JPS5956576A - Formation of thin film - Google Patents

Abstract

PURPOSE:To make the thickness of the coating to be deposited on wafers uniform, by supplying reactive gases to the inside of a reaction chamber through a pipe of which the one end is connected to means for supplying the reactive gases and the other end is positioned near the wafers. CONSTITUTION:The gaseous H2 which is the gas of the 1st group is supplied from gas source 28, 29, 210 to inflow ports 219, 220, 221 positioned near IC wafers 25. On the other hand, gaseous WF6 which is the gas of the 2nd group is supplied to the inside of the reaction chamber near a door 22 from a gas source 27. Electric current is conducted from a current source 24 to a heating coil 23 to control the temp. in the reaction chamber and H2 and WF6 are brought into reaction, whereby a coating layer is formed on the wafers 25. Since the mixture of the reactive gases is permitted only near the wafers 25, the uniformity in the thickness of the coating is improved.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to CV l) (che+n1 cal vapo
The present invention relates to a thin film forming method for forming a circuit pattern, etc. on a wafer using a method of forming a circuit pattern or the like on a wafer.

Conventionally, efforts have been made to miniaturize ICs. Reducing the circuit dimensions has the advantage of increasing the number of circuits per unit area, allowing more complex circuits to be fabricated on a single Ic wafer. This reduction in circuit size also allows the fabrication of high speed circuits that require less effort per circuit unit.

Unfortunately, however, scaling down dimensions creates a number of fabrication problems that are physically and qualitatively different from fabrication methods suitable for larger circuit configurations. If the measure is given by a factor, then the measure of the thickness of the interconnecting wires is given by the square root of the factor. Under such scaling criteria, the cross-sectional area of the interconnection wire decreases at a greater rate than the length of the wire, so the resistance of the interconnection wire will increase as the feature size decreases. This increase in resistance increases the power required for the IC wafer, which also increases the heat dissipation requirements for such wafer. The larger the resistor in the interconnection wires, the greater the risk of current spikes in the wires, and the greater the voltage drop caused by such spikes, the greater the risk of mis-switching the various transistors in these circuits. It is therefore imperative to use low resistance materials in interconnect terminations to minimize problems arising from these.Traditionally, aluminum interconnects One of the main reasons aluminum has been used as a commercial interconnect is its low resistance.

As a circuit becomes smaller, the current required by the circuit decreases at a slower rate than the cross-sectional area of the interconnecting wires.
Current density within these connections tends to increase with decreasing size. I with a connection width on the order of t microns
In C, the resulting current density causes electron migration within the aluminum interconnection wires, thereby creating voids and cracks within the wires. To eliminate these problems, refractory metals such as tungsten, molybdenum, beryllium, and tantalum are used for interconnect connections. This is due to the fact that these metals do not cause electron transfer (because they have a large ability to carry large currents). The use of refractory metals has the additional advantage of allowing high temperature treatment after the metallization step.

Among refractory metals, tungsten is the most promising choice for interconnect terminations (J, M.

Written by Shaw and J. M. Amlc [Vapo
r-(lepositedtungsten as
a + netallization an (
i i++tercnnn echionmate
real 1'or 5ilicon +Iev
ices-JltCA Iteview 197(1
(See June issue). Tungsten has a fairly low value of resistivity (approximately three times that of aluminum), has a relatively high melting and silicide formation temperature, and has a relatively high activation energy for self-expansion. There is. The high formation temperature of the silicon compound is critical since silicon compound films often crack due to volume changes associated with the formation of the silicon compound. Tungsten also has a coefficient of thermal expansion almost equal to that of silicon, so it
The stress inherent in the formed tungsten interconnect wire is relatively low (normal) when the Lower is preferred. Tungsten is generally quite corrosion resistant, so tungsten interconnects will not corrode even if the substrate is subsequently soaked in or cleaned with acidic or deionized water.

For the reasons pointed out above, work began in the late 1960's to develop a method for using tangents as an interconnection connection for 7% of IC ways. However, it has been difficult to form tungsten as a layer of uniform thickness.7 Thickness uniformity is important for creating interconnect fasteners of uniform width. Typically, during the etching step that defines the interconnect connections, the portions of the tungsten layer that are not to be removed are masked from the etchant.

However, even in the plasma etching method,
The tungsten under the mask undergoes some etching, so the width of the interconnect lines becomes narrower and less uniform.

If the deposited tungsten layer is not uniform, it is possible that the interconnect traces will be narrower and hollowed out in areas where the tungsten layer is thinner than average. When the width of the connection is 1 micron, the width becomes narrow like this, or a cut/darcut occurs, creating a fragment.As a result, tI'-(rt 7
The resistance per unit μ increases by 71
It becomes a line segment shape with reduced adhesive force.

Until now, metallization has most commonly been applied by sputtering or evaporation. Due to the high melting point of refractory metals, deposition by evaporation is incompatible with other steps in the IC manufacturing process (high temperatures of 4 degrees are necessary, good for sputtering), and deposition is inherently slow. In addition, the inherent stresses and resistances of the interconnecting connections are generally unsatisfactory. Therefore, the deposited refractory metal CV l) (ch
emical van>ordepos mouth 1 on)
can be mentioned.

Most conventional CVD methods have utilized a cold wall system in which the IC wafer is placed over a two-safter that is heated to the desired reaction temperature by radiofrequency heating.

Conventional reaction chambers using such seven stems are J,
“Vap” by M. Shaw and J. A-Amic.
or -1) eposited Tungsten a
s a Mctallization and Inter
Connection Mat, erial fa
r 5iliconDevices J and V page l
97 (according to the ILCA Review of June 1999, it is shown in FIG. 1 of the reference drawings.

JLT So 1F, 1) Kawa de Sei 1 Sen-1
(2) The number of wafers that can be processed is small. Therefore, for many TC'', the contact area between the wafer and the boat-shaped vessel is reduced compared to the contact area between the cisceptor and the cisceptor. The method using a boat-shaped container is
Heating a boat-shaped container with high-frequency heating cannot heat it sufficiently.

Therefore, the method using a boat-shaped vessel is the high-temperature wall method.

Illustrated in Figure 1 is a conventional hot-wall CVD system suitable for depositing tungsten on integrated circuit wafers by a reaction that reduces tungsten fluoride with hydrogen. ).

The reaction chamber in which the CVD reaction takes place is surrounded by a cylindrical wall 11, which is closed at one end by a door 12. The door 12 can be opened and closed freely so that the IC wafer 14 can be put in and taken out, and the door 12 can also be closed to seal the reaction gas. A cocurrent flow from the 1 ゛ tree 4. The IC wafer j4 is supported in the boat-shaped container 15 at its edge.

Reactant gases tungsten and hydrogen and carrier gas argon are supplied from source vessels 16, 17 and 18, respectively. These gases are fed to the gas flow sewing regulator 19
, ] l O and Ill, these regulators controlling the relative rates at which these gases are fed to the reaction chamber. These gas flow nodes 81 are also connected to the outside γ1 via piping systems ], ], 2, Il 3 and 114 to the joint chamber 115. External mixing chamber 1
The mixed gas mixed in 15 is supplied to the reaction chamber via pipe 116. Unreacted anti-Lv1 substances and reaction products that did not deposit on the wafer or chamber wall are discharged into the discharge boat 1.
17 from the reaction chamber. vacuum pump 11
8 is connected to a discharge boat to suck out these reaction products and unreacted gases. Vacuum pump 118 and gas flow regulator 19, +10.

111 is excited to determine the partial pressure of each reaction gas and (IG gas feed).

In order to increase the number of wafers in a chamber (CVT), the wafers are closely spaced on a boat-shaped vessel, so that the gap between adjacent wafers (Q) through which the reactant gas must enter is narrow.

As illustrated in the aforementioned literature, both cold-wall and hot-wall methods typically mix the reactant gases outside the reaction chamber and feed the reactant gases to one end of the reaction chamber while the reaction products are mixed. and unreacted gas are drawn out at the other end of the reaction chamber. Therefore, as the reactant gas concentration decreases as the reactant gas flows down the reaction chamber, the concentration of the reactant gas in the vicinity of the wafer near the end of the reaction chamber where the reactant gas is supplied will decrease as the gas is withdrawn from the reaction chamber. The reactant gas concentration near the end of the reaction chamber is higher than that in the vicinity.

Such non-uniformity in concentration results in variations in the final wafer coat and ink thickness between wafers. Furthermore, since the gas flow flows along the reaction chamber in a direction targeted to the wafer, the reaction gas enters between adjacent wafers on the boat primarily by diffusion.

This concentration results in a higher concentration of reactant gas near the edge of the wafer, which results in variations in the final coating thickness even within the same wafer. There will be variations in the connection width, and the coating will become thinner locally.
ll (minutes occur.

In the p c v n method, which utilizes a reaction with a relatively fast reaction rate, the CV[
) The reaction increases, thereby increasing the quaeno-to-wafer variation and therefore the wafer-to-wafer and intra-wafer variation potential. This increased reaction rate also increases the likelihood that CV l) reactions will occur at the edges of the wafers before sufficient time has elapsed for the reactant gases to diffuse into the gaps between adjacent wafers in the boat. I will do it. One existing method (N1cl+o
las Ishi.

Miller and I 5rael Being
by ass [Hot-WallCVD i'ungsl
en for V [S IJ, 5nlid 5tate
1'ecln+nlogy, l December 980, V
ol, 23. No. 12), the reaction rate is reduced by lowering the pressure and temperature of the CV D process and by using inhibitors. (However, this method not only sacrifices process speed, but also results in a 10% wafer-to-wafer and 8% within-wafer variation in coating thickness. There is a need for a method that maintains high productivity with respect to the method used and provides uniform coaching within and between wafers.

In accordance with embodiments of the present invention, a method and apparatus for uniformly coating an integrated circuit wafer by chemical vapor deposition is disclosed. This method provides a highly uniform coating suitable for manufacturing integrated circuits having interconnecting connections on the order of one micron in width.

+1 in preferred embodiment for high level of productivity
A hot wall chemical vapor deposition process is used in which the integrated circuit wafers to be coated are edge supported on a boat to form a linear array of parallel wafers. The wafer is therefore accommodated in a cylindrical region whose cross-section corresponds to the transverse cross-section of the wafer to be coated.

A single reaction that decomposes indoors and deposits one of the decomposition products'
This reactant is introduced in the vicinity of the wafer in order to improve the uniformity of coating thickness in the case of teeth. In order to improve the uniformity of the coating thickness formed by the reaction between the reactant gases, the reactant gases are introduced into the reaction chamber in small quantities (both in two separate groups). The CVD reaction begins only after the gases have been able to mix in the reaction chamber.Since the first group of reactant gases are introduced into the reaction chamber in the vicinity of the wafer, the CVD reaction occurs primarily in the vicinity of the wafer. become. Preferably, the reactant gases in this group are fed into the reaction chamber via several inlets located in the vicinity of the wafer such that the reactions in this group are uniformly distributed. The area containing the wafer is surrounded by a cylindrical liner, and the inlet I' is directed to flow the reactant gases into the liner for turbulent mixing and heating of the reactant gases. There is.

To improve the heating of such gases, the liner is formed of a fine material that is a good thermal conductor and is opaque to infrared light. In one embodiment, the wafers are transported between the discharge boat and the
Reaction products and unreacted gases are drawn from the reaction chamber through an exhaust port located at a point such that it is located between the inlet for the group of reaction gases.

The first group of anti-I gases is selected to include a first reactant gas in a CVD reaction that significantly affects the rate of the CVD reaction (ie, the rate-determining step). The first reactant gas has a 4 fl
Since the uniformity of the final coating thickness is strongly dependent on the uniformity of the concentration of the first reactant gas in the vicinity of the wafer, or is essentially dependent on this uniformity. It will depend only on

Since the -th anti-t6 gas is also selected to have the smallest molecular weight among the reaction gases appearing in the rate-determining step, the diffusion rate of the first reaction gas is small (both the rate-determining This rate control is slow relative to any of the other reactant gases that enter the stage (thus increasing the uniformity of the concentration of this rate control reactant gas in the vicinity of the wafer).

The inlet is located near the edge of the cylindrical region containing the wafers and allows the flow of these reactant gases in the first group of reactant gases to have a significant effect on the rate of deposition of CVD products on the surface of the wafer. Reduce variations in concentration to a minimum.

The flow of the inlet and the flow through the inlet are also chosen to minimize variations in the concentrations of these gases.

The present invention will be explained below using examples.

It is a figure showing an example. The reaction chamber is surrounded by a wall 21 and a door 22. A heating coil 23 is provided adjacent wall 21 to provide heat to the reaction chamber. Current is supplied to the coil from an IW flow source 24, which current is adjusted to control the salinity of the reactant gas within the reaction chamber. The IC wafers 25 are supported by their edges in a container 26, and each IC wafer 25 is arranged in parallel. The apparatus illustrated in FIG. 2 is suitable for supplying the reactant gases tungsten, oxidase flumeride and hydrogen to a reaction chamber; Other reactions can occur, such as the reduction of other components including refractory metals such as aluminum, molybdenum, and tantalum.

In the example illustrated in FIG. 2, the reactant gas is gas i*27.
28, 29 and 21O. The relative flow of these gases is controlled by gas system controllers 211, 212, 213 and 214. These controls are preferably ``I'' ylan free-flow seed controls, but other types such as needle valves are also capable of r+f capability.

In the case of decomposition c v 1) reaction, the reaction gas to be decomposed is injected into the warping chamber close to the wafer.

In the case of a reaction between several reactant gases, the reactant gas;
The uniformity of the thickness of the final coating deposited on the IC wafer is improved because it is fed into the reaction chamber in two groups that allow mixing only in the vicinity of the wafer. In the example of Figure 2, hydrogen is the first group of gases and tungsten hexafluoride is the second group of gases. In this example, the tungsten hexafluoride is supplied through tube 215; 215 delivers tungsten to fluoride to the reaction chamber near the door (i.e., closer to the door than the IC wafer, e.g., within a few centimeters of the door). The gases belonging to a group of reactive gases are located close to the IC wafer via -W216 to 218 (i.e., much closer to the IC wafer than the door, e.g., within 2.3 cm from the edge of the wafer). A set of inlets 219-221
supplied to A carrier gas such as argon can be fed parallel to each of the reactant gases as shown in FIG.
? It can be improved. However, for a constant inlet size and a constant reactant gas flow rate, the primary effect of the carrier gas is to increase the velocity at which the gas exits the inlet. In some cases, the increased flow rate can be beneficial because the reactant gas reaches the wafer fabrication region before the reactant gas concentration gradient is created.

However, in the apparatus shown in FIG. 1-1, a sufficiently uniform coating can be applied without introducing a carrier gas. Inlets 219-221 connect tubes 216-218 to flow rate controllers 212-214, so that these controllers control the flow rate of each inlet from 1 to 1a to 116.
1 can be changed separately. This individual control allows the flow rate to be varied as conditions change to maintain optimal uniformity. The diameter of each inlet [1] for a given flow rate determines the velocity of the inlet gas.

Unreacted reaction gas, any carrier gas, and undeposited reaction products are removed from the reaction chamber via exhaust boat 222. The discharge boat 222 is connected to a vacuum pump 223, which controls the gas flow rate.
1 to 214 to also control the partial pressure of the gas. In particular, vacuum pumps can run CVD reactions at or below atmospheric pressure, if desired for a given reaction. In this example, the discharge boat has an inflow of IC wafers D2
19 to 221 and the discharge boat 222. In the second embodiment shown in FIG.
It is location. The embodiment of FIG. 2 creates a convective flow of a first group of reactant gases parallel to the surface of the wafers, thereby enhancing the uniformity of the concentration of these gases within the narrow space between adjacent wafers. There are advantages.

Also, in order to allow the gases within the reaction chamber to reach the exhaust boat, the boat 26 can be permeated by the gas flow or the boat can be supported away from the exhaust boat so that these gases are not obstructed. A path must be provided to allow water to flow to the discharge boat without being washed away. The embodiment illustrated in FIG. 3 has the advantage of being in the form of a now commercially available reaction chamber and furnace apparatus, so that the cost of such an apparatus is much lower than that of the apparatus illustrated in FIG. In the embodiment of FIG. 3, the discharge boat 2
22 into the high temperature region generated by the heating coil 23. Despite the disadvantage of producing convection that is generally orthogonal to the Ueno, devices such as the one depicted in Figure 3 can produce tungsten coatings with less than 10% variation in coating thickness within the Ueno. It was successfully applied on 75 wafers at a time. Therefore, this apparatus is capable of uniformly applying large coatings while maintaining processing speed. 4 and 5 are enlarged views showing the spatial relationship between the inlets for the first group of reactant gases and the inlets for the first group of reactant gases. Although only five wafers are shown to aid in the process, in reality more wafers can be processed than this number. In addition, the spacing between wafers has been highlighted for better visibility. The wafer 25 includes two end wafers. The first wafer is shown as the leftmost wafer in FIG. 4 and the last wafer is shown as the rightmost wafer in FIG. The wafers are shown with the side edges as dashed lines and the ends as the first and last wafers.
It is housed in a circular right cylinder 41 defined by . A group of reference axes are represented in FIG.
The axes are oriented along the direction of the rows of wafers, the Y axis is oriented along the horizontally oriented lateral wafer length, and the Z axis is oriented along the vertically oriented lateral wafer length. The origin of these axes is located at the center of this volume, so for a given length/x and R (R is typically on the order of between 50), this volume is x from x=lx
&ko, and Y2+Z2=0 to y2+Z2= It2
It has spread to (・ru. General K, ueno・hahako C)
) Ueno C/) l value 110,000t7ij (i.e. Y-Z
4 dogs and the same shape as the cross section on the plane)? It is housed in a conical volume with a cross section of y dimension.

In the embodiment illustrated in FIG.
1 is Y-0 and Z=l! Is it a straight line arrangement along the straight line 42 located at z? Position 6, here 12
is a length greater than l (inflow 1'J + mama t
They are symmetrically spaced around the points X=Q, y=o and Z=/'. The inlet is Ueno・σ〕upper, 2
．． Since it is located at 3α, there is no need to first temporarily move the inlet of the Type 41 container containing Ueno 1 out of the passageway for Ueno 1 into the reaction chamber or into the reaction chamber. It can be slid out of the room. Having this gap between the inflow [1] and the gap also means that the reactant gases supplied through each inflow [1] are The supplied inflow location center can be diffused in approximately Gaussian distribution in 114 directions.

For a selected reactant gas belonging to the first group of reactant gases, the final concentration is determined as a function of It is illustrated in FIG. Inflow IJ22
The concentration distributions of 1, 22+3 and 219 are illustrated as solid curves i61, 62 and 63, respectively. The total concentration distribution is illustrated as a dashed curve 64. A similar pattern results for a straight line of pressure having constant Y and Z 0H in the volume containing Ueno. For the linear arrangement of inlets illustrated in FIG. 4, the maximum concentration within the volume containing the wafer occurs on a straight line where Y=o, Z=approximately. Similarly, the minimum density Y, =O.

It occurs on the straight line of Z=-4t.

The difference between the maximum and minimum concentrations within the volume containing Ueno is determined by the gas flow rate through each inlet, the gas flow rate through each inlet, and the X, Y, and second positions of each inlet. is a function of

The lj flow rate is controlled by the controllers 211 to 214 or the third controller in FIG.
The quality of the ridges can be adjusted by the ridge controllers 311 to 314 shown in the figure. For a constant meteor rate, the velocity can be controlled by controlling the dimensions of each inlet. Since variations in wire width and undercut are proportional to variations in coating thickness, the selection of at least some of these parameters is typically selected to minimize variations in coating thickness.

Gases diffuse at different rates in different layers, so
The concentration distributions illustrated in FIG. 6 are similar but not identical for each gas. The variation in coating thickness (ie, the difference between the maximum and minimum coating thicknesses) produced by a single batch of Ueno® is generally a function of the concentration of one or more gases. Theoretical predictions of the optimal choice of parameters to minimize coating thickness variations are made very easily if the rate of reaction depends on the concentration of a single reactant gas. In such cases, the variation in thickness is proportional to the variation in the concentration of this single reactant gas within the volume containing the urethane.The rate-limiting step is often a product of a previous step in the CVD reaction. Since several chemicals are involved, the relevant concentrations can be a complex function of the rate at which the reactant gases are supplied to the reaction chamber. In some methods, the rate-determining step includes at least one reactant gas. , the final thickness can be made a function of the concentration of a single reactant gas by making the rate-determining step dependent essentially only on the concentration of this reactant gas.

In reactions of the formula n p A, l+...4- n k A k which form products (or intermediate products in multi-step reactions), the rate is generally (Az)...[
Ak: Proportional to II.

Here, [A''] is the concentration of the chemical substance numbered A[]1, and nj is the actinic coefficient of Aj in the reaction.

However, these gases with ratios of their concentrations to the concentrations of any other gases reacting in the rate-determining step that are significantly higher than the ratio of their associated actinic coefficients will become saturated, so that the reaction rate will be become essentially independent of their concentration. In a reaction that turns the reactant gas blue in the rate-determining step,
The gas flow angle is such that the overall rate of the reaction is much smaller than the ratio of the relevant chemical coefficients of those chemicals to the concentrations of other chemicals within the rate-determining step of the concentration of that gas. By choosing the rate, it can be made dependent only on the concentration of that single reactant gas. The extent to which reaction rates are essentially independent of the concentrations of other reactant gases is an increasing function of the ratio between the concentrations of these other reaction gases to the concentrations of reactant gases over which the rates are to be varied. Therefore, the concentration of the reactant gas in this bender can be sufficiently small compared to the others, so that the reaction is saturated to any desired degree. The extent to which the reaction is not saturated with other reactant gases will be expressed as a portion of the rate that does not vary linearly with the concentration of this reactant gas.

Even in cases where the reaction rate depends on the concentration of a single reactant gas, predictions can be made by known calculations of diffusive and convective gas flows, but these calculations are rather complex and are therefore not optimal. This will generally be done empirically. The empirical work can be simplified somewhat by optimizing only with respect to the minor cents of the parameters. For example, in Fig. 3, in order to arrange linearly, the Z component of each inflow 00 position must be common (straight t!2
Since it is necessary that it is equal to and Y−(),
The optimization is done only with respect to the X position of each inlet. Therefore, for each inlet, the variation in coating thickness is minimized for X, Y, and Z (or some subset of these parameters, as in the case of a 1α linear geometry for inlet 1). Then, the selection of the position of the inflow l'' to be ejected near the first group of QIC wafers υ is determined. For reactant gases and carrier gases that do not belong to the first group of reactant gases, one or several shallow inlets can be separately supplied in the vicinity of the sea urchin, but the inlet I. There is no need to provide duplicates. if,
If the inflow carrying these other gases] introduces these gases into the reaction chamber away from the vicinity of the wafer but not in the region between the exhaust port and the wafer, then
In this case most of these lJ gases would be swept out of the reaction chamber before entering the volume of l1g volume), and these gases would then move by convection and diffusion. They can be distributed fairly uniformly at the edges of the volume containing the gases, so that these gases have a uniform concentration within this volume. Since these gases are supplied separately from the gases of the first group, they do not react with the reactant gases of the first group until they reach the vicinity of the volume containing the gases. Never start. Therefore, as in FIG. 1, a concentration gradient due to reaction far away from Way/S does not occur.

In Fig. 7, the inlet 3 is located along a constant Y and Z straight line.
The concentration distributions 71-73 produced by gases from 2, 1, 320 and 319, respectively, and the total concentration distribution 74 are shown. Since the discharge port creates a flow with a pressure gradient, the X positions of the inlets 319 to 321 in Figure 5 are compared to the positions in Figure 6, where the inlets are located symmetrically with respect to the plane of X = 0. Similarly, the concentration curve 7 is shifted in the -X direction.
1 to 73 are not symmetrical in the X direction. However, since the difference between the maximum coating thickness and the bulk factor 1 is still a function of the X position of each inlet and the gas flow rate with respect to this inlet, these parameters are σ-reputation F's 77fufutsuji=zu5i can be chosen to be 5 to minimize this difference. The choice to optimize will usually be made empirically. However, the optimum values of the parameters can also be determined computationally using a model of convective and diffusive gas flow from a group of inflows 1'' through an arrangement of parallel plates II.

According to known empirical techniques, the final variation in coating thickness can be determined for each wafer. If coating jJ for a batch of wafers
If the variation in the flow rate is greater than expected, another inlet can be added. For example, 31 or more inlets may be included in a linear array on the way;
It is also possible to form a two-dimensional array 11 by including another inlet in one of the wafers. Similarly, inlets on the sides (ie in the planes Y = It' and y = -it', where R' is greater than 11) can be included. In the embodiment illustrated in FIG. 3, another inlet is placed below the wafer (i.e.
ce," is greater than +t), but even if another way is provided in the plane 1=-2z" for the embodiment of FIG. Coating thickness will have little effect. In all of these cases, the inlet geometry is modified to simplify insertion into the wafer reaction chamber. It should be possible to insert the wafer into the reaction chamber without having to do so.

If multiple inlets are required, then why not pass the gas through a separate inlet as shown in Figures 2 and 3? Alternatively, the inlets can be made as separate holes within the nozzle as shown in FIG. In this figure, the inflow () is represented as a hole 82 in the -(Ill IrT) of the hollow nozzle 81. Gas belonging to the first group of reactant gases is supplied to the nozzle 81 by an inlet 83 and passes through the hole 82 to the wafer. As in the case of separate inflows 1, the inflows [] (and the arrangement should be chosen to minimize the variation in coating thickness applied to one batch of wafers). With such a nozzle, a large number of inflows can be achieved without complicating the structure.
The advantage of having a ! However, there is a limitation in that the flow through the meeting cannot be adjusted individually. Such nozzles can also be provided on the sides, top and bottom of the wafer, as in the case of separate inlets. Advantageously, the reactant gases other than the first group of reactant gases are introduced 1' at a point 11 further apart upstream of the region containing the wafer. This allows these reactant gases to reach the surface of the area containing the wafer (where the reaction begins) with a more uniform distribution than if they were supplied closer to the wafer. It is from. This results in a more uniform distribution in the wafer containing volume and also reduces thickness variations caused by concentration variations.

In order to reduce the thickness variation within a batch of wafers destroyed in the furnace as shown in @3 to 10% or less, the inlet arrangement as shown in Fig. 9 is adopted. used. (2) Wafers from a batch of 751 m were arranged in a boat-shaped container to form a circumferential substrate-like area for accommodating a wafer having a length 2 (/x) of 56σ and a radius 1t of 5CTL. The pressure is l), 1 to +1.5 torr
r (m is in the range of 1 (g), and the temperature is 400
The temperature ranged from 475°C to 475°C. At lower pressures, the reaction rate is slower (as the rate equations discussed above allow), so the reactant gases flow by convection and diffusion, creating a more uniform coating than at higher reaction rates. be able to.

WF6 was fed at 200-5000° and H2 was fed at 30° to 500° via inlet flow. 112
is supplied through tubes 91 () and 911 with a diameter of 635 CrrL (1/41 nch). The tube 91O terminates at its end and has tubes 9I and 92 with a diameter of 318 m (1/81 nch), providing an inlet for the H2 gas.
75(:m(1781nch) tubes 93 and 94/J
・'17911& is connected to supply II2 gas.

The container is Il, 03556 (, u (fl, (1141n
cb ) diameter opening [-] (i.e. inflow 1.195
~98) Gas is supplied into the reaction chamber at +1fi L. The tubes 91 to 94 are bent and remain facing R''l1 outside the volume containing the wafer, allowing the inflow 1'' to be inserted without moving the wafer.

A circular cylinder liner with a diameter of 14 CrrL surrounds the sides of the volume housing the clip, with inlets oriented to flow the reactant gas away from the X-axis. This changes the contact of the gas with the hot liner, improving heating of the gas.

Low pressures reduce the thermal conductivity and thus the rate of heat transfer to the reactant gases. This is important in tungsten precipitation reactions at low pressure. At such low pressures, the liner is heated primarily by infrared radiation from the heating coil, so the liner must be made of a thermally conductive material that is capable of absorbing infrared radiation and sufficiently heats the reactant gases. should be formed. Suitable materials include molybdenum, stainless steel and silicon carbide. There are several advantages to using liners. Conventional CVD reaction techniques include a one-plying apparatus with walls that are both thermally conductive and UV transparent. It can be adopted without the need for extensive modification or construction. Similarly, the liner can be easily removed from the reaction chamber by sliding it out the reaction chamber door. The liner can then be removed from the reaction chamber to remove the reaction products that have deposited on the liner. If this deposited product is not removed periodically, heat transfer into the reaction chamber is reduced and the deposited product can flake off the walls and fall onto the wafer, causing damage. . If a liner is not used,
The reaction chamber must then be periodically disassembled in order to remove these deposited products from the walls of the reaction chamber.

The inlets 95-98 lie approximately i'ρ in the plane Z=0. Inlets 95 and 97 allow gas to flow in the minus Y and minus Z directions, while inlets 96 and 98 allow gas to flow in the plus Y and +Z directions. With such an arrangement, the inflow L has an inner circumferential direction of J(through), and the gases from these inlets not only collide with the liner fins, but also flow around the axis of the liner. The flow of the gas (including the bath 5 flow in the X-axis towards the exhaust port) is therefore generally a helical movement, and by the helical movement the reactant gas accommodates the wafer. It can diffuse from the entire surface of the region into the gaps between adjacent wafers.

The valley inflow is approximately 6 cm from the gland on the X axis (i.e. IC'rrL from the side of the cylindrical area containing the wafer).
If you are at the place. Optimum uniformity is obtained when the inlets 95-98 are 11X = -221, with
Placed at -18cIn10 and 41 (and possibly deposited at -14).

When tungsten is deposited onto a wafer by the CVD method as described above, tungsten may be deposited onto the wafer over areas of doped silicon (such as in the vias), but the tungsten does not oxidize. Substantially no deposits are made on areas of silicon. Looking more closely at the wafer, some of the tungsten deposits occur on oxide regions at points where the oxide acts as nucleation centers. For example, some deposits of Damagesten are observed to occur at the sharp edges of the oxide regions. Therefore, to facilitate uniform chemical vapor deposition of tungsten on the wafer surface, it is necessary to create nucleation centers on top of the silicide region. This means that a thick (20,000 Å) layer of tungsten is deposited on a portion of the wafer before it is placed in a CVD furnace.
This is achieved by sputtering a 0 angstrom) layer. A typical method involves using a thin layer of nickel (300
A tungsten layer (Angstrom) is first sputtered onto the wafer to serve as an etch stop for the plasma etching of the tungsten layer. A thin layer of tungsten is then sputtered onto the nickel and finally a thick layer of tungsten is deposited by chemical vapor deposition.
u will be rubbed. - This method has two additional effects: (1) The nickel layer prevents erosion of the oxidized white area by the layer covering the tungsten. (2) The nickel layer prevents erosion of the oxidized white area by the layer covering the tungsten. 1) The sputtered tungsten layer avoids oxidation of the nickel during transport to the chamber. This is a hassle since it is difficult to etch the nickel oxide.

[Brief explanation of the drawing]

FIG. 1 is a schematic diagram of a conventional CVD apparatus. ? Figures PJ2 and 3 respectively show the first CvT (CvT) sandwiching method used in the thin film forming method of the present invention. FIG. 2 is a schematic diagram showing a second embodiment. FIG. 4, FIG. 5, FIG. 8, and FIG. 9 are enlarged views of the tubes of the CVD apparatus used in the thin film forming method of the present invention. Figures 6 and 7 show CVs used in the thin film forming method of the present invention.
An explanatory diagram of the D device. 16.17. L deficiency 3, 27, 28, 29, 210: Gas source, 13.23: Coil, 19.110, 111, 211. 212, 213, 21
4,311,312゜313.314: Gas flow rate controller, 24.223: ? I flow source +18,223: Vacuum pump applicant Yokogawa Huyulet Packard Co., Ltd. Attorney Tsuguo Hasegawa FIG 6 FIG θ FIG 7 FIG 9

Claims (1)

[Claims] A thin film forming method in which a wafer is placed inside a reaction chamber, and a reaction gas is supplied into the reaction chamber through a tube whose one end is connected to a reaction gas supply means and whose other end is located near the wafer.