DE2513945A1 - PROCESS FOR PASSIVATING THE SURFACES OF SEMI-CONDUCTIVE COMPONENTS - Google Patents

Description

2 5 1 3 9 /, Dipl.-Ing. H. Sauenland · Dr.-lng. R. König · Dipl.-Ing. K. Bergen Patent Attorneys ■ 4ooo Düsseldorf 3D · Cecilienallee 76 ■ Telephone 43273a

25 _β March 1975 29 386 B.

RCA Corporation, 30 Rockefeller Plaza, New York, NY 10020 (V.St.A ₀ )

"Process for passivating the surfaces of semiconductor components"

Although silicon dioxide has hitherto been used on a large scale to passivate the exposed surfaces of silicon semiconductor components, in particular to passivate the exposed edges of PN junctions, it has proven necessary for many applications to provide additional protection to give component, such as longer durability against humidity and other contaminants "However, it is also desirable different between regions conductivity type, such as for example between the base and collector regions of bipolar transistors, as low as possible or _e ensure lower surface leakage currents ,,

Additional protection against contamination and lower leakage currents have so far been achieved in that the Components have been coated or encapsulated with various plastics and glasses. When pulling over or encapsulation with glass, which is often required before encapsulation with plastic, has proven to be special proved desirable, dispersions of very small glass particles in a carrier liquid

509844 / Q731

to be used to deposit coatings of glass particles on the component surfaces and then to a temperature just high enough to fuse the glass particles into a continuous layer.

Glass, which can be used for the passivation of semiconductor components, certain properties must have, for example, ₀ compatible with the respective semiconductor component melting temperature, good adhesion to the component surface, a thermal expansion coefficient which is at least approximately equal to that of the semiconductor, low porosity and the absence of components or Impurities that could adversely affect the electrical properties of the component, »

There have been known various methods, "with which glass particles may be applied to the surface to be coated ₀ One of these methods is that a glass-made powder slurry is removed by a doctor blade on the corresponding surface of a semiconductor wafer, which produced a large number by known diffusion method Contains components, and that the liquid is then evaporated.In another method, the glass particles are allowed to settle from a liquid dispersion onto the component surface.

Applying with a doctor blade has various disadvantages. In most cases it has to be carried out by hand, which is very time-consuming; also can be treated only on one face of the disc, that is _e a simultaneous coating of both sides is not possible "Since in this method takes place a mechanical scraping, occur relatively often sliver breakage and other damage a. In addition, this

509844/0731

Method of covering limited or selected areas only with the aid of a very specific surface shape take place, i. this process is based on coating recessed or recessed areas limited.

The settling process is very time-consuming, requires a disproportionate amount of glass particles and can cannot be carried out selectively, i.e. it cannot be restricted to certain areas of the semiconductor component come into use, Another one that has also been used so far Method is electrophoresis,, in this one In the process, the component (with either positive or negative polarity) is immersed in a dispersion of glass particles, which is charged opposite to the component to be coated, and a current through the dispersion directed. Glass can only be deposited selectively on the conductive parts of the component surface »Except there is the disadvantage that it is not possible to deposit glass selectively on insulating areas Another difficulty with this method is that it does not produce thick layers satisfactorily due to a lack of liability. In order to be able to work selectively, this: em procedure in addition, conductive liquids are used, which can dissolve more impurities than non-conductive liquids Liquids do it, so that the liquids used here pose a greater risk of component contamination represent,,

The object of the present invention is to propose a method that does not have the aforementioned disadvantages, in particular the production of relatively thick, reliably adhering glass coatings on certain, selected areas allowed. This task will

5098A4 / 0731

solved according to the invention in that the insulating material is charged with a certain polarity that the charged component is immersed in a liquid that is an insulating carrier liquid and dispersed Contains glass particles that carry a charge "of a certain polarity, so that the glass particles are selectively precipitate either on the exposed semiconductor areas or on the areas covered with insulating material, and that the glass-coated component is removed from the liquid, dried and at a temperature fired high enough to fuse the glass »

The invention will be explained with reference to the accompanying drawings, in which preferred embodiments are shown explained in more detail. Show it:

FIG _o 1 a portion of a semiconductor wafer, in which a plurality of mutually separated by grooves transistors are located, in cross-section;

2, 3 and 4 successive steps in the application of an embodiment of the method according to the invention in connection with the processing of a wafer according to FIG. _Q 1;

5 to 8 method steps of another possible application of the method to a semiconductor wafer;

9 to 12 show another possible application of the method according to the invention, with another component being coated;

13 shows a charger suitable for carrying out the method according to the invention, in side view; and

5098AA / 0731

FIG. 14 is an elevational view of a portion of the apparatus shown in FIG.

Although the inventive method for production is suitable for a large number of different types of semiconductor components, it will first be described below using the example the fabrication of a large number of bipolar transistors on a single semiconductor crystal wafer.

As can be seen from FIG. 1, a large number of transistors 4 can be produced in a wafer 2 made of monocrystalline semiconductor material, such as silicon, by means of known diffusion processes or a combination of epitaxial growth and diffusion processes. Each transistor 4 has an emitter region 6, for example n-conductive, a base region 8, which can be p-conductive, and a collector region 10, which in turn can be η-conductive _o A layer 12 of insulating material covers the upper surface 14 of each transistor _e A further layer 13 of insulating material covering the lower surface 15 of the disc. The emitter region 6 and the base region 8 are separated from one another by a pn junction 16. The base region 8 and the collector region 10 are separated from one another by a further pn junction 18. In the form of a grid of the component slots 20 are provided in the upper surface 14, extending to below the pn junction 18, wherein the pattern is applied so that the grooves are located on the four sides of each transistor 4 _e It is advantageous to cover the exposed areas of the pn junctions with a dielectric material in order to reduce leakage currents through the junctions and prevent destruction by the surrounding atom

509844/0731

- Ό -

atmosphere or impurities during manufacture to prevent. It is much more economical to have the necessary protective layer over the pn junctions Provide time at which the components 4 are still part of the original disc 2, as each Treat the component after it has been separated from the wafer.

If grooves are etched into the silicon of a mesa wafer, the SiO ₂ covering the wafer (if this is the insulating material) is first applied in accordance with the groove geometry, using known lithographic processes. Then the silicon is etched SiO _{2 is used} as the pattern-defining cover mask. Since undercutting occurs during etching, an overhanging plate of SiO _{2 is created} . When the glass is applied in the manner according to the invention, cavities can arise under the plate, which can lead to a deterioration in the electrical properties during the further manufacturing steps. It is therefore advantageous _{to remove the overhanging SiO 2} before loading and glass coating takes place.

The overhanging parts can be removed in at least two ways. For example, the oxide can be etched, with the wafer remaining immersed until the overhanging end is removed. In this case, although the thickness of the oxide on the mesa surface is also reduced by a fraction, this is not harmful. For example, a three-minute etch in buffered HF solution would be sufficient to remove an overhang of 25 ii , while at the same time only an un-

509 844/0731

Around 3000 2. of the originally 12000 Å thick SiO ₂ layer on the mesa surface must be removed. Another way to remove the overhang is to use ultrasound. For example, a Branson ultrasonic cleaner (model 41-4000) was used to remove the overhang in 30 seconds with water and in 10 seconds with Freon TF «.

According to a preferred embodiment of the invention the parts of the uncovered silicon having Scheibchens which are exposed within the grooves 20 are, with a glass protective layer as follows covered ₀ The disc 2 is to put into a device for glow discharge, which consists of two parallel arrays of thin, vertical arranged wires that are held in an insulated frame. The major surfaces of the disk should be carefully aligned so that they are parallel to the wire assemblies. The wires are preferably about 0.04 mm thick, made of tungsten, and spaced about 12.5 mm apart. The distance between the parallel wire arrangements is selected so that when there is a semiconductor wafer 2 in between, the wires are preferably 5.6 mm away from the closest surfaces of the wafer. The level of charge on the wafer can be varied via the distance between the wires and the surface to be charged, although it is usually desirable to charge the insulating area to saturation.

In order to obtain the desired charging results, a charger should preferably be used, the exactly parallel and in terms of the distance the same positioning of the surfaces of the disc

098 ^^ / 0731

25139AB

chens with respect to the two wire arrangements in the charger allowed »A suitable charger is shown in FIGS. 13 and 14

As can be seen from FIG. 13, the charger 58 has a base plate 60 on which two frames 62 and 64 are made attached to insulating material such as methyl methacrylate resin are. The frames are set so precisely with regard to their spacing on the base plate 60, that they are directed towards each other. Each of the frames 62 and 64 has a foot part 66 and 68, respectively, a plate 70 and 72, respectively, which are fixed vertically along one edge of the foot parts 66 and 68, and a top plate 74 or 76, which are each mounted on the upper narrow side of the vertical plates 70 and 72, so that they extend parallel to the foot parts 66 and 68.

Each frame 62 and 64 is provided with a wire assembly, and 80, each consisting of nine mutually equidistant arranged wires which werc.en held in a vertical plane under the train _o One end of each wire is near the outside edge of the foot part 66 and 68 attached. The other end of each wire is attached to one end of a tension spring 86. Set screws 88 are retained in latches 92 and 94, which in turn are embedded in plastic blocks 82 and 84, respectively, which extend along the outer edges of top panels 74 and 76, respectively.

The wires of assemblies 78 and 80 are routed over the outer edges of top plates 74 and 76, respectively.

The latches 92 and 94 are provided with a not shown High voltage source connected.

5098 ^ / 0731

2,513 9

During loading, the disc 2 is held vertically between the wires 78 and 80, for example by means of a suspension 96 made up of four supports 98, a pair of horizontal metal guides 100 and 102 arranged parallel to and above the upper plates 74 and 76 with smooth , sloping inner surfaces 104 and 106 and a disc _{holder 108 (Fig o} 14) consists o

The disc holder 108 has a V-shaped metal block 110, on the underside of which a pair of pivotable ones Grippers 112 and 114 is attached. The gripper 112 has a grooved one at its lower end Finger 116 on, while the gripper 114 two spaced similarly grooved fingers 118 and 120 are arranged from one another. The fingers 116, 118 and 120 are arranged so that they can receive the disc 2 and hold it in a vertical position.

The block 110 has two smooth side surfaces which run at an angle that corresponds to the slope of the surface 104 and 106 of guides 100 and 102 corresponds. In this way, the block 110 in the guides 100 and 102 are carried and moved, with electrical contact being made at the same time »The guides are grounded

During loading, the wafer holder is preferably reciprocated, at a speed of about 7.5 cm / sec, in order to uniformly load the wafer surfaces _{or the like}

A negative voltage of approximately 6200 volts is applied to the wires. The atmosphere is air or stick

5 09844/0731

material, but the relative humidity is controlled to around 30% at 25 C if silicon dioxide is used as the insulating material. The charging voltage should preferably be as high as possible without causing an arc, since it has been found that better edge delimitation of the charge on the disc 2 and a uniform charge distribution can be achieved in this way. The upper permissible limit of the relative humidity is a function of the time elapsing between the loading of the surface of the wafer and the subsequent immersion in a glass dispersion, for which a period of one second has been found to be particularly advantageous _o If the elapsing time is very short, The relative humidity can be higher because there is less time for the charge to degrade ₀ The relative humidity can be higher even if the insulating material to be charged is hydrophobic. When using a photoresist sold under the name Waycoat SC 180 as insulating material, the relative humidity can be 65% or even more.

The disc located in the holder is grounded through its contact, which it has on its edge with the holder and is exposed to discharge for approximately 5 to 15 seconds. As can be seen from Fig. 2, this leads to a layer 22 of negatively charged gaseous ions on the insulating layer 12, which is on the is located upper surface 14 of the component assembly, and to a similar ion layer 24 on the insulating Layer 13 | those on the lower surface 15 of the disc 2 is provided.

The charge can either be negative or positive, depending on whether the

509844/0731

2 5Ί 3 9 4b

hitting glass particles in the liquid have a negative or positive charge and whether the glass is on the Insulating layers 12 and 13 or on the uncovered, exposed silicon should be provided. In the present example, the corona charge is negative.

The ions 22 and 24 of the corona discharge reach the surfaces of the disc 2 and strike the Surfaces of the insulating layers 12 and 13 down (Fig.2). The ions reaching the semiconductor surfaces are however, they are neutralized and do not form a surface charge there. Although the bare silicon surfaces are the grooves 20 immediately after its contact with air a very thin oxide layer 26 with a thickness of about 20 2. begin, this layer 26 is so thin that it does not act as an insulating material against the gas ions works. Therefore, no charge layer is formed within the grooves 20. To hold cargo and to avoid it of tiny holes, the oxide layer should be at least 1000S thick.

A suitable dispersion of the glass particles is preferably made up about a day before charging. Suitable types of glass for passivating semiconductor components in the context of the present invention are commercially available under the following names: No. 7723 from Corning Glass Co ₀ , IP 760 and IP 820 from Innotech Co. The first-mentioned type of glass has approximately the following composition: 30% PbO, 7 % Al ₂ O, 13% B ₂ O, and 50% SiO ₂ . IP 760 glass is roughly composed as follows: 45.75% PbO, 2.65% Al ₂ O ₃ , 1.60% ZnO, 10.75% B ₂ O ₃ and 39.25% SiO ₂ . IP 820 glass is finally made up as follows: 50% PbO, 10.1% Al ₂ O, and 39.9% SiO ₂ . The percentages mean all

509844/0731

2 b13 9

Weight percent. In the case of the types of IP glass, around 75% by weight of the glass particles have a size below 12% and around 25% have a particle size of less than 3%

The glass powder is dispersed in an insulating liquid containing a charging agent. A suitable insulating liquid is a halogenated hydrocarbon such as Genesolve D (Allied Chemical Co.) or Freon TF (du Pont). Both are 1,1,2-trichloro-1,2,2-trifluoroethane. Another suitable liquid is Isopar G (Exxon Corp.). This is a mixture of liquid isoparaffinic hydrocarbons and can also be referred to as very high-purity isoparaffinic hydrocarbons which are narrowly fractionated. Genesolve D and Freon TF are preferred as carrier liquids because they are denser than Isopar, evaporate quickly, and are non-flammable. When using Freon TF, the evaporation takes place so quickly that the disc can be put into the oven immediately. When using Isopar G, it is necessary to let the disc dry for a few minutes before putting it in the oven _e

The charging agent can be a surface treatment agent, for example one of those in the table below specified.

5098 ^ 4/0731

2 5

3 9

- 13 -

Table treatment agents

1. Zirconium octoate

2. Petroleum magnesium sulfonate

3. Petroleum barium sulfonate

4. Zinc octoate

5. * OLOA 1200 (Chevron, Inc.)

6. Petroleum ammonium sulfonate

7. N-alkylpiperizene - monoalkenyl succinimide

8. RNHCH ₂ CH ₂ NH ₂ , where R is a polybutyl _{radical 0}

The charge induced on the glass particles

negative

positive (changes to negative after a few days) negative negative negative positive

positive negative

* OLOA 1200 has the following structural formula:

- C

(C.

1 '

^Χ HHH

whereby

ρ

, H-C-H

I H

with nitrogen 2.1% by weight and an alkalinity value of 43,

509844/0731

2 5 Ί 3 9 4 S - 14 -

On some samples of the IP-820 glass, it turned out to be turned out to be advantageous, an ultrasound treatment of about 5 minutes, the energy introduced into the glass dispersion being about 100 Watt was to break up agglomerates of the finest particles. Unless this is done, the Agglomerates to give the applied glass layer a bumpy appearance and what is particularly disadvantageous is to partially adhere to areas where glass is not desired.

For the examples described below, a base solution was prepared by first adding 100 grams of OLOA 1200 was dissolved in 100 ml of Freon TF and, secondly, 8 ml of this solution was added to 392 ml of Freon TF.

example 1

In order to apply IP 760 glass to power transistor slices of the mesa structure, the mesa surfaces are provided with a double coating of photoresist, with a total thickness of 5 _<t ° glass must be applied as a layer in the grooves, as already described.

A mixture of glass powder is made by adding 7 ml of the base solution given above to 450 ml of Freon and then adding 4 g of glass powder and shaking for two minutes prepares »This mixture becomes the Stabilization left for at least one day »

To apply the glass particles, the pane is first negatively charged and then in the eirm for 6 seconds the aforementioned mixture containing beaker made of Jena glass immersed. The contents of the beaker become moderate with

Β098ΛΑ / 0731

251394b

Stirred speed while the glass settles.

As shown in FIG. 3, a glass layer 28 is deposited in the grooves 20, but not on the insulating layer Layers 12 and 13.

After the glass is deposited, the volatile constituents (including photo resist) β After are evaporated at about 525 ⁰ C or burnt off and the glass is then melted to form a layer 28 * (FIG. 4) the fusion has the glass layer 28 · a thickness of about 40 ₁ .

Example 2

In this case, IP 820 glass was provided on thyristor disks (not shown) with a mesa structure. Grooves were made on both surfaces of the disk so that mesa structures were on both sides. The mesa bodies were coated with a 1.2 / t thick layer of SiO ₂ .

A glass powder dispersion was prepared by adding 10 ml of the base solution to 400 ml of Freon and then adding were shaken for two minutes after adding 8 g of glass powder.

After the glass dispersion had been able to stand for a day, the application of the glass was carried out in the manner described in Example 1 with the exception that the loaded slices are placed in the glass mixture for 10 seconds have been dived. After the glass was deposited, the volatiles were evaporated and the glass fused.

5098Λ4 / 0731

2,513 - 16 -

The parts made in this way had excellent junction coverings with a layer of molten glass approximately 30 / <thick. Studies related to the electrical properties showed that the collector-base junction having a breakdown voltage of more than 1000 volts and leakage currents of about 4 AA at room temperature and about 38 AA at 100 ⁰ C.

Example 3

In certain cases it is desirable to passivate a semiconductor device with a type of glass which is chosen very obvious thermal expansion coefficient due to their the semiconductor ₀ Such a glass may contain boron and require high melting temperature, so that the Dotiermuster of the silicon is impaired. In such cases it is desirable to deposit the glass on a SiOp layer instead of directly on the silicon in order to shield the silicon from the doping influence of the glass.

In this example (FIG. 5), the wafer 2 has a SiOp coating 30 with a thickness of 8000 Å, namely only in the grooves 20. This can be achieved by known masking methods. Besides, the disc is because of the air surrounding it, inherently on the upper mesa surface with a very thin oxide layer 31 while a similar oxide layer 33 is on the lower surface. That coated Slice is then placed in dry air (relative humidity 27%) exposed to a positive corona discharge. This leads to a layer of positive gas ions 32 (FIG. 6) on the SiOg layer 30 in the grooves 20 the thin oxide layers 31 and 33 are not charged.,

B098AWQ731

A mixture is made from 4.5 ml of the basic solution of solvent and charging agent by adding to 300 ml of Freon manufactured. To this solution, 9 g of Corning No. 7723 glass powder are added and the mixture for two minutes shaken to disperse the powder.

The charged disc is immersed in the aforementioned dispersion for 10 seconds, resulting in a precipitate a layer 34 made of glass particles, specifically only on the SiOp layers 30 (FIG. 7).

The disc is then removed from the dispersion and the remaining volatile constituents are removed by heat treatment driven out and the glass fused to form a layer 34 '(Fig. 8). The thickness of the fused Layer is about 35 /. .

Example 4

In this case, a glass layer is provided in the grooves of a diode wafer of the mesa type in order to reduce the breakdown voltage to increase the diodes »

According to FIG. 9, a silicon diode wafer 36 has a p-conducting lower layer 38 and an n-conducting one upper layer 40, which are separated by a pn junction 41. In the washer 36 there is a network of grooves which extend to the p-type layer 38.

A relatively thick SiO ₂ layer 42 covers the silicon in the grooves, and the wafer is positively charged in the manner described, so that a layer of positively charged ions as in Example 3 on the

509844/0731

18- 2 5 Ί 3 9 4 p

Forms surfaces. Set the upper mesa surfaces 46 a very thin oxide layer 48, while the lower surface 50 of the wafer is of a very thin one Oxide layer 52 is coated; however, these thin layers do not act to store charges Layers. To apply the glass, a dispersion was used which was obtained by adding 10 ml of the base solution of carrier liquid and charging agent to 400 ml of Freon and then adding 6 g of IP 760 Glass powder with subsequent two-minute shaking.

The charged disk was immersed in this dispersion for 6 seconds while the mixture was stirred. The disk was removed from the dispersion with a layer of glass particles 44 which had only _{deposited on the thick SiO 2} layers 42. Here, too, the volatile constituents were evaporated and the glass fused to form a layer 44 '(FIG. 10). The charging and depositing of glass was then repeated to apply a second layer of glass particles 54 to the first layer of glass 44 '(Figure 11); this second layer was fused so that the two layers formed a glass composite layer 56 (Fig. 12). In the glass dispersion that is used to deposit the particles for the second glass layer, the glass particles carry a charge which is opposite to the charge on the first glass layer.

With diodes manufactured in this way, breakdown voltages of up to 1700 volts were measured at room temperature »

The environment for the charging process can be both drier Nitrogen and dry air are used,

5 0 9 8 4 4/0731

251394b

dry nitrogen is somewhat preferable in this respect when the formation of ozone due to the discharge is considerably lower, and there is somewhat higher Voltages can be used without the risk of arcing.

The amount of charge agent in the carrier liquid should just be measured so that on each glass particle the same sign of charge is developed when the strongest glass deposit is desired. if more than the optimal amount is added, excess ions are introduced, resulting in a decrease of the possible glass precipitation, since the excess leads to the charge on the insulating areas neutralize the disc. However, the thickness of the glass deposit can be determined by the amount of the charge agent be varied in an advantageous manner.

To determine the amount of charge agent to be used, a suitable glass dispersion is prepared, a small amount of the charging agent added and a disk with a charge pattern in the dispersion submerged. If the amount of charge is too low, glass will adhere to both the charged, the insulating surfaces as well as on the uncharged surfaces. It will then be another sample of the Taken dispersion and added a slightly larger amount of charging agent. Again a disc with a Charge samples immersed in the dispersion and the results examined. Loading equipment will be slightly in each case increasing levels of fresh amounts of dispersion are added until the glass is only either on the charged one Areas or the uncharged areas - depending on the charge polarities - precipitates. Once the optimal amount of loading equipment for one

5 (19844/0731

2 5 1 3 9 A

given weight of a certain glass in freon has been found, it is possible to use the amounts of loading equipment required for other quantities of glass to approximate a linear extrapolation.

Although silicon dioxide and photoresists are useful materials for insulating layers in the examples described have been mentioned, other insulating materials that are conventional in semiconductor components can also be used Find use, be used, such as silicon nitride. When the insulating material is an organic substance (as in Example 1), it must be removed before the glass particles are fused.

The method can also be used to deposit glass particles simultaneously on the insulator-coated areas on one side of a semiconductor wafer (where the insulator is made of inorganic material) and on the bare, uncovered semiconductor areas on the opposite side of the wafer. This can be achieved by placing charges of opposite polarity on the insulator-related areas on the opposite sides and then dipping the wafer in a glass dispersion _e

509844 / Q731

Claims (12)

RCA Corporation, 30 Rockefeller Plaza, New York , NY 10020 (V. 1-iA) Patent claims: f 1.J Method for the selective application of a glass layer either on exposed semiconductor areas or on areas of a semiconductor component coated with an insulating layer, which has both types of regions, characterized in that the insulating material has a certain polarity is charged that the charged component is immersed in a liquid which is an insulating carrier liquid and contains dispersed glass particles which carry a charge of a certain polarity, so that the Glass particles settle selectively either on the exposed semiconductor areas or on those with insulating material precipitate coated areas, and that the glass-coated component removed from the liquid, dried and firing at a temperature high enough to fuse the glass. 2. The method according to claim 1, characterized in that the charges on the insulating material have the same sign as the charges on the glass particles, so that the glass particles precipitate on the exposed areas of the semiconductor. 3. The method according to claim 1, characterized in that the semiconductor is silicon and the insulating material is silicon dioxide. 5098A4 / 0731 4. The method according to claim 2, characterized in that the insulating material is a photoresist. 5. The method according to claim 1, characterized that the insulating material is silicon nitride. 6. The method according to claim 1, characterized in that that the insulating liquid is 1,1,2-trichloro-1,2,2-trifluoroethylene. 7. The method according to claim 6, characterized in that the liquid is a charging agent contains, preferably zirconium octoate. 8. The method according to claim 6, characterized in that the liquid is a Contains charge means of the following structural formula: HO I // R-C-C C-C I \ H Ό HH
I i
"N- (CC I I H H HHH , i-A- H H whereby
R. with nitrogen 2.1% by weight and an alkalinity value of 509844/0731 9. The method according to claim 1, characterized in that the glass contains approximately 30% PbO, 7% Al ₂ O ₃ , 13% B ₂ O ₃ and 50% SiO ₂ (all percentages by weight). 10. The method according to claim 1, characterized that the relative humidity of the environment is kept so low that a discharge of the charged surface of the insulating Material in the period between loading and immersion in the dispersion liquid is avoided. 11. The method according to claim 10, characterized in that that the relative humidity is about 25 to 30%. 12. The method according to claim 1, characterized that the gas ions and the glass particles are charged with opposite polarities are so that the glass particles settle on the insulating material. 5 098AA / 0731 ι Le he