2016-03

2016-03Edit

New wafers BPC9 & BPCT9

As BPC8 was badly injured at step 3, I decide to process 2 new wafers. 
Note that the conclusion drawn from BPC8 was that we should not use PMGI baked at 200°C, because de development differential becomes too big with AZ : after 5'15, the PMGI is not fully etched in all patterns, while the aluminium layer is already etched from the developper.
Consequence: go back to LOR30B, and bake it a bit harder than the test performed at 170°C, or use PMGI and bake it lower.
See corresponding page BPC9

LOR or PMGI development time versus temperature

After having messed up with BPC8, I decide to make a test of development versus baking temperature, to define the optimum (planarization / dev time)

The spinning procedure is now well defined:

wafer Si/SiO2 standard
bake the wafer @ 110°C for water removal
tape the wafer on blue clean room tape, cut to a ~3" size
dispense 1-1.5mL resist slowly from a pipette (well centered on wafer)
2" @ 300rpm acc 250rpm/s
3" @ 800rpm acc 500rpm/s
60" @ 1500rpm acc 500rpm/s
Cut the edges of the spun resist on all sides of the wafer before untaping

NB: some condition, either acc on final spin step or resist quantity
acc=750 and/or >2mL makes the tape bend upward (tensile strength)
acc=500 and/or <1.5mL resist makes the tape bend downward (compressive)
We should probably be compressive, to insure the bridge does not open more when we open it
OBS: it happens that on one side, the tape is lifted up, on the other side lifted down. 
Where it is lifted up, there is a lot of resist (continuous film) wherea where lift down
there are only angel hairs.
The resist quantity is probably responsible for tensile / compressive strength

characterization procedure:

after spin
bake 5' @ XX °C
develop in MFCD26 until end of intereferences (MFCD26 equivalent to MF726 which is empty)

• PMGI SF15

According to observations on BPC8: baked at 200°C, after 5'15 still residues in some patterns, should indeed double the "interference" stop time for actual development.
Even at 160°C this is not really secure for Al (would mean 3'10 dev time)

• LOR30B
  resist is perempted since 2006, alsmost nothing left in the main bottle, and Camille bottle seems spoiled with crystals!!! Bottle was not really closed !!
  at 2000rpm (with same procedure as before) ellipso tells 6.5µm (according to DS shoudl be 4µm)

• LOR20B
  resist is perempted since 2010, main bottle is almost full
  I take Camille bottle for spin tests. It is clean. Spin really perfect, except the edges are crusty (probably because resist perempted)
  at 1500rpm (with same procedure as before) ellipso tells 4.55µm in the center, 5µm on the edge
  (according to DS shoudl be 3µm)

Note: the dev turns red, so cannot be used for 2 wafers. Use scarcely as there is almost no left.

PMGI SF15 - AZ development time

• Procedure

spin SF15 as above
bake 5" @ 110°C -> bake 5' @ 150°C
spin AZ5214E 4krpm
bake 1' @ 110°C
exp dose 5" @ 9.5mW/cm2 (47.5mJ/cm2)
bake xx @ 120°C
flood exp 45" @ 9.5mW/cm2
dev MFCD26 @ 21.2°C

Tests: (14/03/2016)

Test 2: focus on top show AZ is damaged. Bottom shows normal undercut (~ PMGI thickness)

Test3: the fact the AZ was too baked rendered development inhomogeneous.
Consequently, dev too much on patterns.

NB: after test 3, observed mask from IEF (transferred) and found out it was shit!!! -> impossible to developp a process with this kind of mask!!.                                                         Right, after etching it 15" in Anne de Vismes Cr etch

Test 4 done with CPBv2 mask layer 3 (should have done with mask layer 2 to have a chance to see interferences)
Problems:
- Patterns are cleared, but see small residues on the bottom. Ozone plasma 100W 200µb for 2 min -> no effect, exactly the same before and after.
- Small cracks all over AZ, probably because layers were too much baked (PMGI was baked 17', and AZ was baked 4')
-> DO NOT EXCEED BAKING TIMES (MOSTLY FOR AZ) OTHERWISE RESIST BECOMES BRITTLE

I had no specific trouble on tests with this process, though when I try it on BPCT8 and BPCT9, the PMGI layer unsticked, and developer ran below the resist over large distances.
As I already saw on BPCv1, this issue is solved only by baking PMGI harder. Maybe TI prime could improve a bit.
I will do one more try with TI prime, but otherwise, PMGI/AZ bilayer is close to be abandonned.

LOR20B - AZ development time

• Procedure

spin LOR20B as above (same as SF15, same thickness)
bake 5" @ 110°C -> bake 5' @ 180°C
spin AZ5214E 4krpm
bake 1' @ 110°C
exp dose 5" @ 9.5mW/cm2 (47.5mJ/cm2)
bake xx @ 120°C
flood exp 45" @ 9.5mW/cm2
dev MFCD26 @ 21.2°C

Tests: (15/03/2016)