Alignment Mark Design FAQ

By popular demand, this document is intended to be a quick guide to good practices for designing, fabricating, and using alignment marks in the EBPG with a minimum of hassle. If you're having trouble finding existing alignment marks, try the alignment section of the troubleshooting guide.

What shape should my alignment marks be?

The EBPG5000+ will accept both rectangular and cross-shaped alignment marks. For various reasons, the system is much better at automatically finding the rectangular marks, so it is highly recommended that rectangles be used if it’s an option. Crosses will work, but require a much more accurate initial position for the system to start looking for the mark than rectangles do.

What size should my alignment marks be?

The EBPG will allow marks to be defined with dimensions ranging from 1 to 100 μm. In practice, you probably want them to be 10 μm wide at minimum, and the larger the better. The reason for this is that larger marks reduce the probability of a random “bite” out of one of the mark edges disrupting the locator algorithm, as well as minimizing the effects of rounded or missing corners.

How much space should I leave around my marks?

Structures near your marks will completely confuse the system, so you want them to be pretty isolated. In most cases, 50 μm of empty space beyond the edges of each side of the mark is enough, although more is always better if possible.

How should I make my marks?

Marks can be either positive (deposited material) or negative (etched holes). See the questions below for the pros and cons of each approach. For positive marks, a metallization/liftoff process is the best way to go, while plasma-etching is the easiest way to make good negative marks. Specifics will vary with materials and processes.

For the lithography step of mark-making, any tool in the MNC will provide you with enough resolution to make good marks. The contact aligner and the stepper are the fastest and easiest ways to pattern them, but if you don’t have a mask made up they can also be done fairly quickly on the Heidelburg laser writer in bay 4, or even on the EBPG.

If I’m using positive (liftoff) marks, what material should I make them out of?

The EBPG isn’t designed to be good at imaging; besides running at a much higher beam energy than a SEM, it also only detects backscattered (as opposed to secondary) electrons. As a result, you’ll want to make your marks out of materials with electron-backscattering properties that are very different from your substrate. Backscatter behavior is highly atomic-number (Z) dependent, so heavy elements like Au or Pt will give you good contrast with a Si substrate. If you’re trying to avoid using metals, Ge is a good choice if Si is your substrate material. SiO2? is not recommended, as its scattering properties are very similar to Si, but can still work if it is thick enough.

Be careful of materials with large grain sizes, like Cr and most magnetic alloys; making marks out of these materials will generally give you rough edges, which will make finding the marks very difficult.

How thick should my positive marks be?

They need to be thick enough to scatter the beam a good amount. For heavier elements like gold, it won’t take much; 30-50 nm will generally give you adequate contrast. Lighter elements will need to be thicker; if you’re using something like SiO2?, 500 nm or more of material may be needed. These are lower bounds; thicker marks will always give you better contrast, although making marks thicker can increase the effect of liftoff problems like sloped sidewalls, which can make marks harder to detect.

What about negative (etched) marks?

Unlike positive marks, etched marks have no material contrast to aid the detection system. As a result, they’ll need to be pretty deep to work well. A good rule of thumb is to use depths of at least 1 μm. One advantage of etched marks is that with a good etch process the sidewalls will generally be very repeatably smooth, without the random roughness or torn edges that can happen in even the best liftoff-based processes.

My mark-generation process kind of sucks. Is there any way to work around the fact that only about 1 in 3 of my marks are usable?

CJOB lets you define an unlimited number of backups for a given alignment mark by adding lists of coordinates (instead of a single coordinate) to the mark location boxes in the different modules. If your mark yield is low, create two or three backups (spaced far enough from the original that they can be separately detected, of course) and add them all to CJOB. CJOB will move down the list of marks until it successfully locates one, only erroring out of the write if none of the marks in the list are readable.

I have three alignment marks around the edges of my large pattern, but my overlay accuracy is inconsistent across the pattern. What can I do?

CJOB will allow you to define alignment marks for pattern objects (single GPF files) and layout objects (arrays or collections of exposable objects), as well as the required exposure-level global marks. Adding marks to your pattern at every possible level of the hierarchy will greatly improve the consistency and accuracy of your alignments. As a bonus, you can reuse local (pattern-level) marks further up the hierarchy by using the “pick” and “generate” commands in CJOB, meaning you can probably get away with just printing lots of local marks and using the outermost ones for die- and global-level alignment. See the CJOB manual for more information on alignment-mark hierarchy and inheritance.

I still can't find my marks!

Check out the troubleshooting section on alignment marks for some tips on getting the system to detect stubborn marks.