Scale compression versus type of capillaries (figure)

31 Aug 2014

Hi guys, 

For your reference, here is the figure for the scale compression (slope of the ETF) versus intercept of the heated gas. The symbols represent the type of system and capillaries used. I thought some of you would enjoy seeing the results here, especially if you were not at the workshop. And thanks to all of the labs who contributed data to this!

Cedric

Cedric's picture
Cedric
Offline
Joined: Feb 6 2011
P.S. some labs are not

P.S. some labs are not plotted here because I don't have their intercept of the HG. So if you have not sent me this info, feel free to still do so and I will update the plot.

Paul Dennis's picture
Paul Dennis
Offline
Joined: Sep 16 2011
The positive correlation

The positive correlation between intercept and gradient of the ETF is expected. Fot a wrg equilibrated at 25 degrees C then there is a simple relationship between intercept and gradient of ETF: gradient of ETF = 0.9/intercept

Plotting the gradient of ETF versus -1/intercept should give a slope of 0.9.

My take home message from this plot is:

(i) the scale compression (gradient of the ETF) is independent of the capillary material. There is complete overlap between data for stainless steel, nickel and fused silica. At UEA we use stainless steel and have never had any issues with these capillaries.

(ii) There is a wide range in intercept for heated gases that may result from one, or a combination of several issues including, but not exclusively:

(a) labs using different reference gases that are not at equilibrium at laboratory temperature.

(b) some instruments may have an offset as a result of problems with the dual-inlet. To understand if this is the case it would be useful to have results of zero enrichment tests in which the wrg is measured against itself.

(c) The heated gas correction is not fully understood or controlled. I know that Brad Rosenheim has pointed out in his paper on the IsoPrime that there is a co-variation between gradient of the HGL and the intercept for those labs all using nominally the same Oztech reference gas.

For reference our current intercept is -0.73 and gradient of the ETF is 1.23.

 

Paul Dennis, Stable Isotope Laboratory, Environmental Sciences, UEA

Cedric's picture
Cedric
Offline
Joined: Feb 6 2011
Paul, I agree, the figure can

Paul,

I agree, the figure can mean many things, and we did discuss that at the workshop. But we also noticed that in general fused silica plots lower than both electroformed nickel and steel. Also, steel shows more scatter. Now some of us (at least three labs) had problems with our steel capillaries. These are not subtle problems, so you would know if you had them. In clear, the heated gas values were indistiguishable to the reference gas, which we interpret as resetting of the clumped value through capillarie flow with minute amounts of waters. Changing the capillaries to electroformed nickel or fused silica did work.

Cedric

Cedric's picture
Cedric
Offline
Joined: Feb 6 2011
Paul,  Just out of curiosity,

Paul, 

Just out of curiosity, what is your reference gas? Not Oztech?

Cedric

Paul Dennis's picture
Paul Dennis
Offline
Joined: Sep 16 2011
Cedric,   No we don't use

Cedric,

 

No we don't use Oztech gas as a reference. We prepare our own in the following manner:

(1) React ca 5gm of BDH marble chips with as supplied 70% orthophosphoric acid. We leave the reaction to go for at least a month. This is to allow the clumped composition of the gas to equilibrate to room temperature.

(2) This gas is then dried by passing it slowly through two traps at -120 degrees C, and through a porapak column at -20 degrees C. To do this, preventing freezing out of the CO2 in the cold traps at their low temperature, we drop the pressure of the gas by expanding it into a large reservoir (ca. 25L) prior to drying.

(3) The gas bulk isotope composition is calibrated against NBS 19 and LSVEC.

I take the point that there might be issues with water adsorbed on the stainless steel capillaries. However, these can be flamed if necessary (something we've never had to do), or baked. If the ms only sees dry gas then there should be no build up of adsorbed water.

Paul Dennis, Stable Isotope Laboratory, Environmental Sciences, UEA

brosenheim
Offline
Joined: Sep 15 2011
Hi all - Cedric - thanks for

Hi all -

Cedric - thanks for posting the plot.  Here are my observations:

1. As Paul pointed out, the spread of data along the x-axis represents, perhaps, a reflection of the large problems previously associated with steel.  The problem has been stated to me that everything collapses to near the 25 degree equilibrated line due to reorganization after contact with water presumably in the rugose steel capillaries.  An intermediate problem of values approaching this line (but not quite getting there) would be effectively scale compression.  I see that there are two steel labs and two nickel labs above 10% compression.  10% is arbitrary, but definitely higher than the majority of labs no matter the capillary. 

2. The spread along the y-axis may be due simply to different reference gases.  The intercept is our model-based estimation of the clumped composition of the reference gas.  I showed last week that with a large heated gas slope and no PBL correction, one runs the risk of a poor estimate of this value.  Labs with lower slopes (not of the ETF, but of the heated gas line) should have a better estimate of the clumped isotope composition of their reference gas.  Therefore it may be interesting to color the points according the values of their reference gases or the slopes of their heated gas lines.

I think a lot can still be learned about these data the way you have plotted them, but there are unexploited riches in the plot.  I think we also need to develop a working set of alternative hypotheses about why some labs have observed this collapse toward 25 degrees in their reference frame.  I would contend that if it didn't happen to us in the swamps of southern Louisiana in a lab with two leaky exterior doors and several 100-y old windows and steel capillaries, it may be a red herring.  Especially when one considers how many times we vented our machine when we were having electrical and/or filament problems.

 

Brad

Brad E. Rosenheim

Assistant Professor

College of Marine Science

University of South Florida

Saint Petersburg, FL, U.S.A. 33701

http://www.marine.usf.edu/rosenheim/BER-Group.html

hagit
Offline
Joined: Dec 21 2011
Paul, Can you explain in more

Paul, Can you explain in more details how do you clean the reference gas? I use Oztech and in my experience also other CO2 gas cylinders are clean. But this is useful information if one wants to use something else i the future, or in general for how to clean a large amount of CO2.

With respect to the plot: as Brad indicates, the intercept of the heated gas line is the compressed value of the reference gas (with a minus sign). so the plot is only meaningful for people using the same gas, though having 25°C equilibrium is pretty close to the Oztech value (in my case the oztech in ~0.02‰ higher than the 25°C CO2-H2O). we should perhaps normalize that somehow.

hagit

Paul Dennis's picture
Paul Dennis
Offline
Joined: Sep 16 2011
Hi Hagit, with reference to

Hi Hagit,

with reference to your second point I agree and that was what I wanted to point out in my first comment above.

As for the reference gas preparation I'll prepare a short document detailing exactly how we go about it and post it here, or to clumpy. This may take a couple of weeks as I'm about to depart on a two week undergraduate field trip and may not have internet access. Briefly we carry out the reaction in a 1L flask using 5gm of UEACMST (BDH marble chips) and orthophosphoric acid. Because it's difficult to freeze out water when the CO2 pressure is close to 1atm we expand our CO2 into a 25L flask to reduce the pressure to close to 40mb. This allows us to cryodistill the CO2 into two 500mL flasks through two traps at -120 degrees C. The first is a 'glass bulb' trap that can accommodate a large volume of water. The second is a loop of 6mm od glass tubing (4mm i.d.). We also have a loop of galss packed with porapak held at -20 degrees C.

I agree with your observation that most commercial gases are clean and dry, also are close to clumped equilibrium at ambient temperatures. Our choice for making our own reference gas was really an issue with the cost of Oztech gases and having the equipment to make our own with a bulk sotopic composition close to the samples we're measuring. 

Paul

Paul Dennis, Stable Isotope Laboratory, Environmental Sciences, UEA

Cedric's picture
Cedric
Offline
Joined: Feb 6 2011
Brad - I can try to use

Brad - I can try to use different symbols for HG slopes, although this might be tricky as some labs use PBL, some not. I will clear this with them.

Paul - it would be fantastic if you could post your receipe here. Just create a new thread with a clear title, this way we keep our discussion on topic. 

Regarding the steel capillaries, John mentioned at the clumped isotope workshop that even flaming his capillaries did not work. It's not about steel working or not - it seems to be about good bactch, or bad batch. And if you have a bad batch there is nothing you can do about it (or so it seems).

Cedric

Paul Dennis's picture
Paul Dennis
Offline
Joined: Sep 16 2011
In the comments above Cedric

In the comments above Cedric pointed out that the silica capillaries plotted with a lower value for the heated gas intercept than the bulk of the observations for stainless steel. Similarly the results for Ni capillaries also tended to plot below those results for stainless steel. The interpretaion was that during flow through the capillaries there was the possibility of re-equilibration as a result of adsorbed water in the capillaries.

I think this is an incorrect interpretation of the data. If there is re-equilibration then the effect would be to compress the scale such that one would see larger slopes on the ETF. If one looks at the spread of the data on the x-axis then there is no general pattern. In fact an instrument with Ni capillaries shows the greatest degree of scale compression, and with the stainless steel capillaries having similar compression to other capillaries.

What is really interesting about this plot is the spread of data along the y-axis, or heated gas intercept. I've replotted the data below and included the expected relationship between intercept and slope of the ETF for heated gases that are measured wrt to a working reference gas (wrg) that is at equilibrium at room temperature. Hagit has observed that most of the commercial gases she has observed are very close to equilibrium at ambient temperatures. I've also observed this. Also I think most laboratories, with the exception of Brad Rosenheim's and mine, are using Oztech gas as their wrg. 

Then one has to ask what does the spread of data along the heated gas intercept tell us? Ordinarily one would take the modulus of the intercept (plus 0.026) as the wrg composition. Instruments that compressed would lie along the trajectory shown. However, what we observe is a very large spread of data about this trajectory. For example instruments with compression between 5 and 10% have intercepts that range over more than 0.2 per mille. To me this can mean one of a few things as I pointed out above:

(1) The wrg in different laboratories are not in equilibrium at ambient temperature. I think this unlikely given Hagit's and others observations that their wrg are close in composition to water equilibrated gases.

(2) There are possible issues with change over valve and capillary performance. i.e. there is fractionation of the sample side wrt to the reference side. This can be tested using zero enrichment tests. 

(3) The nature of the correction for non-linearity is poorly constrained. Brad Rosenheim has pointed out that when he corrects his data using PBL or by a linear extrapolation of heated gas data to zero d47 he gets different results for the intercept.

I think it crucial that one has a better understanding of the origins of the scatter in this plot if one is to understand some of the issues of differences between laboratories. It indicates to me that there are many possible sources of measurement error which may sytematically skew the results. It may be that these all drop out when one uses the ETF but I'm not sure this has been adequately demonstrated yet.

Out of interest I've plotted the MIRA data where we have a heated gas intercept of -0.73 and scale compression of 1.23. This is at the high end of compression and is something we are actively exploring to see if we can lower this value.

Paul Dennis, Stable Isotope Laboratory, Environmental Sciences, UEA

hagit
Offline
Joined: Dec 21 2011
Paul, We should remember that

Paul,

We should remember that the intercept of the HGL varies with time in most labs, so there is no reason to expect it to be the same lab to lab. The intercept is the negative of the reference gas compressed. namely, it varies when scale compression varies. we usually see small variations over time, but sometimes big jumps (e.g., we had to bake the machine following a long power outage and unplanned venting. this reduced our scale compression significantly and it took a few month for the source to compress again).

hagit

Paul Dennis's picture
Paul Dennis
Offline
Joined: Sep 16 2011
Hagit, I understand this but

Hagit, I understand this but the intercept you observe should move along the trajectory I plotted in the figure. Does the intercept of your 25 degree C equilibrated gases also show such small variations, and occasional big jumps. If the problem is re-equilibration due to the presence of adsorbed water in the capillaries, source or where ever then it should not affect the value of 25 degree C equilibrated water samples which is close to zero wrt the wrg.

If the heated gas data do not plot on the trajectory of the intercept-ETF slope that I've indicated then I think there might be some other unspecified effects coming into play here. I don't know what these are.

Paul Dennis, Stable Isotope Laboratory, Environmental Sciences, UEA