Method for Applying a Uniform Coating to Integrated Circuit Wafers by Means of Chemical Deposition

FIG. 1 shows a typical prior art device suitable for applying tungsten to IC wafers by means of the chemical vapor deposition process in which tungsten hexafluoride is reduced by hydrogen.

FIG. 2 shows an embodiment of the present invention for the case of tungsten deposition to IC wafers by means of the chemical vapor deposition process in which tungsten hexafluoride is reduced by hydrogen.

FIG. 3 shows an alternate embodiment of the present invention for the case of tungsten deposition to IC wafers by means of the chemical vapor deposition process in which tungsten hexafluoride is reduced by hydrogen.

FIG. 4 shows an expanded view of the placement of inlets with respect to the IC wafers for the embodiment presented in FIG. 2.

FIG. 5 shows an expanded view of the placement of inlets with respect to the IC wafers for the embodiment presented in FIG. 3.

FIG. 6 shows the concentration distribution of a typical reactant gas in the first group of reactant gases resulting from the embodiment presented in FIG. 2.

FIG. 7 shows the concentration distribution of a typical reactant gas in the first group of reactant gases resulting from the embodiment presented in FIG. 3.

FIG. 8 shows a nozzle suitable for dispersing the first group of reactant gases into the vicinity of the integrated circuit wafers.

FIG. 9 shows the preferred embodiment of the configuration of the inlets.

In FIG. 2 is shown an embodiment of the CVD system suitable for practicing the invention. As in the prior art device presented in FIG. 1 and discussed in the background, the reaction chamber is enclosed by a wall 21 and a door 22. A set of heater coils 23 are provided adjacent to the wall 21 to provide heat to the reaction chamber. Current is supplied to the coils from a source of current 24 which is regulated to control the temperature of the reactant gases in the reaction chamber. The IC wafers 25 are supported on edge in a boat 26 to form a linear array of parallel IC wafers. The system shown in FIG. 2 is suitable for supplying the reactant gases tungsten hexafluoride and hydrogen to the reaction chamber but the system can clearly be altered to supply other gases to the reaction chamber to implement other reactions such as reduction reactions of other compounds containing any of the refractory metals including tungsten, beryllium, molybdenum and tantalum. In the example shown in FIG. 2, the reactant gases are supplied by sources of gas 27, 28, 29 and 210. The relative flow rates of the gases is controlled by gas flow controllers 211, 212, 213 and 214. These controllers are preferrably mass flow controllers such as Tylan mass flow controllers but they could also be other types such as needle valves.

In the case of decomposition CVD reactions, the reactant to be decomposed is injected into the reaction chamber in the vicinity of the wafer. In the case of a reaction between several reactants, the reactant gases are supplied to the reaction chamber in at least two groups which are allowed to mix only in the vicinity of the wafers so that the uniformity of the thickness of the resulting coating which deposit on the IC wafers is improved. In the example in FIG. 2, hydrogen is supplied in a first group of gases and the tungsten hexafluoride is supplied in a second group of gases. In this embodiment, the tungsten hexafluoride is supplied through a tube 215 which supplies the tungsten hexafluoride to the reaction chamber in the vicinity of the door (ie. nearer to the door than to the IC wafers--typically within a few cm of the door). In contrast, the gases in the first group of reactant gases are supplied through tubes 216-218 to a set of inlets 219-221 positioned in the vicinity of the IC wafers (ie. much nearer to the IC wafers than to the door and typically within a few cm of the edges of the wafers). Carrier gases such as argon can be supplied along with each of the reactants as shown in FIG. 1 and an external mixing chamber can be included to improve the homogeneity of mixtures of gases. However, for fixed sizes of inlets and flow rates of reactants, the primary affect of the carrier gas is to increase the velocity at which gases exit the inlets. Such increased flow can be beneficial in some cases because it distributes the reactant gases throughout the region containing the wafers before reaction between reactants can produce a notable concentration gradient. However, in the case shown in FIG. 2, adequate coating uniformity was achieved without introducing carrier gases. Because separate tubes 216-218 connect inlets 219-221 to flow controllers 212-214, these controllers can individually vary the flow rates through each inlet. Such individual control enables flow rates to be varied as conditions change in order to maintain optimum uniformity. The diameter of each inlet for a given flow rate will determine the velocity of the entering gases.

The unreacted reactants, any carrier gases and the undeposited reaction products are removed from the reaction chamber through at least one exhaust port 222. The exhaust port is connected to a vacuum pump 223 which cooperates with the gas flow controllers 211-214 to control the partial pressures of the gases. In particular, the vacuum pump enables the CVD reaction to occur at below atmospheric pressures if such is desired for a given reaction. In this embodiment, the exhaust port is located so that the IC wafers are situated between the inlets 219-221 and the exhaust port 222. In an alternative embodiment shown in FIG. 3, the exhaust port 322 is located at the end of the chamber opposite to the door 32. The embodiment shown in FIG. 2 has the advantage of producing a convective flow of the first group of reactant gases parallel to the surfaces of the wafers thereby increasing the uniformity of the concentration of these gases in the narrow spaces between adjacent wafers. In this embodiment, in order to provide access of the gases in the reaction chamber to the exhaust port, either the boat 26 must be permeable to gas flow or the boat must be supported away from the exhaust port to provide an unobstructed path for these gases to flow to the exhaust port. The embodiment shown in FIG. 3 has the advantage of having the shape of reaction chambers and furnace systems now available so that the cost of such a system is reduced from that shown in FIG. 2. The embodiment in FIG. 3 also avoids the need to introduce the exhaust ports 222 through the hot zone created by heating coils 23. In spite of the disadvantage of producing a convective flow generally perpendicular the the wafers, a system such as that presented in FIG. 3 has successfully produced tungsten coatings on a batch of 75 wafers having an intra-wafer variation in coating thickness of less than 10%. This apparatus therefore enables the coating of large batches with uniform coatings while retaining processing speed. The furnace in FIG. 3 is therefore the current preferred embodiment for the invention.

Magnified figures showing the spatial relationship between the wafers and the inlets for the first group of reactant gases are presented in FIGS. 4 and 5 for the embodiments presented in FIGS. 2 and 3, respectively. Only 5 wafers are shown in these figures to aid the clarity of the illustration although typically many more wafers (on the order of 100 wafers) than this will be processed in one batch. In addition, the spacing has been accentuated to further assist visualization. The set of wafers 25 include two end wafers: a first wafer shown as the left-most wafer in FIG. 4 and a last wafer shown as the right-most wafer in FIG. 4. The wafers are contained in a circular right-cylinder 41 having side edges shown as the dashed lines and having end faces defined by the first and last wafers. A reference set of axes are presented in FIG. 4 and are oriented so that the x-axis is aligned along the direction of the array of wafers, the y-axis is aligned along the lateral wafer length directed horizontally and the z-axis is aligned along the lateral wafer length directed vertically. The origin of the axes is positioned in the center of this volume so that the volume extends from x=-to +1.sub.x and from Y.sup.2 +Z.sup.2 =0 to Y.sup.2 +Z.sup.2 =R.sup.2 for some length 1.sub.x , and radius R (typically R is on the order of 50mm). In general the wafers will be contained in a cylindrical volume having a cross-section of the same shape and size as the lateral cross-section (i.e. the cross-section in the y-z plane) of the wafers.

In the embodiment shown in FIG. 4 the inlets 419-421 are positioned in a linear array along a line 42 located at y=0 and z=1.sub.z where 1.sub.z is greater than R. The inlets are also spaced symmetrically about the point x=0, y=0 and z=1.sub.z. The inlets are positioned a few centimeters above the wafers to enable the boat containing the wafers to be slid into or out of the reaction chamber without having to first temporarily move the inlets out of the way. This spacing also allows the reactant gases supplied through each of the inlets to diffuse laterally in an approximately Gaussian distribution centered on the inlet from which those gases are supplied before those gases are drawn into the spaces between adjacent wafers.

For a selected one of the reactant gases in the first group of reactants, the resulting concentration as a function of x along any line having constant y and z coordinates within the volume containing the wafers is shown in FIG. 6. The concentration distributions from inlets 221, 220 and 219 are shown respectively as solid curves 61, 62 and 63. The total concentration distribution is shown a dashed curve 64. A similar pattern results for each line of constant y and z within the volume containing the wafers. For the linear array of inlets shown in FIG. 4, the maximum concentration in the volume containing the wafers will occur on the line y=0, z=R. Likewise the minimum concentration will occur on the line Y=0 Z=-R.

The difference between the maximum concentration and the minimum concentration within the volume containing the wafers is a function of the rate of gas flow through each of the inlets, the velocity of gas flow through each of the inlets and the x, y and z positions of each of the inlets. Note that the rates of gas flow can be regulated by the mass flow controllers such as controllers 211-214 in FIG. 2 or 311-314 in FIG. 3. For constant rates of flow, the velocities can be controlled by the controlling the sizes of each of the inlets. Since the variation in line width and undercut is proportional to the variation in thickness of the coating, the choices of at least some of these parameters are typically selected to minimize the variation in coating thickness.

Because gases of different masses diffuse at different rates, the concentration distributions shown in FIG. 6 will be similar but not identical for each of the gases. The coating thickness variation (i.e., the difference in thickness between the maximum and minimum coating thickness) produced on a batch of wafers is generally a function of the concentration of more than one gas. A theoretical prediction of the optimum choice of parameters to minimize variation of coating thickness is most easily made if the rate of reaction is made to depend on the concentration of a single reactant. In such a case the variation in thickness is proportional to the variation within the volume containing the wafers of the concentration of this single reactant. Often the rate limiting step involves some chemicals which are products of previous steps in the CVD reaction so that the relevant concentrations can be a complicated function of the rate at which each reactant is supplied to the reaction chamber. In some processes, the rate limiting step includes at least one reactant gas so that the resulting thickness can be made a function of the concentration of a single reactant gas by making the rate of the rate limiting step depend essentially only on the concentration of that reactant gas.

In a reaction of the form nl A.sub.1 +. . . nk A.sub.k react to form products (or intermediate products in a multistep reaction), the rate is generally proportional to [A.sub.1 ]nl . . . [A.sub.k ]nk where [A.sub.j ]is the concentration of the A-th chemical and nj is the stoichiometric coefficient of Aj in the reaction. However, these gases having a significantly higher ratio of concentration to the concentration of any of the other gases reacting in the rate limiting step than the ratio of their associated stoichiometric coefficients will be saturated so that the reaction rate is essentially independent of their concentrations. In a reaction involving a reactant in the rate limiting step, the overall rate of reaction can be made to depend essentially only on the concentration of that single reactant by selecting gas flow rates which makes the ratio of its concentration to the concentration of the other chemicals in the rate limiting step much less than the ratios of their associated stoichiometric coefficients. Since the degree to which the reaction rate is made essentially independent of the concentrations of the other reactants is an increasing function of the ratios between the concentrations of those other reactants to the concentration of the reactant on which the rate is to vary, the concentration of this latter reactant can be made small enough compared to the others so that the reaction is saturated to any degree desired. The degree to which the reaction is not saturated in the other reactants will show up as that part of the rate that does not vary linearly with the concentration of this reactant gas.

Even in this case of the rate depending on the concentration of a single reactant, although a prediction can be made by well known calculations of diffusive and convective gas flow, the calculations are rather complicated so that in general the optimization will be made empirically. The empirical work can be simplified somewhat by optimizing with respect to only a subset of the parameters. For example in FIG. 3 the linear array requires the z component of position of each of the inlets be equal to a common value 1.sub.z and that y=0 so that the optimization is made only with respect to the x position of each inlet. The minimization of the variation in coating thickness with respect to x, y and z for each inlet (or some subset of these parameters as in the case of a linear array of inlets) therefore defines the choice of inlet positions to inject the first group of reactants into the vicinity of the IC wafers. The inlet or inlets for the reactant and carrier gases not in the first group of reactants can similarly be separately supplied into the vicinity of the wafers but such a duplication of inlets is not required. If the inlet carrying these other gases introduces these gases into the reaction chamber away from the vicinity of the wafers but not in the region between the exhaust port and the wafers (in which case a large fraction of such gases would be swept from the chamber before entering the volume containing the wafers) then these gases will be enabled to move by convection and diffusion to produce a fairly uniform distribution at the edges of the volume containing the wafers so that a uniform concentration of these gases results in that volume. Because these gases are supplied separately from the first group of gases, they do not begin to react with the first group of reactants until arriving near the volume containing the wafers and therefore the gradient in concentration due to such reaction away from the wafers produced in the prior art embodiment shown in FIG. 1 is avoided.

In FIG. 7 are shown along lines of constant y and z the concentration distributions 71-73 produced by gases from inlets 21, 320 and 319 respectively and the total concentration distribution 74. Because the exhaust port produces a pressure gradient flow, the x-positions of the inlets 319-321 in FIG. 5 are shifted in the -x direction compared to the positions in FIG. 6 where the inlets are positioned symmetrically about the plane x=0. Similarly the concentration curves 71-73 are not symmetric in the x direction. However, the difference between the maximum and minimum coating thicknesses is still a function of the x-position of and the gas flow rate through each of the inlets so that these parameters can be chosen to minimize this difference. Typically, the optimum choice will be made empirically. However, if desired, the optimum values of the parameters can still be determined by straight forward but complicated calculations using models of convective and diffusive gas flow from a set of inlets through an array of parallel plates.

By means of well known empirical techniques, the resulting variation in coating thickness can be determined for each wafer. If the variation in coating thickness for a batch of wafers is larger than desired, then extra inlets can be added. For example, more than 3 inlets can be included in a linear array above the wafers or extra inlets can be included above the wafers to form a two dimensional array. Likewise, inlets on the sides (i.e., in the planes y=R' and y=-R'0 where R' is greater than R) can be included. In the embodiment shown in FIG. 3, extra inlets can be included below the wafers (i.e., in a plane z=1.sub.z " where 1.sub.z " is greater than R), but extra wafers in the plane z=-1.sub.z " for the embodiment in FIG. 2 would have little effect on the coating thickness because of the convective flow toward the exhaust port. In all of these cases, in order to simplify insertion of the wafers into the reaction chamber the inlets should be arranged to allow the wafers to be inserted into the reaction chamber without having to move the inlets.

If a large number of inlets is required, then instead of supplying the gases through distinct inlets as in FIGS. 2 and 3, the inlets can be produced as distinct holes in a nozzle as in FIG. 8. In that figure the inlets are manifested as holes 82 in one surface of a hollow nozzle 81. The gases in the first group of reactants is supplied to the nozzle by an inlet 83 for dispersal through holes 82 into the vicinity of the wafers. As in the case of separate inlets, the number and arrangement of the inlets can be selected to minimize the variation in coating thickness produced on a batch of wafers. Such nozzles have the advantage of providing a large number of inlets without a complicated structure but have the limitation that flow through each inlet can't be individually varied to adjust for changing conditions. Such nozzles can also be provided on the sides, above and below the wafers as was discussed above for the case of distinct inlets. Note that it is advantageous to introduce the reactants other than the first group of reactants at a point upstream and away from the region containing the wafers because this enables these reactants to arrive at the surface of the region containing the wafers (where reaction begins) in a more uniform distribution than if supplied closer to the wafers. This produces a more uniform distribution in the region containing the wafers helping to reduce any variation in thickness due to this variation in concentration.

To achieve less than a 10% variation in thickness within a batch of wafers coated in a furnace like that in FIG. 3, an arrangement of inlets as shown in FIG. 9 was utilized. A batch of 75 wafers were aligned in a boat to form a cylindrical region containing the wafers having a length 2(1.sub.x) of 56 cm and a radius R of 5 cm. The pressure was in the range 0.1-0.5 torr. and the temperature was in the range 400 degrees C. --475 degrees C. The low pressure reduces the rate of reaction (as can be seen by the rate equations discussed above) so that reactants can flow by convection and diffusion to produce a more uniform coating than at higher reaction rates. The WF.sub.6 was supplied at 200-500 cc/min. and H2 was supplied through the inlets at 300-500 cc/min. The H.sub.2 was supplied through tubes 910 and 911 having 1/4 inch diameter. Tube 910 was terminated at its end and had 1/8 inch diameter tubes 91 and 92 providing inlets for the H.sub.2 gas. Similarly 1/8 inch tubes 93 and 94 were connected to tube 911 to supply H.sub.2 2 gas. Each tube supplied gas into the chamber through a 0.014 inch diameter opening (i.e., inlets 95-98). Tubes 91-94 were curved to remain outside the volume containing the wafers to enable the wafers to be inserted without moving the inlets.

A circular cylindrical liner of 14 cm diameter encircles the sides of the volume containing the chips and the inlets are oriented to direct the reactant gases away from the x axis so that the gases impinge on the liner. This improves contact of the gases with the hot liner thereby improving the heating of the gases. This improvement in heating is especially important at the low pressures found useful in the tungsten deposition reaction because low pressures reduce heat conductivity thereby reducing the transfer rate of heat to the reactant gases. Since at such low pressures, the liner is heated primarily by infrared radiation from the heating coils, the liner should be formed of a material which absorbs infrared radiation and in addition is thermally conductive in order to adequately heat the impinging reactant gases. Suitable materials include molybdenum, stainless steel and silicon carbide. The use of a liner has several advantages. Conventional CVD reactor systems can be employed without the need to modify or construct a system to line walls which are thermally conductive and UV opaque. Also, the liner can be easily removed from the chamber by sliding it out the door of the chamber--this allows the liner to be removed from the chamber in order to remove reaction products which have deposited on the liner. If such deposited products are not periodically removed, heat transfer into the chamber is reduced and deposited products can flake off of the walls and deposit on the wafers thereby producing defects. If a liner is not utilized then the chamber must be periodically dismantled to enable these deposited products to be removed from the chamber walls.

Inlets 95-98 are located approximately in the plane Z=0. Inlets 95 and 97 direct gases in a minus y and minus z direction whereas inlets 96 and 98 direct gases in a positive y and positive z direction. This arrangement results in the inlets having a common circumferential direction so that the gases from the inlets not only impinge on the liner, they also receive a rotational motion around the axis of the liner. The total motion of the gases (including the flow along the x-axis toward the exhaust port) is therefore a generally spiralling motion which enables the reactants to diffuse into the gaps between adjacent wafers from the entire surface of the region containing the wafers.

Each of the inlets is located approximately 6 cm from the x-axis (i.e. 1 cm from the side of the cylindrical region containing the wafers). Optimum uniformity is achieved with inlets 95-98 placed approximately at the x-positions x='22 cm , -18 cm, 0 cm and 4 cm respectively.

When tungsten is deposited on a wafer by the CVD process described above, it is observed that the tungsten deposits on the wafer on regions of doped silicon (such as in vias) but does not deposit substantially on regions of silicon dioxide. Closer inspection of the wafer shows that some deposition of tungsten occurs over oxide regions at points of the oxide which serve as nucleation centers. For example, at sharp edges in the oxide regions it is observed that some deposition of tungsten occurs. Therefore, to facilitate the uniform chemical vapor deposition of tungsten onto a wafer surface, it has been found to be necessary to produce nucleation centers on top of the oxide regions. This is achieved by sputtering a thin (250 Angstroms) of tungsten onto the wafer before placing the wafer into a CVD furnace for deposition of a thick (20,000 Angstrom) layer of tungsten. In a typical process, a thin (300 Angstrom) layer of nickel is first sputtered onto the wafer to serve as an etch stop for plasma etching of the tungsten layer. The thin tungsten layer is then sputtered on top of the nickel and finally the thick layer of tugsten is applied by chemical vapor deposition. This process has two additional advantages: (1) the nickel layer prevents attack of the oxide regions by the overlaying tungsten layer and (2) the sputtered tungsten layer avoids the oxidation of the nickel when the wafer is transferred from the sputtering chamber to the CVD chamber--this is important because nickel oxide is difficult to etch.