Patent Application
20100159804
1. Field of the Invention
The present invention relates to a method of determining the pattern evolution of the down force (DF) and shear force (SF) and from them the coefficient of friction (COF) during chemical mechanical polishing between a semiconductor wafer and the polishing pad using Fast Fourier Transform (FFT) or Fourier Transform (FT) spectra of periodic down force, shear force and/or coefficient of friction measurements.
2. Description of the Prior Art
Currently under research and development are processing methods for improvement in density and miniaturization in production of ULSI semiconductor devices. One of the methods, CMP (chemical mechanical polishing) technology, is now a technology essential in production of semiconductor devices; for example, for polishing an interlayer dielectric film, isolating a shallow trench device, forming a plug or preparing embedded metal wiring.
Generally, in chemical mechanical polishing, a polishing pad is first fixed on the rotary polishing pad of a polisher, while an irregular-surfaced semiconductor wafer is fixed on a polishing head. Chemical mechanical polishing is performed by pressing the polishing head onto the revolving polishing pad, while CMP slurry is supplied to the polishing pad. Irregularity on the wafer present before polishing is eliminated by chemical mechanical polishing, and the wafer surface is planarized. The polishing should be terminated immediately after the surface is planarized to ensure maximum uniformity or until unwanted top layers are removed. Until today, the only way to do this job in microelectronic manufacturing is CMP.
A time management method for maintaining a constant polishing period and a method of determining the polishing endpoint detection has been used for making the thickness of the surface-planarized film constant after polishing of a semiconductor wafer, but a method for monitoring pattern evolution in SF, DF or COF is advantageous because it allows the operator greater flexibility in when to end or alter the operation conditions of a process and provides additional useful and precise information as to the progress of the polishing. Such a method can also be used as a diagnostic tool for the polishing process. When polishing using the same set of consumables, one should expect a similar pattern evolution progress. If the pattern evolution is shown to be altered during polishing, it indicates that there is something wrong during polishing making it possible for the operator to take action without further harming the overall process. In polishing a semiconductor wafer carrying an integrated circuit pattern the shear force will vary, according to the material composition, structure and surface conditions of the film being polished. Heretofore, methods of using shear force in the endpoint detection method have been disclosed; for example, in U.S. Pat. No. 5,046,015 and Japanese Patent Application Laid Open No. 8-197417 both incorporated herein by reference and endpoint detection has lead to improvement in the reproducibility of the extent of and thickness of material removed by chemical mechanical polishing.
In the polishing method above, the shear force provides a torque to the polishing table. Thus, it is possible to determine the shear force by measuring the electric current of the motor driving the polishing table. The shear force F, the torque Tq generated on the polishing table, and the distance r between the position of the shear force applied to the polishing table and the rotational center of the polishing table have the relationship: Tq=F×r. However, the position r of the semiconductor wafer on the polishing table is variable as it moves during polishing, and thus, the shear force F cannot be determined only by motor current. As described above no method of directly measuring the shear force generated between a revolving semiconductor wafer and a polishing pad that can be performed easily industrially has yet been disclosed.
For example, when the conditioning, that is to say, the surface roughening of the polishing pad is performed simultaneously with polishing, a torque is applied to the motor driving the polishing table, and the motor current changes. In addition, the load created by the polishing table itself is applied to the motor, and contribution of the shear force between the semiconductor wafer and the polishing pad to the motor torque becomes relatively smaller. Thus, determination of the shear force between semiconductor wafer and polishing pad from the motor current leads to expansion of error.
In US Patent Application No. 2008/0200032, incorporated herein by reference, a polishing method for measuring the COF during polishing of a semiconductor wafer and using the change thereof in determining the polishing end point is disclosed. This patent detects the endpoint by means of significant alterations in the amplitude of frequency peaks in fast fourier transform spectra of shear force versus time. However, this patent application provides no method or disclosure concerning the analysis, based on variance of shear force, variance of down force, FFT of shear force, FFT of down force and FFT of COF of polishing pattern evolution prior to polishing endpoint.
The present invention is method of determining pattern evolution of a semiconductor wafer during chemical mechanical polishing prior to polishing end point by measuring the periodic change in the variance and FT or FFT frequency spectra of shear force and change in variance or FT or FFT frequency spectra of COF, shear force and/or down force between the semiconductor wafer and the polishing pad.
Using the method of the present invention, it is possible to determine these pattern evolutions within a single film or sequentially within multiple films during the chemical mechanical polishing of a wafer easily and determine characteristics of individual wafers or series of wafers to aid manufacturers in wafer development or to help determine optimal polishing conditions for particular wafers or types of wafers.
In the method of polishing a semiconductor wafer according to the present invention, a semiconductor wafer is polished while pressed on a polishing pad fixed on a revolving polishing table. Simultaneously, a slurry is supplied to the space between the polishing pad and the semiconductor wafer. The semiconductor wafer may be held by a polishing head, and the polishing head may be rotated by a driving unit, independently from the polishing table.
In the polishing method according to the present invention, the pattern evolution is determined from the change in variance and the frequency peaks of the FFT spectra of the shear force and COF between the wafer and polishing pad during polishing. The coefficient of friction COF between the wafer and polishing pad during polishing is represented by the ratio of the shear force Fshear applied to the wafer and the polishing pad to the load applied to the wafer Fnormal (Fshear/Fnormal). Fnormal is a value in proportion to the load applied to the polishing head, and thus, the coefficient of friction COF is in proportion to the shear force Fshear when Fnormal is constant.
In directly determining the shear force Fshear applied to the wafer and the polishing pad (hereinafter referred to also as SF), a force in the “horizontal” direction parallel to the surface of the polishing head (x-y direction) generated on the polishing table or the polishing head may be measured. The method of determining the coefficient of friction COF by the force in the x-y direction generated on the polishing head and the load applied via the polishing head onto the polishing table will be described with reference to the drawings.
The center of the semiconductor wafer 15 is fixed by the polishing head 16 and is placed at a distance from the center of the polishing pad 13 on the polishing table 12, and thus, a shear force in the x-y direction is applied by friction with the polishing pad 13. The shear force generated on the semiconductor wafer 15 is transmitted, through the polishing head 16, motor 2, and slide plate 17, to the load cells 19b and 19c. The load cell 19b detects the depth-direction component (“y component”) of shear force, while the load cell 19c the lateral-direction component (“x-component”) of shear force perpendicular thereto; and these components are transmitted to the recorder 20 and FFT 21.
The ratio Fshear/Fnormal and the coefficient of friction COF are calculated from the shear force in combination of these two components and the load in the “z-direction”.
The method of measuring the COF in the x-y plane generated on the polishing table is the same in principle as method of determining the SF described above. The voltage signals obtained for shear force and down force are transmitted to the recorder 20 and the FFT signal processing unit 21 and processed to output COF and variance of COF.
Although the polishing head presses the polishing table downward in
The shear force is measured in real time, and all components including direct-current to high frequency components are determined according to the frequency characteristics of the load cell or strain gauge. The friction coefficient obtained from the shear force also includes a high frequency component, and it is possible to analyze the down force, shear force and coefficient of friction at each frequency by FT or FFT.
The COF depends on the physical properties of the film to be polished, the CMP slurry, and the polishing pad. Conditioning using a dresser such as a diamond conditioner disc may be needed to maintain the roughness of the polishing pad surface constant, and such conditioning is performed during or after polishing.
When there is irregularity on the surface of the film to be polished, the load concentrates on the prominent or elevated regions. The area upon which load is concentrated widens as surface irregularity is reduced by the progress of polishing, and the load is at last applied uniformly on the entire semiconductor wafer surface after the surface has been planarized completely. The measured shear force varies by the change of the area exposed to concentrated load by planarizing resulting from polishing, and of course the various features of the pattern evolution prior to endpoint can be determined and using this change.
For example, as shown in part 1 of
The pattern evolution of the present invention which comprises the change in pattern of the wafer surface over time and the change over time of the FT or FFT spectra of the frequency of shear force measurements but which may also include the change in variance over time of both the SF, DF and COF is due to a combination of physical changes that occur at the wafer surface prior to endpoint as CMP polishing proceeds that include but are not limited to elimination of the initial surface roughness, the representation of which may be observed between Parts 1 and Parts 2a and 2b but additionally to uneven features or dishing 36 that develop as a result of the depth and physical properties of structures 38 that have not yet been uncovered by CMP. Ideally the progress of CMP would flow from Part 1 to Part 2a and then to 3a with the result of an entirely flat surface both during the process and at the end. Analysis of the pattern evolution can indicate how well this is progressing and also provide information about the wafer being polished which can lead to alterations or improvements in the settings or conditions of polishing for that wafer. What often tends to happen is that some surface features persist as in 3b. Determination of the endpoint is useful in its own right but does not provide the diagnostic and potentially corrective enabling information of the present invention.
It is clear that the present invention monitors at least two processes that occur during polishing: the elimination of initial roughness of the wafer surface and the development of dishing or other surface features while still removing the initial layer that are indicative of material conditions of the layer or the presence and dimensions of deeper structures.
SF and DF are measured over a series of short intervals. These intervals are preferably between 2 and 15 seconds in length and have the same length. Variance is calculated for each of them. The longer the interval the better the variance figure but the shorter the interval, the sharper the picture of the SF and COF pattern evolution that emerges. There is a point beyond which lengthening of the interval ceases to generate improvements in the variance figure commensurate with the loss of pattern definition. That point is typically about 5 seconds and that is the preferred value for interval length.
When the shear force at each frequency is determined by fast FT or FFT of the shear force, the maximum (peak) intensities appear at a particular frequencies. The peak intensity varies in relation to the degree of irregularity of the wafer surface, and thus, it is possible to determine the evolution in pattern by monitoring the change in peak intensity. The peak frequency, which is influenced by the location, shape and dimension of irregularities of the surface of the wafer and CMP conditions is determined separately for each semiconductor wafer produced.
Before the CMP end-point, the fluctuation in certain frequency peaks is related to the pattern evolution or, as we can also say, the step height evolution, or even more basically, step height reduction and it is also related to processes or features at the polishing surface that can be used as a diagnostic of the polishing process in real time.
In regard to the type of slurry used in the present invention, typically, the selection of the slurry is not particularly limited and is based on the slurry performance, that is to say the removal rate, within reasonable limits of wafer-non-uniformity. It is possible to detect the pattern evolution with the present invention when slurries are used that are effective for various CMP processes. Any slurry used in CMP processing may be used as the slurry of the present invention.
The COF and variance of shear force figures results obtained using the present invention may either increase, decrease or otherwise vary with time depending upon the wafer and the conditions under which CMP is carried out.
The polishing conditions may be altered from those specified in the above embodiment; however, once conditions are established, it is preferred to maintain them constant during the CMP operation and for comparative runs. However, for example, the load applied to the wafer or the rotation rate of the polishing pad, conditioner or wafer may be altered. In the event that the rate of any of these is altered during the run it is preferred that they be altered in a constant or at least consistent manner.
Results (spectra) obtained by Fast Fourier transformation of the shear forces acquired during the polishing intervals in
The polishing method according to the present invention is not limited to the materials described in the foregoing example and is suitable for determining pattern evolution for any overburden removed by CMP processes including for example copper and Ta/TaN. It is a characteristic of the method of the present invention that it is primarily concerned with pattern evolution during the period of removal of overburden before the underlying layer in wafers subjected to CMP polishing have been exposed.
Hereinafter, the present invention will be described with reference to Examples.
The position of semiconductor wafer 15 is fixed by polishing head 16 and is placed away from the center of polishing table 12, and thus, a shear force in the x-y direction is applied by friction with polishing pad 13. The shear force generated on semiconductor wafer 15 is transmitted, through polishing head 16, motor 2, and slide plate 17, to load cells 19b and 19c. The load cell 19b detects the depth direction component of shear force, while the load cell component 19c the width-direction component of shear force, and these components are transmitted to the recorder 20 and FFT device 21.
A test pattern having the cross sectional structure shown in
The operational condition of the CMP polisher is as follows: polishing table rotational frequency 93 per minute, polishing head rotational frequency 97 per minute, polishing head pressure 22 kPa, diamond conditioning disc load: 26 N, and diamond conditioner disc rotation rate: 30 per minute. Conditioning was performed simultaneously during polishing. The rate of application of CMP slurry was 200 ml/min.
Spectra obtained by FFT of the shear force are shown in
Observation of peak A at about 2 Hz shows that it remained consistent over time but increased slightly in intensity. Peak B at about 5 Hz increased in intensity from T2 to T3 and then declined at T4, Peak C at about 8 Hz differentiated from Peak B from T2 to T3 and then increased in intensity at T4. Peak D demonstrated increased shouldering at T3 and then increased in intensity by T4. Peaks E and F appear to have fused by T3 and then differentiated even more significantly by T4. This may in fact be complete shifts of minor peaks from overlapping one peak to overlapping an entirely different peak or peaks by another time period. Peak G is not strongly represented at T2 but appears to represent a modest intensity by T4 and may represent a process indicating changes in surface features of the wafer as the outer layer is removed. Peak H is not clearly present at T2 but becomes very sharp though of large intensity by T4. Peak I remains sharp throughout the process until T4 and then diminishes indicating that some feature of the wafer surface has perhaps disappeared by this point. Finally, Peak J is all but nonexistent at T2, prominent at T3 and returns to being a shoulder to Peak I by T4. It is interesting that there are peaks that appear to exist during certain time periods in the polishing process and not in others and also that at T3 and perhaps over the entire period there is considerable shifting of the maximum Hz values for the peaks.
Interestingly the FFT graph of the COF shows much the same features of pattern evolution as that of the shear force suggesting, perhaps unsurprisingly, that in this example variation of frequency maxima over time is related to the shear force. The development of peaks K, J and L roughly parallel those of A, B and C and M, N and O those of D, E and F. Finally, Peaks P and Q parallel peaks H and I in development and peaks for G and J are not clearly paralleled in the COF FFT graphs suggesting possibly that they are not due to the shear force component of COF.
Finally, the transient variance of the shear and down forces suggest a decrease in shear forces from T0 to about T2 and from there a continuous increase through T3 with a brief drop and recovery in shear force prior to T4. It is important that the FFT or FT spectra of COF and SF be considered in light of and in conjunction with the variance of shear force because this allows an added dimension to efforts to tie physical effects at the wafer surface to frequencies and to identify and explain these physical effects.
1 is a CMP polisher
2 is a Motor
3 is a stand
11 is a drive motor
12 is a polishing table
13 is a polishing pad
15 is a semiconductor wafer
16 is a polishing head
17 is a slide plate
19
a is a load cell
19
b is a load cell
19
c is a load cell
20 is a recorder
21 is an FFT or data processing device
31 is a silicon wafer
32 is a pad oxide layer
33 is a silicon nitride stopper film
34 are trenches
35 is an HDP SiO2 film
36 is dishing or other surface anomaly
38 is an underlying structure
The method of the present invention allows considerable detailed investigation to be carried out in real time into the factors controlling the removal of material and the structure and characteristics of the wafer surface during CMP. Once the spectra are analyzed and features of the spectra coupled to some degree with surface phenomena on the wafer during CMP it is possible to understand how the wafer surface changes under given CMP conditions and ultimately in addition to providing a deeper understanding into the mechanics of the CMP process this will allow a certain amount of process optimization and, since the method may be used in real time, feedback for the process in situ may be possible as well once the characteristics of a particular wafer surface under particular conditions are well understood using the method of the present invention. So in essence the method of the present invention provides both extremely in depth and cost effective monitoring and ultimately a means of controlling critical aspects of the CMP process to improve it and obtained better and more consistent results.
The method of the present invention also provides a diagnostic measure of the ‘health’ of the polishing process. When polishing using the same set of consumables, one should expect the similar pattern evolution progress. If the pattern evolution is shown to be significantly altered during polishing, it indicates that there is a potential problem or anomaly occurring during polishing and this allows the operator to take the prompt action minimizing further harm to the overall process.
