JP7616022B2 - Method for evaluating oxide film thickness and method for manufacturing silicon substrate with oxide film - Google Patents

Description

The present invention relates to a method for evaluating the thickness of an oxide film and a method for manufacturing a silicon substrate with an oxide film.

In the manufacturing process of single crystal silicon wafers for semiconductor devices, their main surfaces are finished by a polishing process. In addition, a cleaning process is carried out to remove abrasives and metal impurities that have adhered to the silicon wafer surface during the polishing process. A cleaning method called RCA cleaning is used in this cleaning process.

This RCA cleaning is a cleaning method that combines SC1 (Standard Cleaning 1), SC2 (Standard Cleaning 2), and DHF (Diluted Hydrofluoric Acid) cleaning depending on the purpose. SC1 cleaning is a cleaning method that uses an alkaline cleaning solution made by mixing ammonia water and hydrogen peroxide water in any ratio, and lifts off attached particles by etching the silicon wafer surface, and removes particles while suppressing reattachment to the silicon wafer by utilizing the electrostatic repulsion between the silicon wafer and the particles. SC2 cleaning is a cleaning method that dissolves and removes metal impurities on the silicon wafer surface using a cleaning solution made by mixing hydrochloric acid and hydrogen peroxide water in any ratio. DHF cleaning is a cleaning method that removes natural oxide films on the silicon wafer surface using dilute hydrofluoric acid. In addition, ozone water cleaning, which has a strong oxidizing power, may also be used to remove organic matter adhering to the silicon wafer surface and to form a natural oxide film on the silicon wafer surface after DHF cleaning. The surface quality of the silicon wafer after cleaning, such as particle removal and surface roughness, is important, so these types of cleaning are combined depending on the purpose.

Semiconductor elements such as MOS (Metal Oxide Semiconductor) capacitors and transistors are formed on the surface of semiconductor silicon wafers. The insulating films formed on these semiconductor elements, such as the gate oxide film, are used under high electric field strength, and silicon oxide film, which is easy to form, is often used as this insulating film.

One method for evaluating the thickness of oxide films on silicon substrates is to use an ellipsometer. An ellipsometer measures the phase difference (Δdelta) and amplitude ratio (Ψpsi) by irradiating polarized light onto a substrate sample and measuring the change in the polarization state between the incident light and the reflected light. For example, in the case of a silicon oxide film on a silicon substrate, the polarization state of the incident light changes as it is reflected by the silicon oxide film on the top surface and the interface between the silicon oxide film and the silicon substrate. There are two types of ellipsometers: a single-wavelength type that uses a laser as a light source, and a spectral type that uses a white light source containing multiple wavelength components. The single-wavelength type measures delta and psi for a specific wavelength (e.g. 633 nm), while the spectral type can measure delta and psi for each wavelength, and it is known that the spectral type, which provides a large amount of information, can be used to evaluate film thickness more accurately.

As mentioned above, the information obtained by ellipsometer measurements is the phase difference and amplitude ratio, and film thickness cannot be calculated directly. To calculate film thickness, a model corresponding to the substrate sample is created, and the delta and psi theoretically obtained from this model are compared with the delta and psi obtained by ellipsometer measurements. The model is created by setting conditions according to the physical properties of the sample, and the set conditions include the substrate and film materials, the film thickness of each film layer, and the optical constants of the substrate and film. Each item is usually set using a known reference corresponding to the sample, a required dispersion equation that shows the wavelength dependence of the dielectric constant and has multiple parameters, etc.

Furthermore, a process is carried out to change the parameters of the dispersion equation and the film thickness of each film layer of the model so that the degree of difference between the two is minimized in the above comparison (also called fitting). The difference between the two is usually found by calculation using the least squares method, and when it is determined that the result obtained by the least squares method has become small to a certain extent through fitting, the refractive index and extinction coefficient of the film are found from the parameter values of the dispersion equation at that time, and the film thickness at that time is specified as the film thickness of the film possessed by the sample, thereby making it possible to find the film thickness. Note that model creation and fitting are generally performed manually or automatically using a computer based on the required program.

When unevenness (also called roughness) exists on the sample surface, the concept of effective medium approximation may be used (for example, Patent Document 1, etc.). This method improves the results of least squares calculations by defining roughness and voids as one flat layer. Effective medium approximation may be applied not only to cases where roughness exists on the film surface of a sample, but also to interface layers when roughness exists at the interface between the substrate and film or at the interface between film layers. Furthermore, effective medium approximation may be used as a technique for analysis to reduce the value of the refractive index, regardless of the presence of roughness. Naturally, the use of effective medium approximation changes the results of least squares calculations, and as a result, the film thickness value also changes, so the operator must determine whether to use effective medium approximation, for example, from the results of least squares calculations.

Patent Document 2 describes that the thickness of a native oxide film on a silicon wafer obtained by an ellipsometer varies depending on the surface roughness. Specifically, the rougher the surface, the thicker the film thickness, and a method is disclosed for quantitatively evaluating the surface roughness from the correlation between roughness and the thickness of the native oxide film. In addition, an atomic force microscope (AFM) is known as a method for evaluating the surface roughness on a silicon substrate. Arithmetic mean heights such as Ra and Sa are often used as roughness indices. Ra is the arithmetic mean height over a reference length and is a two-dimensional roughness index, while Sa is a parameter that expands Ra to a surface and is a three-dimensional roughness index.

As a method for evaluating roughness in more detail, conversion to the spatial frequency domain by spectral analysis can also be performed. This method can extract specific wavelength components from the measured surface profile, and can be expressed, for example, as a specific spatial wavelength and a parameter related to the amplitude intensity at that wavelength, such as PSD (Power Spectrum Density). By performing PSD analysis in this way, the spatial frequency of the dominant roughness can be identified. In addition, the haze value obtained by a particle counter can be used as an index of roughness. Haze is expressed as a cloudiness, and is widely used as an index of roughness on silicon surfaces. A high haze level indicates that the wafer surface is rough.

A dense silicon oxide film with high insulating properties is produced by thermally oxidizing silicon wafers. However, because silicon wafers shipped with the silicon wafers have a natural oxide film formed during cleaning to prevent particle adhesion, thermal oxidation is often performed on silicon wafers on which a natural oxide film has already been formed. In this case, it is known that the thickness of the thermally oxidized film is affected by the film quality (thickness and structure) of the natural oxide film before thermal oxidation (Patent Document 3).

In recent years, with the miniaturization and multi-layering of semiconductor integrated circuits, there is a demand for further thinning of various films, including the insulating films that make up the elements. This thinning requires the formation of extremely thin insulating films, i.e. silicon oxide films, that are uniform and reproducible within a surface or between substrates. To achieve this, it is necessary to control the quality of the natural oxide film at the time of shipment of silicon wafers, particularly the film thickness, which affects the quality of the silicon oxide film. In general, if the natural oxide film is thick, the thermal oxide film will also be thick. If you want to make the thermal oxide film thinner, it is better to have a thin natural oxide film, and if you want to make the thermal oxide film thicker, it is better to have a thick natural oxide film. Therefore, in recent years, there has been a particular demand for reproducible control of the thickness of the natural oxide film within a certain range.

Patent Document 3 describes the relationship between silicon wafers cleaned under various conditions and the oxide film thickness after thermal oxidation. Specifically, it discloses that increasing the NH ₄ OH concentration in the SC1 cleaning solution increases the amount of OH groups contained in the native oxide film, resulting in a thicker film after thermal oxidation, and that a method of controlling the film thickness after thermal oxidation is disclosed by using the correlation between the composition (film quality) of the native oxide film and the film thickness after thermal oxidation.

JP 2005-283502 A Japanese Patent Application Publication No. 6-163662 Patent No. 6791453

As mentioned above, there is a need to control the thickness of the native oxide film and thermal oxide film on silicon substrates. Generally, in the manufacturing process of silicon substrates, the surface roughness of the substrate is formed by polishing and subsequent cleaning. SC1 cleaning, hydrofluoric acid cleaning, and ozone water cleaning are used to clean the substrate after polishing, and it is known that the surface is mainly roughened by SC1 cleaning, which has an etching effect, during the cleaning process.

Patent document 3 describes the surface roughness Ra after SC1 cleaning or ozone water cleaning, which is approximately 0.06 to 0.12. This Ra value is the roughness value of silicon substrates that are used in recent years.

Patent document 2 discloses that the surface roughness of the substrate affects the thickness of the native oxide film measured by an ellipsometer, but the surface roughness value in this case is 0.22 to 2.05 nm in terms of Ra value measured by AFM, which is very high compared to the above-mentioned surface roughness Ra value of 0.06 to 0.12 nm.

In addition, the thickness of a natural oxide film is generally known to be about 1 nm, but in Patent Document 2, when the Ra value is 0.22 nm, the thickness of the natural oxide film is 0.097 nm, when the Ra value is 1.23 nm, the thickness of the natural oxide film is 1.586 nm, and when the Ra value is 2.05 nm, the thickness of the natural oxide film is 3.313 nm, all of which are far from about 1 nm. In this way, the surface roughness and thickness of the natural oxide film described in Patent Document 2 are significantly different from the surface roughness and thickness of silicon substrates used in recent years. The reason for this is thought to be that the invention described in Patent Document 2 intentionally roughens the surface using a mixture of hydrofluoric acid and nitric acid, which is not used in normal silicon substrate cleaning solutions. In other words, it is difficult to discuss the roughness and thickness of the natural oxide film in the range of Ra of 0.06 to 0.12 nm, for example, using the correlation disclosed in Patent Document 2, and it is presumed that it can be applied to very rough surfaces, such as those with an Ra value of more than 1 nm.

Here, when focusing on the AFM Ra value and the thermal oxide film thickness obtained by spectroscopic ellipsometry when the NH ₄ OH concentration in SC1 cleaning is varied as described in Patent Document 3, it is found that the higher the NH ₄ OH concentration, the higher the AFM Ra value and the thicker the thermal oxide film tends to be (Figure 9 of Patent Document 3). Patent Document 3 discloses that the composition (film quality) of the natural oxide film (chemical oxide film) formed in the cleaning process, for example, the amount of OH groups measured by the ATR (Attenuated Total Reflectance)-FT (Fourier Transform)-IR (Infrared Spectroscopy) method, is correlated with the thermal oxide film thickness, and describes that the higher the NH ₄ OH concentration, the greater the amount of OH groups, resulting in a thicker thermal oxide film.

However, as mentioned above, the results can also be interpreted as suggesting a correlation between the Ra value obtained by AFM measurement and the film thickness after thermal oxidation obtained by an ellipsometer. Thus, the present inventors have found that there are factors that affect the film thickness of the oxide film, such as the surface roughness of the substrate formed in the manufacturing process of silicon substrates used in recent years (hereinafter simply referred to as "film thickness influencing factors"), and that these factors may affect the thickness of the natural oxide film and thermal oxide film obtained by an ellipsometer, but there are no publicly known documents that describe the film thickness influencing factors. If the surface roughness of the substrate formed in the manufacturing process of silicon substrates, such as an Ra value of 0.06 to 0.12 nm, is one of the factors that affect the oxide film thickness, it can be considered to be an important quality in controlling the film thickness of the natural oxide film and thermal oxide film, just like the composition (film quality) of the natural oxide film described above.

The present invention has been made to solve the above problems, and aims to provide a method for evaluating the thickness of an oxide film by determining whether the thickness value of a silicon substrate being evaluated measured by an ellipsometer is a thickness value that includes the influence of previously unknown factors affecting the thickness, such as the surface roughness of the silicon substrate formed during the manufacturing process of the silicon substrate, and by which the thickness caused by the factors affecting the thickness is evaluated with high accuracy.

The present invention has been made to achieve the above object, and provides a method for evaluating an oxide film on a silicon substrate having a surface roughness Sa value of 0.5 nm or less as measured by an atomic force microscope, the method comprising: a silicon substrate to be evaluated on which an oxide film to be evaluated is formed; and a reference silicon substrate having a thickness of ³ or less; a first film thickness measurement step of measuring a film thickness of an oxide film on the silicon substrate to be evaluated by an ellipsometer; an oxide film removal step of completely removing the oxide films on the silicon substrate to be evaluated and the reference silicon substrate; an oxide film formation step of forming an oxide film under the same conditions on the silicon substrate to be evaluated and the reference silicon substrate after the oxide film has been removed; a second film thickness measurement step of measuring, by an ellipsometer, the film thicknesses of the oxide film on the silicon substrate to be evaluated and the oxide film on the reference silicon substrate after the oxide film formation step; and a film thickness evaluation step of evaluating a film thickness caused by a film thickness influencing factor in the overall film thickness of the oxide film on the silicon substrate to be evaluated obtained in the first film thickness measurement step, based on the film thicknesses of the oxide film on the silicon substrate to be evaluated and the film thicknesses of the oxide film on the reference silicon substrate obtained in the second film thickness measurement step.

With this method for evaluating the thickness of an oxide film on a silicon substrate, it is possible to determine whether the thickness value of the silicon substrate being evaluated measured with an ellipsometer is a thickness value that includes the influence of factors that affect thickness, such as the surface roughness of the silicon substrate, and to accurately evaluate the thickness of the film caused by factors that affect thickness.

In this case, the oxide film removal process can be an oxide film evaluation method in which the oxide film is removed by cleaning with hydrofluoric acid.

Hydrofluoric acid can etch only the native oxide film on a silicon substrate, so it can stably and completely remove only the native oxide film, and the surface roughness does not change significantly before and after cleaning with hydrofluoric acid, making it particularly effective for evaluating the effects of surface roughness.

In this case, the method can be a method for evaluating the thickness of an oxide film formed by cleaning with ozone water or hydrogen peroxide water in the oxide film formation process.

Ozone water and hydrogen peroxide water have a strong oxidizing effect and can stably form a natural oxide film of about 1 nm, making it possible to more stably evaluate film thickness caused by factors affecting film thickness.

In this case, the method can be such that the thickness of the oxide film to be evaluated on the silicon substrate to be evaluated is 25 nm or less.

When the film thickness is 25 nm or less, the effect of the surface roughness of the silicon substrate becomes more pronounced, so the present invention can perform more accurate evaluations when the film thickness is thinner.

In this case, the film thickness influencing factors in the film thickness of the oxide film include the surface roughness of the silicon substrate being evaluated, and in the film thickness evaluation process, when the difference film thickness obtained by subtracting the film thickness of the oxide film on the reference silicon substrate obtained in the second film thickness measurement process from the film thickness of the oxide film on the silicon substrate being evaluated obtained in the second film thickness measurement process is 0.02 nm or more and 0.20 nm or less, the method can be used to determine that the film thickness of the oxide film on the silicon substrate being evaluated obtained in the first film thickness measurement process includes the film thickness of the oxide film caused by the surface roughness of the silicon substrate being evaluated.

This type of evaluation method makes it possible to more accurately determine whether the film thickness value obtained by the ellipsometer is affected by surface roughness.

At this time, in the film thickness evaluation process, the difference film thickness is subtracted from the film thickness of the oxide film on the silicon substrate to be evaluated obtained in the first film thickness measurement process, thereby making it possible to provide an oxide film film thickness evaluation method that evaluates the film thickness of the oxide film excluding the influence of the surface roughness of the silicon substrate to be evaluated.

This type of evaluation method makes it possible to evaluate the thickness of the oxide film without taking into account the effects of surface roughness.

In this case, the surface roughness can be evaluated as a thickness of an oxide film having a roughness component with a spatial frequency of 60 to 90/μm.

This allows the effect of surface roughness to be evaluated with greater precision, because roughness components with spatial frequencies of 60-90/um affect the oxide film thickness.

In this case, the surface roughness can be evaluated as a method for evaluating the thickness of the oxide film, which is the roughness component formed by SC1 cleaning.

In this way, the film thickness evaluation method according to the present invention allows for more accurate evaluation of the effects of surface roughness caused by a specific cleaning process. This is because roughness with a spatial frequency of 60 to 90/μm is formed particularly by SC1 cleaning.

In this case, cleaning conditions and/or oxidation conditions before the formation of an oxide film on a silicon substrate are set based on the film thickness caused by the film thickness influencing factors evaluated by the oxide film thickness evaluation method, and the silicon substrate is cleaned and an oxide film is formed on the silicon substrate using the cleaning conditions and/or oxidation conditions, thereby manufacturing a silicon substrate with an oxide film.

This makes it possible to control the thickness of the oxide film with greater precision and produce silicon substrates with an oxide film.

The present invention also provides a method for evaluating the thickness of an oxide film formed on a silicon substrate, the method including a surface roughness of the silicon substrate as a thickness-influencing factor in the thickness of the oxide film, and determining that the thickness of the oxide film formed on the silicon substrate includes a thickness of the oxide film caused by the surface roughness of the silicon substrate, which is a thickness-influencing ^factor , when an average value of the power spectral density at a spatial frequency of 60 to 90/μm on the surface of the silicon substrate is 0.15 nm3 or more.

The oxide film thickness evaluation method of the present invention makes it possible to quickly and easily determine whether the oxide film thickness value measured by an ellipsometer is affected by the surface roughness of the substrate.

As described above, the oxide film thickness evaluation method according to the present invention can determine whether the film thickness value of the silicon substrate being evaluated measured by the ellipsometer is a film thickness value that includes the influence of previously unknown film thickness influence factors, such as the surface roughness of the silicon substrate formed in the silicon substrate manufacturing process, and can accurately evaluate the film thickness caused by the film thickness influence factors. In addition, by calculating the film thickness value caused by the surface roughness, etc., it is possible to analyze the film thickness of the natural oxide film and the film thickness after thermal oxidation in more detail. By utilizing the results of this analysis, the oxide film thickness on the silicon substrate can be controlled with high precision.

1 is a flowchart showing an example of a method for evaluating a thickness of an oxide film on a silicon substrate according to the present invention. 1 shows a flowchart relating to an investigation of the relationship between the surface roughness of a silicon substrate and the thickness of an oxide film. 3 is a graph showing the relationship between the surface roughness (haze) of the silicon substrate, which has been roughened by CMP and SC1 cleaning, and the thickness of the native oxide film and the thickness of the 5 nm oxide film. 3 is a graph showing the relationship between the surface roughness (haze) of the silicon substrate subjected to the roughening treatment of FIG. 2 by single-wafer cleaning and the thickness of the native oxide film and the 5 nm oxide film. 2 is a graph showing the relationship between haze and native oxide film after cleaning in the roughening treatment of FIG. 2 with a solution composition of NH ₄ OH:H ₂ O ₂ :H ₂ O=1:1:10, a cleaning temperature of 80° C., and cleaning times of 0, 3, 6, and 12 minutes. The AFM measurement results and PSD curves of each sample are shown. An example of the evaluation and analysis method according to the present invention will be shown when the thickness of the native oxide film is changed by changing the SC1 cleaning.

The present invention is described in detail below, but is not limited to these.

As described above, there was a need for a method for evaluating the thickness of an oxide film on a silicon substrate measured with an ellipsometer to determine whether the thickness value includes the effects of previously unknown factors affecting the thickness, such as surface roughness formed during the manufacturing process of the silicon substrate, and to accurately evaluate the thickness of the oxide film caused by the factors affecting the thickness.

As a result of intensive research by the inventors into the above-mentioned problem, they have discovered a method for evaluating an oxide film on a silicon substrate having a surface roughness Sa value of 0.5 nm or less as measured by an atomic force microscope, comprising a substrate preparation step of preparing a silicon substrate to be evaluated on which an oxide film to be evaluated is formed and a reference silicon substrate having an average value of power spectrum density of 0.1 nm3 or less at a spatial frequency of 60 to 90/ ^μm on the substrate surface, a first film thickness measurement step of measuring the film thickness of the oxide film on the silicon substrate to be evaluated by an ellipsometer, an oxide film removal step of completely removing the oxide films on the silicon substrate to be evaluated and the reference silicon substrate, an oxide film formation step of forming an oxide film under the same conditions on the silicon substrate to be evaluated and the reference silicon substrate after the oxide film removal step, a second film thickness measurement step of measuring the film thicknesses of the oxide film on the silicon substrate to be evaluated and the oxide film on the reference silicon substrate after the oxide film formation step by an ellipsometer, and The inventors have found that a method for evaluating an oxide film thickness comprising: a film thickness evaluation step of evaluating a film thickness caused by film thickness influencing factors in the overall film thickness of the oxide film on the silicon substrate to be evaluated, obtained in the first film thickness measurement step, based on the film thickness of the oxide film on the silicon substrate to be evaluated and the film thickness of the oxide film on the reference silicon substrate, obtained in the film thickness measurement step, makes it possible to determine whether the film thickness value of the silicon substrate to be evaluated, measured by an ellipsometer, is a film thickness value that includes the influence of film thickness influencing factors such as the surface roughness of the silicon substrate, and to accurately evaluate the film thickness caused by the film thickness influencing factors, thereby completing the present invention.

The present inventors have also found that a method for evaluating the thickness of an oxide film formed on a silicon substrate, the method including the surface roughness of the silicon substrate as a thickness-influencing factor in the thickness of the oxide film, and determining that the thickness of the oxide film formed on the silicon substrate includes the thickness of the oxide film caused by the surface roughness of the silicon substrate, which is a thickness-influencing factor, when the average value of ^the power spectrum density at the spatial frequency of the surface of the silicon substrate is 0.15 nm3 or more, enables a quick and extremely easy determination of whether or not the thickness value of the oxide film measured by an ellipsometer is affected by the roughness of the substrate surface, and completed the present invention.

The following explanation will be given with reference to the drawings.

The inventors conducted extensive research into factors that affect oxide film thickness. As a result, they discovered that the roughness formed during the wafer manufacturing process, specifically the roughness formed during the polishing and cleaning processes, and the quality of the oxide film affect the thickness of native oxide films and thermal oxide films measured with an ellipsometer. In particular, they discovered that with respect to the surface roughness of the substrate, when the average value of the power spectrum density (intensity) in a certain frequency band exceeds a certain value, the thickness of the oxide film increases.

First, we will discuss the relationship between the various surface roughnesses formed during the manufacturing process of silicon substrates and the thickness of the oxide film. Figure 2 is the investigation flow chart. The silicon substrates prepared were subjected to roughening treatment to form roughness by changing the CMP processing conditions and SC1 cleaning conditions, and multiple levels of silicon substrates were prepared. Next, the oxide film was completely removed by hydrofluoric acid cleaning in a batch cleaning machine, and then a natural oxide film was formed by cleaning with ozone water. Since the oxide film formed by the roughening treatment was completely removed by hydrofluoric acid cleaning in both levels, and the oxide film was formed by subsequent ozone water cleaning, it can be interpreted that the oxide film was formed under the same conditions on multiple levels of silicon substrates. After that, haze measurements were performed using a particle counter, and some of the silicon substrates were thermally oxidized with a target thickness of 5 nm, and the thickness of the natural oxide film and the oxide film formed with a target thickness of 5 nm were evaluated by spectroscopic ellipsometry.

Figure 3 is a graph showing the relationship between the surface roughness (haze) of silicon substrates subjected to the roughening treatment of Figure 2 using CMP and SC1 cleaning, and the thickness of the native oxide film and the oxide film aimed for a thickness of 5 nm. At the CMP level (■), the thicknesses of the native oxide film and the oxide film aimed for a thickness of 5 nm were the same even when the haze value exceeded 10 ppm, but at the SC1 cleaning level (●), there was a tendency for both the native oxide film and the oxide film aimed for a thickness of 5 nm to become thicker as the haze increased. Since the oxide film was formed under the same conditions, it was assumed that the film thickness would be the same as with the CMP level, but this was not the case with the SC1 cleaning level.

Furthermore, Figure 4 shows the relationship between the surface roughness (haze) of the silicon substrate, which was subjected to the roughening treatment shown in Figure 2 by single-wafer cleaning using a combination of hydrofluoric acid and ozone water cleaning, and the thickness of the native oxide film and the oxide film aimed at 5 nm. In addition, the hydrofluoric acid cleaning and ozone water cleaning after the roughening treatment were also performed by the single-wafer method, not the batch method. In this case, since the oxide film formation method is different from the batch method of ozone water cleaning, it is not possible to compare the film thickness of the above-mentioned SC1 and CMP levels with the single-wafer cleaning level, but the influence of haze within the single-wafer cleaning level can be discussed. As a result, the single-wafer cleaning level was the same in the thickness of the native oxide film and the oxide film aimed at 5 nm even if the haze changed, just like the CMP level. To summarize the above results, it was newly discovered that the roughness formed by CMP and single-wafer cleaning does not affect the thickness of the oxide film, and the surface roughness of the substrate formed by SC1 cleaning has an effect of making the oxide film thicker.

Here, the results of an additional investigation into the SC1 cleaning level will be described. SC1 cleaning for roughening treatment was performed as a batch cleaning with a liquid composition of _NH4OH : _H2O2 : _H2O = 1:1:10, a cleaning temperature of 80°C, and cleaning times of 3 _, 6, and 12 minutes, the oxide film was completely removed by hydrofluoric acid cleaning, and then cleaning with ozone water was performed. The results show the thickness and haze value of the native oxide film after that, as shown in Figure 5. The thickness of the native oxide film without cleaning time (set as 0 minutes) was 1.207 nm, which is the reference, but with a cleaning time of 3 minutes it was 1.258 nm, with a cleaning time of 6 minutes it was 1.258 nm, and with a cleaning time of 12 minutes it was 1.261 nm. Since the difference between the average film thickness of 1.259 nm for cleaning times of 3, 6, and 12 min and the film thickness of 1.207 nm for no cleaning (cleaning time 0 min) is 0.052 nm, the amount of film thickening caused by SC1 cleaning in the roughening treatment is considered to be about 0.052 nm. Considering that haze tends to increase as the cleaning time is longer, while the amount of film thickening is the same for cleaning times of 3, 6, and 12 min, it is considered that a specific roughness component formed by SC1 cleaning affects the film thickening behavior, and the film thickness is the same for 3, 6, and 12 min because the roughness components involved in the film thickness (thickening) are the same.

To verify these results, the surface roughness of substrates at typical CMP, SC1 cleaning, and single-wafer cleaning levels was evaluated using an AFM (atomic force microscope). The observation field was 1 μm x 1 μm, and in addition to the three-dimensional calculated average height Sa, a PSD curve was obtained from the spectral analysis of the surface profile data. Figure 6 shows the AFM measurement results and PSD curves for each sample.

First, we focus on the CMP level. In terms of CMP level, we evaluated the level with the smallest haze value (CMP-1) and the highest level (CMP-2) in Figure 3. In CMP-2, the power spectrum density (intensity) in the low frequency band (1-10/μm) was very high, and roughness on the low frequency side was predominant. The AFM image also showed a large undulating image, which was consistent. As mentioned above, the roughness created by CMP does not affect the oxide film thickness, so it can be said that this low frequency component does not affect the oxide film thickness.

Next, we focus on the SC1 cleaning level. For the SC1 cleaning level, three levels were evaluated: the solution composition NH ₄ OH:H ₂ O ₂ :H ₂ O = 1:1:10, and the cleaning temperature and cleaning time were 60°C/3 min, 80°C/3 min, and 80°C/12 min. These are the levels at which an oxide film formed under the same conditions would be thicker by the same thickness. As shown in Figure 6, the roughness in the high frequency band (10 to 100/μm) is dominant in all three levels compared to CMP-2, which coincides with the fine granular roughness obtained in the AFM image. Therefore, it can be said that the roughness components (spatial frequency band) formed by CMP and SC1 cleaning are significantly different.

Finally, we focus on the single-wafer cleaning level. In the single-wafer cleaning level, the level with the smallest haze value (single-wafer cleaning-1) and the level with the largest haze value (single-wafer cleaning-2) in Figure 4 were evaluated. The power spectrum density (intensity) of the single-wafer cleaning level was approximately halfway between CMP and SC1 cleaning in the high frequency band (10-100/μm).

We will now consider these roughness evaluation results and their effect on the oxide film thickness. In particular, while the Sa value for both SC1 cleaning at 80°C/3 min and single-wafer cleaning at 2 is 0.108 nm, the oxide film thickness is thicker with SC1 cleaning at 80°C/3 min, but not with single-wafer cleaning at 2. Looking at the PSD curves for both, the power spectrum density (intensity) in the low frequency band (1-10/μm) is the same, whereas the power spectrum density (intensity) in the high frequency band (especially 50/μm or higher) is greater with SC1 cleaning at 80°C/3 min than with single-wafer cleaning at 2. Therefore, although there is no significant difference between the two in the AFM images, the PSD curves indicate that roughness in the high frequency band is more dominant with SC1 cleaning at 80°C/3 min. Furthermore, considering that the three levels of SC1 cleaning shown in the figure (60°C/3 min, 80°C/3 min, 80°C/12 min) all result in the same thickness (approximately 0.05 nm), it can be seen that the power spectrum density (intensity) in the spatial frequency band of 60-90/μm is the same for all three levels, and is higher than that of single wafer cleaning-2. Therefore, it has become clear that the roughness component in the 60-90/μm range is the roughness component that affects the thickness of the native oxide film and the 5 nm oxide film.

In addition, the power spectrum density (intensity) of the spatial frequency band of 50/μm or less is within the three levels of SC1 cleaning.
80℃/12min > 80℃/3min > 60℃/3min
The relationship between the magnitudes of the Sa values is as follows:
80℃/12min > 80℃/3min > 60℃/3min
The results also coincide with those of the comparative example 1, and it is believed that the Sa value in this case is dominated by roughness information on the low frequency side, where intensity is high, which is the reason for the difference in oxide film thickness, despite the fact that the above-mentioned single wafer cleaning-2 and SC1 cleaning-80°C/3 min have the same Sa value of 0.108 nm.

To summarize the above results, it is considered that the oxide film will become thick when the average value of the power spectrum density (intensity), which is the roughness component in the spatial frequency band of 60 to 90/μm, is equal to or greater than the threshold value. Here, the average values of the power spectrum density (intensity) in the range of 60 to 90/μm for the three levels of SC1 cleaning were calculated to be 0.16 nm ³ for SC1 cleaning at -60°C/3 min, 0.18 nm ³ for SC1 cleaning at -80°C/3 min, and 0.17 nm ³ for SC1 cleaning at -80°C/12 min. On the other hand, the average value for the single-wafer cleaning-2, where the oxide film does not become thick, was 0.11 nm ^3. Therefore, it is considered that the average value of the power spectrum density (intensity) of 60 to 90/μm, 0.15 nm ^{3, is} the threshold value, and it can be determined that the thickness of the silicon oxide film on the silicon substrate with the average value of the power spectrum density (intensity) of 0.15 nm 3 or more is influenced by the surface roughness of the substrate, and includes the film thickness caused by the surface roughness of the substrate.

From the above findings, it can be interpreted that a specific surface roughness is one of the factors that affect the oxide film thickness (film thickness influencing factors). Here, the film thickness influencing factors of the natural oxide film and the thermal oxide film include, for example, the structure of the natural oxide film (chemical oxide film) described in Patent Document 3 and the surface roughness of the silicon substrate described above. Therefore, it is considered that the film thickness of the natural oxide film and the thermal oxide film can be controlled by adjusting the structure of the natural oxide film (chemical oxide film) and the surface roughness of the substrate. Furthermore, if it is desired to change the cleaning conditions to thicken or thin the natural oxide film and the thermal oxide film, if it is possible to analyze which factor, the structure of the natural oxide film or the surface roughness of the substrate, is more strongly influenced by the film thickness of the silicon substrate after cleaning, this will be useful knowledge in determining the cleaning conditions. The oxide film thickness evaluation method according to the present invention is very effective in such cases, and an example showing its effectiveness will be specifically described below.

Here, as an example, an example of controlling the thickness of an oxide film by varying the SC1 cleaning conditions is shown. SC1 cleaning was performed on a silicon substrate as shown in FIG. 7A. The cleaning conditions were a liquid composition of NH ₄ OH:H ₂ O ₂ :H ₂ O=1:1:10, a cleaning time of 3 min, and three cleaning temperatures of 45, 60, and 80° C. A plurality of substrates were processed for each level. Next, half of the wafers at each level were cleaned with hydrofluoric acid to remove the native oxide film, and then cleaned with ozone water to form an oxide film under the same conditions. The thickness of the native oxide film was then evaluated by spectroscopic ellipsometry. As a result, as shown in FIG. 7B, the higher the cleaning temperature, the thicker the film after SC1 cleaning (■). Here, the factors that cause the film thickness to become thicker at higher cleaning temperatures after SC1 cleaning (■) are analyzed in more detail.

First, the effect of the surface roughness of the silicon substrate will be described. The film thickness after HF→O ₃ cleaning (●) was 1.207 nm without SC1 cleaning, 1.206 nm at a cleaning temperature of 45°C, 1.258 nm at a cleaning temperature of 60°C, and 1.258 nm at a cleaning temperature of 80°C. 1.207 without SC1 cleaning and 1.206 nm at a cleaning temperature of 45°C can be interpreted as equivalent film thickness. In contrast, at cleaning temperatures of 60 and 80°C, the film thickness due to surface roughness is interpreted as being thicker. When the amount of film thickening due to surface roughness was calculated, the difference between 1.258 nm at a cleaning temperature of 60°C and 1.207 nm without SC1 cleaning was 0.051 nm, and the difference between 1.258 nm at a cleaning temperature of 80°C and 1.207 nm without SC1 cleaning was 0.051 nm. This 0.051 nm is the amount of film thickening due to roughness.

Next, the film thickness excluding the effects of surface roughness (▲) is calculated by subtracting the amount of thickening due to surface roughness, 0.051 nm, from the film thickness after SC1 cleaning (■) at cleaning temperatures of 60°C and 80°C. There is no effect of roughness at a cleaning temperature of 45°C, so it does not need to be taken into consideration. The results were 1.058 nm at a cleaning temperature of 60°C and 1.072 nm at a cleaning temperature of 80°C. Therefore, since the film thickness is thicker at the higher cleaning temperature of 80°C than at 60°C even after removing the effects of surface roughness, this difference can be interpreted as being due to the structure of the native oxide film.

If we interpret the difference in thickness between cleaning temperatures of 60°C and 80°C and a cleaning temperature of 45°C after removing the thickness affected by surface roughness as the thickness caused by the structure of the native oxide film, as shown in Figure 7(C), we can distinguish between the roughness-caused and structure-caused film thickening amounts for cleaning temperatures of 60°C and 80°C and an SC1 cleaning temperature of 45°C. Since the roughness-caused film thickening amounts for cleaning temperatures of 60°C and 80°C are equivalent, it has been newly discovered that when the cleaning condition is set to a cleaning temperature of 80°C, the thickness caused by the structure of the native oxide film becomes thicker, resulting in a thicker overall film thickness after cleaning. This is an example in which the roughness-caused film thickness is saturated, but depending on the conditions, it can be estimated that the roughness and structure-caused film thickening amounts are not saturated. In such cases, if it is required to manufacture oxide films more stably and with greater control over their thickness by understanding the factors that cause them to thicken using this evaluation method, saturating the amount of thickening caused by roughness, such as at SC1 cleaning temperatures of 60°C or 80°C, as exemplified here, will result in the only factor affecting the film thickness being the structure of the oxide film, making it possible to manufacture films with less variation in film thickness. For these reasons, a method for evaluating the film thickness of an oxide film on a silicon substrate that takes into account the surface roughness of the substrate is extremely useful.

First Embodiment
Based on the above, a method for evaluating the thickness of an oxide film on a silicon substrate according to a first embodiment of the present invention will be described in detail below. 1 is a flow chart showing an example of a method for evaluating the thickness of an oxide film on a silicon substrate according to the present invention.

[Method for evaluating oxide film thickness]
The first embodiment of the method for evaluating the thickness of an oxide film according to the present invention is a method for evaluating an oxide film on a silicon substrate having a surface roughness Sa value of 0.5 nm or less as measured by an atomic force microscope, the method including: forming an oxide film to be evaluated on a silicon substrate; and determining whether the average value of the power spectrum density at a spatial frequency of 60 to 90/μm on the substrate surface is 0.1 nm or less. The method for evaluating the thickness of an oxide film includes a substrate preparation step of preparing a silicon substrate to be evaluated and a reference silicon substrate having a thickness of ³ or less, a first thickness measurement step of measuring the thickness of an oxide film on the silicon substrate to be evaluated by an ellipsometer, an oxide film removal step of completely removing the oxide film on the silicon substrate to be evaluated and the reference silicon substrate, an oxide film formation step of forming an oxide film under the same conditions on the silicon substrate to be evaluated and the reference silicon substrate after the oxide film removal step, a second thickness measurement step of measuring the thicknesses of the oxide film on the silicon substrate to be evaluated and the oxide film on the reference silicon substrate after the oxide film formation step by an ellipsometer, and a thickness evaluation step of evaluating the thickness of the oxide film on the silicon substrate to be evaluated obtained in the first thickness measurement step due to a thickness influencing factor in the overall thickness of the oxide film on the silicon substrate to be evaluated obtained in the second thickness measurement step, based on the thicknesses of the oxide film on the silicon substrate to be evaluated and the oxide film on the reference silicon substrate obtained in the second thickness measurement step. Each step will be described in detail below.

(Substrate preparation process)
First, as shown in S1 of FIG. 1, a silicon substrate to be evaluated and a silicon substrate to be used as a reference (hereinafter referred to as a "reference silicon substrate") are prepared. There are no limitations on the conductivity type, diameter, or sample shape of either. There are no limitations on the type of silicon oxide film on the silicon substrate to be evaluated, and examples of such films include natural oxide films, thermal oxide films, and CVD oxide films. The thickness of the oxide film to be evaluated on the silicon substrate to be evaluated is not particularly limited, but is preferably 25 nm or less. When the film thickness is 25 nm or less, the effect of the surface roughness of the silicon substrate becomes more significant, so in the present invention, evaluation can be performed with higher accuracy when the film thickness is thinner.

However, there are restrictions on the surface roughness. First, the surface roughness of the silicon substrate to be evaluated must be 0.5 nm or less in terms of Sa value. The reason for this is that if the Sa value is greater than 0.5 nm, the thickness value of the natural oxide film will deviate significantly from 1 nm as in Patent Document 2. Specifically, the composition of the SC1 cleaning solution was diluted more than usual to NH ₄ OH:H ₂ O ₂ :H ₂ O = 1:1:1000, and two levels of silicon substrates were prepared in which the surface was intentionally roughened with a chemical solution that was made etching-dominant. When the roughness Sa was evaluated by AFM and the natural oxide film was evaluated by spectroscopic ellipsometry, the film thickness was 3.34 nm at a Sa value of 0.65 nm and 5.12 nm at a Sa value of 1.05 nm. If there is roughness with a Sa value of more than 0.5 nm, the oxide film thickness calculated by spectroscopic ellipsometry will deviate significantly from about 1 nm, so the Sa value must be 0.5 nm or less. In addition, as a general polishing process for silicon wafers, the front side is subjected to CMP (single-sided polishing) after DSP (double-sided polishing), and cleaning is performed after each polishing process. Generally, the Sa value of the front side of a silicon wafer after CMP is 0.1 nm or less, and the Sa value of the back side (DSP surface) is about 0.2 to 0.4 nm, so that a surface roughness with an Sa value exceeding 0.5 nm is not formed in a normal silicon substrate manufacturing process. At least, a wafer after CMP processing following DSP processing can be used as the silicon substrate to be evaluated in the present invention. In this way, the present invention can be applied to a silicon substrate to be evaluated having a surface roughness with an Sa value of 0.5 nm or less.

The reference silicon substrate must have an Sa value of 0.5 nm or less and an average value of the power spectrum density (intensity) in the spatial frequency range of 60 to 90/μm of 0.1 nm3 ^or less. If a silicon substrate having an average value of the power spectrum density (intensity) in the spatial frequency range of 60 to 90/μm of more than 0.1 ^nm3 is used as the reference silicon substrate, it becomes difficult to separate the thickening of the oxide film caused by the surface roughness. As described above, since the roughness component of 60 to 90/μm is formed by SC1 cleaning, not CMP or single-wafer cleaning, when preparing a reference silicon substrate, it is necessary to restrict the cleaning conditions after polishing. Usually, a large amount of abrasive particles adheres after polishing, and a cleaning process is essential. In this case, the average value of the power spectrum density (intensity) in the range of 60 to 90/μm can be made 0.1 nm3 or ^less by cleaning with a combination of hydrofluoric acid and ozone water in a spin cleaning machine (single-wafer cleaning) or cleaning with a combination of hydrofluoric acid and ozone water in a batch system. Furthermore, when SC1 cleaning is performed, in the case of a liquid composition of NH ₄ OH:H ₂ O ₂ :H ₂ O=1:1:10, the average value of the power spectrum density (intensity) can be reduced to 0.1 ^nm3 or less by setting the cleaning temperature to 45°C or less. Cleaning using a combination of hydrofluoric acid, ozone water cleaning, and SC1 may also be performed. A reference silicon substrate can be prepared using this flow. At this time, it is more preferable to perform AFM measurement of the prepared silicon substrate and confirm that the average value of the power spectrum density (intensity) in the range of 60 to 90/μm is 0.1 ^nm3 or less.

(First film thickness measurement process)
Next, as shown in S2 of FIG. 1, the oxide film thickness of the silicon substrate to be evaluated is measured by an ellipsometer. As described above, there are two types of ellipsometers: a single-wavelength type that uses a laser as a light source, and a spectral type that uses a white light source containing many wavelength components. There is no restriction on the light source, but it is more preferable to use a spectral type that has a large amount of information that can accurately evaluate the film thickness. In addition, the measurement method of the ellipsometer can be performed by a known method. In order to obtain the film thickness, it is necessary to create a model according to the substrate sample, and this model is created by setting conditions according to the physical properties of the sample, and the items of the conditions that are set include the material of the substrate and film, the film thickness of each film layer, and the optical constants of the substrate and film. In particular, it is necessary to quote known literature values for the optical constants, and in this case, there are multiple reported examples of literature values. Therefore, in the present invention, it is more preferable to make the models and cited optical constants of the first film thickness measurement process and the second film thickness measurement process described later the same, and by making them the same, the film thickness can be evaluated more accurately. At this time, the film thickness obtained by the first film thickness measurement may be affected by the surface roughness of the substrate described above. Therefore, using the method described later, it is verified and determined whether the film thickness caused by this surface roughness is included.

(Oxide film removal process)
Next, the effect of the surface roughness of the silicon substrate is investigated. As shown in S3 of FIG. 1, the oxide film of the evaluation target silicon substrate and the reference silicon substrate are completely removed. Although the removal method is not particularly limited, it is more preferable to perform hydrofluoric acid cleaning. As the action of hydrofluoric acid is to etch the oxide film but not to etch Si, only the natural oxide film can be easily and stably removed completely, and the surface roughness does not change significantly before and after hydrofluoric acid cleaning, so it is particularly effective for evaluating the effect of surface roughness. In this case, there is no limit to the concentration of hydrofluoric acid or the cleaning time, and it is sufficient that the oxide film is completely removed. For example, when the oxide film is completely removed, the bare surface is exposed, and it becomes a water-repellent surface. On the other hand, when an oxide film exists on the surface, it becomes a hydrophilic surface, so it is possible to judge whether the oxide film has been removed from the surface state after cleaning. An example of the conditions is, for example, a hydrofluoric acid concentration of 0.3 to 5.0 wt%, a temperature of 10 to 30° C., and a cleaning time of 60 to 360 seconds.

(Oxide film forming process)
Next, as shown in S4 of FIG. 1, an oxide film is formed under the same conditions on the silicon substrate to be evaluated from which the oxide film has been completely removed and on the reference silicon substrate. By forming the oxide film under the same conditions in this way, the influence of the surface roughness of the substrate can be discussed from the results of the second film thickness measurement described later. In this case, there is no particular limitation on the method of forming the oxide film, but it is more preferable to oxidize the substrate surface with ozone water or hydrogen peroxide water. Ozone water and hydrogen peroxide water have a strong oxidizing effect, and can easily and stably form a natural oxide film of about 1 nm, so that the influence on the film thickness caused by the film thickness influence factor can be evaluated more stably. In particular, if a spin cleaning and batch cleaning machine is used, the above-mentioned hydrofluoric acid cleaning and the subsequent ozone water or hydrogen peroxide water cleaning can be performed in one batch, and the labor is reduced. It is more preferable to perform ozone water cleaning, which has a stronger oxidizing effect and is more stable than hydrogen peroxide water. This is because it is preferable that the variation in the thickness of the oxide film between the silicon substrate to be evaluated and the reference silicon substrate in this process is as small as possible.

(Second film thickness measurement process)
Next, as shown in S5 of FIG. 1, the thicknesses of the oxide films formed on the evaluation silicon substrate and the reference silicon substrate under the same conditions are evaluated by an ellipsometer. The same method can be used.

(Film thickness evaluation process)
Next, as shown in S6 of FIG. 1, the film thickness caused by the film thickness influencing factors is evaluated for the entire film thickness of the oxide film on the silicon substrate to be evaluated, which is obtained in the first film thickness measuring step.

First, we will explain the interpretation of the second film thickness measurement results. In the second film thickness measurement, the film thickness of an oxide film formed under the same conditions is evaluated. Therefore, if there is no influence between the substrates (each level) due to factors that affect film thickness, such as surface roughness, the two should have the same film thickness. If the surface roughness of the substrate affects the film thickness, the film thickness of the silicon substrate being evaluated will be thicker than that of the reference silicon substrate. In such cases, it can be determined that the film thickness of the oxide film on the silicon substrate being evaluated obtained in the first process includes the film thickness of the oxide film caused by the specific surface roughness on the silicon substrate being evaluated.

Specifically, when the difference in thickness obtained by subtracting the thickness of the oxide film on the reference silicon substrate obtained in the second film thickness measurement process from the thickness of the oxide film on the silicon substrate to be evaluated obtained in the second film thickness measurement process is 0.02 nm or more and 0.2 nm or less, it is preferable to determine that the thickness of the oxide film on the silicon substrate to be evaluated obtained in the first film thickness measurement process includes the thickness of the oxide film caused by the surface roughness of the silicon substrate to be evaluated. According to the inventors' investigation, when the difference in thickness is 0.02 nm or less, it is possible that the difference is caused by the variation in the ellipsometer measurement, so that the accuracy is improved by setting the lower threshold to 0.02 nm or more. In addition, when multiple silicon substrates with different cleaning conditions were prepared and the difference in thickness obtained from the second film thickness measurement was calculated, the difference was at most 0.2 nm, so that it is practical to set the upper threshold to 0.2 nm. By performing the evaluation in this manner, it is possible to accurately determine whether the film thickness value obtained by the ellipsometer is particularly affected by the surface roughness.

Furthermore, by subtracting the difference in thickness, i.e., the thickness caused by the surface roughness, from the thickness of the oxide film on the silicon substrate to be evaluated obtained in the first thickness measurement process, the thickness of the oxide film excluding the influence of the surface roughness of the silicon substrate to be evaluated can be calculated and evaluated. In particular, when there are multiple levels of silicon substrates to be evaluated, the film thickness variation factors between the levels as shown in Figure 7 can be relatively evaluated.

The surface roughness in the oxide film thickness evaluation method according to the present invention can be a roughness component with a spatial frequency of 60 to 90/μm. Since the roughness component with a spatial frequency of 60 to 90/μm has a stronger effect on the oxide film thickness, the effect of the surface roughness can be evaluated with higher accuracy. The surface roughness can be a roughness component formed by SC1 cleaning. This is because the roughness component with a spatial frequency of 60 to 90/μm is formed particularly by SC1 cleaning, and the effect of the surface roughness caused by a specific cleaning process can be evaluated with higher accuracy.

[Method of manufacturing silicon substrate with oxide film]
A method for manufacturing a silicon substrate with an oxide film is provided, which includes setting cleaning conditions and/or oxidation conditions before the formation of an oxide film on a silicon substrate based on the film thickness caused by the film thickness influencing factors evaluated by the oxide film thickness evaluation method according to the present invention, cleaning the silicon substrate and forming an oxide film on the silicon substrate using the cleaning conditions and/or oxidation conditions, and manufacturing a silicon substrate with an oxide film. According to the oxide film thickness evaluation method according to the present invention, since it is possible to distinguish the amount of thickening caused by the film thickness influencing factors, it is possible to analyze, for example, which factor, the structure of the native oxide film or the substrate surface roughness, is more strongly affected by the thickness of the oxide film on the silicon substrate, and the result of this can be used to adjust the structure of the native oxide film, the surface roughness of the substrate, and/or the oxidation conditions to precisely control the thickness of the oxide film and manufacture a silicon substrate with an oxide film.

Second Embodiment
Next, a method for evaluating the thickness of an oxide film according to a second embodiment of the present invention will be described in detail.

[Method for evaluating oxide film thickness]
A method for evaluating the thickness of an oxide film is provided, ^which determines that the thickness of an oxide film formed on a silicon substrate includes the thickness of the oxide film caused by the surface roughness of the silicon substrate, which is a thickness-influencing factor, when the average value of the power spectrum density in the spatial frequency range of the silicon substrate is 0.15 nm3 or more. As described above, when the average value of the power spectrum density (intensity) in the spatial frequency band of 60 to 90/μm is 0.15 ^nm3 or more, the thickness of the native oxide film measured by an ellipsometer includes the thickness caused by the surface roughness of the silicon substrate. Therefore, by performing AFM measurement of the silicon substrate and acquiring a PSD curve from the spectrum analysis of the surface profile data, it is possible to easily determine whether or not the thickness of the oxide film caused by the surface roughness of the silicon substrate, which is a thickness-influencing factor, is included.

The method for evaluating oxide films on silicon substrates described above makes it possible to evaluate the thickness of oxide films obtained using an ellipsometer while taking into account factors that affect film thickness, such as surface roughness. Furthermore, it is also possible to distinguish between the factors that cause variations in oxide film thickness and the surface roughness of the silicon substrate and the structure of the native oxide film, making it possible to evaluate the thickness of oxide films on silicon substrates with higher accuracy than ever before.

The present invention will be described in detail below with reference to examples, but the present invention is not limited thereto.

Example 1
As the silicon substrate to be evaluated, a silicon wafer was prepared which had been polished by CMP, cleaned by single-wafer cleaning, and then cleaned by SC1 cleaning. The specific conditions are as shown in Table 1.

AFM measurement was performed in advance on the substrate of level 0 with an observation field of view of 1 μm×1 μm, and a PSD curve was obtained from the spectrum analysis of the profile data. Sa was 0.0521 nm, which is less than 0.5 nm, and the average value of the power spectrum density (intensity) of the spatial frequency of 60 to 90/μm was 0.081 ^nm3 , which is less than 0.1 ^nm3 , so this level 0 was used as the reference silicon substrate. Levels 1 to 6 were used as the silicon substrates to be evaluated. Next, the first film thickness measurement of the silicon substrates to be evaluated was performed using a spectroscopic ellipsometer M-2000V manufactured by J. A. Woollam Co., Ltd. As a result, level 1 was 1.033 nm, level 2 was 1.109 nm, level 3 was 1.121 nm, level 4 was 1.169 nm, level 5 was 1.114 nm, and level 6 was 1.109 nm.

Next, the oxide film was removed from a total of seven levels of substrates, six levels of silicon substrates to be evaluated and one level of reference silicon substrate, by cleaning with 0.5 wt% hydrofluoric acid at a cleaning temperature of 25°C for a cleaning time of 3 minutes. Since the surface condition after cleaning was water-repellent, it was confirmed that the oxide film had been completely removed. Next, all seven levels of substrates were cleaned with ozone water with a concentration of 20 ppm at a cleaning temperature of 25°C for a cleaning time of 3 minutes, and a natural oxide film was formed on the surface under the same conditions.

Next, the second film thickness of the silicon substrates at seven levels was measured using a J. A. Woollam M-2000V spectroscopic ellipsometer. The results were: level 0: 1.199 nm, level 1: 1.202 nm, level 2: 1.248 nm, level 3: 1.250 nm, level 4: 1.291 nm, level 5: 1.206 nm, and level 6: 1.202 nm. The difference in film thickness between each level and level 0 in the second film thickness was calculated, and the results were: level 1: 0.003 nm, level 2: 0.049 nm, level 3: 0.051 nm, level 4: 0.092 nm, level 5: 0.007 nm, and level 6: 0.003 nm. Since the difference in film thickness between levels 1, 5, and 6 was 0.01 nm or less, it was determined that the first film thickness at levels 1, 5, and 6 did not include film thickness due to surface roughness. Because the difference in film thickness for levels 2, 3, and 4 is greater than or equal to 0.02 nm and less than or equal to 0.20 nm, it was determined that the first film thickness for levels 2, 3, and 4 includes film thickness due to surface roughness.

Furthermore, the film thickness excluding the influence of surface roughness from the first film thickness was calculated. Levels 2, 3, and 4 were calculated by subtracting the difference film thickness between the first film thickness and level 0 in the second film thickness from the first film thickness. For example, in the case of level 2, 1.060 nm is obtained by subtracting the difference film thickness of the second film thickness of 0.049 nm from the first film thickness of 1.109 nm, which is the film thickness excluding the influence of surface roughness from the first film thickness. Note that since levels 1, 5, and 6 are not affected by roughness, the first film thickness was used as is, and the film thickness excluding the influence of surface roughness from the first film thickness was set to the same value as the first film thickness. As a result of this calculation, the film thickness excluding the influence of surface roughness from the first film thickness was 1.033 nm for level 1, 1.060 nm for level 2, 1.070 nm for level 3, 1.077 nm for level 4, 1.114 nm for level 5, and 1.109 nm for level 6.

Here, when paying attention to levels 1, 2, 3, and 4, the magnitude relationship of the film thickness excluding the influence of the surface roughness from the first film thickness and the magnitude relationship of the first film thickness are both as follows:
Level 4 > Level 3 > Level 2 > Level 1
Since levels 1, 2, 3, and 4 are levels where the SC1 cleaning conditions were changed, it was found that the first film thickness value fluctuated due to the effect of both the surface roughness and structure of the native oxide film depending on the cleaning conditions.

Furthermore, Table 2 shows the results of an analysis of the factors that make the film thicknesses of levels 2, 3, and 4 thicker than level 1, which has the thinnest first film thickness. Here, the difference in film thickness between the second film thickness and level 1 is defined as the film thickness due to surface roughness (α), and the difference in film thickness between levels 2-4 and level 1, which is the film thickness excluding the effects of surface roughness from the first film thickness, is defined as the film thickness due to the structure of the native oxide film (β).

As a result, α was 0.046 nm in level 2, 0.048 nm in level 3, and 0.089 nm in level 4, and β was 0.027 nm in level 2, 0.037 nm in level 3, and 0.044 nm in level 4. Furthermore, α/β was calculated to be 1.703 in level 2, 1.297 in level 3, and 2.022 in level 4. Considering these results, since the values of α are almost the same in levels 2 and 3, it is considered that the film thickness caused by surface roughness is saturated in levels 2 and 3. It was found that the reason why the film thickness in the first measurement is thicker in level 3 than in level 2 is because the film thickness caused by the structure of the native oxide film is thicker in level 3. Level 4 had the highest α/β, which is thought to be due to the solution composition being changed from 1:1:10 to 1:1:100, which became a more etching-dominant chemical solution, causing the film thickness caused by surface roughness to become dominant. In this way, the factors behind the film thickness variation when the cleaning conditions were changed were analyzed in detail.

Example 2
In Example 2, the thickness of the thermal oxide film was evaluated. Silicon substrates that had been polished by CMP and then cleaned by single-wafer cleaning or SC1 cleaning were prepared as silicon substrates to be evaluated. The specific conditions are as shown in Table 3.

The prepared silicon substrate was thermally oxidized to a film thickness of 5 nm. After that, AFM measurements were performed on the level A substrate with an observation field of view of 1 μm x 1 μm, and a PSD curve was obtained from the spectrum analysis of the profile data. As a result, Sa was 0.011 nm, which was 0.5 nm or less, and the average value of the power spectrum density (intensity) at spatial frequencies of 60 to 90/μm was 0.094 ^nm3 , which was 0.1 nm3 or ^less. Therefore, this level A substrate was used as the reference silicon substrate, and the four levels of substrates B, C, D, and E were used as the silicon substrates to be evaluated.

Next, a first film thickness measurement of the silicon substrate to be evaluated was performed using a J. A. Woollam M-2000V spectroscopic ellipsometer. The results were: Level B: 5.185 nm, Level C: 5.269 nm, Level D: 5.272 nm, and Level E: 5.318 nm.

Next, a total of five levels of substrates, four levels of silicon substrates to be evaluated and one level of reference silicon substrate, were cleaned with 5 wt% hydrofluoric acid at a cleaning temperature of 25°C for a cleaning time of 10 minutes. Because the film thickness was thick, the concentration and cleaning time were changed from Example 1. Since the surface condition after cleaning was water-repellent, it was confirmed that the oxide film had been completely removed. Next, a total of five levels of substrates were cleaned using ozone water with a concentration of 20 ppm at a cleaning temperature of 25°C for a cleaning time of 3 minutes, and a natural oxide film was formed on the surface under the same conditions.

Next, the second film thickness of the silicon substrates of five levels was measured using a J. A. Woollam M-2000V spectroscopic ellipsometer. The results were 1.201 nm for level A, 1.201 nm for level B, 1.249 nm for level C, 1.249 nm for level D, and 1.285 nm for level E. The difference in the second film thickness between level B, C, D, and E and level A was calculated, and the results were 0.000 nm for level B, 0.048 nm for level C, 0.048 nm for level D, and 0.084 nm for level E. Since the difference in the film thickness of level B was 0.01 nm or less, it was determined that the film thickness due to surface roughness was not included in the first film thickness of level B. Since the difference in the film thickness of levels C, D, and E was 0.02 nm or more and 0.20 nm or less, it was determined that the first film thickness of levels C, D, and E included the film thickness due to surface roughness.

Furthermore, the film thickness excluding the effect of surface roughness from the first film thickness was calculated. Levels C, D, and E were calculated by subtracting the difference in film thickness between the first film thickness and level A at the second film thickness from the first film thickness. Since level B is not affected by roughness, the first film thickness was used as is, and the film thickness excluding the effect of surface roughness from the first film thickness was set to the same value as the first film thickness. As a result of this calculation, the film thickness excluding the effect of surface roughness from the first film thickness was 5.185 nm for level B, 5.221 nm for level C, 5.224 nm for level D, and 5.234 nm for level E.

Here, when paying attention to the levels B, C, D, and E, the magnitude relationship of the film thickness excluding the influence of the surface roughness from the first film thickness and the magnitude relationship of the first film thickness are both as follows:
Level E > Level D > Level C > Level B
Because levels B, C, D, and E are levels where the SC1 cleaning conditions were changed, it was found that the first film thickness value fluctuated due to the effect of both the surface roughness and structure of the native oxide film depending on the cleaning conditions.

Furthermore, Table 4 shows the results of an analysis of the factors that make the film thicknesses of levels C, D, and E thicker than level B, which has the thinnest first film thickness. Here, the difference in film thickness between the second film thickness and level B is defined as the film thickness due to surface roughness (α), and the difference in film thickness between levels C, D, and E and level B, which is the film thickness excluding the effect of surface roughness from the first film thickness, is defined as the film thickness due to the structure of the native oxide film (β).

As a result, α was 0.048 nm for level C, 0.048 nm for level D, and 0.084 nm for level E, while β was 0.036 nm for level C, 0.039 nm for level D, and 0.049 nm for level E. Further calculation of α/β gave results of 1.333 for level C, 1.230 for level D, and 1.714 for level E. In this way, it was possible to perform a detailed analysis of the factors behind film thickness variations when cleaning conditions were changed.

As described above, according to the embodiment of the present invention, it is possible to determine whether the film thickness value includes the influence of film thickness influencing factors such as wafer surface roughness formed during the silicon wafer manufacturing process, and to accurately evaluate the film thickness caused by the film thickness influencing factors.

The present invention is not limited to the above-described embodiment. The above-described embodiment is merely an example, and anything that has substantially the same configuration as the technical idea described in the claims of the present invention and exhibits similar effects is included within the technical scope of the present invention.

Claims (10)

A method for evaluating an oxide film on a silicon substrate having a surface roughness Sa value of 0.5 nm or less as measured by an atomic force microscope, comprising the steps of:
a substrate preparation step of preparing a silicon substrate to be evaluated on which an oxide film to be evaluated is formed, and a reference silicon substrate having an average value of power spectrum density of 0.1 ^nm3 or less at a spatial frequency of 60 to 90/μm on the substrate surface;
a first film thickness measuring step of measuring a film thickness of the oxide film on the silicon substrate to be evaluated by an ellipsometer;
an oxide film removing step of completely removing oxide films on the evaluation target silicon substrate and the reference silicon substrate;
an oxide film forming step of forming an oxide film under the same conditions on the evaluation silicon substrate and the reference silicon substrate after removing the oxide film;
a second film thickness measuring step of measuring the film thicknesses of the oxide film on the evaluation silicon substrate and the oxide film on the reference silicon substrate after the oxide film forming step by an ellipsometer;
and a thickness evaluation step of evaluating a thickness caused by a thickness influencing factor in the overall thickness of the oxide film on the silicon substrate to be evaluated, which is acquired in the first thickness measurement step, based on the thickness of the oxide film on the silicon substrate to be evaluated and the thickness of the oxide film on the reference silicon substrate, which are acquired in the second thickness measurement step. The oxide film evaluation method according to claim 1, characterized in that in the oxide film removal step, the oxide film is removed by washing with hydrofluoric acid. The method for evaluating the thickness of an oxide film according to claim 1 or 2, characterized in that in the oxide film formation process, the oxide film is formed by cleaning with ozone water or hydrogen peroxide water. The method for evaluating the thickness of an oxide film according to any one of claims 1 to 3, characterized in that the thickness of the oxide film to be evaluated on the silicon substrate to be evaluated is 25 nm or less. a film thickness influencing factor in the film thickness of the oxide film includes a surface roughness of the silicon substrate to be evaluated;
5. The method for evaluating an oxide film thickness according to claim 1, wherein, in the film thickness evaluation step, when a difference film thickness obtained by subtracting the film thickness of the oxide film on the reference silicon substrate obtained in the second film thickness measurement step from the film thickness of the oxide film on the evaluation silicon substrate obtained in the second film thickness measurement step is 0.02 nm or more and 0.20 nm or less, it is determined that the film thickness of the oxide film on the evaluation silicon substrate obtained in the first film thickness measurement step includes a film thickness of the oxide film caused by a surface roughness of the evaluation silicon substrate. The oxide film thickness evaluation method according to claim 5, characterized in that in the film thickness evaluation step, the difference film thickness is subtracted from the film thickness of the oxide film on the silicon substrate to be evaluated, obtained in the first film thickness measurement step, to evaluate the film thickness of the oxide film excluding the influence of the surface roughness of the silicon substrate to be evaluated. The method for evaluating the thickness of an oxide film according to claim 5 or 6, characterized in that the surface roughness is a roughness component with a spatial frequency of 60 to 90/μm. The method for evaluating the thickness of an oxide film according to any one of claims 5 to 7, characterized in that the surface roughness is a roughness component formed by SC1 cleaning. A method for manufacturing a silicon substrate with an oxide film, comprising: setting cleaning conditions and/or oxidation conditions before forming an oxide film on a silicon substrate based on the film thickness caused by the film thickness influencing factors evaluated by the method for evaluating the thickness of an oxide film according to any one of claims 1 to 8; cleaning the silicon substrate and forming an oxide film on the silicon substrate using the cleaning conditions and/or oxidation conditions; and manufacturing a silicon substrate with an oxide film. A method for evaluating the thickness of an oxide film formed on a silicon substrate, comprising the steps of:
A method for evaluating the thickness of an oxide film, comprising the steps of: including surface roughness of the silicon substrate as a thickness-influencing factor in the thickness of an oxide film; and determining that the thickness of an oxide film formed on the ^silicon substrate includes a thickness of the oxide film caused by the surface roughness of the silicon substrate, which is a thickness-influencing factor, when an average value of a power spectral density at a spatial frequency of 60 to 90/μm on the surface of the silicon substrate is 0.15 nm3 or more.

Images