Dear MEMS colleagues,

Thank you all for taking time to respond on my email. At this moment I
have received over 10(!) mails, so the MEMS network seems to become
quite effective. Now I have read all these mails I am able to draw a
premature conclusion. Because you were so kind to take time for an
answer I think it is appropriate to give you all a collection of all
the responses. Although, I have not yet tackled the problem, there is
one answer (#6) which could prevent the forming of cracks in our
institute. I hope that this reply is also helpful for you. But
whatever is the outcome of your suggestions, Thanks!

1)  Some colleagues suggested to use aluminium or titanium instead of
chromium. Indeed these layers are much easier to apply than chromium.
Unfortunately, for us this is not an option. We are planning to use
chromium as one of the metals of a thermocouple since chromium has the
highest Seebeck coefficient. 2)  Others suggested that the problem
could the postbaking of the polymer. When the solvent is not
completely removed, there might occur out gassing during the
deposition. Indeed this is also critical, but we have baked the
polyimide up to 450 degrees in an oven with nitrogen purge to prevent
burning of the polymer. This procedure was not able to prevent the
cracks. 3)  A related solution was to preheat the sample inside the
evaporator or to slow down the deposition rate in order to prevent a
thermal shock in the polymer during evaporation. We have observed that
the cracks occur at the moment the evaporator is vented to room
pressure. This means that not the increase in temperature of the
polymer-chromium sandwich is critical but the ramping downwards. We
have tried to do this as slow as possible but with now result. We have
also tried to prevent the temperature of the sandwich to rise too
high. This was achieved by the deposition of a very thin layer and by
using a thermal aluminium block carefully attached to the sample by
way of lots of vacuum grease. The cracks survived this procedure! 4)
Another suggestion was to deposit a very thin chromium layer or to
deposit the complete layer in a few subsequent steps. First of all a
very thin layer is not an option because this would increase the
electrical resistance of this layer too. A high resistance makes it
much more difficult for the electronics to measure the electrical
signal coming from the thermocouple sensor (The signal to noise ratio
is getting much worse for a higher input impedance of the first
amplifier). Of course, the deposition of many layers could still be an
option. We tried this but saw directly that the deposition of 25nm of
chromium is already impossible; cracks all over. 5)  A fifth option
which was put forward was to deposit an intermediate metallic layer.
Again this is not an option in our case because this would shortcut
the thermocouple with a different thermocouple. Moreover, this mixture
of metals can slowly degrade the performance of the thermocouple in
time due to bulk migration. 6)  A special trick was suggested to one
of our colleagues: to roughen the polymer prior to the metal
deposition. But that person found no improvement at all. We have not
yet tried this solution ourselves but in our opinion this procedure
will improve the attachment but we do not see why it would prevent
cracks (After all, this rougher surface does not create a "corrugated
zone" or something similar.). 7)  The last explanation given by Mr/Mrs
I.Kanno might be very helpful (S)he wrote:

" I think that your Cr layer has tensile stress. So, If you deposit
the Cr layer with compressive stress, you will be able to prevent
cracking. I used to deposit 2um-thick Cr films on 2um-thick polyimide
layers using rf-sputtering, but I also was in trouble with the cracks
of the Cr. But I confirmed that the Cr films with compressive stress
were very smooth surface without cracks even on a polyimide layer.
Stress of the Cr film is easily controlled by the sputtering gas
pressure, that is, deposition under low gas pressure (high vacuum)
makes metal films compressive. Good luck, I.Kanno."

I fully agree with this solution. With all the experiments we have
done so far there seems to be a strong indication that the heart of
the problem lays in the stress of the chromium and has nothing to do
with the polymer directly. It is known that residual gases like oxygen
and water vapour out gassing from the vessel walls of the evaporator
can easily be incorporated into the deposited layer. For example
aluminium which is deposited at 3*10^-6 Torr is almost without stress
with respect to an underlying silicon layer. When the background
pressure is higher, a compressive (!!) stress is found due to the
larger number of foreign molecules in the deposited aluminium. When
the pressure is lower, the aluminium will shrink due to the expansion
coefficient. When starting a deposition run for a chromium layer it is
observed that the background pressure is quickly falling. This is
caused by the gettering effect. This means that lots of residual gas
is consumed by chromium particles travelling from the source to the
sample. So, our experience somehow contradict your findings. You found
for chromium that a higher background pressure will result in
compressive stress. We found for aluminium exactly the opposite.
Nevertheless, your suggestion to vary the background pressure makes
sense and we have looked at what kind of vacuum we are able to reach
in our (MESA) institute. In MESA we have access to an e-beam
evaporator which is able to reach approximately 1*10^-6 Torr. We have
also a sputter machine with a cryogenic pump able to reach 1*10^-7
Torr. So, my question towards Mr/Mrs I.Kanno is: 1)  What is the
vacuum we need to be able to deposit a film under compressive stress?
2)  Do you know why my information concerning when the stress is
compressive or tensile when varying the background pressure
contradicts with yours? I found this information for example in:
A.Kubovy and M.Janda, The influence of residual gas pressure on the
stress in aluminium films, The research institute of electrochemical
ceramics, 28 September 1976.

Thank you,

Henri Jansen, Cristina Neagu
MESA Research Institute,
University of Twente,
The Netherlands.

Dr. Cristina Neagu
Associate Researcher
MESA Research Institute
University of Twente / EL-TTM
P.O. Box 217, 7500 AE Enschede
The Netherlands
Phone:  x31-53-4892805
Fax:    x31-53-4893343
email: c.neagu@eltn.utwente.nl