BE585434A - - Google Patents

Description

       

   <Desc / Clms Page number 1>
 
 EMI1.1
 



  0 PUPARATIOX OF PURIFIED SEMI'CONDUCTOR PROFESSIONS ste

 <Desc / Clms Page number 2>

 
The present invention relates to the purification of germanium and silicon compounds containing chlorides. Of particular interest as intermediate products in the preparation of the corresponding elements are purified halides according to the invention.

   Although in general, due to the rapid increase in the field of application of semiconductor devices, the interest will be centered on the ordinary preparation of these elemental materials, the purified halides prepared according to the present invention are of interest for '' other techniques,

   for example for the preparation of ultra pure silane
Today's severe design requirements for semiconductor devices such as transistors and rectifiers have resulted in an increasingly pure demand for * semiconductor material *. The growing importance of this field has had a considerable influence on ancillary fields of metallurgical treatment leading to the development of purification processes reducing the levels of impurities of the materials thus treated to infinitesimal limits, which it was believed until now that it could not be obtained.

   Procedures have been developed to reduce the impurity content of these materials to one part per thousand to one part per trillion and even lower orders of magnitude, and there are still separation problems which need to be resolved before many devices. the most recently designed semi-conduc- tor can be produced commercially

 <Desc / Clms Page number 3>

 
A particular problem arises in the treatment of silicon and germanium, the impurities of which must have these distribution coefficients in
 EMI3.1
 matter between liquid phase and solid phase which ssgcochent of the unit. In these cases, the use of ordinary zonal infusion techniques does not provide a complete answer.

   In particular, where materials undergoing treatment melt at high temperature and / or react with one are contaminated with well-known crucible materials, such as
 EMI3.2
 so that the use of molten regions not in contact with the solids retaining materials is required that the zone melting process must form the form of a floating zone technique or other related process, the problem posed is serious.
 EMI3.3
 



  When such aolide-liquid distribution coefficients are unfavorable and when, for others: 3tisons, the area-wise melt treatment becomes a step - treatment;
 EMI3.4
 cotizeuae, the possibility of purifying Haat gas phase compositions from a semi-cosmetic material was investigated.



    One of these processes, reduction by hydrogen does not ma halogenated titles such as sodium tetrachloride and the like. halides containing one or more atoms of hydrogen, such as silicon chloroform, in widespread commercial use. Others from 28
 EMI3.5
 The processes include the hydrolysis of germanium tetrachloride to obtain the corresponding oxide.

   We found that such
 EMI3.6
 reduction of the tantalum or tungse filaments, followed by leaching of the filament material from the crystal, and ultimately by floating zone melting to produce the desired crystal configuration, results in the practical efficient removal of all
 EMI3.7
 169 1apuret48 important elemental product. ra

 <Desc / Clms Page number 4>

 The favorable character of the distribution or vapor phase distribution coefficients of many impurities in these halogenated semiconductor materials has resulted in other procedures using these intermediates as well.

   An example of this procedure is the removal of arsenic acid from germanium titrating by distillation as well as by shrinkage procedures II. and depending on the nature of the starting materials in the procedures employed in the preparation of the chlorinated starting materials, that certain impurities are chlorinated together with the main material and that some of these materials are only removed with difficulty. Impurity materials of this nature, of particular importance from the point of view of a semiconductor device, are boron and phosphorus, both of which form chlorides which are reduced on tantalum and tungsten by reduction with hydrogen at the same time as the semiconductor.

   Although some of the most readily available conductive materials are produced by this process, the continued presence of small amounts of these impurities is a limitation on the use of the device.



  This limitation in the hydrogen reduction process is considered to be of growing importance as the modern design of devices becomes ever more demanding.



   According to the invention, techniques have been found to reduce the content of impurities to levels which have not been achievable on a production scale heretofore. Elementary materials such as silicon, produced according to the invention, can

 <Desc / Clms Page number 5>

 see resistivity levels of up to 10,000 ohm-cm or more and the content of towering impurities can drop as low as 1 x 10 12 atoms per cubic centimeter.

     The procedures described herein, ensuring an end product improved in this way, include the treatment of the liquid phase of an initial halogenated material and may also include a perfect reduction process which dispenses with the use of filament materials. such as tantalum and tungsten which are generally used
According to the invention, it has been found that elemental germanium and silicon having an extremely low content of boron and phosphorus can be prepared by the treatment of corresponding liquid halogenated materials such as tetrachlorides and various chlorosilanes. with certain absorbent materials. These operating processes are preferably carried out in absorption columns.

   Although in general the apparent effect of these absorbents is to remove equal amounts of boron and phosphorus, a subsequent pre-treatment process described herein can lead to greater boron removal which further purifies the end product. . A further aspect of the invention involves the reduction with hydrogen of these halogenated materials on the filaments or thin rods of semiconductor dobbies themselves. Another pre-treatment process described herein, too, results in the efficient removal of the sulfur-containing material, thus avoiding any complications which may be introduced by these contaminants during the subsequent treatment.



   Accordingly, the invention is a method of purifying a liquid phase compound containing chlorine from a

 <Desc / Clms Page number 6>

 element selected from the group comprising silicon and germanium, containing as contaminant at least one chloride of an element of the group comprising boron and phosphorus. In the present invention, the liquid material is treated with at least one absorbent selected from the group consisting of charcoal, hydrated oxides and hydrated silicates, such that intimate contact is produced between the material and the absorbent. , A final step is the reduction of the material thus treated by reduction with hydrogen in the vapor phase, on a mass of a chosen element, comprising silicon or germanium., Corresponding to the element chosen from the same group mentioned above.



   The invention will be understood more easily by referring to the accompanying drawings in which:
Figure 1 is a schematic front elevational view, partly in section, of one type of absorption column which has been found to be suitable for practicing the invention, and
Figure 2 is a front elevational view, partly in section, of an apparatus suitable for use in the hydrogen reduction of halogenated materials on silicum or germanium filaments.



   Referring further to Figure 1, there is shown an absorption column 1 made of pyrex or other suitable material, which is filled to the indicated height with absorbent material 2, and which further contains a small amount of quartz wool or other filter material 3 at its lower end to prevent escape of absorbent material. In operation, the halogenated material 4 which is to be treated is introduced into the tank 5 which is provided with a

 <Desc / Clms Page number 7>

 plug 6 having a vent.

   A venting arrangement 7 allows a uniform flow of the calcined material 4 from the tank 5 to the column 1, where this material may first be at the top of a group part 2 at the top of the column. column in region 8, depending on flow conditions. The flow in the column is controlled by the tap 10 in the siphon tube 11 which connects the column 1 to the column reservoir 12. The material flowing through the tap 10 passes through the flow nozzle.
13 and through tube 14 to enter column reservoir 12. Blockages are prevented by vent tube 15.



   Figure 2 illustrates one type of apparatus suitable for use in the preferred process for reducing purified halogenated materials to their single elements according to the invention. This apparatus, although preferred over ordinary tantalum or tungsten filament apparatus, is not necessary for the invention. Ordinarily reduced materials may have sufficient purity to be suitable for use in current devices.



   The apparatus depicted in Figure 2 is suitable for both reduction and float zone melting. It comprises a quartz tube 20 lined with inlet and outlet tubes 21 and 22 respectively, by means of which a stream of halogenated material and hydrogen is maintained in the tube 20 and it is equipped with a heating means 23 which can be an induction heating coil connected to a high frequency source not shown, the equipment being movable relative to the column 20 Inside, the tube 20 is gan of two crystal holders, an upper retainer 24 and a lower retainer 25.

   The upper morning device 24

 <Desc / Clms Page number 8>

 is a nickel rod which passes through a closure such as a brass plug 26 plated with gold or nickel, and threaded at its upper part and is provided with a knurled nut 27 provided for vertically adjusting the support. The lower support 25 can be of quartz, it is suitable for receiving a support crystal 28 at 29 and is in turn held by a mandrel 30 which can be made of nickel. Mandrel 30 is attached to spindle 31 which is mechanically attached to head 32 which may be gold or nickel plated brass and which may be mounted on a drive carriage not shown, in aorta to move. the whole assembly in relation to the induction heating device 23.

   Alternatively, the head 32 may be attached to a stationary element, providing for the movement of the heater 23.



  The remaining part of the apparatus shown in Figure 2 has been found to be suitable for the introduction of hydrogen and halogenated material. This apparatus comprises a bottle 33 containing silicon tetrachloride or other halogenated metal 34 provided with an immersed inlet tube 35 which is connected to a source of purified hydrogen, not shown, the flow of which is controlled by a tap. 36.



  Hydrogen passing through tube 35 passes bubble to bubble through fluid 34, and is saturated there with halogenated material which in turn passes through tube 37, ball joint 38 and tube 21 into column 20 The vapor pressure of the halogenated material 34 is kept low using a dewar bottle 39 containing cooling material 40.



   The apparatus shown makes use of a small heater * 23 particularly suitable for the sound melting function described. For the simplicity of the device, the same heater can be used for

 <Desc / Clms Page number 9>

 the reduction function and the spreading zone by moving the heater up and down relative to the tube 20. Alternatively, a fixed heater of greater length can be used during reduction.



   A brief description will be given below of the operating manners used in accordance with the present invention. Their exact nature is evident from the examples listed which follow.



   Although in general the improvement in the impurity content and the final resistivity of the elemental matter comes from an absorption column process, another improvement can come from the pre-treatment and subsequent processes that 1''ll write about it too.



  Each of the procedures described is useful for the purposes pursued and can of course be practiced independently. Some of the auxiliary procedures were first devised to avoid random contamination and all kinds of other conditions which could make it difficult to compare the levels of impurities of materials treated by the present absorption technique and those treated by ordinary means. The first of these auxiliary procedures results in the removal of any sulfur compounds which may be present in the silicon tetrachloride or other halogen starting material.

   the silicon tetrachloride is refluxed in contact with the copper turn over a period of several hours. Although relative amounts of tetrachloride and copper turnings have not been found to be critical, it has been found that using 30 grams of copper turnings

 <Desc / Clms Page number 10>

 2000 cubic centimeters of tetrachloride, sufficient sulfur removal is achieved. In general, any sulfur-containing contaminant included in the starting material is removed during some refining process such as, for example, during the zone smelting which follows when it is desired to produce an elemental material. The presence of this contaminating agent in the halide is easily detected since, on hydrolysis, it is released as hydrogen sulfide.

   Sulfur is generally not known to have an adverse effect on semiconductor materials to be used in semiconductor devices, perhaps because much of this sulfur is lost during any kind of processing. However, from a scientific point of view it has been found advisable to treat a starting material as pure as possible, thus avoiding the creation of equilibria which could affect the absorption mechanism either advantageously or annoyingly.



   When halogenated materials according to the invention were to be reduced to their original elements, it has been found desirable to employ a process discussed above, resembling the ordinary hydrogen reduction of silicon tetrachloride on a. heated tantalum filament. This subsequent treatment which has been developed to avoid contamination by tantalum or other foreign filamentary element, uses a semi-conductive material resulting from the reduction. A suitable apparatus for carrying out this subsequent treatment is shown in figure 2 and is described in the text corresponding to that figure.

   According to this process, a single crystal filament of high purity silicon,

 <Desc / Clms Page number 11>

 about an eighth of an inch in diameter and six inches long, was suspended in a quartz tube.



  A second silicon support rod, which can also be a single crystal, was fixed in the lower support and was made to abut the suspended crystal. Because high purity silicon is too high a resistivity to be inductively heated, the lower end of the crystal was notched and a piece of heavily doped silicon containing boron was introduced and fused together to the support material. .



    After fusion, the upper crystal called the filament, was welded to the lower stopper rod using induction heating, conducted along the length of the rod and originating at the place of lower resistivity created by fusion with the boron-containing material.



  Hydrogen was passed to flow through silicon tetrachloride maintained at a temperature of about -20 ° C The combined flow of hydrogen plus silicon tetrachloride collected in the passage was caused to flow overall and over the yarn in%. The central region of the filament was heated to about 1400 ° C. A three inch region was heated by moving the assembly containing the rod in the induction coil heater up and down this distance. The reduction of silicon tetrachloride by hydrogen on the part of the filament thus heated resulted in a deposition of elemental silicon on this surface.

   After creating an appreciable deposit of silica, a liquid zone was established in the ifilamen near the deposit and a zone pass was made through the deposited material to produce a single crystal. It was found that silicon proudit

 <Desc / Clms Page number 12>

 by reduction with hydrogen in this manner had, relative to the purified material on tantalum, a lower impurity content by a factor of at least 10, judging by the resistivity measurements.



   In the absorption procedure itself, it has been found that a wide range of absorbent materials results in a significant improvement in the impurity content. From the following examples, it will be seen that the impurity content * has been calculated in many cases by retrospective process from resistance measurements made on the final reduced materials. This way of proceeding, however, was rather complicated and took time. We have turned to another method to determine the class of absorbents suitable for use of the method according to the invention.

   According to this procedure, high purity silicon tetrachloride or other halogenated material of the highest commercial purity was doped with one percent by weight of each of the two materials, boron richloride and sodium trichloride. phosphorus. A measured amount of material so contaminated interceptually was allowed to drip into a 10 inch long column of the absorbent product under investigation. After remaining in the column for a prescribed time, the halogenated material was removed from the column also at a prescribed rate in advance, and the washed sample was analyzed by an infrared radiation absorption technique. The detection limit of this infrared ray method is approximately 0.01 percent.

   The analytical results thus obtained were verified on standard untreated halogenated samples. Pod we know that the efficiency of the absorption of such

 <Desc / Clms Page number 13>

 column may vary depending on the level of impurities, so that the results that can be obtained by treating these spiked samples may not really indicate the results that can be obtained on undoped samples with very low impurity content a verification test was made by comparing the degree of separation resulting from treatment of doped material with the calculated results of resistivity measurements made on a non-doped scaffold, all tests using the absorbent lead.

   As can be seen from the examples which are appended, it was found that the infrared technique used in the spiked samples gave a good indication of the separation results obtained on low impurity samples as well.



   Various factors indicate that the absorption phenomenon shown here is at least in part chemically or almost chemically saturated, as it has been found that absorbents cannot be regenerated by flushing with dry hydrogen at temperatures of 1. 'order of 120 C. Also, when observations were made on the nature of the banding phenomena which occurred in the column, as could easily be done when phosphorus was used as one of the intentionally added impurities, (see examples 6 and 7 below), it was found that all the band which formed at the beginning very close. from the top of the column did not move down the column as is so often seen in chromatographic work.

   Instead, the strip was found to widen with the lower end of the strip moving downward and the upper end remaining stationary.
In the same experiment, it was found, as might be expected, that the efficiency of: the separation was in

 <Desc / Clms Page number 14>

 largely dependent on the surface area, with improved separation observed on large area materials such as gel structures, compared to similar particle size granular materials of the same gel nature in which the surface had become. deactivated, for example in the case of silica gel washed with sodium hydroxide.
 EMI14.1
 



  When the absorbents have a large surface area, it has been found that good separation is obtained by the use of a wide range of materials. All of these materials are considered suitable for carrying out the invention. These materials include oxides and silicates, both in hydrated form, as well as various forms of activated charcoal, including activated charcoal.



  Examples of these hydrated materials are t
Aluminum oxide
Silicon dioxide
Titanium dioxide
Ferric oxide
Fuller's earth (aluminum calcium silicate)
Manganese dioxide
Chromium oxide
Calcium oxide
Rare earth oxides such as dtyttrium oxide L t mangnesium oxide
Zeo Dur (registered trademark for hydrated silicate treated glauconite and manufac- tured by the Permutit Company, USA
 EMI14.2
 from America).



   All of the above materials have a gel structure or present. otherwise a very porous surface These materials are well known absorbents. Surface activating treatments for these materials are well known
 EMI14.3
 apdcialiatea tograph1. and tochniquists

 <Desc / Clms Page number 15>

 neighbors. All of these known methods of improving the surface activity of any of these absorbents are effective in the practice of the invention.



   Some additional materials have been found to be effective in removing boron and phosphorus from halogenated samples, although the actual degree of separation is not so great. While all of the materials listed lead to the reduction of the level of impurities to the limit values of detection or below these values for the infrared radiation process, the following materials have been found to lower the level of impurities only about an order of magnitude.

   These materials include tungstic acid; Celite 545 (Trademark, manufactured by Johns-Manville company United States of America) (diatomaceous earth containing more than 90 percent silica); and Decalso (Trademark, sodium aluminum silicate gel), manufactured by Pemutit Componay, United States of America. It should be understood that although the efficiency of the separation which results from the use of the Although any late material death was somewhat poorer than that achieved using any of the first category materials, the degree of separation was nevertheless clearly noticeable.

   Efficient separation can occur even with the latter materials even by using a longer column or by activation or other treatment designed to increase surface activity.



   An important class of absorbents which have been found to be ineffective in the practice of the invention is the group of ion exchange resins. Dried resins which have been tested include a cation exchanger from

 <Desc / Clms Page number 16>

 weak carboxylic resin type. a strong acfde type cation exchanger, and a strong base type anion exchanger. It can be concluded from this that the absorption mechanism is not in its essence an ion exchange mechanism.



  However, due to the fact that the Zeo Dur (trademark) weak acid type cation exchanger was ineffective as stated above, it must be recognized that ion exchange materials; having surface characteristics suitable, i.e. high porosity, and of suitable chemical nature, i.e. hydrated oxides or hydrated silicates, may nevertheless be suitable for the operation.



   In the examples given below, some indication is given as to the type of surfactant aggregates found suitable. When some of the gel materials listed above could not be commercially available, and were manufactured to determine the utility of the entire class of absorbent listed, the process followed is described. In the case of silica, it has been found desirable to take a sample of absorbent material and check it against commercially available grades. This was done in the manner set out below. As a whole, however, it will be recognized that the knowledge of those skilled in the art for the various absorbent techniques will be used herein.

   It is not considered necessary to completely define the desired surface characteristics which contribute to a useful absorbent.



  It is believed that anyone experienced in this technique, comfortable with the list of suitable absorbent materials given above, will know how to obtain and how to process any of these materials in order to make them suitable for the purpose pursued. .

 <Desc / Clms Page number 17>

 



  Accordingly, when referring here to an absorbent material by its chemical nature, it should be understood that one is to refer here to a material in a physical form such as to have the desired surface area.



  By way of example, although silica gel is stated to be a suitable absorbent material, it will be understood by one skilled in the art that such material to be suitable must retain a porous structure. Physical forms of silica which would not be thought to be suitable and which have been found to be inactive include material calcined at 1000 ° C for a period of 20 hours, resulting in the removal of most of the water and rupture of the structure of the gel, which severely limits the surface area. Similarly, silica washed in sodium hydroxide led to inactivation of the surface, rendering it unusable as well.

   As is well known, the silica intended to be used as an absorbent is advantageously washed in an acidic material such as normal hydrochloric acid for its purification and to increase its surface action.



  Materials thus prepared have been found to be suitable. as can be seen from the experimental results recalled here, boron and phosphorus were invariably removed * in approximately equal quantity based on infrared detection methods as well as on resistivity measurements. In removing it, the boron can be improved by pre-treating the halogenated material with aluminum chloride as well as with chlorine. According to this process, powdered aluminum chloride along with liquefied chlorine was added to the halogenated material and allowed the material to be held for several hours after which the

 <Desc / Clms Page number 18>

 Material thus treated was poured and introduced into the absorption column in the ordinary manner.

   This process is related to that described in J.M. Whelan, United States Patent No. 2,821,460, in which the combination of materials was shown to give effective removal of phosphorus from silicon tetrachloride by distillation. Several areas of addition of aluminum chloride and chlorine as discussed herein are effective herein. In general, 1 to 100 grams per liter of aluminum chloride have been found to be suitable. The chlorine is advantageously added in liquid form and from 1 to 20 cubic centimeters per liter has been found to be a satisfactory proportion. Larger amounts of aluminum chloride and chlorine are not harmful, but they do not aid the reaction. Smaller amounts of either additive have a diminishing effect.



   As seen from Example 1, 50 grams per liter of aluminum chloride and 10 cubic centimeters of liquid chlorine were used. The advantages gained in the more efficient removal of boron from a halogenated product treated by a method known to remove only phosphorus are not understood. It should be noted, however, that some chances exist to indicate the existence of a complex of boron trichloride and of phosphorus trichloride in parallel with the complex of boron trichloride and of phosphorus pentachloride, the existence of which is known.

   It is also known from the work of Whelan (see patent citation above) that the removal of phosphorus from a halogenated product by treatment with aluminum chloride and chlorine results from the formation of a complex between the pentachloride of phosphorus and aluminum chloride. In that the boron removal is improved by the treatment.

 <Desc / Clms Page number 19>

 temet with aluminum chloride and chlorine, it can be assumed that the complex of aluminum chloride and phosphorus chloride is more stable than the complex
 EMI19.1
 of chloride of boron and of phosphorus.

   This, '* one can imagine, indicates that' boron trichloride '' ''? ' '' <. '.' '' t ...,, thus released from = its complex is more easily absorbed by the column., - It ', should be clearly understood that the processes' described here which make use of aluminum chloride and chlorine in pretreatment are based on experimental results and in no way attaches to the theoretical explanation given above.



   The following examples contain guidance on the procedures followed to determine the scope of suitable absorbent materials. Only satisfactory tests have been shown. Materials which have been found not to be suitable are mentioned above and are excluded from the class listed.



   The examples given below were chosen from experimental tests as demonstrating the suitability of a wide range of absorbent materials. Some
 EMI19.2
 examples of the techniques of pre-treatment and subsequent treatment described above; When this auxiliary treatment is included in an ox * mp2. Understand that parallel trials using the same absorbent materials have been conducted - without making use of this ancillary treatment.

   These tests
 EMI19.3
 additiorinela which are not included in the description. to save space, indicated that the main purification was obtained from the absorption technique '.' ,, and not from one or more of the auxiliary treatments., '' x. , j. , ..;,; ¯, aj p,; .- ,. ¯,. ¯ ,,. ,, j The main exception to this is leprosy-treatment -with '*.' '.-' '"* $"' .y bzz aluménium chloride and chlorine which as indicated .'-...... .... 3 ## '"", 5 eideaatuf. a / led a serisible augaentat3ori of] 1; Éi ± '* r ^' x3Â.V, .. é "" 'ß' ,: '' ", t" ',. 3'1l "' hf - '' ... w3 .. "! r ..

 <Desc / Clms Page number 20>

 



   Boron removal. Methods using this pre-treatment with aluminum chloride and chlorine are therefore considered to be preferable according to the invention.
As discussed above, other auxiliary treatments have been designed to clarify the results and to simplify comparison with untreated materials. As a result, the treatment with the copper turn resulted in the removal of sulfur compounds.

   Reduction of silicon or germanium with hydrogen, while being a useful and certainly preferable procedure to the method now in common use, whereby the reduction is carried out on tantalum, does not give a new puri - fication of the silicon tetrachloride or the resulting element, but only avoids the introduction of new impurities from tantalum or other foreign matter inherent in the reaction system which includes the leaching agents used to remove the material filamentary. All of the ancillary methods disclosed are believed to be novel and constitute a part, albeit less or optional, of the present invention.



  Example 1
2220 grams of silicon tetrachloride from the Stauffer Chemical Company containing traces of boron, phosphorus and a heavy metal chloride as impurities had refluxed overnight in contact with 30 grams of copper tide to remove sulfur compounds. The removal of these contaminants is indicated by the blackening of the copper due to the formation of copper sulfide. The copper tounide was then removed and 50 grams per liter of anhydrous aluminum chloride and 0.001 percent radioactive phosphorus trichloride were added to the silicon tetrachloride.

 <Desc / Clms Page number 21>

 at the same time as 10 cubic centimeters per liter of liquefied chlorine.



   The radioactive phosphorus was added so that trace studies could be made during the purification as set out in Example 7. As stated above, this addition results in the formation of a coordinating compound. nation between phosphorus pentachloride and aluminum chloride according to the teachings of JM Whelan Patent No. 2,821,460. The silicon tetrachloride, along with the added materials, was 'left to stand for a period of 18 hours'! approximately to ensure completion of the reaction.

   The silicon teuchlurre was then added to the reservoir of an absorption column of the type shown in FIG. 1, by means of a siphon containing corked glass wool to entrain all the solid aluminum chloride. This column had an interior diameter of 2.22 cm and was packed to a height of 20.5 cm with A1203 passing through the 80-200 mesh F-20 grade sieve supplied by the Aluminum Company of America, USA. United of America. Prior to use, A1203 was activated by heating it in air at 270 ° C for 18 hours.

   The column was filled with silicon tetrachloride and allowed to stand for 16 hours, after which the liquid was allowed to pass through the column at a rate of one cubic catalyst per minute. Silicon tetrachloride was then used to prepare a silicon rod by hydrogen reduction on tantalum in the known manner (see R. Hölvling, Zeitschrift für angewandte Chemie 40 page 655 (1927) using tantalum tubing 1.27 in diameter, having a wall thickness of 0.076 mm and a length of 28 cm The technique used is described in more detail in the Bell Telephone Labroatories Record of September 1955, pages 327-330.

   The resulting silicon

 <Desc / Clms Page number 22>

   @ washed in 48 percent hydrofluoric acid to remove tantalum, it was etched in a mixture of 3 volumes of 70 percent nitric acid and one volume of 48 percent hydrofluoric acid and washed in deionized water The silicon rod was then passed thirty times through a float zone apparatus in a hydrogen atmosphere to remove residual phosphorus and to grow a single crystal. The float process is described elsewhere. See Journal or Matais, volume 18 pagea 1316-19.

   The crystal thus treated was found to be of the p-type with a conductor side resistivity of 85,000 ohm-cm. and a resistivity of 10,000 ohm-cm. further down the rod, when boron separation was not as effective. In comparison, a control run using only distillation purified silicon tetrachloride gave a resistivity at the conductor end of 4,700 ohm-cm and 1,200 ohm-cm further along the rod. In both cases, the higher resistivity at the end of the conductor corresponds to the most efficient zone refining region.



  Example 2
A sample of silicon tetrachloride was treated with copper as described in Example 1.



  It was then purified by passing it through a column as above, except that the absorbent was 6-15 mesh passing silica gel obtained from the Amend Drug Company, USA. Before use, the gel was leached with normal hydrochloric acid 6 for 24 hours, washed with deionized water to remove hydrochloric acid, dried at 110 C and activated for 4 hours at 300 C. Silicon tetrachlorurs

 <Desc / Clms Page number 23>

 was then reduced by reduction with hydrogen on a silicon filament in an appardl of the type d written in FIG. 2.



  According to this process, a single crbtal filament of high purity silicon, about 3.2 mm in diameter and 15.3 cm long, was suspended in a tube.
 EMI23.1
 '"" "' <tW) Mtt <MMM .. nmxtm)) to a nickel support. A silicon rod then fixed in the quartz support at the bottom of the apparatus and was held in place by a mandrel of nickel in the lower head. Since high purity silicon cannot be heated directly by induction, the lower end of the silicon rod was notched, a piece of boron containing heavily doped silicon was introduced into the notch, and the lower end of the silicon rod was notched. This doped region could not be heated directly and the heat would be transferred to the crystal without fusion.

   The lower rod was then brought to abut against the silicon filament and it was matched by adjustment made by manipulating the upper head in the ball joint and by adjusting the knurled knob of the threaded part of the upper support A weld was then made in dry hydrogen by passing heat from the lower end of the rod to the junction with the filament using the induction coil and a five megacycle generator. A bottle containing about 250 cubic centimeters of silicon tetrachloride was then attached to the apparatus and cooled to about -20 C The hydrogen was made to flow through the silicon tetrachloride and from there to pass through a quartz tube at a flow rate of approximately 2 liters per minute.

   The central region of the filament was heated to 14000C by passing heat from the
 EMI23.2
 lower rod by moving all 11 apla reil in one direction and the other in the. one-portion induction coil

 <Desc / Clms Page number 24>

 7.6 cm center of the filament, the speed of the relative movement being about 0.5 mm per minute. A dense layer of deposited silicon occurred increasing the diameter of the rod to about 1.0mm in about four hours.



  The stream of hydrigen was then reduced to a rate of 0.5 liters per minute and the filament was melted just below the deposited region to produce a molten zone which was then moved up the rod at a rate. - tsse of about 1.27 mm per minute to melt the deposited material and transform it into a single crystal.



  To prevent the transfer of gaseous impurities from the quartz paboi of the apparatus and to prevent the deposition of silicon on the wall, the tube constituting the furnace was cooled by a curtain of water produced by directing a jet of water on the wall. tube several inches above the induction coil. A drainage tube was provided to collect water at the lower end of the apparatus. The resulting silicon body was then zone-melted once-through and it was found to be a p-type having a resistivity of 44,700 ohm-cm.



  Example 3
185 cubic centimeters of t 'type silicon tetrachloride bonded in the previous examples was doped with 0.001 percent radioactive phosphorus chloride and passed through a column packed with 6-12 mesh silica gel. The dimensions of the column in this example were 1.27 cm in diameter with a stuffed part length of 25.4 cm. The adsorbent was activated as shown in Example 2 and the silicon tetrachloride was allowed to pass through the column at a flow rate of about 0.5 cubic centimeters per minute.



  Radioactive tracer techniques have indicated

 <Desc / Clms Page number 25>

 that the phosphorus trichloride content was reduced to 7.5 x 10-5 mol. percent. Radiographs of the column indicated that even these results could be improved by proper adjustment of the particle size and packing conditions, since the indications showed that a considerable channel effect was occurring in the absorbent.



  Example 4
Example 3 was repeated, substituting alumina for the silica gel increasing to 6-12 meshes. The results were similar to those of Example 3.



  Example 5
Example 3 was repeated by replacing the 6-12 mesh silica of Example 3 with 60-200 mesh silica gel (Davison Chemical Company, Grade 950, USA and using a column of 2.22 cm inside diameter stuffed over a length of 50.8 cm.



  The apparatus dimensions and processing conditions were otherwise the same as those given in Example 3. The smaller particle size silica * resulted in a reduction of PC13 from 0.001 vou percent to 5.3 x 10. -6 mol. Percent.



  Example 6
60-200 mesh activated alumina was substituted for the silica of Example 5. A 185 cubic centrimete sample of silica tetrachloride was spiked to 0.001 percent phosphorus chloride. phore as above. By passing this sample through the aluminas column as described in the examples above, a reduction in phosphorus chloride content of 5.9 x 10-8 mol was obtained. percent, the limit of detection for capture techniques, - radiactive, After pass of 1500 cubic centimeters

 <Desc / Clms Page number 26>

 of silicon tetrachloride thus treated, the amount of phosphorus which flowed out was increased to 3.4 x 10-6 mol. percent phosphorus chloride.

   The column x-ray showed a high concentration of phosphorus about 7.6 cm down from the top of the stuffed part and no visible channel effect.



  Example
The procedure followed in Exampleµ was repeated, however with the addition of 10 cubic centimeters of liquid chlorine per liter of silicon tetrachloride to the previous sample to pass through the absorption column. The liquid coming out did not show a detectable phosphorus, indicating a reduction to less than 6.6 x 10-7 mol. percent. After passing 3000 cubic meters of so-doped silicon tetrachloride, x-rays of the stuffing showed a tight band of phosphorus N extending 1.27 cm down from the top of the column.

   On this basis, it was estimated that the column would be capable of purifying at least 30,000 cubic centimeters of silicon tetrachloride before flowing silicon tetrachloride containing phosphorus. detectable. example
A sample of silicon tetrachloride was prepared as in Example 6. 50 grams per liter of anhydrous aluminum chloride, along with 10 cubic centimeters of liquid chlorine per liter, in each case based on tetrachloride. silicon were added and the sample was allowed to stand for 24 hours.

   The flowed liquid thus treated was exept of detectable phosphorus. As in Example 6, the passage of 3000 cubic centimeters of this tetrachloride

 <Desc / Clms Page number 27>

 of silicon pretreated through the column as described gave a tight band of phosphorus 1.27 cm down from the top of the packed column.



    Example 9
To verify the relative effectiveness of different absorbents for the removal of boron trichloride and phosphorus trichloride from silicon tetrachloride, the following method was used. 50 milliliter burettes 61 cm long and 1.27 cm inside diameter were stuffed to a height of 25.4 cm by various absorbents tested. Samples of 30 cubic centimeters of silicon tetrachloride doped each time with one percent boron trichloride and phosphorus trichloride were allowed to drip into each of the burettes filled with pipettes at a flow rate of about 1 cubic centimeter. per minute. After such an addition, the columns were allowed to rest for one hour.

   Samples of 10 cubic centimeters of silicon tetrachloride so treated were withdrawn from each burette at a rate of about 0.8 cubic centimeter per minute. A second set of samples, including all of the silicon tetrachloride that could flow out, was collected. Untreated and run-off samples are analyzed by infrared absorption techniques.

   A strong absorption band for PC13 had a wavelength of 1313 cm -1 and bands of Bc13 were found at 1415, 1368 and 1345 cm-1. The detection link for these bands, with the method used. site was 0.01 percent.

 <Desc / Clms Page number 28>

 
 EMI28.1
 x, i lt s 13
 EMI28.2
 
<tb> SiC13 <SEP> elapsed <SEP> Sic!
<tb>
 
 EMI28.3
 n Absorbai 10 premiera ce drainabla '' "îV S11116., 3 43eOl% BC13 ¯il1 .. activ '260 * C in 16 0.01% Pol3 O, OLGPC13 hours 11 Al 0 to 8..4 <0.01% BOl, 3 <0.01% BC1.



  ¯xl "1e. Ac- tira A 260 * CO, Ol Pol,", 0, Ol% P013 for 16 hrs 12 Charcoal d..0 ::: O, Ol B013 0.01% BC13 coco bzz, meshes activated at 30,000 z6 Pol 0.01% Pci3 in 4 hours 13 811108 gel 6-16 <O, Ol; d Bel3. :: 0.01% B013 meshes washed A iïC1 dry ac- c O, o196 Pol3 <=. 0 , 01% PC1 .. tiv6 A 300'C
 EMI28.4
 
<tb> in <SEP> 4 <SEP> hours
<tb>
 
A faster process has been developed to determine the effectiveness of a large class of absorbent materials.
 EMI28.5
 bantes. Following the prooé3 1000 cubic centimeters of silicon tetrachloride were doped with 1 percent of
 EMI28.6
 Bowl 3P at the same time as 1 percent PC.3.

   20 cubic centimeter portions of this starting material, along with 5 cubic centimeters of the absorbent to be tested were placed in Erlenmeyer flasks. The vials containing these samples were stoppered and left to stand for one hour in an ice bath with shaking.
 EMI28.7
 sometimes. At the end of this period, the silicon tetrachloride was emptied from the absorbents here; filran through glass wool. The treated material was analyzed by an infrared ray technique as in Examples 9 to 12.

   The results are shown below:

 <Desc / Clms Page number 29>

 
 EMI29.1
 .m. 11 .. 24
 EMI29.2
 
<tb> Content <SEP> in <SEP> impu-
<tb>
 
 EMI29.3
 , 9; 42% Dlt absorbent retained after lift treatment 9! lica gs1 â 6-12 stitches "0, Ol Bel
 EMI29.4
 
<tb> leaaivé <SEP> HC1 <SEP> washed, <SEP> ac-
<tb>
 
 EMI29.5
 tlv4 at 2? 0 C in 16 hours 0.01% PC13 15 Alumina at 0-14 mesh, 0.1% BOl leached with washed HCl, 3 activated at 270 ° C in 16 hext- 01% PC1
 EMI29.6
 
<tb> res <SEP> 3
<tb>
 
 EMI29.7
 16 Alumina 9-14 mesh, 0.1% Ber less NaOH, washed activated at 270 ° C in 16 nra 0.1% PC1- 17 Gel from 1'12 to about <0.01 B01 '; \ 12 mesh .. <: 0 Old P01- 19 hsp3 at about 12 meshes <0.01% Bel3 0.01% Pei3 19 Fuller's earth at about 0.01% BC1 100 mesh ...:

   0.01% POl)
 EMI29.8
 
<tb> 20 <SEP> Tungstic <SEP> <SEP> to <SEP> approximately <SEP> 0.1% <SEP> BC1
<tb>
<tb>
<tb>
<tb>
<tb>
<tb>
<tb>
<tb> 100 <SEP> meshes <SEP> 0.1% <SEP> PC13
<tb>
 
 EMI29.9
 21 Mg (OH) 2 to about 100 0.01 BCl mailI'8 0.01% Pci 3
 EMI29.10
 
<tb> 22 <SEP> Zeo <SEP> Dur <SEP> (<SEP> Brand <SEP> of <SEP> manufacturer
<tb>
 
 EMI29.11
 as, silicate hydrated at 0) 01 BC1-
 EMI29.12
 
<tb> approximately <SEP> 6 <SEP> meshes) <SEP> prée <SEP> 0,01% <SEP> PC! 3
<tb>
<tb>
<tb>
<tb>
<tb> paré <SEP> by <SEP> Permutit <SEP> Company,
<tb>
 
 EMI29.13
 United States of America.
 EMI29.14
 
<tb>



  23 <SEP> Celite <SEP> 545 <SEP> (<SEP> Brand <SEP> of <SEP> manufacturer
<tb>
<tb> that <SEP> earth <SEP> of <SEP> diatoms <SEP> to
<tb>
<tb> approximately <SEP> 100 <SEP> meshes <SEP> containing <SEP> 0.1% <SEP> BC13
<tb>
 
 EMI29.15
 at least 9C of silica) pre- 0.1% P013
 EMI29.16
 
<tb> decked out <SEP> by <SEP> Johns <SEP> Manville <SEP> come
<tb>
<tb> pany, <SEP> United States <SEP> d (America.
<tb>
 
 EMI29.17
 24 Decalso (Trademark,
 EMI29.18
 
<tb> silicate <SEP> of aluminum <SEP> and <SEP> of <SEP> 0.1% <SEP> BC13
<tb>
 
 EMI29.19
 sodium approximately 100 mesh 0.1; 6 PC1'1
 EMI29.20
 
<tb> made <SEP> by <SEP> Permutit <SEP> Com
<tb>
 
 EMI29.21
 pany, U.S.A.

   Example 25
About 200 cubic centimeters of tetrachloride
 EMI29.22
 germanüma. were prepared by treatment of gormnidr dioxide 29

 <Desc / Clms Page number 30>

 known to be unsuitable for use in semi-
 EMI30.1
 conductoura with normal hydrochloric acid 6, after which the germanium tetrachloride was removed by distillation. To further contaminate the halogenated material, 5 milligrams of As203 dissolved in 10
 EMI30.2
 centieette * cube of hydrochloric acid 12 times normal.

   The material was allowed to stand for 18 hours, after which the hydrochloric acid was separated from the tetrachlo-
 EMI30.3
 ride of gormanium by extraction. 25 cubic centimeters of previously prepared CaC14 were set aside as a sample of
 EMI30.4
 control. 50 cubic centimeters of the same Mtierw were allowed to pass through a 1.27 cm column of silica gel and packed to the height shown in Example 1. The flow rate was 0.5 cubic centimeter per minute. ; The silica gel absorbent used had been activated with normal dilute hydrochloric acid 6 for about 70
 EMI30.5
 hours, washed with ddaîoniede water, dried at 120 OR and activated at 2700C for 16 hours.

   The control sample and a 25 cubic centimeter sample of flowed material were hydrolyzed separately in 150 cubic centimeters
 EMI30.6
 deionized water. The resulting germanium dioxide was washed and dried at 120 ° C overnight. The germanium dioxide samples were then trans-
 EMI30.7
 formed into hydrogen at 65,000 in a quartz pod.



  Elemental germanium was first melted and turned into solidified ingots in a directional direction. The relative purity of the two ingots was established by means of resistivity measurements. The average resistivity of the unpurified material was 0.003 ohm-cm compared to 0.40 ohm-cm for the purified material, indicating an increase. mentation of the purification of two orders of magnitude, i.e. an improvement factor greater than 100.

   Matter

 <Desc / Clms Page number 31>

   he-131 tittée was considered to be of suitable quality for the semicondutuer Example 26
The method of Examples 14 to 24 was applied to a sample of silicon choroform (SiHCl) which had been doped with 1 percent each time BC13 and PC13 The use of a 6-16 mesh silica gel results in
 EMI31.1
 knows the reduction of both BC13 and PC13 to a level of 0.1 percent.



  Example 27
The process of Example 26 was repeated on a sample similarly doped with silicon chloroform using 80-200 mesh alumina, activated at.
 EMI31.2
 270 * G for lS hours. The levels of BC13 and pal) were reduced each time to about 0.1 percent.



  Example 28
A sample of silicon chloroform was purified using the column described in Example 25, packed with 60-200 mesh silica gel (Davison Chemical Company, USA, Grade 950), and activated at 270 ° C for 18 hours. The silicon chloroform was allowed to pass through the column at a rate of 0.55 cubic centimeters per minute and was then reduced on a silicon rod using the method of Example 2. The material in the deposited region was. of type n with a resistivity of about 850 ohm-cm. A similar rod made from a control sample of the initial material was of the p conductivity type and with a resistivity of around 1400 ohm-cm.

   After performing these measurements, the two rods were given an additional zone pass through a vacuum of 5 x 10-6 millimeters of mercury by the floating zene technique to remove phosphorus. The rod pro-

 <Desc / Clms Page number 32>

 picked by the chloroform to purified silicon was p-type of about 1500 ohm-cm while the control rod was p-type with resistivity of about 1000 ohm-cm. From these data, it was concluded that the silicon chloroform used was quite pure from the start, containing races of boron choride and phosphorus chloride. From the conductivity type measurements and the comparative resistivity measurements, before and after zone melting, it appeared that the absorption technique did not improve, in particular in the removal of boron.



  Example 29
The following experiment was performed to determine the efficiency of the column technique for the removal of heavy metal chlorides from silicon tetrachloride: After passing 6000 cubic centimeters of silicon tetrachloride through the gel column Silica of Example 1, the column was emptied of residual liquid and one inch samples of absorbent were removed at various points on the column. These samples were hydrolyzed in deionized water, dried at 150 ° C for 16 hours and analyzed spectrochemically.

   Tests made for six metals on samples taken from these places in the column were recalled. below
 EMI32.1
 
<tb> Metal <SEP> Section <SEP> to <SEP> 1 <SEP> pou- <SEP> 4 <SEP> inches <SEP> dece <SEP> above <SEP> then <SEP> the <SEP> top <SEP> section <SEP> of the
<tb> down, <SEP> to <SEP> 1 inch
<tb> Aluminum <SEP> VST <SEP> ND <SEP> ND
<tb>
<tb> Bore <SEP> VST <SEP> VST <SEP> ND
<tb>
<tb> Brass <SEP> VST <SEP> VST <SEP> ND
<tb>
<tb> Ver <SEP> ST <SEP> VST <SEP> ND
<tb>
<tb> Mangnesium <SEP> ST <SEP> VST <SEP> VST
<tb>
<tb> Manganese <SEP> VST <SEP> VST <SEP> ND
<tb>
 

 <Desc / Clms Page number 33>

 
The designations given in the form of the table above are commonly used in the technique of spectrochemical analysis.

   The debitions are as follows:
ST - slight trace, less than 0.005% by weight
VST - very slight trace, less than 0.001% by weight
ND - not detected Example 30
Silicon tetrachloride was treated with copper turnings as described in Example 1. Using the same absorbent material and column as in Example 1, liquid was introduced into the column reservoir and allowed to pass through the column as before. The liquid material was poured in and the steps outlined in Example 1 were performed.

   After 36 passes through a floating zone melting, the silicon was found to be p-type with a resistivity of 8000 ohm-cm at the conductive end and 1800 ohm-cm throughout, Example 31
The procedure of Example 30 was repeated with activated silica gel passing through a 6-16 mesh screen as in Example 2 as the absorbent. The resulting silicon rod was p-type with a resistivity of 11,000 chm-cm on the conductor side and 4000 ohm-cm along the rod.



   For the studies set forth herein, it was found desirable to test two gel-based materials which apparently were not commercially available. As a check on some of the tests that were performed, it was also decided to prepare a commercially available gel structure and then check the absorbency.

 <Desc / Clms Page number 34>

 of the material synthesized with that of the commercial material (see Example 32). The processes actually used for the preparation of these three absorbent materials are given below.

   References in parentheses to an example number found in each description indicate that the absorbent prepared in that example was used.



  Example 32
Silica gel passing approximately through a 6 to 12 Mile sieve prepared as shown below was washed with deionized water, dried 4110 C and activated at 270 C for 18 hours and used as an absorbent in a vacuum cleaner. reissue of Example 14. A similar purification was obtained.



   Preparation of titanium dioxide gel (example 17) -
25 cubic centimeters of tetraisopropyl titanate were stabilized against rapid hydrolysis by the addition of 5 cubic centimeters of hydrochloric acid 12 normal liver to which was added with stirring a stabilized solution of 75 cubic centimeters of deionized water . The mixture was then heat treated at 80 ° C. to bring it into solution. Without cooling, ammonium hydroxide was added, allowing it to flow dropwise until a gelatinous precipitate was obtained. The material was comfortable at rest and cooled to room temperature.



  On standing for several hours the curdled masses of the precipitate stiffened The material was made strongly alkaline (pH about 10) by the addition of ammonium hydroxide, after which it was left to stand until the following day. . After washing with deionized water, the curd was dried for two days in a vacuum desiccator. after which we air-dried at 100 C

 <Desc / Clms Page number 35>

 for 24 hours and activated in air at 270 ° C. for 18 hours. The material was then put into use as an absorbent.



  Preparation of ferric oxide (FE2030 (example 18)
About 100 grams of Fe (OH) 3 was dissolved in excess normal hydrochloric acid 6. The solid matter which formed was removed by filtration. The remaining liquid was made basic by the addition of ammonium hydroxide (pH about 10) which resulted in the formation of a gelatinous precipitate. The precipitate was filtered and washed in deionized water to remove all ammonium chloride and ammonium hydroxide. The material was then dried for 24 hours in air at 20 C and activated in air at 270 C for four hours.



  Silide preparation (example 32)
50 cubic centimeters of silicon tetrachloride were poured slowly dropwise into 500 cubic centimeters of deionized water at a temperature of about 0 C. The resulting opalescent liquid set in the form of a rigid gel in two hours at room temperature.



  After waiting until the next day, the material was broken up with a stirrer rod, washed with deionized water to remove hydrochloric acid and dried for two days in a vacuum desiccator. Finally, the material was brought to air at 110 C for 24 hours and activated at 270 C in air for 18 hours.



   It is interesting to note that all of the absorbent materials of the preceding examples contain water of hydration. That all absorbents useful here for the purposes pursued must contain such binding water has been clearly established. As a result, while

 <Desc / Clms Page number 36>

 silica gel activated at 270 ° C and approximate formula
SI2. 0 18H20 was effective, calcined silica containing little or no water of hydration was ineffective. Probably in the case of silica there is a breakage of the associated gel strength so that the loss of potency can be explained from the physical characteristics.



  However, the same phenomenon has been observed in the case of a non-gel-shaped structure of the abeerbant, the case of magnesia. In this case, it was found that although the anhydrous magnesium oxide was ineffective and did not give rise to any noticeable decrease in the level of impurities, this mutor containing a water of hydration molecule, Mg (OH2) was quite effective in reducing the level of boron and phosphorus to the approximate detection level of the infrared analysis technique used with this measurement. (See example 21).



   Much of the experimental work described here was carried out on samples which had been purposely contaminated with im that can only be removed from material with difficulty. As a result, boron and phosphorus were added to the halogenated silicon materials, such as silicon tetrachloride and silicon chloroform, while arsenic was added to the corresponding germinium material. In all of these experiments, the level of impurities thus increased decreased appreciably, often to below the detectable limit of the particular analytical process used. Some of the tests reported below, however, such as Examples 2, 30 and 31, relate to work which was carried out on the purest commercially available silicon tetrachloride.



  In these tests, of course, the standard analytical process was unnecessary to determine the levels of impurities.

 <Desc / Clms Page number 37>

 In these cases, the purified material finally obtained is reduced to its elemental composition by reduction with hydrogen, either on tantalum or on silicon as described, then applying one or more passes by the floating zone technique.

   Resistivity measurements made on samples are naturally of great importance especially when the material thus purified is reduced and used in a semiconductor device.
However, it is not expected that the commercial use of the present invention will be limited to the preparation of the elementary materials for such use.



  For example, silicon tetrachloride can be an initial or intermediate product in a process intended for the production of silane. (See Journal of American Chemical Society, Volume 69, pages 2692-6, November 1947).



   It is known that the main impurities in the halogenated starting material were boron and phosphorus.



  These two materials are important impurities in the silicon system, they are such as to produce conductivities of opposite types and have different distribution coefficients between liquid and solid phases in a silicon system. As a result, it is possible after passing the material through a determined number of zone passes through a body of that material and after making resistivity and conductivity measurements at different positions in the direction of progression of these areas, do a backward calculation and determine the total number of accepting impurity atoms (boron) and the total number of donated impurity atoms (phosphorus). These calculations were made for the final materials corresponding to Examples 1 and 30 to 32.

   These calculations are summarized below.

 <Desc / Clms Page number 38>

 
 EMI38.1
 
<tb>



  Number <SEP> of a
<tb> Number <SEP> of atonies <SEP> tomes <SEP> ce
<tb>
<tb> accepting <SEP> - <SEP> giving <SEP> - <SEP>
<tb>
<tb> Atoms <SEP> by <SEP> cent- <SEP> Atoms <SEP> by
<tb>
 
 EMI38.2
 Sx'ap Absorbent cubic timeter one hundred. cube 1 Ai 203 l, 4x 1012 1.9 x 1012
 EMI38.3
 
<tb> (A1C1 + C13)
<tb>
 
 EMI38.4
 30 AI 203 7.2 x 10 12 3.1 x 1014 il sio2 5.6 x 10 12 5.8 x 10 12 32 sio2 - Carbon 1.4 x 1012 1.1 x 10I4
 EMI38.5
 
<tb> from <SEP> wood
<tb> Aloi <SEP> 3+ <SEP> C12
<tb>
 
As will be seen from the foregoing description, any broad group of absorbent materials is suitable in the practice of the invention.

   Regardless of the choice of absorbent material or the combination of these materials, it may be found that the separation is improved or that subsequent contamination is avoided by employing any of the pre- treatments and subsequent treatments described. The most salient of the preliminary treatments is that in which chlorine and aluminum chloride are added first to the halogenated material. This pre-treatment led not only to the expected improved banding of phosphorus, now as pentachloride, but also to improved boron removal by a mechanism not yet fully understood.



   As seen from a comparison of Examples 2 and 31,] Contamination during reduction after the absorption treatment or after any of the old treatments, can be reduced by reducing substantially on a silicon rod rather than on a material. foreign

 <Desc / Clms Page number 39>

 handles such as tantalum which must be removed afterwards.



   An apparatus suitable for this use is shown in Figure 2 and is described in the relevant text subject.
Absorbent materials which have been found useful in the practice of the present invention are all hydrated materials. These materials which are all oxides or silicates with the sole exception of charcoal, are designed to have hydroxyl groups attached to their surfaces. As is known to those skilled in the art, charcoal can have absorption characteristics similar to these hydrated oxides and silicates.



  In addition to having the desired chemical nature, as described herein, and as illustrated by the examples, it is of course necessary that the materials chosen have the physical characteristics desired for a good absorbent. No attempt has been made to fully define these characteristics, all of which are well known to those skilled in the art of chromatography. However, some of these characteristics are inherent in this description. For example, it has been observed that a gel structure has an effective absorbent surface area suitable for these purposes. It has also been noted, however, that this material must be dried and activated to remove what is generally considered to be physically embedded water.



  It has also been noted that the absorbency of these gel materials is significantly lowered on calcination, which leads to a rupture of the gel structure and a consequent decrease in the effective surface area.



  It is also seen that the treatment of an absorbent material with a reagent which has the effect of reacting with the free surface or of the blocking agent destroys its power.

 <Desc / Clms Page number 40>

 absorbent so that leaching of the silica gel with sodium hydroxide renders this material unusable for the purposes of the invention. Whether or not the chosen absorbent material has a gel structure, it is
 EMI40.1
 naturêllemelltJ! é1 Mntrante-dJ.ut .-: Uiser .. J1! 1Arl -impSQP. as fine as possible.

   Thus, for example, it can be seen from a comparison between Examples 3 and 5 and also 4 and 6 that the use of a 60-200 mesh sieve material leads to a significant improvement in the separation compared to what occurs with material passing through a 6-12 mesh screen, otherwise chemically and physically identical.



   It is believed that the examples listed and the other work described are of sufficient scope to lead to the conclusion that the absorbents are used in the present invention.
 EMI40.2
 horsefly are those which belong to materials including hydrated oxides hydrated silicates and also including charcoal and other carbonaceous materials which are known to exhibit surface reaction characteristics of materials in this group. When we
 EMI40.3
 As "t6r 'to" 8 materials of this class in the description and in the following claims, it should of course be understood that the materials so referred to must, in other respects, have both physical and chemical characteristics. which make these materials suitable for this use.

   Requirements of this nature include saddles described in the previous paragraph. Other requirements. * Also considered obvious are that the material chosen should not be soluble in a large amount or capable of reacting with the halogenated material which is subjected to treatment and that this absorbent should not contain any contaminating material considered. as harmful to the system being treated.

 <Desc / Clms Page number 41>

 



   It can also be seen that mixed absorbent beds are ineffective in the process according to the invention. In this regard, a bed of charcoal SiO2 and A1C3 has been effectively used in an experiment which in other respects was similar to that of the example, with good results.



   Other ordinary column techniques can be used with advantage in conjunction with the methods described. For example, the flow can be improved by cooling the column. Dimensions can be varied or one can use successive columns.



   R e v e n d i c a t ions
1. Process for treating a compound containing chlorine in the liquid phase of an element chosen from the group comprising silicon and germanium, as an agent contaminating at least one chloride of an element chosen from the group comprising boron and phosphorus, comprising treating this material with at least one absorbent selected from the group consisting of charcoal, hydrated oxides and hydrated silicates, such that intimate contact is produced between the material and the absorbent.


    

Claims (1)

2. The method of claim 1 wherein the treatment consists of passing the material in question to pass a packed column containing said absorbent. 3 The method of claim 2 wherein the absorbent is at least one hydrated oxide selected from a EMI41.1 group which includes Si02, A'2 0 3p Ti02, ro23 and lego. 4. The method of claim 3 wherein the absorbent comprises silica gel. 5. The method of claim 3 wherein the absorbent contains aluminum gel. 6. The method of claim 3 wherein <Desc / Clms Page number 42> the absorbent comprises titanium dioxide gel. 7. The method of claim 3 wherein the absorbent comprises iron oxide gel. 8. The method of claim 3 wherein the absorbent comprises magnesium hydroxide. 9. The method of claim 2 wherein the absorbent comprises sodium silicate. 10. A process for purifying a compound containing liquid phase chlorine and silicon comprising adding chlorine to this material, adding aluminum chloride to the material and passing the material through a packed column, having a pad comprising at least one absorbent selected from the group consisting of hydrated oxides, hydrated silicates and charcoal, and finally reducing the material so treated by reduction with hydrogen in the vapor phase on a body of silicum 11. The method of claim 10 wherein chlorine and aluminum chloride are added to the material before passing it through the column. 12. A process for purifying a silicon compound containing liquid phase chlorine, comprising adding chlorine and aluminum chloride to this material, removing unreacted aluminum chloride, and passing through. The material thus treated by a packed column having a packing comprising at least one absorbent selected from a group which includes hydrated oxides, hydrated silicates and charcoal. 13. A process for reducing a silicon compound containing chlorine to the state of elemental silicon comprising passing an atmosphere essentially comprising this compound together with hydrogen. <Desc / Clms Page number 43> on a heated silicon surface so as to reduce the compound and produce the deposition of elemental silicon on this silicon surface.