Tuesday, August 7, 2007

Microsemi and others seek Metallurgical Bond Evaluation Methods for Diodes

Metallurgical bond evaluations of diodes have been a subject
of interest for many years. It has stimulated the use of additional
evaluation tools and methods over the last decade that
are still evolving. This MicroNOTE will provide some added
background and information on this topic.
Metallurgical bonding requirements for Category I, II, or III
are defined in MIL-PRF-19500, Appendix A, paragraph
A.3.19.
Destructive Physical Analysis (DPA) procedures also
exist to determine when bonds are acceptable. This has
involved the MIL-STD-1580A and B revisions entitled
“Destructive Physical Analysis for Electronic,
Electromagnetic, and Electromechanical Parts”, as well as
method 2101 in MIL-STD-750 for diodes. In some respects,
there have been notable differences between them that may
have also resulted in different dispositions of DPAs.
However with recent changes in MIL-STD-1580B including a
section for “Detailed Requirements for Glass Bodied Diodes”,
these two methods now better support one another. We will
review some of the previous possible conflicts and how they
are now resolved.
The oldest method for bond evaluation involving MIL-STD-
1580A provided for cross-sectional analysis with a minimum
of 50% bond criteria of the available area. It did not separately
recognize any of the three different metallurgical bond categories
now identified in MIL-PRF-19500. This criterion may
have been misleading unless multiple cross-sections were also
performed.
The MIL-STD-750, Method 2101 is entitled “DPA Procedures
for Diodes”. It supports MIL-PRF-19500 by recognizing all
three metallurgical bond categories and their different requirements.
Although cross sectioning is still included for design
verification in method 2101, it is not intended for the exclusive
disposition of bond voids in the same manner as MILSTD-
1580A previously did. For example, method 2101 also
provides a scribe-and-break procedure for glass-axial-lead and
surface-mount diodes. It also recognizes thermal impedance
as a bond indicator, particularly for Category III where crosssectional
analysis or scribe-and-break methods may not be as
definitive.
This latter Category III example is where a “diffusion bond” is
provided only between the outermost metallization layer of the
elements being joined as defined in MIL-PRF-19500M,
Appendix A. For Category I or II, the scribe-and-break
method typically reveals the strength of the bond regions by
showing pulled silicon at a die interface. If the glass package
envelope contacting the die causes the silicon to break in an
undesired location, the broken device may be further analyzed
as described below.
The "Silver button with braze" construction is an example
specifically defined in Method 2101 as a Category II metallurgical
bond. Bonding in these products by Microsemi is
achieved with braze preforms between the die and plugs. The
preforms are designed to melt in a controlled manner during
the sealing/braze process. When evaluations were sometimes
made with the older MIL-STD-1580A, random voids in individual
cross-sectional planes could appear to be 50% or
greater. However when the entire die is removed and the actual
bond is measured, there has been consistently greater than
60% bonds in Microsemi evaluations. This has also been confirmed
by cross sectioning at a 2nd and 3rd plane (see figures
1, 2, and 3 examples) as well as by using an additional scribe,
break, and dig procedure.
Figure 1

(25% location through die)
2
Although not a requirement of Method 2101 for scribe and
break, Microsemi has also evaluated each bond by scraping or
digging through all of the bond interfaces. All of these evaluations
reveal bonding areas well beyond the requirements of
Method 2101. To demonstrate this, Microsemi evaluated 3
devices that were cross-sectioned from a JANS assembly lot.
These cross-sectional photos can sometimes exhibit bond-area
voids at a single plane that exceed 75% as seen in figures 4
and 5. Microsemi then performed a scribe, break, scrape, and
dig procedure on additional samples to evaluate all of the
device interfaces such as in figures 6 and 7. By taking the
photo of the worst interface (figure 7), and running an imaginary
line through it in various planes to simulate a cross-section
photo, one can see the pitfalls of cross-sectional analysis
by itself as an acceptance tool as previously referenced in the
older MIL-STD-1580A if done only on one plane.
Cross sectioning in a single plane is useful, but not as a standalone
method for bond evaluation. This handicap also applies
to other assembly designs including solder-bond contacts
defined in Category II or diffused bonds in Category III. The
MIL-STD-750 Method 2101 recognizes these limitations and
offers other methods of confirming bond integrity. This
includes scribe-and-break and thermal impedance testing to
ensure thermal transfer qualities out of the die-bond region as
described in MIL-STD-750, Method 3101. These additional
analytical tools are also now cross-referenced in the latest
revision of MIL-STD-1580B as described earlier.
Figure 2
(50% location through die)
Figure 3
(75% location through die)
Figure 4
Figure 5
3
There are also many other considerations for a good quality
bond analysis when using cross sectioning for glass encapsulated
voidless construction such as those with Tungsten slugs
for Category I bonds. For example, scribe-and-break techniques
may not be practical for bond analysis since silicon
will often shatter with these voidless designs. During cross
sectioning, particular care must be taken to prevent inducing
voids from die to slug during initial grinding when micro fracturing
of silicon silicides occur from variations in the type of
abrasive, grinding direction, pressure, etc. Also care must be
taken to prevent crack damage during cross sectioning. For
more details, refer to Microsemi MicroNOTE 035 entitled
“Metallographic Cross-Sectioning Techniques for the
Performance of DPA of Glass Encapsulated Voidless
Construction Diodes”.
Figure 6 Figure 7
by: Kent Walters & Al Russo
The above article is provided to you by http://www.escs9120.com/ FYI
ESCS provides information to make sound business decsions. (hope this helps)