CA1108772A - Method for forming isolated regions of silicon - Google Patents

Abstract

METHOD FOR FORMING ISOLATED REGIONS OF SILICON

Abstract of the Disclosure A method for isolating regions of silicon involving the formation of openings that have a suitable taper in a block of silicon, thermally oxidizing the surfaces of the openings, and filling the openings with a dielectric material to isolate regions of silicon within the silicon block. The method is particularly useful wherein the openings are made through a region of silicon having a layer of a high doping conductivity.

Description

11 Background of the Invention 12 The invention relates to methods for dielectrically 13 isolating regions of monocrystalline silicon from other 14 regions of monocrystalline silicon.
Description of the Prior Art -16 In the monolithic integrated circuit technology, it 17 is usually necessary to isolate various active and passive 18 elements from one another in the integrated circuit structure.
19 These devices have been isolated by backbiasing, PN junctions, partial dielectric isolation and complete dielectric isolation.
21 The dielectric materials used have been silicon dioxide, glass, 22 and so forth. The preferred isolation for these active 23 devices and circuits is some form of dielectric isolation.
24 The dielectric isolation has the substantial advantage~over the PN junction isolation because it allows the butting of 26 the circuit elements against the isolation and thereby 27 result in greater density of packing of the active and 28 passive devices on the integrated circuit chip.

FI9-77-032 ,-1-..
. : . . . _ .

' ' ~.: ' . :
`

; :: ,~: .
. . .

srlz 1 One form of dielectric lsolation involves the

2 formation of grooves or depressions in silicon where

3 the isolation regions are to be forrned. During the

4 groove formation, the remainder of the silicon surface is protected by a protective film which is substantially 6 unaffected by the silicon etch used to form the grooves.
7 The usual protective layer is a silicon nitride, silicon 8 dioxide sandwich layer. Following the formation of the 9 grooves by conventional chemical etching, the silicon body is subjected to a conventional oxidation step where-11 by the silicon in the groove area is oxidized and the 12 silicon dioxide fills up the groove as well as oxidizing 13 further into the silicon to form the isolation region.
14 One of the ma~or problems with this process is what is known as "bird's beak".
16 The "bird's beak" is a non-planar silicon dioxide 17 formation at the top periphery of the groove and is caused 18 by the lateral oxidation underneath the silicon nitride 19 layer. Since the oxidation of a specific thickness of silicon requires an almost equivalent amount of free space 21 to expand into, and since the Si3N4-limits the unrestricted 22 expansion, the result is an up-pushing of the silicon 23 nitride at the edge of the groove. The final consequence ~-24 of this is a general stress in the perimeter region of the groove as well as difficulties in subsequently achiev-26 ing good butted diffusions against the vertical portion of 27 the silicon dioxide. This non-butting capability defeats FI9-77-032 ~ -2-.
.

., ~

ta,~i~2 1 a major benefit of the original purpose of the silicon dioxide region. This process is described more fully by E. Kooi U.S.
Patent 3,970,486, Clevenger U.S. 3,534,234, Peltzer U.S. 3,648,125 and I. Magdo et al, Canadian Patent Application No. 143,3~8, filed May 30, 1972.
Another technique for forming d-ielectric isolation is described in the V. Y. Doo U.S. Patent 3,386,865 and "A Composite Insulator-Junction Isolation" by R.E. Jones and V. Y. Doo, published in Electrochemical Technology, Vol. 5, No. 5-6, May-June 1967, pp. 308-310. This technique involves the formation of a silicon dioxide layer or similar type of layer on the substrate in the region where dielectric isolation is desired. An epitaxial layer is grown upon the substrate in all regions except where the silicon dioxide is lo-cated. The surface of the epitaxial layer and the sides of the openings are partially thermally oxidized. The openings are then - filled by vapor deposition of polycrystalline silicon, silicon dioxide or similar materials. This technique has some disadvantages. Selec-` - tive epitaxy, as required by this technique, is very sensitive to f the area relationship between silicon dioxide and silicon regions.
~ 20 For example, two different s-ize silicon regions would tend to fill in ; at a different rate so that at-the end of a process, the regions are - filled in to a different extent.- Also, in mesa-type depositions, crystallographic faceting tends to occur. This results in pyramid-like growth and tends to widen the isolation regions beyond the original lithography ,capabilities. The slanted e~

DLM/W6 c ,'' , ,. ~

~ . _, ..... . .

1 silicon/silicon dioxide interfac~ will again cause 2 difficulties in achieving reliable butted diffusion 3 against the silicon dio~ide region.
4 The formation of grooves and the filling of such 4

5 . grooves have been described by other publications such

6 as the G. L. Kuhn, U.S. Patents 3,892,608 and 3,969,168.

7 In these patents~ chemical etching is used to form a V

8 groove, a rounded bottom groove or a rectangular evacuated

9 space. There is little detail as to how the groove is

10 formed but it is clear that the groove would be limited by t

11 the nature of the chemical etching step. The process does J

12 not necessarily yield a planar surface and it requires

13 photolithography after the formation of the grooves. D. K.
~14 Roberson U.S. 3,956,033 describes a similar chemical etch 15 followed by filling with polycrystalline silicon. Here t~
16 again, the groove is limited by the chemical etching 17 technique and it is unclear how the overgrowth of the 18 polysilicon is removed. U.S. Patents K. E. Bean et al 19 3,725,160 and W. R. Morcom et al 3,979,237 also show filling of grooves. In these patents, the effect of 21 chemical etching is more clearly brought out where it is 22 shown that monocrystalline silicon are preferentially ~.
23 etched chemically to provide grooves having symmetrical 24 sidewalls sloped at precise angles depending upon the ' particular face crystal to which the silicon surface is 26 aligned.
27 The Brand U.S. Patent 3,979,765 also describes the 28 chemical etching to open rectangular grooves and the ' ``
.

%
1 filling of the grooYeS with insuldtor material. ~lowever, it is dif-ficult to unclerstand how some of the principal steps such as etching and filling are actually e-ffected from the description.
The formation of rectangular yrooves have been rnade in the Hoch-berg U.S. 3,966,577 and T. Kaji et al U.S. 3,997,378 and S.A. Abbas, IBM* TDB Vol. 20, No. 1, p.l44, June 1977 entitled "Recessed Oxide Isolation Process", by reactive ion etching techniques. None of these references describe the problems and detailed solutions for forming reactive ion etched grooves in s;licon. Neither do they involve t~emselves with the problem of effectively filling the groove to form the best possible isolation for the silicon regions. There are des-criptions of the reactive ion etching processes in the "A Survey of Plasma-etching Processes", R. L. Bersin, published in Solid State Technology, May 19, 1976, pp. 31-36 and particularly for silicon in J. M. Harvilchuck published German patent application number 26,174,834, published December 9, 1976 for "Reactive Ion Etching of Silicon".
However, the details of the reactive ion etching in these publications do not show how the reactive ion etching would be utilized in the formation of dielectric isolation.
Summary of the Pre_ent Invention In accordance with the present invention, a method for forming dielectric isolation is described wherein tapered sided isolation structures are formed in a specific manner. In the present text, we use openings, *Registered Trade Mark ., _, . . ~ , ;:: ' ' :

r~

1 channels, holes, grooves and trenches interchanyeably.
2 Tapered sided openings of less than about 4 micrometers 3 in width at the surface are formed in a silicon body in 4 the areas where dielectric isolation is desired. The formation of the tapered openings is by reactive ion 6 etching. The taper of the tapersided openings can be 7 from near vertical to as much as 20 from the vertical 8 and terminates into the substantially flat bottom of the 9 openings. The surfaces of the openings are then thermally oxidized to silicon dioxide. The oxidized openings are 11 filled with a suitable dielectric material to fully 12 isolate the regions of silicon. The dielectric material 13 from above the surface of the body is removed to provide

14 uniformly filled isolation pockets at the surface of the silicon body. The precise taper of the opening is 16 important so as to balance the need for greater density 17 of devices against the completeness of groove filling by 18 the CVD method. Grooves of insufficient taper will 19 exhibit a CVD SiO2 which could contain an opening or poor quality dielectric region in the center of the 21 filled groove.
22 The method for dielectrically isolating regions of 23 silicon further overcomes the problem of reactive ion 24 etching through the silicon body which is composed of a layer of highly doped silicon. The highly doped silicon 26 is etched more isotropically under most conditions of 27 the reactive ion etching than regions of lower doping.
28 The conditions of reactive ion etching through such a : , ' r -~ ~jr~ 7'2 1 hiyhly doped layer of silicon involves the use of a 2 reactive chlorine specie ambient having a pressure of 3 between about 2 to 50 micrometers with the chlorine 4 specie percentage in the gas of be-tween about 2 to 10 and etch rate of between about 0.03 and 0.08 micrometers.
6 srief-Description of thc ~ra~.~ings 7 FIGURES lA - lE illustrate a method for forming the 8 dielectric isolated structure of the invention;
9 FIGURE 2 is a graph showing the problem of over-etching highly doped silicon regions at various etching 11 conditions;
12 FIGURES 3 and 4 illustrate the problem of filling 13 the tapered hole of varying tapered angles from the 14 vertical;
FIGURE 5 illustrates the problem of filling tapered 16 holes of one taper from the vertical for different widths 17 of holes;
18 FIGURE 6 is experimental data of the groove taper 19 angle change as a function of chemical vapor deposited silicon dioxide filling thickness;
21 EIGURE 7 is experimental data of the extent of 22 crevice above or below the surface versus groove width 23 for different tapers; and 24 FIGURES 8, 9 and 10 are graphical illustrations which indicate the characteristics of filling the open-26 ings with dielectr:ic material.

7%

1 Dcscrlptioll of the Prefcrl-ecl ~mbodimcnts 2 Referring now particularly to FIGU~ES 1~ , the 3 manufacturing steps for the dielect:ric isolation of one 4 form of the invention are shown. The struc-ture of FIGURE lA includes the monocr~stalline silicon substrate 6 10 which is shown as P- conductivity for illus-tration 7 purposes, an N+ layer 12 over the substrate 10 and an 8 N- conductivity layer 14 on the layer 12. For the 9 purposes of the invention, either all or some of the layers 10, 12 and 14 could be of opposite conductivity 11 from the conductivity types indicated. However, it is 12 preferred to have the layer 12 to be in a high conduc-13 tivity region where it will ultimately be the collector 14 of a bipolar transistor. This structure can be fabri-cated by various techniques. However, the preferred 16 technique is to provide a P- monocrystalline silicon 17 substrate and to diffuse an N+ blanket diffusion into 18 the substrate by using conventional diffusion or ion 19 implantation of an N type impurity such as arsenic, antimony or phosphorus to produce an N+ xegion with a 21 surface concentration of between about 1 x 1019 or 22 1 x 10 1 atoms/cc. The layer 14 is subsequently grown 23 onto the 10, 12 structure by means of epitaxial growth.
24 This may be done by conventional techniques such as the use of SiCL4/H2 or SiH4/H2 mixtures at growth temperatures 26 of about 1000C to 1200C. The N+ layer may have a 27 typical thickness of between about 1-3 microns whereas 28 the epitaxial layer has a thickness of from 0.5 to 10 ~ _ . . . .

1 microns, the e~act thicknesses depending upon the device 2 to be built.
3 ~lternatively, the s-tructure could be made by 4 various combinations of thermal diffusion, ion implantation and/or epitaxial growth which would include the formation 6 of a varied subcollector region where subsequent formation 7 of bipolar devices is desired.
8 In certain device structures, buried highly doped 9 regions or layers are not necessary and can therefore be omitted. This would be true for FET type devices.
11 Alternatively, multiple buried highly doped regions of 12 different dopant types could be formed by multiple epitaxial 13 and diffusion processing. These structures could be needed 14 for buried subcollector, as well as buried conductor lines.
The next series of steps shown in FIGURES lA and ls 16 are directed to the technique for reactive ion etching 17 of tapered sidewall openings or channels in the silicon 18 structure. A silicon dioxide layer 16 is formed by the 19 conventional techniques of either thermal growth at a temperature of 970C in a wet or dry oxygen ambient or 21 by chemical vapor deposition. Other mask materials can 22 also be used such as silicon nitride and aluminum oxide 23 or combinations thereof and so forth~ Openings 18 are 24 formed in the oxide in the regions where dielectric isolation is desired. These openings are formed by the 26 conventional photol:ithography and etching techniques.
27 The FIGURE lA structure is now ready for the reactive 28 ion etching process. This process may be more fully -1 understood by reference to the J. `~. Ilarvilchuck e-t al 2 patent application referred to above. The ~ induced 3 plasma is reactive chlorine, bromine or iodine specie 4 as specified in the Harvilchuck patent application. The 5 thic~ness of the masking layer 1~ is between about 2,000 6 to 20,000 Angstroms, the exact thickness depending on the 7 depth requirement of the silicon groove. The precise 8 description of the RF glow discharge apparatus is given 9 in the beforementioned patent application. The reactive ion etch or plasma ambient is preferably a combination of 11 an inert gas such as argon and a chlorine specie. Appli-12 cation of suitable power in the order of about 0.1 to 13 0.75 watts/cm from an RF voltage source will produce 14 sufficient power density to cause the reactive ion etching operation of silicon to be carried out at a rate of about 16 0.02 to 0.08 micrometers per minute. The desired result 17 of the etching is shown in FIGURE lB wherein the openings 18 or channels at least partially penetrate through the P-19 conductivity region 10. The channels or openings may go substantially through the N+ region 12.
21 It is important that the openings or channels be 22 tapered at greater than about 2 from the vertical.
23 This is because the subsequent dielectric filling deposi-24 tion process results in a slightly thicker deposition near the top of the groove as opposed to the bottom of the 26 groo~e. Thus, in case of vertical grooves, there is at 27 one stage, an overgrowth of the remaining narrow groove 17~Z

1 which results in poor dielectric material quality in 2 the region belo~ the overgrown area. In case of a 3 sufficiclltly tapered groove, -the groove is filled up 4 from the bottom. The preferred amount of taper, adequate for appropriate chemical vapor deposition filling of a 6 dielectric material such as silicon dioxide will in part 7 depend on the groove width as will become clear from 8 FIGURE 6. The taper of much greater than 20 from the 9 vertical will take up an undue amount of space on the surface of the semiconductor device. This formation of 11 the tapered structure depends upon two principal items.
12 The primary influence on the sidewall formation is the 13 angle of the sidewall of the masking layer 16 in the 14 opening 18. The second dependence is upon the etch rate difference between the masking material and the substrate 16 material. The higher substrate/masXing material etch rate 17 ratio favors the more vertical walls in the silicon 18 substrate.
19 Standard lithography techniques tend to result in slightly tapered resist window openings. Whe~ reactive 21 ion etching is used to open the underlying silicon dioxide 22 film through these tapered resist windows, and when the 23 etch rate ratio between the resist and silicon dioxide 24 is near unity, the taper of the resist window is trans-ferred to the silicon dioxide window. This taper is 26 then, in turn, transîerred into the silicon unless a 27 high etch rate ratio exists between silicon and silicon 28 dioxide. The taper in silicon dioxide mask is preferably , , . , . - :
.

1 within the range of 5 to 20 from vertical. If the 2 lithography process permits vertical resist window 3 openings, then the silicon dioxide window opening is 4 vertical and under such conditions the tapered shape of the silicon opening would tend to be near vertical 6 and generally insensitive to the etch rate ratio between 7 silicon and silicon dioxide.
8 The change in etch rates also influences the under-9 etching of highly doped N+ or P+ regions such as region 12. At etch rates of about 0.07 micrometers per minute, 11 vertical grooves with no lateral N+ undercut etching are 12 formed. Lowering the etch rates will cause increasingly 13 more taper, as indicated in FIGURE 2. FIGURE 2 is a graph 14 which shows the effect of the silicon etch rate in micro-meters per minute versus the percent of the chlorine specie 16 in argon for various system pressure conditions. Curve 20 17 is at a pressure of 10 micrometers. At this pressure and 18 at the etch rates indicated, there is virtually no under-19 cutting in the N+ regions no matter what percent of chlorine specie is used.
21 The taper of curve 20 goes from virtually vertical 22 sidewalls at 10 percent chlorine specie in argon to a 23 20 taper from the vertical at about 3 percent chlorine 24 specie in argon. The power is 0.16 watts/cm2 in all of the FIGVRE 2 experiments and the cathode is silicon dio~ide.
26 Curve 22 shows that at an etch rate of 0.06 micrometers 27 per minute and approximately 3 percent chlorine in argon, 28 a vertical sidewall groove is obtained. As one moves up FI~-77-032 -12-7~

1 on the curve to the e~ch ra-te of 0.10 micrometers per 2 minute and 5 percent chlorine specie in argon, ~e see 3 undercutting in the N+ region. Eurther, as we proceed 4 up the curve to 0.14 micrometers per minute and approx-S imately 7 percent chlorine specie in argon, we see ex-6 tensive undercutting which is very serious. The curve 7 24 shows the situation at pressures of 40 micrometers;
8 at 2 percent chlorine specie in argon and at about 0.06 9 micrometer/min. etch rate, the N+ undercutting is not a problem. However, as one moves up the curve to 0.08 micro-11 meters, the undercutting begins to become more apparent.
12 Further up the curve it is expected that further undercutting 13 will occur. The point 26 is reactive etching at 90 micro-14 meters total pressure and shows extensive undercutting

15 which produces an unsatisfactory product. It can be seen -~

16 from this graph of FIGURE 2 that the useful operative

17 pressure range is from about 2 to 50 micrometers with a

18 chlorine specie concentration of between about 2 and 10

19 percent in the gaseous ambient and an etch rate between about 0.04 and 0.08 micrometers per minute. The chlorine 21 specie which will operate in this manner are CL2, CCL4, 22 CHC13 and o~her chlorine containing species.
23 The key difficulty with the undercutting of the N+
24 region is that it limits the ability of how close one isolation region can be placed next to another. If 26 significant underetching occurs and two isolation regions 27 are located very close to each other, a total underetch of 28 the region 14 will occur. Furthermore, the N+ collector ~..

-7~

1 region will be entirely removed and thus no transistor 2 could be built. A still further proble~l would occur for 3 any undercut region in that such non-linear tapered 4 grooves would not be properly filled in with chcmical vapor deposited dielectric material such as silicon 6 dioxide. The consequence is a filled groove with a buried 7 hole or channel.
8 Referring now to FIGURE lC, the next step in the 9 process is to thermally oxidize the openings or channels by subjecting the body to an oxidation ambient ~hich may 11 be for example 970C in wet oxygen. The body is subjected 12 to the ambient for about 10 to 30 minutes to produce the 13 preferred silicon dioxide thickness within the opening or 14 channel of about 500 to 2000 Angstroms. The purpose of the thermal oxide 30 is to assure good silicon/silicon 16 dioxide interface properties, the qualities of which are 17 usually not as good with chemical vapor deposited dielectric 18 material~ Good quality dielectric Material is necessary to 19 permit the subsequent abutting of diffused junctions against the dielectric isolation.
; 21 The minimum thickness is nominally 500 Angstroms so 22 that a good thermal silicon dioxide layer is formed. Any 23 thinner layer could present difficulties because of pin 24 holes in the oxide and therefore cause electrical integrity problems.
26 The maximum thickness is primarily dictated by the 27 time required at elevated temperatures. Long times at ~: ' -1 high temperatures will tend to move any diffused junction 2 within the silicon regions. Very thic~ oxide films formed 3 at such temperatures will also cause stress problerns in 4 the silicon material.
The grown oxide 30 will follow almost precisely the 6 taper of the sidewalls in the opening formed by the 7 reactive ion etching step. The next step, involviny the 8 filling of the opening with a suitable dielectric material, 9 is shown completed in FIGURE lD wherein the opening or channel is filled with vapor deposited silicon dioxide 11 layer 32. In order to completely fill the opening or 12 channel, it is necessary to cover the surface of the layer 13 30 not only in the channel or opening, but completely over 14 the surface of the silicon body. The preferred filling process is a chemical vapor deposition of silicon dioxide 16 using gas mixtures of Co2/SiH4~N2 or N2O/SiH4/N2 at 800 -17 1000C.
18 Typical deposition rates are of the order of SG - 100 19 Angstroms per minute and total deposition thicknesses are nominally 3 micrometers for 2 micrometers wide grooves if 21 a near planar surface is desired. The specific relation-22 ship of planarity and chemical vapor deposited silicon 23 dio~ide thickness is shown in FIGURE 8.
24 The problems involved in the filling of openings or channels can be more fully appreciated with reference to 26 FIGURES 3, 4 and 5. These figures show the critical 27 importance of the sidewall taper and the problem of si~e 28 of the opening at the silicon surface. As can be seen in 1 FIGURE 3, the center area of the chemical vapor deposited 2 silicon dioxide filled opening shows a small crevice region 3 40. This crevice is present only after silicon dioxide 4 etching of a cross-sectional surface. The crevice forma-tion implies a poor oxide in that region and experiments 6 indicate that this is caused by the oversrowth of the 7 silicon dioxide over opening dimensions which are less 8 than about 0.2 micrometers in width and having a wall 9 angle of less than about 20. It has been determined that the tapered angle of the opening decreases as the 11 opening filling progresses (FIGURE 6). Tilis can affect 12 the overcoating in different ways. The specific effect 13 being related to the groove width and taper. These effects 14 are shown in FIGURES 3, 4 and 5. They illustrate by a series of lines, which represent equal amounts of chemical 16 vapor deposited silicon dioxide layers, the progression of 17 filling in of the grooves. As can be seen in FIGURE 3, 18 the etched out crevice region 40 tends to be buried further 19 down into the filled in groove as the groove width widens and the taper increases. FIGURE 4 shows only the taper 21 effect for the same size grooves. Again, the more tapered 22 groove tends to bury the crevice (or poor silicon dioxide 23 region) deeper. FIGURE 5 shows the groove width effect 24 for a specific taper angle. As the groove widens, the poor oxide region tends to occur at a higher position.
26 These results are summarized in FIGURE 7 in which the 27 extent of the crevice above or below the silicon wafer FI9-77-032 , -16-z 1 surface is plotted for differen-t groove wldths and 2 different taper angles. ~s is clear, successfully 3 buried poor silicon dioxide regions can be achieved 4 for groove geometries of narrow groove widths and 5 . nominal taper angles. As the groove width widens, 6 the taper must be increased accordingly in order to 7 keep the poor oxide buried.
8 FIGVRE 6 is experimental data indicating the change 9 of a taper angle of a groove as groove filling proceeds.
It is obvious that for vertical grooves, any deposition 11 will cause negative tapered grooves and consequently 12 result in physical silicon dioxide voids.
13 FIGURE 8 is experimental data on the necessary 14 amount of chemical vapor deposited silicon dioxide re-quired to fully planarize the surface over a groove.
16 The amount to achieve this is related to the width of j 17 the groove at the top of the groove.
18 FIGURE 9 shows similar experimental data to FIGVRE
19 8 on groove filling and planarity and points out that the planarization is strictly dependent on the groove 21 width and not on the taper angle.
22 FIGVRE 10 is still another means of expressing the 23 surface planarity of a filled groove, this time showing 24 the effect of groove width and different overcoating thicknesses.
26 The final step of the process is the reactive ion 27 etching of the silicon dioxide layer 32 shown in FIGURE
28 lD to produce the structure of FIGURE lE. The excess . ~

._ ................ .

--`L~3772 1 silicon dioxiae is conveniently removed by reactive 2 ion etching and with the help of an optical film thick-3 ness monitoring system or by the knowledge of the e-tch 4 rate of the silicon dioxide. The system used for this process would be of the low pressure sputter etch type 6 system with the wafer positioned on a silicon cathode 7 cover plate. A fluorinated hydrocarbon such as CP~
8 would be used as an etchant so that an SiO2/Si ratio of 9 approximately 1:1 results. The gas pressure could run from 10 to 70 micrometers with gas flow rates of 2 to 11 50 cc/min. The RF power level would run from 0.1 watts/cm2 12 to 0.5 watts/cm2.
13 The result of the reactive ion etch thinning of the 14 silicon dioxide layer is the possible exposure of the inadequately buried poor oxide region in the center of 16 the groove. This is a potential problem because any 17 chemical wet etching of the wafer surface with such ex-18 posed regions of poor silicon dioxide would cause crevice 19 formation in those regions. Such crevices could become potential traps for dirt or process residues and could 21 negatively influence the device characteristics.
22 An alternative embodiment to overcome some of the 23 undercutting problems would be to form a highly doped 24 region 12 in a manner that this region would be inter-rupted and set back from the areas where the openings or 26 channels are to be formed. Therefore, a lower doped 27 region of P-, which is to be reactive ion etched, would 28 surround the region. Thus, there would be no undercutting ~, . - .

-,,3L,~,~s"~i~377~

1 problem. This alternative requires special oxidation 2 and photolithographic and etching steps to allow for 3 this interrupted region.
4 The formation of a P+ region underneath the isola-5 . tion region may be useful when tlle substrate is P-. In 6 such cases, the P- region has a tendency to change its 7 resistivity, even to the extent of inverting to N-type 8 material, when it is thermally oxidized. A P+ implant 9 prevents such inversion possibility. This may be formed by use of a P+ ion implantation of a dopant such as ~oron 11 before the thermal oxidation step of the groove. The 12 preferred technique is the use of a thin chemical vapor 13 deposited silicon dioxide coating of the groove. Such 14 coating of between about 500 to 800 Angstroms will permit the implantation of for example, boron, through the bottom 16 of the groove into the silicon, but not through the 17 silicon dioxide on the tapered walls. This is true since 18 the slanted walled silicon dioxide represents a much 19 thic~er silicon dioxide than its actual thickness because of the directionality of the 90 implanting ions. After 21 the implantation and appropriate annealing, the chemical 22 vapor deposited siliFon dioxide is removed and the normal 23 process sequence (i.e. FIGURE lC) is taken up.
24 Another approach to modify the fabricatior process is to utilize a heat anneal in a steam atmosphere of the 26 sample after process step FIGURE lE~ This anneal at 27 about 900-950C would be to convert any e~posed poor ,, .
~ FI9-77-032 -19-. `
`~

-31~72 1 quality silicon dioxide in the center region of the 2 groove to a good quality silicon dioxide_ The benefit 3 of this process modification is that thectaper angle 4 does not become as important for assuring buried poor 5 . quality silicon dioxide regions and ther~fore permits 6 higher device density fabrication. This process can 7 cure poor silicon dioxide in, for example, the smaller 8 taper angles, such as 2 to 4.
9 While the invention has been particularly shown and 1~ desc~ibed with reference to the preferred embodiments 11 thereof, it will be understood by those skilled in the 12 art that various changes in form and de~ail may be made 13 therein without departing from the spiri-t and scope of 14 the invention. For example, devices oth~r than a bi-polar transistor could be used advantage~usly in the 16 isolated monocrystalline silicon pockets~formed by the 17 process. Such devices would include pas~ive devices 18 such as resistors and active devices suc~ as ~îOSFET
19 devices.

GOS:jr ~/9/77 .~ , .

Claims (9)

The embodiments of the invention in which an exclusive pro-perty or privilege is claimed are defined as follows:

1. A method for isolating regions of silicon comprising:
providing a silicon body composed principally of one conductivity and having a layer of highly doped silicon;
etching tapered sided openings in said body and through said highly doped layer in a silicon reactive chlorine specie ambient having a pressure of between about 2 and 50 micrometers, chlorine specie of gas between about 2 and 10 percent, and etch rate between about 0.03 and 0.08 micrometers per minute;
thermally oxidizing the surfaces of said openings;
filling said thermally oxidized openings with a dielec-tric material wherein the said regions of silicon are iso-lated.

2. The method of Claim 1 wherein the power applied for said etching is between about 0.1 to 0.75 watts/cm2.

3. The method of Claim 1 wherein the said chlorine specie is chlorine gas and the said chlorine in said ambient is about 2 to 8 percent.

4. The method of Claim 3 wherein the remaining component of said ambient is argon.

5. The method of Claim 1 wherein the said layer of highly doped silicon has a doping level greater than about 2 x 1019 atoms/cc.

6. The method of Claim 1 wherein said highly doped layer is within the said silicon body.

7. The method of Claim 1 wherein the said dielectric material is removed from above the surface of said body wherein the said openings are uniformly filled at the said surface.

8. The method of Claim 7 wherein a further heating of said structure at a temperature of about 900 to 950°C to improve the structure of said dielectric material.

9. The method of Claim 6 wherein a said layer is N+, said body is P and a P+ implant is made at the bottom of said openings prior to said thermally oxidizing said openings.