2014-04

2014-04Edit

mardi 08/04

wafer BPC2

- spin AZ5214E 4000rpm 60"
/!\ no more in the bottle. Have to get it from one user (Elodie)'s bottle
- bake 110°C, 1'
- expo 3" @ 9.2mW/cm2
- bake 2' @ 120°C (setting 120°C, with a big beaker on top)
- flood expo 30" @ 9.2mW/cm2
- dev 30" MIF 726 @ 20°C + rinse 1' in beaker / water tap -> still have residues, but very difficult to say for small parts
- add 10" MIF726 -> still residues
- add 2'10 -> seems OK for large areas but not for small regions
- add 30", now some whole areas have been removed

Reprocess with a different bottle
same thing! The resist is way too old (per. 2009), and we could not find any other bottle...

mercredi 09/04

reprocess again with a third bottle. Same thing.
Try changing the developper from MIF 726 to AZ351B 1:4. All the resist goes away in 2'.

The differential of solubility is very bad. Either the first expo is too low (likely because we don't see the patterns after reversal bake) or the resist has been exposed in the bottle and there is no chance we can make this litho.

I suspect this is a problem of first expo too low. Test this idea on a standard wafer.

- spin 4krpm
- bake 1' @ 125°C
- measure with interferometer (take N1 N2 and N3 cauchy coefficients from AZ5214E datasheet)
- flood expo 30"@ 9.2mW/cm2
- dev AZ351B : ODI 1:4, by steps of 15", while measuring the remaining thickness
In 1'15 everything is gone

unexposed wafer: the fit is better with exposed resist parameters.

! Didn't have the correct model for the fit. In fact there is 500nm SiO2 (BTW, the refractive index of AZ is close to SiO2)
Do it again.
NB: with unexposed wafer right after baking, the fit is better with exposed resist cauchy parameters than with unexposed parameters...

end @ 2'10. Maybe because the developper is saturated with this second wafer (used 500mL in total, 100mL dev + 400mL ODI), or because the development is not really cumulative (still active while out of dev bath???)

Now Do the same but with a first expo 10" + bake 2'30 @ 125°C + flood 30"

vendredi 11/04

Reprocess BPC2 with a brand new AZ5214E
WORKS VERY WELL!!!

evap in new canon
- ion mill 150V, 150mA (I anode = 24mA), 2'
- Ti 50nm on backside of wafer. Psas_after = 5e-8mb
- Al 100nm @ 1nm/s (should do 2nm/s). Psas_ev = 1.4e-7mb
- Ag 5nm @ 0.2nm/s. Psas_ev = 8e-8mb

lift in remover PG.
OK for lift but something went very wrong. The Al looks granular as if it were etched in remover OR as if it was lifted with resist residues.
Beneath the pads that hold the wafer during evap, the layer is clean! This means that it could come from the ion milling...
Have to remove the remaining Al in Al etch -> there remain some residues.
I have to put it again 2' in Al etch pure (theretical etch rate 40nm/s), then in ozone plasma for 5 minutes, but there are still some residues.

lundi 14/04

Calibrate ion milling with an oxidized wafer
measure oxide layer thickness with ellipsometer before etching: 506 to 508nm
ion mill 2' @ 150V, 150mA, 40°, rotation 12°/s.

No change in oxide layer thickness...
ion mill seems useless to remove the oxide layer, but it can still be used to remove water and other contaminants on the surface.

New wafer BPC3 + wafer test BPCT2

Spin resist BPCT2

- spin AZ5214E 4000rpm 60"
- bake 110°C 1'
- expo 3" MJB4
- bake 2'30 @ 125°C (setpoint)
- flood expo 25"
- dev 1' in MIF726

remove resist residues
20" ozone plasma @ 100X, 0.2mb

deposit BPCT2

- ion mill 30" @ 150V, 150mA (24mA anode) 40°, rotation 16°/s
- Al 100nm @ 1nm/s @30°, rotation 16°/s
- Ag 5nm @ 0.2nm/s @30°, rotation 16°/s

lift in aceton: difficult but OK

mardi 15/04

deposit BPC3 pumped over night

- ion mill 30" @ 150V, 150mA (24mA anode) 40°, rotation 16°/s
- Ti pump (50nm @ 0.2nm/s)
- Al 100nm @ 2nm/s
- Ag 5nm @ 0.2nm/s

Lift difficult! CATASTROPHE when I turn on ultrasonic at low power, the wafer cleaves in 4 pieces!
Anyways, the edge of the layer is turned up - the undercut was not good enough

I will do a bilayer process

Spin BPCT3 (oxidized standard wafer)

- spin LOL2000 1500rpm 60"
- bake 130°C, 10min
- spin AZ5214E 4000rpm 60"
- bake 110°C 1'
- expo 3" MJB4
- bake 2'30 @ 125°C (setpoint)
- flood expo 25"
- dev 40" in MIF726

The undercut is around 800nm (on can see the double undercut)

- add 20" in MIF726

The undercut is around 2µm. The LOL between the gates  has been completely removed. I hope the mask is not collapsed.

Ozone plasma 20" @ 100W, 0.2mb -> OK

Evap BPCT3

- 100nm Al @ 2nm/s, 30°, spin 16°/s
- 5 nm Ag @ 0.2nm/s, 30°, spin 16°/s

Lift goes very well in a few minutes.

I suspect the Ag layer to have oxidized very quickly in remover and in aceton in the previous trials
I leave it in aceton for the night

mercredi 16/04

Indeed, the  layer is badly oxidized and looks crap -> should use another capping metal.

Nonetheless
The edge has a nice profile
-> the step angle is atan(105nm/1600) ~ 4°
-> the patterns are now 1.6µm larger than the design (/!\ yes, only 1.6, not 2x1.6. In fact it is only 0.8µm larger on each side)
-> one part has been shadowed by the foot of the AZ. Indeed the angle does not seem constant over  the step, the first few 100nm having a steeper slope.

-> the displacement of 0.8µm implies that the well undercut (more than 30°) layer thickness is around 1.4µm (0.8/tan(30)).
Since we had a problem on the preceeding test, this could mean that the flanks of the resist had a backbending, and only reducing the angle to below 22° (atan(800-100/1700); 800=achieved displacement, ~100=effect of LOL, 1700=measured AZ layer thickness) might have solved the problem already.
-> It seems mandatory when one wants to do such thing to measure the layer thickness prior to evaporation

So now, the angle of the metal layer in fact depends on the AZ resist layer thickness also (not on the LOL alone) Should have applied much less angle if we  wanted a more controlled pattern width (eg if we want 500nm increase of pattern size, with that AZ thickness -> only 10° needed).

Another way to do this is
(i) to avoid having undercut in the AZ (increase the dose of the first insulation),
OR (ii) to dilute the AZ but that is probably much less reproducible,
OR (iii) use MaN ??

Note: the measured frontier of the LOL (here 1.2µm away from the edge of the metal) is consistent with the optically measured undercut of 2µm (1.2+ 0.8 = 2µm)

Now I try this bilayer process on the true BPC wafer

I have huge problems!!!

- Have to spin (and reprocess) 4 times the LOL because there are lots of imperfections on the spin, whereas the wafers looks OK
- Then I do 2 times the lithography process which fails (all the small patterns -eg circles in ground plane- go away at development)

-> must be a problem with adhesion of the LOL onto Si, compared to SiO2

add a primer
BPC4

- spin TI prime 4000 rpm 60"
- spin LOL2000 1500rpm 60"
- bake 130°C, 10min
- spin AZ5214E 4000rpm 60"
- bake 110°C 1'
- expo 3" MJB4
- bake 2'30 @ 125°C (setpoint)
- flood expo 25"
- dev 50" in MIF726

-> undercut of 1.6µm, clear, no pattern went away!!

Updated the process optical resist bilayer

Ozone plasma 30" @ 100W, 00.2mb -> OK

jeudi 17/04

Evap BPC4

Pumped over night, P_sas = 1.4e-7mb. Open valve, -> 8.7e-8mb
- Ti pump 20nm @ 0.2nm/s -> P_sas = 3.5e-8mb (gun on) -> 2.8e-8mb (gun off)
- 100nm Al @ 2nm/s, 30°, spin 16°/s, P_ev = 2e-7mb (-> 7e-8mb gun off)
- 10 nm Pt @ 0.1nm/s, 30°, spin 16°/s, P_ev = 5e-8mb

Lift goes very well in a few minutes in aceton. Put 10" low power US at the end.

/!\ The US power is much less in hood on the right than in left hood

Add 2' in warm remover PG
Then 1' in ozone plasma

In optical microscope, looks good. The pattern size increase (of metal parts) is by 1µm. (not 1.6)
/!\ In fact, the spin was different between BPCT3 and BPC4, because Benoit Roche noticed there was a very slow acceleration of 200rpm/s^2 set on the spinner (so 20" to reach 4000rpm) which might explain why the AZ was 1770nm instead of 1400...

Step 2, spin BPC4

spin TI prime @ 4000rpm 60"
bake 120°C 1'
spin PMGI SF8 (batch 12060407, exp. 7/1/2013) @ 3000rpm 60"
bake hot plate setpoint 175°C, 5' (measured: 168°C), under beaker
spin PMMA A6 (from Youri bottle -I think it is the same batch as last spin in 01/14) @ 6000rpm 60"
bake hot plate setpoint 175°C, 15', under beaker
spin UV III (bottle VS 10/2013) @ 4000rpm 60"
bake hot plate 140°C, 90"

Problem with the first spin, many dusts appear after spinning PMMA...(I first took B. Roche's bottle, as there was not anymore PMMA A6 bottle). Then discussing with Youri, he took the end of the A6 bottle yesterday, so 2nd spin is with Youri's bottle.
Have to reprocess: remover (1'), IPA, aceton, IPA, then ozone plasma 1'.
The layer has turned a bit darker... but nothing drastic.
2nd spin is good

mardi 22/04

So in fact, observing a chip, I see that the layer has been damaged by the remover (?) -> take pictures
There are many little holes in the layer!

I test the possibility to developp PMGI with MIF726. Even by diluting the MIf, the result is the same.
Al is etched in 3min roughly.

Now I don't really want to use this layer for CPB, a priori.
But we still need to determine its Tc so I will anyhow measure a resonance on it.

In the litterature from shipley (AZ), document called Aluminum etching, one can see that:
- the etch rate in TMAH developper is 50-100nm/min
- the AZ developper is Al compatible (has a lower Al etch rate)
- If one puts Al in the AZ remover with only traces of water, the Al is strongly etched !!!
The document recommends then to use organice removers such as NMP
Maybe remover PG has the same issue?

Process again BPC2

I will make another layer with thicker cap layer -> need to use another material.
As well, I will use less angle so that only AZ is needed.

step1, spin BPC2

- bake 110°C, 1'
- spin AZ5214E 4000rpm
- bake 110°C, 1'
- expo 3" @ 9mW/cm2
- bake 120°C (setpoint) 2'30
- flood expo 30"
- dev MIF726 50"

Undercut looks good

mercredi 23/04

step1, evap BPC2

pumped over night, with valve open.  P_sas=5.2e-8mb

- Ti pump, 10nm @ 0.2nm/s -> P_sas = 2.4e-8mb
- Al 130nm @ 2nm/s, 10°, planetary 16°/s, P_ev ~ 2.5e-7mb
- Ti 20nm @ 0.5nm/s, 10°, planetary 16°/s, P_ev ~ 2e-8mb
- Au 10nm @ 0.1nm/s, 10°, planetary 16°/s, P_ev ~ 4.6e-8mb

lift aceton, easy
remover PG, 10' @ 60°C, rinse ODI
ozone plasma 1', 100W

step 2, spin BPC2

spin TI prime @ 4000rpm 60"
bake 120°C 1'
spin PMGI SF8 (batch 12060407, exp. 7/1/2013) @ 3000rpm 60"
bake hot plate setpoint 175°C, 5' (measured: 168°C), under beaker
spin PMMA A6 (from Youri bottle) @ 6000rpm 60"
bake hot plate setpoint 175°C, 15', under beaker
spin UV III (bottle VS 10/2013) @ 4000rpm 60"
bake hot plate 140°C, 90"

obs:

• Ramp length is around 300 nm
• first steep step is around 55nm, total thickness is around 195nm. Evaporated thickness: 130+20+10=160nm

BPC2T1

Test sample from BPC2 (23/04/2013).

Idea: Same as "Strain5" from 02/04/2014 (see March-section)

Exposure:  30 kV, Spot 1, "DC_strain_red_DoseTest1"
dev: as "Strain5" with MIF726=60s, stiring the sample very gentle.
obtical microscope: Generally looks good, no cracks visible.
Evap:  Ar milling and TwoAngleAu.prg evaporation, each 30 nm.
Lift-Off:  20 min in PGRemover at 60 deg C, no US. Peeled off easily.
Obs: SEM images

• Junction is asymmetric for all 48 SQUIDS, typically right JJ is 50x90 nm^2 and left one is 50x180 nm^2.
• Two anlge evaporation was probably done in the wrong order. Le crabe should be first shifted down and the island in the second evaporation up. Maybe this is also the reason for the "overshifting" of the island.
• Undercut is clearly visible and showing the right size.
• Overall the sample look fine and all 48 SQUIDS looked very similar although a dose test was performed within the 48 SQUIDS (see "Strain5" for foruther information). This shows good reproducibility within a single chip.