Patent Application
20080057608
The adhering manner by which stripe direction S is perpendicular to rotation direction G is shown in
That is, the manufacturing method of substrate 1 being a group III nitride substrate in one embodiment of the present invention, as shown in
More desirably, the range of the compressibility of pad 51 is 1%-10%. More preferably, the range is 1%-3%.
More preferably, the pressure applied from pad 51 to substrates 1 is 300 g/cm2 (29.4 kPa)-1000 g/cm2 (98 kPa).
More preferably, the range of pH of abrasive liquid 54 is pH=1.5-10. Further suitably, pH=2-6. It is desirably acid. Alkali selectively corrodes a crystal defect gathering region 13. This produces protrusions and recesses on the surface. Accordingly, acid abrasive liquid 54 is suitable.
When employing an acid abrasive liquid, inorganic acid such as hydrochloric acid, sulfuric acid, nitric acid or the like can be used. Organic acid such as citric acid, malic acid or the like can also be used. Organic acid has a weak effect and does not corrode crystal defect gathering region 13, and therefore advantageous in maintaining the flatness of substrates 1. Inorganic acid and organic acid may be salts of inorganic acid and salts of organic acid, as combined with metal elements.
Indices for evaluating a mirror wafer after grinding, lapping and polishing are flatness, over-abraded circumference, surface roughness and the like.
Flatness means the proportion of the area having an off angle θ of less than 0.1° relative to the total area of low defect single crystal region 11 (including C-plane growth region 12, if any). The off angle is an angle formed between direction T perpendicular to stripe direction S and the surface of the substrate (wafer). Now, substrate 11 that is a stripe type substrate having the (HZ)m structure shown in
Flatness (%)=a/z×100
Flatness 80% means that the area having an off angle θ of less than 0.1 in direction T perpendicular to the stripes occupies 80% of the total low defect single crystal region 11. Herein, the definition is different than the normal wafer flatness.
Flatness is an index for evaluating a wafer, and it is not a definition of grinding, lapping and polishing. Only the flat portion can be used for manufacturing devices. That is, in a wafer having low flatness, the area that can be used for manufacturing devices is small. The present invention requires that the flatness of a wafer after abrading is at least 40%. Further preferably, it is at least 60%. Still further preferably, it is at least 80%.
Over-abraded circumference means that the circumference of a wafer is abraded and thus becomes lower than the surface, whereby the whole wafer is formed in a convex shape. This also reduces the effective area. On the other hand, flatness is slightly different from over-abraded circumference. As to the stripe structure, since crystal defect gathering region 13 is physically and chemically weak, in some cases it becomes a recess by grinding, lapping and polishing. This is evaluated by the flatness. As the recess herein is the one formed at each crystal defect gathering region 13, it is different from over-abraded circumference that is limited to the circumference.
Surface roughness is also an evaluation index of a mirror wafer. There are various surface roughness such as Rmax, Ra, Rz, Ry, Rms (JIS standard) and the like. Herein, Ra is employed as an index of surface roughness. It is obtained as an average of absolute values of differences between crests and troughs formed on the wafer surface. The extent of demand for the surface roughness varies depending on the purpose. The present invention requires that the Ra of a wafer after polishing is at most Ra 2.0 nm. More desirably, it is at most Ra 0.9 nm.
In some cases, an epitaxial wafer in which a group III nitride layer is epitaxially grown on the mirror wafer is manufactured to examine photoluminescence (PL) intensity for evaluation.
In some cases, a device is manufactured by epitaxially growing a group III nitride layer on the mirror wafer and providing electrodes. By actually supplying power to the device so that it emits light, the performance of the device is examined. Thus, how the wafer is finished is checked. When devices are manufactured, the proportion of conforming devices is referred to as a yield. The same wafer provides different yields depending on the target device. The present invention requires that the yield when manufacturing blue laser (430 nm) is at least 35%.
If stripe direction S of a group III nitride crystal wafer (substrate 1 or 2) having the stripe structure is oriented parallel to rotation direction G of abrading holder 53, physically and chemically weak crystal defect gathering region 13 is selectively abraded and/or polished. This results in poor flatness and surface roughness. In contrast, according to the present invention wafers are adhered so that stripe direction S is perpendicular to rotation direction G of abrading holder 53. Since abrading holder 53 is rotated on its own axis and abrading surface plate 52 is revolved about a prescribed point, not always the relative movement direction between substrate 1 and pad 51 is rotation direction G of abrading holder 53.
On the other hand, when averaged for a period of time, it can be seen that the relative movement direction between substrate 1 and pad 51 is rotation direction G of abrading holder 53. Pad 51 relatively moves with reference to substrate 1 so as to be perpendicular to stripe direction S of substrate 1, and therefore the particularly weak crystal defect gathering region 13 is not selectively corroded and abraded. Therefore, according to the present invention, the reduced amounts of low defect single crystal region 11, crystal defect gathering region 13 and C-plane growth region 12 are averaged and the flatness is maintained. Surface roughness can also be prevented from becoming poor.
Setting of pH of abrasive liquid 54 is also important. The strong alkaline abrasive liquid selectively erodes crystal defect gathering region 13 that is chemically weak. Accordingly, crystal defect gathering region 13 is recessed. This results in poor flatness and surface roughness of substrate 1. Therefore, the pH range of abrasive liquid 54 is 1-12. More preferably, the pH range of abrasive liquid 54 is 1.5-10. Selective corrosion of crystal defect gathering region 13 is small when acid abrasive liquid 54 is employed and, therefore, more preferably the range is 2-6. Thus, respective thicknesses of low defect single crystal region 11, crystal defect gathering region 13 and C-plane growth region 12 are reduced in substantially the same proportion, whereby the flatness is maintained and the surface roughness is low. Strong acid of pH<1 is not preferable, since crystal defect gathering region 13 is also corroded to form a recess.
If the compressibility of pad 51 is too high, pad 51 enters a recess formed by corroded crystal defect gathering region 13, whereby reduction in the thickness of crystal defect gathering region 13 further progresses. On the other hand, if the compressibility of pad 51 is low, pad 51 is hard and does not expand and contract. When pad 51 is too hard, a shock is likely to occur and a scratch easily occurs in substrate 1. Substrate 1 may possibly be damaged. Based on such reasons, while lower compressibility of pad 51 is suitable, too much hardness is disadvantageous. Accordingly, the compressibility of pad 51 should be at least 1%.
Therefore, the compressibility range of pad 51 is 1%-15%. In this range, crystal defect gathering region 13 is not selectively worn. Hence, flatness is protected. More preferably, the compressibility range is 1%-10%. Most preferably, it is 1%-3%. Compressibility of pad 51 can be determined by the following expression, using a thickness T1 that is a thickness one minute after an initial load W1 is loaded, and a thickness T2 that is a thickness one minute after the load is increased to load W2:
Compressibility (%)=(T1-T2)/T1×100
100 g/cm2 is employed as W1, and 1800 g/cm2 is employed as W2.
Pressure is 100 g/cm2 (9.8 kPa)-1500 g/cm2 (147 kPa). More preferably, pressure is 300 g cm2 (29.4 kPa)-1000 g cm2 (98 kPa).
When a wafer (substrate 1 or 2) is polished under such conditions, a mirror wafer having flatness of at least 40% and surface roughness of at most Ra 2.0 nm can be obtained.
When a group III nitride layer is epitaxially grown on the mirror wafer, an epitaxial layer having excellent crystallinity and morphology can be formed. When LDs are manufactured, the yield of at least 35% can be achieved.
The group III nitride mirror wafer (group III nitride substrate) of the present invention can be used as a substrate for semiconductor devices such as follows.
There are semiconductor devices such as light emitting elements (light emitting diodes, semiconductor lasers), electronic elements (rectifiers, bipolar transistors, field effect transistors, HEMTs), semiconductor sensors (temperature sensors, pressure sensors, radiation sensors, visible-ultraviolet light detectors), SAW devices, acceleration sensors, MEMS components, piezoelectric oscillators, resonators, piezoelectric actuators and the like.
Substrate 7 that is a group III nitride substrate with an epitaxial layer in one embodiment of the present invention includes, as shown in
Group III nitride layer 3 is not particularly limited, and for example it may be a GaxAlyInl-x-y N layer (0≦x, 0≦y, x+y≦1). Also, the method of epitaxially growing group III nitride layer 3 is not particularly limited, and for example it may preferably be the HVPE (Hydride Vapor Phase Epitaxy, the same applies hereinafter) method, the MBE (Molecular Beam Epitaxy, the same applies hereinafter) method, the MOCVD (Metal Organic Chemical Vapor Deposition, the same applies hereinafter) method and the like. Before epitaxially growing group III nitride layer 3, etching and/or annealing of substrate 1 or 2 being the group III nitride substrate can be performed in an apparatus for the epitaxial growth, so as to modify the property of the surface of substrate 1 or 2.
A semiconductor device in one embodiment of the present invention includes at least one layer of group III nitride layer 3 formed on at least one main surface of substrate 1 or 2 being a group III nitride substrate, and an electrode formed at the group III nitride substrate (substrate 1 or 2) or group III nitride layer 3. The semiconductor device exhibits a high performance, since it is provided with at least one layer of group III nitride layer 3 which is an epitaxial layer being excellent in morphology and crystallinity on at least one main surface of group III nitride substrate.
A manufacturing method of a group III nitride substrate with an epitaxial layer in one embodiment of the present invention includes, as shown in
A manufacturing method of a semiconductor device in one embodiment of the present invention includes, as shown in
(The relationship between the wafer adhering direction, stripe direction S and rotation direction G)
An alumina block measuring 135 mm outer diameter and 30 mm thickness was employed as an abrading holder (abrasive plate). A wafer (substrate) to be subjected to grinding, lapping and/or polishing was a stripe type GaN substrate measuring 50 mm diameter and 0.5 mm thickness. Crystal defect gathering region 13 had a width h of 50 μm. Low defect single crystal region 11 had a width z of 350 μm. Pitch p was 400 μm. There was no C-plane growth region 12 (y=0). Three wafers to be simultaneously subjected to grinding, lapping or polishing constituted one set of samples.
The three GaN wafers were adhered to the abrading holder using thermoplastic solid wax. The abrading holder was heated to the temperature higher by 30° C. than the softening point of the wax so as to melt the wax. The three stripe GaN wafers were regularly adhered at the position where each periphery was distanced 5 mm from the circumference of the abrading holder (abrasive plate). The adhering direction relative to the stripes is as shown in Table 1.
“Parallel” means that stripe direction S of substrate 1 and rotation direction G of abrading holder 53 are parallel to each other as shown in
Four sample sets (samples 1, 2, 3 and 4) each constituted of three wafers were subjected to grinding. In samples 1 and 3, stripe direction S was perpendicular to rotation direction G (
After substrates 1 were adhered to the abrasive plate (abrading holder 53), samples 1-4 were subjected to grinding with diamond grindstone No. 2000. Samples 3 4 were subjected only to grinding. After grinding, samples 1 and 2 were further subjected to lapping using diamond loose abrasive grain having an average grain size of 2 μm. The maximum scratch depth, thickness of a work-affected layer, flatness after the machine working were measured.
Table 1 shows the measurement result. A scratch is a linear scar made by grinding, abrading and the like. Since samples 3 and 4 were subjected only to grinding with coarse bonded grindstone, scratches were deep. While the maximum scratch depth of sample 4 was 290 nm, the maximum scratch depth of sample 3 was 95 nm.
The thickness of work-affected layer of sample 4 was thick, being 7 μm (average). The thickness of work-affected layer of sample 3 was 4 μm, being reduced substantially by half. This is attributed to the difference in the adhering orientation of stripe direction S to abrading holder 53. It is desirable that the scratch is shallow and the work-affected layer is thin. Sample 3 showed a favorable result than sample 4. Hence, it is preferable that stripe direction S is perpendicular to rotation direction G (which is referred to as S perpendicular G).
Crystal defect gathering region 13 is the site where dislocations densely gather, which is physically and chemically weak. When ground or abraded, it is easily scarred. In contrast, low defect single crystal region 11 and C-plane growth region 12 are hard and sturdy, being physically and chemically robust. They are not easily ground or abraded.
When stripe direction S and rotation direction G are parallel to each other (
When stripe S and rotation direction G are perpendicular to each other (
Samples 1 and 2 were subjected also to lapping. In lapping, abrasive grain is finer than in grinding. Furthermore, since loose abrasive grain is used, scratches are abraded and become shallow. The maximum scratch depth of sample 1 was 21 nm, and that of sample 2 was 49 nm. The scar was shallow since in sample 1 rotation direction G was perpendicular to stripe direction S, and the regions having high surface hardness and the regions having low surface hardness were alternately present in the grinding and lapping direction. In sample 2, the scratch was deep since rotation direction G was parallel to stripe direction S (referred to as “S parallel G”), and the portion being low in the surface strength was firstly hollowed deeply. The work-affected layer of sample 1 was 0.5 μm, and that of sample 2 was 2 μm. Sample 1 was excellent also in this respect. Since rotation direction G was perpendicular to stripe direction S and the strength was averaged (HZHZH . . . or, HZYZH . . . ), the affected layer was thin.
Samples 1 and 2 were further subjected to CMP (Chemical Mechanical Polishing) using colloidal silica. After the processing, the surface roughness of sample 1 was Ra 0.5 nm and that of sample 2 was Ra 1.4 nm. The flatness of sample 1 was 65% and that of sample 2 was 45%. Flatness is an index of percentage of the area having an off angle of less than 0.10 relative to low defect single crystal region 11. A higher value is better. Sample 1 being “S perpendicular G” was lower in the surface roughness and higher in flatness than sample 2 being “S parallel G”. This is because stripe direction S is perpendicular to the rotation direction G of abrading holder 53.
In sample 1, crystal defect gathering region 13 was not recessed after machine abrading, and therefore showed no flatness problem. Among samples 1-4, sample 1 being “S perpendicular G” was the best. Comparing the samples subjected only to grinding, sample 3 was superior to sample 4. This can be explained similarly. That is, since sample 3 was “S perpendicular G”, the movement direction of the grindstone crossed crystal defect gathering region 13 and low defect single crystal region 11 and grinding was averaged.
With sample 1 (“S perpendicular G”) in Example 1, conditions were varied (six types) and CMP was performed. Specifically, an examination was carried out varying pH of CMP abrasive liquid.
Sample 1 was prepared as follows. On an alumina block measuring 135 mm outer diameter and 30 mm thickness as abrading holder 53, three GaN wafers (substrates) measuring 50 mm φ diameter and having stripes with 400 μm pitch were adhered by solid wax so that stripe direction S was perpendicular to rotation direction G (radially). Then grinding and lapping were performed. After grinding and diamond lapping were performed with diamond grindstone No. 2000 (JIS standard R6001 microgrits for precision abrading #2000), CMP was performed under the conditions shown in Table 2. Samples 5-14 were obtained.
For each of the samples 5-14, the loose abrasive grain was colloidal silica (SiO2). pH was 0.8 for sample 5; 1.0 for sample 6; 1.5 for sample 7; 2.0 for samples 8, 9 and 14; and 6.0 for sample 10. That is, they were acid. pH was 10 for sample 11; 12 for sample 12; and 13 for sample 13. That is, they were alkaline. Additives for pH adjustment were hydrochloric acid for samples 5-7, nitric acid for samples 8 and 9, carbonic acid for sample 10, potassium hydroxide for samples 11-13, and malic acid for sample 14.
The compressibility of the pad was 15% for samples 5-7 and 9-14. It was 8% for sample 8. The pressure on a wafer (substrate) was 300 g/cM2 (29.4 kPa) for each of samples 5-14. After CMP, the flatness and surface roughness Ra of the samples were examined. Flatness is the proportion of the area having an off angle of less than 0.1°. At least 40% is necessary. At least 60% is desirable.
While it is not simple in consideration of other conditions such as pad compressibility, as based solely on this result, samples 5 and 13 respectively showed 38% and 35%, indicating low in flatness and not being conformable. Samples 6 and 12 respectively showed 45% and 40%, being conformable. Samples 7 and 11 respectively showed 55% and 58%, being further excellent. Samples 9 and 10 respectively showed at least 60%, being further excellent Samples 8 and 14 respectively showed at least 80%, being further excellent. In order to improve flatness, the abrasive liquid must not be strong acid or strong alkaline, pH of 1-12 provides flatness of at least 40%. pH of 1.5-10 provides flatness of at least 55%. Desirably, the abrasive liquid is acid. When at least 60% flatness is to be obtained, pH should be 2-6 (acid). Flatness can further be improved by using organic acid rather than inorganic acid.
As to surface roughness, samples 5 and 13 showed 1 nm or higher, being relatively poor. Samples 6, 7, 11 and 12 were relatively good. Samples 8, 9, 10 and 14 showed 0.6 nm or lower, being further excellent. When abrasive liquid is strongly acid or alkaline, selectively corrodes crystal defect gathering region 13, whereby surface roughness and flatness become poor.
Suitable pH is 2-6. Depending on the purpose, pH of about 1-12 may be used. As to acid abrasive liquid, flatness is more improved by use of organic acid such as malic acid or citric acid than by use of inorganic acid. While inorganic acid selectively corrodes crystal defect gathering region 13, organic acid less exhibits such an action. This is considered to be contributing to the flatness.
After CMP was performed, an epitaxial layer was deposited thereon through the MOCVD method. The PL (photoluminescence) light emission was examined. Epitaxial property is PL strength (in arbitrary unit). Samples 5 and 13 respectively showed 35 and 39, being particularly weak. Samples 6, 7, 11, and 12 respectively showed 48, 61, 62 and 45, exhibiting sufficient PL strength. Samples 8, 9, 10 and 14 showed considerably high PL strength (84, 73, 68, and 86). Samples 5 and 13 had their crystal defect gathering region 13 corroded by strong acid or alkali, resulting in poor flatness and high roughness. This resulted in poor quality of the epitaxial layer and low PL strength.
Further, various nitride layers were epitaxially grown on the substrate, and electrodes were provided to obtain blue laser devices having a wavelength of 430 nm. It was separated into chips and property of LD light emission and the yield of conforming items were examined. Samples 5 and 13 were respectively 28 and 32, being particularly poor. This is because of the strong acid and alkaline abrasive liquids. Samples 11 and 12 were respectively 53 and 37, being relatively poor. This may also be understood that the alkaline abrasive liquid corroded crystal defect gathering region 13 and flatness and surface roughness became poor. Sample 14 showed a 74% yield, being excellent. Samples 8, 9 and 10 respectively showed 73%, 62%, and 57% yields, from which it can be seen that excellent substrates were obtained.
With sample 1 (“S perpendicular G”) in Example 1, conditions were varied (five types) and CMP was performed. Specifically, pad compressibility was varied. Since crystal defect gathering region 13 is weak and easily abraded, a pad that easily deforms may enter the abraded and recessed crystal defect gathering region 13, thereby further deeply abrade crystal defect gathering region 13.
Sample 1 was prepared as follows. On an alumina block measuring 135 mm outer diameter and 30 mm thickness as abrading holder 53, three GaN wafers (substrates) measuring 50 mm φ diameter and having stripes with 400 μm pitch were adhered by solid wax so that stripe direction S was perpendicular to rotation direction G (radially). Then grinding and lapping were performed. After grinding and diamond lapping were performed with diamond grindstone No. 2000, CMP was performed under the conditions shown in Table 3. Samples 15-21 were obtained.
For each of the samples 15-21, the loose abrasive grain was colloidal silica (SiO2). For each of the samples 15-21, pH=11 (alkaline). It had been known that pH=11 was excessively strong alkaline and not in a suitable range in Example 2. On the other hand, this knowledge was based on the high pad compressibility such as 15% or 8%. Therefore, considering that a lower pad compressibility might yield a favorable result, a further experiment was conducted with the bad condition of pH=11. The pH adjusting component of the abrasive liquid was KOH.
Pad compressibility was 0.8% for sample 15; 1.0% for sample 16; 1.5% for sample 17; 3% for sample 18; 10% for sample 19; 15% for sample 20; and 20% for sample 21. That is, the compressibility is increased in this order.
The pressure on a wafer was 300 g/cm2 (29.4 kPa) for each of samples 15-21.
After CMP, the flatness and surface roughness Ra of the samples were examined. Sample 21 showed flatness 35%, being the lowest. Sample 20 showed flatness 40%, and sample 19 showed flatness 60%, being improved. Comparing samples 21, 20 and 19 with each other, it can be seen that an excessively high compressibility fails to attain flatness. When compressibility is at most 15%, flatness is at least 40%. In order to achieve the flatness of at least 60%, compressibility must be at most 10%. In order to achieve the flatness of at least 80%, compressibility should be at most 3%.
Sample 15 showed roughness Ra 2.3 nm, being the highest. Sample 16 showed Ra 0.9 nm, being improved. With the pad having low compressibility, the abrasive grain strongly strikes the group III nitride substrate during abrading and the surface roughness becomes high. Accordingly, pad compressibility must be at least 1%.
Sample 21 showed roughness Ra 1.4 nm, being relatively high. Sample 20 showed Ra 0.55 nm, being improved. Samples 18 and 19 showed excellent surface roughness (Ra 0.42 nm and Ra 0.45 nm, respectively). A pad having high compressibility may easily deform and enter the recessed crystal defect gathering region 13. Accordingly, crystal defect gathering region 13 may particularly be corroded and the recesses and protrusions may become prominent. The surface loses flatness, and the surface roughness becomes high. Therefore, pad compressibility must be at most 15%. More preferably, it is at most 10%. Further preferably, it is at most 3%.
After CMP was performed, an epitaxial layer was deposited thereon through the MOCVD method. The PL (photoluminescence) light emission was examined. Epitaxial property is PL strength (in arbitrary unit). Samples 15 and 21 respectively showed 35 and 36, being particularly weak. Sample 17 showed 90, exhibiting sufficient PL strength. Samples 16 and 18 showed considerably high PL strength (75 and 84). Pad compressibility for sample 1.5 was small and the abrasive grain strongly struck the group III nitride crystal, whereby the surface roughness became high. This resulted in weak PL strength. Pad compressibility of sample 21 was high and the pad easily deformed. Therefore, the pad entered the recessed crystal defect gathering region 13 and deeply abraded the same. This resulted in high surface roughness, poor flatness, occurrence of over-abraded circumference, and weak PL strength.
Further, various nitride layers were epitaxially grown on the substrate, and electrodes were provided to obtain a laser device. It was separated into chips and property of LD light emission and the yield of conforming items were examined. Sample 21 showed a 27% yield, being poor. This may also be understood that the crystal defect gathering region 13 was abraded because of the high compressibility, resulting in poor flatness and surface roughness. Sample 15 showed a 26% yield, being poor. This may be attributed to the high surface roughness. Samples 16, 17 and 18 respectively showed 70%, 78%, and 71% yields, being excellent. Samples 20 and 19 respectively showed 42% and 56% yields.
Based on the foregoing result, pad compressibility must be at most 15%. More preferably, it is at most 10%. Further preferably, it is at most 3%.
With sample 1 (“S perpendicular G”) in Example 1, conditions were varied (nine types) and CMP was performed. Specifically, pressure was varied. Crystal defect gathering region 13 is weak and easily abraded. Low pressure on the pad may reduce the abrading rate of low defect single crystal region 11 and C-plane growth region 12 and increase the removal ratio of crystal defect gathering region 13 being poor in chemical resistance, resulting in low flatness. High pressure may deform the pad, whereby the pad may deeply abrade crystal defect gathering region 13, resulting in low flatness. The force which presses the abrasive grain may also become great, resulting in an increased surface roughness Ra. Therefore, the effect of pressure must also be examined.
Sample 1 was prepared as follows. On an alumina block measuring 135 mm outer diameter and 30 mm thickness as abrading holder 53, three GaN wafers (substrates) measuring 50 mm φ diameter and having stripes with 400 μm pitch were adhered by solid wax so that stripe direction S was perpendicular to rotation direction G (radially). Then grinding and lapping were performed. After grinding and diamond lapping were performed with diamond grindstone No. 2000, CMP was performed under the conditions shown in Table 4. Samples 22-30 were obtained.
For each of the samples 22-30, the loose abrasive grain was colloidal silica (SiO2). For each of the samples 22-30, pH=2.5 (acid). Additive for pH adjustment was HNO3 (nitric acid). It had been known that pH=2.5 being acid was a suitable pH range in Example 2. Since pressure was to be varied, pH of the suitable range was selected.
Pad compressibility was 3% for each of samples 22-30. The pressure on a wafer was 80 g/cm2 (7.8 kPa) for sample 22; 100 g/cm2 (9.8 kPa) for sample 23; 150 g/cm2 (14.7 kPa) for sample 24; 300 g/cm2 (29.4 kPa) for sample 25; 500 g/cm2 (49 kPa) for sample 26; 800 g/cm2 (78.4 kPa) for sample 27; 1000 g/cm2 (98.0 kPa) for sample 28; 1500 g/cm2 (118 kPa) for sample 29; and 1600 g/cm2 (157 kPa) for sample 30.
After CMP, the flatness and surface roughness Ra of the samples were examined. Sample 22 showed flatness 38%, being the lowest. Sample 23 showed flatness 50%, being improved. Samples 26 and 27 respectively showed flatness 85% and 82%, being excellent. Sample 30 with high pressure showed flatness 52%, being low. Too high or low pressure on the pad impaired flatness.
Sample 30 showed surface roughness of Ra 2.5 nm, being too high. Sample 29 showed Ra 2.0 nm, being improved. Samples 24 and 23 showed excellent surface roughness (Ra 0.39 nm, Ra 0.40 nm, respectively). Higher pressure resulted in higher surface roughness, whereby the quality of the surface was deteriorated.
Too low pressure reduces the abrading rate of low defect single crystal region 11 and C-plane growth region 12, which are physically and chemically strong, while only weak crystal defect gathering region 13 is abraded. Thus, flatness becomes low. Too high pressure deforms the pad, whereby weak crystal defect gathering region 13 is greatly abraded and the flatness is impaired. Since the pressing pressure is great, the surface roughness is also greatly deteriorated.
Based on the foregoing result, the pressure must be 100 g/cm2 (9.8 kPa)-1500 g/cm2 (147 kPa). More preferably, it must be 300 g/cm2 (29.4 kPa)-1000 g/cm2 (98 kPa).
After CMP was performed, an epitaxial layer was deposited thereon through the MOCVD method. The PL (photoluminescence) light emission was examined. Epitaxial property in the table is PL strength (in arbitrary unit). Sample 30 showed 12, being particularly weak. Samples 26 and 27 showed 91 and 88, exhibiting sufficient PL strength. Samples 25 and 28 showed considerably high PL strength (78 and 68). Pressure applied from the pad on the wafer for sample 30 was too high, whereby crystal defect gathering region 13 was abnormally abraded. Since flatness was low and surface roughness was great, the quality of the epitaxial layer was deteriorated.
Further, various nitride layers were epitaxially grown on the substrate, and electrodes were provided to obtain laser devices. It was separated into chips and property of LD light emission and the yield of conforming items were examined. Sample 30 showed a 8% yield, being extremely poor. Sample 22 showed a 34% yield, being poor. Sample 25 showed a 75% yield, being excellent. Based on the foregoing result, it can be seen that the pressure should not be too high or low.
The pressure applied on the pad must be 100 g/cm2 (9.8 kPa)-1500 g/cm2 (147 kPa). More preferably, it must be 300 g/cm2 (29.4 kPa)-1000 g/cm2 (98 kPa).
Although the present invention has been described and illustrated in detail, it is clearly understood that the same is by way of illustration and example only and is not to be taken by way of limitation, the spirit and scope of the present invention being limited only by the terms of the appended claims.
| Number | Date | Country | Kind |
|---|---|---|---|
| 2006-239764 (P) | Sep 2006 | JP | national |
