How Does Millsian's Model Claim to Instantly Simulate Complex Molecules?

[JustinLevy]

Does anyone know how BlacklightPower's latest non-sense is able to spit out answers that are close to correct? ( http://www.millsian.com/ )
They actually are willing to release a trial program, so I figured someone by now has dug through it and determined how it does it. Based on the shear number of tests molecules the company used, I'm wondering if it is just essentially a huge tight binding fit to many molecules.

They released a "paper" comparing their program to a quantum chemistry package (SPARTAN), in which they clearly compare the wrong values in order to claim spartan isn't even order of magnitude correct (I can run similar calculations on the PC GAMESS quantum chemistry package and see that the Millsian people must not have grabbed the correct value from the output or something ... while I can't prove they did it on purpose, I of course suspect it). Because they attack standard quantum chemistry packages, and are trying to make money on what appear to me to be a fraud (and the company is taking in millions from unsuspecting investors), I assumed someone would have investigated and reported this.

Suprisingly, some google searching didn't bring anything up.
Does anyone know any details?
How can they get away with this?

 

[alxm]

Does anyone know how BlacklightPower's latest non-sense is able to spit out answers that are close to correct?

I recall looking at their bogus claims. And these are no less bogus. The fact that they don't give details about their 'method' is reason for suspicion itself. Given that there are fundamental errors on the level of basic math in their 'alternative' to quantum theory (whatever it was they called it..), I'd be quite interested in seeing how they actually got these values, or any values, directly from their 'theory'. The few details that are given, are great cause for doubt. IIRC they claim it's based on the principle that the total heat of formation of a molecule can be written as a sum over heats of formations of the functional groups, which they say is justified by their 'alternative physics'. (Suspiciously enough: the mathematical details of that justification aren't given. So even if one took their theory at face value, it would be unproven)

More importantly, it's a well-established fact that heats of formations can not, consistently and accurately, be calculated as a sum-of-parts. If it were, it'd have been discovered empirically and justified theoretically ages ago, and would probably be taught in undergrad chemistry by now.

They actually are willing to release a trial program, so I figured someone by now has dug through it and determined how it does it. Based on the shear number of tests molecules the company used, I'm wondering if it is just essentially a huge tight binding fit to many molecules.

I considered looking at it, but they want registration and crap and I really wasn't going to jump through hoops to debunk something I already know is false. But yes, my intuition was also this: They've just done a fit over a set of systems large enough to give reasonable values for similar simple organic molecules.

They released a "paper" comparing their program to a quantum chemistry package (SPARTAN), in which they clearly compare the wrong values in order to claim spartan isn't even order of magnitude correct

They didn't even get the values right? That's worse than I thought. I recall seeing that and noted that they were comparing heats of formation to Hartree-Fock/6-31G. Essentially the worst method that still gives qualitatively correct results.

But heats of formation is only a single parameter. If you've seen what real benchmarks of QM methods do, they compare a lot more than just that. I'd like to know how their 'method' does at calculating bonding distances, bond angles, ionization potentials, electron affinities, vibrational spectra, etc! And of course, do so without using any experimental data at all. After all, they're the ones claiming to replace quantum theory.

Because they attack standard quantum chemistry packages, and are trying to make money on what appear to me to be a fraud (and the company is taking in millions from unsuspecting investors), I assumed someone would have investigated and reported this.

Well, they're deluded. I think Mills honestly believes the snake-oil he's selling. So it's hard to say if it really qualifies as 'fraud'. I doubt anyone would really buy their software. Anyone who knows enough to need such software should, IMHO, know enough to know they're full of s--t. Mostly its their investors getting defrauded. Small potatoes compared to what Bernie Madoff did though.

Blacklight got a fairly decent debunking in http://www.spectrum.ieee.org/jan09/7127" in IEEE Spectrum back in January.

 

[JustinLevy]

I recall seeing that and noted that they were comparing heats of formation to Hartree-Fock/6-31G. Essentially the worst method that still gives qualitatively correct results.

Sure, Hartree-Fock is not the most accurate for absolute values but it is amazing it is that close starting from first principles and simple approximations. And even then, it surely gives better results than what they claim.

Here's an example:
Dimethyl sulfoxide
they list the experimental value as 35.4348
they list 6-31G* has giving 1125.4243 for a 3068.3% error

That is just a blatant lie. No if ands or buts.
That is misrepresentation, to make their product look better, and to attact investors.
That is fraud.

Mostly its their investors getting defrauded. Small potatoes compared to what Bernie Madoff did though.

I heard it is on the order of tens of millions of dollars (I heard > $40,000,000). Yes, small potatoes compared to what Bernie Madoff did ... but still a sizable fraud.

I can understand getting away with fraud if you keep your "perpetual motion machines" away from the public at large to debunk, but how can you get away with it if you start releasing demos. It's like they are getting bolder and think they can get away with anything. I hope this is signing their death certificate as a "company".

Well, they're deluded. I think Mills honestly believes the snake-oil he's selling. So it's hard to say if it really qualifies as 'fraud'.

I get the impression he does as well. But based on the formulaic way they follow similar scams in the past, I get the impression several running the company must be aware. Regardless, his theories can't predict what they claim ... so the program must (while convoluted) boil down to look up table comparison. That would be lying about what they have done to their investors, lying about the results to gain investment, and therefore clearly fraud. I don't care what agreement people have to sign ... that doesn't protect the company from criminal liability. How can they be getting away with this!?

Do you know some good examples of molecules which have heats of formation very far from the sum of their parts. If so, then even without registering for the program ... it can be proved the program doesn't do what they claim by looking at their publically released results.

 

[alxm]

Do you know some good examples of molecules which have heats of formation very far from the sum of their parts. If so, then even without registering for the program ... it can be proved the program doesn't do what they claim by looking at their publically released results.

Well the textbook example would be something like benzene. If you were to calculate the energy of a single C-C bond and a C=C double bond through such a model (looking at ethane, ethylene, etc), you'll probably find it'll work as an okay-ish approximation for simple aliphatic hydrocarbons. But is benzene then 3 single + 3 double bonds? Not at all. Due to the conjugation/resonance of the pi bonds.
(after which the textbook goes on to show how even a fairly crude, approximate model built on quantum theory- Hückel theory - gets a quite good result)

While that alone is enough to disprove the principle, it probably wouldn't foil their program, which I'm guessing has a pre-programmed experimental value for benzene. But anything sufficiently far from whatever experimental values they used should cause problems. Organometallic compounds (d-orbitals! spin states!) could probably lead to some dismal failures.

 

[JustinLevy]

If someone wanted to report fraud, where would they even report it to?
I'm very tempted to send an email. Sure, others have most likely tried to report the company as well, but shear volume of complaints many be necessary at the end of the day. I'm willing to throw a drop in the bucket to help it overflow.


The fact that this "company" has survived this long is very troublesome.

 

[holversb]

JustinLevy posted:
---------
They released a "paper" comparing their program to a quantum chemistry package (SPARTAN), in which they clearly compare the wrong values in order to claim spartan isn't even order of magnitude correct

Dimethyl sulfoxide
they list the experimental value as 35.4348
they list 6-31G* has giving 1125.4243 for a 3068.3% error
That is just a blatant lie. No if ands or buts.
---------

BH: All values were taken from Spartan's precomputed database using 3-21G and 6-31G* basis sets at Hartree-Fock. I encourage you to double-check the results and post them on this forum.

Alxm posted:
----------
More importantly, it's a well-established fact that heats of formations can not, consistently and accurately, be calculated as a sum-of-parts.
----------

BH: On pages 58-85 of the below we present the results of our calculations based on linear combinations of bond energies corresponding to functional group composition. Agreement with experimental values is spectacular. This paper has been accepted for publication.

http://www.millsian.com/papers/modeling%20paper012709S.pdf

Alxm posted:
----------
I'd be quite interested in seeing how they actually got these values, or any values, directly from their 'theory'.

I'd like to know how their 'method' does at calculating bonding distances, bond angles, ionization potentials, electron affinities, vibrational spectra, etc!
---------

Millsian has posed a comprehensive discussion of each solved functional group (including energies, angles, distances) and how they are linearly combined to compose molecules, in the Volume 2 of Mills' book, available for free download at:

http://www.millsian.com/book.shtml

You can also find a complete collection of excel spreadsheets used in the calculations of energies, bond angles and bond distances, which are used as the basis for the molecular modeling software, for free download at:

http://www.millsian.com/spreadsheets.shtml

best,

Brett Holverstott
Millsian, Inc.

 

[JustinLevy]

In the future, please use the quote feature. I find it makes the posts easier to read.

BH: All values were taken from Spartan's precomputed database using 3-21G and 6-31G* basis sets at Hartree-Fock. I encourage you to double-check the results and post them on this forum.

I can't believe I'm wasting my time on this, but sure, I'll double check the RHF calculations.
I'll use a simplier molecule to speed up things.

From your company's paper (all energies in eV):

Boron trichloride, bonding energy.
Experimental value 13.8075
RHF 3-21G (using experiment for elemental binding) 324.7150 -> percent error 2251.7%
RHF 3-21G (using RHF 3-21G for elemental binding) 680.2961 -> percent error 4827.0%


Now, doing the RHF 3-21G calculations to check, and I get (in Hartrees):
Cl -457.2765517601
B -24.3897617087
BCl3 -1396.5331129536
(I just copy pasted these from a computational program output, and obviouisly the calculation is not accurate to that many decimal places as the last few digits will change between different round off errors, implementations of the RHF algorithm, etc. But these ARE the RHF 3-21G results.)

Binding energy = -1396.5331129536 - (-24.3897617087) - 3*(-457.2765517601) = -0.3136959646

In eV this is 8.54 eV
This is the correct order of magnitude. Which for such calculations, is all one can expect from RHF which is an approximation of the true quantum answer (it isn't even the full non-relativistic answer. And as an aside full quantum calculations give answers more precise for hydrogen than even Millsian's claims at a fundamental theory do ... their theory, even with their bogus calculations of it, are already ruled out by experiment). Furthermore, 3-21G is a very small basis. Order of magnitude is all one should expect from such a calculation, despite what Millsian implies.


What did Millsian claim? Yep, they claimed the RHF 3-21G value of 680.2961 eV for a 4827.0% error.

THIS IS A BOLD FACE LIE.
The very fact that you noticed that the answer are orders of magnitude off, should have made you realize you should double check how you are applying some of the values from table, etc. or double check them yourself. Look, it took me only ~ 15 min to do those calculations and write this up. Itstead you blatantly misrepresent the RHF energy results.

That you continue representing this instead of double checking when it was pointed out, is not a good sign for your company's intentions.

BH: On pages 58-85 of the below we present the results of our calculations based on linear combinations of bond energies corresponding to functional group composition. Agreement with experimental values is spectacular. This paper has been accepted for publication.

http://www.millsian.com/papers/modeling%20paper012709S.pdf

The "theory" behind this is non-sense.
Your "solutions" aren't even solutions to your own theory!

Brett Holverstott
Millsian, Inc.

You should be sick of yourself. Preying on people that don't know enough science or math to know any better. Collecting over $50 million in funds for a project based on objectively false statements.

-------------

I don't understand how such companies can continue existing. Does anyone know where to report such businesses? With enough reports, maybe state or federal agencies will finally react.

 

[zincshow]

Boron trichloride, bonding energy.
Experimental value 13.8075
RHF 3-21G (using experiment for elemental binding) 324.7150 -> percent error 2251.7%
RHF 3-21G (using RHF 3-21G for elemental binding) 680.2961 -> percent error 4827.0%

I would appreciate some clarification. It appears you have no problem with millsian calculation of the energy, but you are concerned with millsian representation of the spartan values, which you feel are being misrepresentated? I have used the spartan program but did not see the spreadsheet. Is it in the program or you had to bring up the molecule and recalculate it? Where did the 324 and 680 numbers come from?

 

[alxm]

BH: On pages 58-85 of the below we present the results of our calculations based on linear combinations of bond energies corresponding to functional group composition. Agreement with experimental values is spectacular. This paper has been accepted for publication.

The paper, which contains numerous mathematical errors (which have been debunked elsewhere) and misstatements of fact (such as QM methods being fitted to empirical values, something which is entirely false for the vast majority of QM methods), does not provide any justification for why the heat-of-formation of a molecule can be divided into a sum-of-parts. There exists no such justification.

Postulate the following:
- Chemical binding energies are due to differences in energy, mostly kinetic, of the electrons of the constituent atoms.
- Electrons repel each other.
- Electrons obey Newton's laws of motion. (True of both quantum and classical theory)
Given the above, no other knowledge is needed to discern that the kinetic energy of electrons is a physical many-body problem. It must, ultimately, be described by some non-linear differential equation. It follows then, purely mathematically (and is easily subject to hard, mathematical proof) that the overall kinetic energy of the system cannot be expressed as a simple sum-of-parts. The parts are interdependent.

You claim that this is true for 'functional groups'. Where and how do you physically define a 'functional group'? Quantum theory of chemistry does not need 'functional groups'. If you're using the ordinary chemistry definition of 'functional group' then your reasoning is circular. A "functional group" in chemistry is a part of a molecule that is reasonably independent of the rest of the molecule. Saying that a molecule can be approximated by the parts of the molecule that are already known to be somewhat independent of the rest of the molecule is saying nothing new at all.

So you're agreement with experiment is hardly 'spectacular'. You're using a set of experimental values (that you chose) to 'predict' a set of experimental values (that you chose). That's not very impressive. I wouldn't even say it's science.

 

[holversb]

Hi Justin Levy,

I double-checked the Spartan values and calculations for dimethyl sulfoxide and boron trichloride. Results are as follows:

Dimethyl sulfoxide is correct. Values obtained from Spartan 2004 were verified, giving 3-21G and 6-31G* values of -548.846594 au and -551.537050 au, respectively. The calculation for 6-31G* using experimental IP's is as follows:

-551.53 * 27.212 au/eV - (10859 eV + 2043 eV + 2*1030 eV + 6*13.59 eV) = 1125 eV.
(1125 eV - 35.4348 eV) / 35.4348 eV = 3076% relative error.

The relative errors are: 210.9% (3-21G theory vs. exp IPs), 27% (3-21G vs. 3-21G IPs), 3068% (6-31G* theory vs exp IPs), and 2901% (6-31G* theory vs 6-31G* IP's). We were also initially impressed by the magnitude of the errors, even when theoretical IP's were used. The relative error in the Millsian calculation (using exp IP's) is 0.25%.

I did find an error in the calculation of the theoretical 3-21G IP's for boron trichloride (an error which effects some of the boron molecules). This changes the relative error of 3827% to 20.3% for 3-21G vs 3-21G IP's. My new calculation gives 2252.9% (3-21G vs exp IPs) 20.3% (3-21G vs 3-21G IPs), 996.4% (6-31G* vs exp IPs) and 29.6% (6-31G* vs 6-31G* IPs). The relative error in the Millsian calculation (using exp IP's) is 0.35%.

As you can see, the magnitude of the relative error of quantum calculations is not a reliable indicator of human error in reporting. While I regret that there is calculation error in the boron molecules, the reality is that it is difficult to avoid some errors in a spreadsheet this large.

Millsian has proposed a fundamental shift in the way atoms and molecules are solved, and has rigorously tested its theoretical predictions of bond energies, distances, angles, and (recently) dipoles against 800 molecules for which experimental data is available. Thousands of predicted values match experiment to a degree of accuracy, and with a simplicity of calculation, unmatched by other methods. We will continue to push forward our technology and its applications for a variety of industries.

Millsian is dedicated to honesty and transparency in reporting. We have an extensive collection of online documents detailing our theoretical calculations in both written and spreadsheet form, see my previous post. I will also see to it that the spreadsheet used in our recent paper appears on the spreadsheets page shortly. I encourage you to investigate this material if you have further questions.

best,

Brett Holverstott
Millsian, Inc.

 

[JustinLevy]

I did find an error in the calculation of the theoretical 3-21G IP's for boron trichloride (an error which effects some of the boron molecules).

Thank you for admitting the error.

While I regret that there is calculation error in the boron molecules, the reality is that it is difficult to avoid some errors in a spreadsheet this large.

I skimmed the table, and randomly picked a simple molecule with large error. It turned out you had made a mistake. If there are only a few mistakes in the entire spreadsheet, I must be one lucky lucky guy. More likely is that there are many errors.

Millsian is dedicated to honesty and transparency in reporting.
...
I encourage you to investigate this material if you have further questions.

Look, let me put this as simply as possible.

1] It is not clear that you are dedicated to honesty and transparency
Many of your statements about quantum mechanics are objectively wrong (as alxm here, and many others elsewhere have pointed out). Furthermore, even if we accept the postulates of your company's "grand unified theory of classical physics", your derivations do not follow. Your math is incorrect.

Furthermore, in the specific case of this publication, you seem to employ purposeful misleading in your conclusions. You choose to compare to a heavily simplified calculations of the full non-relativistic theory, and claim victory. Instead, why not compare your "exact" theory to full calculations of what physicists currently consider the closest we have to an "exact" theory of electrodynamics.

Scatterring experiments support quantum mechanics, and not your "the electron is a disk, or orbitsphere" ideas. And your theory, despite claiming to be exact, does not match experiment perfectly, while QED fits completely to experimental limits. Which brings me to my next point

2] Your theory is already ruled out by experiment.


So my questions to you are:
A] Do you personally believe GUT-CP is an exact theory as your company claims, despite not agreeing with experimental results?

B] Every scientist needs to consider the possibility that previous hypotheses could be incorrect. Since there are already numerous complaints about mathematical errors in the theory, and even disagreements with experiment if we just ignore that and take your calculations results verbatum... What more would you have to see before you personally would come to understand GUT-CP is wrong?

The fact that you don't already see GUT-CP is wrong is troubling, but if there is actually nothing that could change your mind, then you have no right to claim you are pursuing science.

 

[alxm]

Mr Holverstott, the saying goes "extraordinary claims require extraordinary evidence". You're not providing any evidence at all, since you haven't explained how you arrived at any of the numbers you claim.

A hand-waving referral to 'papers' that are both logically incoherent and mathematically wrong, doesn't cut it. It's not my job to convince you that you're wrong. You better convince me why you're right.

How about this: Provide me/us with a real practical example right here. Demonstrate how you, from first principles (being whatever you think your theory is), calculate something simple. Omitting no steps, and using no experimental evidence other than fundamental constants (electron mass, Planck's constant, fundamental unit of charge).

Let's make it the ground-state energy of singlet helium - the simplest polyelectronic atom. Show me how you arrive at this result (in exact, closed form, as per your own claims). Omitting no steps. Then repeat the procedure from the same initial model to get the ground-state energy of parahelium.

Show me your work. Show me how to get these two simplest-case values. You could have a million results that 'match experiment', but that doesn't mean s--t if you don't show me how you arrived at them.

Finally, let's be clear about this: Quantum mechanics is exact, insofar atoms and molecules are concerned. QM calculations done at the highest level of theory are consistently within experimental error. As Dirac said as early as 1929: "The underlying physical laws necessary for the mathematical theory of ... the whole of chemistry are thus completely known". 80 years of calculations with this theory have confirmed that opinion. You simply can't improve on quantum mechanics insofar atoms and molecules are concerned, because there is nothing to improve on. There are no known deviations from theory that are not within the limit of calculation errors. The fact that you chose to compare yourself to Hartree-Fock, the crudest approximate method within quantum chemistry, rather than methods known to be more exact, such as CI, suggests you're either dishonest, or simply ignorant of the subject. There is certainly a lot of ignorance of the subject obvious in those 'papers' of yours. You simply haven't even bothered to understand the theory you're claiming to displace.

 

[holversb]

Let's make it the ground-state energy of singlet helium - the simplest polyelectronic atom.

This topic is treated in the paper:

Exact Classical Quantum Mechanical Solutions for Atomic Helium Which Predicts Conjugate Parameters from a Unique Solution for the First Time - R.L. Mills, Physics Essays, Vol. 21(2), (2008), 103–141.

and also in Chapter 7 ("Two-electron Atoms") of Mills' book, available for free download online at: http://www.millsian.com/bookdownload.shtml

Scatterring experiments support quantum mechanics,

This topic is treated in Chapter 8 ("Classical Photon and electron Scattering") of Mills' book, available for free download online at: http://www.millsian.com/bookdownload.shtml

And also in the introduction to:

Maxwell's Equations and QED: Which is Fact and Which is Fiction, R.L. Mills, Physics Essays, Vol. 19, No. 2, (2006), 225-262.

I encourage you to take advantage of these resources, and if you have further questions, please post them on the Society for Classical Physics forum:
http://tech.groups.yahoo.com/group/SocietyforClassicalPhysics/

best,

Brett Holverstott
Millsian, Inc.

 

[JustinLevy]

I encourage you to take advantage of these resources, and if you have further questions, please post them on the Society for Classical Physics forum:
http://tech.groups.yahoo.com/group/SocietyforClassicalPhysics/

We have already read enough to see that your theory is flawed mathematically, and even if we ignore that, the theory is proven wrong experimentally.

As Alxm and I mentioned earlier, quantum mechanics is exact to the best ability of measurement while the theory Mills is promoting is already outside of experimental bounds. There is no debate on this. So the only remaining question is whether you admit this opennly, and whether you consider the theory wrong or just an "approximation" of something better (unlike the claims the theory is exact by Mills).

You signed up to this board to respond to comment about Mill's theory. You have already admitted one mistake, but most questions you avoid and instead just keep referring to more papers with more flaws.
If you won't answer our questions here, why would we go to a board where our questions will be lost amongst the yelling from hundreds of crackpots?

In this case, your silence speaks volumes.
When your company is taken down with fraud charges, I hope you are included on the list of defendants... for it is clear you have no interest in pursuit of truth, but just money from those that don't know any better. I implore you to do some soul searching. And even if you still believe the theory, at least consider the possibility that you are wrong so we can discuss it opennly here so you can learn.

 

[alxm]

This topic is treated in the paper:

Exact Classical Quantum Mechanical Solutions for Atomic Helium Which Predicts Conjugate Parameters from a Unique Solution for the First Time - R.L. Mills, Physics Essays, Vol. 21(2), (2008), 103–141.

And that's not what I asked for. I've already explained that Mills' book is incoherent, as far as I can tell, incorrect on many simple matters of fact, mathematically wrong and logically inconsistent. And even looking at you reference I see numbers without proper explanations for how they were derived.

What I asked was for you to provide this simple example right here and now. You're the one who purports to understand this insanity. Although I can't say I see much evidence of that - given that the few 'answers' you do give amount to 'see the book'.

The fact that you've chosen to silently ignore most criticisms speaks volumes though.

 

[Eli Rabett]

One correction, http://en.wikipedia.org/wiki/Benson_group_increment_theory" has been a long used method for calculating heats of formation of molecules. It has long been my assumption that Millsian is such a thing, coupled with a molecular mechanics program based on some sort of classical Hamiltonian for calculating bond lengths and angles, but I like the idea of it being a large look up table.:rofl:

 

[alxm]

One correction, http://en.wikipedia.org/wiki/Benson_group_increment_theory" has been a long used method for calculating heats of formation of molecules. It has long been my assumption that Millsian is such a thing, coupled with a molecular mechanics program based on some sort of classical Hamiltonian for calculating bond lengths and angles, but I like the idea of it being a large look up table.:rofl:

Well, yes perhaps I overstated the case somewhat. But let's get this clear: group additivity is an empirical tool, not an actual theory of chemistry. It doesn't in itself contain any insights or assumptions of how molecules are formed, etc. Theoretically it's unjustified - and for reasons explained unjustifiable, as it's fundamentally a non-linear property. At best it's an approximation, an approximation that often works, but for reasons that are again, strictly empirical (but subtly so). The energy is divided into parts - on what grounds? Well, on purely empirical ones - we divide the energy into such-and-such components, because that's what works best, purely empirically. There's no underlying theory that says why this is the case. So you have no way of predicting when the approximation is vaild or not.

I have no problem with approximations or semi-empiricism. But once the amount of empirical information you're using is on the same scale as your 'predictive' abilities, then you're not really doing science anymore. You're just repeating existing knowledge.

The big issue here is that Mills & Co aren't claiming to have created an empirical or semi-empirical tool. They're claiming to have disproved Quantum theory and suggest replacing it with their own new fundamental 'theory' of matter. And they're claiming their results are exact, although I see no reason why 'group addition' is any more theoretically justified in their theory. At least not if I take their word at the fact that it obeys classical physics. (Mills has been floating his crackpot ideas since 1990 at least, and they've been repeatedly refuted as being mathematically incorrect, and in violation of both classical physics, quantum results, and empirical fact ever since)

I kind of doubt "Millsian" is a simple look-up table. It probably has some heuristic. Maybe it's even a good heuristic. It seems they're not interested in telling. But regardless, it doesn't mean anything if it's purely empirical - which all evidence suggests it is. I've seen no justification for it from 'first principles' (either theirs or those of real physics). So it could give perfect results and it still wouldn't validate Mills' crazy theories.

Likewise, comparing their (empirical) results to purely theoretical quantum ones, that are known to be approximate does nothing to invalidate quantum theory, no matter how lousy the approximation is. Because it's a theory. I can explain why Hartree-Fock is not exact. I can predict when its results will be more and less accurate. I can put bounds on the size of the error.

That's something very different from an empirical theory.

 

[mkw]

http://millsian.com/papers/MillsianImp081809WebS.pdf

 

[kmarinas86]

IIRC they claim it's based on the principle that the total heat of formation of a molecule can be written as a sum over heats of formations of the functional groups

Are you saying that their claim is that each functional group has a heat formation? I don't believe they said that.

(Suspiciously enough: the mathematical details of that justification aren't given. So even if one took their theory at face value, it would be unproven)

Justification is not always purely mathematical. This is especially true in the case of starting from so-called "first principles". One can apply them to see if they impose severe restrictions on what is physical possible. The path of deductive reasoning should ideally be close 100% mathematical. An example of that ideal would be Einstein's approach to predicting special relativity. The discovery of the periodic table by Mendeleev does not seem to have been predicted in such a left-brain dominant manner.

More importantly, it's a well-established fact that heats of formations can not, consistently and accurately, be calculated as a sum-of-parts. If it were, it'd have been discovered empirically and justified theoretically ages ago, and would probably be taught in undergrad chemistry by now.

Do you mean that Hess's law is wrong? If not, then where is the origin of this inaccuracy? Is it because the parts themselves cannot be measured accurately? Does it involve summing the wrong parts or not accounting for their changes during the reaction such as the loss of an electron?

I considered looking at it, but they want registration and crap and I really wasn't going to jump through hoops to debunk something I already know is false. But yes, my intuition was also this: They've just done a fit over a set of systems large enough to give reasonable values for similar simple organic molecules.

That doesn't count as an achievement?

They didn't even get the values right? That's worse than I thought. I recall seeing that and noted that they were comparing heats of formation to Hartree-Fock/6-31G. Essentially the worst method that still gives qualitatively correct results.

Well if Hartree-Fock/6-31G is the worst qualitatively correct method of obtaining those results, it still works, right?

But heats of formation is only a single parameter. If you've seen what real benchmarks of QM methods do, they compare a lot more than just that. I'd like to know how their 'method' does at calculating bonding distances, bond angles, ionization potentials, electron affinities, vibrational spectra, etc! And of course, do so without using any experimental data at all. After all, they're the ones claiming to replace quantum theory.

You said you didn't want to jump through the hoops. It's a bit like saying you want a clean house yet don't want to do any housework. That's strange approach you have there. Why do you believe that you like to know all of that when you know you don't have the time for it?

Well, they're deluded. I think Mills honestly believes the snake-oil he's selling. So it's hard to say if it really qualifies as 'fraud'.

Just qualify what you mean by "it", and then it is easy, not hard, to qualify "it" as a fraud. Do you mean the company? Do you mean the theory? Do you mean their promises? Do you mean their collaborations? Do you mean their licensing strategy? What say?

I doubt anyone would really buy their software. Anyone who knows enough to need such software should, IMHO, know enough to know they're full of s--t. Mostly its their investors getting defrauded. Small potatoes compared to what Bernie Madoff did though.

Most of their revenue does not come from their software. They are primarily a licensing company.

Blacklight got a fairly decent debunking in http://www.spectrum.ieee.org/jan09/7127" in IEEE Spectrum back in January.

Only fairly decent? Why is it not excellent?

 

[kmarinas86]

Well, yes perhaps I overstated the case somewhat. But let's get this clear: group additivity is an empirical tool, not an actual theory of chemistry. It doesn't in itself contain any insights or assumptions of how molecules are formed, etc. Theoretically it's unjustified - and for reasons explained unjustifiable, as it's fundamentally a non-linear property.

The supposedly "fundamental" non-linearity may be due to how you calculate the heat of formation in the first place. For example, do two hydrogen atoms automatically have the same heat of formation? What if one hydrogen atom has an extra electron, and the other does not? I would say the heat of formation is different between the two atoms because the adding of an electron to one of them implies a change of energy state.

At best it's an approximation, an approximation that often works, but for reasons that are again, strictly empirical (but subtly so). The energy is divided into parts - on what grounds? Well, on purely empirical ones - we divide the energy into such-and-such components, because that's what works best, purely empirically. There's no underlying theory that says why this is the case. So you have no way of predicting when the approximation is vaild or not.

What should prevent a scientist from coming with a more fundamental explanation for the heat of formation of these components?

I have no problem with approximations or semi-empiricism. But once the amount of empirical information you're using is on the same scale as your 'predictive' abilities, then you're not really doing science anymore. You're just repeating existing knowledge.

Predictive abilities may lag or lead experimental evidence. This may affect the credibility of scientific theories and their support within the scientific community, but certainly not their validity. One famous example was the observation which vindicated Einstein's work of General Relativity. It was thanks to the foiling of attempt to observe a solar eclipse in 1914 that gave Einstein the chance to correct his theory. By the eclipse in 1919, he had this corrected his theory. Good timing made him a superstar. To have been late in correcting this would have a blow to his credibility and his subsequent fame. My standing on this issue is that the ordering of events affects the outcome of a scientist's career, which implies that the cause of any increase of one's reputation in the scientific community (and in the public in general) has less to do with objectivity and more to do with social psychology.

The big issue here is that Mills & Co aren't claiming to have created an empirical or semi-empirical tool. They're claiming to have disproved Quantum theory and suggest replacing it with their own new fundamental 'theory' of matter. And they're claiming their results are exact

It might have been better to qualify by what they meant by "exact".

although I see no reason why 'group addition' is any more theoretically justified in their theory. At least not if I take their word at the fact that it obeys classical physics. (Mills has been floating his crackpot ideas since 1990 at least, and they've been repeatedly refuted as being mathematically incorrect, and in violation of both classical physics, quantum results, and empirical fact ever since)

In your opinion, what is the least it would take to see such a reason?

I kind of doubt "Millsian" is a simple look-up table. It probably has some heuristic. Maybe it's even a good heuristic.

You did not see any heuristic?

It seems they're not interested in telling.

It appears that they are. Instead saying they have heuristics, they say they have a "theory". The problem is that these heuristics are not parsed and given special names. So the pattern of thought in the book is hard to observe. I have looked at their book for hours over the past year. I will tell you upfront that there is no way you can find these "heuristics" in a day or two. This is especially true if you do not account for the various changes that have been made to their book over many the months or years that critics you have read from have known about this.

To me, there is little difference between a theory and a heuristic. A theory is supposed to have been validated. A heuristic is a fancy way to say "folksy assumption". To be sure, many have attempted to come up with formal definitions of the words "theory" and "heuristic" which do little to make the two seem alike (which they are).

But regardless, it doesn't mean anything if it's purely empirical - which all evidence suggests it is.

Define "all evidence".

I've seen no justification for it from 'first principles' (either theirs or those of real physics). So it could give perfect results and it still wouldn't validate Mills' crazy theories.

If perfect results were obtained with a crazy theory, then that would be a strange quandary. Ask yourself, why would one have better results with nonsense?

Likewise, comparing their (empirical) results to purely theoretical quantum ones, that are known to be approximate does nothing to invalidate quantum theory, no matter how lousy the approximation is. Because it's a theory.

Invalidation has do to with assumptions involved. You could make a theory that makes thousands of horrible predictions, but then claim that you can explain the errors away. You generate heuristics to cover up the flaws. I believe that Occam's razor is a better idea. However, you say:

I can explain why Hartree-Fock is not exact. I can predict when its results will be more and less accurate. I can put bounds on the size of the error.

What are your heuristics?

That's something very different from an empirical theory.

Does your approach make use of Hartree-Fock/6-31G more scientific? Explain.

 

[alxm]

Are you saying that their claim is that each functional group has a heat formation? I don't believe they said that.

I believe so. Specifically their http://www.millsian.com/theory.shtml" says:

Further, electrons are localized in molecules to specific regions, i.e. functional groups which act as building blocks, or independent units, in larger structures. Using two basic equations, Millsian has solved the important functional groups of chemistry, allowing molecules of arbitrary size and complexity to be modeled trivially and almost instantly on a personal computer.

To me, this seems to imply that they're doing exactly that - summing up over 'functional groups'. Which would in fact be valid if they were truly independent, but they're not. They didn't seem to dispute this earlier in this (quite old, btw) thread.

Justification is not always purely mathematical.

Nevertheless, they claim to have one. Only it does not justify this, and as I explained in as simple terms as I could, it can't be justified mathematically even if only the most basic assumptions about how electrons act are true.

Do you mean that Hess's law is wrong?

No I don't, and nothing I said contradicts it.

That doesn't count as an achievement?

Scientifically, no. Because it's only telling you something you already know.

Well if Hartree-Fock/6-31G is the worst qualitatively correct method of obtaining those results, it still works, right?

Not for predicting anywhere near correct values for heat of formation.

If your intention is to prove that quantum mechanics is wrong (which they claim), one should be comparing to the best values calculated. Not from the crudest approximation in use.

You said you didn't want to jump through the hoops. It's a bit like saying you want a clean house yet don't want to do any housework.

No it's not, because I'm not particularly interested in their claims. It's not my job to go around and convince delusional people that they're delusional.

Just qualify what you mean by "it", and then it is easy, not hard, to qualify "it" as a fraud. Do you mean the company? Do you mean the theory? Do you mean their promises? Do you mean their collaborations? Do you mean their licensing strategy? What say?

All the above.