Calculation of localization energies L(+)

Hi,

I'm having a lot of trouble coming up with an aromatic nitration definition that matches the facts on the ground.

Both bromobenzene and biphenyl have a strong preference to nitrate in the para position.  Bromobenzene has these energyE() values 2.5061;2.5366;2.5104 for the ortho, meta, and para C atoms, and biphenyl has 2.4;2.5441;2.4471 for the same.  Based only on these values, one would expect biphenyl to nitrate in the ortho position, but it doesn't because of steric encumbrance.

To try to consider the effects of sterics, I use the rules below.  The first finds all aryl atoms within a certain energy of the lowest.  The second chooses the least sterically hindered of these.

REACTIVITY: energyE(ratom(1)) - min(energyE(reactant(0), filter(reactant(0), "aromaticAtom()"))) <= 0.0475

SELECTIVITY: -stericHindrance(ratom(1))

Unfortunately, because the energyE() of the meta C atom in bromobenzene is closer to that of the para C atom than the para C atom in biphenyl is to the ortho C atom, no matter what value I set in the reactivity rule, I get the wrong answer for one or the other of the two.

I think that what I really need is a way of combining the stericHindrance() and energyE() values in a single expression.  Any ideas?  Does the stericHindrance() value accurately reflect the effect of the Ph substituent on its neighbors?  Please share your thoughts.

I should note that these substrates are not particularly exotic.  If I can't write a uniform equation that takes care of these cases, I begin to lose hope that I can write anything general at all.

Yes. The localization energy is just an electronic factor, so steric issues should be considered as well. The stericHindrance function calculates hindrance from the lowest energy conformer, so its value ddepends of the geometry of that conformer. The stericEffectIndex function is a fast topological alternative, independent from conformational variance, but too raw to be applicable for this calculation. I plan to have a specific calculation that combines this quick topological index with some connectivity information.

You can combine electronic a steric calculations for example by separating them like this:
SELECTIVITY: energyE(...); stericHindrance(...)
TOLERANCE: ##; ##

First the localization energy will be calculated with its given tolerance. If there are still more than one potential sites then the second function, steric hindrance will be calculated for these remaining sites with its own tolerance value.

I'm still having problems that seem to be more related to the calculations than the screens.  For example, anisole and 4-methylanisole:

In anisole, the energy of the para and ortho C atoms are lower than that of the meta C atom.  Good enough.  But in 4-methylanisole, why is the energy of the C meta to MeO significantly lower than that of the C ortho to MeO?  This result contradicts everything I know about aromatic chemistry.  The Me should not have nearly as large an effect as the MeO.

-- Bob

Hi, it's now two months ago, I have shown that there are severe failures in the huckel localization energy L(+) calculation, please see:

http://www.chemaxon.com/forum/ftopic3612.html />
and there the post Sat Feb 07, 2009 11:46 am. There is a detailed evaluation for many examples.

I will attach here the L(+) results for phenol (the results for anisole are the same), and for the methyl substituted systems. As seen for toluene, the methyl group leads to rather low L(+) values for ortho and para, the L(+) value for benzene is 2.54. And how to understand the low value at the Me-group itself ?

The Marvin L(+) energies for phenol (anisole) are much too low. The correct values, using the same heteroparameters for phenol as Marvin are shown in blue. The blue values show the difference in HMO energy between the total pi-system minus localized pi-system, using the same parameters as Marvin for phenol (See the detailed discussion in the other topic).

And in ftopic3612 you can see also the result, benzaldehyde has lower Marvin L(+) values as phenol. This is not acceptable.

Regards, Hans-Ulrich

The same problem occurs for N-acylanilines.

[bob@epoch questionData]$ evaluate -e "energyE()" "CC(=O)Nc1ccccc1" -p 4

;0.8237;4.1209;3.0964;1.8291;1.1994;1.4244;1.2534;1.4244;1.1994

;0.7302;4.1083;3.0893;1.5947;1.1987;1.1488;1.4836;1.8302;1.1488;1.1987

But not for the parent anilines.

[bob@epoch questionData]$ evaluate -e "energyE()" "Nc1ccccc1" -p 4

1.5495;3.0854;1.9785;2.5541;2.0602;2.5541;1.9785

[bob@epoch questionData]$ evaluate -e "energyE()" "Nc1ccc(C)cc1" -p 4

1.5158;2.8244;1.9489;2.2509;2.2638;1.8037;2.2509;1.9489

Hi,

Calculated result depends on the model's parameters.

The actual model does not  know everything about aromatic chemistry...

Now I decreased the hyperconjugation effect of the "Me" group.

New results are on the attached figure.

In the future I think it would be better to create

a "teachable" reaction calculator for creating model's parameters automatically.

Jozsi

I think that's a great idea.  It's quite awkward for us to have to download a new version of JChem every time we find that a parameter needs to be adjusted, and it's only possible once a year for us to update the version that the commercial version of ACE Organic uses.  It's also not the best use of your time to be constantly fiddling with parameters.  

You could either have a teachable calculator, or you could simply make the parameters editable by the owner of JChem.  Or you could put them in a file, and then we could wait for experts like Hans to modify them, and then the rest of us could download the improved versions.  

Hi,

Yes. I wanted to say exactly this.

Jozsi

So, will you discuss this idea internally within ChemAxon and let us know whether you plan to implement it and how soon?

Hi,

Yes. We are going to discuss this in Chemaxon and inform you.

Jozsi

When will these new parameters be available?

Hi, I have many comments here. Let me begin with the first quote:

"The actual model does not  know everything about aromatic chemistry..."

What does that mean? "Aromaticity" is the result of HMO and not the input. And the simple Hueckel Molecular Orbital model does not use this virtual category which chemists have invented to get "containers" to put some molecules in.

As already mentioned in the HMO-topic http://www.chemaxon.com/forum/ftopic3612.html HMO-calculations have

only three things as input : 1) Topology of the pi-system     2) Some hetero parameters    3) The total number of pi-electrons.

Look in the HMO-matrix: There are many many zeros and at the diagonal some "alpha"-values for the hetero-atoms and some non-diagonal "beta"-values for the bonds. That's all.

By diagonalization you get the eigenvalues and eigenvectors. And there comes the chemical interpretation of these numbers, in this topic the so called "energy": Multiplying the eigenvalues with the number of electrons in these "Molcular Orbitals" and summing the results. That's the "HMO-pi-energy."

In the next post there comes the "localization energy L(+)".

Regards, Hans-Ulrich

Hi, now the "Localization energy L(+)".

On page 335 of Streitwieser's book there is Wheland's definition: " ... a cation localization energy is defined as the pi-bonding energy required to isolate two electrons at the position r from the remainder of the pi-network. Such isolation results in a new pi-system of>Using the internet program SHMO I have done these calculations for phenol (anisol will give the same results), please see the attached image.

As can be seen the correct L(+) energies for phenol, given in blue, show lower values in ortho and para position as compared with benzene ( L(+) = 2.54). The chemical interpretation is: Electrophilic substitution for phenol is easier at ortho and para position than for benzene. OK.

The Marvin calculation results, given in red, are wrong. And my evaluation showed: Marvin uses different hetero parameters for phenol h(O)=2.0 and the localised cationic pi-systems h(O)=2.5. That's wrong. And it leads to the very low L(+) values given in red.

Comparing the phenol/anisole L(+) energies with the aniline L(+) values, given seven posts before, you will see, the L(+) energies for aniline are much higher than for phenol. This is not acceptable.

The L(+) values for aniline are for ortho 1.98 and for para 2.06. That's lower as for benzene 2.54 and leads to the correlation: It easier to subtitute aniline derivatives in ortho and para the benzene. Here Marvin uses the same parameters for aniline and the localized cations.

But there is another problem: I have shown, Marvin uses for amino groups the same parameters as for aza-pi-centers (like pyridine). That's also wrong. Please see the topic http://www.chemaxon.com/forum/ftopic3612.html.

Please correct these failures before Marvin 5.3, as you have written in an answer.

Regards, Hans-Ulrich

Hi, and now:

"Hyperconjugation".

In the other topic, I have discussed the problems of using hyperconjugation models in the very very simple HMO model.

There was the example, Marvin includes the middle atom of triphenyl-methane as an atom with a pi-orbital = hyperconjugation. ???

And as can be seen in the image attached here, the Marvin hyperconjugation parameters for methyl, ethyl, isopropyl and t-butyl groups are the same. ???

My suggestions are:

1) Don't use the hyperconjugation model in the Marvin Hueckel calculations at all. Please look in Rauk's book. The  internet SHMO has no parameters for hyperconjugation. To use hyperconjugation in Hueckel calcultions may be appropriate in very special cases comparing seperated special systems. But a general use for all molecules and ions is too problematic to be used at all.

2) Don't use special paramaters for special categories. In very early posts I have shown, Hueckel calculations are not restricted to "aromatic" systems. The Simple Hueckel MO model is designed to planar pi-systems, and nothing more.

.

These failures are a problem for "teaching", but these are also wrong for using the Marvin and JChem programs to analyse "many thousand compounds" (as mentioned in other topics).

Regards, Hans-Ulrich

I have deleted this post from this topic. It was only a problem with the post editor. Hans-Ulrich

Jozsi will answer the issues next week when he will be back in the office. Thank you for your patience.

(Thank you for reporting the forum editor problem, I think, that it has been fixed.)

Hi,

When will these new parameters be available?

The next patch release will be availble in middle of May.

Jozsi

Hi Jozsi,

it is not only a parameter problem, it is a problem of the algorithm to calculate the electrophilic localization energy L(+).

And the algorithm which produces the red L(+) values shown above must be in many cases wrong, f.e. acetanilid, phenol, anisol.

How to explain the minimum value L(+) = 0.82 for the C-atom of the carbonyl-group of acetanilid ? It is unacceptable to have this low localisation energy as a model for the transition state of the electrophilic attack at a positively charge center.

Comparing the low values in the ring with the values for aniline itself leads to the consequence: Acetanilid is easier attacked by electrophiles than aniline. ???

And: Why is the methyl group in the acetanilid not included, but in the methyl substituted "aromats" it is included in the total pi-system? The (dubious) hyperconjugation should be more effective at the carbonyl group than at "aromats". But to say it once more: Omit the hyperconjugation at all: See the sequence methyl-ethyl-isopropyl-tertbutyl shown above.

In the next post there is a detailed analysis of the localisation energy calculation for acetanilid.

Regards, Hans-Ulrich

Hi,

Yes, you are right.  I fixed one more  bug. L(+) is now 4.4.

Jozsi

Now the detailed analysis for acetanilide.

In the upper image there are the total L(+)-results for acetanilide. In red the output of MarvinSketch with the inacceptable values for the carbonyl-C and the much too low values in ipso, ortho and para-position. In blue the values calculated with the "standard hetero parameters" (Streitwieser, Heilbronner/Bock, Isaacs ...). Now the blue values show: The easiest attack is in the ortho-position, than comes the para-position and than the others. The value for the meta-position corresponds to the standard value for benzene, 2.54.

All other L(+) values are much higher. To simulate this attack with the Hueckel-model is not defined using localisaton energies for an attack at the nitrogen of acetanilide. The result of this attack is an ammonium ion and the positiv charge is not in the pi-sytsem. May be, those positions should be omitted from the calculation of L(+) values.

The second image below, shows the results for the various pi-systems which are the basis for the L(+) calculation. My evaluation of the Marvin results lead me to the conclusion, the parameters are not the same for the different pi-systems. That's wrong.

The localisation energy model for the electrophilic substitution must have the same hetero parameters for the original pi-system as for the localised cations (See the very instructive examples in Rauk's book). The localisation energy L(+) is simply the difference between the pi-energy of the total pi-system and the pi-energy of the localised cation.

To change the hetero parameters for different pi-systems is outside the simple Hueckel MO method.

In this context a question. Why is the Hueckel-calculation blocked for single atoms? So I could not calculate the Marvin-pi-energy for system number 2.

In the other now very very long Hueckel topic http://www.chemaxon.com/forum/ftopic3612.html

I have shown how to program the simple Hueckel MO method, and I will repeat it here shortly:

pi-topology - pi-hetero paramaters - Total number of pi-electrons,   that's all.

And the hetero atoms must be restricted to C N O F P S Cl Br I.

This leads for these 9 atom types to maximal (9 * 10) / 2 = 45 heteroparameters. But in reality not all of these make sense, f.e. a bond value for Cl and Br does not make sense.

Regards, Hans-Ulrich

Hi,

Why is the Hueckel-calculation blocked for single atoms? So I could not calculate the Marvin-pi-energy for system number 2.

I will check why result is not  appear for single atoms.

I have a  question:

Nitrogen atom has "+" charge  in the "system number 2". It means that the number of pi electrons decreased with only "1".  This is not correct anyway.   Electrophilic attack requires "2" pi electrons to be excluded from the delocalized  system.

How can you explain the depicted structure of  the "system number 2" ?

Thanks.

Jozsi

Hi Jozsi, yes you are right:

To draw "system number 2" with a "+" charge at the nitrogen is bad. I have long time tried to draw a mesomeric resonance structure for a cation which is produced by a electrophilic attack at the carbonyl C: I didn't find a reasonable structure.

The original total system had 10 pi-electrons. The electrophile "catches" 2 pi-electrons and so the rest pi-system has 8 electrons, distributed over 8 pi-centers with 7 single positive cores and 1 double positive core at N. So there is a charge balance: 1 positive charge for the rest pi- system, which is correct: The electrophile comes with 1 postive charge and bonds to the system, and the postive charge goes to the rest pi-system. This is easy to draw for the normal attack at an aromatic ring and leads to the well known "sigma-complex". This "sigma-complex" is a simple model for the transition state of the electrophilic substitution.

But what is the result of an attack of a positively charged electrophile at the positively charged C-Atom of the carbonyl group ? Chemically this does not make any sense. But formally the electrophile catches the 2 pi-electrons of the carbonyl-bond. So the oxygen gets 1 positive charge ? The most electronegative atom in the total cation should have the positive charge ? I defeated to draw such a structure. So I put the positive structure to the nitrogen which had 2 pi-electrons in the original molecule. But this is also a bad structure, as you have commented.

The only way to get a reasonable picture for this absurd cation is to make a MO-calculation, and a simple HMO calculation with the same "Standard parameters" as above shows a better picture, see the attached image.

The cation has 4 doubly occupied MOs, see the image. On of these MOs in not conjugated to the rest system, it is an AO at the oxygen having two electrons and so the oxygen gets 1 negative total charge. So the rest pi-system gets 2 positive charges, distributed over the rest system. The nitrogen has +0.54, the ipso-position +0.45, the para-position +0.45 and the rest is ortho and meta. How to draw a reasonable mesomeric resonance structure for such a system ?

The whole analysis shows: An electrophilic attack at the carbonyl-C is absurd. And the high L(+) value 3.13 fits to this analysis.

O, this post got rather long.

Regards, Hans-Ulrich

Hi Bob,

it would be fine to give this topic a new title  ?

"Calc. of localization energies L(+)" or something like this.

Regards, Hans-Ulrich

Sure.  Most of this discussion is over my head, anyway.  I am not a theoretician.

Here is what I think is the most important outcome of this discussion:  The constants (parameters) that JChem uses to make its calculations of energies should be located in a file, perhaps even in evaluator.xml.  Hans and other experts can refine the parameters, the rest of us can update them immediately, and ChemAxon can incorporate the improved parameters into new JChem versions as they are released from time to time.   (Alternatively, or in addition, the JChem API should include methods for modifying the parameters.)  Then Jozsi can focus on programming instead of worrying about localization energies.

Jozsi, have you had a chance to discuss this concept internally yet?

BTW, our local newspaper had a very nice article about the Hungarian racehorse Overdose this week.  There is local interest here because Lexington, KY is a major center of the racehorse world.  

Hi Bob,

I am also not a "theoretician". I was a synthetic organic chemist, having a lot of experience in chromatography, NMR, mass-spectroscopy, X-Ray and other experimental methods. The computers are a tool, but I don't accept all results the programs put out, especially when the programs are not OpenSource. I remember a time, I got the source code and could read what is really done. Now times have changed and I must "evaluate" what the programs are doing.

To the contants (parameters) I will continue the discussion in the next post.

Regards, Hans-Ulrich

PS: Please change the title, so other readers of this (super!) forum can see, what we are really discussing here.

Hi Jozsi,

now we have at least two problems here:

1) Hyperconjugation: Please delete it, kill it. See the inacceptable examples above.

2) The algorithm for calculating localisation must be wrong. It is not allowed to use different parameters for the original pi-system and for the localised pi-system (there are at least four book references in my posts).

3) Please use the consistent parameters table given by F. A. Van-Catledge, J. Org. Chem. 1980, 45, 4801-4802.

And do not change these parameters to special compounds. Doing this means to fit the output of the program to the wanted result. It is inacceptable, to have every new version a new parameter set. The Hueckel MO model is much too simple to expect results will fit to all experimental facts. Even the "ab initio"-user don't get always the desired output.

You know I am a "fan" of the Marvin software. But the number of bugs must be reduced.

Best regards, Hans-Ulrich

Hi Bob,

Jozsi, have you had a chance to discuss this concept internally yet?

I mentioned this concept to the right people. We did not discussed the details deeply yet. But I think that everybody see now in Chemaxon that this is an up to date  problem what we need to solve asap.

Jozsi

Hi Hans-Ulrich,

How to draw a reasonable mesomeric resonance structure for such a system ?

I tried to draw resonance structures for the "system number 2".

Your double cationic system would requires 4 charged atoms as it is shown on the attached resonance structures.  This strong charge separation is no plausible I think so.

I think the  "oxygen cationic" type resonance structure is a "better" model molecule for studying this "forbidden" electrophil attack at the carbonyl carbon.

Jozsi

P.S. I will ask a system administrator to change the name of  this topic.

Hi Jozsi,

first I must comment your mesomeric resonance structures, pease see the image LE_Acetanilid6.

The upper formulas you have drawn are not resonance structures for the localised sigma-complex, which is the result of the attack of the electrophil. To say it once more: The E+ "catches" two electrons from the original pi-system, and when the oxygen remains negativelly charged, these two electrons must come from the aniline-pi-system. This aniline had 8 pi-electrons, and when 2 of these are withdrawn, it has only 6 and so it is doubly charged. That's very bad of course.

The lower part of the image contains a comment of Bob, given in an other topic, concerning mesomeric resonance structures for carbonyl groups, please see ftopic4332.html

The discussion here is once more about the problem how to translate MO-results to VB-structures. This was the reason I hesitated to draw resonance structures for the sigma complex. But I will try it in the next post.

Now possible mesomeric resonance formulas for the sigma-complex resulting from the attack of the electrophil. The electrophil catches two blue electrons to form a sigma-bond. The resulting pi-electons are drawn in red.

When the most electronic atom oxygen remains negatively charged, then the ring system with the amino-group must be doubly positively charged. The result for the cation which results from the attack of E+ must be a total charge of (+1) = (-1)oxygen + (+2)ring. This analysis corresponds to the MO description given above. Please compare the MO-pi-charges with the resonance formulas.

Hi Jozsi,

but now I must come to the real problem: Bob gets immediately answers to his requests, but there is no real answer to my severe failure reports concerning the Hueckel calculations here and in the other long topic about Hueckel-calculations   www.chemaxon.com/forum/ftopic3612.html  .

I made nearly three months ago (Feb 07, 2009) in this old ftopic3612 many posts and suggestions to correct the Hueckel calculation algorithms. But there was no answer.  And there are more and older topics about Hueckel calculation problems, but I will not repeat these here. Some of these are corrected, some are not.

When you need it, I will discuss these problems with your colleagues, too. And I will repeat a sentence from above:

"You know I am a "fan" of the Marvin software. But the number of bugs must be reduced."

Otherwise I would not invest so much time to minimize the failures.

Best regards, Hans-Ulrich

Hi Hans-Ulrich,

I want to jump in here in defence of our development and appreciation of your efforts.

At this time we develop features very much guided by users and the market and we find that Hueckel calculations are not a high priority for users, so do not focus effort here.

I am sorry that the features/bugs are not being ironed out as you wish. We are also unhappy, partly because of these failures, but also because we appreciate your effort and would not want to discourage your taking the time to test and let us know where we need to improve.

Can I ask you to continue but consider that as a "response based" development company we are not completely in control of where we can spend resources.

Cheers

Alex

Hi Alex,

thank you for your answers. So I would like to continue the discussion here with some comments.

"Hueckel calculations are not a high priority" - The first post in this topic was on reactivity of electrophiles attacking aromatic compounds. The word "Hueckel" came up in this topic when I showed the basics of "energyE()" i.e. the localisation energy L(+) calculated using the HMO-model. So it's not a discussion on Hueckel calculations, but on the use of reactivity-parameters in Chemoinformatics. And this should be very important for Marvin/JChem users.

"the features/bugs are not being ironed out as you wish" - The failures in the L(+)-calculation for phenol are now fixed in Marvin version 5.2.2 coming May 27, 2009. See the attached image phenol.png. Now the L(+) energies perfectly fit the values which have to come using the standard hydroxy-parameters (Streitwieser, Heilbronner/Bock, Isaacs ...)

(BTW: There was no information of this bug fix in the list for this new realease 5.2.2 , and no information in this topic here, too).

But now Marvin is using different N-parameters comparing aniline and acetanilide. For aniline Marvin still uses the standard parameters but for acetanilide there are now new parameters (It's easy to analyse this with various HMO-programs). This policy is inacceptable. It leads to as many parameters as many compounds you have. Then one could make a data bank and store the reactivity itself.

Question: Which parameters are used for acetanilide ?

The handling of hyperconjugation is still inacceptable. There are also new parameters. But this will lead to a new topic ASAP.

Regards, Hans-Ulrich

Now that Marvin 5.3 has been released, I would like to ask if it addresses the problems that Hans-Ulrich and I have identified in previous posts.  Specifically, will it give accurate calculations of localization energies in aromatic compounds?  If it does not, when you will fix these problems?  We have developed new features in our software that rely on these calculations to be accurate, and they *must* be accurate if we are to ask students and instructors to use these features.

Hi Marvin developer and documenter,

before I will post here my evaluation of the newest MarvinSketch version 5.3.0.2 I have questions:

What is the difference of "HMO Localization energy" and "Localisation energy" ?

What is the difference of "HMO Electron density" and "Electron density" ?

What is the difference of "HMO Charge density" and "Charge density" ?

I cannot find the definitions of these new options.

Please see the screenshots.

Regards, Hans-Ulrich

Hi,

The "HMO" refers to the  parameter set that was pointed out by you earlier. I attach the  "HMO.pdf" file including k_ij and h_i parameters. 

Thanks in advance for testing.

Jozsi

Hi Jozsi,

it's good to have now the self contained parameter set developed by F. A. Van-Catledge, J.Org.Chem. 1980, 45, 4801-4802. In collaboration with Arvi Rauk I have optimized SHMO: http://wchem.cup.lmu.de/SHMO . It uses the same parameter set as default und you can now see the h and k parameters immediately after changing an atom to a heteroatom. (The former version of SHMO showed these parameters only in the Data Table). A detailed analysis will come in a further post in this topic.

The documentation should contain the reference to Van-Catledge.

And: Evaluating the Hueckel options without "HMO" showed me, these are obviously the old versions, containing failures mentioned in my earlier posts. I will document these also in a further post.

Regards, Hans-Ulrich

Hi Jozsi,

evaluating the Huckel options not beginning with "hmo..." I found in the cxcalc-documentation the confirmation of my analysis above: The Huckel options not beginning with "hmo..." are deprecated. (As a Java programmer I know this attribute very well ).

But not all cxcalc options have this attribute. The following cxcalc options should be also defined as deprecated:

chargedensity - electrondensity - electrophiliclocalizationenergy -
huckel - huckeltable - localizationenergy - nucleophiliclocalizationenergy

For "picharge density. Deprecated, use "electrondensity" calculation."and "totalchargedensity. Deprecated, use "chargedensity" calculation." the pointer should be changed to the "hmo..." options.

And the attribute deprecated should also be shown in the menu. Or, better: Delete these from the MarvinSketch menus. We don't need these in the GUI.

So I will stop my evaluation of the deprecated huckel options and I will concentrate on the new options.

Regards, Hans-Ulrich

Hi Jozsi,

as already mentioned in earlier posts of this topic the now "deprecated" Huckel calculations tool still use wrong parameters. The h- and k-parameters for nitrogen in pyrrol and pyridin cannot be the same, see the analysis shown in the image.

Once more a reason to delete these deprecated methods, at least from the GUI.

Regards, Hans-Ulrich.

Huckel calculations are now marked with the HMO prefix. The rest of the functions are not considered Huckel calculations, they are special calculations used for reaction predictions. They must be available in the Marvin user interface to be able to calculate these values as well. We will consider to move them away to another plugin sometimes, however.

Hi,

there is another good result of using now a new version of the Huckel calcualtions.

"Hyperconjugation" is now omitted, and that's very good.

The results of the old now deprecated Huckel implementation where very inconsistent and it is an "overuse" of the simple Huckel method.

Regards, Hans-Ulrich

Hi Jozsi and Gyuri,

now our posts are "crossing" each other and we have nearly a "blog".

But I cannot agree to your argumentation. Of course many users want to continue the calculation methods there are using the years before. But what is, when the methods are inconsistent and sometimes wrong ?

At least you should clearly seperate the deprecated methods from the new. Only to have the prefix "HMO" is not enough. And in the menu it must be clearly visible that the user uses a deprecated method.

A first easy correction would be, to add to "Localisation energy" the word "deprecated".

Of course this not a good marketing point. But Sun has no problems with it. The compiler says "deprecated" and the user is now prompted to change his code.

And these changes could be done in only three lines of your Java code.

Regards, Hans-Ulrich

Hi Zsolt,

why did you delete your post at 10:44 ?

Regards, Hans-Ulrich

Gyuri answered your question before I did.

Zsolt

OK. Regards, Hans-Ulrich

Hi Gyuri,

the post Feb 11, 2010, 10:31 in this topic is unanswered. To emphasize this parameter problem I have created a new image presentation, see attachement. I think the image is self-explaining.

As already mentioned in the HMO-topic http://www.chemaxon.com/forum/ftopic3612.html HMO-calculations have only three things as input : 1) Topology of the pi-system     2) Some hetero parameters depending on the neighbor topology (not the total formula)    3) The total number of pi-electrons.

Which algorithm does Marvin use to set these parameters ?

Regards, Hans-Ulrich

Hi,

I think, that Jozsi is the one who can provide with the most accurate answer regarding the algorithm used. He will answer asap.

Best regards,

Gyuri

Hi,

Which algorithm does Marvin use to set these parameters ?

Simple Huckel calculation

I  am glad to see that the HMO localization energy is "OK"  according to your  comparison for the two test molecules. Assembling of the "secular determinant" from topology and then calculation of  the eigenvalue-eigenvectors is a "simple task" recently.Finally ,after all of these, calculation of required parameters can be done.  

Reactor calculation

On the other hand the old calculation result ,which is invoked by the "reactor" simulation,  is significantly different from the standrad HMO. The algorithm basically the same than above. The parameters and their dependency on the environment ,however, completely different from the SHMO.   Additionally the "hyperconjugation" and all elements of the "periodic system" allowed in calculation. 

Jozsi

Hi Jozsi,

thank you for the information. But there is also a question in the attached image:

What do you mean about: Imidazol has the same localisation energy in postion 4 and 5 ?

And in benzimidazol there is the same "right/left" problem.

Are these reactivities really equal ? Or is it the result of fast tautomerization ? (Is this really fast?)

Regards, Hans-Ulrich

Hi Jozsi,

before answering to your "Reactor calculation" comments I will post here two examples more for which the "Localization energy calculation" (without prefix HMO) uses inconsistent hetero parameters for nitrogen.

Are the localization energies really equal on the left and ride side of these unsysmmetrical molecules ?

Regards, Hans-Ulrich

And now comments to the sentence

"The parameters and their dependency on the environment ,however, completely different from the SHMO."

As far as I have evaluated these parameters these are the standard parameters coming from Streitwieser, f.e.:

C-atoms with 3 neighbors: C pi-orbital with 1 default electron and h(C)=0.00


N-atoms with 2 neighbors: N1 pi-orbital with 1 default electron and h(N1)=0.50


N-atoms with 3 neighbors: N2 pi-orbital with 2 default electrons and h(N2)=1.50

k(C-C)=1.00, k(C-N1)=1.00, k(C-N2)=0.80

The only problem here seems to be: the allocation, classification of the nitrogen atoms contributing 2 pi-electrons (N2) and having three sigma-bond neighbors (- NH -   or   - NH2) is wrong: Here the "aza-parameters" are taken from N1.

How and why does the environment change the N2-"amino-nitrogen" to a N1-"aza-nitrogen" ?

But this is an old problem: http://www.chemaxon.com/forum/ftopic3612.html

For my last post in this topic   Sat Feb 07, 2009 1:01 pm   there was no answer.

Regards, Hans-Ulrich

Hi,

And now comments to the sentence

Certain parameters are allowed to be standard literature data.

How and why does the environment change the N2-"amino-nitrogen" to a N1-"aza-nitrogen" ?

   I will check. (may be a bug).

But this is an old problem: http://www.chemaxon.com/forum/ftopic3612.html


For my last post in this topic   Sat Feb 07, 2009 1:01 pm   there was no answer.

   One of the reason why the new simple HMO method was born is to give answer to all of  your earlier questions (e.g. "kill hyperconjugation")  

   The most important question now is that the new HMO method can be used for reactor porposes or not? Please note that,as you know, the "hyperconjugation" and  some atom type not included in the (e.g. Br,I ...etc) self-consistent parameter set given by Van-Catledge.

May be adding of the "hyperconjugation" and parameters (h,k) of some other atoms will be enough for reactor simulation. 

If yes. What kind of parameter values  can be assigned to these? Can you help in clarification of this question?

Thanks.

Jozsi

Hi Jozsi
now we come together.

To your sentences "The most important question now is that the new HMO method can be used for reactor porposes or not? Please note that,as you know, the "hyperconjugation" and some atom type not included in the (e.g. Br,I ...etc) self-consistent parameter set given by Van-Catledge."
my answer is:

We can try to implement "hyperconjugation". There are 3 models in the Streitwieser parameter list (please see Table 3, page 135). We can make test calculations with pi-SCF or other quantumchemical calculations and the we can adapt the parameters h and k.

Further elements could be: Br, I and Se. But other would be very problematic.

But before we do this, I must comment a new/old problem: The new Hueckel implementation in MarvinSketch 5.3.0.2 is no more able to handle mesomeric resonance formulas. After a long discussion MarvinSketch 5.2.0.x was able to handle those formulas. It's terrible.

I must open a new topic, because this topic is now to long to discuss this.

Regards, Hans-Ulrich

BTW.: Also the new website has lost options in the post reply editors. Where are the buttons bold, underline etc. ? And the are no more new lines. - Now, two ours later, the editor works and I could edit the post once more.

Hi Bob,

this topic was originally coming from you.

What is your meaning about an upgrade of localization energy calculations ?

Regards, Hans-Ulrich

Hi web site administrator team, would you help me once more ?

This topic is now so long, that it's devided into two pages. This is not a "feature" it's a real hindrance.

While editing a topic I have often the same topic open in two firefox windows. In the window on the left side I'm scrolling up and down using the middle mouse wheel and in the window on the right side I'm editing my next post, always looking and analysing the earlier posts in this topic. But when this topic is partitioned  it is no more possible to scroll it: a real hindrance.

And there is another negative result of this partitioning in pages. New readers of a topic scroll down the first page and then they don't recognize the little tiny "Go to page 1,2 Next" in the right down corner. So they never see the real end of a may be long discussion (and often a topic is not at end and still open).

This difference of reading a book with pages and reading a long topic without having to change pages is an advantage of screen scrolling !

Splitting long lists in pages is another thing. F.e. the topic list
of a forum member can be partitioned. Here a long list contains
different subjects not depending on each other.

Technically I cannot see any advantage of this partitioning. The cache of the browser is large enough, the memory of the computer is large enough, and having pages often produces a reloading when I have to change the pages.

Please avoid the partitioning of topics. Technically this would be easy for you. Simply enlarge the length when the partition-mechanism starts. And all is OK.

Regards, Hans-Ulrich

Thank you, I forward your proposal to the webmasters.

We already solved it. Now 100 post are displayed per page, while previously 50 posts were displayed.

Zsolt

Super. Thank you.

Best regards, Hans-Ulrich

Hi, in the attached image it is shown, that calculating the localization energies using the Hueckel method gives for a pi-donor substituent (like the amino NR2) the following order:

ortho 2.279  <  para 2.323  << meta 2.543  (unsubstituted benzene has 2.536)

The reason for having ortho lower than para is in the topology: the localization in ortho position leads to a pi-conjugated chain with the donor amino group at one end. This topology is more stable (8.763) than the para-localized system having the donor amino in the middle (8.719), see the attached image.

Or to say it in another way: the linear chain pi-system is more stable than the branched pi-system. Hexatriene C=C\C=C\C=C has the pi-energy 6.988 and 2-vinylbutadiene C=CC(=C)C=C has 6.899.

But the chemical reactivity is often: para is more reactiv than ortho. It seems, the HMO based localization energy calculation cannot simulate this.

May be it would be possible to simulate the experimental reactivity using a frontier orbital model: The squared coefficients in the HOMO (Highest Occupied Molecular Orbital) of the anilin example (same parameters) are in the order: para 0.230 > ortho 0.137 >> meta 0.026. So an electrophile should prefer an attack at the para position.

I will post this here as a call for opinions.

Regards, Hans-Ulrich

Do you think, that we can expect a correct prediction of the ratio of regioisomeric products from the electronic parameters only? The steric effects (hindrance) can influence the ratio of the ortho/para ismorers as well, but they are not included in the current rules.

Yes, steric factors need to be taken into consideration as well.

We have been using the following definition:

        <property dictRef="REACTIVITY" title="REACTIVITY">
          <scalar><![CDATA[energyE(ratom(1)) - min(energyE(reactant(0), filter(reactant(0), "aromaticAtom()"))) <= 0.0675]]></scalar>
        </property>
        <property dictRef="EXPLAIN_REACTIVITY" title="EXPLAIN_REACTIVITY">
          <scalar><![CDATA[Aromatic atoms must have localization energy close to the compound's minimum to provide the main product. ]]></scalar>
        </property>
        <property dictRef="SELECTIVITY" title="SELECTIVITY">
          <scalar>-stericHindrance(ratom(1))</scalar>
        </property>
        <property dictRef="EXPLAIN_SELECTIVITY" title="EXPLAIN_SELECTIVITY">
          <scalar><![CDATA[Among atoms determined by the reactivity rule, aromatic atom with the least steric hindrance provides the main product. ]]></scalar>
        </property>
        <property dictRef="TOLERANCE" title="TOLERANCE">
          <scalar>0.04</scalar>
        </property>
        <property dictRef="EXPLAIN_TOLERANCE" title="EXPLAIN_TOLERANCE">
          <scalar><![CDATA[Permits ortho products for O(-) and OH.]]></scalar>
        </property>

The "0.675" is an arbitrary number chosen from looking at the energyE() values in various substituted aromatic compounds.  Unfortunately, with JChem 5.2 parameters, the algorithm always fails for at least one compound.  At 0.675, chlorobenzene gives both meta and para products; reduce the value to 0.475 to exclude the meta product, and t-butylbenzene gives an ortho product.  You can see the origins of these results from the following calculations:

[bob@epoch bob]$ evaluate -e "energyE()" "CC(C)(C)c1ccccc1" -p 4

;2.9809;;;3.9168;2.4682;2.546;2.5354;2.546;2.4682

[bob@epoch bob]$ evaluate -e "energyE()" "Clc1ccccc1" -p 4

4.049;2.5849;2.4979;2.5367;2.5035;2.5367;2.4979

So, does JChem 5.3 give different values from those above?  Can you post the results here?  I need the energyE() values that Reactor would use.  

Also, if anyone has a better idea about how to set up the reactivity and selectivity parameters, I'm all ears.

[bob@epoch bob]$ evaluate -e "energyE()" "O=[N+]([O-])c1ccccc1" -p 4 

3.5583;5.9826;3.5583;2.9764;2.6481;2.6506;2.7416;2.6506;2.6481

Why, pray tell, is the energy of the meta C atom *between* those of the ortho and para C atoms?

Hi, trying to answer these questions, I have problems and questions:

I cannot reproduce the results of Marvin using my own HMO program.

1) In the "HMO"-calculations Marvin uses parameters h and k for the nitrogroup I cannot understand. As can be seen in the attached image both oxygens are "O1e" and the nitrogen is "N2e" in the nomenclature of Van-Catledge. What parameters does Marvin use ? Obviously different parameters: The calculated pi-Energies are different (15.809 vs. 14.839)

2) For the rather high "Old"-Localization Energies I have the suspicion Marvin uses once more different parameters for the non-localized and the localized system.

Could you please show me the parameters for these calculations, especially for the "HMO"-version. I must have these to continue my evaluation.

Regards, Hans-Ulrich

Hi,

Your pictures are really help to find bugs. Thanks!

Parameters (h and k) are specified correctly in our program.

However, the number of the electrons of the "Nitro Nitrogen" added into the "Pi system" was badly refenced in our program (only "1e-" was assigned).  This is the search of the "Pi enrgy" difference. I have fixed this now.

Localization energy,however, is not Ok yet.

Jozsi

Why are you guys drawing pentavalent N?  Ugh.

Is energyE() the localization energy or the pi energy?

Earlier in this discussion or another one, we talked about giving the user access to the parameters so that he or she can adjust them without waiting for a new JChem or Marvin release.  Any progress toward that goal?

Hi, yes now I can reproduce your HMO calculation results, see the image and the yellow HOMO at the right side.

You should correct your sentence to "HMO Localization energy, however, is now Ok yet."

But the "Old Localisation energies" are not to understand. The values are very high compared with the unsubstituted benzene (here the "Old LE(+)  is OK = 2.54).

In the attached image you see the definition of the localisation energies, i.e. the difference of the pi-energy of the original pi-system and the localized cation (produced by the attack of the electrophile E+). It is not restricted to aromatic rings. Of course you can calculate this LE(+) for other pi-systems, too. F.e. butadiene LE(+) at C-1 = 1.644 and at C-2 = 2.472 (correctly calculated by both options, the new and the old one).

Comparing the SHMO-energies with 3 digits and the Marvin-energies with 4 digits there is "ein Rundungsfehler" (what is that in English ?)

So my conclusion is: Delete the "Old" options and methods. I can see no reason to have it any more.

Regards, Hans-Ulrich

Hi,

Why are you guys drawing pentavalent N? 

This is an allowed , common representation of the nitro group.

Is energyE() the localization energy or the pi energy?

energyE() is the localization energy.

Any progress toward that goal?

I proposed to do this inhouse a long time ago. Unfortunately I don't know about any progress toward this goal. 

Jozsi

Why are you guys drawing pentavalent N? 

It may be common, but it violates a fundamental rule of chemistry.  As for "allowed", you can write whatever you want and I can't stop you, so I guess that makes it allowed.  But it's not "allowed" in the sense of having any contribution to the real nature of the compound, and to an organic chemist, it's like fingernails on a blackboard.  

Is energyE() the localization energy or the pi energy?

OK, so how do I tell Reactor to use the "HMO localization energy" or the "new localization energy", the parameters that Hans has identified as giving much better results?

Hi,

however, is now Ok yet."

  Numerically it is Ok, however, practically (?). Bob complained about a contradiction if I see well.

  Why is the "meta value" between the "ortho and para value" ?

  "ein Rundungsfehler"

   "agreement" (?)  may be result of  "truncation error"  or "round off error"

Jozsi

Hi Jozsi,

to the sentences "Numerically it is Ok, however, practically (?). Bob complained about a contradiction if I see well" and "Why is the "meta value" between the "ortho and para value" ? I will prepare a detailed answer. It is not "a contradiction", it is a consequence of the topology of the pi-system.

Regards, Hans-Ulrich

I was using JChem 5.2 with that calculation that gave the bad data.  I have not yet downloaded JChem 5.3.  Before I do so, I want to know that Reactor gives better results.  If it does not, I see no point in upgrading.

Hi, I think these HMO-options are already implemented, see your examples:

~/ChemAxon/MarvinBeans/bin$ cxcalc -h
Calculator, (C) 1998-2010 ChemAxon Ltd.
version 5.3.1

Anilin: cxcalc hmoelectrophiliclocalizationenergy "Nc1ccccc1" -p 4

id      Localization energy L(+)

1       3,0417;2,8376;2,2791;2,5428;2,3232;2,5428;2,2791

Nitrobenzene: cxcalc hmoelectrophiliclocalizationenergy "O=[N+]([O-])c1ccccc1" -p 4
id4     Localization energy L(+)
1       2,8108;5,8695;2,8108;2,8443;2,5247;2,5442;2,5718;2,5442;2,5247

In the next post my comments to the nitrobenzene results.

Regards, Hans-Ulrich

Hi, now my comments on the electrophilic localization energies of nitrobenzene.

As can be seen in the attached image, the localization energies for nitrobenzene
are very high compared with the unsubstituted benzene (2.536).
Nitrobenzene has an acceptor substituent, so the localized cation (corresponding
to the Wheland sigma complex) is not stabilized, as it is for the donorsubstituted aniline.

The mesomeric resonance VB-formulas show "awful" dicationic pi-systems in the
localized ring structure (It would be easier to show the destabilization using MO).
Seeing these VB-formulas it is reasonable, that the order is ortho < meta < para.

So the new HMO localization energy calculation results are correct.

And the old version should be deleted. The old version had many bugs and failures.

Now we have to look for parameters for Br and I,
and how to handle hyperconjugation of alkyl substituents.

Regards, Hans-Ulrich

Thanks, Hans.

Question for Jozsi: I see in your JChem 5.3 documentation, Reactor will use the old parameters when it is asked to evaluate energyE(), but it will use the new oneswhen it is asked to evaluate hmoelectrophiliclocalizationenergy().  I just want to confirm that this is indeed the case before I upgrade.

Hi,

Thank you Hans for this explanation.

I just want to confirm that this is indeed the case before I upgrade.

Yes Bob this is the case now.

Jozsi

Hi,

I see some contradiction here.   Benzene E(+) = 2.536 

                                                   Nitro Benzene in ortho E(+) = 2.524

How is it possible that "Nitro Benzene" is more reactive in ortho position than "Benzene" ? (Of course this can  not be true.)

I sketched one more alternative sites  for the "ortho, meta and para" attack.   It can be seen that the stabilization of the "nitro group"  is more favored in the "meta and para" product since more resonance path is allowed.

On the other hand the "ortho" product is less favored in respect of the stabilization of the "nitro group".

Now, I think that the "ortho<meta<para" order is not sure. We need to reinvestigate the parameters of the "nitro group".

Jozsi

Hi, yes, it is a problem of the parameters.

I have changed the parameters: hO=1.5, kNO=1.0, kN=1.5 and kNC=0.8 and now the reactivity order is L(+) m<o<p. Ugh.

And for benzaldehyd there is the same situation: the hO value must be higher  than 1.0 to get the order L(+) m<o<p.

So I must agree: "We need to reinvestigate the parameters" to calculate the reactivity orders.

Regards, Hans-Ulrich

Hi Bob,

while doing some "ab-initio"-calculations to get experience in localistaion energies I have a question:

Which method do you use when you calculate the reactivity of different compounds? F.e., are you using the difference L(+)-aniline minus L(+)-benzene or are you using the quotient L(+)-aniline devided L(+)-benzene ? And, are you using L(+)-benzene 2.5359 as standard ?

This information would help to compare the "ab-initio"-results with the Marvin-L(+) results.

Regards, Hans-Ulrich

We're using this definition for Freidel-Crafts reactions:

        <property dictRef="REACTIVITY" title="REACTIVITY">
          <scalar><![CDATA[energyE(ratom(1)) - min(energyE(reactant(0), filter(reactant(0), "aromaticAtom()"))) <= 0.0675 && energyE(ratom(1)) >= 1.1]]></scalar>


        </property>
        <property dictRef="EXPLAIN_REACTIVITY" title="EXPLAIN_REACTIVITY">
          <scalar><![CDATA[Aromatic atoms must have localization energy close to the compound's minimum to provide the main product. ]]></scalar>
        </property>
        <property dictRef="SELECTIVITY" title="SELECTIVITY">
          <scalar>-stericHindrance(ratom(1))</scalar>
        </property>
        <property dictRef="EXPLAIN_SELECTIVITY" title="EXPLAIN_SELECTIVITY">
          <scalar><![CDATA[Among atoms determined by the reactivity rule, aromatic atom with the least steric hindrance provides the main product. ]]></scalar>
        </property>
        <property dictRef="TOLERANCE" title="TOLERANCE">
          <scalar>0.04</scalar>
        </property>
        <property dictRef="EXPLAIN_TOLERANCE" title="EXPLAIN_TOLERANCE">
          <scalar><![CDATA[Permits ortho products for O(-) and OH.]]></scalar>
        </property>

In the reactivity rule, both the numbers 0.0675 and 1.1 are empirically derived.  Basically, I choose a parameter and then look at how it affects the products that are calculated for various substrates.  So far, unfortunately, I haven't found parameters that work for all common cases.

Using benzene as a benchmark is a good idea, but it would still require empirical calibration of the numbers, because the energyE() units are, I think, arbitrary.  

Using JChem 5.3.1, here's an erroneous calculation:

bob@epoch:web$ evaluate -e "hmoElectrophilicLocalizationEnergy()" "Cc1ccc(cc1)[N+]([O-])=O" -p 4

;2.5718;2.5442;2.5247;2.8443;2.5247;2.5442;5.8695;2.8108;2.8108

bob@epoch:web$ evaluate -e "hmoElectrophilicLocalizationEnergy()" "c1ccc(cc1)[N+]([O-])=O" -p 4

2.5718;2.5442;2.5247;2.8443;2.5247;2.5442;5.8695;2.8108;2.8108

Note that JChem claims that the C meta to the NO2 group is less positive (i.e., more electron-rich) than the one ortho to the NO2 group.  This is just not true.

Also:

bob@epoch:web$ evaluate -e "hmoElectrophilicLocalizationEnergy()" "Cc1ccc(cc1)" -p 4

;2.5359;2.5359;2.5359;2.5359;2.5359;2.5359

It looks like hmoElectrophilicLocalizationEnergy() does not take the electron-donating ability of alkyl groups into account, so it is not acceptable for calculating regioselectivities in electrophilic aromatic substitution reactions.  Is there another function that we should use?

Hi Bob and Jozsi,

reading the last two posts here, I have the suggestion to continue the development of the "old" L(+)-calculations, too.

There are some problems (f.e. the handling of imidazol, pyrazol a.o., see my posts Feb.15, 2010 8:43 am and Feb 16, 2010 9:28) and also the handling of alkylsubstituents (CH3, CH2H5, CH(CH3)2 and C(CH3)3 are handled with the same parameters), but these could be corrected.

I am looking for a correlation of more sophisticated quantumchemical calculations with parameters for L(+) calculations.

Regards, Hans-Ulrich

Hi,

while continuing the testing of the Localization Energy calculations I found a new (and old) problem:

The
L(+) calculations for the three important mesomeric resonance
structures for the chromophore of Michler's Hydrol Blue give three
different results, see the attached image. But only the old L(+) in the bottom are wrong,
the results for the "HMO L(+)" calc. are correct and there are the same
values for each of the three mesomeric resonance structures.

I
have also tested the calculations for the "HMO Charge density" and also
the old "Charge density", here the results are both correct and the
symmetry of the cation is preserved.

BTW: While developing the L(+) calculation procedures I have the suggestion to rename the "old" Huckel tools as

"Reactor Localization Energies" and "Reactor Charges densities" etc.

Regards, Hans-Ulrich

Hi,

I have the suggestion to rename the "old" Huckel tools as "Reactor Localization Energies"

Yes, we will rename.

old L(+) in the bottom are wrong,

The fix will be available in the 5.3.3 version.

Jozsi

Will JChem 5.3.3 realize that the C atom ortho to a nitro group is considerably more positive than a C atom meta to it?  See the calculations below.

bob@epoch:~$ evaluate -e "energyE()" "O=[N+]([O-])c1ccccc1" -p 4 

3.5583;5.9826;3.5583;2.9764;2.6481;2.6506;2.7416;2.6506;2.6481

bob@epoch:~$ evaluate -e "energyE()" "O=[N+]([O-])c1ccc(C)cc1" -p 4 

3.9676;5.5951;3.9676;3.2559;2.5291;2.6265;4.3326;3.391;2.6265;2.5291

Hi,

Will JChem 5.3.3 realize that the C atom ortho to
 a nitro group is considerably more positive than a C atom meta to it? 

   Yes.

Jozsi

Hi, some more comments in this long topic.

1) There may be some misunderstanding here: The localization energy is not an electronic property as cited above: "the C meta to the NO2 group is less positive (i.e., more electron-rich) than the one ortho to the NO2 group" and "the C atom ortho to a nitro group is considerably more positive than a C atom meta to it". The localization energy describes the stability of the "Wheland sigma complex" coming after the transition state of the first step of an electrophilic aromatic substitution reaction. Of course the calculated energy is calculated using an electron model, the Hueckel molecular orbital simulation, but this energy is not a property of the molecule attacked by the electrophile.

2) So it might be interesting, to have a method to estimate the reactivity of a pi-system by calculating its electronic properties, f.e. the electron densities or the net charges at the different positions. In the upper attached image one can see, here the ortho and para postions are (a little bit) positive and the meta position is neutral. So an electrophile should prefer an attack at the meta postion: OK. But this is only true for the calculation using the HMO options of MarvinSketch.

3) Using the "old" huckel calculation options the (astonishing high) net charges show the order ortho > meta > para, see the lower attached image. How to understand ? I have tried to simulate these results with many different parameters, but I could not verify the MarvinSketch results. So I have the question here: Jozsi, please send me the hetero-parameters, MarvinSketch is using here. I want to understand these results.

4) Both images show, something is wrong in the total number of electrons in these pi-systems. For the "HMO" results, only the net charges and electron densities at the oxygen atoms are wrong. But in the lower example I could not analyse the failures.

5) Question to Bob: Do you use net charges or electron densities to simlulate the reactivity of molecules with pi-systems ? It would be another possibility to simulate the special reactivity of the different positions. And this model would analyse properties of the starting molecules and not of the intermediate.

BTW: I have also edited the presentation of the nitro group a little bit using delocalised bonds, see the images.

Best regards, Hans-Ulrich

Hi,

Only 9 electrons used both in the "HMO" and  in the  "old" calculations.

The parameters of the "old" version are these:

k(NO)= 0.7

k(NC)= 0.7

h(O)= 1.5

h(N)= 2.0

k(CC)=1.0

h(C)=0.0

The fix will appear in the 5.3.3 version.

Jozsi

I have been using energyE(), modified by stericHindrance(), to simulate the reactivity of aromatic compounds in electrophilic aromatic substitution reactions.  I use energyE() because that is the parameter that Gyuri used for electrophilic aromatic substitution reactions in his paper describing how to use Chemical Terms Language to predict reaction products.  I am willing to use any calculated property at all, as long as it gives correct predictions for the relative nucleophilicities of the different C atoms of a substituted aromatic ring.  The hmoElectrophilicLocalizationEnergy() could be useful, except it doesn't account for hyperconjugative or inductive effects.  (If it did, the ortho, meta, and para C atoms of toluene would not have the same value.)  No one has yet pointed me toward another JChem calculation that will give the right values.

Hi Bob,

while evaluating the possibilities for new "Reaction localization energy" parameters (using high tech quantum chemistry, QC) I have a question: Do you know a literature reference having experimental results for electrophilic aromatic substitition reactions, or a review about it? Especially the nitro-group is very difficult to handle, not for QC but for HMO.

Regards, Hans-Ulrich

Neil Isaacs, Physical Organic Chemistry, 2nd ed., pp 473-498 has an excellent explanation with lots of data and references.

Will JChem 5.3.3 realize that the C atom ortho to
 a nitro group is considerably more positive than a C atom meta to it? 

JChem 5.3.4 results:

bob@epoch-virtual:~$ evaluate -e "energyE()" "O=[N+]([O-])c1ccccc1" -p 4

4.5583;6.9826;4.5583;2.9764;2.5563;2.5693;2.7416;2.5693;2.5563

bob@epoch-virtual:~$ evaluate -e "energyE()" "O=[N+]([O-])c1ccc(C)cc1" -p 4

4.7447;6.3723;4.7447;3.1632;2.5145;2.5934;4.3098;3.1681;2.5934;2.5145

As you can see from the last two figures, JChem 5.3.4 does NOT realize that the ortho C of the nitroaromatic is considerably more positive than the meta C.

When are these very serious bugs going to be fixed?

If the problem is one of parameters, I BEG you, please make the parameters user-readable and writable, so we can fix this problem ourselves.  We cannot ask students to use features that give such inaccurate results.

Hi,

When are these very serious bugs going to be 
fixed?

I have fixed these bugs. The fix will be available in the 5.3.6 in july, I think so.

The best parameter combination  of  the nitro functional group given in the attached abstract. I have adopted these parameters (5 parameters in table 1) . 

...please make the parameters user-readable and 

writable...

This will be implemented in the near future.

Jozsi

When are these very serious bugs going to be 
fixed?

I have downloaded JChem 5.3.6, and I am still getting exactly the same results for nitrobenzene as I did previously.  I made sure to update my classpath to point to jchem5.3.6/lib before running Evaluator from the command line.  Did you implement the fix in JChem 5.3.6?

-- Bob

Hi,

My result with the 5.3.6 version is:

C:\cygwin\home\marvinhome\marvin\scripts>evaluate -e "energyE()" "O=[N+]([O-])c1ccccc1" -p 4
3.9966;9.0904;3.9966;3.1111;2.5694;2.547;2.6423;2.547;2.5694

Here you can try the Marvin 5.3.6 (GUI version) as well. You need to obtain the same result as given with the "evaluator" above.

http://www.chemaxon.com/marvin/sketch/index.php

Please let us if your result  will be  different. Thanks.

Jozsi

I realized my problem.  I had changed the classpath variable to point to JChem 5.3.6, but not the path variable.  Now I am getting the right answers.  

I like it that I can access these calculations from the Marvin 5.3.6 GUI now.