IMPROVEMENTS RELATING TO A DEPOSITION PROCESS
Field of Technology
This invention relates to a method and apparatus for improvements relating to deposition processes. The invention is particularly useful for, but not limited to, chemical vapour deposition (CVD) for the production of semiconductor devices.
Background
The technique of chemical vapour deposition (CVD) is a well established method for depositing thin layers of material on a surface of a substrate (or wafer) , particularly for the production of semiconductor devices. During a CVD process, a gaseous compound of the material can be used to deposit the material on a heated substrate, such as silane for depositing silicon. This gaseous compound, referred to as the "reactant gas" , may be a single compound or a mixture of several compounds, the ratio of which can be adjusted to change the composition of the deposited material . The reactant gas may be diluted by the addition of another gas, referred to as the "dilution gas". The dilution gas can be passive or it may also have a cleansing function to reduce undesirable material (contaminants) becoming incorporated in the growing film. An etching gas may also be included in the process gas to control selectively the growth rate of the material to be deposited, thus allowing growth on one part of a mixed material substrate but not on another.
In all cases the process results in the deposition of the desired material (e.g. silicon) on the substrate and the liberation of gaseous by-products (e.g. hydrogen) which are either pumped away by a vacuum pump or are flushed away by the flow of dilution gas.
CVD processes can proceed at atmospheric pressure but various reduced pressure regimes are typically used for the deposition of certain materials to adjust the material
quality, rate of deposition and deposition uniformity.
The rate of deposition of material or a substance in CVD (often referred to as "growth rate" or "deposition rate") has a strong correlation with the surface temperature of the heated substrate and the arrival rate of the gaseous reactant molecules, amongst other factors. The relationship between growth rate, pressure and temperature for a typical deposition process (e.g. silicon from silane) can be shown on a graph called an Arrhenius plot 10, reproduced in figure 1.
The growth rate 12 as a function of inverse temperature 14 is plotted to yield a family of constant pressure curves 16, 18, 20 as shown in figure 1, where pressure Pι<P2<P3. In the low gradient portion 22 of curve 16 the growth rate is limited by the arrival rate of the reactant species at the substrate surface and is only weakly affected by variation of the substrate temperature. This is variously termed the mass transport limited zone, pressure limited zone or the P zone. In the high gradient 24 portion of curve 16 the growth rate is limited by the rate of thermally driven processes at the surface and changing the pressure has very little, or no effect. This is variously termed the kinetic limited zone, the temperature limited zone or the T zone.
When materials are deposited at relatively high temperatures (for example, silicon at >1000 degrees
Centigrade) the reaction is generally taking place in the P zone so relatively high reaction chamber pressures (atmospheric or just below) can be used to maximize growth rates . However, a high heat load can have detrimental effects on the substrate and any material already deposited thereon. For example, in the semiconductor industry it is increasingly desirable to deposit a series of thin layers of different materials on top of one another to very accurate thicknesses (to form quantum well structures, for instance) and with an extremely well defined boundary between each layer. It is extremely difficult to achieve these onerous requirements when depositing layers at high temperatures because an unacceptable amount of diffusion between the
layer boundaries occurs. Thus, the deposition reaction may have to take place in the T zone (low temperature) in which the growth rate i-s not increased by raising the pressure.
Furthermore, in the manufacture of modern semiconductor devices the uniformity of the deposited film's thickness and composition across the substrate surface can be crucial . When a deposition system is operated in the P zone of the Arrhenius curve, gas flow, gas mixing and reactant depletion characteristics strongly affect the deposition uniformity. Turbulence in the reactant gas stream is best avoided since it is difficult to control; the flow velocity across every part of the substrate needs to be optimized and the reactant gas must not be too strongly depleted as it passes over the substrate . High flows of dilution gas are sometimes employed and substrate rotation is normally used to minimise the effects of such potential process disturbances. When films are deposited in the T zone of the Arrhenius plot the deposition rate is almost entirely controlled by the temperature (as is evident from figure 1) . Thus, it is necessary to maintain a high degree of temperature uniformity across the surface of the substrate.
Several published patents and patent applications attempt to partially address some of these issues in different types of deposition systems. For example,
US5261960 describes a CVD apparatus with a reduced cross section. The velocity profile of the gas passing in a linear flow pattern over a substrate is controlled to try to prevent deposition of material in unwanted locations in the chamber .
US6143077 describes a source gas delivery system which injects gas from above the substrate. The delivery system is disposed above the substrate and generally cone shaped for controlling gas flows from the center of a circular substrate and the radial flow velocity of the gas as a function of distance from the center of the substrate.
US5294778 describes a substrate platen heater system for low pressure CVD systems which allows the temperature of a substrate platen in a CVD apparatus to be adjusted so that
a uniform or non-uniform temperature profile is provided across the platen.
Each component in the reactant gas mixture has an effect on the deposition characteristics represented by an Arrhenius curve. The deposition characteristics of individual reactant components may either be independent of, or dependent upon the other materials in the gas mixture, with varying degrees of dependency. Thus, fully characterizing and controlling all interdependencies within a complex process is an extremely daunting task. Once a deposition apparatus is constructed it becomes necessary to evaluate the so called "process space" of the apparatus and to choose the best process conditions for the required deposition. By "process space" we mean a multi-dimensional environment in which a growth process occurs, where each dimension is a variable which affects the growth characteristic .
When using conventional CVD apparatus such process space studies require the deposition and measurement of many test films, each grown on separate substrates, with system adjustments of temperature, pressure, gas flow rate and gas mixture being made between each deposition. This is very labour intensive and costly, which means that studies are usually very restricted in their extent. Data points are usually taken relatively distant from one another and it is presently assumed that deposition characteristics vary smoothly between the data points .
Summary of the Invention A first aspect of the present invention resides in a method of depositing a material onto a substrate, comprising; a) heating the surface of the substrate so as to provide a substantially smoothly varying temperature gradient across at least a part of the surface, b) exposing said surface of the substrate to a process gas for deposition of material onto said surface, c) determining the deposition rate of said material onto said surface, as a function of temperature, which is variable across the
surface, d) identifying, from the determined deposition rates, a critical temperature T of the substrate's surface, at which the growth rate switches between a substantially pressure sensitive/temperature insensitive region to a substantially temperature sensitive/pressure insensitive region; and e) depositing material onto the or a second substrate substantially at the critical temperature T. The present invention also provides a method for generating data for use to characterise a process of depositing material on a surface of a substrate in a process chamber, comprising, a) disposing said substrate in said chamber, b) heating said surface so as to provide a substantially smoothly varying temperature gradient across at least a part of the surface, c) exposing said surface to a process gas for deposition of a substance on said surface, wherein the process gas comprises a reactant gas, d) determining the deposition rate of said material onto said surface, as a function of temperature, which is variable across the surface, at a plurality of points upon the surface, e) generating a set of data corresponding to the deposition rate of the material on said surface with respect to temperature, whereby the set of data is used to determine an optimal deposition rate.
A further aspect of the present invention resides in a method of depositing a material onto a substrate, comprising: a) heating the surface of the substrate so as to produce a substantially smoothly varying temperature gradient across at least a part of that surface; b) exposing the surface of the substrate to a process gas at a plurality of partial pressures P, for deposition of material onto the said surface; c) determining the deposition rate of the said material onto the surface of the substrate as a function of temperature at a plurality of points upon the surface, for each of the plurality of partial pressures; d) identifying, from the determined deposition rates, a critical partial pressure Pc for the process gas, defined as the partial pressure P, for a given temperature T, at which the growth
rate switches from a substantially pressure sensitive, temperature insensitive region to a substantially temperature sensitive, pressure insensitive region; and e) depositing material onto the or a second substrate substantially at or above the critical partial pressure Pc. The present invention still further provides a method for generating data for use to characterise a process of depositing material on a surface of a substrate in a process chamber, the method comprising; a) disposing said substrate in said chamber, b) heating said surface so as to provide a substantially smoothly varying temperature gradient across the plurality of surface points, c) exposing said surface to a process gas for deposition of a substance on said surface, said process gas comprising a reactant gas, d) determining the deposition rate of the material onto the surface at a plurality, n, of points on the surface, e) obtaining a value for the deposition rate as a function of temperature T from the determined deposition rate for at least some of the n points, and f) generating a set of data corresponding to the deposition rate of material at the at least some of the n points, whereby the set of data is useable to determine an optimal deposition rate.
Advantageously, the surface of the substrate can be heated unevenly, but smoothly, so that a range of growth rates can be determined for various process gas parameters and substrate temperatures . This can then be used to determine an optimum growth rate required for a user's requirements. Also, since the temperature gradient is generally smooth, any inflections in the growth rate curve close to the growth rate of interest to the user, can be detected or determined with a relatively high level of accuracy. The growth rate at discrete points on the substrate can be determined, and because each point is at a different temperature to neighbouring points by virtue of the uneven heating of the substrate, the growth rate as a function of temperature can be derived for the substrate.
Advantageously, further substrates can be subsequently processed with the chamber parameters set to the optimum
condition, the substrate being heated uniformly to an optimum temperature. Preferably, one or more further substrates can be processed in the same chamber which is used to determine the critical temperature or pressure for best growth conditions. Thus, no further characterisation of the process chamber should be required, unless, of course, different deposition processes are to be carried out in the chamber.
In this way, a deposition process can be optimised using a single combinatoral experiment, thus greatly reducing the time necessary to evaluate the chamber's characteristics. The same chamber is then used for wafer production, thus enhancing the growth capability of deposition chambers; material can be deposited on a substrate to a greater accuracy or with greater reliability from one substrate to another. Furthermore, an optimised process chamber should greatly reduce any waste products, or unused process gases, thus potentially reducing cost of manufacture and release of environmentally damaging by- products. This is achieved by optimising the consumption of reactant gas which interacts with the substrate's surface. Of course, any reactant elements which do not react to deposit material on the substrate pass through the chamber to the exhaust. By optimisation of the reactant gas consumption, we mean that substantially all of the reactant molecules interacting with the surface of the substrate are reacting to deposit material thereon.
A yet further aspect of the present invention resides in a controller for controlling a process used to deposit a layer of material on a substrate disposed in a process chamber; the process chamber comprising an inlet for supplying a process gas which comprises a reactant gas, to the substrate disposed in said chamber, and an outlet; the controller comprising a look-up table containing data for use by the controller to determine at least one chamber condition for use when processing said substrate in the chamber; wherein the controller is arranged in communication with at least one of a plurality of process control devices, the plurality of devices comprising, one or
more sensor disposed in the chamber for determining conditions therein, a heater for heating the substrate, a throttle for adjusting the process gas flow rate through, or pressure in, the chamber, and wherein, the controller is arranged to monitor and adjust the at least one chamber condition so that material is deposited on the surface of the substrate at an optimal deposition rate.
The controller can control and adjusts the chamber conditions to achieve efficient use of the reactant gas. The process gas can comprise one or more reactant gases and any number of other gases. The controller can control the flow rate of each of the gas species which constitute the reactant gas on an individual basis. Also, as previously mentioned, the substrate can be heated unevenly to a range of temperatures across the substrate's surface. In this way, the deposition rate of material on the substrate can be determined with respect to various changing process parameters. Advantageously, the controller can be arranged to maintain an optimal process conditions, in accordance with a user's requirement. Several further substrates can then be processed in the optimised chamber, with the advantages previously described.
A computer programme which carries out the method step when run on a computer is also provided.
Description of the Drawings and Preferred Embodiments
Embodiments of the present invention are now described by way of example, with reference to the accompanying drawings, in which:
Figure 1 is an Arrhenius plot showing film growth against substrate temperature at three different reactant gas pressures;
Figure 2 is an enlarged area of the plot of figure 1; Figure 3 is a three dimensional representation of the plot of figure 1;
Figure 4 is a three dimensional representation of the deposition rate of a substance on an unevenly heated substrate;
Figure 5 is a schematic diagram of an apparatus embodying the present invention;
Figure 6 is an Arrhenius plot similar that shown in figure 1; Figure 7 is a graph of pressure on a substrate's surface with respect to position on the surface;
Figure 8 is a graph of film thickness deposited on a substrate with respect to position on the substrate's surface ; Figure 9 is a plot of reaction gas partial pressure with respect to position on a substrate's surface at various depletion rates; and
Figure 10 is a flow diagram of process steps embodying the present invention. To achieve a robust production process in semiconductor manufacturing a full understanding of, and an ability to adequately control, complex deposition processes are essential when relatively thin layers of material are required. There is value in optimizing the deposition process to run under conditions that are on, or close to the knee of the Arrhenius curve (that is the point where the T zone and P zone meet as indicated by numeral 26 in figure 1) . This results in deposition of a uniform film with the minimum consumption of reactant gas (if depositing in the T zone) or at optimum growth rate with minimum thermal budget for the substrate (if depositing in the P zone) .
Our studies of deposition rates in this region of the Arrhenius curve show that a relatively simple deposition process, such as deposition of polycrystalline silicon from silane (SiH4) , proceeds with an unexpectedly varying growth rate over a small temperature range at or close to the knee of the Arrhenius curve. These results have not, to our knowledge, been observed before and are significant if growth of material at or near to the Arrhenius curve knee is desired.
This growth rate phenomenon is shown in figure 2 which shows that the temperature dependence of the growth mechanisms is more complex than previously thought . As a
result, determining the effect of complex growth kinetics on characteristics of the deposited material (such as thickness, smoothness, step coverage, stress, adhesion, crystallinity and electrical or optical qualities) is now considered vital for the production of the optimized films for modern and future semiconductor devices when the films are grown at the knee. If the fine details of the Arrhenius curve, such as is represented by the solid curve in figure 2, are overlooked, process conditions may be chosen which are sub-optimal and/or which can result in poor process stability; the product of the process is most likely to be unsatisfactory. It is not yet clear why the growth rate curve at the knee includes these sub-features, or inflections. A description of how the data is obtained for use in the plot of figure 2 (and figure 4) is described later.
All of the variables mentioned above can be plotted as axes on a graph to show the effect of variations in the process parameters on selected characteristics of the deposited material . Such graphs provide maps of the "process space" described above .
The Arrhenius curve shown in Figure 1 is a two dimensional representation of a process space, showing the effect on growth rate of varying the substrate temperature. If the changes in the Arrhenius curve as a function of another variable such as pressure or gas composition are plotted on a third axis, a three dimensional representation of the process space 50 is created, as is shown schematically in figure 3. The front face 52 of this data box maps out an Arrhenius curve, while the third dimension 54 maps out the variation of the Arrhenius curve as a function of pressure. The data at the left side face 56 show how deposition rate increases with pressure at only relatively high temperatures in the P zone. Other process variables, such as the effect of gas flow rate and pressure on film uniformity (amongst others) , could also be mapped in a similar way. The process space is therefore n dimensional, where n represents the number of variables in the process. Thus, it can be seen that such
process spaces can be highly complex with large number of variables .
We have found that a solution to the problem of mapping this highly complex process space is to conduct combinatorial experiments on a substrate in a process chamber. The experiment involves generating a matrix of different process conditions simultaneously at different points on a single substrate which is processed in a single deposition run (or on multiple spatially distributed substrates if more than one substrate is used) .
Combinatorial techniques have been used in the past to study epitaxy processes and examples can be found in the following documents.
US6364956 describes an apparatus wherein a programmable flux from each of a number of evaporation sources is applied to a substrate or substrates in a way which deposits a film containing many different material compositions in a single experiment. The apparatus may also be used to homogeneously mix the two or more target materials at an atomic level, since the materials are deposited onto the substrate simultaneously.
JP01329366 (Japan Science & Technology Corp) describes a mask which is moved over a substrate while film composition and reaction conditions are varied, thus producing a pattern of films on different areas of a single substrate.
US6344084 describes a combinatorial molecular layer epitaxy apparatus which is useful for synthesizing organic, inorganic or metallic superstructures comprising different materials and substances, and which is especially useful for efficiently searching for new substances in a short period of time. The apparatus is capable of synthesizing, in a series of reactions, a group of substances in an epitaxially grown super lattice structure and in a systemically controlled manner.
WO0208487 describes a chemical vapour deposition apparatus with a segmented gas injection showerhead that exposes different parts of the substrate to various gas flow environments during a single experiment . The shower head
design allows impinging gas fluxes and compositions to be varied independently as a function of position on a substrate's surface. Each segment of the showerhead includes a gas inlet and exhaust to minimise inter-segment mixing and enable a high degree of spatial distribution control of gas flux seen by the substrate.
EP1161986 describes a molecular beam epitaxy apparatus which combines different fluxes through a mask system. While these known systems are each valuable for experimentation in their respective applications, their complexity renders them unsuitable for use in semiconductor production. Thus, knowledge gained from their use would need to be transferred to different apparatus for semiconductor production. This requires further equipment cross- calibration and is most likely be very time consuming, costly and prone to error.
Referring back to figure 2 (and previous descriptions of that figure) , it can be seen that the variation in growth rate as a function of temperature for polycrystalline silicon grown by CVD does not change smoothly. The relatively low temperature resolution of test equipment (and the data previously produced by the equipment) used to generate Arrhenius plots has previously prevented these fine details from being observed. If deposition were performed using parameters at a point near one of the inflections, or steps shown in figure 2, a very small variation in the temperature of the substrate could have a significant impact on the deposited film thickness. This can result in a process going outside an acceptable "process window" and semiconductor device yield being reduced. In a worst case, devices on a substrate will be lost or rendered useless, ultimately resulting in a large economic loss for the device manufacturer.
Also, the temperature dependent processes that cause these inflections in the growth rate may also have an effect on other important characteristics of the growing films, such as surface smoothness, selectivity, defect density, refractive index, step coverage, dopant incorporation rate, grain size and preferred grain orientation. Thus, to ensure
all these conditions are optimized, it is important to have knowledge of the process space and how chamber process parameters affect the space. Therefore, it is both valuable and important to be able to observe any such characteristics of film growth, then measure their effect on the physical, optical or electrical properties of the deposited films and finally choose and set process conditions that allow mass production of films, or semiconductor devices, with the required characteristics. The conditions necessary for running an optimized process in the T zone can involve programming a process to run at point A of figure 2 while avoiding point B. Another desired set of film characteristics may require setting a process to run at points B or C while avoiding A or D. A process that has to be run in the P zone (for example to allow variation of the reactant gas partial pressure to regulate the stoichiometry of a multi-component film) but which must be conducted at the lowest possible temperature to minimise the thermal burden imposed on the substrate, requires selecting point D and avoiding the inflection to point C.
In embodiments of the present invention, it is possible to observe a highly resolved temperature section of the Arrhenius curve on a single substrate by generating a programmable and quantifiable temperature profile across the substrate, the extent of which is referred to as the "temperature window" .
Each area of the substrate is subjected to material growth at a slightly different temperature to that of the neighbouring areas and the temperature gradient across the wafer preferably varies smoothly. This allows deposition of a substance to be studied under a wide range of smoothly varying temperature conditions in a single experiment and on a single substrate. By mapping the deposited film's characteristics (after growth) it is possible to directly correlate growth rate against the various temperatures of the substrate.
Mapping the temperature dependency of the growth process with high resolution is a first important feature of
the present invention. A temperature range that produces material growth characteristics of interest is then identified. More specifically, the temperature range within which the knee of the Arrhenius curve for a given deposition reaction is identified. The dynamic range of the temperature range should be chosen so that sub-features of the knee (i.e. inflections) be distinguished to suitable degree of resolution.
It is then possible to position the temperature window in the temperature region of interest and conduct experiments with variations of other process parameters such as gas flow or pressure, and map out a multi-parameter process space to determine further important process interdependencies on growth rate. It is preferable to perform such experiments on a single substrate by varying the other process conditions such as pressure or gas flow as a function of time, thereby producing a three dimensional matrix of different process conditions which varies laterally as a function of temperature and in depth as a function of time. There are some process changes with time which may not be determinable from these analytical methods. For instance, it would not be possible to detect changes of flow rate of a reactant gas which only contains silane because changes the deposition rate (i.e. mm per hour) of only silane on a substrate would not be detectable (unless the reactant gas was changed at a set time so that a 'marker' layer were deposited) . However, changes of dopant levels with time can be detectable if the concentration of dopant is controlled in a known fashion, for instance. As a result, the dopant concentration in the material deposited varies in a known way with time. Thus, knowledge of the depth of material deposited over a known period with a certain dopant concentration allows one to calculate the growth rate of the material; where the concentration changes occur indicate a change of growth rate.
Depth sensitive analytical techniques can then be used to characterise the results of deposition under a large number of combinatorial conditions, rapidly adding detail to
the process space data set with minimum expense and consumed materials. Analytical techniques suitable for this task include secondary ion mass spectroscopy, x-ray diffraction and cross-section electron microscopy. These techniques are usually carried out after the substrate has been processed. The electron microscopy method requires the substrate to be cleaved before it is analysed.
Different precursor gases require different energies for decomposition when used in CVD processes. As a result, different reactant gases are sometimes chosen to optimise deposition rates in different temperature regimes (for example, by using disilane instead of silane for silicon deposition) . The temperature dependence of the deposition reaction and etching reaction is different if deposition processes involving mixtures of deposition reactants and etching reactants are used (such as might be used when controlling deposition selectivity) . When these reactions take place simultaneously during a process they can interact in a way which makes it more difficult to optimise the process, especially as a function of temperature.
Therefore, the ability to map out the selectivity of the deposition process in fine detail with a smoothly varying temperature profile across the substrate is an advantage provided by the present invention. Referring to figure 4, growth rate on a substrate with respect to relative positions on a substrate's surface is represented by a three dimensional plot 70 of our experimental data. It can be seen that the growth rate is at a maximum near to the centre of the substrate and tails off towards the edge. The substrate is disc shaped and material is deposited on its surface which is heated in a non-uniform manner. The temperature profile of the substrate's surface on which deposition occurs is generally dome shaped, with the centre of the substrate having a higher temperature than the edge. A method and apparatus for achieving this heating characteristic is described below.
From figure 4 it can be seen that the growth rate varies as a function of substrate temperature. The point at which growth changes from the T zone to the P zone is
clearly visible and a large variation of growth rate gradient can be seen. This change in gradient also has subtle variation associated with it, as can be seen at steps 72 and 74 in the growth rate of figure 4. [The plot in figure 2 is taken from normalised data of the growth rate shown in figure 4, at the knee in the growth rate curve] . A substrate heater comprises a multi-zone radiative heater element disposed behind the substrate to directly heat the surface of the substrate on which deposition does not occur. A central zone of the heater has a generally circular envelope in which a graphite heating element meanders. Preferably two outer heater elements surround the central elemental concentrically. These outer elements can be configured in a spiral, or in a circular shape, and are also made from graphite. Each heater element has independent power control units so that a radial temperature gradient can be produced on the substrate's surface. Of course, the heater can also be arranged to provide a substantially uniform temperature over the substrate's surface, taking account of heat loss at the edges of the wafer.
The heater needs to be calibrated before meaningful results from experiments can be obtained. This is achieved by using temperature calibration substrate, such as the commercially available Sensarray (RTM) substrate, which has thermocouples embedded in its surface.
In use, the power to each heater element can be varied, the resulting change in temperature (and temperature gradient) can be monitored and the results can be stored in a look-up table or database for future reference. The stored data should include a variety of temperature dynamic ranges, variations in temperature and a variety of mean temperature values on the substrate's surface.
Referring to figure 5, a chemical vapour deposition (CVD) apparatus 110 embodying the present invention is shown in highly schematic form. A single substrate 112 is mounted in a process chamber 114 in which the CVD process takes place .
A mixture of gases 116 enters the apparatus 110 at
input port 118 from a gas delivery system, not shown. The delivery system is arranged so that the concentration of each gas in the gas mixture can be varied independently from one another. This is achieved by a series of automated valves disposed on each gas supply to the delivery system. Each valve is controlled by a central controller 132, and thus the concentration of each gas component in the gas mixture is controlled and can be varied as required. Each gas mixes with one another, either in a dedicated mixing region, or as the gases pass along a pipe or conduit to the input port . The flow of each gas can be monitored using a dedicated mass-flow monitor unit disposed on each gas supply pipe. Such monitors are useable in conjunction with the valve 120 to control concentration levels and flow rates. Valve 120 9s controlled by the controller 132 and can be adjusted to vary the flow into the chamber 114.
A more preferable arrangement is to use a mass-flow controller on each gas supply pipe. Each mass flow controller comprises an automated valve which receives command instruction from the controller 132. The controller's instructions are used to control the flow of gas through the valve. A flow signal is returned to the controller for display purposes. The total flow of gas into the chamber 114 is determined by the sum of flows for each gas through its dedicated mass-flow control unit.
Thus, the gas flow into the chamber 114 is set by the controller and no other flow control element is required. Further control elements, such as valve 120 downstream of the input part 118 can cause gas build-up problems if the valve is set to restrict the flow of input gases to a level below that set by respective mass flow controllers.
An alternative method to control the concentration and flow of gases into the chamber is to control the partial pressure, or concentration, of each gas in the volume before the valve 120. A partial pressure sensor, such as a mass spectrometer, can be used to monitor partial pressures of each gas, and then adjust the automated valve upstream to control gas flow and/or concentration levels of the constituent gases into the mixing space before valve 120.
The valve 120 is controlled by controller 132 and can be adjusted to vary gas flow into the chamber.
Another alternative would be the use of mass flow controllers to control relative proportions of gases flowing into the mixing space before valve 120. A pressure gauge is required to monitor the overall pressure in the mixing space and this overall pressure value is feedback to the controller so that it can control the mass-flow controllers and maintain a constant pressure in the mixing space . The flow rate valve 120 is then used to control the overall flow into the deposition chamber 114.
The gas mixture contains the reactant gas or gases and preferably other process gases, such as an appropriate etching gas, for example, and each gas constituent having different, variable concentration levels with respect to the other gases comprising the mixture. The mass flow controllers or valve 120 (depending on the gas concentration or flow arrangement used) act to limit the gas flow into the chamber during substrate processing. The gas mixture flow across the chamber as indicated by dashed arrows 122 in a generally left to right direction in figure 5. The gas mixture reacts to deposit a layer of material on the substrate surface.
The substrate 112 is mounted on a heater 124 arranged to heat the back surface of the substrate. It is preferable for the surface on which material is being deposited to face down and away from the heater elements. This is advantageous because any contaminants from the heater fall onto the back surface of the substrate and do not contaminate the front surface on which deposition takes place.
The gas mixture exits the chambers at exhaust port 126. Incorporated with the exhaust port is a variable conductance valve 128. This valve acts to restrict or limit the flow of gas leaving the chamber. The valve 128 is used to maintain the chamber pressure at a predetermined set-point. The reaction chamber pressure can be measured using a sensor 134. Capacitance manometers are suitable for this sensing task. The controller compares this signal to the predetermined set-point and adjusts the valve 128
accordingly so that the chamber pressure matches the set- point .
The gas mixture is drawn through the apparatus by a pump 130 located to blow or suck the mixture through the system. Alternatively, two or more pumps could be arranged, at either ends, or on both sides, of the apparatus to draw the gaseous mixture through the apparatus .
The controller 132 is arranged to monitor and control conditions in the chamber, such as the substrate temperature, chamber gas pressure and the gas mixture composition. The pressure of the gas in the chamber is detected by a sensor 134 coupled to the controller. Alternatively, or additionally, a flow-meter (not shown) can be incorporated in the chamber to detect the flow of gas through the chamber. This flow-meter can be coupled to the controller to provide it with appropriate data.
The controller has access to a look-up table or database 136 which contains data which can be used to optimise deposition process by controlling the flow valve, pressure valve, and/or the gas pump. The type of data required for the optimization process is described below. The data in the look-up table provides means for the controller to calculate reference, or theoretical chamber conditions to which the actual chamber conditions can be compared. The comparison allows the controller to ensure the actual chamber condition do not drift or deviate from preset parameters. Should the chamber conditions deviate outside the parameters, then the controller adjusts the appropriate chamber device (substrate heater, inlet valve, or the pump speed, for example) to that the chamber conditions returns to the desired reference value. Precautionary measures are installed to prevent adverse feedback in the control loop.
The deposition rate at discrete points on a substrate is dependent on the temperature at that point, as discussed above. The deposition rate is measured by exposing a non- uniformly heated substrate (i.e. a substrate which is heated so that it has a temperature gradient across its surface) to deposition chemicals for a fixed period of time and then measuring the thickness of material deposited at each
discrete point on the surface. Because each point is at a different temperature during deposition, the deposition rate changes from point to point. Thus, the growth rate across the substrate ' s surface can be derived as a function of temperature. This growth rate data is used to determine the optimal conditions in the chamber for deposition.
As previously described, when a CVD process occurs in a chamber operating substantially in the T zone, the uniformity of the deposited layer across the substrate is almost entirely controlled by the uniformity of temperature across the substrate's surface. Generally, there are more than enough reactant molecules hitting the substrate surface to feed the thermally regulated chemical reaction deposition. Thus, if the gas flow and chamber pressures are set at arbitrary values, too much reactant gas can flow through the chamber without reacting to form a deposited layer and a substantial proportion of the reactant gas is wasted.
An optimum condition occurs at a given temperature when the process is conducted at a pressure which is sufficient to keep the growth rate in the T zone whilst a minimal amount of reactant gas is wasted. This optimal condition occurs at a critical pressure, Pc. Embodiments of the present invention identify the critical pressure for certain operating conditions and deposition recipes and cause the CVD system to operate at, or near, this pressure, thus simultaneously achieving optimum process uniformity and optimum utilisation of the process materials.
Referring now to figure 6, a substantially similar plot to that shown in figure 1 is provided. However, in figure 6 P2 = pc, P3>PC and P!<P0. Vertical broken line 160 indicates the given temperature T at which a growth rate associated with a chamber operating at the critical pressure deposits materials in the T zone and achieves optimized process conditions for P2. A chamber operating with a pressure P3 would also achieve the same growth rate, as indicated by the horizontal broken line 162 since a chamber operating at P3 would deposit material in the T zone of a growth rate curve. However, if the pressure in the chamber drops below Pc, to
Pi for example, then the growth rate also drops to an amount indicated by a second broken horizontal line 164.
With reference to figures 7 and 8, a plot of reactant gas pressure and deposited film thickness as a function of position across the substrate's surface are shown respectively. Broken line 170 indicates the pressure of reactant gas across the substrate's surface and shows the reactant gas pressure is always above the critical pressure, Pc shown by broken line 172. Thus, the thickness of a layer deposited on the substrate is uniform, as indicated by line 174 in figure 8. However, should the pressure across the substrate's surface drop to a level indicated by solid line 176 in figure 7, which has a central position above Pc and lateral portions below Pc, then the thickness of material deposited on the substrate is uniform where the reactant gas pressure is above Pc, and is reduced (because of the reduced growth rate associated with lower pressure, as shown in figure 6) at the edges of the substrate, as indicated by line 178 in figure 8. Hence, the growth must be conducted within a narrow range of pressures whilst ensuring the reactant gas pressure always remains above Pc at all points on the substrate's surface to ensure optimal use of reagents.
Another factor which needs consideration is the rate of reactant depletion; as the reactant gases react to deposit a layer of materials on the substrate, the reaction components become depleted. Thus, the arrival rate (and hence reactant gas pressure) reduces across the substrate's surface as a function of distance from a leading edge of the substrate being processed. Figure 9 shows a plot of reactant gas partial pressure over the substrate's surface for three different flow rates. The leading edge of the substrate 180 is on the right hand side of figure 9 and the gas mixture flows from right to left, as indicated by arrow 182. Previous CVD systems fix gas flow rates through the reaction chamber to a pre-determined setting depending on the reaction. The total pressure (that is the pressure of all the gases in the chamber, including the reaction gases and any additional gases) is adjusted by controlling the gas
delivery system or by varying the speed of the pump. This can be controlled using feedback from a pressure gauge. Thus, using an arbitrary setting for flow rate, either too much reaction gas is used (resulting in large waste yields) or not enough gas is used (resulting in an uneven deposition of material on the substrate) . Hence, optimization of the process should also account for the inter relationship pressure of gas in the chamber and the flow rate.
The flow rates in figure 9 represent three different gas injection rates and the reaction gas partial pressure is set to be equal to, or just above Pc at the exit point 184 of the substrate. The consumption rate of reactant gas for the first flow, indicated by line 186, is greater than the rate of consumption for the second 188 and third 190 flow rates. In this instance, the reactant gas in the first flow 186 is passing over the substrate at a slower rate than the second or third gas flows. Thus, the rate of gas flow through the chamber also needs to be optimised to ensure minimal wastage of reactant components . Altering the flow rate of one of the gas components, even an inert component, has an effect on the deposition characteristics associated with other gas components. The optimisation should also account for any parasitic deposition on surfaces in the chamber other than the substrate . The above description is directed to a single component deposition, for example, the deposition of silicon using silane, a single component reactant gas. It can be desired to deposit a multi component matrix to a prescribed stoichiometry, or to add a dopant with varying concentration as a function of depth in the deposited layer. Since each component of a reaction gas mixture for a multi component deposited matrix exerts its own partial pressure on the substrate, a plot similar to that shown in figure 3 exists for each component. Thus, to optimise a multi component process, each component's partial pressure needs to be accounted for in the process space optimisation.
Furthermore, if a varying concentration of deposited material with depth is required, then one or more of the reactant gases should be operating in the P zone of their
respective plots; varying the pressure changes the growth rate, and hence concentration, of the dopant in the substrate. Unintentional variations in the partial pressure of a reactant species as the gas flows across the substrate surface cause variations in deposition rate, as previously discussed. Thus, the gas flow rate of a reactant component being deposited in the P zone of the plot shown in figure 8 must be increased to an extent so that a required partial pressure uniformity is achieved across the substrate, and hence a layer of uniform thickness is deposited. This tends to reduce the reactant gas transit time over the substrate, and so the depletion of the reactant gas is reduced resulting in inefficient use of the reactant gas. An optimised process therefore requires careful control of all the gas flows to achieve the critical process pressure Pc for the T zone depositions, whilst, at the same time, the transit time of reactant gas across the substrate must be controlled so that the species being deposited in the P zone are deposited with the requisite uniformity and thickness. Therefore, in order to optimise the CVD process, it is necessary to know the dependence of growth rate on temperature and pressure, for a given deposition specie, and the corresponding gas depletion characteristic for a given apparatus (to account for any parasitic deposition process unique to the apparatus) . This data can be Obtained using the combinatorial techniques described above and can be stored electronically in the database (136 in figure 5) for , access by the controller 132.
A method of operating the CVD system described above allows for a batch production run. It is assumed that the optimum conditions required for growth have already been obtained using the combinatorial techniques described above and all the necessary data has been transferred to the lookup table . With reference to figure 10, an operator can enter at step 200 the desired growth rate for a substrate to be processed into a suitably programmed computer to act as the controller which then determines the optimum flow rate, pressure and substrate temperature 202 from data stored in
the database.
The substrate temperature is then set and controlled by the controller and temperature monitor at step 204. The heater is set so that the substrate's surface is evenly heated at the required temperature for the growth rate required for optimal processing. The controller determines the flow rate necessary to avoid depletion effects from the pre-determined data in the data base at step 206. The valves and/or variable pump parameters are set by the controller at 208, thus ensuring the correct gas mixture and flow rate/pressure in the reaction chamber.
The system is preferably run up to the determined optimum conditions using an inert gas at 210. The controller adjusts the valves (and pump) to achieve and maintain the optimum conditions whilst the inert gas is flowing through the chamber. Once the optimum conditions are met and maintained for a pre-determined period (for example 10 seconds) , the controller automatically switches the gas input from the inert gas to the process gas, and hence growth or deposition commences at step 212. Once the desired growth has been completed, the process is shut down. Processed substrates can now be removed from the reaction chamber and replaced with one or more unprocessed substrates . Thus, the same apparatus used to determine optimal process conditions is also used in mass production batch runs. This has several advantages, including improvements in device manufacture, reproducibility of devices from batch- to-batch and reduction in waste products of manufacture. The chamber conditions are continually monitored during substrate processing and the process parameter may require adjustment whilst deposition is commencing to maintain optimal chamber conditions. Preferably the controller continually monitors the chamber conditions (that is, pressure, temperature of substrate, mixture contents of the reaction gas and flow rate of the gas mixture) during the growth phase and adjusts any of these parameters, if and when necessary. For example, if the substrate temperature falls outside a tolerable range, then the controller makes
the necessary adjustment to the substrate heater to bring the temperature back within the tolerable range. Typically, at a substrate temperature of 730 C, silicon is deposited at a rate of 20 nm/min from silane reaction gas ("data from size growth kinetics and doping in reduced pressure chemical vapour deposition", JM Hartmann et al , Journal of Crystal Growth, 236 (2002) plO 20) .
The system should be calibrated for growing doped materials or alloys. The re-calibration is necessary since various compositions and doping levels may be required over a range of temperatures and growth rates. Furthermore, each alloy or doping material will have deposition or growth parameters which are invariably different to other CVD reactions . The controller is preferably automated by a suitable computer programme and can utilise a display and known input means to allow human interfacing with the controller. The programme can be coded to carry out the pertinent steps shown in figure 10 and can be stored electronically, on a floppy disk, for example. Safeguards which prevent adverse feedback between the controller and any of the chamber device which it controls should be incorporated in the computer programme. In this way, damage to the CVD system, or the substrate being processed may be reduced in the event of an accident, or system runaway.
A further combinatorial method which can be exploited with the present invention is to determine the rate of consumption of the different reactant gases while varying the temperature of the substrate. This can be done by first adjusting the heater to produce a uniform substrate temperature profile. Then the exhaust gas stream can be monitored with a mass spectrometer (or the like analysis tool) to determine the concentration of the precursors and the reaction by-products as a function of time as the substrate temperature is adjusted. The rate of change of the reactant partial pressures as a function of time can be used to map out the temperature dependency of each reaction.
Depending upon the design of the reaction chamber and the probability of parasitic reactions taking place on
surfaces other than the substrate, it may be necessary to conduct an experiment without any substrate present to provide a baseline characteristic for the reaction vessel itself. This can then be subtracted from the data obtained with the substrate (s) present to yield a data set which is characteristic of the reactions on only the substrate (s) . Depending upon the design of the reaction vessel and gas sampling method, it may be possible to monitor only the gas that has flowed across the substrate surface and thereby suppress from the data any characteristics of parasitic reactions taking place elsewhere in the reaction vessel.
The required settings for temperature, pressure, gas flow rate and gas ratios in multi-component depositions can be programmed into the apparatus' control system to allow fully optimised uniform deposition for semiconductor production.
The embodiments of the present invention are described above with reference to silane deposition processes. Of course, the invention equally applies to deposition using other compounds, such as SiF , SiH2Cl2 or SiCl , or alloys or dopants for semiconductor industrial applications, for example. Likewise, the representations of our studies only relate to silane decomposition; other deposition processes are, of course, most likely to have different growth characteristics. The invention has been described with reference to a deposition chamber suitable for processing a single substrate. Of course, the invention can also be used in chambers which accommodate and process more than one substrate .