Patent Application
20060009127
1. Field of the Invention
The present invention relates to a method for estimating and controlling a polishing profile or polishing amount in the polishing process of flatly polishing a surface of an interconnect material or an insulating film formed on a polishing object, such as a wafer, in the manufacturing of a semiconductor device, and a polishing method and a polishing apparatus which employ the above method in carrying out polishing. The present invention also relates to a program for controlling a polishing apparatus, and a storage medium in which the program and data have been stored.
2. Description of the Related Art
In the CMP process of flatly polishing a surface of an interconnect material or an insulating film laminated on a substrate in the manufacturing of a semiconductor device, the polishing conditions employed in the operation of the manufacturing line are previously optimized, and successive polishing operations of substrates are carried out repeatedly under the same optimized polishing conditions until the wear of a polishing member reaches its limit. However, in the course of wear of the polishing member, the surface topology of the interconnect material or insulating film on the substrate after polishing, herein referred to as polishing profile, changes with time in accordance with the degree of wear of the polishing member. In general, a change of polishing member is set at a time before the change in polishing profile with time begins to affect the device performance.
Semiconductor devices are becoming finer these days, and the processing speeds of devices are becoming higher by multi-level lamination of interconnects. With such semiconductor devices, the surface topology of an interconnect metal or an insulating film after polishing, i.e., the polishing profile, is required to be made flat with higher accuracy. Thus, an acceptable change in polishing profile with time is narrower for devices with finer and advanced multi-level interconnects. This necessitates more frequent changes of worn polishing members. However, consumable members for use in CMP are generally very costly, and therefore an increase in the frequency of change of consumable members significant affects the device cost.
A method is known conventionally which comprises measuring a thickness of a film on a wafer before and after polishing in a CMP process and, based on the results of the measurement, setting polishing conditions for the next wafer to be polished (see, for example, Published Japanese Translation of PCT international Publication No. 2001-501545). According to this technique, a polishing coefficient, indicating a polishing rate per unit surface pressure, is determined as an average value without a distribution on a wafer based on the results of measurement, and such polishing time and polishing pressure for the next wafer are set that will provide a desired average polishing amount. This is because the polishing coefficient changes with the condition of polishing (including the wear of consumable member, the condition of slurry, temperature, etc.), and therefore it is necessary to update the polishing coefficient and thus the polishing time and polishing pressure as needed by using the results of measurement. However, techniques for detecting the end point of polishing are fully developed nowadays, and it is now possible to automatically terminate polishing when a desired film thickness has been reached despite a change in the state of polishing. Accordingly, it is not necessary now to employ the above-described technique.
Further, since this conventional technique merely updates the polishing time and polishing pressure so that a desired average polishing amount can be obtained, it is not possible to correct a change in the polishing profile with time due to the wear of polishing member.
Another known technique involves monitoring and calculating a thickness of a remaining film thickness during polishing in a CMP process, and changing each of the pressures of pressure chambers so as to enhance the flatness of the remaining film, thereby correcting a change in the polishing profile with time due to a change with time in the slurry or polishing pad used (see, for example, Japanese Patent Laid-Open Publication No. 2001-60572). This technique is intended to be applied to a wafer polishing process in which a thickness of a film is measured with an optical sensor. The number of measurement points is inevitably limited by the spot size of the optical sensor and the rotational speed of a polishing table. This technique thus has the problem that sufficient information cannot be obtained for setting the chamber pressures that are to be changed to flatten the remaining film after polishing. Further, when this technique is applied to a wafer polishing process employing a high polishing rate, there is a case in which the response time from the measurement of the thickness of a remaining film till the feedback of a corrected value is longer than the time until the termination of polishing. Thus, the polishing can be terminated before the control achieves flattening of the remaining film.
The present invention has been made in view of the above situation in the related art. It is therefore an object of the present invention to provide a polishing method which, in the polishing process of flatly polishing a surface of an interconnect material or an insulating film laminated on a substrate in the manufacturing of a semiconductor device, can automatically reset the polishing conditions according to the state of a polishing member based on data on the polishing profile changing with time, thereby extending the life of the polishing member and obtaining flatness of the polished surface with higher accuracy, and to provide an apparatus adapted to carry out the polishing method.
In order to achieve the above object, the present invention provides a polishing apparatus comprising a top ring for holding and rotating a polishing object, such as a wafer, and pressing the polishing object against a polishing member to polish the polishing object. The top ring includes a plurality of concentrically-divided pressing portions, and is designed to be capable of independently setting a pressure for each pressing portion, whereby the pressure between the polishing object and the polishing member can be controlled. When the polishing profile of a polishing object is not flat, it is possible, for example, to apply such an additional pressure to a portion deficient in polishing amount as to compensate for the deficient amount, thus providing a flat polished surface with high accuracy.
The pressure of each processing portion of the top ring is generally set so that the polished surface of an interconnect metal or an interlevel insulating film formed on a polishing object becomes flat. The pressure setting, in many cases, has conventionally been practiced according to an engineer's empirical rule. With such an empirical rule, it is usually necessary to previously polish several polishing objects for adjustment in order to establish conditions for planarized surface of the polishing object.
The present invention employs a first simulation software which estimates and calculates the polishing profile of a polishing object through input of pressure setting conditions for each pressing portion of the above-described top ring. It has been found that the results of simulation with the first simulation software only produce a 1-5% error with respect to the actual polishing profile. The present invention can avoid waste of polishing objects, which is necessary for adjustment of pressure setting in the conventional method, and can estimate a polishing profile in a very short time by using the simulation software, thus shortening time for adjustment of pressure setting.
According to the first simulation software, by merely updating a polishing coefficient (coefficient involving the influence of polishing pad, slurry, etc.) which can be determined from the results of measurement of the thickness of a remaining film (or polishing profile) at a relatively small number of measurement points, it is possible to estimate the thickness of the remaining film after polishing at its numerous points other than the measurement points. This makes it possible to easily correct the influence of changes in a slurry and a polishing member, such as a polishing pad, and to estimate the polishing profile to be obtained under the corrected reset polishing conditions. In the case where the updating of polishing coefficient is made by using the results of polishing carried out under polishing conditions close to the polishing conditions set in the first simulation software, the error can be made as low as about 1 to 3%. In a practical semiconductor device manufacturing line in which polishing objects (wafers) are polished successively, there is no significant difference in the set values of polishing conditions between successive polishing objects, enabling a high-accuracy simulation. When the number of measurement points for the measurement of polishing profile is relatively small, it is desirable to utilize a curve interpolating the measured values to determine a polishing coefficient.
The present invention obtains a desired polishing profile by making the remaining film on a wafer into one having a desired thickness. For this purpose, according to the present invention, desired set pressures of the respective pressing portions of the top ring are calculated with a second simulation software by inputting desired polishing time, average polishing amount and configuration of remaining film (or polishing profile) so as to satisfy these conditions. The second simulation software incorporates the first simulation software as a module. An estimated polishing profile at a set pressure is calculated with the first simulation software and the estimated profile is compared with a desired polishing profile. Based on the comparison, a corrected set pressure is calculated. By repeating the calculation of estimated polishing profile and the calculation of corrected set pressure with the second simulation software, it is possible to calculate a desired set pressure that provides a polishing profile approximating to the desired polishing profile.
In practical, a set polishing time may be used as a reference value (target value), and polishing may be terminated when the actual amount of a remaining film being monitored has reached a desired value (end point detection manner).
Unlike the conventional technique that stabilizes an average polishing amount, the present invention can also control and stabilize the surface flatness after polishing or the thickness of remaining film. For this purpose, according to the present invention, after processing preferably one test polishing object and updating the polishing coefficient, optimized polishing conditions for providing desired polishing time, average polishing amount and thickness of remaining film, are obtained using the second simulation software. A polishing object is polished under the optimized polishing conditions. The polishing coefficient is updated as needed according to the wear of a polishing member, and the polishing conditions are re-optimized to stably provide the desired polishing time, average polishing amount and configuration of remaining film.
By feeding back the polishing conditions of a polished polishing object in carrying out polishing, it becomes possible to ensure the quality of a polished polishing object with higher accuracy, taking account of accuracy of the flatness of a remaining film after polishing and accuracy of the feedback control which is influenced by the polishing conditions. When a failure occurs in the polishing apparatus, or a polishing member (consumable member) wears out and reaches its use limit, a desired polishing profile may not be obtained even if the polishing conditions are adjusted. In such cases, according to the present invention, the operation of the polishing apparatus can be stopped or a warning can be issued based on the polishing conditions calculated with the second simulation software. This can increase the product yield and extend the life of a polishing member to its use limit.
It is possible with the present invention to obtain the data of polishing profile not only for a film measurable with an optical measuring device, but also for a metal film by using a metal film-measurable device and perform feedback control. The present invention is thus highly versatile with no limitation on its application to polishing processes. Furthermore, data on film thickness can be obtained by any suitable method, such as a method of measuring a film thickness with a measuring device capable of monitoring it during polishing, a method of transporting a wafer to a measuring device for measurement after polishing, or a method of measuring a film thickness outside the polishing apparatus and transferring and inputting the film thickness data to the polishing apparatus. It is also possible employ a combination of these methods. For example, data on film thickness before and after polishing may be obtained by different methods to facilitate the operation.
In addition, by reading a program for executing the simulation tool of the present invention from a computer-readable storage medium into a computer for controlling the polishing apparatus, it becomes possible to expand the function of the conventional polishing apparatus.
A polishing method and a polishing apparatus (CMP apparatus) according to embodiments of the present invention will be described below with reference to drawings. First, a polishing apparatus according to an embodiment of the present invention will be described using
As shown in
According to the polishing apparatus shown in
After completion of wafer polishing process, the shutters are closed to separate the SMIF or FOUP from the polishing apparatus, and the SMIF or FOUP is transferred automatically or manually to another processing process. It is therefore necessary to keep the internal atmosphere of the SMIF or FOUP clean. For that purpose, there is a down flow of clean air through a chemical filter in the upper space of an area C, which a wafer passes through right before returning to the wafer cassette 1. Further, since the linear motor is employed for traveling of the transfer robot 4, scattering of dust can be reduced and the atmosphere in the area C can be kept clean. In order to keep wafer in the wafer cassette 1 clean, it is possible to use a clean box that may be a closed vessel, such as a SMIF or FOUP, having a built-in chemical filter and a fan, and can maintain its cleanness by itself.
Two cleaning apparatuses 5, 6 are disposed at the opposite side of the wafer cassettes 1 with respect to the travel mechanism 3 of the transfer robot 4. The cleaning apparatuses 5, 6 are disposed at positions that can be accessed by the hands of the transfer robot 4. Between the two cleaning apparatuses 5, 6 and at a position that can be accessed by the transfer robot 4, there is provided a wafer station 50 having four wafer supports 7, 8, 9 and 10.
An area D, in which the cleaning apparatuses 5, 6 and the wafer station 50 having the wafer supports 7, 8, 9 and 10 are disposed, and an area C, in which the wafer cassettes 1 and the transfer robot 4 are disposed, are partitioned by a partition wall 14 so that the cleanliness of the area D and the area C can be separated. The partition wall 14 has an opening for allowing semiconductor wafers to pass therethrough, and a shutter 11 is provided at the opening of the partition wall 14. A transfer robot 20 is disposed at a position where the transfer robot 20 can access the cleaning apparatus 5 and the three wafer supports 7, 9 and 10, and a transfer robot 21 is disposed at a position where the transfer robot 21 can access the cleaning apparatus 6 and the three wafer supports 8, 9 and 10.
A cleaning apparatus 22 is disposed at a position adjacent to the cleaning apparatus 5 and accessible by the hands of the transfer robot 20, and another cleaning apparatus 23 is disposed at a position adjacent to the cleaning apparatus 6 and accessible by the hands of the transfer robot 21. Each of the cleaning apparatuses 22, 23 is capable of cleaning both surfaces of a semiconductor wafer. All the cleaning apparatuses 5, 6, 22 and 23, the wafer supports 7, 8, 9 and 10 of the wafer station 50, and the transfer robots 20, 21 are placed in the area D. The pressure in the area D is adjusted so as to be lower than the pressure in the area C.
The polishing apparatus shown in
The polishing tables 34, 35 include, besides the mechanical dressers 38, 39, atomizers 44, 45 as fluid-pressure dressers. An atomizer is designed to jet a mixed fluid of a liquid (e.g. pure water) and a gas (e.g. nitrogen) in the form of a mist from a plurality of nozzles to the polishing surface. The main purpose of the atomizer is to rinse away polished scrapings and slurry particles deposited on and clogging the polishing surface. Cleaning of the polishing surface by the fluid pressure of the atomizer and setting of the polishing surface by the mechanical contact of the dresser can effect a more desirable dressing, i.e. regeneration of the polishing surface.
The dresser 48 has a rectangular body longer than the diameter of the polishing 36. The dresser head 97 is swingable about the swing shaft 98. A dresser fixing mechanism 96 is provided at the free end of the dresser head 97 to support the dresser 48. The dresser fixing mechanism 96 and the dresser 48 make a pivot motion to cause the dresser 48 to move like a wiper for wiping a windowshield of a car on the polishing table 36 without rotating the dresser 48 about its own axis. The polishing tables 36, 37 may comprise the scroll-type table.
Returning to
The reversing devices 28, 28′ have a chuck mechanism for chucking a semiconductor wafer, a reversing mechanism for reversing a semiconductor wafer, and a semiconductor wafer detecting sensor for detecting whether the chuck mechanism chucks a semiconductor wafer or not, respectively. The transfer robot 20 transfers a semiconductor wafer to the reversing device 28, and the transfer robot 21 transfers a semiconductor wafer to the reversing device 28′.
In the area A constituting one of the polishing chambers, there is provided a linear transporter 27A constituting a transport mechanism for transporting the semiconductor wafer between the reversing device 28 and the top ring 32. In the area B constituting the other of the polishing chambers, there is provided a linear transporter 27B constituting a transport mechanism for transporting the semiconductor wafer between the reversing device 28′ and the top ring 33. Each of the linear transporters 27A, 27B comprises two stages linearly movable in a reciprocating fashion. The semiconductor wafer is transferred between the linear transporter and the top ring or the linear transporter and the reversing device via the wafer tray.
On the right side of
The semiconductor wafer 101 transferred to the top ring 32 is held under vacuum by a vacuum attraction mechanism of the top ring 32, and transported to the polishing table 34. Thereafter, the semiconductor wafer 101 is polished by a polishing surface composed of a polishing pad or a grinding stone or the like attached on the polishing table 34. The first polishing table 34 and the second polishing table 36 are disposed at positions that can be accessed by the top ring 32. With this arrangement, a primary polishing of the semiconductor wafer can be conducted by the first polishing table 34, and then a secondary polishing of the semiconductor wafer can be conducted by the second polishing table 36. Alternatively, the primary polishing of the semiconductor wafer can be conducted by the second polishing table 36, and then the secondary polishing of the semiconductor wafer can be conducted by the first polishing table 34.
The semiconductor wafer 101, which has been polished, is returned to the reversing device 28 in the reverse route to the above. The semiconductor wafer 101 returned to the reversing device 28 is rinsed by pure water or chemicals for cleaning supplied from rinsing nozzles. Further, the wafer holding surface of the top ring 32, from which the semiconductor wafer has been removed, is also cleaned by pure water or chemicals supplied from cleaning nozzles.
Next, processes conducted in the polishing apparatus shown in
The other semiconductor wafer is processed in the following route: the wafer cassette (CS2)→the transfer robot 4→the wafer support 8 of the wafer station 50→the transfer robot 21→the reversing device 28′→the wafer stage 901 for loading in the linear transporter 27B→the top ring 33→the polishing table 35→the polishing table 37 (as necessary)→the wafer stage 902 for unloading in the linear transporter 27B→the reversing device 28′→the transfer robot 21→the cleaning apparatus 23→the transfer robot 21→the cleaning apparatus 6→the transfer robot 4→the wafer cassette (CS2).
In two cassette parallel processing in which three-stage cleaning is perfor5med, one semiconductor wafer is processed in the following route: the wafer cassette (CS1)→the transfer robot 4→the wafer support 7 of the wafer station 50→the transfer robot 20→the reversing device 28→the wafer stage 901 for loading in the linear transporter 27A→the top ring 32→the polishing table 34→the polishing table 36 (as necessary)→the wafer stage 902 for unloading in the linear transporter 27A→the reversing device 28→the transfer robot 20→the cleaning apparatus 22→the transfer robot 20→the wafer support 10 of the wafer station 50→the transfer robot 21→the cleaning apparatus 6→the transfer robot 21→the wafer support 9 of the wafer station 50→the transfer robot 20→the cleaning apparatus 5→the transfer robot 4→the wafer cassette (CS1).
The other semiconductor wafer is processed in the following route: the wafer cassette (CS2)→the transfer robot 4→the wafer support 8 of the wafer station 50→the transfer robot 4→the reversing device 28′→the wafer stage 901 for loading in the linear transporter 27B→the top ring 33→the polishing table 35→the polishing table 37 (as necessary)→the wafer stage 902 for unloading in the linear transporter 27B→the reversing device 28′→the transfer robot 21→the cleaning apparatus 23→the transfer robot 21→the cleaning apparatus 6→the transfer robot 21→the wafer support 9 of the wafer station 50→the transfer robot 20→the cleaning apparatus 5→the transfer robot 4→the wafer cassette (CS2).
In serial processing in which three-stage cleaning is performed, the semiconductor wafer is processed in the following route: the wafer cassette (CS1)→the transfer robot 4→the wafer support 7 of the wafer station 50→the transfer robot 20→the reversing device 28→the wafer stage 901 for loading in the linear transporter 27A→the top ring 32→the polishing table 34→the polishing table 36 (as necessary)→the wafer stage 902 for unloading in the linear transporter 27A→the reversing device 28→the transfer robot 20→the cleaning apparatus 22→the transfer robot 20→the wafer support 10 of the wafer station 50→the transfer robot 21→the reversing device 28′→the wafer stage 901 for loading in the linear transporter 27B→the top ring 33→the polishing table 35→the polishing table 37 (as necessary)→the wafer stage 902 for unloading in the linear transporter 27B→the reversing device 28′→the transfer robot 21→the cleaning apparatus 23→the transfer robot 21→the cleaning apparatus 6→the transfer robot 21→the wafer support 9 of the wafer station 50→the transfer robot 20→the cleaning apparatus 5→the transfer robot 4→the wafer cassette (CS1)
According to the polishing apparatus shown in
A plurality of wafer trays are assigned to loading wafer tray for holding a polishing object to be polished and unloading wafer tray for holding a polishing object which has been polished. Therefore, the polishing object to be polished is transferred not from the pusher but from the loading wafer tray to the top ring, and the polished polishing object is transferred from the top ring not to the pusher but to the unloading wafer tray. Thus, the loading of the polishing object to the top ring, and the unloading of the polishing object from the top ring are conducted by respective jigs (or components), i.e. the wafer tray, and hence the abrasive liquid or the like attached to the polished polishing object is prevented from being attached to a common support member for performing loading and unloading the polishing object. As a result, the solidified abrasive liquid or the like is not attached to the polishing object to be polished, and does not cause damage to the polishing object to be polished.
An inline monitor IM is provided in the appropriate place in the area C of the above-described polishing apparatus. The wafer after polishing and cleaning is transferred to the inline monitor IM by the transfer robot 4, where a film thickness or a polishing profile of the wafer is measured. The inline monitor IM is actually disposed above the transfer robot 4. The motion of the whole polishing apparatus is controlled by a control unit CU. The control unit CU is provided with a connector to be connected to a storage medium reader for reading a control program and data from an external storage medium by connecting the storage medium reader to the control unit CU as necessary. The control unit CU may be provided in the polishing apparatus, as shown in
As is known from Preston's equation, the polishing amount of a wafer is approximately proportional to the pressure of the surface of the wafer on a polishing pad. In order to determine the pressure, however, it is necessary to perform modeling of a top ring having a complicated structure and take account of the non-linearity of a polishing pad which is an elastic material, the large deformation of a wafer which is a thin plate, and the stress concentration which is especially marked at the edge of a wafer. It is therefore difficult to obtain an analytic solution of a distribution of the pressure of the surface of the wafer in mathematically. On the other hand, the use of a finite element method or a boundary element method for determining the pressure involves dividing these objects into a large number of elements, leading to a vast amount of calculation. This necessitates a lot of computation time and a high computational capacity. Moreover, to obtain appropriate results, it is necessary for the operator to have the expert knowledge of numerical analysis. It is therefore virtually impossible from a practical viewpoint and also in view of the cost to use such a numerical analysis method as a reference in carrying out a simple adjustment in the work site or to use the method by incorporating it into the polishing apparatus.
In the case where a profile control-type top ring is employed in the polishing apparatus of the above-described construction, this problem becomes more complex. The “profile control-type top ring” is a generic term for top rings having a plurality of pressing portions. Examples of such top rings include a top ring having a plurality of pressing portions comprised of air bags or water bags partitioned concentrically with membranes, a top ring having a plurality of pressing portions, comprised of partitioned air chambers, for directly pressing on the back surface of a wafer with air pressure by independently pressurizing the respective air chambers, a top ring having pressing portions that press on a wafer by springs, and a top ring having localized pressing portions including one or more piezoelectric devices. A top ring having a combination of such pressing portions can also be used. As interactions of these pressing portions are added to the above problem, it is not easy to determine the pressure of the surface of the wafer. Then, according to the present invention, a distribution of the pressure of the surface of the wafer is determined using a first simulation as described below. The following description illustrates a top ring having a plurality of concentrically-partitioned air bags as pressing portions.
Thus, as shown in
The number of the air bags of the top ring T is not limited to 4, but may be increased or decreased according to the size of the wafer. Though not shown in
According to the present invention, a set of a distribution of the pressure of the front surface of the wafer W corresponding to a combination of pressures applied by the air bags E1 to E4 and the retainer ring E5 on the back surface of the wafer W and on the surface of the polishing pad around the wafer W, is calculated and stored in advance in a memory of the above-described control unit CU of the polishing apparatus. Assuming that the distribution of the pressure of the front surface of the wafer W can be regarded as substantially linear (i.e. the superposition principle substantially holds true) if, in a polishing process, the practical pressure setting range for the pressures of the air bags on the back surface of the wafer and for the pressure of the retainer ring on the polishing pad are 100 to 500 hPa and the air pressure is within the range of ±200 hPa, the distribution of the pressure of the front surface of the wafer W, corresponding to any of intended pressures of the air bags on the corresponding areas of the back surface of the wafer, can be determined within the back surface pressure setting range of ±200 hPa by synthesizing the distribution of the pressure of the front surface of the wafer, corresponding to a combinations of three back surface pressures, 100 hPa, 300 hPa and 500 hPa.
A description will now be given of a method of synthesizing the pressure of the front surface of a wafer W from pressures applied from the air bags E1 to E4 on the wafer W and from the retainer ring E5 on a polishing pad (hereinafter referred to as back surface pressures), in the case where the top ring T is designed to be capable of controlling the five pressures, i.e. the pressures of the four air bags E1 to E4 on the wafer W and the pressure of the retainer ring E5 on the surface (polishing surface) of the polishing pad around the wafer W, by referring to
First, data on a distribution of the pressure of the wafer front surface on the polishing member (polishing pad) is obtained and stored in advance. In the case of the above-described five regions and three pressures, the number of combinations of the back surface pressures is total 35=243. Of these combinations, 27 combinations are selected as necessary combinations for synthesizing the distribution of the pressure of the wafer front surface. Assuming that pressures Z1, Z2, Z3, Z4 and Z5 (unit: hPa), respectively denoting the pressures of the air bags E1 to E4 on the wafer and the pressure of the retainer ring E5 on the surface of the polishing pad around the wafer, can each take either one of the values 100, 300 and 500, the 27 combinations of the Z1-Z5 values, which are to be stored in a memory of the control unit CU, are as follows:
Z1−Z5=100 (1)
Z1−Z5=300 (2)
Z1−Z5=500 (3)
Z1=100, Z2−Z5=300 (4)
Z1=100, Z2−Z5=500 (5)
Z1=300, Z2−Z5=100 (6)
Z1=300, Z2−Z5=500 (7)
Z1=500, Z2−Z5=100 (8)
Z1=500, Z2−Z5=300 (9)
Z1=Z2=100, Z3−Z5=300 (10)
Z1=Z2=100, Z3−Z5=500 (11)
Z1=Z2=Z3=Z4=500, Z5=300 (27)
The distributions of the pressure of the front surface of the wafer, corresponding to the above 27 combinations of the set pressures on the wafer back surface, can be calculated in advance using, for example, a finite element method. The calculated distributions of the pressure of the front surface of the wafer and the 27 combinations of back surface pressures correspond to the calculated pressures, are stored in a memory of the control unit CU. The combinations of the set pressures and the corresponding distributions of the pressure of the wafer front surface may be stored in the memory of the control unit CU by reading the information from a storage medium with a storage medium reader connected to the control unit CU, or by storing the information in advance in a ROM set in the control unit CU and reading the information out of the ROM.
Various distributions of the pressure of the wafer front surface corresponding to various changes in the wafer back surface pressures are then synthesized by using the 27 combinations stored in the memory. To give a specific example, in the case of applying the following pressures: 150 hPa by the air bag E1; 200 hPa by the air bag E2; 150 hPa by each of the air bags E3 and E4; and 250 hPa by the retainer ring E5, i.e., in the case where the set pressures to be calculated are: Z1=150, Z2=200, Z3=Z4=150 and Z5=250, the intended set pressures can be expressed in the vector form: Zp=(150 200 150 150 250]T, wherein the symbol T represents transpose of matrix. Thus, similarly, the above 27 combinations of pressures can also be exposed by vector form. For example, the combination of pressures of the above item (4) can be expressed by the vector Zc2=[100 300 300 300 300]T. The suffix (e.g. C2) is a serial number indicative of conditions.
In determining the distribution of the pressure of the wafer front surface, corresponding to the intended set pressure vector Zp, 5 combinations are selected from the above 27 combinations of the back surface pressures applied by the air bags so as to respond to changes in the set pressures of adjacent areas. For example, the following 5 combinations expressed by the vectors are selected in order to realize the above-described set pressure application conditions of Z1=150, Z2=200, Z3=Z4=150 and Z5=250:
Zc1=[100 100 100 100 100]T
Zc2=[100 300 300 300 300]T
Zc3=[300 300 100 100 100]T
Zc4=[100 100 100 100 100]T
Zc5=[100 100 100 100 300]T
Using these vectors, the set pressure vector Zp can be expressed as follows:
Zp=f1×Zc1+f2×Zc2+f3×Zc3+f4×Zc4+f5×Zc5 (1)
Zp=[150 200 150 150 250]T
In the equation (1), f1 to f5 are constants. The following 5 equations with f1 to f5 unknown can be obtained from the above equation (1):
150=f1·100+f2·100+f3·300+f4·100+f5·100
200=f1·100+f2·300+f3·300+f4·100+f5·100
150=f1·100+f2·300+f3·100+f4·100+f5·100
150=f1·100+f2·300+f3·100+f4·100+f5·100
250=f1·100+f2·300+f3·100+f4·100+f5·300
From these equations, f1 to f5 can be determined. Since f3 is equal to f4 (f3=f4) in the equations, the number of equations and the number of unknowns are both four.
In other words, when using a matrix with the 5 vectors as its elements, i.e. Mc=[Zc1 Zc2 Zc3 Zc4 Zc5], the relationship between the intended set pressure vector Zp and the coefficient vector f=[f1 f2 f3 f4 f5]T can be expressed as follows:
Zp=Mc·f (2)
The equation (2) indicates that the set pressure vector Zp, to be calculated, can be expressed as a linear combination of the vectors of the combinations of set pressures stored in the memory of the control unit CU. From the equation (2), the coefficient vector f can be determined by the following equation:
f=Mc−1·Zp
There is a case in which the matrix Mc includes a row or column that is not linearly independent, causing inconvenience for determining the inverse matrix Mc−1. In such a case, the matrix can be converted into an inverse matrix-determinable form by appropriate replacement or addition and subtraction of the row or column. Such arithmetic processing is an ordinary mathematic processing and does not need any special measures to be taken.
After the coefficients f1 to f5 are thus determined, the pressure distribution Pc of the wafer front surface, corresponding to the intended set pressure Zp, can be obtained by multiplying the date on the distributions of the pressure of the wafer front surface(Pc1 to Pc5), corresponding to the pre-selected combinations of pressures on the wafer back surface (i.e. the five combinations Zc1 to Zc5), by the respective coefficients f1 to f5 and then adding the all terms together, as follows:
Pc=f1·Pc1+f2·Pc2 . . .
In the manner as described above, the distribution of the pressure of the wafer front surface, corresponding to intended set pressures on the wafer back surface, can be determined, without a complicated calculation as by a finite element method, by adopting set pressures on the wafer back surface in such a pressure range that a change in the pressure of the wafer front surface can be regarded as being substantially linear (i.e. the superposition principle holds true), preparing data on pre-calculated distributions of the pressure of the wafer front surface in a number of cases (27 cases in the above example) and appropriately selecting some cases from them and synthesizing the selected data.
The distribution of the pressure of the wafer front surface can thus be determined in accordance with the procedures described above. A simulation tool for obtaining the pressure distribution of the wafer front surface, corresponding to the set pressures on the wafer back surface, can be produced by thus storing the procedures in a computer.
It is also possible to determine the coefficient matrix by a method comprising calculating in advance all the combinations of 5 areas and 3 pressures, i.e. 35=243 combinations, formulating the equation Zp=MCall·fall using the matrix MCall=[Zc1 Zc2 . . . Zc242 Zc243] including the all combinations and the coefficient vector fall=[f1 f2 . . . f242 f243] representing 243 coefficients, and determining the coefficient vector by fall=MCall−1·.Zp using the pseudo inverse matrix of MCall. Thus, there is no particular limitation on methods for determining an appropriate coefficient. Since superposition in a pressure range, in which a pressure change can be regarded as being linear, is utilized, any linear algebraic method can be used to determine coefficients corresponding to the coefficients f1 to f5.
The range of pressure on the wafer back surface and particular pressures adopted in the pressure range, which are to be calculated in advance, are not limited to the range of 100 to 500 hPa and the three pressures 100, 300 and 500 hPa described above. For example, the five pressures (100, 200, 300, 400 and 500 hPa) may be adopted only for the areas corresponding to the air bag E4 and the retainer ring E5.
After the distribution of the pressure of the wafer front surface is thus determined, an estimated polishing profile of the wafer can be determined by multiplying the pressure distribution and the data on the distribution of polishing coefficients on the wafer front surface, previously determined for the wafer to be polished. As is known from the Preston's empirical equation, the polishing amount Q of a wafer is approximately proportional to the product of the pressure P of the wafer front surface, the relative speed v of contact surface and the polishing time t:
Q=k·P·v·t
wherein k is a proportionality constant as determined by the material of the polishing pad, the material to be polished, the type of the slurry used in polishing, etc.
The relative speed v of contact surface on the wafer front surface (i.e. the relative velocity between the wafer front surface and polishing pad) differs at various points on the wafer front surface, and the polishing time t differs depending on the polishing conditions. Taking polishing coefficient as polishing rate per unit pressure, the polishing coefficient corresponds to Kv in the Preston's equation. By determining the distribution of Kv values on the wafer front surface in advance, an estimated polishing amount Qest on the wafer front surface can be determined by the following equation:
Qest=Kv·Pc
Further, an estimated polishing amount per unit time, i.e., estimated polishing rate QestΔt can be determined by the following equation:
QestΔt=Qest/t
Since the estimated polishing amount (estimated polishing rate) of a wafer can be determined by such a simple calculation, the results of calculation with the simulation tool can be used as a reference in carrying out a simple adjustment in the work site, or the simulation tool can be incorporated into the polishing apparatus (CMP apparatus).
The data on the distribution of polishing coefficients on the wafer front surface can be given in an arbitrary manner. According to the simplest method, the polishing rate can be given as a value which is proportional to the distance r between the center of the wafer and any point on the wafer if a difference Δω in rotating velocity between a polishing pad and the wafer is constant, since the relative speed v is approximately proportional to the distance r and to the difference Δω.
First, in step 1, the surface topology of a film on a wafer is measured in advance. Next, in step 2, the wafer is actually polished under particular set pressure and polishing time conditions. In step 3, the distribution of the pressure of the wafer front surface under the set pressure conditions is calculated in advance using the simulation tool. The surface topology of the polished film on the wafer is re-measured and, from the difference before and after polishing, the distribution of the polishing amount on the wafer front surface is calculated (step 4). Next, in step 5, the calculated distribution of the polishing amount is divided by the polishing time and the calculated pressure distribution to determine the distribution of the polishing rates per unit pressure and unit time at various points on the wafer front surface, i.e. the distribution of polishing coefficients on the wafer front surface. It is also possible to divide the calculated distribution of the polishing amount only by the calculated pressure distribution without division by the polishing time, thus determining the distribution of the polishing rates per unit pressure.
It is also possible to pre-calculate the distribution of polishing coefficients for a polishing pad at the time of its initial use, after its use to a certain degree and near its use limit, and to store the data on the change in polishing coefficient with time in the control unit CU.
It has been confirmed experimentally that the results of estimation of the polishing amount or polishing rate of a wafer by the above-described method for the profile control-type top ring are approximately equal to the results of actual polishing of the wafer. In some cases, the polishing profile in a peripheral annular region of a wafer, the region having a width of about 10 mm from the peripheral end, differs slightly from the pressure distribution profile of the wafer front surface. This is because the annular region of the wafer is influenced, during polishing, by a reaction force due to deformation of a polishing pad, which is an elastic body, and by a peripheral bevel portion of the wafer, in addition to the influence of the pressure applied from the wafer back surface. However, such other influences than the pressure distribution can also be modeled by determining the polishing coefficient from the pressure distribution and the actual polishing profile. This makes it possible to estimate and calculate the polishing profile of the entire front surface of the wafer with high accuracy.
In the case where it has been confirmed that the polishing profile of a peripheral region of the wafer front surface has a particular relationship with a physical factor different from the pressure distribution, it is possible to combine the above-described estimation method with a method for estimating the polishing profile of the peripheral region of the wafer using the particular relationship. Assuming, for example, that a difference between the pressure E5p of the retainer ring E5 and the pressure E4p of the air bag E4 located on the outermost peripheral region of the wafer back surface, in association with the flow conditions of slurry, affects the polishing coefficient of the outermost 10 mm-width region of the wafer. In this case, it is difficult only with the polishing coefficient calculated from the pressure distribution of the wafer front surface and from particular polishing conditions to estimate with high accuracy the polishing profile with a large change in the pressures E4p and E5p. However, in case it has been confirmed that the flow of slurry changes in proportion to a relative change in pressures of E4p and E5p, for example, (E4p-E5p)/|E4p|, the polishing coefficient of the outermost region of the wafer can be corrected by multiplying the polishing coefficient by an appropriate correction coefficient which is:
1+m·(E4p−E5p)/|E4p|
wherein m is an appropriate proportionality constant.
In particular, the appropriate proportionality constant m is determined by comparing a polishing coefficient calculated from the results of polishing carried out under particular conditions with a polishing coefficient calculated from the results of polishing carried out by changing only the pressure of the retainer ring E5. The polishing profile of the peripheral region of the wafer is estimated by using the proportionality constant m thus determined. By thus correcting the polishing coefficient using a physical factor not associated with the surface pressure, such as the flow of slurry, temperature distribution, the concentration distribution of slurry, etc., the polishing profile can be estimated more accurately.
A wafer has, near its peripheral bevel portion, a region which has a relatively poor flatness compared to the wafer central region and whose shape is deviated from an ideal shape. For example, a roll-off can be formed in the outermost region of a wafer having a surface oxide film due to roll-off of the bare wafer. The term “roll-off” herein refers to a shape deviated from an ideal configuration of wafer edge region. The degree of roll-off can be defined as ROA which is a measured deviation from a reference plane at a point on the wafer front surface e.g. at 1 mm distance from the peripheral end. The roll-off and ROA of bare wafer are described in M. Kimura, Y. Saito, et al., A New Method for the Precise Measurement of Wafer Roll Off Silicon Polished Wafer, Jpn. J. Appl. Phys., Vol. 38 (1999), pp. 38-39.
Though the ROA of a bare wafer is at most about 1 μm and the degree of roll-off of the oxide film is also at the same level, the roll-off affects the pressure distribution in the peripheral region with a width-of about 5 mm from the peripheral end of the wafer. The ROA differs between wafers and between wafer lots, which causes variation of polishing in the peripheral regions of wafers. An edge shape (usually an ideal edge shape) modeled for a finite element method usually differs from the actual edge shape of a wafer to be polished. The polishing profile can therefore be estimated more accurately by correcting the polishing coefficient of the outermost region with ROA values measured before and during polishing. The polishing coefficient may also be corrected by using an indicator other than ROA, which can indicate the configuration or degree of roll-off.
For measurement of ROA, for example, a contactless measuring method using a laser beam may be employed. Such a method can be carried out by using, for example, an edge roll-off measuring device LER-100 manufactured by Kobelco Research Institute, Inc. Further, for measurement of roll-off configuration, a measuring method may be selected from an optical method, a stylus method, an electrical method using, for example, an eddy current sensor, a magnetic method, an electromagnetic method, and a fluidic method, and the like. A roll-off configuration measuring device may either be installed in the polishing apparatus or provided separately from the polishing apparatus. In the case of installing the roll-off device in the polishing apparatus, the roll-off device may be installed adjacent to the inline monitor IM shown in
In an edge region of a wafer having a surface metal film, the metal film in the outermost region of the wafer is removed, or the metal film is not formed in the outermost region right from the start, for example, for the purpose of preventing contamination. The configuration of the end portion of the metal film is also not flat and thus requires correction of the polishing coefficient. The correction can be made in the same manner as in the case of roll-off of oxide film.
As will be appreciated from the foregoing description, application of the present method is not limited to a profile control-type top ring using air bags. If a force acting on the wafer back surface is found, the pressure distribution of the wafer front surface can be determined and the polishing profile can be estimated therefrom. Thus, the present method can be applied to top rings having various types of pressing portions, including air bags capable of holding a pressurized gas, liquid bags capable of holding a pressurized liquid such as pure water, partitioned air chambers which are directly pressurized with a pressurized gas, pressing portions which generate pressures by elastic bodies, for example, springs, and pressing portions which press on by piezoelectric devices, and the like. Top rings having a combination of such various types of pressing portions may also be used.
According to the present invention, the top ring is designed to be capable of setting a polishing pressure independently for each of the plurality of pressing portions, i.e., the air bags E1 to E4 and the retainer ring E5 and, using the above-described simulation tool, pressures that are necessary to set for the respective pressing portions in order to obtain the intended polishing profile are calculated, and the calculated pressure values are fed back to a wafer to be polished later. With this method, even when the polishing profile changes with time due to wear of a polishing member, the change can be corrected as needed. This makes it possible to stably obtain the desired polishing profile. An example of control flow for achieving this will now be described with reference to
First, the surface topology of a wafer before polishing, i.e., the thickness distribution of an interconnect metal or an insulating film on the wafer, is measured with a film thickness measuring device, such as the inline monitor IM, and the measurement data is stored in a memory (step 1). This measurement is carried out on at least one point in the wafer each of the areas corresponding to the air bags E1 to E4 and the area corresponding to the retainer ring E5. At first, back surface pressures are set arbitrarily for the respective areas, and the set back surface pressures are stored in a memory (step 2). The wafer is then polished under polishing conditions including the set pressures (step 3).
Next, the surface topology of the wafer after polishing, i.e. the thickness distribution of the interconnect metal or the insulating film on the wafer is measured with a film thickness measuring device, such as the inline monitor IM, and the measurement data is stored in a memory (step 4). This measurement may be carried out with the inline monitor IM installed in the polishing apparatus or with a measuring device installed outside the polishing apparatus. Downloading of the measurement data may be performed either online or via a storage medium. This measurement is carried out on at least one point in the wafer each of the areas corresponding to the air bags E1 to E4 and the area corresponding to the retainer ring E5.
Based on the measurement results, polishing pressure conditions for creating the intended polishing profile are calculated by the following procedure. First, the intended polishing profile is set. This setting may be performed, for example, by designating a plurality of points, at which control of polishing amount is desired, on the wafer front surface, and setting a polishing amount QT or a polishing rate QTΔt=QT/t for each designated point. The following description illustrates the case of setting polishing amount QT. Thus, a desired polishing amount is inputted and stored in a memory, and a desired polishing amount QT corresponding to a measurement point is calculated.
Based on the measurement data stored in the memory in steps 1 and 4, a polishing amount Qpoli is calculated for each of the areas of the wafer after polishing, corresponding to the air bags E1 to E4 and the retainer ring E5 (step 5). The calculated polishing amount Qpoli for each point is divided by the polishing pressure P, set before polishing and stored in the memory in step 2, of the area including that point to calculate the polishing amount per unit surface pressure QpoliΔP=Qpoli/P (step 6).
Next, a target polishing amount QT at a point nearest to a measurement point is extracted, or a target polishing amount QT is approximated linearly from two points near a measurement point. For each point, the polishing amount difference ΔQ between the target polishing amount QT and the polishing amount QPoli, ΔQ=QT−QPoli, is determined (step 7). The polishing amount corresponding to the polishing amount difference ΔQ is divided by the polishing amount per unit surface pressure QpoliΔP calculated in step 6 to calculate a correction polishing pressure ΔP of the back surface pressure, ΔP=ΔQ/QpoliΔP (step 8).
The correction polishing pressure ΔP calculated in step 8 is added to the pressure P set before polishing in step 2 to determine a recommended polishing pressure value Pinput=P+ΔP (step 9). In the case where an area includes a plurality of measurement points, the pressure values calculated for the plurality of points are averaged, and the averaged value is taken as the recommended polishing pressure value Pinput of the area.
The recommend polishing pressure value Pinput calculated in step 9 is inputted into the simulation tool of the present invention (step 10), and a polishing amount is calculated and for each point in the above-described manner to determine an estimated polishing amount Qest. Then, the polishing amount difference ΔQ between the estimated polishing amount Qest and the target polishing amount QT, ΔQ=QT−Qest, is calculated for each point (step 11).
Decision is made as to whether the polishing amount difference ΔQ between the estimated polishing amount Qest and the target polishing amount QT, calculated for each point in step 11, is within the allowable range (step 12). If the polishing amount difference ΔQ is within the allowable range, the recommended polishing pressure value Pinput is stored in a memory, and is fed back to step 2 and applied to a wafer to be actually polished (step 13). If the polishing amount difference ΔQ is out of the allowable range, the procedure is returned to step 6 with replacement of Qpoli=Qest, P=Pinput, and the procedure from step 6 to step 11 is repeated until the polishing amount difference ΔQ becomes within the allowable range to determine the recommended polishing pressure value Pinput.
The “polishing” in step 3 shown in
The feedback cycle can be set arbitrarily. For example, a method can be employed which involves carrying out the measurement for every wafer and feeding back the estimation results to the next wafer to be polished. According to another usable method, the estimation results are not fed back when the wear of a polishing member is small because of small change in the polishing profile, and are fed back after the wear of the polishing member has reached a certain high level. In the latter method, the measure may be carried out for arbitrarily selected wafers, and application of particular polishing conditions fed back after the measurement of a selected wafer may be continued until the next measurement of another selected wafer. The feedback cycle may be shortened as the wear of the polishing member progresses.
In the case of setting polishing rate instead of polishing amount, the polishing amount Qpoli is divided by polishing time t in step 6. Further, in the case of taking account of polishing rate, the above-described relationship with the distance r and the relative velocity difference Δω may be employed. Polishing conditions (polishing pressure, polishing time, polishing rate), which can provide a desired polishing profile, can thus be determined by using the simulation tool.
When a failure occurs in the polishing apparatus, or a polishing member (consumable member) wears out and reaches its use limit, a desired polishing profile may not be obtained even if the polishing conditions are adjusted. In the case where the polishing amount difference ΔQ between the estimated polishing amount and the target polishing amount, calculated in step 7, changes extremely from the previous calculation, or the recommended polishing pressure Pinput falls outside a range feasible with the polishing apparatus, the operation of the polishing apparatus can be stopped or a warning can be issued. Conventionally, a polishing member (consumable member) is changed with a new one after its use in a certain number of polishing runs so as not to adversely affect the device performance. According to the present invention, it becomes possible to use a polishing member to its use limit without being influenced by the number of polishing runs, thus decreasing the frequency of change of polishing member. Further, the present invention can be used also for failure diagnosis, and can therefore increase the yield of polished products.
Instead of the correction of polishing coefficient made in consideration of the influence of the edge configuration of a wafer, it is possible to correct the back surface pressure based on the results of measurement of the edge configuration after the calculation of the recommended pressure value so as to correct the polishing profile of the edge portion. This can reduce variation of polishing in the peripheral regions of wafers due to variation of edge configurations. For example, in the case of a wafer having a surface oxide film, the recommended polishing pressure value of the outermost retainer ring E5 may be multiplied by a pressure correction coefficient according to the degree of roll-off (corrected retainer ring pressure value=pressure correction coefficient×recommended retainer ring pressure value). The pressure correction coefficient can be created, for example, by actually polishing wafers having known roll-off values with various retainer ring pressures in advance. Alternatively, the pressure correction coefficient may be created by calculating the relationship between the pressure and the degree of roll-off by a finite element method.
The degree of roll-off of a wafer momentarily changes during polishing, due to polishing of the wafer. Accordingly, it is possible to correct the pressure during polishing by measuring the degree of roll-off during polishing with a measuring device installed in the polishing apparatus. The pressure can be corrected without measurement of the degree of roll-off during polishing by creating a pressure correction coefficient also taking the polishing time into consideration.
In the case of a wafer having a surface metal film, the configuration of the end portion of the metal film can be corrected by the same method as the above-described method for correcting the roll-off of an oxide film. The method for correcting an edge configuration with a pressure correction coefficient is also applicable to the case of not carrying out the above-described calculation of recommended pressure values.
The polishing apparatus, by replacement of its top ring, can be applied to a variety of polishing objects. When a top ring is replaced with another one to change a polishing object with another one, it is generally necessary to change a group of pressures (pressure distribution) of the front surface of the former polishing object, the pressures having been calculated for the polishing object according to the configuration of the former top ring, with another group of pressures (pressure distribution) calculated for the latter polishing object according to the configuration of the later top ring. The new data setting may be performed by reading the calculation results of a group of set pressures and pressure distribution data from a computer-readable storage medium, as described above. It is also possible to input parameters, such as the number of the air bags of the top ring, their pressure ranges, etc., upon the start-up of the polishing apparatus, calculate pressure distributions of the front surface of the polishing object, corresponding to the parameters, within the polishing apparatus, and store the data in the control unit.
As described hereinabove, it is possible with the present invention to formulate not only a recipe for flatly polishing an object but also a recipe for polishing an object into a particular configuration. Thus, even when the surface topology of a film on a wafer before polishing is not flat, a recipe can be formulated which, in consideration of the topology, can provide the remaining film after polishing with a flat surface. Further, unlike the conventional practice of optimizing polishing conditions by resorting to an Engineer's empirical rule, the present invention makes it possible to calculate optimum polishing conditions for providing a desired polishing profile. As compared to the conventional adjustment method of polishing a number of test wafers before setting polishing conditions, the present invention can save labor, time and cost. Furthermore, by reading a program according to the present invention into a computer for controlling a polishing apparatus, it becomes possible to add a new function to the polishing apparatus and respond to enhancement of the performance by replacement of a top ring.
| Number | Date | Country | Kind |
|---|---|---|---|
| 2004-202970 | Jul 2004 | JP | national |
