The document discusses the fabrication process for micromachined hot wire anemometers. It involves depositing platinum or titanium wires on a silicon substrate using evaporation or sputtering. Contact pads and prongs are patterned using lithography and dry etching. The wires are released by vapor phase HF etching of the sacrificial silicon dioxide layer. Challenges include stress in evaporated platinum wires which can cause breakage at low temperatures. Sputtered titanium and wire design modifications help address this issue. The probes are tested inside a cryogenic probe station to calibrate their response to fluid flow.

1. Very Low Temperature Anemometry
Device microfabrication
13/07/2017 Hot Wire Anemometer - CVD 1

2. Hot Wire Anemometer
Hot Wire Anemometer - CVD 213/07/2017
Hot Wire Anemometery :
measure of the amount of heat transferred from a metallic wire
heated by joule effect and the flowing fluid
In the stationary state there is balance between:
• Joule effect
• Forced convection
• Other heat transfert mechanisms (usually neglected)
If the elecrical resistance of the wire changes with the
temperature it is possible to evaluate the fluid velocity :
!"#$%
& '():
+,
+, − +.
/0
= 2 + 4 5
Several possible configurations:
• CCA (Constant Current)
• CTA (Constant Temperature)
• …
I
U L≈1mm
d≈1÷10μm
Smaller devices could have better resolution :
• Mass↘ → temporal resolution↗
• Length↘ → spatial resolution↗
with l/d>>200 in order to neglect thermal conduction
Micromachined anemometer
(NSTAP – Nano-Scale Thermal Anemometry Probe)
« Fabrication and characterization of a novel Nanoscale thermal anemometry probe »
M.Vallikivi, A. J. Smits
IEEE Journal of microelectromechanical systems
L≈ 50÷100μm
d≈1÷4μm
e≈100nm

3. Fabrication process
Hot Wire Anemometer - CVD 313/07/2017
1
Substrate
preparation
2
Wire
deposition
(layer 1)
3
Contact pads
deposition
(layer 1B)
4
Backside
etching
(layer 2 )
5
Wire
release

4. Masks
Hot Wire Anemometer - CVD 413/07/2017
Mask 1 :
wires
Mask 2 :
prongs
Mask 1B :
contact Pads

5. Hot Wire Anemometer - CVD 513/07/2017
substrat
• Wafer Si 4 pouces
Substrat
ALD
• ALD : 5/10 nm Al2O3 (optional)
PECVD
• PECVD : 300/500/1000 nm SiO2
• Substrat : wafer Si 4 inch
– Thickness 500 ±25µm or 400 ±25µm
trade-off between mechanical resistance and ease of fabrication
– Single or double side polished
(single side polished is enough)
• Etch stop : Aluminum oxide
– Up to 20nm Al2O3 by Atomic Layer Deposition
it will protect the membrane from chemical etching by fluoride
during Reactive Ion Etching
But : we have to etch it at the end of the process (chlorine
plasma, TMAH, …)
• Membrane : Silicon Oxide
– Up to 1000 nm SiO2 by Plasma Enhanced Chemical Vapor Deposition
Easy to deposit, but brittle and stressed film (compression)
– Alternatives :
Thermal silicon oxide (constraining, alumina etch stop would not
be possible)
Silicon nitride by LPCVD (more difficult fabrication process)
Compensated Silicon nitride by PECVD (not readily available)

6. Reminder: Lift-off
Hot Wire Anemometer - CVD 613/07/2017
Positive resist double layer lift-off
1 - Optical lithography
2 - Metal Deposition3 - Cleaning
a. Lift-Off Resist (LOR)
spin coating
b. Positive resist
spin coating
c. UV Exposition
d. Development
And LOR etching

7. Hot Wire Anemometer - CVD 713/07/2017
e-beam
evaporation
•Etch Ar+ : 24s@250V
•Ti : 10nm @0,1nm/s
•Pt : 100nm @0,1nm/s
Lift-off
• Soak : Acétone
• IPA + US rinsing
• N2 drying
• soak : AZ326MIF
• EDI rinsing
• N2 drying
• Solvent cleaning
• Dehydration : 2min@150°C
• LOR10A: 4000/2000/50
• Bake : 5min @170°C
• UV5(0,6) : 4000/2000/60
• Bake : 1min 30sec @130°C
• Exposure MJB4 DUV : 1,3ec:
• Bake : 1min @130°C
• soak : 2min en AZ326MIF
• EDI rinsing
• N2 drying
DUV
Lithography
Pt wires by evaporation
SiO2
LOR+UV5
SiO2
Pt
Double layer lithography After lift-off:
Pt
Pt
After e-beam evaporation:
e-
e-beam evaporation

8. Pt wires by evaporation
Hot Wire Anemometer - CVD 813/07/2017
Suspended Pt layers
PtSi
SiO2
Pt
Transversal image
Structure Zone Model
In « zone T » :
• Competitive growth
• Inhomogeneous film
« granular » layer
The layer structure is determined by
kinetic limitations :
• Nucleation
• Cristal growth
• Grain growth
Tdeposition / Tmelting
Si
SiO2
Pt

9. !"#
!$
= &' − &#
)#
1 − +#
T↘
• Intrinsic stress :
it comes from the structure of the layer and it is linked to the
deposition process
In evaporated polycristallin films it comes from the coalescence of
cristals → quite important for Pt layers
Pt stress and brittleness
Hot Wire Anemometer - CVD 913/07/2017
• Evaporated thin Pt films are brittle
Pt
« Nanoscale size effects on the mechanical properties of platinum thin films and cross-
sectional grain morphology »
K. Abbas et al., J.Micromrch. Mictoeng. 26 (2016) 015007
• Thermal stress :
The film is made at a different temperature than the temperature
at which the device will be used. Differences in linear thermal
expansion coefficient between film and substrate induce a stress in
the film
« Stress and strain in polycrystalline thin films»
G.C.A.M. Janssen, Thin Solid Film 515 (2017) 6654-6664
Lowering the temperature the film will develop a tensile if it wants
to shrink more than the substrate allows it.

10. Pt - What can we do?
Hot Wire Anemometer - CVD 1013/07/2017
•Decrease the deposition rate :
• This should reduce the stress level of the Pt layer
•Increase the thickness:
• Stress modification
• The layer should be less brittle
•Heat the substrate during deposition / Bake :
• Recrystallization of the layer
• Increase the grain size
•Change the wire design :
• In order to allow deformation of the wire
•Change material
• We can try to use another metal
(Au? Ni? W?)
•Change deposition technique :
• Sputtering for compressive layer
To improve reliabity :
Pt PROBLEM :
Pt wires can work at room
temperature
BUT
they break at low
temperature

11. Hot Wire Anemometer - CVD 1113/07/2017
e-beam
evaporation
•Etch Ar+ : 24s@250V
•Ni : 100/400nm @0,25nm/s
Lift
Lithography
Ni wires by evaporation
AFC21
AFC24 AFC28 :
Ti/Ni/Ti
AFC29 :
Cr/Ni/Cr
Adhesion problem :
Interface layer is
needed

12. Hot Wire Anemometer - CVD 1213/07/2017
Pulvérisation
•Ti : 5000nm
PDC : 300-500W
p : 3-5 10-3mbar
Lift
Lithographie
Ti wires by sputtering
Double layer lithography After lift-off:After e-beam evaporation:
Energetic particle bombardment of the
layer during deposition can generate a
compressive stress of the layer. This is
suitable for our devices, in order to
compensate thermal stress.
The main parameter that control atomic
peening is pressure during deposition.
Lowering the pressure leads to a higher
polarization voltage for a given current
and an higher mean free path for
particles in the chamber, which means
more energetic particles striking the
surface of the layer.
ATOMIC PEENING:
Ar
-
+
Sputtering

13. Hot Wire Anemometer - CVD 1313/07/2017
Ti wires by sputtering - SEM
p↗
P↘
AFC27
AFC31
AFC31
PDC : 500W
p : 3*10-3mbar
AFC27
AFC27
AFC27
AFC26
PDC : 300W
p : 5*10-3mbar

14. Hot Wire Anemometer - CVD 1413/07/2017
Au pads
Evaporation
•Etch Ar+ : 24s@250V
•Ti : 10nm @0,1nm/s
•Au : 200nm @0,1÷ 0,5nm/s
lift
• Acetone soak
• IPA with US rinse
• N2drying
• AZ326MIF soak
• EDI rinsing
• N2 drying
• Solvent cleaning
• Dehydration : 2min@150°C
• LOR10A: 4000/2000/50
• Bake : 5min @170°C
• AZ1512HS: 4000/2000/60
• Bake : 1min 30sec @100°C
• Alignment
• Exposure MJB4 UV : 25sec
• 2min30 sec in AZ326MIF
• EDI rinsing
• N2 Drying
Lithographie
UV
PbSn and Au are a ternary eutectic
systems : when you put melted
PbSn on the Au pad you melt the
layer
• Add a coating on the pad
(e.g. Cr : 10nm @0,1nm/s)
• Use another solder (silver
conductive glue or other )
Pt
LOR+UV5
Au
Optional step:
Au contact pads in order to reduce the
series resistance of the wire
PROBLEM
Pb-Sn welding on Au :
Double layer lithography After lift-off:

15. Hot Wire Anemometer - CVD 1513/07/2017
Reminder : Deep Reactive Ion Etching
(DRIE)
Litho Bosch cycle…… clean
Passivation
Etching
Boost
Anisotropic
etching of CF-
polymer
Main
Isotropic etching
of silicon
isotropic
deposicion of CF-
polymer
Etching mask
fabrication
(usually thick resist or hard
mask)
Repetition of bosch
cycle
parameter adjustment could be
necessary
CF- polymer and etching mask
removal
..…repeat

16. Back side shaping
Hot Wire Anemometer - CVD 1613/07/2017
1. DRIE
0
50
100
150
200
250
0 50 100 150 200
Depth(µm)
width (µm)
ARDE - Aspect Ratio Dependent Etching
2. isotropic etch
3. cleaning
4. smoothing

17. Hot Wire Anemometer - CVD 1713/07/2017
Back side: prongs (1)
Prongs
Structuration
• DRIE with ramping to etch
properly almost through the
entire wafer
• “soft” DRIE to reach the SiO2
layer
• Solvent cleaning
• Dehydration bake: 2min@150°C
• HMDS (back) : 4000/2000/50
• Bake : 1min @110°C
• AZ4562 (back) : 4000/2000/70
• Bake : 5min @110°C
• AZ4562 (front) : 4000/2000/90
• Bake : 2min @110°C
• Back face alignement
• Exposition MA8 :20sec
• 3-4 min en AZ326MIF
• EDI rinsing
• drying N2
Backside
Lithography
•Control the slope in
order to keep the Si
structure as lond as
possible
•Avoid « grass »
formation
Litho : During DRIE
Focus
« Up»
Focus
« down »

18. Hot Wire Anemometer - CVD 1813/07/2017
Back side: prongs (2)
Nettoyage
• O2 plasma cleaning
Fin gravure
et lissage
• Isotropic etch 1 to attack the Si
sacrificial structure
• Isotropic etch 2 to smooth the
surfaces without damaging the
SiO2 membrane
•Do not damage the
Si02 layer
•Fixation method of
the sample on the
holder has to be
improved for a
better thermal
contact
Nettoyage
• O2 palsma cleaning
Nettoyage
• Plasma O2 cleaning (front)

19. Probe release
Hot Wire Anemometer - CVD 1913/07/2017
HF vapor
• Dehydration : 5 min @170°C
•HF vapor 100/150 nm
retrieving
• Retrieve the devices by
breaking the links
Si
SiO2
Pt
Before HF etching After HF etching
After HF etching

20. Probe release – new samples
Hot Wire Anemometer - CVD 2013/07/2017
AFC30
AFC31
Sputtered Ti (AFC30, AFC31).
New mask 2 - without links between wafer
and sensors.
The sensors stands on the silicon membrane
until HF-vapor etching .
Much better yield (up to 59 chips over 60)

21. Probe release – new samples
Hot Wire Anemometer - CVD 2113/07/2017

22. Shaped prongs
Hot Wire Anemometer - CVD 2213/07/2017

23. canne cryogénique
Tests
Hot Wire Anemometer - CVD 2313/07/2017
Puces montées sur leur support
Tête de la canne cryogénique
Pour l’immersion des puces dans le He liquide et
la mésure de résistance et température
Intérieur du cryostat d’étalonnage
la puce tourne à vitesse connue pour
simuler l’écoulement et retrouver les
paramètres de la loi de King
canne cryogénique Cryostat d’étalonnage

24. What to do ?
Hot Wire Anemometer - CVD 2413/07/2017
•Optimize DRIE parameters:
•We should keep in place the Si Structure as long as possible
•Modify Mask 2
•Inhomogeneities compensation
•Get rid of useless parts in the mark
•Reshape the links between probes and wafer ?.
•Increase etch selectivity towards SiO2
•Modify the etching parameters
•Bake the substrate
•Increase thickness
•etch-stop
•Effects on the stress levels should be verified
•Improve sample mounting and cleaning in the ICP :
•Better thermal contact during etch
•Avoid polymer redeposition on the front
•Use ar non-metallic material for the wires
•e.g. : insulating NbN
We can deposit it by reactive sputtering at SBT
Matthiessen’s law « non-métallic »
resistivity
T
ρ
0 T
ρ
0
• Improve sensibility at low temperature
• Improve mechanical reistance
→ New substracive process needed
(deposition-lithography-etching)

25. Hot Wire Anemometer - CVD 2513/07/2017
The End

26. Labs
Hot Wire Anemometer - CVD 2613/07/2017
CEA
INAC (Institute for Nanoscience and Cryogenics)
SBT (Low Temperatures service)
LRTH (Refrigeration and Thermohydraulics Laboratory)
“Turbulence” group
Main interest :
behavior of helium at very high Reynolds number
r est un nombre entre 0 et ½ qui
indique l’inverse du nombre dans
lequel se scindent les tourbillons.
η est l’échelle de dissipation.
Pour SHREK elle est de l’ordre du
micron jusqu’à quelques dizaines
SHREK
700m2 class 1000 (ISO 6) clean room
• 350m2 at 10.05
• 350m2 at BCAi (50A),
More infos : pta-grenoble.com
→ Lithography
→ Deposition
→ Etching
→ Caracterisation
Plateforme
Technologique
Amont

27. • masse↘ → résolution temporelle↗
• longueur ↘ → résolution spatiale ↗
avec l/d>>200 pour minimiser la conduction
thermique
Anémomètre à fil chaud
Hot Wire Anemometer - CVD 2713/07/2017
Fil métallique résistif traversé par un courant dans un fluide en
mouvement → refroidissement éolien (wind chill)
Dans un état stationnaire, on a équilibre entre :
• Effet joule
• Convection forcé
• …
Si la résistance du fil varie avec la température on peut évaluer
la vitesse du fluide :
!"# $% &#'( ∶
*+
*+ − *-
./
= 1 + 3 4
Différents montages possibles :
• CCA (Constant Current)
• CTA (Constant Temperature)
• …
I
U
• Anémomètre microfabriqué
« Fabrication and characterization of a novel Nanoscale thermal anemometry probe »
M.Vallikivi, A. J. Smits
IEEE Journal of microelectromechanical systems
L≈1mm
d≈1÷10μm
L≈ 50÷100μm
d≈1÷4μm
e≈100nm

28. Bilan de puissance
Hot Wire Anemometer - CVD 2813/07/2017
I
U
L≈1mm
d≈1÷10μm
Dissipation de la chaleur :
• Conduction vers le fluide
• Convection vers le fluide
• Conduction vers les supports
• Rayonnement thermique
On peut négliger les derniers deux
Flux de chaleur (loi de Fourier) :
!" = −%&'(
À intégrer sur A=πdL
̇* = +
,
!"-. = ℎ. (0 − (1
h=coef de transfert
Fil métallique résistif traversé par un courant dans un
fluide en mouvement → refroidissement éolien (wind
chill)
Principe : la puissance importée par le fluide donne une
mesure indirecte de la vitesse du fluide
Bilan de puissance :
23
24
= ̇W − ̇Q
Où:
•
23
24
: variation d’énergie stockée sous forme de
chaleur dans le fil
• ̇W = 70 89 : puissance dissipée par effet Joule
• ̇Q : puissance transférée depuis le fil vers l’exterieur

29. Nombre de Nusselt (Nu)
Hot Wire Anemometer - CVD 2913/07/2017
I
U
L≈1mm
d≈1÷10μm
Nombre de Nusselt :
!" =
ℎ %
&'
Efficacité du transport par convection par rapport
à la conduction
!" =
(")**+,-. /0+,*1é0é. /3/+4.
(")**+,-. /0+,*1é0é. (+0 -3,%"-/)3,
̇6 = 7
!" &'
%
89 − 8; = π 4 &' 89 − 8; !"
Le bilan devient :
dE
dt
= @9AB − π 4 &' 89 − 8; !"
Si CDE
CF
G; alors
MN
MO
= 0
Donc
@9AB = π 4 &' 89 − 8; !"
il faut savoir comment varie Nu en
fonction de U

30. Nu=Nu(Re)
Hot Wire Anemometer - CVD 3013/07/2017
Nombre de Reynold basé sur le fil !"# :
!"# =
% &
'(
Où
• % : vitesse « à l’infini » (quelque d à l’amont du fil )
• '( : viscosité cinématique à la température du film
On va considérer que Nu ne dépend que de Re.
Pour un nombre de Pradl )* = ⁄,
- ≃ 1
On a 0 ≃ 01
Avec les hypothèses d ’écoulement 2D potentiel et
stationnaire :
Lois de KING :
Nu = 1 + 2 6 !"7#
NB : Nu ∝ !"#
⁄9 :
→ écoulement laminaire ( à l’échelle
du fil)
Avec des Hypothèses moins restrictives :
Loi de KRAMER :
Nu = 0,42 )* ⁄9 >
+ 0,57 !#
⁄9 :
)* ⁄9 A
En général :
BC = DE + FE !"#
Le bilan devient :
!#G:
= π I J( K# − KE M
N
DE
+ FE !"#
Expérimentalement on a accès
à " = !#G
Si !# = !# K on peut avoir
accès à la température du fil

31. Nu=Nu(Re)
Hot Wire Anemometer - CVD 3113/07/2017
Relation fondamentale de
l’anémométrie
Loi de King
!"#$
!" − !&
= ( + * +
Avec :
( =
, - ./
0 !&
(&
* =
, - ./
0 !&
1
23
*&
Pour une dépendance
linéaire :
!" 4
= !& 1 + 0 4" − 4&
⇒ 4" − 4& =
!" − !&
0 !&
Avec 0 =
7
89
:89
:;