The turbomolecular pump arrived today. The power supply should be here in a few days. I have not decided yet whether to use an oilbased roughing pump with a foreline trap and valve or go with an oilless diaphragm pump, such as is offered by Pfeiffer. An oilbased pump will allow a lower vacuum, but may not be worth the extra trouble.
MIZUNO REPLICATION AND MATERIALS ONLY


Where's everyone gone?
Personally I'm for the most part playing the lottery with different materials and experimental conditions (electrolysis) than Mizuno, following Edmund Storms' suggestions from his papers, so what I'm doing isn't really intopic for this thread. Storms believes that a Mizunostyle burnished material may produce energy in an electrolytic cell as well.
I'm not geared to measure any heat in the watt–fewwatts range but if it works it might produce lowlevel radiation emission which I could detect with a Geiger counter. Since a few weeks ago in different experimental attempts I found it to be sensitive to air flows due to charged radioactive particles from Radon progeny accumulating on its exposed high voltage portions, this time I put it inside a wellclosed plastic box that is not exposed to any air flow.
I found out recently that Storms had something similar occurring as well and he reported it in one of his papers in 2012. As an air flow calorimeter is used in the original Mizuno R20 experiment which might circulate air around any radiation detector in the environment, this could be an intopic observation.
Excerpt from https://www.lenrcanr.org/acrobat/StormsEnatureofen.pdf

Tricky that  the oxide layer formed by presumably intense heat is  potentially at least  physically different to that formed at low temperatures it may for example be glassy in nature, rather than abrasive. But Ed is a careful and skilful experimenter, and has probably considered things like that.
Alan, you are correct. In my case, I could have formed a glassy layer that would not allow transfer of Pd. I was simply trying to see what would happen without going to a lot of effort. A more complete study would test the effect of the oxide produced at different temperatures. In any case, the Mizuno effect would not work unless Pd were stuck to the surface while being mixed with inert inclusions, which could be particles of NiO. This condition would require the oxide layer to have some special characteristics, which I have not yet explored. According to my theory, the presence of and the nature of the inclusions is one of the critical variables. I might explore the various variables at the end of the summer. Perhaps by then the effect will have been replicated and someone might be interested in funding my study.

An oilbased pump will allow a lower vacuum, but may not be worth the extra trouble.
If you have a good turbo pump then an oilfilled pump is fine. I'm using a single stage Edwards oilfilled roughing pump with a turbo and we have seen 4.5 x 10^{8 }mB without a foreline trap though I have one on the way.

Any concern about oil vapor getting into the turbo etc without the foreline trap? I'm in the same situation, waiting for foreline trap to arrive.

Any concern about oil vapor getting into the turbo etc without the foreline trap? I'm in the same situation, waiting for foreline trap to arrive.
Well, having changed the filaments, rebaked the QMS and pumped it down for a couple of days we are now at 3.4x108mB  and the foreline trap only arrived today and isn't fitted.
So maybe depending on your roughing pump you don't have to worry. They are mostly to trap water vapour anyway.

Well, having changed the filaments, rebaked the QMS and pumped it down for a couple of days we are now at 3.4x108mB  and the foreline trap only arrived today and isn't fitted.
So maybe depending on your roughing pump you don't have to worry. They are mostly to trap water vapour anyway.
Rather than worry about oil contamination, I purchased a Pfeiffer MVP2023DC dry diaphragm pump. Pfeiffer Vacuum recommends this type of fore pump for their TMU262 T/M pump. Based on Pfeiffer's data, it should be possible to attain 1e8 mbar with this combination of pumps.

Attached is a plot of Log excess power vs 1/T with an uncertainty of ±2 watts shown for the Mizuno data. A linear relationship is clearly produced, which is consistent with previous similar plots of excess power published in J. Cond. Matter Nucl. Sci. 20, 8199 (2016) and in my second book. The behavior is not consistent with the behavior being caused by error. This behavior is consistent with my theory and provides another way to test whether the excess power results from LENR. I realize the skeptics will propose many different prosaic sources, but I suggest a more positive interpretation would be more useful, at least initially.

Ed,
Great work! That is very interesting. It is an Ea of 0.174eV or 4kCal/mol.
I almost said I thought that tipped things very strongly towards this being some chemical (or chemically mediated LENR) reaction. It certainly could do that with a bit more data. But alas it does not yet.
I think it would put the balance of probabilities firmly against calorimetry errors, and towards some reaction, for the R19 results, if we can include the 23C values (rightmost on graph).
My comment however relates to the error bound of the 23C figures which at roughly 2W has a very large error bound on the log scale  effectively going to infinity. Since 0 excess might be expected here from any number of mundane calorimetric errors I think we need to consider the R value of all the other points linear on the Arrhenius plot versus some other relationship typical of calorimetry. Given more points we could easily disambiguate this.
The error bars on your graph are much too large, corresponding to +/ log10(2) (a +100%, 50% error). You meant to do the +2W before not after the log!
Looking at all the other points a linear fit on Arrhenius graph does not quite seem right. It would be fascinating to see what sort of relationship these made on some other graph and compare R values.
More investigation of this relationship (and checking that it remained for R20 new style not so good but still works) would strengthen the paper. That removing just one set of (contentious, due to large error bound) points here completely alters the conclusion is annoying. It would not be true if we had more points.
Best wishes, THH

Taking (I am too lazy to do all the R19 points)
Power out Reactor temperature (C) 38.6 239 78.3 386 10.4 145 41.8 234 41 229 39.6 232 76.1 320 102 383 97.6 383 92 381 92 375 35 235 34 236 2.1 24 2.9 23 35 232 we get an R value of 0.95 looking for a linear correlation:
R Calculation
r = ∑((X  M_{y})(Y  M_{x})) / √((SS_{x})(SS_{y}))
r = 56635.95 / √((206873.938)(17132.04)) = 0.9513
Meta Numerics (crosscheck)
r = 0.9513
Removing the 23C and 24C room temperature points we get a higher R value of 0.98.
Now look at these points on an Arrhenius plot:
P T/C 1/T/K log10(P) 38.6 239 0.001953 1.586587 78.3 386 0.001517 1.893762 10.4 145 0.002392 1.017033 41.8 234 0.001972 1.621176 41 229 0.001992 1.612784 39.6 232 0.00198 1.597695 76.1 320 0.001686 1.881385 102 383 0.001524 2.0086 97.6 383 0.001524 1.98945 92 381 0.001529 1.963788 92 375 0.001543 1.963788 35 235 0.001969 1.544068 34 236 0.001965 1.531479 2.1 24 0.003367 0.322219 2.9 23 0.003378 0.462398 35 232 0.00198 1.544068 R Calculation
r = ∑((X  M_{y})(Y  M_{x})) / √((SS_{x})(SS_{y}))
r = 0.004 / √((3.957)(0)) = 0.9907
Meta Numerics (crosscheck)
r = 0.9907
A very good negative correlation, consistent with Ed's above (I wanted to check my working, and that not using all the points was OK).
Now remove the contentious (because high error bound on log scale) 24/23 C points:
R Calculation
r = ∑((X  M_{y})(Y  M_{x})) / √((SS_{x})(SS_{y}))
r = 0.001 / √((0.969)(0)) = 0.9768
Meta Numerics (crosscheck)
r = 0.9768
The Arrhenius correlation is only much more significant than a linear one if you include points that (on the Arrhenius graph) have a very large error. To be fair, you should include these points for the linear match and exclude them for the Arrhenius one. In which case Arrhenius wins. But then it is unfair in a different way, the linear fit is needed over a much larger temperature range than the Arrhenius one. So on balance I think the best we can do is remove these points that do not work on an Arrhenius graph from both the fits.
For a better comparison you would need to do a fit weighted by the error bound of each point. But that is more complex. Really the big difference is the ones I've removed, which have the same error as the others on a linear graph but a much much higher error bound on a log(P) graph.
It would also be instructive to do a comparison between power in and power out to see whether the relationship there was closer to linear, although in this system we would expect some nonlinear corrections to calorimetry errors due to (for example) the nonlinear natural convection cooling of the calorimeter case, and higher order case temperature dependent errors.

Ed,
Great work! That is very interesting. It is an Ea of 0.174eV or 4kCal/mol.
I almost said I thought that tipped things very strongly towards this being some chemical (or chemically mediated LENR) reaction. It certainly could do that with a bit more data. But alas it does not yet.
I think it would put the balance of probabilities firmly against calorimetry errors, and towards some reaction, for the R19 results, if we can include the 23C values (rightmost on graph).
My comment however relates to the error bound of the 23C figures which at roughly 2W has a very large error bound on the log scale  effectively going to infinity. Since 0 excess might be expected here from any number of mundane calorimetric errors I think we need to consider the R value of all the other points linear on the Arrhenius plot versus some other relationship typical of calorimetry. Given more points we could easily disambiguate this.
The error bars on your graph are much too large, corresponding to +/ log10(2) W (a +100%, 50% error). You meant to do the +2 before not after the log!
Looking at all the other points a linear fit on Arrhenius graph does not quite seem right. It would be fascinating to see what sort of relationship these made on some other graph and compare R values.
More investigation of this relationship (and checking that it remained for R20 new style not so good but still works) would strengthen the paper. That removing just one set of (contentious, due to large error bound) points here completely alters the conclusion is annoying. It would not be true if we had more points
Best wishes, THH
Thanks THH for your efforts. Yes, the data could be treated in the detail approach as you applied if an exact value for the activation energy were sought. Instead, we are only trying to determine whether the data shows a basic logarithmic behavior, which it does although not with the exactitude we both would like. Other similar plots of other data show a more exact behavior.
According to my theory, the process that limits the rate of power production is the rate at which D can diffuse in the PdD to reach the NAE where the nuclear reaction takes place. In other words, the rate of diffusion is the throttle that controls the rate at which the nuclear process can take place, rather like the rate at which gasoline can be delivered to the location where the chemical reaction takes place in an engine. (I need to point out, regardless of how the nuclear process operates, this chemical effect will have some influence.) Therefore, the activation energy would be expected to close to the activation energy for diffusion, which it is. So yes, a chemical process is controlling the rate of overall energy production from a nuclear process that occurs very rapidly at many NAE sites. The measured power is the average from millions of sites and the temperature is the average of all these sites. Consequently, the calculated activation energy will be complicated by some sites being nearer a source of D than others and some sits being hotter than others. We can only obtain a value for the average. Nevertheless, the value is near that produced by diffusion of D in PdD. I suggest this process needs to be considered when explaining the behavior of LENR. As you suggest, hopefully the people studying the burnishing process will take the time to explore the temperature effect in more detail.

OK  done properly with ALL 55 R19 points, and with everything except the zero power in point that has v high errors bars on the Arrhenius plot.
2nd order fit given as well only when this is much better than linear fit.
NB  Pout below = EXCESS output power
Pin vs Pout
Correlation coefficient = 0.9757817798012672
1000/(Treact/K) vs log10(Pout)
Polynomial degree 1, 55 x,y data pairs.
Correlation coefficient = 0.9855366276187667
Polynomial degree 1, 53 x,y data pairs.
Correlation coefficient = 0.9706890155329876
Treact vs Pout
Polynomial degree 1, 55 x,y data pairs.
Correlation coefficient = 0.9416562739977635
Polynomial degree 2, 55 x,y data pairs.
Correlation coefficient = 0.9831476333714042
Polynomial degree 1, 53 x,y data pairs.
Correlation coefficient = 0.9807341370418086

Fascinating, but I don't see why one would excruciate over decimal places in a correlation coefficient, when in broader terms, we still aren't sure which direction to rub our mesh in !
Having said that, the Mizuno plot of exponentially increasing power output with temperature is excellent motive to try a number of different combinations of direction !

A detailed study of the high (3kW) power output Ni/Pd mesh structure and how to reproduce it effectively would now seem to be the way forward.

Fascinating, but I don't see why one would excruciate over decimal places in a correlation coefficient, when in broader terms, we still aren't sure which direction to rub our mesh in !
Having said that, the Mizuno plot of exponentially increasing power output with temperature is excellent motive to try a number of different combinations of direction !
For any replicator. If you can get more (say at least 4 definite clusters) accurate points on the Arrhenius graph  for R19 Mizuno has two clusters, + a few outliers  we can see clearly whether an Arrhenius relationship fits the power out better than something characteristic of calorimetric errors are mistakes. That provides a lot of additional validation, and even if it does not fit Arrhenius the shape of the graph may provide insight into what is going on.
You don't need as many points at each power level as Mizuno provides in the R19 data. If he had done the same number of tests, but split equally between 50W, 100W, 150W, 200W input, the data would be much more interesting. That change would be no more effort. (Note that zero power in / room temperature is a difficult point to measure because the calorimeter power measurement error translated into a very large error bar on the Arrhenius plot when the power out is low).
More than 4 points would be good, but you need multiple tries at each point to estimate the errors/variability. I'd say after 5 runs on any given set of parameters it would be worth changing temperature while keeping otehr parameters the same, if you have capability of more runs. I'd also note that stepping temperature over 6 hour or whatever increments is an effective way to get out temperature dependence in data, once you think you have at least some excess power effect.
From mizuno's data it seems that meshes provide varying power out, but each mesh is pretty stable. It is that stability that allows parameteric testing, something of great use in understanding what the results mean.
You will do better testing one mesh properly than experimenting with 10 different meshes (quite apart from the cost!)
THH

From mizuno's data it seems that meshes provide varying power out, but each mesh is pretty stable. It is that stability that allows parameteric testing, something of great use in understanding what the results mean.
As you point out here, even from the same person, conducting successful tests, the results are varied. I totally agree with what you are saying per se, but the emphasis must surely be on parametrically testing the various stages from material aquisition, through preparation, to running of the reactor, succesfully or not. This seems to boil down to consistency of known impurities, or physical morpholgy, which either help or hinder. That is the only thing that will kill the perenial ailment of this project, which is lack of consistency and, therefore, lack of reliable and consistent repeatability.
There seem to be good islands of knowledge that link the various types and levels of success, but it is the lack of predictive theory that functions in the practical area which means the only way this is going to get reliably sorted out, is to devise experiments which will compensate for the lack of effective overall understanding.
You don't need as many points at each power level as Mizuno provides in the R19 data. If he had done the same number of tests, but split equally between 50W, 100W, 150W, 200W input, the data would be much more interesting.
Again, in the microscopic sense of gaining maximum information from the data available, I totally agree. But there seems little doubt about the success of TM's runs, and the magnitude of the COP. I can't agree that high resolution analysis of the available data, or saying how it could have been better done, is going to tell us much more than we already know. It is how to get these results, and better,(or worse), and why, that is the focus of the task in hand. IMHO, of course.

Again, in the microscopic sense of gaining maximum information from the data available, I totally agree. But there seems little doubt about the success of TM's runs, and the magnitude of the COP. I can't agree that high resolution analysis of the available data, or saying how it could have been better done, is going to tell us much more than we already know. It is how to get these results, and better,(or worse), and why, that is the focus of the task in hand. IMHO, of course.
OK  so this is a big disagreement between us. If you want replication data to be generally accepted, having parametric dependence of excess heat on temperature in a way that confirms an activation energy mediated reaction would be a strong reason to take the work more seriously.
It really depends on whether you are doing it for yourself, or wanting also to interest others.
Of course, if R20 style results prove replicable none of this matters  and the world will change very shortly.
THH

Quite.

A detailed study of the high (3kW) power output Ni/Pd mesh structure and how to reproduce it effectively would now seem to be the way forward.
Unfortunately, that will be delayed. It turns out the mesh is still in the R20 cell, and the cell has broken and cannot be opened. It will cost $1,500 to open it and repair it without damaging the mesh. It may be a while before we come up with the money.
I reported earlier that the mesh was already replaced. That was a mistake. (Mizuno's mistake, not mine, for once.) He labeled a graph "R20 new mesh." He meant the latest cell, R21. R21 is very similar to R20.
It may also be delayed because we are looking for: A top notch lab, that will do a detailed analysis, with state of the art instruments, for free. That's a rare combination of attributes. Probably nonexistent. Seven_of_twenty has assured me that topnotch labs will line up to evaluate something like this. I wish he would produce a few of them.

Unfortunately, that will be delayed. It turns out the mesh is still in the R20 cell, and the cell has broken and cannot be opened. It will cost $1,500 to open it and repair it without damaging the mesh. It may be a while before we come up with the money.
I reported earlier that the mesh was already replaced. That was a mistake. (Mizuno's mistake, not mine, for once.) He labeled a graph "R20 new mesh." He meant the latest cell, R21. R21 is very similar to R20.
It may also be delayed because we are looking for: A top notch lab, that will do a detailed analysis, with state of the art instruments, for free. That's a rare combination of attributes. Probably nonexistent. Seven_of_twenty has assured me that topnotch labs will line up to evaluate something like this. I wish he would produce a few of them.
I know this will sound like a blinding flash of the obvious but couldn’t McKubre move some influences at SRI to do the much needed in depth analysis of that unique palladium rubbed mesh?