Talk:Hamiltonian mechanics

Accessibility
I have a master in math. I understand that Euler-Lagrange ODE is the result of solving the Gateaux derivative of a functional to zero. So I can see how the Euler-Lagrange ODE is useful for solving, e.g., optimal control problems. I seem to not know what the purpose of the Euler-Lagrange ODE is outside of minimizing a functional, and how this is useful for modelling, though I tried several times. I knew before reading this wikipedia article that the Hamiltonian principle is an equivalent to the Lagrangian principle. And I do understand that the Lagrangian principle relates to the Euler-Lagrange ODE. Now after reading this wikipedia article I do know exactly as much as I did before, hence, these questions remain unanswered by the article:

1) what is the principle exactly? e.g., an equation? If so, what is the recipe for composing it?

2) what is the problem that the principle solves?

3) how does the principle solve the problem?

4) how does Lagrangian principle solve the problem? (or refer to where lagrangian's principle is applied to answering 1-4; alas, it is no shame to reproduce info from another page to have a self-contained presentation...)

5) give an example without short-cuts. E.g., you pose the hamiltonian but do not explain which procedures you followed to pose it.

Surely, it is no-one's job to make the article useful to anyone. And I see it is frustrating to see that it appears useless even to optimally prepared readers at the present stage. But I want to encourage everyone who is able to make it accessible. — Preceding unsigned comment added by 2A02:C7F:EACB:CE00:6892:E17F:5044:49D6 (talk) 17:45, 17 November 2021 (UTC)

It would be great to give the nomenclature, complete with SI base units - that would help general readers understand the article. It's probably a grat textbook summary for the specialist but surely someone can at least give an introduction for the reasonably-educated generalist ? — Preceding unsigned comment added by 51.219.114.246 (talk) 12:00, 10 July 2023 (UTC)

Intro
Hmm, wouldn't an introduction to Hamiltonian mechanics WITHOUT starting from Lagrangians and starting from symplectic spaces and Poisson brackets be more natural? Phys 19:09, 5 Sep 2003 (UTC)

Er, NOT to most physicists chemists and engineers... :-) Linuxlad 19:01, 17 Mar 2005 (UTC)

Page move
The move of page title isn't a good idea. We generally prefer general titles (Hamiltonian mechanics), to more special ones, such as particular equations.

Charles Matthews 10:36, 16 Jun 2004 (UTC)

For what it's worth, I agree with Charles. Also, can someone please explain the current mess of Talk pages involved? Especially Talk:È§£æžåŠ›å... (Mozilla won't let me type the whole of it).

Taral 17:42, 17 Jun 2004 (UTC)

What I want to know is why the Hamiltonian view of mechanics is more useful than the classical version? What can engineers and scientists do with it that they cannot do w/o it?


 * Good question. Hamiltonians are great for solving problems that involve transfer of energy, and momentum, between rigid bodies, and in some circumstances allow problems that would be hard to solve using just Newton to be solved in very few lines of equations. Examples include things like compound pendulums and so on. However, very few engineers use Hamiltonians in their day to day lives, to a large extent they have been replaced by kinematic programs like MSC ADAMS, which have the advantage of being able to easily handle flexible bodies, non ideal contact with damping, and friction. I did not study them at uni, and must confess that this page is absolutely no help at all to me, and I work in the area of dynamics. I'm sure it makes sense to physicists. Greglocock 04:48, 3 January 2007 (UTC)

Norm
Liouville's equation here shows total d/dt equal to the Poisson bracket. A physicist would expect to write partial d/dt here, because the essence of Liouville is that total d/dt, meaning the convective derivative taken with the particle, is zero. See main Liouville's theorem (Hamiltonian)

Linuxlad 12:28, 9 Nov 2004 (UTC)

Looking at this issue again, Goldstein's Classical Mechanics

(Later Note - this is Goldstein 1964 edition ie 1950 2nd reprint)

at the ready, there appear to be two differences from what I'd expect:-

In the absence of any further constraints on f, I'd expect (cf Goldstein eqn 8-58) that:-

a) the convective/total time derivative of f equalled the Poisson bracket of f & H PLUS the partial time derivative of f.

b) In the _particular_ case of phase space density (or probability) it is possible to show that the convective derivative is zero, so that the partial derivative equals minus the Poisson bracket (Goldstein eqn 8-84) - but note that this result does NOT follow trivially from the result for general f as implied.

(So I reckon that's 2/3 violations of my naive physicist's expectations) - I hereby give notice that I may edit accordingly :-)

Linuxlad 10:18, 10 Nov 2004 (UTC)

This page certainly needs some work. For example it doesn't give (and neither does the page linked to) the classical expression of the Poisson bracket. As far as I can see, though, the definitions are the standard ones, such as are given in Abraham and Marsden, Foundations of Mechanics, though.

Charles Matthews 11:08, 10 Nov 2004 (UTC)

Just to be clear, I'm looking at the first part of the section entitled Mathematical formalism. and specifically the two equations for f and for ρ

So

Equation 1 For general f :-



\frac{df }{dt}=\frac{\partial f }{\partial t}+\sum_{i=1}^{N}\left[ \frac{\partial f }{\partial q_{i}}\dot{q}_{i}+\frac{\partial f }{\partial p_{i}}\dot{p}_{i}\right] $$

and on substituting Hamilton's equations for terms 2 & 3 we get the Poisson Bracket of f & H plus the partial time derivative.

So, (from standard maths methods viewpoint), first equation has partial df/dt missing.

2nd equation (for ρ) - Liouville's theorem is TOTAL d by dt of  phase space density is zero, which does NOT follow directly from above, and is not what's given there anyway (you need to change total time derivative to partial)!

I have now edited acccordingly

Bob

Linuxlad 11:30, 10 Nov 2004 (UTC)

Hamiltonian systems
There is a redirection from "hamiltonian system" to this page, which is certainly better than nothing, but not really satisfying in my opinion of view.

There should be Hamiltonian systems, integral of motion, integrable systems, Lax pairs, ... Amateurs, please try to create some stubs... MFH 17:48, 17 Mar 2005 (UTC)


 * I utterly agree. A Hamiltonian system is never defined anywhere.  I vaguely remember what it is from my lectures ten years ago but that's not enough.  I try a stub please correct me! --131.220.68.177 16:34, 19 August 2005 (UTC)


 * the Hamiltonian system page already exists. This problem is solved. Jorgecarleitao (talk) 14:21, 5 March 2013 (UTC)

Classical Relativistic Electromagnetic Hamiltonian
This article seems to have sprouted a section:


 * Classically (non-quantum mechanical), for a particle of charge q and mass m in an electromagnetic field with vector potential A and scalar potential &phi;, the relativistic Hamiltonian is
 * $$ H = \sqrt{(\mathbf{p}-q\mathbf{A}/c)^2 c^2 + m^2 c^4} + q\phi $$
 * where the term q&phi; is potential energy, the radical term is total energy of the particle, in terms of the energy-momentum relation, with momentum replaced by what may be construed as a "gauge covariant momentum": it is called kinetic momentum.

However, I don't beleive this has anything to do with hamiltonian mechanics. Well, its an example of a hamiltonian, but its one of a thousand examples, and doesn't seem to be a particularly good one. If the goal is to add examples, something simpler should be given: e.g. ball on an inclined plane or something like that. Perhaps this belongs in its own article? linas 00:16, 26 October 2005 (UTC)


 * Don't fully agree with the above. Formulating the classical Hamiltonian for a particle in an e/m field, is a standard way (eg 3rd year undergraduate level, I recollect) of working to the important quantum mechanical case. Linuxlad 11:14, 11 November 2005 (UTC)


 * We have distinct articles for things like the simple harmonic oscillator and the hydrogen atom. I strongly urge you to start an independent article on the --classical relativistic electromagnetic Hamiltonian or maybe relativistic charged particle. Yes, I am well aware that this Hamiltonian is treated in many books, including the awe-inspiring development in Landau and Lifschitz. However, Hamiltonian mechanics has many, many facets to it that stretch far beyond what the third-year undergraduate is aware of. A distinct article for this will allow it to be expanded as it deserves (after all, L&L found 100+ pages of things to say about it). linas 22:36, 18 November 2005 (UTC)

I don't think you've made your case here! The example neatly extends the idea from the pure mechanical variables into e/m using probably the most important example of a single-particle system. But perhaps you're right - this is really a page for people whose searching for deeper mathematical insights should not be sullied with examples of the real world :-) Linuxlad


 * I frankly don't understand why we are arguing about this. There are thousands of Hamiltonians, and while relativistic charged particles are interesting and important for many reasons (and deserve their own article (hint hint)), I can't imagine why they are "important" for hamiltonian mechanics. Surely a ball running down an inclined plane is a "more important" example (as it illustrates contact geometry, sub-Riemannian geometry and lagrange multipliers all at the same time? Maybe a pendulum, as that can illustrate the transition to chaos,  the KAM torus and the phase-locked loop at the same time? The lorenz attractor? An example from fluid dynamics? Why  charged particles?


 * As to the quality of this article; at this current time, it is rather incomplete at this time. However, that's a different issue. linas 01:13, 19 November 2005 (UTC)


 * (Have you ever read the Landau and Lifshitz treatment of the classical relativistic charged particle?) linas 01:19, 19 November 2005 (UTC)

Proposed mergefrom Hamilton's equations
The result was merge sbandrews (t) 18:51, 23 June 2007 (UTC)

This article is supposed to be the category leader for Hamiltonian mechanics. But it offers no orientation for a general reader coming in who has never heard of either Hamiltonians or Lagrangians.

Such a reader might have been directed here by Equipartition theorem for instance (currently the object of scientific peer review at Wikipedia:Scientific_peer_review/Equipartition_theorem)

My belief is that this would be best fixed by merging in most of the content from Hamilton's equations at the top of the article, to give an overview, and an idea of where we're trying to get to, and why; before we get into the algebra fest.

The two articles appear to cover the same subject, namely what are Hamilton's equations, and where do they come from, and what new view of the world do they give one.

So it seems to me a fine idea to merge them, and gather together the best content from each. Jheald 14:09, 3 April 2007 (UTC)


 * That sounds like a good idea, and I'd support the move. &mdash; Rebelguys2 talk 03:54, 11 April 2007 (UTC)


 * make sense, I'll have a go... sbandrews (t) 10:36, 10 June 2007 (UTC)


 * ok I'm done, comments or reverts welcome... sbandrews (t) 11:00, 10 June 2007 (UTC)

Derivation 1
Isn't the form of Lagrange's equations given there plain wrong? Why is the generalized force entering the equation at all? 131.215.123.218 (talk) 23:03, 10 June 2008 (UTC)
 * It seems that derivation is wrong. The lagrange equation is not
 * $$\frac{\mathrm{d}}{\mathrm{d}t} \frac{\partial \mathcal{L}}{\partial {\dot q_i}} - \frac{\partial \mathcal{L}}{\partial q_i} = F_i$$
 * this but
 * $${d \over dt} {\partial T \over \partial \dot{q}_\sigma}  - {\partial T \over \partial q_\sigma } = F_\sigma \qquad \sigma = 1, \; \cdots, \; 3N-k$$,
 * where T is kinetic energy. see Lagrange's Equations -- from Eric Weisstein's World of Physics.
 * Also, the Hamilton's equation seems wrong. The right form is

\frac{\partial \mathcal{H}}{\partial q_j} = - \dot{p}_j, \qquad \frac{\partial \mathcal{H}}{\partial p_j} = \dot{q}_j, \qquad $$.
 * see Hamilton Equations -- from Wolfram MathWorld. --StarLight (talk) 10:52, 25 July 2008 (UTC)

Mnemotechnics
One of the subsections on this page is titled "Basic physical interpretation, mnemotechnics". Can anyone explain the relevance of mnemotechnics to this section? AstroMark (talk) 13:39, 15 August 2008 (UTC)
 * No? I shall remove it then. AstroMark (talk) 14:03, 27 August 2008 (UTC)

Example?
Would it be useful to provide an example or two, to demonstrate how Hamiltonian mechanics is used, and what insight they provide? Perhaps the examples from Lagrangian mechanics could be brought over and redone, highlighting the difference between the two. I would do it, but I don't really know the answers to my own questions. User:!jimtalk contribs 22:45, 13 December 2008 (UTC)

DS1000's added material
I again removed an erroneous statement added to the section Hamiltonian mechanics by. He said "The above remains valid http://academic.reed.edu/physics/courses/Physics411/html/411/page2/files/Lecture.10.pdf< /ref> at relativistic speeds with the observation that: $$p_i=\gamma_im_i \dot{x}_i$$" (I broke the "ref" to keep this in-line). The erroneous part of this statement is that the "above remains valid" which is only true for $$ \mathcal{H} = \sum_i \dot{x}_i p_i - \mathcal{L} \,,$$ but not for the remainder of the section which is non-relativistic. The reference given uses a non-conventional definition of the dot (time derivative) which makes it the derivative with respect to proper time rather than coordinate time. It is possible that he got the erroneous formulas from the part of the reference which contains (10.36) which is a different (from the usual relativistic Hamiltonian) which the author calls "functionally equivalent". But that is for his special purpose. Please do not re-add this material without discussing it here first. JRSpriggs (talk) 05:12, 18 December 2008 (UTC)
 * You aren't making any sense, first you admit that it is correct for relativistic domain and then you complain that it isn't applicable for the non-relativistic domain.Of course it isn't.What is with the stuff you are writing about not using the derivative wrt proper time? The derivative wrt proper time is gamma*derivative wrt coordinate time. Didn't you see the gamma? This is a repeat of the situation we had with your editing the relativistic force derivation . Is it that you want the relativistic Hamiltonian separated out? I did that, I hope that this is satisfactory for you. DS1000 (talk) 05:36, 18 December 2008 (UTC)
 * Your many editings are completely incorrect, please stop to use the wikipedia to defend your incorrect points in sci.physics.relativity http://groups.google.com/group/sci.physics.relativity/msg/e80d792ecfc7b42f< /ref>. First, the square Hamiltonian is not valid for relativistic velocities, second your momentum $$p_j=\gamma_j m_j \dot{x}_j$$ is completely wrong for the given Lagrangian, because lacks the magnetic contribution. You were explained this before in both USENET and BLOGGER and corrected. You were explained what is the correct momentum and relativistic Hamiltonian JuanR (talk) 14:19, 18 December 2008 (UTC).
 * What are you rambling about? You aren't making any sense. DS1000 (talk) 16:42, 18 December 2008 (UTC)
 * Of your incorrect editing of relativistic Hamiltonian deleted by at 21:16, 16 December 2008, you resubmitted and was deleted again by by  at 04:54, 18 December 2008. This is also archived in this section above where  wrote I again removed an erroneous statement added to the section Hamiltonian mechanics by .


 * The more recent version does not contain the wrong Hamiltonians and momentum you wrote in many previous incorrect versions but now contains the correct momentum and square root Hamiltonian was given to you so early as the day 12 Dec. However, you more recent editing is not still acceptable.


 * First, you use notation is incompatible with the rest of the page. for example the relativistic Hamiltonian in Hamiltonian mechanics does not reduce to the square root Hamiltonian given in section Hamiltonian mechanics because you are using a different definition for $$p_j$$. There is not reason to change notation in the middle of page. Momentum can be given as $$p_j=m \gamma \dot{x}_j + eA_j$$ and for low velocities reduces to the $$p_j=m \dot{x}_j + eA_j$$ given in previous section.


 * Second, your claimed equivalent and much nicer representation as function of the relativistic momentum makes no sense. You do not understand the Legendre transformation. The state variable in the Hamiltonian is your given $$P$$ not your $$p$$. Moreover your 'representation' $$H = \gamma m c^2 + e\Phi$$ is not a function of $$p$$ as you claim because $$\gamma$$ is a function of your $$P$$. Again you are mixing apples and oranges.


 * Third, your claim that the Hamiltonian can be viewed as the sum between the total relativistic energy E and the potential energy is also incorrect. The total relativistic energy $$E$$ is obtained from the zero component of the four momentum in a covariant approach. I.e. $$p^\mu = (E/c,\vec{p})$$. Evidently $$E$$ is not $$H - V$$, except for a free particle $$V = 0$$.


 * Please stop from submitting more incorrect editings of this page, your 47 incorrect versions given in the last 4 days are enough!! JuanR (talk) 12:38, 19 December 2008 (UTC)


 * I see, you are some kind of stalker-crackpot, you can't understand the math and you compensate by stalking. Find a different place to conduct your stalking, ok?  —Preceding unsigned comment added by DS1000 (talk • contribs) 14:48, 19 December 2008 (UTC)


 * It is evident you cannot learn but fortunately continues correcting your wrong editings. He has today corrected your incorrect claim that $$E$$ is the total energy http://en.wikipedia.org/w/index.php?title=Hamiltonian_mechanics&oldid=259112015 as I warned in this talk before.


 * However, the notation is still inconsistent with that given in previous non-relativistic sections. Notation in all the page would be unified! Moreover, symbol $$E$$ usually denotes the zero component of four-momentum. The Hamiltonian is better splinted as $$H = mc^2 + T + V$$, where $$T$$ is relativistic kinetic energy (see Goldstein or any other textbook for details).


 * The claim that exist an equivalent representation as function of the relativistic (kinetic) momentum is still not acceptable. The Legendre transformation changes state variables $$(x,v)$$ into your $$(x,P)$$. The kinetic momentum $$p$$ is not state variable in Hamiltonian mechanics (except when there is not magnetic potential, of course because then $$P=p$$), the final expression is not a Hamiltonian and the symbol $$H$$ would not be used for that function. One alternative notation could be $$H=H(x,P)$$ for the Hamiltonian and $$h=h(x,p)$$ for the last expression.


 * I apologize by not editing by myself the article, but I still need to practice more with this system. JuanR (talk) 13:57, 20 December 2008 (UTC)


 * The derivation is a matter of very simple math, it is not my fault that you are unable to follow it. Your sole reason for registering seems to be trolling DS1000 (talk) 18:33, 20 December 2008 (UTC)


 * I do not understand why you insist on either signed or unsigned ad hominem and insults when this talk is about the physics and math in the article. As reported above your last editing http://en.wikipedia.org/w/index.php?title=Hamiltonian_mechanics&oldid=259115585< /ref> still contains some incorrect claims reflecting your misunderstandings about the Legendre transformation, the state variables, and Hamiltonian mechanics in general. I have suggested several modifications to your last editing. Also the notation you choose would be revised because is inconsistent with the rest of the article. I apologize by not editing by myself the article, but I still need to practice more with this system. JuanR (talk) 13:39, 21 December 2008 (UTC)


 * Now that you have separated the material into its own section, the new version (but not the old one) appears to be correct. However, it would be good to say something about what the Lagrangian is in this case. See Lagrangian.
 * I said that the dot was used unconventionally in the reference (the pdf file). I did not say that you used it that way in your addition; I saw the gamma. JRSpriggs (talk) 06:12, 18 December 2008 (UTC)

section on deformable bodies
First, there are no sources cited for this section, and second, shouldn't this be in the article on Hamilton's principle? —Preceding unsigned comment added by 72.10.133.106 (talk) 01:15, 12 January 2009 (UTC)

derivation by time
Hello. You have mistake in the article - the right equation is that the total derivative of H by time is minus the partial derivative of L by time. 23:31, 10 February 2009 (UTC) —Preceding unsigned comment added by ירון (talk • contribs)

Article about hamiltonian
Am I blind or isn't there any article about classical hamiltonian itself (a function)? Magdulewicz (talk) 13:49, 23 June 2010 (UTC)


 * I think that the function (in classical physics) is adequately covered in this article. Is it not? What else would you have us say about it which is not already in this article? JRSpriggs (talk) 06:39, 24 June 2010 (UTC)


 * I guess I'd talk about effective potentials, but that's just my two cents. Effective potentials using the hamiltonian are an interesting thing that you can do without talking about the dynamics of the situation.--24.7.197.142 (talk) 08:53, 3 February 2011 (UTC)

Symbols
Does anyone know the origin of the abbreviations T and V that are "traditionally" used for kinetic and potential operators? The closest guess I can come up with is translational T, but I haven't a clue for V. Laburke (talk) 21:07, 12 January 2011 (UTC)
 * According to Marsden & Ratiu (1999) p. 231 the notation T and V seems to go back (at least) to Lagrange. In a footnote they refer to further possible sources on historical background. -- Crowsnest (talk) 06:45, 14 January 2011 (UTC)
 * Merci, mais je reste avec la même question, pourquoi V? <>. Thank you for looking into this, but I didn't see what V stands for. Perhaps "champ Vectoriel"?Laburke (talk) 20:22, 13 March 2011 (UTC)


 * Frankly at this point in time, it does not matter for what word "V" is short, if indeed there is any such word. Given the immense multitude of concepts and the limited number of mathematical symbols available to represent them, re-use of symbols (overloading) and arbitrary choice of symbols (merely because they are different from some other symbols one has chosen to use) are bound to occur. JRSpriggs (talk) 02:22, 14 March 2011 (UTC)

Relating the time derivatives of the lagrangian and the hamiltonian.
This is really nitpicky. I'm fairly sure you can prove that the total derivative of the hamiltonian equals the negative of the partial derivative of the lagrangian. I don't mean to be weird about this because I look at the proof given in the article, and I'm convinced of the statement about the partial derivatives of the two equaling each other. Though, I'm pretty sure it is actually the way that I have written it here. This is a much deeper statement. I think this should be changed :). Anybody interested in finding a source for this?

$$\frac{d \mathcal{H}}{d t} = - {\partial \mathcal{L} \over \partial t}$$ --24.7.197.142 (talk) 08:49, 3 February 2011 (UTC)


 * Well, I'm stupid. Somebody else mentioned this problem in this discussion page. I'm feeling stupid now for not reading through this discussion page more carefully, but I think I'll just leave this here.--24.7.197.142 (talk) 08:55, 3 February 2011 (UTC)


 * I did more research and the partial of the hamiltonian is equal to the total derivative of the hamiltonian in time. So, everybody is right :). Here's what it is:
 * $$\frac{d \mathcal{H}}{d t} = - {\partial \mathcal{L} \over \partial t} = \frac{\partial \mathcal{H}}{\partial t}$$--24.7.197.142 (talk) 09:00, 3 February 2011 (UTC)


 * See V.I. Arnold (1978) Mathematical methods of classical mechanics, p. 67, for the total time derivative of $$\mathcal{H}(\boldsymbol{p}(t),\boldsymbol{q}(t),t)$$:


 * $$\frac{d\mathcal{H}}{dt}

= \frac{\partial \mathcal{H}}{\partial \boldsymbol{p}} \dot{\boldsymbol{p}} + \frac{\partial \mathcal{H}}{\partial \boldsymbol{q}} \dot{\boldsymbol{q}} + \frac{\partial \mathcal{H}}{\partial t}    = \frac{\partial \mathcal{H}}{\partial \boldsymbol{p}} \left( - \frac{\partial \mathcal{H}}{\partial \boldsymbol{q}} \right) + \frac{\partial \mathcal{H}}{\partial \boldsymbol{q}} \frac{\partial \mathcal{H}}{\partial \boldsymbol{p}} + \frac{\partial \mathcal{H}}{\partial t}    = \frac{\partial \mathcal{H}}{\partial t}. $$


 * -- Crowsnest (talk) 09:20, 3 February 2011 (UTC)

Approach
Somewhere in the article it is written: "One thing which is not too obvious in this coordinate dependent formulation is that different generalized coordinates are really nothing more than different coordinatizations of the same symplectic manifold."

As an engineer I have no idea what a symplectic manifold is. It makes it rather difficult when one is trying to refresh one's ideas about Hamiltonians that one should now have to go and try and understand what "Symplectic manifolds are". Please - When writing articles proceed from the simple to the complex, from the concrete to the abstract and from the particular to the general and not vice versa Phillip (talk) 06:04, 21 April 2011 (UTC)..

Hamiltonian for economics
I'm and economist and I was looking for an economic interpretation for Hamiltonian. I didn't found anything useful here, but I did find a classic paper on the subject, "Introduction to Hamiltonian Dynamics in Economics". Do you think this should be a new section in this article? I think that hamiltonian function shouldn't redirect tho Hamiltonian mechanics, so other areas would only look for the use of Hamiltonian function. Best,Luizabpr (talk) 17:19, 4 July 2011 (UTC)


 * I'm now sure what an economics interpretation of the Hamiltonian would be. This is an article about Hamiltonian mechanics, which is a branch of physics and not related to economics in any way. About the Hamiltonian function, it is the same thing as the Hamiltonian, and hence it has to redirect here (or if Hamiltonian function can mean more things, a disambiguation page should be created). —Kri (talk) 19:32, 13 August 2011 (UTC)


 * Many of the top economists today have advanced degrees in physics. These economist are able to apply techniques of mathematical physics in work on economics. Reviewing this paper it is an application of the technique - not even a generalization and should be included here to illustrate that Hamiltonian techniques can be used in other disciplines. Some pure matematical uses are mentioned in http://terrytao.files.wordpress.com/2008/03/hamiltonian.pdf OrenBochman (talk) 09:44, 11 February 2012 (UTC)
 * The Hamiltonian is pretty basic to dynamic models in economics these days. I'm surprised that six years after these comments there is still no mention of it (or even a link or disambiguation) here. I would encouage someone to add something helpful. Winter84 (talk) 17:35, 16 August 2018 (UTC)


 * Read Hamiltonian (control theory) and Hamiltonian system. Lantonov (talk) 10:21, 27 October 2020 (UTC)

Re-write section of "beauty", also a pinch negative
I have changed the subsection Notes under Simplified overview of uses, tagged by No original research:


 * "Hamilton's equations are appealing in view of their beautiful simplicity and (slightly broken) symmetry." <--- Are they not symmetric in $$\pm q \rightleftharpoons \mp p$$ and $$\pm \dot{q} \rightleftharpoons \mp\dot{p}$$? A little too much hype?
 * "They have been analysed under almost every imaginable perspective, from basic physics up to symplectic geometry." What does "basic physics" mean? A simple physical system with one/very few degrees of freedom? Elementary physics at school level? I can't think of anything.
 * "A lot is known about solutions of these equations, yet the exact general case solution of the equations of motion cannot be given explicitly for a system of more than two massive point particles." <--- The sentence is sort of fine but partially redundant. Generally solutions to differential equations can be analysed qualitatively and quantitatively, with approximations and in limiting cases. They can’t always be found explicitly.
 * "The finding of conserved quantities plays an important role in the search for solutions or information about their nature." <--- This is randomly stuck in, and its easy to read conservation laws from the lagrangian (finding cylic coords). The Hamiltonian itself is conserved for time-independent Lagrangian functions.
 * "In models with an infinite number of degrees of freedom, this is of course even more complicated." <--- Fine, but redundant. Why talk about infinite degrees of freedom when a finite number is easier to comprehend? Also arn't physical models supposed to simplify the siuation and have as few degrees of freedom as possible? Not always, but usually.
 * "An interesting and promising area of research is the study of integrable systems, where an infinite number of independent conserved quantities can be constructed." <--- Source? Why is it so helpful to have an infinite number of conserved quantities? What will a reader get from this?

I really do not think the current version adds anything helpful to the article. All that needs to be said (if anything) is:


 * "Hamilton's equations are symmetric in $$\pm q,\mp p$$, and hence $$\pm \dot{q},\mp\dot{p}$$. Naturally, the more degrees of freedom the system has, the more complicated its behaviour (predicted by the solutions), since the degrees of freedom correspond to the configuration of the system i.e. (generalized) positions, momenta and the rates at which these change (time derivatives). As such, for more than two massive particles the solutions cannot be found exactly - the many-body problem. It is still possible to obtain qualitative knowledge about the system by approximative analysis of the differential equations."

This has replaced that section with no heading (no need) and the tag has been removed. Furthermore


 * "Generally, these equations do not provide a more convenient way of solving a particular problem in classical mechanics. Rather, they provide deeper insights into both the general structure of classical mechanics and its connection to quantum mechanics as understood through Hamiltonian mechanics, as well as its connection to other areas of science." <--- Are the equations not convenient due to their symmetry and that they are 1st order differential equations? There is no problem in using more than one formulation of mechanics, so this has a slightly negative impression on the equations. I changed it to be more neutral;


 * "Sometimes, but not always, these equations provide a convenient way of solving a particular problem in classical mechanics. Also, they provide deeper insights into both the general structure of classical mechanics and its connection to quantum mechanics as understood through Hamiltonian mechanics, as well as its connection to other areas of science."

-- F = q(E + v × B) 12:25, 16 February 2012 (UTC)


 * Also I re-wrote the instructive/textbooky written section the using H equations section. WP is not supposed to be written like this (WP is NOT...). More edits of this nature can probably be done but I'm too busy for now. -- F = q(E + v × B) 12:41, 16 February 2012 (UTC)


 * Forget the crossed out content above - I stated the obvious and it was unhelpful/wrong, so the paragraph in the lead will be re-phrased to the original wording. =( -- F=q(E+v×B) ⇄ ∑ici 10:07, 18 March 2012 (UTC)

Geometry of Hamiltonian systems
The section Geometry of Hamiltonian systems looks accurate enough, with proper links, but could surely be beefed up a little bit, perhaps with a diagram or figure with all the "ingredients". YohanN7 (talk) 14:11, 5 October 2012 (UTC)

I don't really see the point of this section. That's a lot of evolved math for nothing but Hamiltonian systems (in the sense of physics). Maybe the symplectic point of view is already enough for the interested (mathematical) physicist. Just say, a Hamiltonian system can be considered as a symplectic manifold together with a smooth function $H$ on it. The manifold, which is usually the contangent bundle of a configuration space manifold $Q$, is symplectic, because the symplectic structure gives us a notion of poisson bracket and thus we can write down Hamilton's eq. of motion as a global flow equation (give that equation + explain). If needed, one can then go over to the time dependent case and set $\omega = d \theta - d E \wedge dt$ as well as $H_{new}= H_{old} - E$, where we take the cotangent bundle of $Q \times \R$. Also this "which comes equipped with a natural symplectic form, and this latter function is the Hamiltonian." does not make sense. Yes, there is a natural symplectic structure (the cartan derivative of the canonical 1-form), but it is not a function, let alone the Hamiltonian. 94.211.47.103 (talk) 20:44, 10 November 2013 (UTC)

What is Hamiltonian mechanics mathematical formalism
I believe this article does not has, neither explain what is, the Hamiltonian mechanics' mathematical formalism. There is no section actually presenting the main mathematical premisses of Hamiltonian mechanics.

For example:


 * 1) Hamiltonian mechanics requires canonical coordinates variables and not just "generalized coordinates". This is not introduced.
 * 2) the fact that one of the generalized variables is attached to the reference frame chosen is also not explained, which leads to a big confusion of the article... The "momentum" is not arbitrary, it is fixed by the canonical conjugate relations...
 * 3) While Hamiltonian was historically connected by the Lagrangian mechanics, this is no longer true. This can be supported by different sources, such as the chapter on Hamiltonian system of Edward Ott's book "Chaos in Dynamical Systems".

Another missing part in the article relates with the difference between Hamiltonian mechanics and classical mechanics...

I recently made a modification on the "overview of uses", which I believe fixes some of these problems. It was reverted based on an argument of "breaking 's", which I believe is far from pragmatic (besides, every was correct). I'm going to wait a bit until the user who reverted it to comment here, I'm sure the leading argument is not the. In case of no response, I will revert my contribution back.


 * Using TeX for inline equations is not popular. If you use HTML markup for the inline equations, then focus might be more on the content part of your changes. YohanN7 (talk) 19:55, 7 March 2013 (UTC)

Please simplify the introduction
The introduction of this article is way too advanced for me. I am an educated amateur not a mathematician. How about an introduction ( or a separate article) along the style of "Hamiltonian Mechanics for Dummies". I got interested in HM from reading about Newton and I am curious about the trail from Newtonian mechanics onwards. My friends agree with me that many of the maths pages on Wiki have no suitable lead in section. Whoever thought of the idea "books for Dummies" was on to a great idea. — Preceding unsigned comment added by 118.208.122.190 (talk) 22:23, 15 March 2014 (UTC)

Sub-Riemannian geometry example
I think that the Heisenberg Hamiltonian is confusing. If we take the coordinates $x,y,z$ like in the Heisenberg group page, it should be proportional to $$(p_x -p_z y/2)^2 + (p_y + p_z x/2)^2$$. I think the one that was meant in the article is the Hamiltonian in an non-holonomic basis, but in this case $$p_x,p_y$$ have nothing to do with the corresponding momenta for coordinates $$x,y$$, which is some what confusing. — Preceding unsigned comment added by 147.122.44.52 (talk) 14:55, 24 July 2018 (UTC)

Charged particle in an electromagnetic field
Section Charged particle in an electromagnetic field says that the formulas are for Lorenz gauge. I believe the Lagrangian is valid for all gauges? Can someone confirm? Ponor (talk) 14:43, 6 May 2020 (UTC)

2020-10-27 edit: this was discussing the 2020-05-06 version of the article, where the Hamiltonian was said to be valid in Lorenz gauge, as if it would be different in other gauges for the vector potential. This has been corrected since. Ponor (talk) 12:41, 27 October 2020 (UTC)


 * Lagrangian and Hamiltonian are valid for all gauges (measurement systems). That's why they are defined in generalized coordinates which do not depend on which coordinate system or gauge is used. In the physical jargon it's called "gauge-invariant". It's just that the particular formulas that are used in that section are valid for Lorenz gauge. Those formulas would be different if another gauge is used. Lantonov (talk) 10:36, 27 October 2020 (UTC)


 * Hi I would not confuse gauge invariance with the choice of generalized coordinates. My concern was only about the minimal substitution part. At that time, the article said the Hamiltonian Lagrangian $$ \mathcal{L} = \sum_i \tfrac{1}{2} m \dot{x}_i^2 + \sum_i q \dot{x}_i A_i - q \varphi$$ was (only?) valid in Lorenz gauge, while it's actually valid in any: take As and φs in whichever gauge you like, put them in, and physics will be the same. Ponor (talk) 12:41, 27 October 2020 (UTC) Corr. Ponor (talk) 14:04, 27 October 2020 (UTC)


 * You mean Lagrangian instead Hamiltonian. Well, at least now the article does not claim that the Lagrangian is valid only for the Lorenz gauge. The formula is also different replacing the first term with its relativistic counterpart (Lorenz transformation). Since the classic Lagrangian is a generalisation for the Newtonian mechanics (low velocities), for relativistic mechanics the expressions should be transformed appropriately for the Langrangian to be valid there too. Lagrangian is also valid in general relativity, that is, in curved space, with still more modifications and transformations like Levi-Chivita connections, covariant derivatives; metric, Riemann, and Ricci tensors, and so on. Lagrangian didn't fit well in quantum mechanics but Hamiltonian did. The problem with the Lagrangian is that it subsumes a preservation of mass. For example, $$\partial L / \partial \dot{\mathbf{q}}$$ is equivalent to the Newtonian momentum $$\mathbf{p} = m \mathbf{v}$$ where only the the velocity $$\mathbf{v}$$ is supposed to be changing. In relativity and quantum physics, this is no longer true: mass may become energy and vice versa. However, the energy conservation laws are valid (at least for special relativity and quantum mechanics) so the Hamiltonian as the expression of the total energy of the system should be preserved there. Lantonov (talk) 13:33, 27 October 2020 (UTC)


 * (Yes, I meant Lagrangian, sorry about that, fixing it now) And yes, there was a discussion about the issue elsewhere (another editor's user page) and it got fixed. I've never seen these relativistic L and H used in any real problems, and I'm only talking about the nonrelativistic ones here. While the Lagrangian may not be relativistically invariant, you still have the liberty of choosing the gauge for your As and φs. Normally you'd be told what E and B are (because they can be measured) and you'd need to know which gauge to use for the potentials in the expressions for L and H. Ponor (talk) 14:04, 27 October 2020 (UTC)

The meaning of "off-shell".
In the section Hamiltonian mechanics, it says "Since this calculation was done off-shell, one can associate corresponding terms from both sides of this equation to yield:". You added a "clarification needed" template to this. Off-shell simply means that the system may be following a path which is not the one required by classical physics, i.e. not the one which extremalizes the action. It does not have to have anything to do with quantum mechanics. JRSpriggs (talk) 17:59, 26 October 2020 (UTC)
 * OK, I realize this, but wouldn't it be simpler and clearer to just say that the system may not be following a path which extremalizes the action. When the reader is sent to On shell and off shell, it immediately reads that "In physics, particularly in quantum field theory, configurations ... ", so on-shell and off-shell are used particularly (mostly) in quantum field theory. Further along the introduction it becomes clear that all the most important classical laws, such as the action formulation, extremal solutions to the variational principle, Euler-Lagrange equations, Noether's theorem, and conservation laws are on-shell. On the other hand, virtual particles, which are a quantum system, are off-shell. So, the term "off-shell" seems somewhat out-of-place when used in an exposition that is strictly limited to classical mechanics. Also, it is not clear why the calculation should be off-shell in order to associate the corresponding terms from both sides of the equation. These terms can be equated just by the equal sign in the equation, regardless of how this equation is calculated. Lantonov (talk) 21:12, 26 October 2020 (UTC)
 * Maybe this A DEEPER DIVE: ON-SHELL AND OFF-SHELL explains better "on the mass shell" and "off the mass shell" with the example of virtual particles on Feynman diagrams. For a person who just wants to learn what a Hamiltonian is, going to Feynman diagrams to explain it would be too much. Lantonov (talk) 09:55, 27 October 2020 (UTC)


 * In order to perform the mathematical step to get
 * $$\frac{\partial \mathcal{H}}{\partial q^i} = - \frac{\partial \mathcal{L}}{\partial q^i}$$
 * we need to set $$d q^i$$ to a non-zero value (so that we can divide it out) while simultaneously setting $$d p_i = d t = 0$$. This means going off-shell because this path is not (in general) classically permitted. Similarly for the other two equations. This would make our assumptions contradictory and the result invalid, if we had already assumed that we were on-shell. JRSpriggs (talk) 15:48, 27 October 2020 (UTC)
 * This is explained clearer and in more detail, e.g., in Landau & Lifshitz, vol. 1, section 43, without referring to mass shell. Lantonov (talk) 18:37, 27 October 2020 (UTC)

Notation
Another, smaller, problem (more like nit-picking) is that the Lagrangian in most popular textbooks and in the related Wiki articles, such as Lagrangian mechanics, Maupertuis's principle, Principle of least action, Calculus of variations, Hamilton's principle, etc., is denoted just $$L$$ while in this article it is denoted $$\mathcal{L}$$. In the same way, $$H$$ is more popular than $$\mathcal{H}$$ (e.g. Schrödinger equation). Lantonov (talk) 07:23, 27 October 2020 (UTC)

How many Hamilton's equations?
Another incongruence:

In the overview it is written that Hamilton's equations are

$$ \frac{\mathrm{d}\boldsymbol{p}}{\mathrm{d}t} = -\frac{\partial \mathcal{H}}{\partial \boldsymbol{q}}\quad,\quad \frac{\mathrm{d}\boldsymbol{q}}{\mathrm{d}t} = +\frac{\partial \mathcal{H}}{\partial \boldsymbol{p}} $$

referring to a relative obscure textbook of Hand and Finch.

On the other hand when deriving Hamilton's equations later, they are


 * $$\frac{\partial \mathcal{H}}{\partial q^j} =- \dot{p}_j \quad, \quad \frac{\partial \mathcal{H}}{\partial p_j} = \dot{q}^

j \quad, \quad \frac{\partial \mathcal{H}}{\partial t } = - {\partial \mathcal{L} \over \partial t}$$

In the beginning we have two Hamilton's equations, and after derivation they become three. Which is true? I think that two is the right number, at least my textbook (Landau & Lifshitz) says so. Lantonov (talk) 08:44, 27 October 2020 (UTC)


 * Typically, the Lagrangian and thus the Hamiltonian are not given as explicit functions of t. So the third equation is implicit, 0 = 0. Here we are dealing with the more general situation where dependence on t is allowed. For example, if the electromagnetic background field (not a part of the system being modeled) varies over time. JRSpriggs (talk) 15:58, 27 October 2020 (UTC)


 * Agreed for the present case. Which is defining separately the Hamiltonian function and the Hamiltonian equations. A better way would be both to be derived at once from the total variation of the action, and then $$dS / dt = L$$ and $$\partial S / \partial t = -H$$. However, when equating partial time derivatives of Lagrangian and Hamiltonian, we should specify that they depend on different parameters: $$L(\dot{q}_i,q_i)$$ and $$H(p_i,q_i)$$. The total derivatives are:
 * $$dL = \sum{\dot{p_i} dq_i} + \sum{p_i d \dot{q}_i} + \frac{\partial L}{\partial t} dt$$
 * $$dH = -\sum{\dot{p_i} dq_i} + \sum{\dot{q}_i dp_i} - \frac{\partial L}{\partial t} dt$$
 * When we compare the partial time derivatives, this should be written like:
 * $$ \left ( \frac{\partial H}{\partial t} \right )_{p,q} = -\left ( \frac{\partial L}{\partial t} \right )_{\dot{q},q}$$
 * with the indices showing that the differentiation by t should be done on the left side with constant p and q and on the right side with constant $$\dot{q}$$ and q. Lantonov (talk) 18:21, 27 October 2020 (UTC)
 * The first set of equations are standard. The second set uses dot notation to signify time differentiation and is equivalent. Lantonov is correct about the constancy of p and q in the partial derivatives, but this is not how we convert from Lagrangians to Hamiltonians, so it is not very important. Like the article says, you convert by finding an expression for the variables needed for either of the two models. IMO, we can just remove the partial H partial t equals negative partial L partial t bit, it serves only to confuse those who otherwise don't understand. Footlessmouse (talk) 21:33, 27 October 2020 (UTC)
 * Agree with Footlessmouse to remove the third equation. It does not belong to the Hamilton's equations not only because it's confusing but also because it defeats the main application of these equations which are the basis of the Canonical transformation and from then on to Hamilton-Jacobi equation and Liouville's theorem (Hamiltonian). All these use only two equations. The third equation is odd also because it alone includes Lagrangian. Hamilton's equations are all about the Hamiltonian and its parameters p and q. About converting from Lagrangian to Hamiltonian, the most direct way (though not necessary the easiest to understand) is through Legendre transformation. This is given in the article but very shortly with intermediate calculation omitted at the second and the third step. The reader is sent to Legendre transformation where they (it?) finds a rigourous mathematical exposition of the subject which excludes Lagrangian and Hamiltonian. The overall definition of the Hamiltonian is confusing (I pointed the most confusing place) because it does not capture the main ideas behind it, such as avoiding constancy of mass ($$\mathbf{p}$$ instead of $$\dot{\mathbf{q}}$$), as a total energy (H = E) its use in wider field, not just classical mechanics, its connection to symmetry, that is amenability to abstraction, through canonical transformations, Poisson brackets, Jacobi identity, Lie groups. If the difference between Lagrangian and Hamiltonian is properly explained in the derivation itself, like the difference between Hamilton's principle and Maupertuis's principle and their role in defining Lagrangian and Hamiltonian, respectively, a lot of confusion will be avoided. BTW, a good attempt to explain the Hamiltonian is done in Hamilton-Jacobi equation. Lantonov (talk) 08:27, 28 October 2020 (UTC).

Relativity requires the more general form in which dependence on t is explicit because you must have "background" fields or constraining objects UNLESS you go whole-hog to a generally covariant theory of everything (which does not exist at present). In other words, anything done using the Hamiltonian formalism must (for the time being) be an approximation based on non-relativistic assumptions. If q0 = t, then p0 = &minus; H. JRSpriggs (talk) 20:38, 28 October 2020 (UTC)

A new subsection Applications in machine learning
So called Energy based model are applied in machine learning, essentially usage of Hamiltonian mechanics. And redirect the page Energy based models to this subsection. --mcyp (talk) 18:39, 5 February 2022 (UTC)

New section on Hamiltonian as the total system energy
I have created a new section to detail the case of $$\mathcal{H} = T + V$$ which is taught in engineering and science courses. This follows on from the section to give more precision and detail, including proofs and citations.

The section also somewhat relates to, which has been getting quite a few edits lately.

Are there are any comments about this new section, or anything that anyone would like me to address? DrBrandonJohns (talk) 02:00, 22 February 2024 (UTC)