Let me try to answer your question a little more directly. There are a lot of assumptions built into the statement that EPD correlates with defect density. I'd suggest a cautious approach here.

Yes, one can automate the measurement of etch pit density (EPD). And, of course you can just simply measure the density using a microscope. But the correlation of EPD to bulk crystalline defect density is highly problematic. It's probably true that a defect that appears on the surface will be more likely to pit, the converse is not necessarily true. That is, an etch pit on the surface does not imply a defect (at least in the conventional sense).

And, the exact method of etching combined with the crystalline material also has an impact. One etchant might be more "aggressive" than another, resulting in differing EPD for the exact same material. Will you etch with an acid? A base? A solvent? Vacuum? Each method of etching may yield a different EPD.

Surface preparation will also play a role in the measured EPD. No surface is truly, atomically smooth. (Well, no extended surface anyway.) The degree of polish and even the method through which it is achieved will play a role in the EPD. A surface with a greater RMS roughness will show a different EPD than one that is smoother. A surface prepared using one method of polishing will have a different EPD than the same material polished with another method, even though they might have the same RMS roughness.

And, that brings up crystal orientation. For example, a (111) surface will etch differently than a (100) surface. Each crystalline face will show a different surface energy and therefore a different propensity to etch without regard to defect density. And, of course, it's extremely difficult to produce a surface that is perfectly oriented to a crystalline face. Cleaving does produce the best faces, but some crystals don't cleave well, if at all (ice for example). There will always be slight misalignments and imperfections producing steps, and other surface defects that are not associated with bulk defect density. Again, these defects will show a different (probably higher) propensity to etch.

All the mentioned problems (and many more) make an absolute correlation between EPD and bulk defect density problematic. That's not to say there is NO correlation. It's just that any measurement is likely imbedded in a multitude of competing, masking, effects. Etch pit density measurements are best used in a carefully controlled, relative way.

Relative etch pit measurements are a reasonable gauge in the context of process control. Here, one probably has a carefully controlled growth process and the exact method of etching is also carefully controlled. So, for a given, carefully, identically cut and exposed face, polished in exactly the same way, one might make some kind of statement about if the process itself being more or less successful at generating a crystal of lower or higher "quality". For example, one can map EPD across the surface of a single wafer and make statements like "since EPD near the edge of our 8" diameter Si waver is greater than in the center, we probably have higher defect density there." Again, a relative measurement.

Unless you're following a particular standard very closely (and such standards exist in the semiconductor industry for Si, Ge, etc.) direct comparison is probably not warranted.

Here is formula for calculating etch pit density: EPD=-(1/Ap)*log(Re/Ro), where Ap is the average etch pit area, Ro is the intensity of the light reflected from the polished crystal (reference point), and Re is the intensity of the light reflected from the etched crystal. In both cases, the beam of light is focussed normally on surface of the crystal. This process can be automated (see attached link) using etched test wafer. The test wafer is suitably moved stepwise in a plane perpendicular to the light beam by motors under computer control so that a map of the EPD values at a preselected number of points on the test wafer is automatically generated.

http://www.google.com/patents/US4925298

Let me try to answer your question a little more directly. There are a lot of assumptions built into the statement that EPD correlates with defect density. I'd suggest a cautious approach here.

Yes, one can automate the measurement of etch pit density (EPD). And, of course you can just simply measure the density using a microscope. But the correlation of EPD to bulk crystalline defect density is highly problematic. It's probably true that a defect that appears on the surface will be more likely to pit, the converse is not necessarily true. That is, an etch pit on the surface does not imply a defect (at least in the conventional sense).

And, the exact method of etching combined with the crystalline material also has an impact. One etchant might be more "aggressive" than another, resulting in differing EPD for the exact same material. Will you etch with an acid? A base? A solvent? Vacuum? Each method of etching may yield a different EPD.

Surface preparation will also play a role in the measured EPD. No surface is truly, atomically smooth. (Well, no extended surface anyway.) The degree of polish and even the method through which it is achieved will play a role in the EPD. A surface with a greater RMS roughness will show a different EPD than one that is smoother. A surface prepared using one method of polishing will have a different EPD than the same material polished with another method, even though they might have the same RMS roughness.

And, that brings up crystal orientation. For example, a (111) surface will etch differently than a (100) surface. Each crystalline face will show a different surface energy and therefore a different propensity to etch without regard to defect density. And, of course, it's extremely difficult to produce a surface that is perfectly oriented to a crystalline face. Cleaving does produce the best faces, but some crystals don't cleave well, if at all (ice for example). There will always be slight misalignments and imperfections producing steps, and other surface defects that are not associated with bulk defect density. Again, these defects will show a different (probably higher) propensity to etch.

All the mentioned problems (and many more) make an absolute correlation between EPD and bulk defect density problematic. That's not to say there is NO correlation. It's just that any measurement is likely imbedded in a multitude of competing, masking, effects. Etch pit density measurements are best used in a carefully controlled, relative way.

Relative etch pit measurements are a reasonable gauge in the context of process control. Here, one probably has a carefully controlled growth process and the exact method of etching is also carefully controlled. So, for a given, carefully, identically cut and exposed face, polished in exactly the same way, one might make some kind of statement about if the process itself being more or less successful at generating a crystal of lower or higher "quality". For example, one can map EPD across the surface of a single wafer and make statements like "since EPD near the edge of our 8" diameter Si waver is greater than in the center, we probably have higher defect density there." Again, a relative measurement.

Unless you're following a particular standard very closely (and such standards exist in the semiconductor industry for Si, Ge, etc.) direct comparison is probably not warranted.

Thank you very much prof.Sergei Sergeenkov and Patrick Bisson for your kind and good suggestions for my question.

with regards

V.Jayaramakrishnan

"It ain't over until the fat lady sung". VJ! Not so fast. This is a topic of direct interest to me. In fact I was looking for precisely this topic and am delighted to find like minded folk on RG. Here is a discussion from LinkedIn that has direct relevance:

http://www.linkedin.com/groupItem?view=&gid=2683600&type=member&item=188581023&qid=a8b760da-5a8c-4d56-8400-260fd1c9eed2&goback=%2Egde_2683600_member_196640300%2Egmr_2683600%2Egde_2683600_member_256708520%2Egna_2683600

Please join us learn & share your knowledge and expertise in this field. You are all pre-qualified, pre-approved and welcome! Let's take advantage of our collective wisdom. Help us understand EPD better.

XRD (NDE) & EPD Compared:

http://www.flickr.com/photos/85210325@N04/8435791663/

[email protected]

Volker! Nice data. What are the etch compositions, time & temperature used? Despite the limitations of the EPD method, users are still able to take advantage of the information from EPD. Never-the-less, the limiting issues still abound. Some of these being:

1. Etchants are still empirical & proprietary! (Highly subjective)

2. Only exposed surface (hkl) is examined.

3. NOT True volumetric measure of defects.

4. Only defects terminating on examination (etch) surface contribute to EPD.

In addition to all this, the fact that the dislocation density is a function of (hkl) adds another kink to the EPD analysis.

As a metallurgist I've used etchants since the late 1970's including electro-chemical etching. Etch composition, time, temperature and surface conditions (entropy) are variables that change etching characteristics significantly. It amazes me that folks are able to get reliable, repeatable information from EPD still for this many years. Kudos!

BTW do any of you have any idea how the EPD is related to the material performance and device yield? What's the purpose of the EPD? Wonder if any of the large manufacturers will share that kind of information. Intuitively, there's got to be a direct relationship.

No matter how fancy the counting technique, the fundamental issues with EPD still remain. i.e., EPD is highly process/operator dependent. It would not be easy to duplicate results by two independent users with identical wafers. Especially if you throw in to the mix the fact that the "off-cuts" may change. The etching kinetics would change depending on the crystallographic orientation of the exposed etch surface.

On top of all this we are observing variation in orientation on the wafers laterally. The solidified crystal does not come with atoms lined up neatly on a single plane. There are long range lattice curvatures in addition to local variations in dislocation densities. See this ZnSe (224) example: http://www.flickr.com/photos/85210325@N04/8435791663/

A rich discussion so far to follow. It has been duly pointed out that the EPD does not reveal all kinds of defects. It is particularly useful to measure density of dislocations (a kind of line defect) rather than point, and planar defects. Even in case of dislocations, etch pits are produced only when the dislocation ends up on the surface, so dislocation loops in the bulk are not revealed.

The immediate answer to V. Jayaramakrishnan's original question about material quality is that a very high EPD in a sample indicates that the material is mechanically weak (very brittle) due to presence of a large number of dislocations. The electrical effects of dislocations have also been studied extensively in semiconductors as dislocation cores tend to getter impurities and eventually may act like very thin conducting wires in the bulk of the material and can lead to "electrical short-circuits". Breakdown voltages have been found to be lowered due to this effect. On the other hand, clean dislocations have been shown to be electrically inactive. Since dislocations can create allowed electronic states in (semiconductor) band gaps, they can act as carrier recombination centres. This leads to drastic degradation of radiative recombination efficiency in materials like GaAs.

Rodney Dangerfield asks a morose looking fellow if he was happy and he replies "Yes". So, Rodney says "Then inform your face"!

I see at least 4 of us contributors besides the host. But, "1 / 0 · 9 Answers · 118 Views". Now there ought to be at least 4 of us that like this question to participate. At least the 4 of us ought to Like or Dislike the question. I hope it is just oversight and not apathy or condescension. It surely couldn't be due to "old news" syndrome as the "fat lady ain't sung yet" regarding this topic. The circulation of the topic would improve thus and we are likely to have other wiser ones participate if we all did our due diligence. Thanks for stepping up! That includes you Kanad (I bet folks in the UK call you Ken?).

1. "Green" or "Red" for the question. Please answer this.

2. "Green" or "Red" for each comment. Presuming you are reading each comment.

Just remember these choices are never final just as the contents. Each of us can change the answers and the choices at will. We shall all promise not to take personal affront to each other's choices or comments.

If it were just Q&A without digression and contention for frolic then the discussion would be boring and mundane. So jump in and state your "informed opinion". For claiming facts in this business may be "short lived" as understanding and knowledge increase. Like the earth used to be "flat" only a few hundred years ago!

Dear All,

Your answers are very awesome and really useful for my work. anyway now only i can realize where is my research ( three decades back) in this scientific world. i hope I am zero in my research, but that there no level zero in research.

VJ! That is the awesome news. Research can turn "lemons" in to "lemonade". The challenges with EPD inspires us to find solutions. Keep on keeping on!

Kanad! "clean dislocations have been shown to be electrically inactive. Since dislocations can create allowed electronic states in (semiconductor) band gaps, they can act as carrier recombination centres. This leads to drastic degradation of radiative recombination efficiency in materials like GaAs"

This is the aspect that interests me most. Are there any studies correlating such morphology of dislocation to ultimate material performance and yield? Please post some links or forward me PDF files when convenient.

[email protected]

Hi Ravi,

The literature on electrical properties of dislocations in semiconductors is extensive. Here are a few references that you may find interesting:

http://www.intechopen.com/books/crystalline-silicon-properties-and-uses/structure-and-properties-of-dislocations-in-silicon

“Gettering of impurities in solar silicon”, I Périchaud, Solar Energy Materials and Solar Cells, Volume 72, Issues 1–4, April 2002, Pages 315–326

“Recombination at Dislocations in Silicon and Gallium Arsenide”, P. R. Wilshaw, T. S. Fell, G. R. Booker, Point and Extended Defects in Semiconductors, NATO ASI Series Volume 202, 1989, pp 243-256

“Electrical properties of dislocations and point defects in plastically deformed silicon”,

P. Omling, E. R. Weber, L. Montelius, H. Alexander, and J. Michel, Phys. Rev. B 32, 6571–6581, 1985

“The electrical properties of dislocations in semiconductors”, A. Ourmazd, Contemporary Physics, Vol. 25, Iss. 3, 1984

Thanks Kanad! Really appreciate the references. Would appreciate PDF copies if available since I may not have access to subscription journals. The structure/prop book reference is fantastic. I've barely scratched the surface.

[email protected]

Bottom-line questions:

1. If the crystallographic orientation varies up to 70-300 Arc Sec across a 3" (75mm) wafer, would it perceivably affect the performance and device yield?

2. Same question for the dislocation density, what is the tolerable range?

3. Where can I find a conclusive/definitive study correlating crystallographic structure to property based on the recent 5 decades of knowledge in crystal growth, materials physics & device manufacture? Of specific interest is the use of multiple techniques to verify the results including EPD, EBIC, TEM, SEM, AFM, RHEED, PL, Optical Microscopy, and other techniques.

Dear Jarayamakishnan, I agree with Kanad Mallik in the fact that the term "crystalline perfection" has a lot of interpretations. In addition of the general definition of point, linear, surface and volume defects as structural lost of symmetry, there are many specific definitions of crystallinity for minerals, organic or inorganic crystals. Must are related to XRD data, such as peak resolution, peak broadening, peak miss-position, and ground X-Ray intensity among others. Perhaps this last experimental criteria is the most general, because all types of defects (except twins and segregation) must cause a contribution to the ground radiation. In this way, the EPD should be related best with linear defects, and you might precise the term "crystalline perfection" used in relation with the results of EPD.

Dear Prof. Carlos,

Thank you for your valuable comments on my question

with regards,

V.Jayaramakrishnan

Volker! I love the English translation side-by-side. I'll review in detail and learn. Thanks! Danke!

Thanks Volker, it is a good paper. There is nothing mysterious, it is only a lack of mutual understanding. Maybe the IUPAP or the IUPAC or the IUCr have to work somewhat in the nomenclature of these terms and the words "nanodefects, microdefects and milidefects" might be the most popular in our field referring to these characteristics of the solid materials; in a similar way as micro, meso and macro are universally accepted in relation to porous materials (this is not a proposition, it is only a joke) . The real question is that there are too many concepts applied to the lack of crystalline perfection, and it is often really difficult to know about what we are talking to in this matter.

Theory meets experiment for GaAs:

https://www.researchgate.net/post/What_are_the_causes_for_ASYMMETRY_in_the_Bragg_X-ray_Rocking_Curve_Profile_RCP_for_a_symmetric_004_GaAs_reflection

http://www.linkedin.com/groupAnswers?viewQuestionAndAnswers=&discussionID=257747275&gid=2683600&commentID=5792923344173019136&goback=&trk=NUS_DISC_Q-subject#commentID_5792923344173019136

http://www.flickr.com/photos/85210325@N04/9430820747/in/set-72157635172219571

How to calculate etch pit density?

Excellent question Ranjit! The real question is, how does it relate to dislocation density?

In practice, some acceptable etch parameters are chosen and the resulting etch pits are measured microscopically. Works best for low dislocation density. Highly subjective compared with XRD which gives sample volume (VOXEL) averages!

Computation of corresponding dislocation density is an art. There are many references in literature. However, using samples of known dislocation density as "standards", it is practically feasible to estimate dislocation density precisely.

Image below contrasts NDE based Bragg XRD Micrographs with EPD Micrographs:

"if values are lower (>15 × 102 / cm2 )what does mean and how to interpret for higher values?"

Lower numbers both for EPD and XRD FWHM indicate lower defect density and better quality crystal. Higher numbers on the other hand would be due to larger defect density in general. However, this must be observed topographically to understand the 2D nature of the strains in the crystal. Grain boundaries and sub-grain boundaries are contributors as well. In fact, larger dislocation densities tend to form their own "cell structure" depending on the growth and processing conditions.