Re: CCL:Error estimation of gibbs free energy



Hi Guosheng,
 > From the top of my head, that formula comes from a Taylor expansion in
 which you neglect the higher-order terms.
 Honestly, the best info I have on error propagation are my hand written
 notes from the courses I followed during undergrad/grad studies.
 However, if you do a search on google with the following terms:
 general formula error propagation
 it gives lots of references.
 The most frequently mentioned reference I see is
 Taylor, J. R., "An introduction to Error Analysis", University Science
 Books, 1982.
 I also think that any book on numerical analysis should have at least a
 chapter that covers error propagation.
 See also:
 http://mulliken.chem.hope.edu/~polik/Chem345-2000/errorpropagation.htm
 Regards,
 Ioana
 ****************************************************************************
 * Ioana Cozmuta, PhD            *					   *
 * NASA-AMES Research Center     *  "Gravitation can not be held
 responsible*
 * Mail Stop 230-3               *  for people falling in love"		   *
 * Moffet Field,CA 94035         *  					   *
 * phone: (650) 604-0993         *                           Albert Einstein*
 * fax:   (650) 604-0350         *                             (1879-1955)  *
 ****************************************************************************
 On Thu, 15 May 2003 WU_GUOSHENG(-(at)-)Lilly.com wrote:
 > Dear Ioana,
 >
 > Could you please point out if there is any reference for your equation?
 >
 > I am sure there must be such kind of thing in some data analysis books,
 > but it's just inconvenient for me to find it at this moment.
 >
 > It seems to me the straightforwad way to get an error estimation is
 > by the basis calculus, like following (with similar presentation to
 > yours):
 >
 > for  f = f(x1, x2, ...),
 >
 > (del_f)= (df/dx1)*(del_x1) + (df/dx2)*(del_x2) + ...
 >
 > Then, (del_f)^2 can be the next form, if one has the interest:
 >
 > (del_f)^2= (df/dx1)^2 * (del_x1)^2 + (df/dx2)^2 * (del_x2)^2
 >          + 2*(df/dx1)* (del_x1) * (df/dx2) * (del_x2) + ...
 >
 > So for your example if f= x+y then (del_f) = (del_x)+(del_y)
 >
 > if f = x*y then simply
 >                 del_f = x * (del_y) + y * (del_x)
 >
 > If one like, (del_f)/f = (del_x)/x + (del_y)/y
 >
 >     or   [(del_f)/f]^2 = (del_x/x)^2 + (del_y/y)^2
 >                        + 2 * (del_x)*(del_y)/(x*y)
 >
 > Of course, for practical usage, one may overlook one or more terms if they
 > are much smaller than others.
 > - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
 > Thanks for your attention.
 >
 > Sincerely,
 >
 > Guosheng
 >
 > Guosheng Wu, Ph.D.
 > Eli Lilly & Company
 >
 >
 >
 >
 >
 > Ioana Cozmuta <ioana(-(at)-)nas.nasa.gov>
 > Sent by: Computational Chemistry List
 <chemistry-request(-(at)-)ccl.net>
 > 05/14/2003 01:32 PM
 >
 >
 >         To:     "'CHEMISTRY(-(at)-)ccl.net'"
 <CHEMISTRY(-(at)-)ccl.net>
 >         cc:
 >         Subject:        CCL:Error estimation of gibbs free energy
 >
 >
 > Hi,
 >
 > The most general formula from which you can always estimate the error
 > del_f of a function f(x1, x2, x3, ...xn) is
 >
 > (del_f)^2= (df/dx1)^2*del_x1^2+(df/dx2)^2*del_x2^2+...
 >
 > where df/dxi are partial derivatives of the function f with respect to xi
 > (i=1,n) and del_xi are the individual errors.
 >
 > So for example if f= x+y then (del_f)^2 = (del_x1)^2+(del_x2)^2
 > if f = x*y then (delf/f)^2= (del_x/x)^2+(del_y/y)^2
 >
 > If x1, ...xn are functions of other variables I find the general formula
 > above (with derivatives) much easier to use.
 >
 > Ioana
 >
 > On Wed, 14 May 2003, VITORGE Pierre 094605 wrote:
 >
 > > d(dG) = square root([d(dH)]^2 + [d(dS)]^2)
 > > cannot be true, if dh=0 and T is constant the above Eq. gives
 > > d(dG) = d(dS)
 > > while it is of course
 > > d(dG) = T d(dS)
 > >
 > > You probably mean:
 > > d(DG) = square root([d(DH)]^2 + [(Td(DS)+(DS)dT)]^2)?
 > > or
 > > d(DG) = square root([d(DH)]^2 + [Td(DS)]^2 + [(DS)dT]^2)?
 > >
 > > Pierre Vitorge
 > > CEA DEN Saclay DPC/SECR/LSRM & UMR 8587 (CEA-CNRS-Universite
 d'Evry)
 > > Universite d'Evry UMR 8587
 > > pierre.vitorge(-(at)-)cea.fr
 > > http://perso.club-internet.fr/vitorgen/pierre/pierre.html
 > >
 > >
 > > -----Message d'origine-----
 > > De: Dr. Richard L. Wood [mailto:rlw28(-(at)-)cornell.edu]
 > > Date: lundi 12 mai 2003 15:58
 > > À: Wong Lai Ho
 > > Cc: CHEMISTRY(-(at)-)ccl.net
 > > Objet: CCL:Error estimation of gibbs free energy
 > >
 > >
 > > Wouldn't the following be true
 > >
 > > d(dG) = square root([d(dH)]^2 + [d(dS)]^2)?
 > >
 > > dG is delta G, d(dG) is the error in delta G, dH is delta H, d(dH) is
 > the
 > > error
 > > in delta H, dS is delta S and d(dS) is the error in delta S.
 > >
 > > This is taken from a simple error propagation analysis as taught in
 some
 > > physical/analytical chemistry laboratories.
 > >
 > > Richard
 > >
 > > Wong Lai Ho wrote:
 > >
 > > > Dear CCLers,
 > > >
 > > > I have a question regarding error estimation of the experimental
 > values of
 > > > gibbs free energy.
 > > > As we knows that:
 > > > delta G(formation)
 > > >    = delta H(formation)
 > > >    - Temp * [entropy(molecule)-sum(entropy(atoms to form the
 > molecules))]
 > > >
 > > > Can I claim that the experimental error bar of the gibbs free
 > energy(delta
 > > > G) is equal to the error bar of the enthalpy change of
 formation(delta
 > H)?
 > > > (since the error introduced by the entropy is usually very small)
 > > > Is there any papers or books talking about this error?
 > > >
 > > > I would be grateful if you can give some comments on this claim.
 > > >
 > > > Larry
 > > >
 > >
 > > --
 > > Richard L. Wood, Ph. D.
 > > Physical/Computational Chemist
 > > Post-doctoral Associate
 > > Department of Chemistry
 > > Trinity University, San Antonio, TX 78212
 > >
 > >
 > >
 > >
 > >
 > >
 > >
 > >
 > >
 > >
 > >
 > >
 > >
 > >
 > >
 > >
 > >
 >
 >
 > -= This is automatically added to each message by mailing script =-
 > CHEMISTRY(-(at)-)ccl.net -- To Everybody  |
 CHEMISTRY-REQUEST(-(at)-)ccl.net -- To
 > Admins
 > Ftp: ftp.ccl.net  |  WWW: http://www.ccl.net/chemistry/   | Jan: jkl(-(at)-)ccl.net
 >
 >
 >
 >
 >
 >
 >
 >