JP2003289058A - Method for polishing compound semiconductor wafer and apparatus for polishing the same - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method for polishing a wafer and a polishing jig wherein no residue is left between a template and a suction pad during polishing a wafer surface. <P>SOLUTION: The suction pad 6 is adhered to the rear surface of a polishing plate 4, and the circular template 7 is adhered to a non-joint part of the inner circumference of the suction pad 6 to prevent the sliding of the wafer. After the wafer is removed, the suction pad 6 and the template 7 are cleaned by rotationally swinging a brush 40 and jetting cleaning liquid. A residue between the template and suction pad can be removed by the brush 40. Cleaning per polishing is desirable, and it may be conducted after polishing several wafers. <P>COPYRIGHT: (C)2004,JPO

Description

Detailed Description of the Invention

[0001]

TECHNICAL FIELD The present invention relates to a method for polishing a compound semiconductor wafer, that is, an InP wafer or a GaAs wafer. A semiconductor wafer is a thin piece obtained by cutting a semiconductor single crystal ingot. When the shape of the ingot is circular, it is often cut as it is to form a circular wafer. If the ingot is not circular, the ingot is ground into a circular shape (cylindrical grinding) to form a circular wafer, or the wafer is cut into a circular shape. However, the wafer is not limited to a circular shape, and a rectangular wafer is sometimes used. In the case of a circular wafer, OF for indicating the orientation is attached to the peripheral portion. In the ingot stage, the crystal orientation is examined by X-ray diffraction to mark the side surface of the ingot to form a wafer, and then the wafer is cut along the cleavage. Alternatively, the side surface may be mechanically ground in the ingot as it is.

A wafer cut from an ingot is called an as-cut wafer. This has a work-affected layer on the surface and has surface irregularities. The thickness also varies. The edges are sharp and easily broken. Therefore, the work-affected layer on the surface is removed by etching. The front surface and the back surface are shaved by lapping to have a predetermined thickness.

Further, when chamfering, the edge of the wafer is applied to a rotary grindstone having a concave surface and the upper and lower edges are slightly shaved off.
This is called beveling (Champha). In the case of a circular wafer, it has a disc shape and an OF.

After that, both sides or one side of the wafer is polished. When polished, the wafer has a metallic luster and reflects light like a mirror. The wafer thus polished is called a mirror wafer. Depending on the intended use, the wafer is polished into a single-sided mirror or a double-sided mirror.

[0005]

2. Description of the Related Art A polishing apparatus has a large circular plate with a polishing cloth on its upper surface, which rotates (revolves), and a wafer which is fixed to the lower surface with wax and pressed against the polishing surface plate to rotate itself (automatically rotate). And a polishing liquid supply device that supplies the polishing liquid to the polishing platen. The polishing liquid is caused to flow in the center of the rotating polishing platen, and the platen is rotated largely while rotating the head. The polishing cloth is composed of a voluminous material of suitable softness. It is a material that can maintain a flat surface while containing a polishing liquid therein. Materials such as polyurethane and artificial leather are used for the polishing cloth.

Polishing methods are divided into mechanical polishing and chemical polishing. Mechanical polishing is mainly mechanically scraping the surface of a wafer with a polishing liquid containing polishing particles such as colloidal silica. Chemical polishing uses a polishing liquid such as alkali or acid to chemically and chemically remove the surface of a wafer to give a mirror surface. Both may be used together depending on the purpose.

Only Si, GaAs and InP are actually manufactured and used in large quantities as semiconductor wafers. S
A large number of wafers having a diameter of 20 cm (8 inches) and a diameter of 30 cm (12 inches) are manufactured and sold in large quantities, and i is actively used for manufacturing semiconductor devices. S
Although the i-semiconductor and other semiconductors are greatly different, the term compound semiconductor is used as a term for indicating a semiconductor other than the Si semiconductor. It is a concept that originally covers a wide range, but here a wide single crystal wafer can be manufactured.
Limited to aAs and InP. In other words, GaAs wafer and In
A compound semiconductor wafer is used as a term meaning only P wafer. It is called a compound semiconductor because there is no suitable term that includes only GaAs and InP. In the case of GaAs, there are limited applications such as AlGaAs laser substrates and high-speed IC substrates.

The polishing method differs depending on the material of the wafer.
For physically robust Si wafers, silica particles (SiO 2
A polishing liquid prepared by dispersing ₂ ) in water is often used.

In the case of GaAs, physical polishing may be performed using a polishing liquid in which silica particles are dispersed in water. Also, chemical polishing may be performed. For example, an aqueous solution of sodium hypochlorite (NaClO) is used as a polishing liquid to chemically remove the surface of GaAs. When the sodium hypochlorite remains, the wafer becomes uneven and cloudy. In order to prevent this, hydrogen peroxide solution (H ₂ O ₂ ) is supplied to the platen before the end of polishing to remove sodium hypochlorite. Alternatively, a sodium thiosulfate solution (Na ₂ S ₂ O ₃ ) may be supplied to the polishing platen at the final stage. For example, it is described in JP-A-7-201786, "Compound semiconductor substrate polishing method and polishing apparatus", JP-A-2-207527, "GaAs wafer polishing method", and JP-A-3-248532 "Semiconductor wafer processing method". . In the case of chemical polishing, it is thus necessary to rapidly remove the chemical polishing material at the end of polishing. In the case of physical polishing, it was necessary to completely remove the abrasive after the polishing was completed due to the change in the adsorption method. I will explain the circumstances.

A method of fixing the wafer to the plate of the polishing head will be described. The polishing head is composed of a circular plate and a head shaft extending upward from the center of its upper surface. The head shaft is a member that is rotated by a rotating mechanism and that applies an appropriate pressure to the wafer. The plate fixes the wafer on its lower surface. The wafer must be firmly fixed to the plate. Also, the wafer must be fixed flat on the plate.

Conventionally, the wafer is fixed to the back surface of the plate using wax. The wax melted by heat is flattened on the plate and the wafer is attached. If the wax has irregularities, the irregularities are transferred to the surface of the polished wafer, which is not preferable. Wax can completely fix the wafer to the head. After polishing is finished, the wax is heated to melt the wax and the wafer is separated from the plate. It is further washed with an organic solvent or the like to remove the wax residue. The wax method has been used for Si wafers, GaAs wafers, and InP wafers.

Wafer fixing with wax has a proven record and is highly reliable. But it still has its problems. S
It has been stated that the i-wafer has a larger diameter. The larger the area, the greater the possibility that dust will adhere to the plate. Even if a small amount of foreign matter is sandwiched, it is transferred to the surface by polishing. In order to eliminate such an influence, it suffices to thicken the wax, but if this is done, the thickness will not be uniform over the entire surface of a wide wafer. A slight distribution of the wax thickness may cause unevenness or distortion of the wafer after polishing.

Another drawback of the wax method is that it takes time to put on and take off. The plate must be heated to melt the wax and flatten to fix the wafer, and after polishing it must be heated to remove the wafer. Waxes contaminate the backside of the wafer and must be thoroughly cleaned, which further increases working time. Although the wax may be fixed while the diameter of the wafer is small, the difficulty of the wafer attaching / detaching process with the wax becomes larger as the diameter of the wafer becomes larger.

Therefore, a vacuum chuck fixing method is used for a 20 cm diameter wafer for a Si wafer and a 30 cm diameter wafer. In this method, a narrow exhaust passage having a hole on the back surface of the plate with the suction pad attached is dug from the plate to the head shaft, and the exhaust passage is connected to the vacuum exhaust device from the head shaft. Since there is no elastic body such as wax, a new flexible suction pad is needed. In addition, it is necessary to form a passage for evacuation in a plate or shaft and connect it to an evacuation device. When the wafer is pressed against the plate and the gas passage is evacuated, the difference in air pressure causes the wafer to be drawn closer to the plate. No horizontal displacement during polishing.

When polishing is completed, the vacuum exhaust device is turned off and air is introduced into the exhaust passage to separate the wafer from the plate. This method does not require heating or cleaning. It does not take time to put on and take off. Problems such as non-uniform wax thickness cannot occur. Since Si has excellent mechanical strength and is robust, the wafer can be firmly fixed to the plate by a vacuum chuck. The vacuum chuck method, however, requires a vacuum exhaust device, and it is necessary to punch exhaust grooves and holes in the polishing plate, which complicates the device and has the disadvantage that it takes a lot of time to evacuate or use atmospheric pressure. . Recently, a simpler wafer holding method than the vacuum chuck has been developed and put into practical use. A urethane plate is attached to the polishing plate, and a ring-shaped template is attached to the pad. The suction pad is well made, and the wafer is firmly held on the polishing plate only by pressing the wafer against the suction pad. Although it may be evacuated, it is excellent in that the wafer is held by the intrinsic adsorption force of the adsorption pad without the evacuation. In such a case, the vacuum exhaust device is unnecessary. There is no need to dig grooves or holes in the polishing plate. After completion of polishing, a spatula may be inserted into the edge of the wafer and lifted to remove the wafer from the suction pads. However, even in that case, it is possible to form a groove or a hole in the polishing plate and connect it to an external pump, and at the time of peeling, air is sent from the side of the polishing plate to lift the wafer and facilitate the peeling. Although such a suction pad system has been developed in the field of Si wafers, it is desired to use it for GaAs wafers and InP wafers. It was found that there are some problems when the vacuum chuck method or the suction pad method described above is applied to GaAs and InP wafers.

[0016]

Si wafers are used in many devices. GaAs wafer is also laser (L
D), light emitting diodes (LEDs), and other high-speed logic devices (ICs). The problem here is not the Si wafer but the compound semiconductor GaAs.
A wafer and an InP wafer. Since GaAs has already been mentioned, the demand for the newest InP wafer will be explained.

InP is similar to GaAs in zinc blende (Zinc
It is a cubic crystal of blende type. Crystals can be grown by the liquid capsule method (LEC; Liquid Encapsulated Czochralski) and the vertical Bridgman method (VB; Vertical Bridgman). Optical communication system (1.3 μm band, 1.
55 μm band photodiode (PD; InGaA)
s, InGaAsP) has been used as a substrate material. It has a band gap in the optical communication band and is suitable as a light receiving element. However, the demand for the InP-based light receiving element is small, and the InP-based light receiving element is 300 μm-
Since the device is 400 μm square and small, a wafer is not necessary so much. Therefore, the demand for InP wafers for optical communication PD is still small. The shipping volume of InP wafers for PD substrates is high. Therefore, a small wafer of about 2 inches is manufactured.

However, new possibilities are emerging for InP. The electron mobility of InP is higher than that of Si. Not only n-type InP substrate but also semi-insulating (SI; Semi-insulatin)
g) InP substrates have also become possible to manufacture. Further VB
The method has made it possible to produce InP crystals with a considerably low dislocation density (20,000 to 10,000 cm ⁻² ). Therefore, the application as a high speed transistor is about to open.

The electronic circuits inside the mobile phone are made by Si devices. However, Si has a low electron mobility and there is a limit to the speedup. For higher speeds, the digital circuit part uses Si devices and the analog part uses high-speed G
It is said that it may be configured in a complementary manner by using an aAs device. In preparation for this, GaAs-based transistors are being further improved. It may be one of the strong options in the future. However, it seems that further miniaturization would be difficult if two types of devices (Si and GaAs) having different substrate materials coexist.

If a device including both a Si-based device and a GaAs-based device exists, it will be possible to make the next-generation mobile phone smaller and faster.
InP is expected as a substrate for such new electronic devices. Although still in the testing stage, InP-based analog / digital transistor devices may be manufactured. If so, the demand for InP wafers should increase. For PD, the size of a chip is small and the size of a wafer is small, and it may be about 2 inches or 3 inches. However, if it is a substrate of an electronic device with a large chip size, a larger diameter will be desired. .

[0021] With such expectations, attempts have been made to improve the manufacturing technique of InP wafers. In the crystal growth method, LE
The VB method has been tried instead of the C method and the HB method, and it has become possible to manufacture low dislocations. Although the dislocations are not low enough to produce high-speed transistors, it will eventually be possible to produce desired low-dislocation InP crystals. Further maturation of crystal growth technology is desired.

The polishing technique must of course also be improved. In terms of polishing, GaAs and InP have similar problems, but here, the InP wafer will be described.
In the case of a wafer having a small diameter, a method of sticking it on the plate of the head with wax is used. InP
Wafers are now fixed to the plate with wax,
Polish in that state. InP wafers are still as small as 1 inch or 2 inches in diameter, which is all that is required. However, with a delay, InP wafers will follow the path of larger diameters. If so, there is a problem with wax fixation. InP substrates will also be manufactured to 4 inches (10 cm).

As described above, it takes time and effort to heat and melt the wax to attach the wafer to the wafer, remove the wafer from the plate by heating, and a step of cleaning to remove the wax is added to remove dust. This is because when sandwiched, there is a problem that it is transferred and unevenness is formed on the wafer surface.

The fixing method used for large-diameter Si wafers, in which a wafer is fixed to a plate by a vacuum chuck and polished, is tried. With the technique of attaching a suction pad to the polishing plate, making a hole, providing a passage for vacuum exhaustion therethrough, connecting the passage to a vacuum exhaust device, pressing the wafer against the plate, vacuum exhausting and fixing the wafer to the plate is there. Furthermore, for the Si semiconductor, it is desired to use the suction pad method that has been put into practical use.

In the case of Si wafer, the successful technology is InP
It turned out that it cannot be applied to a wafer as it is. In
P is less rigid than Si and is brittle. Si is more robust than the original, and the wafer does not break even if the force of the vacuum chuck is increased. However, since InP is inferior in mechanical strength to Si, it cracks if the force of the vacuum chuck is increased too much. So you can't pull too hard. The InP wafer cannot be fixed only by the vacuum chuck, and the InP wafer slides on the plate. If the wafer slides on the lower surface of the polishing plate, the wafer cannot be polished well.

Therefore, it is necessary to devise to supplement the vacuum chuck. It is pulled in a vacuum and does not come off in a direction perpendicular to the surface.
So you shouldn't skid. To prevent this, a ring may be attached to the plate concentrically with the outer circumference of the wafer to prevent the wafer from slipping. It is possible to prevent the sideslip by attaching a ring having an inner diameter slightly larger than the outer diameter of the wafer and slightly thinner than the wafer to the plate and vacuum chucking the InP wafer inside the ring. Such a skid prevention ring is called a template.

It is desired to polish a large-diameter InP wafer using a polishing plate having a structure in which a suction pad is attached to the back surface of a plate having a vacuuming hole and a passage, and a ring-shaped template is attached to the outer peripheral portion of the plate. S is a large diameter
4 at best, not 12 inches (30 cm) like i
InP with an inch (10 cm) diameter.

Ga, which has a lower strength than Si, like InP
Also in the case of As wafer, a vacuum chuck fixing method using a template can be used. Therefore, GaAs will also be a vacuum chuck as the diameter increases. Furthermore, the application of the suction pad system will be demanded.

The vacuum chuck method and the suction pad method of GaAs and InP are still in the testing stage and cannot be said to be in practical use. The polishing methods include physical polishing and chemical polishing, GaAs is often chemically polished (sodium hypochlorite), and Si and InP are polishing liquids containing hard polishing particles such as colloidal silica. Many. However, conversely, GaAs may be physically polished and InP may be chemically polished. It is difficult for any of the polishing liquids to remain on the polishing plate and the suction pad after polishing. The present invention is a device for avoiding the residual polishing liquid, and is the same problem in both chemical polishing and physical polishing, but the problem when colloidal silica is used will be described below.

The abrasive particles are not dissolved in the polishing liquid but merely dispersed. Therefore, when the polishing liquid dries, the silica particles will stick to the polishing equipment and polishing cloth. When the InP wafer is polished between the rotary platen and the plate while supplying the polishing liquid, silica particles adhere to the plate surface. When not using the template, the abrasive particles could be easily removed by simply wiping the plate surface.

However, if the template is pasted on the plate, silica particles will remain at the corners of the joining portion between the template and the plate. It sticks in the corner, so it's not easy to remove. In addition to the shape problem of attaching the template to the plate, there are conditions under which silica particles are likely to remain.

In the case of a Si wafer, the polishing time is about 1 to 2 minutes, which is short. Even if the polishing particles adhere to the plate, the amount of residual particles is small because the polishing time is short. It's easy to remove particles because you don't use a template.

In the case of a GaAs wafer, either physical polishing or chemical polishing is performed. When using the polishing plate or template, the polishing liquid should not remain on the suction pad or template.

In the case of an InP wafer, the polishing time is as long as 10 minutes. Since the polishing time is long and the physical polishing uses colloidal silica, it is easy for the polishing particles to adhere to the plate surface. A suction pad is attached to the plate, and a template is attached to the outer peripheral portion of the suction pad. Since there is a very narrow recess between the wafer and the template, silica particles collect in this recess. Even if the InP wafer is removed, thick abrasive particles (colloidal silica or the like) remain inside the template.

The next InP wafer cannot be attached to the plate while leaving the residue (abrasive particles or wafer fragments) as it is. One reason is that the residual particles raise the edge of the InP wafer and the wafer tilts, so that the wafer cannot be fixed by the vacuum chuck. Secondly, hard debris contained in the residue may reach the lower surface of the wafer during polishing and damage the wafer. A linear scratch formed by the hard particles rubbing the wafer surface is called a scratch. Further, there is a possibility that the residue may be scattered around the wafer surface and stain the wafer surface.

InP also has a disadvantage in the occurrence of scratches. Since Si has a Vickers hardness of about 700 and is hard, the surface of the wafer is not easily scratched. But I
Since nP has a Vickers hardness of about 400 and is soft, scratches are easily generated by the hard substance in the residue. In the case of Si, as described above, the wafer can be temporarily fixed to the plate by simply using the vacuum chuck method or the suction pad method without using the template, so that there is no inconvenience that silica particles are accumulated on the edge of the template.

In the case of GaAs, a template is used when temporarily fixing with a vacuum chuck method or a suction pad method.
Even with chemical polishing, the solid matter on the edge of the template may remain. When the InP wafer is temporarily fixed and polished by the vacuum chuck method or the suction pad method, it is a physical polishing using a polishing liquid containing a suspended substance, and the polishing time is long and a bad condition such that a template is used is prepared.

For this reason, in the vacuum chuck holding method and the suction pad holding method using the suction pad and the template, residues such as polishing particles are apt to accumulate on the inner circumference of the template, and the disadvantages due to the residues are large. Therefore, it is necessary to cleanly remove the residue such as abrasive particles from the plate every time one piece is polished.

At present, InP wafers are produced on a trial small scale and are not in mass production. Therefore, every time one InP wafer is polished, an operator manually rubs the back surface of the plate with a brush to scrape off abrasive particles stuck to the template and the plate surface. Clean the plate surface by pressing the rotating brush against the plate surface by hand. The plate is cleaned for about 20 seconds while rotating the brush at about 30 rpm. 5 seconds is not enough. It seems that 10 seconds is not enough. Since it is a manual work, the angle of the brush can be freely changed. By pressing the brush against the plate in different orientations, it is possible to completely scrape off any dirt that has stuck to the edges of the template.

However, this is costly and inefficient because it requires manpower. We would like to automate the process of polishing the InP wafer starting from the temporary fixing of the InP wafer with a vacuum chuck.

As described above, the mainstream polishing of InP wafers is still to fix the plate to the wax.
The vacuum chuck method described here is still in the trial stage. Without solving the problem of cleaning the back surface of the plate, the vacuum chuck method and the suction pad method cannot be actually used for InP wafer polishing. Such problems also exist in GaAs wafers.

[0042]

According to the polishing method of the present invention, a suction pad is attached to the back surface, and a ring-shaped template is attached to the suction pad so as to have a non-bonding portion so as to prevent lateral displacement. The GaAs wafer is pressed against the suction pad or fixed by a vacuum chuck, the polishing plate is pressed against the rotating surface plate, and the InP or GaAs wafer is polished by rotating the rotating surface plate and the polishing plate while supplying the polishing liquid, After polishing, lift the wafer from the surface plate and remove the wafer. While spraying the cleaning solution on the back surface of the polishing plate, rub the back surface of the polishing plate with a brush that has diagonally extending bristles and remain at the corners of the joint surface between the template and suction pad. The solid matter is removed.

Since a residue is likely to remain between the suction pad and the template, it is important to rub the suction pad and the template with a brush to physically remove dirt and residue.
Depending on the size and model of the brush, the swing or rotation speed is 3
It is often 0 rpm to 100 rpm. The time required for one cleaning is about 5 to 20 seconds. Most preferably, the pad and template are cleaned after each polish. However, the pad and template may be washed every two or three sheets.

The width of the unbonded portion of the template with respect to the pad is about 0.5 mm to 7 mm. Non-bonded part is 0 mm
If so, scratches appear on the wafer in the next polishing, which is not preferable.

The cleaning liquid may be pure water, alkali or weak acid. Pure water may be used because the polishing particles and the polishing pieces of the wafer adhere to the concave portions of the template. Pure water is suitable if the polishing liquid contains pure water and polishing particles only (physical polishing). When the polishing liquid contains not only silica particles but also chemicals (when used in combination with physical and chemical polishing), chemicals may remain. In that case, the cleaning solution may be a weak alkaline solution or a weak acid solution in order to neutralize the action of the chemicals.

[0046]

BEST MODE FOR CARRYING OUT THE INVENTION The present invention uses an InP wafer or Ga on a polishing plate to which a suction pad and a template are attached.
The As wafer is fixed without using wax by the internal suction force or vacuum suction force of the suction pad. When using a vacuum, a vacuum exhaust path and suction holes are provided in the polishing plate and polishing head. If the suction force of the suction pad is intrinsic, the vacuum exhaust path, hole, groove, etc. are unnecessary. The wafer is polished while being sucked by the suction pad.
After completion of polishing, the back surface of the polishing plate is rubbed with a brush to clean. Since the residue and dust of the polishing liquid accumulates on the non-bonded portion between the suction pad and the template, the brush is used to forcibly remove it.

FIG. 1 is a vertical sectional view of a polishing apparatus used in the present invention. Although described as an InP wafer, the same applies to a GaAs wafer. A polishing cloth 2 made of resin or cloth is attached to the rotary platen 1. This rotates (revolves) around the rotation axis. A polishing head 3 is a jig for pressing the wafer on the rotary platen in a state where the wafer is attached. This is composed of a disk-shaped polishing plate 4 and a head shaft 5 following the upper part thereof. The upper end of the head shaft 5 is rotatably supported by a bearing (not shown) and is rotated by a motor (not shown). Polishing plate 4 and head shaft 5
Rotate together (rotate). A resin (for example, polyurethane) suction pad 6 is attached to the lower surface of the polishing plate 4. The suction pad 6 has a function of sucking the wafer and softening the impact force between the wafer and the polishing plate 4. A ring-shaped template 7 for determining the position of the wafer and preventing lateral displacement is fixed to the suction pad 6 by adhesion or thermocompression bonding. The outer peripheral portion of the template 7 is completely fixed, but the inner peripheral portion is a non-bonded portion that is not partially bonded.

An InP wafer (or GaAs wafer) 8 is formed on the surface of the suction pad 6 surrounded by the template 7.
Attach. Vacuum chuck without using wax. Therefore, the head shaft 5 has an exhaust passage 10 formed in the axial direction.
The polishing plate 4 also has an exhaust passage 11 that is radially branched. The end of the exhaust passage 11 is a suction hole 12 continuing to the outside. The suction hole 12 is the polishing plate 4
It is also worn on the pad 6. The exhaust passage 10 of the head shaft 5 is connected to an external vacuum pump (not shown). The back surface 13 of the InP wafer 8 contacts the suction pad 6 and is fixed by vacuum. Although FIG. 1 illustrates the case of the vacuum chuck method, a suction pad method of holding a wafer by the suction force of the suction pad can also be applied. The wafer is naturally adsorbed by pressing it against the adsorption pad. Even in this case, the suction pad may have a hole, the polishing plate may have a passage 12, and the shaft 5 of the polishing head may have a passage 10 to be connected to a vacuum device or a pump.
When attaching the wafer, it is sufficient to press it on the suction pad, but when peeling it off, the passages 10, 11 and 1 from the pump side are removed.
Air is sent to the suction pad through 2 to raise the wafer and make it easy to remove. It becomes a useful mechanism when detaching (releasing) a wafer. However, if a spatula is used when removing the wafer, the holes 10, 11 and 12 for exhaust and intake and the holes of the suction pad as shown in FIG. 1 are not necessary. The present invention includes such three modes (vacuum chuck suction, suction pad suction + air desorption, suction pad suction + spatula desorption), and FIG. 1 shows vacuum chuck suction, suction pad suction + air. It represents two cases of desorption. In the case of adsorption pad + spatula detachment, the holes of the adsorption pad and the passages 10, 11 and 12 of the polishing plate are removed from FIG. A narrow concave portion 9 is formed between the inner peripheral surface 14 of the template 7 and the outer peripheral surface 15 of the wafer, and polishing particles and the like remain therein.

In this way, the polishing head 3 holding the InP wafer 8 on the lower surface is pressed against the rotary platen 1, and while the polishing liquid is supplied, the rotary platen 1 and the polishing head 3 are rotated to flatten the surface of the wafer. Polish smoothly. The polishing liquid contains abrasive particles such as colloidal silica for physical polishing.

After the polishing is completed, the polishing head 3 is lifted and separated from the rotary platen 1, and the cleaning liquid 20 is sprayed onto the wafer surface 16 from the cleaning nozzle 20 to clean the wafer. That is Figure 2
Is the state shown in. After that, the wafer is removed from the suction pad. The desorption method is slightly different in each mode.
In the case of the vacuum chuck method, the vacuum chuck is released in the vacuum device to eliminate the vacuum, and the wafer is removed from the polishing head. In the case of adsorption pad + air desorption, air is sent to raise the adsorption pad and remove the wafer. In the case of adsorption pad + spatula detachment, insert the spatula into the edge of the wafer and lift to remove the wafer. As it is, the polishing head 3 needs to be cleaned because it contains abrasive particles and liquid. Spraying cleaning water on the back of the polishing head will not clean it. This is because the particles of the polishing liquid remain, and the particles firmly adhere to the inner peripheral portion of the template.

It is necessary to physically clean the plate as it will remove sticky solids. So you need to rub it with a brush. Manual work can be done by any method of rubbing depending on the judgment of the operator. Therefore, the dirt can be completely removed. However, this is not the case when automating. The brush will repeat a simple movement. Since solid residue adheres to the template and the concave portion of the plate surface, it is difficult to remove it. It doesn't stick to the other plate surface very much, and even if it does, it's a flat part and can be taken immediately. However, the inner edge of the template becomes a dent and is difficult to remove. Therefore, the tips of the brushes (bristles) enter into the concave portions of the template.

For that purpose, one method is to use a brush whose tip is deformed outward. When the bristle tip is pressed against the plate surface, the tip bends between the plate and the template, and the residual solid matter clogging the gap can be removed.

The brush rotates, swings, or reciprocates linearly. The angle between the rotation axis of the brush and the normal is called the tilt angle Θ. The brush is a rotating shaft (or swing shaft)
May be parallel to the normal to the plate (Θ = 0). Even in that case, the tip can enter the recess.

Even if Θ = 0, there can be two relative motions. One has a common axis (Fig. 4). The plate can be cleaned by reversing the rotation direction of the plate and the brush. The other is that the brush rotation axis deviates from the plate center axis, and the brush moves in a planetary motion (Fig. 6). The present invention can be used in either case.

Alternatively, the brush rotation oscillating shaft is inclined with respect to the plate shaft. Θ ≠ 0 (FIG. 5). In that case, the tip can be strongly inserted into the recess by pressing it against the plate.

The rotation and swing axes of the brush may be perpendicular to the plate axis. Θ = 90 degrees. In that case, the tip of the hair in the middle portion does not enter the recess. Instead, the tips of the hairs at the ends enter the recesses and scrape off the solid matter. Therefore, the cleaning ability seems to be slightly inferior.

FIG. 3 shows an example in which the pad and template are cleaned by rubbing the back surface of the polishing plate with a cylindrical brush. The circular cylinder brush 26 includes a shaft 25, a circular cylinder portion 27 provided at the tip of the shaft 25, and the bristles 2 provided on the circular cylinder portion 27.
Have eight. The cleaning liquid 21 is sprayed from the cleaning nozzle 20 onto the suction pad 6 and the template 7 on the back surface of the polishing plate 4. The shaft 25 is rotated, the polishing head 3 is also rotated, and the bristles 2
The pad and the template are rubbed by the hair tips of 8. By the action of the cleaning liquid 21 and the brush, the residual liquid, dirt, and particles on the pad or the template are removed.

FIG. 4 shows an example of cleaning the pad and template by rubbing the back surface of the polishing plate with a vertical brush. Here, the vertical brush is a brush in which bristles are planted almost vertically on the disk, and is called so as to be distinguished from the cylindrical brush of FIG. The vertical brush 30
A large number of thin bristles (33) are planted on the upper surface of the disc 31. The bristle tips 34 hit the pad or the template to clean it. A thinner one is easier to enter a narrow gap. Therefore, it is better to have hair tips of 1 mm or less. The support bar 35 supports the disc 31 from the side.

This brush rotates around the central axis of the disc 31. The normal 36 of the pad surface and the rotation axis 37 of the brush are parallel. After that, the pad surface normal 36 and the inclination Θ of the rotation axis 37
Is called the tilt angle. In this example, a rotation transmission mechanism and a bearing are provided inside the support rod 35 so that the disc is rotated around a rotation shaft 37 centered on the disc. When the brush is pressed against the pad, the bristle tips spread and enter the non-bonded portion of the inner peripheral portion of the template, and the non-bonded portion can be cleaned.

FIG. 5 shows an example of cleaning the pad and template by rubbing the back surface of the polishing plate with an inclined vertical brush. As mentioned above, the vertical brush is a brush in which bristles are planted almost vertically on the disc, and does not mean that the support rod and the disc are vertical. This is called by the relationship between the bristle tips and the disc to distinguish it from the circular cylinder brush of FIG. This vertical brush 40 also has a large number of thin bristles (43) planted on the upper surface of a disc 41. The tips 44 clean the pads and template. The support rod 45 supports the disc 41 from the diagonal side.

This brush rotates around the central axis of the disc 41. The normal 36 of the pad surface and the rotation axis 37 of the brush are not parallel but inclined. The rotation about the central axis of the disc is the same as in FIG. 4, but the rotational movement is also inclined because the disc itself is inclined. The tip of the hair is slanted, so it is suitable for cleaning the non-bonded portion between the template and the pad. Also in this example, a rotation transmission mechanism and a bearing are provided inside the support rod 45 so that the disc is rotated around the rotation shaft 37 centered on the disc. When the brush is pressed against the pad, the bristle tips enter the non-bonded portion of the inner peripheral portion of the template and can be cleaned there. If the inclination angle Θ is too large, the effective ground contact area is reduced, which is not effective. It is understood that the range of the useful tilt angle Θ is about 0 ° to 45 ° in the later experiment.

FIG. 6 shows an example of cleaning the pad and template by rubbing the back surface of the polishing plate with a smaller vertical brush. This vertical brush 50 is also a disk 5
A large number of thin hairs are transplanted (53) on the upper surface of 1.
The point that the cleaning liquid 21 is sprayed on the pad surface is the same as in the previous examples. Brush tips 54 clean the pads and template. The support rod 55 supports the disc 51 from the diagonal side. This brush rotates about the central axis of the disc 51. The direction of rotation is parallel to the normal 36 of the pad surface. However, since this is a smaller brush than that of FIG. 4, the brush can be moved on the pad surface to more finely clean the template and the pad.

FIG. 7 is an enlarged sectional view showing a contact state between the template 7 and the suction pad 6. The template and the suction pad are bonded by an adhesive material, or an appropriate material is selected and thermocompression bonded. Although the thickness of the joint is small,
In FIG. 7, the thickness of the adhesive material 17 is emphasized and drawn. This is because the non-bonded portion 29 is clearly shown. Template 7
If the entire back surface of the wafer is attached to the suction pad, the adhesive may stick out to press the edge of the wafer, or the protruding portion may attract polishing residues. This can cause scratches on the surface of the wafer. Therefore, a part of the inner peripheral portion of the template is formed as a void so as not to be bonded to the pad. Non-joint part 29
Call here.

The presence of the non-bonded portion 29 causes a drawback that the abrasive particles enter and accumulate in this portion. FIG. 8 shows such a state. After polishing, polishing particles and polishing liquid remain in the recesses 9. Since the template is flexible, it is raised due to the residue 22 getting in. In FIG. 8, it is shown exaggeratedly that the inner periphery of the template 7 is partially raised by the residue 22 (19).

FIG. 9 shows a state in which the bristle tips 64 of the brush-implanted hair 63 penetrate deep into the non-bonded portion 29 to forcibly remove the residue. In this way, it is important that the flocked part is flexible and bends to enter the non-bonded part. In addition, the non-bonded part is narrow, so the tip of the hair must be thin. In this figure, the disc portion 61 is inclined. On the other side, the slope is opposite. Such inclination is impossible unless the disc portion 61 is narrow. So it has to be a small brush. For large brushes, the tilt angle Θ must be small.

FIG. 10 shows a polishing plate that has been devised to facilitate removal of the residue at the non-bonded portion. The cylindrical portion 70 is extended below the polishing plate 4, and the peripheral portion of the pad 6 is attached to the lower end 71. A flexible balloon 72 capable of contracting and expanding is housed in an internal space 74 surrounded by the cylindrical portion 70 of the polishing plate and the pad 6. A flexible guide plate 75 is provided between the balloon 72 and the pad 6.
Is provided. A groove 76 is cut in the guide plate 75 so that gas can flow between the pad hole 78 and the internal space 74.

A gas passage 79 different from the exhaust passage 10 is bored in the head shaft 5. The gas passage 10 communicates with the internal space 74 via the small groove 73. Gas passage 79
Is connected to a different vacuum pump (not shown). If the gas is sent from the vacuum pump to the balloon 72 via the gas passage 79, the balloon becomes larger and larger. When the gas is pulled out, the balloon is degenerated. By extracting the gas in the internal space 74 of the polishing plate 4 from the exhaust passage 10 to the outside through the small groove 73, the wafer 8 can be adsorbed to the pad. It should be noted that there are two vacuum systems in this way. The path of the groove 76 of the guide plate 75 is determined according to the hole of the pad.

With the balloon 72 unfolded and the pad horizontal, the wafer is sucked to polish the wafer surface. After polishing, the head is lifted and washed with water, the wafer is removed, and the pad is exposed. The balloon is retracted to deform the pad 6 into a concave surface. FIG. 11 shows such a state. When the balloon 72 becomes smaller, the central portion of the pad is lifted up, the periphery of the pad is bent and the non-bonded portion is opened, so that the residue can be easily removed.

FIG. 12 shows another polishing plate improvement. Also in this case, the pad is deformed during cleaning to facilitate cleaning. A central portion of the polishing plate 4 becomes a columnar portion 82, and an annular balloon 84 is provided on the outer peripheral surface 83 thereof. The balloon 84 contacts the flange 88 of the polishing plate. A retaining ring 85 is provided below the balloon 84 on the outer peripheral surface 83 of the columnar portion 82. Lower surface 8 of the retaining ring 85
A peripheral portion of the suction pad 6 is attached to the pad 6. The central portion of the suction pad 6 is attached to the lower surface 87 of the columnar portion 82. An annular template 7 is attached to the suction pad 6.

An exhaust passage 89 for the vacuum chuck is bored inside the polishing plate 4. In accordance with this, the suction hole 90 is also formed in the pad 6. Balloon 8
4 also has exhaust passages 91 and 92 connected to it inside the polishing plate.

When the wafer is adsorbed, the balloon is retracted to make the pad flat. When the wafer is removed and the pad is washed and cleaned, the balloon 84 is inflated as shown in FIG. Since the retaining ring 85 sinks, the pad 6 deforms. The non-bonded portion 29 expands due to the deformation. As a result, the bristle tips easily enter the non-bonded portion. Therefore, the residue can be eliminated more completely.

[0072]

EXAMPLE The plate after polishing the InP wafer under various conditions was cleaned by changing the type of polishing liquid, the presence / absence of template, the type of cleaning liquid, the presence / absence of a brush, the motion mode of the brush, and the like. The number of each sample is 10 to several tens. After polishing the wafer, the polishing head is lifted to spray cleaning water on the lower surface (front surface) of the wafer to clean the surface of the wafer, remove the wafer from the plate, and place the wafer in the wafer carrier by an automatic carrier. Table 1 shows the polishing conditions, plate cleaning conditions, and polishing results.
Shown in.

In this table, from the left, the sample number, polishing liquid, presence / absence of template, presence / absence of cleaning mechanism, cleaning liquid,
It indicates the degree of dirt / defects, backside oxidation, and scratches. Backside oxidation is the oxidation of the backside of the wafer. The polishing liquid may reach the back surface of the wafer and oxidize it. This is not the case in the case of physical polishing, but the back surface of the wafer may be oxidized in the case of chemical polishing using an oxidizing chemical. Scratch-like scratches mean that linear scratches were generated on the surface of the next wafer that was attached to the same polishing plate and polished. It's not that the polishing plate was scratched.

[0074]

[Table 1]

[Sample 1] The polishing liquid does not contain particulate matter. In other words, only chemicals. There is no template on the polishing plate and it is pasted with wax.
The wafer was not removed and cleaned after polishing the wafer. So this is outside the scope of the invention from the beginning, but is included for comparison. In 70% of the samples, the back surface of the wafer was oxidized. Next, when the wafer was attached to the same polishing plate and the wafer was polished with the same polishing liquid, scratch-like scratches did not occur. This is a natural result because a polishing liquid containing no solid particles was used.

[Sample 2] Similar to Sample 1, the polishing liquid does not contain particulate matter. A suction pad is attached to the polishing plate so that it can be temporarily fixed (sucked) by a vacuum chuck. A resin template is attached to prevent skidding. Although the structure of the polishing plate falls under the conditions of the present invention, this is also outside the conditions of the present invention because the polishing liquid does not contain such particles as silica particles.
This is also listed for comparison. There is a polishing plate cleaning mechanism. The cleaning mechanism is simply spraying the cleaning liquid. The cleaning liquid is tap water. This was found to be the polishing plate stain for 20% of the samples. The wafer backside was oxidized for 10% of the samples. This is because the chemical polishing liquid was used. When the wafer was attached to the same polishing plate and the wafer was polished with the same polishing liquid, scratches did not occur. This is also natural because a polishing liquid containing no solid particles is used.

[Sample 3] A polishing liquid containing no silica was used for chemical polishing. Similar to the sample 2, a suction plate is attached to the polishing plate and a resin template is fixed. The wafer was temporarily fixed to the plate with a vacuum chuck and polished. In this case, after polishing the wafer, the polishing plate was washed with pure water. There were no stains or defects on the polishing plate. Wafer backside oxidation was seen for the 10% sample. No scratches occurred. This is natural because the polishing liquid does not contain silica or the like.

[Sample 4] As in Samples 1 to 3, the polishing liquid does not contain silica. The polishing plate has a suction pad on which a template is attached. After polishing the wafer, the wafer was washed with a 0.1% KOH solution (weak alkaline solution). There were no stains or defects. However, the backside oxidation of the wafer occurred about 10% as in Samples 2 and 3. No scratch was generated in the subsequent polishing of the wafer.

[Sample 5] Similar to Samples 1 to 4, the polishing liquid is a chemical liquid containing no particulate matter (such as silica). The polishing plate has a suction pad and a metal template. After fixing with a vacuum chuck and polishing and removing the InP wafer, the polishing plate was washed with pure water. There were no stains or defects. Oxidation on the backside of the wafer is 10
% Of the samples found. This is similar to Samples 2-4. 9 scratches on the next wafer polishing
% Of the samples found. Since the polishing liquid does not contain particles such as silica, the polishing liquid should not cause scratches. It is presumed that this is because a part of the metal template or a fragment of the wafer becomes a small solid particle fragment which remains in the concave portion of the template and wraps around the surface during the next polishing to damage the wafer. The soft wafer may be partially damaged by the collision of the metal and the wafer, and the fragments may remain. This means that rigid metal templates are not very good.

[Sample 6] As in Samples 1 to 5, the polishing liquid is a chemical liquid containing no particulate matter. A suction pad is attached to the polishing plate, and a ceramic template is fixed. After polishing the InP wafer and removing it, the polishing plate was washed with pure water. There was wafer backside oxidation for 10% of the samples. 6
Scratch-like scratches were found on% of the samples. It is also possible that a part of the ceramic template or a part of the wafer crushed by colliding with the template remains and scratches the wafer in the next wafer polishing. Sample 5
As is the case, the hard ceramic template is not suitable for pasting.

[Sample 7] Almost the same as Sample 5, except that
The difference is that the template is made of resin-coated metal. Wafer backside oxidation occurs for 10% of the samples. This is similar to Samples 1 to 6 because it occurs because the polishing liquid is a chemical. When this was also washed and the next wafer was polished, the occurrence of scratches was 0%. This point is different from Samples 5 and 6. It is probably because the template was coated with resin and part of the template could not be scraped, or the surface was given elasticity so that the wafer did not come into contact with the template and part of it was damaged and crushed. Therefore, there is no residual solid matter, and it seems that scratches did not occur on the wafer surface in the next wafer polishing.

[Sample 8] Almost the same as Sample 6, except that
The difference is that the template is made of resin-coated ceramic. Wafer backside oxidation occurs for 10% of the samples. This is similar to Samples 1-7. When this was also washed with pure water and the next wafer was polished, scratch-like scratches were found to be 0%. In that respect, it differs from Samples 5 and 6,
Same as sample 7. This is probably because the ceramic template was coated with resin so that part of the template could not be scraped, or the surface was given elasticity so that the wafer did not come into contact with the template and part of it was damaged and crushed. Therefore, there is no residual solid matter, and it seems that scratches did not occur on the wafer surface in the next wafer polishing. When the samples 3 and 5 to 8 are compared and weighed, the template may be made of resin, resin-coated metal, resin-coated ceramic, or the like, but may be made of bare metal,
It turns out that ceramic templates are not good. It is likely that some of the fragile InP is more crushed by contact with the template than Si and the like, and some of the InP is left behind. The above Samples 1 to 8 all use the polishing liquid containing no particulate matter, and there is no possibility that the particulate matter remains in the polishing fluid. However, it has been clarified that a part of the template or the wafer may become a granular material and scratches may be generated on the wafer.

[Sample 9] For samples after this, physical polishing was adopted and a polishing liquid containing a particulate solid substance was used.
The InP wafer is polished while supplying a polishing liquid containing colloidal (colloidal silica particles). A suction pad is attached to the polishing plate, and a resin template is attached to the periphery. The width of the non-bonded portion between the template and the polishing plate is 5 mm. Why are there unjoined parts? As mentioned above, if this is 0, then there is squeeze out of the adhesive material from the template, which interferes with wafer attachment. Therefore, instead of attaching the inner edge with an adhesive, a part of the inside of the template is an area without an adhesive. If so, the wafer cannot be lifted up, but on the other hand, there is a drawback that polishing particles remain on the non-bonded portion. The wafer was temporarily fixed to the back surface of the plate by a vacuum chuck, and polished with a template to prevent side slip. After polishing the wafer, remove the wafer from the plate,
Normally, the back surface of the plate is washed, but it was not washed here. Then, in the next polishing of the wafer, the back surface of the wafer was oxidized in 5% of the samples.

This does not mean that 5% of the total area of one sample was oxidized. If even a small portion of one sample is oxidized, it counts as an entire one. This means that 5% of all samples were backside oxidized. It becomes an important evaluation criterion because it should not be oxidized on the back side even if a little.

The physical polishing is mainly made of colloidal silica and strong chemicals are not used, but if the plate is not washed, the residue may adversely affect the back surface of the wafer and oxidize the back surface. 4 of the whole sample
About 0%, scratch-like scratches appeared on the surface of the wafer. This is because polishing particles (silica particles) remain on the back surface of the plate, and the residue may wrap around the wafer surface and damage the wafer surface at the next polishing. Soft In because it does not mix particles with a regular particle size, it also mixes wafer fragments.
The P wafer is scratched. As described above, the InP wafer is S
Like the i-wafer, it is physically polished with a polishing liquid containing particles, but the InP wafer is weaker than the Si wafer and is therefore easily scratched. On the other hand, the polishing time for one wafer is S
The i-wafer requires a short time of 1 to 2 minutes, whereas the InP requires 10 minutes, and the polishing time is long, so that it is easily affected by plate contamination.

Sample 9 is not possible because of the backside oxidation.
In addition, it is useless because scratch-like scratches appear. In other words, it is impossible to clean the back surface of the plate after each polishing. The surface oxidation of wafers that can be shipped must be 0%. Also, scratches must be 0%.

[Sample 10] A polishing liquid containing colloidal silica was used. The InP wafer is temporarily fixed to a plate with a pad attached by a vacuum chuck. The template is made of resin and attached so that the non-bonded portion has 5 mm inside. The InP wafer was physically polished by such a polishing liquid. After polishing, the wafer was lifted from the rotary platen and washed with water to wash the surface (lower surface) of the wafer. The wafer was removed, and the inside of the template and the template were washed with washing water. But no brush was used. Colloidal silica particles will remain between the template and pad because they are not rubbed by a brush.

Then, the next polishing was performed. In the next polishing, 1% of the wafers had backside oxidation. Further, scratch-like scratches were generated on the surface at a rate of 30%. The backside being oxidised would mean that the residue left on the template caused a chemical reaction on the backside of the wafer. The occurrence of scratches probably means that silica particles and the like remaining on the inner edge of the template go around to the surface of the wafer and damage the surface of the wafer. Scratch scratches should not occur as much as 30%. It means that the stain around the template hinders the polishing of the next InP wafer. That means that the stain around the template must be completely eliminated.

[Sample 11] An InP wafer was temporarily fixed on a pad having a resin template by a vacuum chuck. The width of the non-bonded portion of the template is 5 mm. The InP wafer was polished with a polishing liquid containing colloidal silica. The InP wafer was detached from the rotary platen, washed, and then removed from the plate. Cleaning water was applied to the plate surface and rubbed with a roll brush to clean the pad and template. The roll brush was also rotated and rubbed between the non-bonded portions. The brush was rotated while holding the rotation axis of the brush parallel to the pad surface of the polishing plate. A roll brush is a brush having bristles planted on a cylindrical surface. The tips are almost perpendicular to the axis. Brush tip diameter is 0.5 mm
φ. Since the roll brush has bristles on the cylindrical surface, the plate surface is rubbed while rotating, but there is a drawback that the bristles do not easily enter into the non-bonded portion.

In the next polishing of the wafer, backside oxidation was 0%. That is, there is almost no residue on the plate surface (pad surface), so the back surface is not oxidized. However, scratch-like scratches appeared on the surface of the InP wafer at a rate of 5%. It is likely that polishing particles and the like, which are slightly left in the gaps at the non-bonded portion of the template, will go around to the surface of the wafer in the next polishing and damage the surface of the wafer. So it is suggested that you should use a brush that can rub strongly on the non-bonded part between the template and the pad. It may be that a roll brush is not enough.

[Sample 12] Similar to sample 11, InP
The wafer was temporarily fixed on the pad having the resin template by a vacuum chuck. The width of the non-bonded portion of the template is 5 mm. The InP wafer was polished with a polishing liquid containing colloidal silica. The InP wafer was detached from the rotary platen, washed, and then removed from the plate. The pad and template were cleaned by applying washing water to the plate surface and rubbing with a vertical brush. A vertical brush is a brush in which a disk is attached to the tip of the shaft and hair is evenly planted on the disk. Although it is a common brush, it is called a "vertical brush" here to distinguish it from a roll brush. Let θ be the angle of inclination between the brush shaft and the normal to the pad surface of the polishing plate. Θ = 0 means that the brush axis and the pad normal are parallel.

The bristle tips are parallel to the axis or slightly inclined. In contrast to roll brushes where the bristle tips were perpendicular to the axis. The diameter of the bristles of the brush is 0.5 mmφ.
The brush was rotated while holding the rotation axis of the brush parallel to the pad surface of the polishing plate. The vertical brush was also rotated and rubbed between the non-bonded portions. Since the vertical brush has the tips of the bristles planted parallel to the disk at the tip of the shaft, turning the shaft parallel to the pad normal (Θ = 0) causes the bristles to bend outwards, and You can enter the joint.
Therefore, the abrasive particles will not remain in the non-bonded portion.

In the next wafer polishing, backside oxidation was 0%.
Met. That is, there is almost no residue on the plate surface (pad surface), so the back surface is not oxidized. The occurrence rate of scratch-like scratches on the surface of the InP wafer was 1%. You can insert the bristle tip into the non-bonded part by holding the shaft perpendicular to the pad surface and rotating it,
It is speculated that this may be because particles and the like can be swept out. If the scratch occurrence rate is 5% or less, it can be said to be a good product. So sample 12 passes that criterion.

[Sample 13] Similar to Samples 11 and 12, an InP wafer was temporarily fixed on a pad having a resin template by a vacuum chuck. The width of the non-bonded portion of the template is 5 mm. The InP wafer was polished with a polishing liquid containing colloidal silica. The InP wafer was detached from the rotary platen, washed, and then removed from the plate. Cleaning water was applied to the plate surface of the polishing head, and the pad and template were cleaned by rubbing with a vertical brush. Vertical brush same as sample 12 (brush tip diameter 0.5 mm
φ), but with the axis tilted 10 degrees from the normal (Θ = 10
The pad surface was rubbed with (). As mentioned earlier, the vertical brush has the tip of the bristles planted parallel to the axis on the disk at the tip of the shaft, but a part of the tip of the bristles is inclined to the outside of the axis. Since the hair is deformed to, the hair tips will get into the non-bonded portion more than necessary. If so, it may be more effective to tilt the axis of the vertical brush. So here we are tilting 10 degrees. When the tip of the brush is pressed firmly against the pad, the tip of the bristle bends appropriately and can reach the non-bonded portion under the template. Therefore, the abrasive particles will not remain in the non-bonded portion.

In the next wafer polishing, backside oxidation was 0%.
Met. That is, there is almost no residue on the plate surface (pad surface), so the back surface is not oxidized. The appearance rate of scratch-like scratches on the surface of the InP wafer was 1%, which was similar to that of Sample 12. Since particles and the like do not remain in the non-bonded portion, they hardly wrap around the surface and damage the wafer surface.

[Sample 14] Similar to Samples 11, 12, and 13, an InP wafer was temporarily fixed on a pad having a resin template by a vacuum chuck. The width of the non-bonded portion of the template is 5 mm. The InP wafer was polished with a polishing liquid containing colloidal silica. The InP wafer was detached from the rotary platen, washed, and then removed from the plate. Cleaning water was applied to the plate surface of the polishing head, and the pad and template were cleaned by rubbing with a vertical brush. Vertical brush same as sample 12 (brush tip diameter 0.5m
mφ), but with the axis tilted 20 degrees from the normal (Θ = 2
The pad surface was rubbed at 0 °). The only difference from sample 13 is that the inclination angle of the brush is increased from 10 ° to 20 °.

Similar to the sample 13, the tip of the slanted brush can enter the non-bonded portion of the template and scrape off the residue. In the next wafer polishing, backside oxidation was 0%. Similar to the samples 12 and 13, there is almost no residue on the plate surface (pad surface), and the back surface is not oxidized. The appearance rate of scratch-like scratches on the surface of the InP wafer was 1%, which was similar to that of Samples 12 and 13. Since particles and the like do not remain in the non-bonded portion, they hardly wrap around the surface and damage the wafer surface.

[Sample 15] An InP wafer was polished using a polishing plate having a resin template attached to a pad. This also polished the InP wafer with a polishing liquid containing colloidal silica. After polishing, the wafer was removed from the polishing plate, and the pads and template of the polishing plate were washed with water using a vertical brush as in Samples 12-14. The inclination angle Θ of the brush was set to 30 °. The other conditions are the same as those of Samples 12 to 14. The width of the unbonded part of the template is 5 mm, and the brush tip is 0.5 mm
It was φ. In the next wafer polishing, the rate of occurrence of backside oxidation of the wafer was 0%. In addition, the rate of occurrence of scratches was 1%. This is also a good product.

[Sample 16] This is also physically polished with a polishing liquid containing colloidal silica. After polishing, the pad and template on the back surface of the polishing plate were washed with water and rubbed with a vertical brush having a tip diameter of 0.5 mmφ. The angle of inclination of the brush was 45 °. In this case as well, no residue remained on the non-bonded portion of the template, and the template could be cleaned well. In the next wafer polishing, surface oxidation was 0%. Scratch-like scratches occurred at a rate of 5%. 5% for scratch
Since the following conditions are imposed, it can be seen that the vertical brush can be inclined up to 45 °. If it is further inclined, the hair tips will not enter the non-bonding portion of the template, and the silica particles accumulated here may not be completely removed.

[Sample 17] Samples 17 to 22 were examined for suitability by changing the width of the non-bonded portion of the template between 0 mm and 7 mm. Sample 17 has a non-joint width of 0 mm
It is what The resin template was snugly attached to the pad of the polishing plate. That is, the inner edge of the template is completely bonded to the pad and there is no unbonded portion. Therefore, the part of the adhesive that sticks out inevitably remains. This may affect the vacuuming of the wafer and cause particle retention. So this is an important test. After polishing the wafer, the back surface of the polishing plate was rinsed and cleaned with a vertical brush (tilt angle: 0 ° to 45 °).

In the next polishing of the wafer, backside oxidation was 0%. The scratch-like scratch was 1%.
In that respect, it is almost the same as the samples 12 to 15. However, in Sample 17, an abnormality was found in the peripheral portion of the surface of the wafer. It is presumed that this is because the adhesive material squeezed out on the inner edges of the template and the pad, and the particles contained in the polishing liquid remained there. It may have reached the surface of the wafer and affected the polishing state of the peripheral portion of the wafer. Although the angle of inclination of the vertical brush was variously changed, there was a peripheral abnormality in any way. Therefore, it can be seen that some non-bonding portion must be provided in the bonding between the template and the pad.

[Sample 18] Also, a polishing plate having a pad adhered thereto and a template adhered thereto with an adhesive was used. The width of the non-bonded portion was 0.5 mm. The InP wafer was temporarily fixed in the template with a vacuum chuck, polished with a polishing liquid containing colloidal silica, the wafer was removed, and the pad and the template were washed with water by rubbing with a vertical brush having a tip diameter of 0.5 mmφ. Backside oxidation is 0% in the next wafer polishing
And scratches on the surface were 1%. There were no peripheral abnormalities. This is a pass. This means that the width of the non-bonded portion needs to be about 0.5 mm. This is to prevent the adhesive material from squeezing out, and to prevent the polishing particles from remaining.

[Sample 19] This is also similar to Samples 17 and 18, except that the width of the non-bonded portion is set to 1 mm when the template is attached to the pad of the polishing plate. It is about twice as much as the previous example. If the non-bonding width is increased, the possibility of the adhesive squeezing out becomes lower, but there is a possibility that the abrasive particles may penetrate further and remain. Therefore, I changed the non-bonding width and investigated such a possibility. After polishing the wafer, remove the wafer and wash it with water in the same way as the previous samples.
The pad and template were rubbed and cleaned by (). Next time, the polishing condition of the wafer was examined, but there was no abnormality in the peripheral portion. Oxidation on the back surface was 0%, and the occurrence rate of scratches on the surface was 1%. This is also a good product.

[Sample 20] The width of the non-bonded portion of the template was 5 mm. It is a fairly wide non-joint.
Therefore, abrasive particles may penetrate deeply. After polishing the wafer, the wafer was removed and the pad and template were rinsed and cleaned with a vertical brush. The same test was performed to examine the polishing state of the next polished wafer. Backside oxidation is 0
%Met. There was no peripheral abnormality. Occurrence of scratches was 1%, which was the same result as in Samples 17 to 19. It means that even if the non-bonded portion is as deep as 5 mm and colloidal silica enters, the particles can be swept out by rubbing with a brush.

[Sample 21] The width of the non-bonded portion was further increased to 6 mm. Since it is a wide non-bonded part, polishing liquid may enter and remain. After polishing the wafer, the pad and template were cleaned with a vertical brush (Θ = 0 to 45 °). In the next polishing of the InP wafer, there was no peripheral part abnormality. Furthermore, scratches are 2% for scratches and sample 17
A little more than ~ 20. The backside oxidation was 0%. This is also acceptable because the allowable percentage of scratches is 5%.
However, it can be seen that the possibility of scratches increases if the unbonded portion is too wide. In addition to that, the lift of the template from the pad becomes large, and the wafer enclosing action of the template may become unstable.

[Sample 22] The width of the non-bonded portion was further increased to 7 m.
Increased to m. After polishing the wafer, the pad and template were rinsed and cleaned with a vertical brush. In the next polishing of the InP wafer, there was no peripheral part abnormality. No backside oxidation
%Met. The rate of occurrence of scratches on the surface was 5%. This is probably because some of the adhered abrasive particles and the wafer shavings entered and remained in the deep non-bonded portion and spilled over to the surface of the wafer during the next polishing, which created scratches. Although the acceptable scratch rate is 5% or less, this is a passing product, but it is presumed that it is not good to expand the non-joint beyond that. From these results, it is understood that the width of the non-bonded portion is preferably about 0.5 mm to 7 mm.

[Sample 23] In Samples 23 to 25, the conditions for cleaning the pad of the polishing plate were changed. The template is provided with a non-bonded portion having a width of 6 mm. Sample 2
In all of Nos. 3 to 25, the non-bonding width is set to 6 mm and the conditions are made uniform for easy comparison. An InP wafer was polished using this polishing plate. After removing the wafer after polishing, the back surface of the polishing plate was rubbed with a vertical brush having a tip diameter of 0.5 mm to clean the surface while discharging pure water.

In the next wafer polishing, oxidation on the back surface was 0% and scratches on the front surface were 0%. This is the first time I have explained that scratches did not enter the surface. Since there was no residue on the back surface of the polishing plate or the non-bonded portion of the template, the next wafer polishing was good. Since the cleaning liquid is pure water and has no chemical action, it has a sufficient cleaning action.

[Sample 24] Similar to Sample 23, an InP wafer was polished using a polishing plate to which a template was attached so that the width of the non-bonded portion was 6 mm. The tip diameter is 0.
A pad and template were cleaned by rotating a 5 mmφ vertical brush. The cleaning liquid is a mixture of pure water and 0.1% KOH. In the next polishing of the InP wafer, the back surface oxidation was 0%. However, scratches on the surface appeared at a rate of 1%. The difference from the sample 23 is that it is pure water or KOH solution. Scratch is 1%, and 5% or less is also a good product.

[Sample 25] An InP wafer was polished using a polishing head in which a template was attached to a pad so that the width of the non-bonded portion was 6 mm as in Samples 23 and 24. After polishing, the wafer was removed, and the pad and the template were cleaned using a vertical brush having a tip diameter of 0.2 mmφ while spraying pure water. I use a brush whose tip is smaller than the previous brush (0.5 mm diameter). This is because it is expected that the thinner tips will penetrate deeper into the non-bonded part. In the next polishing of the wafer, backside oxidation was 0% and frontside scratches were 0%. We were able to give very good results. From the results of Samples 23 and 25, even if the tip of the brush is as thick as 0.5 mmφ,
It means that a thin one such as 0.2 mmφ is also effective.

[Sample 26] All the items described so far are obtained by cleaning and washing the pad after each polishing. In order to investigate whether or not the pad cleaning should be performed every time the wafer is polished, Samples 26 to 28 were changed in cleaning frequency. In both cases, a polishing plate is used in which a resin template is bonded to a pad so that the width of the non-bonded portion becomes 6 mm. After polishing the wafer, the pad was cleaned with a vertical brush with a 0.5 mm tip diameter. For sample 26, the polishing plate and pad were cleaned and cleaned with a brush and pure water every time the wafer was polished. According to this, in the next polishing, the back surface oxidation of the wafer was 0%, and the scratch-like scratch generation on the surface was 1%. This is acceptable because it is 5% or less.

[Sample 27] Under the same conditions as Sample 26 (width of non-bonding portion: 6 mm, brush bristle tip diameter: 0.5 mm), but pad cleaning was not performed at each polishing, and 1 was obtained after polishing three sheets.
I decided to wash and clean only once. Cleaning the pad by rotating the vertical brush remains unchanged.
In the cleaning cycle for every three wafers, the state of the third wafer was examined. According to this, backside oxidation of the wafer was 1%. The residues of the previous polishing and the previous polishing may remain on the intermediate portion or the non-bonded portion between the template and the wafer, and they may adhere to the back surface of the wafer in the next polishing. Further, scratches on the surface occurred at a rate of 3%. This is also because the residue wraps around.

[Sample 28] The conditions are the same as those of Samples 26 and 27, but 1 is obtained every time 6 sheets are polished instead of after every polishing.
I decided to clean it with a vertical brush while spraying pure water only once. In the cleaning cycle for every six wafers, the state of the sixth wafer was examined. Backside oxidation was 2%, and the occurrence rate of surface scratch scratches was 6%. Since the allowable rate of scratches is 5%, it means that cleaning every 6 sheets is not enough. It was found that it is best to clean the pad and template every time, but it is also possible to clean every three sheets.

[0114]

The polishing technique for Si wafers has been matured so far. However, GaAs wafers are required to have a new polishing technique as the diameter increases. Further, until now, the polishing technique has not been settled for the InP wafer, which has only a small demand as a substrate for the light receiving element. The possibility of demand other than the light receiving element has emerged, and 2-inch InP wafers have been manufactured. For polishing, a systematic technique has been required in place of the manual operation. GaAs and InP wafers have lower hardness than Si wafers and are difficult to polish. As with Si, physical polishing is often performed using a polishing liquid containing particles. Polishing takes longer than Si (about 1 to 2 minutes) (about 10 minutes)
Therefore, the abrasion of the polishing cloth is significant and scratches on the wafer surface become a problem.

In the case of a GaAs or InP wafer having a small diameter, it was attached to a polishing plate with wax and polished. After polishing, the wax was melted by heating and the wafer was removed from the polishing plate and washed with an organic solvent to remove the wax. That was fine as long as the production was low.

However, when the diameter is increased to 2 inches and 3 inches and the production amount is increased, the waxless plate fixing is desired. If it exceeds 4 inches, I definitely want to do waxless polishing. But Ga
Waxless polishing of As and InP wafers has never been seen before. This is because there are many disadvantages in the case of GaAs and InP, which have lower rigidity than Si and are easily chipped and damaged.

The present invention is aimed at fixing a GaAs or InP wafer by waxless fixing so that it can be polished. However, a GaAs or InP wafer (compound semiconductor wafer) is placed on a pad and evacuated. May cause a lateral shift. You need a template to prevent that. From the standpoint of preventing misalignment, the gap between the template and the wafer should be narrow. However, if the gap is narrow, protrusion of the adhesive material of the template becomes a problem.
The protrusion of the adhesive material is a problem because it obstructs the close adsorption of the wafer and promotes the stubborn adhesion of dust. Therefore, it seems better to make a part of the inner circumference of the template a non-bonded part. When the non-bonded portion is formed, the adhesive material can be prevented from protruding, but polishing dust (residue) enters into the inner depth, which causes a new problem that it is difficult to remove.

According to the present invention, not only the washing water is sprayed but also the brush is forcibly cleaned so that the pad and the template are not left with abrasive particles and the like, and the next polishing can be performed well. It is an excellent invention.

[Brief description of drawings]

FIG. 1 is a cross-sectional view of a vacuum chuck type polishing plate, a wafer and a polishing platen used for polishing an InP wafer in the present invention.

FIG. 2 is a cross-sectional view showing a state in which after polishing an InP wafer, a polishing plate is lifted and a cleaning liquid is sprayed from a cleaning nozzle onto a polishing surface of the wafer to clean the wafer surface.

FIG. 3 is a perspective view showing a state in which a wafer is removed after polishing an InP wafer, and a pad and a template are cleaned and washed by a rotating brush having bristles on a cylindrical surface while spraying a cleaning liquid.

[FIG. 4] After polishing the InP wafer, the wafer is removed, and a pad and template are cleaned and cleaned by spraying a cleaning liquid and making a rotating brush and a normal line of the vertical brush parallel to the disk parallel to each other. FIG.

FIG. 5: After polishing the InP wafer, the wafer is removed, and a pad and template are cleaned and washed by spraying a cleaning liquid and vertically rotating brushes of a vertical brush on the disk so that the rotation axis and the polishing plate normal line are inclined. FIG.

[FIG. 6] After polishing the InP wafer, remove the wafer, spray a cleaning liquid, and use a smaller vertical brush that brushes the disk substantially perpendicularly so that the rotation axis and the polishing plate normal line are parallel to each other, and the pad and template are attached. The perspective view which shows the state which is cleaning and washing.

FIG. 7 is a cross-sectional view showing a bonded state of the template and the pad on the back surface of the polishing plate. The outside part of the template is adhered to the pad by an adhesive material, but there is a non-bonded part that is not adhered inside the template.

FIG. 8 is a cross-sectional view showing a state in which a residue such as polishing particles or a wafer polishing piece generated by polishing enters and remains in a non-bonding portion between the template and the pad on the back surface of the polishing plate.

FIG. 9 is a cross-sectional view showing that a brush tip of a brush enters a non-bonded portion between a template and a pad on the back surface of the polishing plate, thereby acting to remove a residue sandwiched by the non-bonded portion.

FIG. 10: A wafer is attached to a pad while the balloon is inflated with the balloon interposed between the polishing plate bottom surface and the pad, the wafer is removed, and then the balloon is retracted to distort the pad into a concave surface. FIG. 7 is a cross-sectional view of the improved polishing plate for expanding the non-bonded portion to remove the residue, when the balloon is expanded.

11 is a cross-sectional view of the improved polishing plate of FIG. 10 with the balloon retracted.

FIG. 12 shows an improved polishing plate in which the polishing plate is made thicker and only the central portion of the pad is attached to the plate surface, and the pad is attached to the attaching member at the peripheral portion of the pad so that it can be expanded and contracted by a balloon attached to the ring member. FIG. 3 is a cross-sectional view of a state in which the balloon is degenerated and the pad is made flat.

13 is a cross-sectional view of the polishing plate of FIG. 12 with the balloon ring expanded and the pad deformed into a concave surface.

[Explanation of symbols]

1 rotating surface plate 2 polishing cloth 3 polishing head 4 Polishing plate 5 head axis 6 adsorption pad 7 templates 8 InP wafer (or GaAs wafer) 9 recess 10 exhaust passage 11 exhaust passage 12 holes 13 Wafer backside 14 Template inner surface 15 Wafer outer peripheral surface 16 Wafer surface 17 Adhesive 18 Template surface 19 Raised part 20 washing nozzle 21 cleaning liquid 22 residue 25 axes 26 cylinder brush 27 round body 28 Flocking 29 Non-joint 30 vertical brushes 31 disk 32 template outer peripheral surface 33 Flocking 34 Hair tips 35 Support rod 36 plate normal 37 rotation axis 40 vertical brush 41 disc 43 hair transplant 44 Hair tips 45 Support rod 50 vertical brushes 51 disk 53 hair transplant 54 Hair tips 55 Support Rod 61 Disc part 63 hair transplant 64 hair tips 70 Cylindrical part 71 Lower end of cylindrical part 72 balloon 73 small groove 74 Internal space 75 Guide plate 76 groove 78 pad hole 79 gas passage 82 Column 83 outer peripheral surface 84 balloon 85 Hold ring 86 Lower surface of retaining ring 87 Lower surface of column 88 flange 89 Exhaust passage 90 pad holes 91 gas passage 92 gas passage

─────────────────────────────────────────────────── ─── Continuation of front page (51) Int.Cl. ⁷ Identification code FI theme code (reference) B24B 37/04 B24B 37/04 Z 55/06 55/06 F term (reference) 3C047 FF08 HH00 3C058 AA07 AB04 AB06 AC05 CB03 DA17

Claims (9)

[Claims] 1. A polishing plate having a flat lower surface, a head shaft continuing above the polishing plate, a suction pad attached to the lower surface of the polishing plate for sucking a wafer, and a suction pad for preventing lateral displacement of the wafer. The InP wafer or GaAs wafer is adsorbed to the adsorption pad surface surrounded by the template of the polishing head, which includes a template in which a part of the inner peripheral portion is a non-bonded portion and the remaining portion is bonded or thermocompression bonded, and the polishing head is polished. The wafer is polished by pressing it against the platen and supplying the polishing liquid to the polishing platen, rotating the polishing platen and the polishing head, and lifting the polishing head to clean the polished surface of the wafer and removing the wafer from the suction pad. The wafer is polished by removing it, and after polishing the wafer, every one or every two or three wafers, the pad of the polishing head and the template are polished. The method for polishing a compound semiconductor wafer is characterized in that the wafer is cleaned with a brush that rotates or swings while spraying a cleaning liquid, and the next wafer is polished. 2. The width of the non-bonded portion of the inner peripheral portion where the template is not bonded to the suction pad is 0.5 mm to 7 mm.
And a method of polishing a compound semiconductor wafer. 3. The brush is a vertical brush in which bristles are planted in a direction perpendicular to the disc, and the vertical brush is rotated such that the inclination angle of the brush rotation axis with respect to the suction pad normal is 0 ° to 45 °. Polishing plate pad by,
The template is cleaned, and the template is cleaned.
A method for polishing a compound semiconductor wafer according to any one of 1. 4. The method for polishing a compound semiconductor wafer according to claim 1, wherein the polishing liquid contains polishing particles. 5. The method for polishing a compound semiconductor wafer according to claim 1, wherein the template is a resin, a metal covered with a resin, or a ceramic covered with a resin. 6. The pad is deformed into a concave shape in order to widen the non-bonded portion and easily remove the residue sandwiched in the non-bonded portion when cleaning the suction pad and the template. 7. A method for polishing a compound semiconductor wafer according to claim 1. 7. A polishing plate having a flat lower surface, a head shaft continuing above the polishing plate, a suction pad attached to the lower surface of the polishing plate for sucking a wafer, and a suction pad for preventing lateral displacement of the wafer. Part of the inner circumference 0.5
mm to 7 mm as a non-bonded portion and a polishing head including a ring-shaped template in which the remaining portion is adhered or thermocompressed, a polishing platen with a polishing cloth attached, and a cleaning liquid is applied to the pad of the polishing plate when the polishing plate is lifted. The cleaning liquid supply device for spraying, and the diameter of the tip of the hair for cleaning the pad and template of the lifted polishing plate are 1 mm.
A polishing apparatus for a compound semiconductor wafer, comprising: a rotating or oscillating brush having flock that can enter the non-bonding portion. 8. The polishing plate has a cylindrical shape with an open lower surface, and a balloon is built in the cylinder of the polishing plate, and a passage for guiding gas to the balloon is provided inside the polishing head, and the polishing plate is connected to a vacuum pump for polishing. The peripheral part of the pad is held by the end of the cylindrical part of the plate, and the central part of the pad is pushed out by the balloon.When the wafer is polished, the balloon expands to make the suction pad a flat surface, and when cleaning the suction pad, the balloon shrinks. The polishing apparatus for a compound semiconductor wafer according to claim 7, wherein the pad surface is made concave to facilitate removal of the residue sandwiched in the non-bonded portion between the template and the suction pad. 9. The polishing plate has a cylindrical portion in the lower central portion, and an annular retractable balloon and a retaining ring are fitted on the outer peripheral surface of the cylindrical portion, and a passage for guiding gas to the balloon is provided inside the polishing head. It is installed and connected to a vacuum pump, the central part of the pad is attached to the lower surface of the column, the peripheral edge of the pad is attached to the lower surface of the restraining ring, and the peripheral portion of the suction pad is pushed out by a balloon. At times, the balloon was degenerated to make the suction pad flat, and when cleaning the suction pad, the balloon was expanded to make the pad surface concave so that the residue sandwiched between the non-bonded portion between the template and the suction pad could be easily removed. The compound semiconductor wafer polishing apparatus according to claim 7.