Abstract:
The original fomblin oil based pump malfunctioned, so an attempt was made at using non-fomblin oil rotary vane pump for a plasma etching (ashing) process by diluting the oxygen infused gas mixture with nitrogen at the pump inlet. The amount of nitrogen dilution did not allow the ashing chamber to maintain low enough vacuum below 700mTorr.
Instead, a different fomblin based pump was used, and a leak was found and fixed. Additionally, the free CERN ultra high vacuum simulator Molflow+ was used to simulate the ashing chamber under ultra high vacuum conditions for fun and giggles.
Introduction:
The plasma ashing process is a common materials processing technique used to etch away polymer/dielectric layers from polymer multi-layer and wound capacitors in order to extrude and reveal the electrodes for termination. However, this process uses a gas mixture of argon and oxygen. Oxygen is highly reactive with many oils used in vacuum pumps. So it is necessary to use inert oils such as fomblin or other polyfluoropolyether oils1.
The plasma ashing process at my work usually uses a rotary vane fomblin pump. However, the pump recently burnt out its motor. Since this is the only ashing chamber at my work, it was critical to get it going once again. So while waiting for a fomblin based pump replacement, an attempt was made to dilute the oxygen gas with nitrogen to have a mixture of at least 80% nitrogen and 20% Oxygen. This is approximately the gas ratio of air which has shown to be fairly inert when it comes to reactive oils.
Unfortunately, the idea did not work and I received a different fomblin based pump after a few days of trying.
Results:
Nitrogen Dilution with Edwards Pump
The set-up: An Edwards E2m40 2-stage rotary vane pump with hydrocarbon oil was used for the oxygen dilution process. A t-section KF flange was used at the beginning of the foreline in which both the foreline and the nitrogen line were connected.
(Disregarding the argon in the process) The necessary nitrogen flow rate needed to equal a 80% Nitrogen to 20% Oxygen mixture was calculated to be 1.2L/min. Sorry, I can't say exactly what the oxygen flow rate for company reasons.
An Omega mass flow controller (MFC), calibrated for N2 gas, was used to leak 1.2L/min of nitrogen into the beginning of the foreline.
Figure 1. Set-up of Edwards E2m40 Pump system
A vacuum pumping curve and pressure vs. time curve of the E2m40 on the ashing chamber with and without the nitrogen leak is shown below (note: the plasma process was not turned on therefore no oxygen or argon flow was introduced. These curves are purely of the pump on the chamber) .
Figure 2: Pumping of Ashing Chamber with Edwards E2m40. Pressure vs time (top) and Pumping speed vs pressure (bottom)
The pressure was written down with respect to time. In turn, the pumping speed was calculated using the formula given by Oerlikon Leybold Fundamentals of Vacuum Technology, Chapter 2.3.1.12.
-dp / dt = Seff * p / V
Seff - Effective Pumping Speed
p - pressure in Torr
t - time
V - volume, in this case the chamber volume is 74L
Notice though that this is the equation for pumping in a rough vacuum. This, however, can be used as a simplified model to obtain an approximate effective pumping speed (Seff) even if we are working between medium and high vacuum. Note that the differential of pressure with respect to time was calculated by a simple difference equation dp = (p2 - p1) and dt = (t2-t1).
The pumping speed curve may look weird at first, as most companies spec their pump with a nice logarithmic graph. The graph they spec, like the one given by Edwards3, is under ideal conditions. It usually does not take into effect the pumping of water vapor, other gases, desorption, etc.
If one looks at the Pressure vs. Time curve, one can see that it's relatively similar to that proclaimed by Charles Bishop in Vacuum Deposition onto webs, films, and foils4. The first corner 12 Torr indicates that along with air, the water vapors are now being pumped. The second corner, which is around 1 to 5 torr indicates that hydrogen is now being additionally pumped.
Conclusion
It is clear that a 1.2L/min N2 leak immensely effected the base pressure of the system. The base pressure jumped nearly 100x with the nitrogen and settled at around 1.7 Torr.
The difference of the base pressures is approximately 1.7 Torr. We can calculate the amount of gas that has settled in equilibrium in the system using the ideal gas law:
PV = nRT
Note that we use the difference in our pressure, which just happens to be approximately the base pressure with N2 leak to 1 decimal place, in order to find the amount of nitrogen contributing to the pressure. First convert pressure from torr to pascal and volume from liters to cubic meters:
P = 1.7 Torr = 226.648 Pa (according to google)
V = 74L = 0.074 m3
n = PV/(RT) = (226.648 Pa) (0.074 m3) / ( (8.314 J K-1mol-1) (273 K))
n = 7.389 e-4 moles
using molar volume density of nitrogen to convert to volume:
v = n * ρ = (7.389 e-4 moles)/(0.0446 moles/L)
v = 0.0165 L of N2 gas left in the ashing chamber
Now converting our flow rate to seconds:
Sleak = 1.2L/min = 0.02 L/sec
So at about 1.7 Torr the pump reached equilibrium with the N2 leak (plus any other leaks and desorption mechanisms). So the pumping speed seems to be about 1.2L/min or, equivalently, 72L/hr. This seems correct.
Unfortunately, I am not quite sure as to why this nitrogen leak caused such a severe effect. I was not able to continue examining it as I received another fomblin based pump to get the ashing chamber into operation.
Fomblin Adixen Pump Characteristics
Because this pump did not require any gas insertion, it easily pumped down to the tens of mTorr range. The figures below show the pump down characteristics.
Figure 3: Pump Down characteristics for Adixen Fomblin pump. Pressure vs Time (top) and effective pump speed vs. pressure (botom) Note: the system had a monomer injection device which was leaking, therefore "no monomer injection device" means no leak.
Looking only at the red line (With monomer injection device), one can see that the Adixen pump reached base pressure much faster than the Edwards. Additionally, the pumping speed looks much more logarithmic. However, it is noticeable again that the both graphs have a corner (node) where it seems the pumping of water vapors and hydrogen gas begins.
After pumping down, I decided to check the leak rate of the system. Lo and behold, the system was leaking almost 90mTorr/min.
So we had a leak. The figures below show the pressure characteristics of the leak.
Figure 4: Leak characteristics of vacuum chamber. Pressure vs. Time (top) and leak rate vs. pressure (bottom)
Now let me explain: after a helium leak check test, I found that an injection device connected to a port on the bottom of the chamber was leaking. This device was used to inject monomer into the chamber in a past experiment. Therefore, the chamber was leaking from a monomer injection device (MID).
Although the graph above shows the leak rate in L/minute, the leak rate according to Oerlikon Leybold Fundamentals of Vacuum Technology, Chapter 5.4.12. (which I like to call the leak throughput) for a pressure rise test is:
QL = V * Δp/Δt
With a constant chamber volume V = 74L, we calculate QL to be
444 mTorr*L/min with the MID still connected. Without the MID connected and instead a KF blank cap on the port the QL is about 4mTorr*L/min. Removing the MID reduced the leak rate by two orders of magnitude, which is quite significant.
Calculating Conductance of Foreline
According to Howard Tring5, the conductance can be approximately calculated (with error) of a straight pipe with the graph shown here. So, the foreline valve I have attached to the ashing chamber and pump is approximately 1.5 meters in length with an inner diameter of 40mm.
If we follow from the top of the chart and pick the 40mm diameter curve, we can see that at 1Torr, the conductance is still in viscous flow and is approximately 300 L/sec. This is for a a length of 1 meter. For 1.5 meters, we divide this result by 1.5 to obtain:
Cfore = 200L/sec
Now, if we combine that with the pumping speed of the Edwards pump shown here at 1Torr we can find the effective pumping speed. For this, we use the Edwards pumping speed of 30m3/hr at 1mbar (or .75 Torr). We convert the pumping speed to L/sec:
S = 30m3/hr * 1000L/m3 * 1 hr/360sec = 83.33L/sec
The pump and foreline are in series so we have to do a series addition of conductance and pumping speed:
1/Seff = 1/S + 1/Cfore = 1/83.33 +1/200 = 0.017
Seff = 58 L/sec
This result is almost 6 times higher than the actual pumping speed we obtain in the above curve. Unfortunately, it is very difficult to take into account all the various aspects which would effect this result, including the fact that the foreline is not a straight pipe and is instead a flexible wire-reinforced tube.
MolFlow+ Simulation
For fun, I decided to try out using the CERN ultra high vacuum simulation software: Molflow+.
This software is not meant for medium to high vacuum like I work with. It uses montecarlo simulation which is a statistical algorithm to simulate the statistical process of molecular gas dynamics, specifically ultra high vacuum pumping. So it is important to understand that these results will have little to no bearing on the above, and I did for fun and to break into new software.
First, I drew up the ashing vacuum system in autocad as shown below. The drawing is a very simple drawing as Molflow+ uses simple solids to compute. The pipe is a 440cm pipe (about 1.5meters) with an inner diameter of 40cm. This is the foreline.
Notice that there is a T-section at the inlet of the foreline. I did this to simulate the N2 gas leak. However, Molflow+ does not have an option for constant gas feed so I had to settle for about
4 mTorr*L/sec.
I added a pumping speed of 10L/sec at the beginning of the foreline.
Figure 5: Drawing of Ashing Vacuum Chamber System in AutoCAD
The results below show the simulation at two different times. One can see that after about 4 minutes (bottom), the system seems to reach equilibrium.
As one can see, the simulation gives a wall pressure of about 0.007mbar to 0.008mbar which is approximately 5mTorr.
Citatiaons: