19 Leslie Court
Whippany, New Jersey 07981
ARSENOSILICAFILM FOR BURIED LAYERS
Arsenosilicafilm is an excellent material for use as a doped oxide diffusion source. It is especially useful for buried layer diffusions. The doping requirements for buried layer diffusions are generally in the range of 10 to 20 ohms per square with layer thicknesses usually 2 to 5 microns. While these electrical requirements are not difficult to achieve, the concomitant requirement that there be no surface pitting of any type requires careful attention to processing details to achieve optimum results. In this note, information is presented on the effects of temperature, time, atmosphere and dilution to achieve various sheet resistivities and junction depths with little or no surface damage.
Arsenosilicafilm is a liquid formulation which on drying forms
an arsenic doped SiO2 layer. The formulation is a true
solution. It should not be confused with glass dispersions in
organic binders or glazes. The solvent in this doping formulation
is ethyl alcohol and the formulation may be diluted with ethyl
alcohol, methyl alcohol, 2-ethoxyethanol and 2-ethoxyethyl
acetate amongst others.
Arsenosilicafilm is available in two different formulations: Arsenosilicafilm Co=1x1020, the standard formulation, and as Arsenosilicafilm 8869. The material designated Type 8869 differs from the standard material in only one way. A small concentration of a polymeric species is added to reduce the hydroscopic nature of the film after spinning. The shelf life of this material is guaranteed for three months from the lot code (date) on the bottle.
Arsenosilicafilm is singularly free of foreign ion contamination. Typical results of analysis show:
|Na less than 1.0 ppm||Fe less than 0.1 ppm|
|Cu less than 0.1 ppm||Mn less than 0.1 ppm|
|Ni less than 0.1 ppm|
Arsenosilicafilm is usually applied by spinning at 3000 rpm. This yields a thin SiO2 film containing Arsenic Pentoxide. After spinning the films are generally air dried and baked in the temperature range of 100o to 200oC. The baking drives off residual solvent. If the baking is omitted, ragged junctions may result. Films formed after spinning are approximately 1500 to 1700 angstroms thick, measured after baking at 200oC, for spin speeds of 3000 rpm. The thickness of the film will vary inversely with the square root of the spin speed up to 6000 rpm.
The diffusion profiles of arsenic in silicon when
Arsenosilicafilm is used as a diffusion source may be
approximated by expressions given in Crank, "Mathematics of
Diffusion", Oxford University Press 1956, p. 37. These
expressions are developed for a single diffusion species
possessing different diffusion coefficients in two plane infinite
media. It has been determined that this model is adequate for
diffusion conditions where no oxygen is present in the ambient
and for conditions where the thickness of the dopant film is
larger than a diffusion length of the dopant in SiO2.
The expression for the concentration profile is:
C o K
2(D 2t) 1/2
Co = concentration of dopant in SiO2 layer.
D1 = diffusion coefficient of dopant in SiO2.
D2 = diffusion coefficient of dopant in Si.
|K =||concentration of dopant in silicon.
concentration of dopant in SiO2
At the interface the concentration in the silicon is:
The pertinent data to fit these expressions have been determined at 1200oC:
Co = 1.5 x 1022 As atoms/cm3 SiO2
D1 = 3 x 10-14 cm2/sec. - diffusion coefficient in SiO2
D2 = 5 x 10-13 cm2/sec. - diffusion coefficient in Si
Co (Si) = 5 x 1019
K = .003
While the above expression holds for diffusions is nitrogen, as will be noted in Table I, different results are obtained in oxidizing ambients. In strongly oxidizing ambients the penetration profile in the silicon will not remain erfc for the time predicted by the expressions above. The SiO2 layer growing at the interface between the silicon and the dopant film eventually separates the source from the silicon causing the diffusion process to revert to a Gaussian type. Also, in the presence of oxygen one realizes higher surface concentrations than in nitrogen, possibly due to pile up of dopant as described by Atalla and Tannenbaum BSTJ, 1960, p. 933 f. When a moderately oxygen rich atmosphere is present during the diffusion the surface concentration obtained at 1200oC is greater than the values quoted above.
|Arsenosilicafilm (Standard)||1 x 1020|
|Arsenosilicafilm 8869||2 x 1020|
The predominant type of surface disturbance observed with
these diffusants is the formation of small hexagonal glassy
deposits which remain on the surface of the silicon wafer when
the SiO2 layer is dissolved in dilute HF solution.
These rosettes, i.e., the hexagonal glassy structures, will
dissolve in concentrated HF solution and frequently when they are
so removed a small pit is left on the silicon surface. The
rosettes are more frequently observed in the masking SiO2
layer in the windows where arsenic doping is desired.
While the exact nature of these rosettes has not been determined, it is believed that they are a crystalline form of SiO2. The crystallization or devitrification of the silica is probably nucleated by the high concentration of arsenic atoms present in the film.
It is general experience that the thicker the dopant layer the greater the density of rosettes per cm2. It is also observed that the density increases with time of heat soak at the elevated temperatures required for diffusion. The greater frequency in the SiO2 layer arises because the solution of the dopant film in the layer provide ideal sites for the development of rosettes.
At 1200oC, the rosettes will increase in diameter with time of heat soak. Rosettes which develop in the windows do leave arsenic rich regions in the silicon. The depth of the pits remaining in the silicon under the masking oxide depends upon the oxygen concentration in the ambient in which the diffusion occurs. Apparently the rosette or crystalline SiO2 exhibits a different diffusion resistance to O2 than amorphous or glassy SiO2. This then yields an unequal oxidation rate at the silicon surface and results in the pit. It has been observed that some pitting in the masked areas do not limit devices yields.
From the observations presented it is apparent that to achieve a minimum amount of surface damage with Arsenosilicafilm, one should employ a thin film consistent with the sheet resistivity required, the ambient atmosphere should be dilute in oxygen and the time of diffusion should be minimum, again consistent with the doping required.
In light of the considerations presented to minimize surface
damage, data is presented on the diffusion characteristics of
Arsenosilicafilm in various atmospheres and at various dilutions.
The first and most striking characteristic shown by the data is
the effect of oxygen in the atmosphere during diffusion to
produce lower sheet resistivities than is observed in
non-oxidizing atmospheres. This is shown in Table I where
variation in sheet resistivity as a function of time of diffusion
in various nitrogen-oxygen mixtures in presented. Note that as
small an oxygen concentration as 3% will yield lower sheet
resistivity than nitrogen alone. Also, one should note that while
the sheet resistivity is lower in the early time of the
diffusion, the sheet resistivity does not continue to decrease
with extended diffusion time as it does in the atmosphere where
the oxygen concentration is lower. In addition, with proper
selection of the oxygen concentration the sheet resistivity will
decrease linearly with the square root of the time characteristic
of erfc diffusion profiles.
Arsenosilicafilm may be diluted with ethyl alcohol or methanol. Dilution yields a thinner film for constant spin speed. However, the concentration of arsenic in the glass layer remains the same. The effect of dilution on the diffusion characteristics is shown in Table II. From the data listed there, it is noted that the thinner the dopant film the longer one may allow the diffusion to proceed before there is serious development of rosettes. With Arsenosilicafilm undiluted, about one hour is the maximum heat soak time before surface deterioration occurs. However, the material diluted one to one will not lead to surface deterioration for four hours of diffusion. For diffusions carried out at richer oxygen mixtures, the diluted films will not remain erfc as long as they do for a 3% O2: 97% N2 mixture. Also, in the oxygen rich ambients, more difficulty with surface pitting in the masked regions will be experienced.
TYPICAL DIFFUSION PROCESSESS
In the light of the preceding discussion, several typical
diffusion procedures may be set up dependent on the sheet
resistivity and depth of penetration required.
A.Arsenic Doped Layers - 6 to 10 Ohms/square
To achieve this low sheet resistivity, Arsenosilicafilm 8869 is recommended. The diffusion is carried out in O2 at 1200o-1250oC depending upon the depth of penetration required. The coated wafers are diffused for one hour in O2. The wafers are then removed, and the arsenic doped layer is removed in dilute HF. The sheet resistivity will be in the range of 10-12 ohms/square. A second layer of Arsenosilicafilm 8869 is applied and diffused for one hour. The wafers are removed and deglazed. The wafers are then returned to the diffusion furnace to achieve the depth of penetration required. While this process entails some wafer handling, one may achieve defect-free surface with sheet resistivities significantly below 10 ohms/squares. The deglazing process consists of dissolving the arsenic doped glass in 10% HF solution. The etch rate of Arsenosilicafilm in 10% HF is approximately 1000 angstroms per minute at room temperature. In addition, about 200 angstroms of the oxide mask should be removed to remove nucleation centers for rosette growth due to arsenic diffusion into the thermal oxide.
B.Arsenic Doped Layers - 10 to 15 Ohms/square
Arsenic doped layers 10 to 15 ohms per square are achieved by a single step diffusion process utilizing either standard Arsenosilicafilm or Arsenosilicafilm 8869. In the first hour oxygen is the ambient. After the first hour, the atmosphere is changed from oxygen to nitrogen and the diffusion may be continued for two or three additional hours if deeper penetrations are required than are realized after one hour. Alternatively, after the first hour of diffusion, the wafers may be removed from the diffusion furnace and deglazed as described above and returned to the diffusion for additional diffusion.
C. Arsenic Doped Layers - 15 to 25 Ohms/square
To achieve arsenic doped layers 15 to 25 ohms per square, Arsenosilicafilm standard or Type 8869 is diluted 1:1 with ethyl alcohol. The diffusion is carried out in 5% O2: 95% N2 or 3% O2:97% N2. After 4 hours of diffusion at 1200oC, one will achieve a mean sheet resistivity of 15 ohms/square and a depth of penetration of 4 microns.
THE EXTRACTION PROCESS
The extraction process for producing buried layers refers to a
process where the dopant film is spun on a unmasked wafer and
photo-etched to leave islands where the doping is required. In
this process a thermal SiO2 layer for masking not
employed. A capping layer of Silicafilm is spun over the
Arsenosilicafilm prior to photo-etching. This capping layer
protects the dopant film during the photo-etching process and is
allowed to remain on the surface during diffusion. In this way
volatilization of arsenic during diffusion is eliminated. An
additional benefit arises thereby in that the reverse side of the
wafer is not doped, reducing difficulties with autodoping in the
subsequent epitaxial process. Since only thin doping and capping
layers are employed, rosette formation is practically eliminated.
No doping will occur in the uncoated areas of the wafer if the diffusion atmosphere contains oxygen and if the dopant film is completely removed from these areas during photo-etching. Soaking the wafers in HC1 at 30o-35oC for several minutes after HF etching will remove adsorbed arsenic atoms and leave a dopant free surface.
After coating, the wafers should be baked for one hour at 350oC to 400oC in air to densify the film and prevent washout of dopant by the subsequent Silicafilm coating. After the application of Silicafilm, the wafers must be baked again for one hour at 350oC to 400oC to densify the SiO2 cap. The wafers are then coated with Photo-resist and exposed and developed. The etchant for the films is 5% HF for 20 to 30 seconds. The wafers are rinsed in DI water and resoaked in fresh 5% HF solution. They are rinsed again and soaked in HCl at 30o-35oC for one or two minutes and washed in DI water. The photo-resist is removed with commercial strippers. The diffusion is carried out as described below.
Arsenic Doped Layers by the Extraction Process
Arsenosilicafilm (either standard or 8869) is spun on the wafer at 5000 rpm. The wafers are baked as described above. Silicafilm is spun on at 3000 rpm. Again the wafers are baked. Photo-etching is carried out as described above. Diffusion is carried out in 10% O2: 90% N2. The following results are obtained:
|Sheet Resistivity||1 Hr.||4 Hrs.||8 Hrs.|
|Ohms per square||15||7||5|
No pitting is observed after 8 hours of diffusion. The extraction process leaves a doped mound rather than a pocket. The height of this mound may be increased if the last 15 minutes of diffusion is carried out in oxygen or steam.
Epitaxial layers have been deposited over arsenic diffused layers produced by the extraction process. These layers exhibited no unusual characteristics and many thousands of devices have been produced from these wafers.
VARIATION OF SHEET RESISTIVITY WITH DIFFUSION TIME IN VARIOUS
O2/N2 AMBIENTS AT 1200oC
|Atmosphere**||1/4 Hr.||1 Hrs.||2 Hrs.||4 Hrs.|
|O2 for 15 min.
followed by N2
|15||10||- -||- -|
*Processing Conditions-Spin Speed 3000 rpm; Wafers heated in air at 200oC for 15 minutes prior to diffusion.
**Flow Rate-10 cu.ft./hr.
EFFECT OF DILUTION ON ARSENOSILICAFILM
|1 Hrs.||4 Hrs.|
|Undiluted||11||None||6||Many in SiO2Mask|
|3 AsSiF: 1 Ethyl Alc.||13||None||8||Many in Oxide Mask|
|2 AsSiF: 1 Ethyl Alc.||25||None||14||Few in Oxide Mask|
|1 AsSiF: 1 Ethyl Alc.||32||None||16||None|
Processing Conditions-Same as those for Table I
|Alphabetic Product Listing||
|Material Safety Data Sheet|