JP2014075489A - Evaluation method for wafer, and polishing method for wafer - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an evaluation method for a wafer capable of evaluating a quality of a polishing surface of the wafer with high accuracy without limiting analysis information at a specific spatial frequency, and thereby, of obtaining an optimum polishing condition uniquely.SOLUTION: An evaluation method for a wafer of evaluating a quality of a polishing surface of the wafer in polishing processing of polishing the wafer by slidably contacting the wafer with a polishing pad of a polishing device, includes the following steps of: measuring a polishing surface shape of the wafer and a surface shape of the polishing pad after polishing; measuring respective power spectrum densities from the measured polishing surface shape of the wafer and the measured surface shape of the polishing pad; acquiring a linear primary approximation formula from a correlation between the power spectrum density of the wafer and the power spectrum density of the polishing pad; and evaluating the quality the polishing surface of the wafer from a gradient of the acquired linear primary approximation formula.

Description

  The present invention relates to a wafer evaluation method and a wafer polishing method for evaluating the quality of a polished surface of a wafer in a polishing process in which a wafer is slid in contact with a polishing pad of a polishing apparatus.

  Due to the recent high integration of devices, the required quality of semiconductor wafers such as silicon wafers is becoming increasingly sophisticated. In particular, improving the flatness and surface roughness of the wafer surface is intended to improve the electrical characteristics of the device. Is necessary.

  For example, it is known that the microroughness on the wafer surface affects the electrical characteristics of the device. For example, if the microroughness is large, the oxide breakdown voltage decreases, and the microroughness is large in the channel under the gate oxide. Then, it is known that electron scattering occurs and the electron mobility becomes small. In particular, it has been found that haze, which has irregularities with a spatial wavelength of about 0.01 to 5 μm on the wafer surface, affects the reliability test of the electrical characteristics of the device, particularly the temporal dielectric breakdown characteristics (TDDB) of oxide films Yes.

On the other hand, undulations having a spatial wavelength of about several mm to 20 mm on the wafer surface are also problematic in photolithography, element isolation, and the like in the device manufacturing process. On the other hand, it is disclosed that the surface shape of the back surface of the wafer is measured, the power spectral density is obtained therefrom, and the power spectral density at a spatial wavelength of 10 mm is set to 10 μm ³ or less (Patent Document 1).

  In addition, a method is disclosed in which, in a wafer polishing process or a cleaning process, the surface shape of a measured wafer is converted into a power spectrum to evaluate microroughness, and polishing conditions and cleaning conditions are adjusted according to the results ( Patent Document 2).

JP 2000-31224 A JP 2006-278513 A

  However, in this evaluation method, a spatial frequency at a certain point that is easily influenced by polishing conditions is captured, and the relationship between the spatial frequency and the quality of the wafer surface is evaluated. Actually, the spatial frequency that varies depending on the polishing conditions is not a single point, but has a certain range. Therefore, in this evaluation method, the evaluation result varies depending on the spatial frequency employed.

  The power spectral density of the wafer surface roughness greatly depends on the polishing tool that is in sliding contact with the wafer surface in CMP processing of the wafer. Examples of polishing tools include polishing pads and polishing slurries, and their polishing action determines the power spectral density, that is, the roughness density for each spatial frequency. However, it is difficult to determine which factor of polishing conditions actually contributes to determine the power spectral density.

As described above, although the spatial frequency affected by the polishing conditions has a certain width, the conventional evaluation methods using the power spectral density pick up one point with the spatial frequency and limit it to that. It remained in analysis technology to evaluate the relationship with quality. Therefore, for example, there is a problem that it is not uniquely determined which point should be adopted when an optimum condition is selected from polishing conditions having a wide fluctuation range of power spectral density.
The present invention has been made in view of the above-described problems. The quality of a polished surface of a wafer can be evaluated with high accuracy without restricting analysis information at a specific spatial frequency, thereby uniquely identifying optimum polishing conditions. It is an object of the present invention to provide a wafer evaluation method that can be obtained in an automated manner.

  In order to achieve the above object, according to the present invention, in a polishing process in which a wafer is slid in contact with a polishing pad of a polishing apparatus for polishing, a wafer evaluation method for evaluating the quality of a polished surface of the wafer, A step of measuring the polished surface shape of the wafer and the surface shape of the polishing pad after polishing, a step of measuring respective power spectral densities from the measured polished surface shape of the wafer and the surface shape of the polishing pad, and the wafer Obtaining a linear linear approximation from the correlation between the power spectral density of the polishing pad and the power spectral density of the polishing pad, and evaluating the quality of the polishing surface of the wafer from the gradient of the obtained linear linear approximation A method for evaluating a wafer is provided.

  With such a wafer evaluation method, power spectral densities that differ for each spatial frequency can be unified as a linear linear approximation gradient, and the chemical and mechanical balance that is the principle of CMP processing is quantitatively evaluated from the gradient. Thus, the quality of the polished surface of the wafer can be evaluated with high accuracy. From this evaluation result, the optimum polishing condition can be uniquely obtained.

At this time, in the step of evaluating the quality of the polished surface of the wafer, it is preferable to evaluate the quality of the polished surface of the wafer on the basis of the gradient when polishing is performed using a polishing slurry without abrasive grains.
In this way, the quality of the polished surface of the wafer can be more easily and accurately evaluated.

Further, in the present invention, there is provided a wafer polishing method for polishing by bringing a wafer into sliding contact with a polishing pad of a polishing apparatus, wherein the quality of the polished surface of the wafer is evaluated by the wafer evaluation method of the present invention, and the evaluation result Accordingly, there is provided a method for polishing a wafer, characterized by adjusting polishing conditions for the next polishing.
Such a wafer polishing method can be carried out under the optimum polishing conditions obtained by the wafer evaluation method of the present invention, and can be polished to a wafer having improved polishing surface quality such as microroughness and ESFQR. .

  In the present invention, each power spectrum density is measured from the measured polishing surface shape of the wafer and the surface shape of the polishing pad, a linear first-order approximation is obtained from these correlations, and the wafer is obtained from the gradient of the obtained first-order linear approximation. Since the quality of the polished surface is evaluated, different power spectral densities for each spatial frequency can be unified as the gradient of the linear linear approximation formula, and the chemical and mechanical balance, which is the principle of CMP processing, can be quantitatively evaluated from the gradient. The quality of the polished surface of the wafer can be evaluated with high accuracy. From this evaluation result, the optimum polishing condition can be uniquely obtained.

It is a flowchart which shows an example of the evaluation method of the wafer of this invention. It is explanatory drawing explaining the mode of the interposition ratio of the polishing slurry in the grinding | polishing surface at the time of wafer grinding | polishing. It is the schematic which shows an example of the double-side polish apparatus which can be used with the grinding | polishing method of the wafer of this invention. It is a figure which shows the power spectrum density of the grinding | polishing surface shape of the wafer in Example 1 and a comparative example. It is a figure which shows the power spectrum density of the surface shape of the polishing pad in Example 1. FIG. 5 is a scatter diagram obtained in Example 1. FIG. FIG. 3 is a diagram showing a gradient value under each polishing condition in Example 1. It is a figure which shows the relationship between the gradient value in Example 1, and RMS. It is a figure which shows the relationship between the gradient value in Example 2, and ESFQR of the wafer after grinding | polishing. It is a figure which shows the result which put together the power spectrum density in a comparative example for every grinding | polishing conditions. It is a figure which shows the relationship between the power spectrum density and RMS in a comparative example.

Hereinafter, although an embodiment is described about the present invention, the present invention is not limited to this.
The wafer evaluation method of the present invention will be described with reference to FIG.
First, the polished wafer and a part of the polishing pad are sampled (FIG. 1A). Here, the wafer is obtained by slicing ingots such as silicon and other compound semiconductors grown by, for example, the CZ method or the FZ method, and appropriately performing processes such as chamfering, lapping, etching, and polishing. .
Next, the polishing surface shape of the wafer and the surface shape of the polishing pad are measured (FIG. 1B). The method for measuring the surface shape is not particularly limited. For example, various surface roughness measuring devices can be used for measuring the polished surface shape of the wafer. Further, for example, a laser microscope can be used to measure the surface shape of the polishing pad.

Next, the respective power spectral densities (PSD) are measured from the measured polished surface shape of the wafer and the surface shape of the polishing pad (FIG. 1C). Thereby, the power spectral density of the polishing surface shape of the wafer with respect to the spatial frequency and the power spectral density of the surface shape of the polishing pad with respect to the spatial frequency are obtained.
Next, a correlation between the power spectral density of the wafer and the power spectral density of the polishing pad is obtained from these obtained power spectral densities by synchronizing the respective spatial frequencies (FIG. 1 (d)). This correlation can be obtained as a scatter diagram, for example. A linear first-order approximation formula is acquired from this correlation, that is, a scatter diagram.

Next, the gradient of the obtained linear first-order approximation is calculated and evaluated (FIG. 1 (e)). From the value of this gradient, as described below, the quality of the wafer polishing surface can be evaluated, and more quantitative polishing conditions can be selected.
Thus, in the wafer evaluation method of the present invention, the power spectral density of the polishing pad is simultaneously monitored in addition to the power spectral density of the wafer. By combining these pieces of information, specifically, by obtaining a scatter diagram of wafer surface roughness density information and polishing pad surface roughness density information for each spatial frequency, different roughness density information for each spatial frequency is linearized. It can be unified as the gradient value of the linear approximation. This is different from the prior art in which only one point of spatial frequency information is taken into consideration in that it takes into account roughness density information over a wide range of spatial frequencies.

Here, the meaning of the above-described gradient value will be described in detail. As described above, the scatter diagram is represented by the power spectral density of the wafer with respect to the power spectral density of the polishing pad when the spatial frequencies of both power spectral densities are matched. The gradient value of the linear first-order approximation formula in this scatter diagram is a scale indicating how much the polishing pad surface roughness component has been transferred to the wafer.
In the polishing process, it is considered to be an ideal state for maintaining the quality of the polished surface that the abrasive grains of the slurry are sufficiently distributed in the wafer surface. At this time, if there are no abrasive grains of slurry between the wafer and the polishing pad, the polishing pad surface roughness component is transferred directly to the wafer polishing surface through only the slurry liquid.

  When abrasive grains of slurry exist between the wafer and the polishing pad, transfer of the polishing pad surface roughness component to the wafer surface is suppressed by the abrasive grains. In this case, the gradient value is smaller than when no abrasive grains are present. That is, the gradient value in the absence of abrasive grains is considered as a reference, and the ratio of how the gradient value under a certain polishing condition is changed relative to this reference can be quantitatively expressed.

That is, the gradient value can be expressed as a transfer rate of the polishing pad surface roughness component based on the abrasive-free condition. Depending on the magnitude of the gradient value, it is possible to compare the proportion of polishing slurry on the polishing surface, that is, the balance between chemical and mechanical. Specifically, as the gradient value is larger, as shown in FIG. 2A, the polishing pad surface roughness component is smaller because the abrasive particle concentration of the slurry between the wafer polishing surface and the polishing pad is smaller or the abrasive particle size is smaller. Thus, it can be determined that the chemical is relatively dominant.
Conversely, as the gradient value is smaller, as shown in FIG. 2B, the abrasive grain concentration of the slurry between the wafer polishing surface and the polishing pad is larger or the abrasive grain size is larger, so that the polishing pad surface roughness component is transferred. The rate is smaller, so that it can be determined that mechanical is dominant.
Therefore, it is possible to quantitatively evaluate the balance between chemical and mechanical fluctuations depending on the polishing conditions, and it is possible to set more accurate polishing conditions.

  For example, polishing is performed in advance under the condition of a reference abrasive-free slurry, and the calculated gradient value is set as a reference (= 1). Thereafter, the gradient value calculated when polishing under various polishing conditions is obtained as a relative value of the reference value, and the relationship with the quality to be improved such as microroughness or ESFQR is obtained. The optimum polishing conditions for improving the quality are selected from the relationship between the gradient value thus obtained and the quality (FIG. 1 (f)).

  Thus, the wafer evaluation method of the present invention can quantitatively evaluate the chemical and mechanical balance, which is the principle of CMP processing, from the gradient, and can evaluate the quality of the polished surface of the wafer with high accuracy. From this evaluation result, the optimum polishing condition can be uniquely obtained.

Next, the wafer polishing method of the present invention will be described. The present invention can be applied to both the double-side polishing in which both surfaces of the wafer are simultaneously polished and the single-side polishing in which one side is polished. Here, the case of performing double-side polishing will be described.
FIG. 3 shows an example of a double-side polishing apparatus that can be used in the wafer polishing method of the present invention. The double-side polishing apparatus 10 includes an upper surface plate 2 and a lower surface plate 3 that are provided to face each other vertically, and a polishing pad 4 is affixed to each surface plate 2, 3. The wafer W is held in a holding hole provided in the carrier 1 and is sandwiched between the upper surface plate 2 and the lower surface plate 3.

  Using such a double-side polishing apparatus 10, first, the wafer W to be polished is inserted into the holding hole of the carrier 1, and the carrier 1 is placed between the upper surface plate 2 and the lower surface plate 3. Then, the upper and lower surfaces of the wafer W are sandwiched between the polishing pads 4 attached to the upper and lower surface plates 2 and 3, and the upper surface plate 2 and the lower surface plate 3 are rotated. While supplying polishing slurry containing abrasive grains to the polishing surface of the wafer, the wafer W is brought into sliding contact with the polishing pad 4 to polish both surfaces.

Thereafter, the quality of the polished surface of the wafer after polishing is evaluated by the wafer evaluation method of the present invention described above, and the polishing conditions for the next polishing are adjusted according to the evaluation result.
With such a wafer polishing method, optimum polishing conditions can be obtained by the wafer evaluation method of the present invention as described above, so that the quality of the polished surface such as microroughness and ESFQR can be improved. Can be polished.

  EXAMPLES Hereinafter, the present invention will be described more specifically with reference to examples and comparative examples of the present invention, but the present invention is not limited to these.

Example 1
The quality of the polished surface of the wafer was evaluated according to the wafer evaluation method of the present invention. Here, microroughness (RMS) was evaluated as the quality of the polished surface.
LSM-3000 manufactured by Taylor Hobson was used for measurement of the polished surface shape of the polished wafer. Laser surface CLSM (1LM21) was used for measuring the surface shape of the polishing pad. As the PSD analysis software, Taylor Map Gold manufactured by Taylor Hobson was used. The RMS value as microroughness was measured by TMS-3000W manufactured by Schmitt.

  First, in order to prepare a plurality of wafers to be evaluated, double-side polishing was performed under a plurality of polishing conditions described below. First, the wafer was a silicon wafer having a diameter of 300 mm. The primary abrasive particle size of the slurry was changed to 12, 35, 74, and 89 nm. The abrasive concentration was changed to 0.5, 1.25, 2.5, and 5.0 wt%. Moreover, the wafer at the time of grinding | polishing using the polishing slurry without an abrasive grain was also prepared. The double-side polishing apparatus used was DSP-20B manufactured by Fujikoshi Machinery Co., Ltd., the polishing pad was MH-S15A manufactured by Nitta Haas, and the polishing slurry was KOH-based colloidal silica slurry manufactured by Fujimi Incorporated. As a carrier of the double-side polishing apparatus, a base material made of Ti and an aramid FRP inserted in a holding hole was used.

The polished surface shape of the prepared polished wafer and the surface shape of the polishing pad were measured, and the power spectral density of each was measured.
4A and 4B show the power spectral density of the wafer measured. FIG. 4A shows the case where the abrasive grain size is constant at 35 nm and the abrasive grain concentration is changed in the range of 0 to 5.0 wt%. FIG. 4B shows the abrasive grain concentration of 5 0.0 wt% is constant, and the non-abrasive grains and the abrasive grain diameter are changed in the range of 12 to 89 nm.
FIG. 5 shows the measured power spectral density of the polishing pad.

Next, the spatial frequency was synchronized from these power spectral densities to obtain a scatter diagram of the power spectral density of the wafer and the power spectral density of the polishing pad. A linear first-order approximation was obtained from this scatter diagram.
Here, the obtained scatter diagrams are shown in FIGS. FIG. 6A shows the case where the above-mentioned abrasive grain size is made constant, and FIG. 6B shows the case where the abrasive grain concentration is made constant.

FIG. 7 shows the gradient of the linear first-order approximation expression under each polishing condition. Here, the gradient value is a relative value when the gradient value is 1 when an abrasive-free slurry is used.
These gradient values shown in FIG. 7 indicate the relative transfer rate of the polishing pad surface roughness component for each polishing condition to the wafer surface as described above, and the larger the abrasive particle size from FIG. Further, it was confirmed that the higher the abrasive concentration, the lower the transfer rate, and the tendency for the abrasive grains of the slurry interposed between the wafer surface and the pad surface to increase.

The result of the relationship between the slope value and RMS is shown in FIG. As shown in FIG. 8, it was found that the smaller the slope value, the better the RMS.
With reference to the result of FIG. 8, the RMS with respect to the obtained gradient value can be evaluated, that is, the quality of the polished surface of the wafer can be evaluated. In addition, referring to FIG. 7 and FIG. 8, it is possible to select optimum polishing conditions (abrasive grain size and abrasive grain concentration of the polishing slurry) for reducing RMS. For example, if the polishing is performed under the polishing conditions in which the gradient value in FIG. 8 is 0.65 or less, the quality target of the wafer can be achieved. As shown in FIG. 7, the polishing conditions for this are polishing conditions with an abrasive grain size of 35 nm or more and an abrasive grain concentration of 0.5 to 5 wt% excluding the conditions of an abrasive grain size of 35 nm and an abrasive grain concentration of 0.5 wt%. is there.

(Example 2)
The silicon wafer was double-side polished according to the polishing method of the present invention. The polishing conditions were the same as in Example 1. However, as the polished wafer, a wafer having a laser mark at a rotational angle of 275 degrees on the outermost peripheral portion was used. The silicon wafer having the laser mark on the outermost peripheral portion has a problem that the undulation is generated in the vicinity of the laser mark after polishing on both sides due to the difference in polishing rate with the surroundings, and the maximum flatness is deteriorated. It has been inferred that the fluctuation factor of the undulation size in the vicinity of the laser mark is the balance between chemical and mechanical during polishing, and in order to improve this, it is necessary to make the mechanical more dominant.

Therefore, in order to improve the ESFQR after polishing, the gradient value under each polishing condition was determined according to the wafer evaluation method of the present invention, and the optimum polishing condition for reducing this gradient value was selected.
FIG. 9 shows the relationship between the slope value and the ESFQR of the polished wafer. As shown in FIG. 9, it was confirmed that ESFQR can be improved as the gradient value is smaller. A condition (specifically, an abrasive grain size of 89 nm and an abrasive grain concentration of 5.0 wt%) having a gradient value of about 0.2 could be selected as the optimum polishing condition.

(Comparative example)
Using the same wafer as that prepared in Example 1, only the power spectrum density of the polished surface shape of the wafer was measured, and the RMS of the wafer polished surface was evaluated according to a conventional evaluation method in which one spatial frequency was selected.
First, similarly to Example 1, the power spectrum density of the polished surface shape of the wafer as shown in FIGS. 4A and 4B was measured.
4A and 4B, it was confirmed that the range in which the power spectral density fluctuates depending on the polishing conditions is as wide as 3.0E-3 to 5.0E-01 [1 / μm]. Among these, three spatial frequencies, for example, f = 2.0E-02, 5.0E-02, and 1.0E-01 [1 / μm] were selected as the spatial frequencies having large fluctuations. The result of putting together the power spectrum density regarding each selected spatial frequency for every polishing condition is shown in FIG.

  Further, FIG. 11 shows the relationship between the power spectral density and the RMS. As shown in FIG. 11, different correlations exist depending on the selected spatial frequency, and appropriate polishing conditions could not be obtained uniquely.

  The present invention is not limited to the above embodiment. The above-described embodiment is an exemplification, and the present invention has any configuration that has substantially the same configuration as the technical idea described in the claims of the present invention and that exhibits the same effects. Are included in the technical scope.

1 ... carrier, 2 ... upper surface plate, 3 ... lower surface plate, 4 ... polishing pad,
10: Double-side polishing apparatus.

Claims (3)

In a polishing process in which a wafer is brought into sliding contact with a polishing pad of a polishing apparatus for polishing, a wafer evaluation method for evaluating the quality of a polished surface of the wafer,
Measuring the polished surface shape of the wafer after polishing and the surface shape of the polishing pad;
Measuring each power spectral density from the measured polished surface shape of the wafer and the surface shape of the polishing pad;
Obtaining a linear first-order approximation from the correlation between the power spectral density of the wafer and the power spectral density of the polishing pad;
And a step of evaluating the quality of the polished surface of the wafer from the gradient of the acquired linear first-order approximation formula.   2. The quality of the polished surface of the wafer is evaluated based on a gradient when polishing is performed using a polishing slurry without abrasive grains in the step of evaluating the quality of the polished surface of the wafer. Wafer evaluation method. A method for polishing a wafer in which the wafer is slid in contact with a polishing pad of a polishing apparatus,
The wafer polishing method according to claim 1, wherein the quality of the polished surface of the wafer is evaluated by the wafer evaluation method according to claim 1, and the polishing conditions for the next polishing are adjusted according to the evaluation result. Method.

Images