Manufacturing method of semiconductor device and semiconductor manufacturing apparatus
In accordance with an embodiment, a manufacturing method of a semiconductor device includes detecting elastic waves, and detecting or predicting an abnormality of the processing object occurring during polishing of the processing object. The elastic waves are generated from the processing object during the polishing. The abnormality is detected or predicted by analyzing the detected elastic waves.
1. A semiconductor manufacturing apparatus comprising:
a rotatable first table configured to hold a processing object;
a rotatable second table having an abrasive cloth and configured to polish the processing object with the abrasive cloth;
an AE sensor configured to detect elastic waves generated from the processing object during polishing of the processing object;
an elastic wave processor configured to:
plot first elastic wave data, second elastic wave data, and third elastic wave data in a multidimensional feature space having at least three axes of respective feature amounts,
the first elastic wave data being data collected when polishing of each of sample processing objects is finished with neither a scratch nor a crack,
the second elastic wave data being data collected when the scratch occurs during polishing of each of the sample processing objects, the second elastic wave data being classified depending on degree of the scratch,
the third elastic wave data being data collected when the crack occurs during polishing of each of the sample processing objects, the third elastic wave data being classified depending on degree of the crack;
categorize the first to third elastic wave data plotted in the multidimensional feature space into a normal polishing mode, plural degrees of scratch fault modes, and plural degrees of crack fault modes;
plot fourth elastic wave data collected concurrently with polishing of a target processing object in the multidimensional feature space; and
perform a pattern recognition to thereby analyze whether the fourth elastic wave data corresponds to either one of the normal polishing mode, the plural degrees of scratch fault modes and the plural degrees of crack fault modes concurrently with polishing of a target processing object in the multidimensional feature space; and
a control computer configured to control the rotatable first table and the rotatable second table based on an analysis result provided by the elastic wave processor.
2. The apparatus of claim 1 ,
wherein the elastic wave processor analyzes the elastic waves by a frequency analysis that uses Fourier transform.
3. The apparatus of claim 2 ,
wherein the elastic wave processor detects the scratch fault mode or the crack fault mode by detecting a change point in time-series power spectrum data obtained by the frequency analysis.
4. The apparatus of claim 1 ,
wherein the scratch fault mode or the crack fault mode of the processing object occurring during the polishing is a scratch or a crack occurring in the processing object.
5. The apparatus of claim 1 ,
wherein the first to third elastic wave data is created by a simulation.
6. The apparatus of claim 1 , wherein, when an occurrence of the scratch fault mode or the crack fault mode is detected or predicted, the elastic wave processor determines a suspension of the polishing process, a change of the polishing conditions or a continuation of the polishing.
7. The apparatus of claim 1 , wherein the first to third elastic wave data has four or more higher dimensions.