2013-12

2013-12Edit

vendredi 6/12

Cleave BPCD7 on scriber
Not a real success!

BPCD7_1
Expose line test on XL30
motif x2000_xnm_0.8_BUsuperp (dose from 0.8 to 1.4)
position list x2000_Xnm_BUsuperp_30keV_noBUfor PMGI.pls

WD=17mm, spot 1 I=21.8pA
30keV, 300µC/cm2, 1200pC/cm
x2000 -> area dwell time 2.202µs (step size 4nm), line dwell time 5.506µs (step size 1nm)

mardi 17/12

Discussion with Denis: PMGI is electrosensitive. PMGI SF8 dev rate in PMGI 101 developper is around 0.7nm/s when baked at 190°C, but increased up to 40nm/s when exposed with e-beam 40% PMMA dose.

dev MIBK / IPA (1/3) 1'15 + stop IPA 30" + dry N2
dev PMGI in MFCD26, 1' + stop ODI 30" + rinse IPA 5" + dry N2

obs: 
No pattern for x0.8
Mostly no undercut up to x1.2
In every dose and every size of wire, the ends of the wires are overdosed, and there is large PE.
Some patterns have pronounced undercut, some don't, some have wobble in the undercut (see pict).
The proximity effect is important and can be seen very clearly because the PMGI layer is e-beam sensitive
On clean patterns which have a good undercut, the size of the undercut is roughly 1.75µm (+/-200nm read error)
It does not depend much on the wire width, at least optically.

We attribute the wobbles to a not completely developped PMMA layer.

ADD 15" dev MIBK / IPA, same procedure
ADD 1' dev MFCD26 (see if the already present undercut is governed by e-beam exposure, or bare etch rate, or both)

obs:
Now some PMGI has been completely removed over large areas, which seem to have collapsed. This is located close to the "path" design for the x1.4 and x1.3 grids
But for most features the undercut has simply slightly increased, up to roughly 2.1µm
See attached file BPCD7_1_exp+dev.pdf

-> Next test for cleavage and angle observation: dev MIBK 1'30 + dev MFCD26 1' or less.

BPCD7_2
Expose line test on XL30
motif x2000_xnm_0.8_BUsuperp (dose from 0.9 to 1.4)
position list x2000_Xnm_BUsuperp_30keV_noBUfor PMGI.pls

WD=17mm, spot 1 I=21.46pA
30keV, 300µC/cm2, 1200pC/cm
x2000 -> area dwell time 2.237µs (step size 4nm), line dwell time 5.592µs (step size 1nm)

dev 1'30 MIBK + rinse IPA 30" + dry N2
dev 30" MFCD26 + rinse 30" ODI + rinse 5" IPA + dry N2

No clearly visible undercut in any pattern, but patterns OK.

Add 30" MFCD26

Optical pictures: See attached file BPCD7_2_exp+dev.pdf

cleavage with standard technique

sputter 2x15" + Angle observation 75° (pictures attached as 7z and table to make)
SEM images

obs:
-wider lines show larger undercut for the same dose than thinner lines -> proximity effect + electrosensitivity of PMGI
-single line and 30 nm line are more sensitive to dose than wider lines
-for relative doses above 1.1 the undercut varies much less in size than for relative doses below 1.1
-the single line shows larger undercut than the 30 nm line -> trouble with the sweep of the e-beam
-linewidth is around a factor of 2 larger than designed -> both a result of the observation method (which opens the mask), the fact that the mask is suspended (constraints on PMMA opens the gap) and the resolution limit of the e-beam ( working at 30keV, either the resist is not fully exposed to the bottom, or it is too wide)
-difference between designed linewidth and extracted linewidth is least affected by the dose for 90 nm desgined linewidth

(attached files "Plots 20131218.pdf" and Summary_20131218.xlsx and pictures on "Bpcd7-2 sur 'iram-he-004557\images\hls\Bpc'")

- The smallest linewidth seem to be obstructed by a thin layer. In principle it shouldn't be the gold layer unless something has gone wrong with the sputtering. I'd rather bet for a layer of PMGI which has remained in the gap during dry. We will check that with the 2nd part of the chip
 

wednesday 18/12

2nd part:
With other part, double angle evaporation in 'vieux canon': double angle ion etching +/-22°, 500V, 3mA (PAr=2e-4 torr) then Au 30nm @ 1nm/s +/- 22°.

Lift-off in clean room: 10' aceton + 2" US + rinse IPA + dry N2 + 10' remover PG 60°C + rinse ODI + dry N2
observation of the double lines

obs: (see attached files BPCD7_2.pdf et BPCD7_2.xlsx)
- the separation between the 2 wires centers is on average 600nm +/- 27nm, which is what is expected.
- the linewidth of the mask extracted from the first evaporated wire does not depend much on the dose, and is roughly between70nm and 100nm larger than designed, assuming a square shaped PMMA mask (but we know this is not the case)


- the size of the wires range from 30nm to 100nm for second evaporation, and from 30nm to 140nm for first evaporation. The mask is opaque under this angle below a linewidth of 130nm.

- Above dose 1.1, the width and the mask linewidth aren't much sensitive on the dose
- Indeed, the smallest patterns, even though clearly developped and undercut and even for the largest doses, never have an evaporated wire. THERE WAS A PROBLEM WITH THE DEVELOPMENT PROCEDURE

Another try with only 30" PMGI dev

BPCD7_3

Expose line test on XL30
motif x2000_xnm (dose from 0.9 to 1.4)
position list x2000_Xnm_30keV_noBUfor PMGI.pls
done another playlist with all the same pattern exposed, and a dose factor
modified the pattern to replace the 30nm rectangle with 2 lines separated by 20nm, and the 50nm rect with 3 lines separated by 20nm. (a path should be equivalent to a 40nm rectangle, but the polygon scanning by e-beam is so badly done that it is in fact much better to make paths than rectangles)

WD=17mm, spot 1 I=21.33pA
30keV, 300µC/cm2, 1200pC/cm
x2000 -> area dwell time 2.25µs (step size 4nm), line dwell time 5.626µs (step size 1nm)

20/12/2013

dev 1'30 MIBK + rinse IPA 30" + dry N2
dev 30" MFCD26 + rinse 30" ODI + rinse 5" IPA + dry N2 (--> Taken pictures with optical microscope, to scan)

Cleavage unsucessful (pushed scriber through lines 0.9 to 1.1, and other doses were cut between two fields, so almost no observable undercut on the whole chip)
Sputtering uncontrolled, but roughly 60" in several steps.

obs:
- mask sealed for all single lines (meaning there is again the PMGI residues in the opening.
Residues are clearly visible in the picture below
It seems possible to generalize that from 100nm and below, PMGI seals the opening, and has difficulties to be removed in the undercut area.

-> maybe use another developper with surfactant such as MIF726 (same 2.4% TMAH content as MF-CD 26, but presence of surfactants to have uniform wetting) ?
-> Or another rinse method

Image below: dose factor x1.4, single line, 30" MF-CD 26 on BPCD7_3 (left), and 1' MF-CD 26 on BPCD7_2 (right)

- for dose factors below x1.1, no polygon opened, whereas the paths are opened (notwithstanding the PMGI residues)
- paths have a larger undercut than polygons with the same opening -> proximity effect + electrosensitivity of PMGI
- linewidth almost independent on dose.
- smaller linewidth than BPCD7_2. This may come from the fact that the sputtering was 60" instead of 30", and that the mask does not deform under observation. We also had hints that the linewidth was smaller than measured in BPCD7_2, because the mask edge isn't square shaped, and the extracted linewidth should even reduce if we assume a rounded mask edge.
left: measured linewidth on mask BPCD7_3; right: comparison between _2 (extracted from Au wire width) and _3

- can't say much about undercut... Only observable for doses above 1.2.
Shows a decrease between 300 and 400nm as compared to the 1' MF-CD 26 development (see below: solid line 1' development, dashed lines 30" development)