voltage across the plasma sheath at the substrate carrying electrode and as a function of amplitude ratio of RF voltages applied to two groups of inventively provided sub-electro- des, Fig. 13: in a schematic representation, a preferred con- struction of the reactor according to the present invention, Fig. 14: schematically the bottom view on the electrode ar- rangement subdivided in bar-shaped sub-electrodes at a reactor according to the present invention, Fig. 15: in a representation according to fig. 13 the sub- electrode pattern with two-dimensionally arrayed sub-electrode squares, Fig. 16: in a representation according to fig. 14 or 15 pro- vision of triangular sub-electrodes, Fig. 17: in a representation according to one of figs. 14 to 16 the provision of groups of honeycomb patterned sub-electrodes, Fig. 18: departing from a sub-electrode pattern according to fig. 15 the bottom view (a) and cross-sectional view (b) of the sub-electrodes with surface enlar- ging convex roof shapes, Fig. 19: in a cross-sectional representation the shapes of a further sub-electrode embodiment with curved sur- face, Fig. 20: an enlarged view of a part of the reactor according to the present invention and to fig. 13 with addi- tional preferred features, Fig. 21: schematically the RF feed of a reactor according to the present invention, Fig. 22: schematically a first preferred realisation mode of RF feed according to fig. 21, Fig. 23: schematically a second preferred mode of RF feed according to fig. 21, Fig. 24: in a schematic bottom view a further structure of sub-electrodes in a centric circular pattern, Fig. 25: in a schematic representation according to fig. 9 a further preferred embodiment of the inventive plasma reactor with border electrode arrangement, Figs. 26 to 28: three different distributions of treatment effect on a 400 mm x 400 mm substrate and in dependency of DC bias applied.

Detailed descriPtion of the invention with reference to the fissures Before explaining the behaviour of the reactor according to the present invention, let us introduce the notations and the method by recalculating the electrical equilibrium of a clas- sical capacitively coupled RF reactor as shown schematically in fig. 1. Such reactor comprises as was said two electrode arrangements, 1, 2 with respective electrode surfaces A1, A2 which are both capacitively coupled to the plasma 3 within the plasma reaction volume via the respective sheath. The driving RF voltage is generated between the respective elec- tric potentials V1 and V2 whereby usually one of these poten- tials, say V2, is reference potential, normally ground poten- tial. The plasma potential Vp is unique because a plasma is a conductor and can be considered conductive as a metal for the driving frequencies considered which are well below the plas- ma frequency. Capacitive RF coupling between one electrode and the plasma depends on the sheath thickness. The sheath can thereby be considered as a vacuum capacitor, the thick- ness thereof varying with RF voltage or potential-drop across the sheath. The most commonly adopted approach consists in describing the sheath thickness as a power law of the local voltage, i.e. potential difference: e1 = e2 V1 - Vp I a, e2 = e0 V2 - Vp I a, (1) where e1 and e2 are the respective sheath equivalent thick- nesses and a is a coefficient depending on the range of plasma condition and on the type of plasma chemistry being used.

The relation between V1, V2 and Vp is easily derived conside- ring that the two sheaths act as a capacitive divider: (V1-Vp)/(Vp-V2) = (A2/e2)/(A1/el), (2) hence IVl-Vpl/lVp-V2l I = (A2/A1)n with n = 1/(1-a). (3) We recover here the well known power law as first derived by Koenig et al. (H.H. Koenig and L.I. Maissel, IBM J. Res. Dev.

14, 168 (1970)). In this first derivation n was estimated to be around 4 (hence a = 0.75). In fact most electronic proc- ess conditions correspond more to n = 1.5 to 2, (a = 0.3 to 0.5). In any case the net result is that most of the RF volt- age is found in front of the smaller electrode surface. The result of this calculation is displayed in Fig. 6 where the sheath RF voltage is expressed in percentage of the total RF voltage across the plasma. The calculation is for a square substrate with some reasonable assumptions for the plasma gap (3 cm) and the edge distance (4.5 cm). It appears that when the substrate size exceeds 40 cm, it is impossible with a classical reactor to develop on one electrode more than 60 W of the RF voltage. It is also impossible to avoid the RF voltage on the larger electrode to be less than 40 W of the total voltage.

As the DC voltage developed across an RF sheath is nearly equal to the local RF voltage amplitude, it is difficult for larger substrates to impose voluntarily a large energy ion bombardment when desired, and it is also difficult to avoid some ion bombardment where such bombardment is not desired.

Hence RIE is difficult to implement in large substrate capa- citive coupling RF-reactors because only half of the potenti- al self bias will be present in front of the substrate. Reco- vering this factor implies increasing the RF voltage nearly by 2, hence increasing the RF power by a factor of the order of 3 to 4. On the other hand, avoiding ion bombardment is necessary for processes which are likely to cause plasma in- duced damages. Since this cannot be reduced by the electrode surface ratio it has to be reduced by reducing the RF ampli- tude at the expense of a longer process time.

In the figures 2 to 5 the different applications of driving RF voltage to the two electrodes of a capacitively coupled RF reactor are shown, thereby with disposal of the workpiece either on the larger electrode for reduced ion bombardment (fig. 2 and 3) or on the smaller electrode (fig. 4 and 5) for exploiting ion bombardment.

With the help of fig. 7 the principal according to the pre- sent invention shall now be explained. Again, the plasma re- action volume 3 of the reactor is confined by two electrode arrangements 10 and 20, principally according to the reactor as shown in fig. 1. In opposition thereto, at least one of the two electrode arrangements, according to fig. 7 one of them, is subdivided into sub-electrodes 12. Subdivision of the electrode arrangement 10 into electrically and mutually isolated sub-electrodes 12 may be realised just in one direc- tion x along the electrode arrangement 10 so that the sub- electrodes 12 are of a bar-like shape or may be realised in two directions x and y, so that the sub-electrodes 12 form a two-dimensional pattern.

By this it becomes possible to group the sub-electrodes 12 as shown by the groups A, B, C in fig. 7 and to apply to each of these groups different electric potentials, especially poten- tials differing generically in amplitude and/or phasing and/or frequency and/or shape, thereby especially such RF potentials. Such different and preferably mutually adjustable driving signals are shown in fig. 7 by V11 to V13.

Some general consideration to this inventive concept: - If both electrode arrangements 10 and 20 are e.g. driven on RF potentials and none is linked to a reference poten- tial as e.g. to ground potential, it is absolutely possi- ble to subdivide both electrode arrangements 10 and 20 into sub-electrodes and to also drive the sub-electrodes of the second electrode arrangement at respectively dif- ferent potentials (not shown).

- In a most preferred embodiment of one of the electrode arrangements as arrangement 20 being driven on a referen- ce potential the workpiece is deposited adjacent to the unstructured and reference potential driven electrode ar- rangement, i.e. adjacent to or on electrode arrangement 20 of fig. 7.

- Although it is absolutely possible so as to exploit all possible desired effects of the novel reactor structure to select the different driving voltages at different frequencies in a most preferred embodiment as proposed presently, the voltages applied differ in amplitude and/or mutual phasing.

- As shown in fig. 7 grouping of the sub-electrodes 12 is preferably realized periodically in that, considered in the direction x, sub-electrodes belonging to the groups A, B, C are arranged periodically alternating. Although this is a most preferred embodiment, it is clear that this periodicity can be omitted or locally interrupted if desired.

- In fig. 7 grouping the sub-electrodes in three respecti- vely RF-driven groups is shown. If desired, more such groups may be realised, but in the today's preferred mode the sub-electrodes are grouped in two differently and further preferably adjustably RF-driven groups.

For the following description we refer to the case according to fig. 7, where the sub-electrodes 12 are grouped periodi- cally in two groups, let us say A and B.

Thereby, the periodicity, i.e. the distance between two sub- sequent sub-electrodes belonging to the same group should be of the order or less than the extent of the plasma gap accor- ding to the distance PG of fig. 7. This will result in the fact that the workpiece is subjected to an "averaged" effect of the several "plasma columns" operated between respective sub-electrodes and the electrode arrangement 20 of fig. 7.

The most general formulation for the electrical signal V11 Vl2 etc. and V20 prevailing at the respective sub-electrode groups and, in case of an undivided single counter-electrode, to that counter-electrode 20 according to Fig. 7 may be writ- ten as follows: V11 = S1 (t) V12 = sect) V20 = 520(t) Thereby, S1, S2 and S20 denote spectral representations of signals, which spectra may vary in time. It must be pointed out that one of the most important spectral amplitudes is that prevailing for zero frequency, thus representing the DC bias-potential at the respective electrode arrangements. As schematically shown in fig. 8 by respective matching boxes MB1, MB2 and MB20 and possibly modulated signal generators Gull, G12 and G20 generating an output signal with a desired spec- trum and varying in time as desired, the inventive electrode arrangement may be operated to achieve a large scale of dif- ferent effects with respect to plasma generation. Thereby, additionally either of the three generators may be omitted and a respective electrode arrangement connected via a re- spective matchbox to reference potential or by further omis- sion of that matchbox, such electrode arrangement may di- rectly be connected to reference potential.

Further, two or three of the generators shown may be realised as one single generator connected via different matchboxes on the respective electrode arrangements. The skilled artisan will recognise the huge number of control possibilities for the electrical signals V , for the minimum number electrode arrangement as shown in fig. 8 and a generalised electrode arrangement of a sub-electrode set 12 with m sub-electrode groups and an electrode 20 also subdivided in n sub-electrode groups.

The matchboxes, which are either passive impedance networks or may possibly include active sources considerably contri- bute to the formation of the DC bias at the respective elec- trode arrangements, i.e. the spectral amplitude at zero fre- quency.

In fig. 9, departing from the example of fig. 7, there is schematically shown a preferred structure of a today's imple- mentation of the reactor according to the present invention.

This preferred structure shall be the basis for the analysis as follows. Thereby, the electrode arrangement 10 is subdi- vided into parallel sub-electrode bars 12, grouped in two groups A and B respectively driven by RF voltages V1l and Vl2 with respect to ground potential applied to the electrode ar- rangement 20. Grouping the sub-electrodes 12 is periodic, as may clearly be seen from fig. 9. The sub-electrodes 12 belon- ging to the respective groups are interconnected to the re- spective input by feed-bus 7A and 7B.

The sub-electrodes of the group A define for plasma coupling surfaces A whereas the sub-electrodes of group B define such surface AB. The electrode arrangement 20 is coupled to the plasma reaction volume via surface A20. Both sub-electro- de groups are capacitively coupled to the plasma via the re- spective sheath.

Let us assume: V11 = VOexp(i+)Vl2 = VOexp(-i+) (4) for the two sub-electrode group driving RF voltages, i.e. the two voltages have the same amplitude, but have a mutual phase shift of 2+. Phase + and/or amplitude V0 may be adjustable or even modulatable. As will be shown, such a reactor according to fig. 7 and more generically according to fig. 8, realised e.g. according to fig. 9 allows to control ion bombardment regardless of workpiece or substrate size and especially in a planar geometry. Indeed, the design according to the figs. 7 to 9 can be extended according to the workpiece extent to be treated.

As may be seen from fig. 9 the sub-electrodes are additional- ly tailored to enlarge the overall coupling surface AA,B of the electrode arrangement 10. Due to a slope 0 of the roof- shaped sub-electrode bars 12, with # = 45° according to fig.

9, the surfaces are as follows: AA = AB = A2o/2cos(0) (5) Note that, as shown in fig. 10, which represents an enlarged part of the reactor structure according to fig. 9, the RF coupling surfaces are evaluated in the central part of the reactor on one unit of the periodic structure. If the work- piece 4 and thus the reactor itself are very large, the bor- der effects can indeed be neglected in this calculation. With respect to accurately controlling even such border effects, we refer to fig. 25 and figs. 26 to 28 as well as to the re- spective description later on.

Please note that in fig. 10 the plates S1l, S12, S20 represent the respective parts of electrode surfaces acting as capaci- tive plates.

The continuity of RF current implies the following equations: (VOexp(+i+) - Vp) C11 + (VOexp(-i+) - Vp) C22 = VpC20 (6) Because AA= AB and / V1l / = / V12 /, then, Vp is real, it has a phase centered between V1l and Vl2. The result is expressed in a rather complex form: tg²(#) = ((X/cos0)2,aX2)/(l+X)2 (7) where X = cos+(VO/Vp)-1.

This relation was calculated for a = 0.5 (n = 2) and the re- sults are displayed in fig. 11. Fig. 11 shows the percentage of the driving RF voltage amplitude which is present at the boundary of the sheath facing the bottom electrode 20 on which the substrate 4 is deposited. The calculation was made for four types of saw tooth profile, hence for four values of the angle 8. For zero phase shift, in other words when all electrode groups are driven by the same RF voltage, one reco- vers the expected results expressed in equation (3). The RF voltage is equally shared when two flat electrodes are facing each other (0 = 0°). When the electrode is corrugated by the saw tooth structure, then the smallest surface electrode is the bottom electrode arrangement 20 and an increasingly large fraction of RF voltage occurs adjacent and facing the work- piece 4. Very impressive is how effective is phase shifting so as to decrease RF voltage in front of the substrate carry- ing bottom electrode 20. When the two driving voltages are in phase opposition, then the RF voltage in front of the work- piece carrying bottom electrode 20 becomes zero.

Although it is convenient to consider a mere phase shift bet- ween the driving voltages for the two electrode groups A and B according to V11 and V12, it is not necessary that the vol- tages on both sub-electrode groups of the array are of the same amplitude, as was explained with the generalised consid- erations to fig. 8, and even such amplitudes - single at one frequency, or multiple at several spectral frequencies - may vary in time. Thus, amplitude modulation, frequency modula- tion or phase modulation may be used. Additionally or instead of the varying phasing it is e.g. also possible to vary the amplitude ratio of the driving RF voltages. For the dual pe- riodic array as shown in the figs. 9 and 10, which is one preferred embodiment of the arrangement according to fig. 8, and still with a phase shift of 2 , varying the one frequency amplitude may be expressed as follows: V1l = VOexp(i+), V12 = aVOexp(-i+). (8) in which "a" is a real coefficient between 0 and 1.

As an example we consider the case of a dual periodic array with no phase shift + (+ = O in equation (8) but only with a variation of the amplitude of the RF voltage at the RF input for group B, V12 . The calculation can be done the same way as above. For illustration, fig. 12 shows the result for a saw tooth dual array with 0 = 450 and n = 2. Fig. 12 displays the variation of the plasma RF voltage above the substrate rela- tively to the amplitude of the main voltage amplitude V0 when the amplitude of the voltage at V12 is varied according to "a" of (8). It is found that the plasma voltage can vary from 15 to 67 W when the amplitude aVO of the voltage V12 is va- ried. Even more striking is the variation the self bias of the main electrode. The self bias is calculated by assuming that the plasma sheaths are perfectly rectifying the RF vol- tage. It is found that the self bias changes sign when the voltage amplitude of V12 is varied.

Before reporting some experimental results a preferred form of realisation of the reactor according to the present inven- tion shall be described with the help of fig. 13.

The bars as sub-electrodes 12 of group A are connected to the common bus-bar 7A and the bars of group B to the common bus- bar 7B. The bus-bars 7 are respectively connected to RF power inputs 8A and 8B. In order to avoid spurious plasma ignition the back of the sub-electrode array and the spaces between the electric connections to the sub-electrode bars 12 are shielded by spacer shields 9. The substrate 4 is deposited on the opposite electrode 20 which is preferably grounded. All electrodes in this embodiment are cooled or temperature-con- trolled by a circulating fluid. The fluid is circulating in conducts 11 in the sub-electrode bars 12 of the arrangement 10 and in the base electrode arrangement 20 in conducts 12.

The plasma 13 is filling the plasma gap between the electrode arrangements 10 and 20 which gap GP being somewhat larger than the distance P between two adjacent sub-electrodes 12.

As was already noted in the preferred embodiment according to fig. 13, but also according to figs. 7 to 10, the two sub- electrode groups are of equal surface, but it is also possi- ble to consider the case of unbalanced surfaces of the sub- electrode groups surfaces. As also shown in the figs. 9 and 10 the surface of the sub-electrode subdivided arrangement 10 is voluntarily corrugated on the side facing the plasma 13.

The purpose of this is to increase the surface of contact between the plasma and the electrode arrangement 10. Note that the profile of a sub-electrode 12 is effective in in- creasing the contact surface if the geometrical features are wide enough for the plasma to penetrate into the resulting corrugation. In the embodiment according to fig. 13 the sub- electrode bars have again 0 = 45". The surface of contact between the electrode arrangement 10 and the plasma 13 is therefore increased by a factor of a compared to the flat base electrode arrangement 20.

In a further preferred embodiment the base electrode 20 may be construed as a chuck for holding by suction flat workpie- ces 4 as glass plates to be treated. To realize such a chuck there is (not shown in fig. 13) realized within electrode arrangement 20 a system of suction channels which abut in a multitude of suctioning holes at the upper surface of elec- trode 20 to snugly hold the workpiece 4 during treatment, or there is provided an electrostatic chuck.

Experiments were performed on a reactor construed as shown in fig. 13. Both extremes with respect to RF driving were per- formed, namely where both electric inputs 8A and 8B were dri- ven together and thus V1l = V12 (see fig. 9) and where one of the inputs 8B or 8,was grounded, thus realizing a = 0 accor- ding to (8). The self bias potential and the RF driving am- plitude were measured. As seen in Table 1, the results are qualitatively very satisfactory. In particular one observes a spectacular change of signum of the self bias when the elec- trode equivalent contact surface is reduced by a factor of 2 when varying the RF voltage applied to one sub-electrode group from nominal to zero. The absolute value of RF self bias does not match the theoretical estimation. This is main- ly due to the fact that the measurements were performed on a rather small reactor of 500 x 500 mm2. The calculation for an infinite reactor as shown in fig. 12 must be corrected to take into account the contributions of the reactor edges to RF coupling. This was done in an approximate way in Table 1.

Nevertheless, the drastic variations of the DC bias are asto- nishing. The RF bias of the processing plasma in front of the workpiece can be drastically changed by relative variation of RF input voltages on the sub-electrode groups. SET UP MEASURED ON REACTOR RATIO RATIOZ5 RFinlets V0 RF Selt bias s ratio correcte calculated two in parallel 306 V + 244 V 1 0.79 0.62 0.33 one grounded 360 V - 142 V I -0-39 -0.59 -0.71 Table 1: Comparison experiment-theory for a dual array with 450 saw tooth. The corrected value is derived from a calcula- tion as above but roughly accounting for the edge effects which are far to be negligible in relatively small reactors.

The sub-electrode design according to figs. 9, 10 and 13 is very simple. It is the most simple periodic structure. From the bottom it appears as parallel stripes as shown in fig. 14 with alternated connections to the two RF inputs.

Such a periodic design can also be made periodic in two di- mensions, x / y of fig. 7. Such an example is shown schemati- cally in fig. 15 for a square-shaped sub-electrode 12.

Similarly, a dual array of triangular sub-electrodes 12 can be made as schematically shown in fig. 16. An even more so- phisticated approach is shown in fig. 17 where a triple array with three RF inputs and three groups A, B, C of sub-electro- des is realized. Again, the three corresponding driving RF voltages or more generic signal spectra can be out of phase and/or of different amplitudes, possibly even of different frequencies and/or voltage-shapes, the signals may have dif- ferent spectra.

The periodic structure of the sub-electrode array can also introduce features with respect to electrode surface as does the tooth pattern which was described. A diamond shape as shown in fig. 18 increases the RF capacitive coupling surface of each sub-electrode and may be applied to different sub- electrode structures e.g. as shown in the figs. 15 to 17. The sub-electrode profile can also be rounded as shown in cross- section in fig. 19. One of the only limitations for the de- sign rules of the electrode array structure is that when sha- ping the sub-electrode units, hollow volumes or volumes open to the plasma reaction volume such as holes, grooves or slots shall be wider than twice sheath thickness or dark space di- stance so that the plasma penetrates deeply into such reces- ses. In a typical plasma process the sheath thickness is a few mm, hence it is reasonable to assume that no hollow in- trusion on the surface of electrode arrangement 10 should be narrower than about 1 cm if such intrusion shall add to the active coupling surface of such electrode.

Considering the reactor according to the present invention and as most generically shown in fig. 8, it must be stated that laying the workpiece flat on the bottom horizontal elec- trode arrangement 20 is the most convenient approach, but the same design can be realized vertically or with the substrate and electrode arrangement 20 at top if appropriate workpiece and especially substrate holding is realized. Such different geometric arrangement can be considered with an eye on parti- cle control.

Some preferred additional features shall be explained with an eye on fig. 20 which shows an enlarged cross-sectional view of the sub-electrode array. In order to limit the capacitive coupling between the sub-electrodes of the e.g. two sub-elec- trode groups the gaps 23 between adjacent sub-electrodes can be provided with spacers 24. These spacers can be made of an insulating material such as of ceramic or of a metal, either at floating potential or at a defined potential intermediate the potentials applied to the adjacent sub-electrodes. Simi- larly, a shield 26 above the sub-electrodes reduces the capa- citive coupling, either between the sub-electrodes and/or from the sub-electrodes to the backing ground plate 25. This shield 26 can be continuous, but then it shall be made of an insulating material such as of ceramic or if it is metallic, it shall be interrupted as shown in fig. 20. Several shiel- ding layers can be interposed to improve capacitive decoup- ling. In fig. 20 a second shield 27 is shown to improve de- coupling between the sub-electrode array and the backing ground plate 25. The shield 27 is actually a grid to allow gas to flow therethrough.

This introduces another advantage of the sub-electrode array.

One can take advantage of the spacings or gaps 23 between the sub-electrodes to allow a gas 31 to flow between the sub- electrodes and into the reaction volume. For this, a process gas 31 is introduced through a piping 28 into the area behind the sub-electrode array. Means may be provided to uniformly distribute the inlet gas 31 to the gas outlets 23 as e.g.

realized by the grid shield 27.

Some further considerations to the RF feed system: From the sub-electrode groups provided, at least one with at least one sub-electrode is to be RF-voltage driven.

The inputs to the groups of sub-electrodes for RF voltage shall be realized through vacuum feed-throughs 41 as shown in fig. 20. Thereby, as is customary for RF feed, as was dis- cussed there is provided a matching box 42 matching the out- put of amplitude and/or phase adjustable or modulatable RF generators 43 and 44 which generate narrow-band or wide-band signal spectra. Thus, one generator 44 may e.g. be adjustable in its sole amplitude of its output RF signal whereas the second generator 43 may be adjustable in phasing at a single spectral frequency. Phase shift of the two RF signals input to two or more groups of the sub-electrode array may also be performed within the matching box 42. For the skilled artisan there are many different ways for implementing the RF drive to the reactor according to the present invention and as shown in fig. 8.

A today preferred technique is schematically shown in fig.

22. There is provided one RF oscillator 47, the one frequency RF output thereof being led via an adjustable phase shifter unit 48 on a power amplifier 49 and additionally directly on the input of a second power amplifier 50. Preferably, the amplifiers 49 and 50 are tailored to accept large reflected power and they are coupled to the RF inputs of the reactor via simple passive matching circuitry in order to roughly match the plasma impedance to the generator impedance. An- other preferred design is shown in fig. 23 where two ampli- fier output stages 55 and 56 are driven by the same - one frequency - RF oscillator but provide for independent adjust- ment of amplitude and/or phasing of the output RF voltage.

The output of the amplifier stage 55 is coupled through a matchbox to the primary winding 51 of a transformer. The sec- ondary winding 60 of the transformer provides an RF voltage difference between the group inputs to the sub-electrode groups according to the present invention. The central point of secondary winding 52 is connected via a matchbox 54 to the output of the second amplifier output stage 56. Phase control and/or amplitude control is preferably performed at amplifier stages 56 and/or 55.

Thus, the most generic aspect of the reactor according to the present invention is to provide at least one sectioned elec- trode. In spite of the preferred examples described it might be interesting not to evenly space the sub-electrodes e.g.

for resolving general etching problems with respect to homo- geneous treatment distribution. Sub-electrodes of annular shape or of frame-like shape may be arranged around a central axis. Such a centric grouping of the sub-electrodes is sche- matically shown in fig. 24.

Further, it must be pointed out that it is absolutely possi- ble to operate one or the other or more than one of the sub- electrode groups at a predetermined non-RF potential such as on DC. It is further evident that the RF generator arran- gements with phase and amplitude and possibly even frequency adjustment may be realized digitally thereby providing a po- wer amplifier output stage to feed the sub-electrode group inputs. Also, as the inventive concept allows for separate independent inputs, it is well suited to drive the different sub-electrode groups by differently shaped voltages according to different signal spectra, briefly, one has complete free- dom to operate the sub-electrode groups at any electric sig- nal desired and mutually independently.

It is further to be pointed out that e.g. for RIE applica- tions the reactor according to the present invention and with corrugated surface enlarging sub-electrodes may be operated with all sub-electrode groups on the same RF voltage whereby the mere fact of enlarged electrode surface at the structured electrode arrangement significantly improves ion bombardment of the substrate deposited on the counter electrode arrange- ment according to arrangement 20 of fig. 7. In RIE applica- tion and making use of a reactor as shown in fig. 13, thereby feeding all sub-electrode groups with the same RF voltage, it was possible to treat glass substrates of 370 x 370 mm extent by carbon deposition with a coating uniformity according to maximum deviations of 5 t from the desired thickens. After treatment, the reactor was cleaned by an oxygen plasma.

Further consideration must be given to cooling and heating of the substrate to be treated. Thus, and according to electrode 20 of fig. 13, cooling/heating is preferably accurately con- trolled so as to reach different substrate temperatures, de- pending whether the substrate is coated as by chemical vapour deposition in the inventive reactor or is etched. For etching substrates, and especially substrates which are provided with a photolacquer layer, the substrate temperature is to be maintained preferably below photolacquer melting temperature.

Thus, for such operation, the substrate, e.g. on electrode 20 of Fig. 13, is cooled by the cooling system 12 to a tempera- ture in the range of 400C to 800C.

If the reactor is used four layer deposition on the substrate as for performing plasma enhanced vapour deposition, then the substrate is preferably heated, e.g. to about 2000C. This considerably improves the quality of the deposited layers in that the layers are deposited with reduced stress, necessita- ting no additional heat treatment. It is thereby rather un- customary to heat the substrates during PECVD to such rela- tively high temperatures.

Looking back on Fig. 7, of V1l, V12 , V13 ... .

at least one comprises an effective RF spectrum all signals at a one-side electrode (substrate or no-substrate electrode) may be the same DC, AC, pulsating ramp and all kinds of signals resulting from super-position may be applied, as DC + AC + ramp.

Especially in the embodiment according to Fig. 9 the sub- electrode groups may be driven at one common signal whereby for such a case the sub-electrodes might be conductively lin- ked to form a corrugated electrode area so as to enlarge sur- face area. The same is valid for structures as shown in Figs.

14 to 19.

Up to now we did not describe any features at the inventive reactor to deal with border effects with respect to homogene- ity of substrate treatment. In fig. 25 there is shown an in- ventive reactor according to fig. 9 and electrically driven generically as was discussed in connection with fig. 8, whereby there is provided besides the two groups of sub- electrodes driven at Vll, Vl2 a border electrode 50 of e.g.

triangular cross-sectional shape as shown. This border elec- trode 50 is generically electrically operated via a matchbox MBB and a generator GB. Thereby, again, the generator GB may generate a signal of predetermined spectrum, possibly fre- quency, phase or amplitude modulated, thereby may generate a single frequency signal or such generator GB may be just omitted, just maintaining the matchbox connected to ground potential. Again the matchbox considerably contributes to the DC bias according to the zero spectral frequency amplitude of the signal V, prevailing at border electrode 50.

By means of the border or peripheral electrode 50, preferably surrounding the subdivided electrode arrangement 10, and ei- ther passively connected preferably via a matchbox to refer- ence potential or actively driven by an additional generator, a considerable improvement with respect to treatment uniform- ity is reached and especially with respect to etching uni- formity at a substrate. Thereby, the border electrode 50 may even be left on floating potential. Preferred operation of the border electrode 50 is nevertheless via a matching box connected to reference potential with or without additional generator GB.

A reactor according to fig. 25 was operated as follows: Vll: RF; 13,56 MHz, 0 V DC bias, lkV RF power V20: linked to ground potential via a passive adjustable matchbox connected to ground potential via a passive adjustable matchbox MBB In a CF4/SF6 a-Si: H-atmosphere, at 5 x 10-2 mbar, the distri- bution of etching effect on 400 mm x 400 mm substrates was tested.

Fig. 26 shows the distribution at: DC bias of V20: -215 V DC bias of VBO: +107 V DC bias of V1l, V12 : 0 V Fig. 27 shows the distribution at: DC bias of V20: -320 V DC bias of VBO: -83 V DC bias of V11, V12 : 0 V Fig. 28 shows the distribution at: DC bias of V20: -312 V DC bias of VBO: +3 V.

DC bias of V11, V12 : 0 V Therefrom, it is clearly seen that the distribution of proc- ess effect along the workpiece may accurately be adjusted by appropriately selecting the respective DC bias signal applied to either of the electrodes.