Silicon Carbide continues to gain acceptance and enjoy widespread adoption in mainstream high voltage applications. Even as both substrate and epitaxial material quality keeps improving, new failure mechanisms can get introduced due to various factors like diameter expansion, process and volume scale up. This forces the industry to evaluate and implement a variety of characterization methods to identify and eliminate any potential root causes for failure modes. Currently every substrate and epitaxial wafer is scanned for defects even in mass production due to the variable presence of various types of killer visual defects and pure crystal defects. Due to the bandgap of SiC, a convenient technique to non-destructively reveal crystal defects in the material is to use excitation light in the Ultraviolet (UV) range. Recently all volume mass production characterization systems have adopted this UV excitation along with detecting crystal defects using either near-ultraviolet (NUV) or infra-red (IR) PL bands [1,2]. Both the widely used KLA CS920 and the Lasertec SICA 88 platforms use UV light excitation using either a laser or a lamp. In our previous reports [3-6] while detecting faint PL signals, we have observed that exposing the wafers to this UV excitation results in material changes and increase in background PL that can be detected via the various PL acquisition channels. This raises concerns on the exact mechanism of the SiC material change. This constantly changing background also causes issues with measurement repeatability and calibration of the tools. In this work, we offer a detailed study of the effects of both UV laser and UV lamp exposure on as-grown SiC epitaxial wafers and explain the changes in the material. Additionally we replicate exactly the changes using various independent processing steps, while also offering various ways to both prevent and remove this effect.

Performing repetitive scans on as-grown epitaxial layers by using either an UV laser or UV lamp results in an observable increase in the background PL. This effect is highly repeatable in samples and results in a 25%-35% change in the signal to noise ratio both in the NUV and IR wavelength ranges. This can either adversely affect or enhance the defect signatures depending on which wavelength range is being captured. This causes repeatability issues for the detecting recipes and algorithms due to the signal variation. A few surface treatments done to the affected surfaces have been successful in restoring the surface PL signals to their original values. From this and from performing EDX on pre- and post HF etch samples the formation of a very thin layer of UV-exposure accelerated oxide was observed. This accelerated oxide occurred only in the present of the UV excitation. The cause for the formation of this oxide is proposed to be the highly reactive surface state of the as-grown epitaxial wafers. Passivating the as-grown surface before repetitive scanning completely eliminates the change in background PL.

This effect is further replicated on epitaxial wafers by growing thermal oxides of different thicknesses and also depositing PECVD oxide. The effect of the thermal oxides very closely mimics the observed effect, while the deposited oxide does not cause a big change in the PL signals. This shows the oxide – SiC interface is an important factor for this effect, and not just the presence of the oxide. The other effect of UV excitation on the expansion of basal plane dislocations (BPD) into stacking faults (SF) [2] was also studied. SF nucleation was only observed after about 150 scans and even then only in a small fraction of the BPDs.

Characterization of the affected surface after repetitive exposures will be presented. High intensity constant exposure low temperature PL spectrum results which are very different in nature to the relative low intensity exposures will also be presented. Both surface treatments to completely recover from the effect and passivation schemes to prevent this effect from UV excitation exposures will be shown. Further characterization of the effect of both thermal and deposited oxides on PL of epitaxial layers will be shown.

References:

[1] T. Tanaka et al. Mater. Sci. Forum, 778-780, 91-94 (2014)

[2] N. A. Mahadik et al. Appl. Phys. Lett. 100, 042102 (2012)

[3] H. Das et al. ECS Transactions, 69 (11) 29-32 (2015)

[4] H. Das et al. ECS Transactions, 75 (12) 233-237 (2016)

[5] H. Das et al. Mater. Sci. Forum, ISSN: 1662-9752, Vol. 897, 222-225 (2017)

[6] H. Das et al. To be published in ICSCRM 2017 proceedings