Chapter 1.9
VLSI Etchants
Notes on Etch Rate Tests
Microlab Manual Chapter 1.9 is based upon Etch
Rates for Micromachining Processing (©1996 IEEE) by Kirt R. Williams and Richard S. Muller, IEEE
Journal of Microelectromechanical Systems, Vol. 5, No. 4, December 1996. The
complete article is available at the link above. All prepared samples prepared
and etch tests for this report were performed in the Berkeley Microlab. Sample
preparation and etchants are discussed on the following pages. Measurement
techniques and other notes are given below.
The original etch report has been expanded and updated. The new etch
report is available at Etch
Rates for Micromachining Processing – Part II (©2003 IEEE) by Kirt
Williams, Kishan Gupta, and Matt Wasilik, IEEE Journal of
Microelectromechanical Systems, Vol. 12, No. 6, December 2003. The updated report
includes substrates, thin films and etchants from several laboratories and not
addressed in the original report. Some highlights of the updated report are
quartz, pyrex, and sapphire substrates; silicon-germanium, gold, and tungsten
thin films; Cu aqueous etchants, and SF6/C4F8 and XeF2 gas etches.
Transparent Films (poly, oxides,
nitrides, photoresists) were coated over the entire wafer and etched unpatterned. Their
thicknesses were measured on the Nanospec using refractive indices determined
by ellipsometry and verified with the Nanospec. These RIs are listed in the
samples section of this report. (The RI of my low-stress nitride films was
significantly different using the Nanospec than the ellipsometer. I used the
ellipsometer RI, which agreed with published data.)
Five points were measured before each etch, the films were etched,
then the same five points (to within a few millimeters) were measured again.
The average of the differences of these five points, divided by the etch time,
determined the etch rates.
Opaque Films (single-crystal
silicon, metals) were patterned with photoresist. In some cases, the photoresist
was left on the wafers for the etch. In others, as appropriate, the material
was patterned and the PR was removed before the etch test. The SCS was covered
with patterned nitride for the hot KOH etch.
A step height near the center of the wafer was measured with the
Alphastep, the film was etched, and then the same step (to within a few tenths
of a millimeter) was measured again. The step height difference and the etch
rate of the photoresist or substrate then determined the etch rate of the film.
Plasma Etching
Plasma etching was done for one-half or one minute with one wafer
in the etch chamber. Care was taken to avoid plasma-hardening effects with the
photoresist samples.
Plasma etch rates on patterned wafers can be quite different from
those listed here for two reasons:
1.
Some plasma etch rates tend to increase when there is less surface
area to be etched, due to higher etch gas concentrations.
2.
Usually be etched under those conditions (e.g., oxide in the poly
etcher, LAM 1). These wafers were etched alone so that no etch gas was consumed
by the normally etched material.
Wet Etching
Faster wet etches were done for one minute (even less for a few
very rapid etches). All slower wet etches were done for at least 10 minutes to
get a more accurate measurement.
Accuracy
For etches in which the computed standard deviation was smaller
than the average rate, that etch rate is listed. In cases where the standard
deviation was larger than the average (or the surfaces were very rough when
Alphastep measurements were used), an upper limit is given (i.e. < 50). Etch
rates of zero are given where the films were thicker after the etch, as often
happened with photoresist in wet etches. In a few cases, the entire film was
etched off in a short time; a lower limit is listed for these etch rates. My
measurements are rounded to two significant figures. I estimate my results to
be good to within ± 10% ± 5 A/min.
Final Notes
Because etch rates will vary with surface area exposed,
cleanliness of chamber, other materials present, age and previous use of
solution, temperature of solution, temperature of chamber, other variables, and
seemingly, phase of the moon, do not expect your results to be the same as
those listed here! I have therefore included etch rates that other lab users
and I have observed for many of the etchant/material combinations to give an
idea of how much etch rates can vary. Note that the top rows (all my own
observations) are with fresh solutions and recently cleaned chambers, while the
"Observed Range" high and low can be for older solutions, etc.
Sample Preparation
Most of the materials listed here
are commonly used in the Microlab; others have been included for comparison.
Three different popular KTI photoresist hardbake times were used to determine
the effect of longer bake times.
SC Si <100>
Single Crystal Silicon with <100> orientation.
Poly n+
In-situ heavily n-doped polycrystalline
silicon. RI = 3.97.
Deposited on a wafer with thermal oxide
already on it to enable thickness measurements.
Deposited in Tylan 11 using recipe
SDOPOLYG: SiH4 = 120 sccm, PH3 = 1 sccm, 650ºC.
No anneal.
Wet Ox
Silicon dioxide grown in water vapor. RI =
1.46.
Grown in Tylan 2 using program SWETOXB: grown at 1100ºC, p = 1
atm, with a 20 minute N2 anneal at 1100ºC.
Dry Ox
Silicon dioxide grown in dry oxygen. RI =
1.46.
Grown in Tylan 2 using program SDRYOXB:
grown at 1100ºC, p = 1 atm, with a 30 minute N2 anneal at 1100ºC.
LTO Undop
Undoped, annealed
low temperature oxide. RI = 1.46.
Deposited in
Tylan 12 using recipe VDOLTOC: SiH4 = 60 sccm, O2 = 90
sccm, PH3 = 0 sccm (no doping), 450ºC, p = 300 mT.
Annealed in N2 in Tylan 2 using
program N2 ANNEAL at 1000ºC for 60 minutes.
PSG Unanl
High-doped
phosphosilicate glass with no anneal. RI = 1.47.
Deposited in Tylan 12 using recipe
VDOLTOC: SiH4 = 60 sccm, Od = 90 sccm, PH3 = 10.3 sccm
(high doping), T = 450ºC, p =300 mT.
PSG Hidop
High-doped,
annealed phosphosilicate glass. RI = 1.48.
Deposited in
Tylan 12 using recipe VDOLTOC: SiH4 = 60 sccm, O2 = 90
sccm, PH3 = 10.3 sccm (high doping), T = 450ºC, p = 300 mT.
Annealed in N2 in Tylan 2 using
program N2ANNEAL at 1000ºC for 60 minutes.
Stoch Nitrid
Stochiometric silicon nitride (Si3N4). RI
= 1.99.
Deposited in Tylan 9 using program SNITD:
NH3 = 75 sccm, SiH2Cl2 = 25 sccm, p = 200 mT, T = 800ºC.
Low-σnitrid
Low-stress silicon
nitride (silicon-rich SixNy).RI = 2.18.
Deposited in Tylan 9 using program SNITD.V
: NH3 =16 sccm, SiH2Cl2 = 64 sccm, p = 300 mT,
T = 835ºC.
Al/2% Si
Sputtered aluminum with 2% silicon in the
target.
Deposited in the CPA at P = 4.5 kW, track
speed = 20 cm/min, p = 6 mT.
Sput Tung
Sputtered
tungsten.
Deposited in the
CPA at P = 4.5 kW, track speed = 10 cm/min, p = 6 mT.
(Most of my tungsten peeled off my wafers
due to residual stress. P = 2.9 - 3.1 kW should reduce this problem.)
Sput Ti
Sputtered titanium.
Deposited in the CPA at P = 4.5 kW, track speed = 10 cm/min, p = 6
mT.
Sput Ti/W
Sputtered 90% titanium/10% tungsten alloy.
Deposited at Stanford; conditions unknown.
Used as an adhesion layer for tungsten.
KTI 20 mn
OCG 825 (G-line) photoresist hardbaked 20
minutes at 120ºC. RI = 1.63.
Deposited using Eaton program 10.
KTI 1 hr
OCG 825 (G-line) photoresist hardbaked 1
hour at 120ºC. RI = 1.65.
Deposited using Eaton program 10.
KTI 1 dy
OCG 825 (G-line) photoresist hardbaked 1
day at 120ºC. RI = 1.67.
Deposited using Eaton program 10.
Olin Hunt
Olin Hunt 6512 (I-line) photoresist
hardbaked 30 minutes. RI = 1.63.
Deposited using Eaton program 15.
Etchants and Tips on their Use
All etches were
done at room temperature (about 20ºC) unless otherwise indicated.
All wet etches
were done with fresh solutions, agitating occasionally.
All plasma etches
were done with recently cleaned chambers.
Silicon Etchant – Polycrystalline Silicon (Bell Labs)
Concentrated hydrofluoric acid (49% by
weight).
From bottle.
Etches oxides very rapidly. Often used to
remove sacrificial PSG when micromachining.
1 : 10 HF : H2O (HF from
bottle).
Typically used for stripping oxide and HF dips.
1:25 HF: H2O (HF from bottle).
Used for HF dips to strip native oxide without removing much of
other oxides on the wafer.
5 : 1 buffered hydrofluoric acid (a.k.a.
buffered oxide etch, BOE).
From bottle.
Pattern with photoresist.
Because this solution is buffered, its etch rate does not vary
much with use. Best for controlled etching of oxides.
Silicon Etchant – Polycrystalline
Silicon (Bell Labs)
This solution is mixed and bottled by
Microlab staff. Bottles are stored in the tall white acid cabinet next to sink
432C (old lab).
Etch rate ~ 100 Å/sec
33% DI water / 3% NH4F / 64%
HNO3
Bottle content:
960 ml DI water
75 ml NH4F
1890 ml HNO3
Phosphoric acid
(85% by weight) at 160ºC.
From bottle.
Heated at
designated bath in Sink 7.
Used for wet
etching of nitride. The nitride is typically patterned with densified low-doped
PSG (densifying at 1000ºC for an hour will not affect low-stress nitride).
If the PSG mask
is not densified, it will be removed faster and may also have microcracks
through which the acid can seep.
The nitride can
also be patterned with poly.
Etch rate varies
significantly with temperature.
The rate (in A/min, T in ºC (!)) fits the
equation 0.0872exp[0.0386T].
Potassium hydroxide solution at 80ºC.
Mixed from 1000 g KOH pellets : 2 liters H2O.
Heated at right side of Sink 3.
Solution is self heating. Let it sit a few
hours before using.
Pattern with nitride.
Used for anisotropic etching of
single-crystal silicon.
Attacks (100) and (110) planes much faster
than (111) planes.
Etch rates listed here are down in the <100> direction.
3:2 25% TMAH : H2O (25% TMAH from bottle)
90ºC
Etch rate ~ 50 micrometers/hour
Occasional addition of water to the bath during long etches is required.
Additional information about this etch is located in the Sink3 manual and the
pocket wafer process module.
Ti Etchant #2
Second titanium etch solution listed in
the lab manual.
Mixed from 20 : 1 : 1 H2O : HF : H2O2.
Titanium wet etch. Due to HF content, it
also etches oxides.
Pattern with photoresist.
Concentrated hydrogen peroxide (30% by
weight).
From bottle.
Used to wet etch tungsten and its alloys, which can be patterned O2
with photoresist.
Piranha solution
at Sink 8.
This consists of
about 5.6 liters of H2SO4 held at 120ºC to which 100 ml
of H2O2 is added immediately before use.
Used to clean wafers. Strips photoresist
and metals, while not affecting silicon, oxides, and nitrides.
Acetone spray strip using MTI program 10.
Used to strip
photoresist.
While acetone
readily stripped all the photoresists listed in this table, its effectiveness
depends on the processing the PR has gone through. Heating the PR by a few tens
of degrees above 120ºC, either while hardbaking or during a process step, will
make it significantly harder. Some plasma processing will have a similar effect
(known as "plasma hardening"). In such cases, an oxygen plasma can be
used to remove the PR.
Technics-c plasma, O2 = 51.1
sccm, 50 W.
One wafer.
Used for "descumming" freshly
developed photoresist (typically for one minute).
Unbaked OCG 825 PR was removed at 330 A/min during a descum test.
Technics-c
plasma, O2 = 51.1 sccm, 400 W.
One wafer.
This oxygen plasma is used to ash (strip) photoresist
for 5 - 10 minutes. A power of 300 W for 7 min is also often used. It has been
argued that the lower power is better because less plasma hardening occurs
stripping. The Technics-c can hold up to four 4-inch wafers. A loading effect,
in which the etch rate decreases when there is more to etch, is seen. In a 400
W PR stripping test, ashing four wafers at the same time was 23% slower than
one alone.
Technics-c
plasma, SF6 = 12.9 sccm, He = 21.0 sccm, 100 W.
One wafer.
Used to plasma
etch nitride. Unfortunately, it removes poly isotropically at about the same
rate.
Pattern with
photoresist.
This etch
exhibits a severe loading effect. It is not only affected by the number of wafers
in the chamber, but also by the fraction of surface area of nitride exposed.
Furthermore, the etch rate varies with position in the chamber, so wafers
should be rotated 3-4 times during an etch (some users also move their wafers
among the four positions in a "planetary" motion).
Plasma etching, especially at higher
power, heats the chamber, which may affect etch rates and thus selectivity.
During all Technics-c tests, the plate temperature varied from 20 to 30ºC.
Technics-c plasma, CF4 (Freon
14) = 10.0 sccm, CHF3 (Freon 23) = 5.0 sccm, He = 10.0 sccm, 200 W.
One wafer.
Another nitride plasma etch. Same gases as LAM 2 uses.
LAM 1 plasma,
standard recipe: CCl4 = 130 sccm, O2 = 15 sccm, He = 130 sccm, gap =
1.5 cm,
p = 280 mT, P =
300 W.
Poly plasma etch.
Pattern with
photoresist. If total etch times longer than about 2 minutes are required,
break the etch up into several shorter times, giving the PR a chance to cool
and thus erode less.
LAM 2 plasma,
standard recipe: CF4 = 90 sccm, CHF3 = 30 sccm, He = 120
sccm, gap = 0.38 cm,
p = 2.8 T, P =
850 W.
Oxide plasma
etch.
Also etches
nitride well, but such use is allowed by special permission only.
Pattern with
photoresist.
If total etch times longer than about 2
minutes are required, break the etch up into several shorter times, giving the
PR a chance to cool and thus erode less.
LAM 2 plasma, standard recipe, but lower
power: CF4 = 90 sccm, CHF3 = 30 sccm, He = 120 sccm,
gap = 0.38 cm, p = 2.8 T, P = 450 W.
Some users have had better results using this lower power.
LAM 3 plasma,
standard recipe: BCl3 = 50 sccm,
N2 =
50 sccm, Cl2 = 30 sccm, CHCl3 = 20 sccm, p = 250 mT, P =
250 W.
All etches
followed by airlock processing: CF4 = 90 sccm, O2 = 10
sccm, P = 400 W, for 1 minute.
Aluminum plasma
etch. For thick layers of Al, either thicker photoresist or especially hardened
PR must be used (use the hardening process known as "PRIST"). The
airlock recipe, necessary to remove chorine from the wafers, is not supposed to
do any etching.
LAM 3 plasma, tungsten recipe: SF6 = 20
sccm, p = 100 mT, P = 125 W. No airlock processing.
Tungsten plasma etch.
Attacks poly isotropically.
By special permission only.
This process may be replaced a tungsten etch in the Tegal etcher.
Tegal plasma etcher is not ready for use
at this date.
It will be used for etching both nitride
and tungsten.
For
IEEE copyrighted material: Personal use of this material is permitted. However,
permission to reprint/republish this material for advertising or promotional
purposes or for creating new collective works for resale or redistribution to servers
or lists, or to reuse any copyrighted component of this work in other works
must be obtained from the IEEE. This material is presented to ensure timely
dissemination of scholarly and technical work. Copyright and all rights therein
are retained by authors or by other copyright holders. All persons copying this
information are expected to adhere to the terms and constraints invoked by each
author's copyright. In most cases, these works may not be reposted without the
explicit permission of the copyright holder.