ECM: Ene gy-Based Recons uc ion o Binding
Con ibu ions in Unknown Gas Mix u es
Abs ac
Mos gas-mix u e expe imen s e eal only wo mac oscopic quan i ies: he ex e -
nally injec ed ene gy Ein and he ou going ene gy Eou . All mic oscopic in e nal e ec s
collapse in o he measu able esidual:
E es =Ein −Eou .
This pape de elops he Ene gy Composi ion Me hod (ECM), a gene al o malism
o econs uc ing o es ima ing in insic binding con ibu ions using only his esid-
ual. Because in e nal s uc u e is ne e di ec ly obse able, ECM classi ies all possible
in o ma ion s a es (Case 0–4) and de e mines which pa s o he composi e esidual
a e eco e able, pa ially iden i iable, o undamen ally insepa able.
The expanded o mula ion demons a es how ECM ex ac s he maximum possible
in o ma ion allowed by physics, wi hou any molecula -le el access, spec oscopy, o
species iden i ica ion.
1 In oduc ion
De e mining binding con ibu ions wi hin a gas mix u e is an indi ec p oblem: mic oscopic
in e ac ions canno be isola ed expe imen ally. Ins ead hey mani es only h ough he
mac oscopic ene gy di e ence
E es =Ein −Eou .
Howe e , his esidual is a composi e quan i y o igina ing om mul iple in e nal sou ces:
E es =Ebind +Ec oss +Eloss,
whe e Ebind e lec s in insic species-le el binding, Ec oss e lec s c oss-species in luence, and
Eloss cap u es equipmen dissipa ion.
1
The cen al challenge is iden i iabili y: Wha in e nal quan i ies lea e dis inguishable
oo p in s on E es? Unde which condi ions can binding be econs uc ed, and when is only
pa ial es ima ion possible?
ECM es ablishes a comple e classi ica ion (Cases 0–4) desc ibing e e y logical in o ma ion
s a e and shows when econs uc ion is exac , app oxima e, o non-unique.
2 No a ion
Symbol Meaning
Ein Injec ed ene gy
Eou Released ene gy
E es =Ein −Eou Ene gy esidual
Ebind In insic binding con ibu ion
Ec oss C oss-species in luence
Eloss Equipmen dissipa ion
xiMole ac ion o species i
BiBinding coe icien o species i
nNumbe o species in he mix u e
3 F amewo k: Residual Decomposi ion
The cen al s uc u al decomposi ion is:
E es =Ebind +Ec oss +Eloss.
I composi ion is known, binding is a linea mix u e:
Ebind =
n
X
i=1
xiBi.
I composi ion is unknown, o i nis unknown, he p oblem becomes a mix u e o pa am-
e e iden i ica ion and in e ence. ECM p o ides sys ema ic econs uc ion ules o e e y
in o ma ion s a e.
2
4 Case-Based Recons uc ion F amewo k
This sec ion p o ides a comple e, s uc u ed econs uc ion logic o all in o ma ion s a es— om
ully known sys ems o ully unknown mix u e s uc u es. ECM ea s each Case as a dis inc
iden i iabili y egime. The goal is always he same:
Ebind om E es =Ein −Eou .
Because he obse able dimension is only one scala (E es), he in e nal eco e abili y
depends en i ely on how much p io in o ma ion is a ailable.
The i e Cases below o m he ull ECM econs uc ion cha .
4.1 Case 0: Single-Species Mix u e (Baseline Iden i iable Case)
This is he only egime in which in e nal binding can be eco e ed exac ly.
Ec oss = 0.
Residual is
E es =Ebind +Eloss.
I he equipmen loss is calib a ed:
Ebind =E es −Eloss.
Why his Case is special. The e a e no mul iplica i e unknowns and no s uc u al
ambigui y. No composi ion, no species in e ac ion, no hidden dimensions. Thus, ECM
eco e s Ebind uniquely.
4.2 Case 1: Mul i-Species Mix u e, Composi ion Known
Gi en: - Numbe o species nis known. - Composi ion xiis known. - Binding coe icien s
Bia e unknown.
Then
Ebind =
n
X
i=1
xiBi.
Residual:
E es =
n
X
i=1
xiBi+Ec oss +Eloss.
3
Iden i iabili y. Since xiis known, he unknowns a e linea in Bi. Mul iple expe imen al
condi ions (kcondi ions) p o ide a sol able sys em:
E(k)
es −E(k)
loss =X
i
xiBi+E(k)
c oss.
C oss e ms canno be di ec ly sepa a ed, bu hey a e bounded and pa ially cancel
ac oss condi ions. Thus ECM p oduces a low-unce ain y es ima e o Bi.
4.3 Case 2: Species Coun Known, Composi ion Unknown
Unknowns: - xi(composi ion) - Bi(binding coe icien s)
Residual:
E(k)
es =
n
X
i=1
xiB(k)
i+E(k)
c oss +E(k)
loss.
Why his Case is di icul . The e m PixiB(k)
icon ains mul iplica i e unknowns
(xiBi). A single expe imen canno sepa a e hem.
Requi emen . ECM equi es a se o Kindependen expe imen al condi ions:
- di e ing empe a u e, - di e ing inpu in ensi y, - di e ing p ocess iming,
which p oduces a ia ion in B(k)
i.
Then he sys em becomes:
E es =A(x1, x2, ..., xn)T+ε
wi h ank(A)≥n equi ed.
Ou come. Exac eco e y is impossible, bu ECM p oduces s able es ima es o bo h
xiand Bi.
4.4 Case 3: Species Coun Unknown
He e e en he dimensionali y o he sys em is hidden.
Residual se :
{E(1)
es , . . . , E(K)
es }
con ains pa e ns caused by di e en species.
Goal: in e n.
ECM uses s uc u al ea u es:
1. Dis inc esponse modes o E es ( empe a u e slopes, p essu e cu a u e) 2. Sin-
gula alue decay (elbow de ec ion) 3. S abili y ac oss epea ed condi ions
4
Once ˆnis in e ed, he sys em educes o Case 2 and he usual es ima ion applies.
Ou come. No exac , bu s uc u ally consis en es ima ion o xi,Bi, and ˆnbe-
comes possible.
4.5 Case 4: Fully Unknown Sys em
No species coun , no composi ion, no coe icien s, no c oss/loss models.
Residual:
Ees
bind =E es −b
Ec oss −b
Eloss.
Es ima ion ools used:
- non-nega i i y cons ain s
Ebind ≥0, xi≥0
- smoo hing ac oss condi ions - physically easonable mono onici y - consis ency il e s -
minimizing esidual misma ch
Ou come. Case 4 does no allow exac eco e y, bu ECM s ill yields he closes
possible es ima e pe mi ed by mac oscopic in o ma ion heo y.
4.6 Es ima ion P inciple (applies especially o Case 2–4)
ECM es ima ion sol es:
min
xi,Bi,Ec oss,Eloss X
k E(k)
es −X
i
xiB(k)
i−E(k)
c oss −E(k)
loss!2
wi h cons ain s:
xi≥0,X
i
xi= 1, Ebind, Ec oss, Eloss ≥0.
This ensu es:
- physically alid solu ion, - ma hema ically consis en es ima e, - lowes -e o econ-
s uc ion.
This p inciple is wha makes ECM a alid in e en ial me hod e en when comple e iden-
i iabili y is impossible.
5
Es ima ion P inciple and Theo e ical Jus i ica ion
Whene e he in e nal deg ees o eedom exceed he dimension o he obse able esid-
ual, exac econs uc ion becomes ma hema ically impossible. This ollows om he basic
iden i iabili y condi ion o linea algeb a:
ank(A)<numbe o unknowns ⇒no unique solu ion.
In hese egimes (Cases 2–4), ECM does no pe o m heu is ic guessing. Ins ead, i applies
a cons ained in e se-p oblem o mula ion ha selec s he physically admissible solu ion
consis en wi h all obse a ions.
The esidual sa is ies
E(k)
es =X
i
xiB(k)
i+E(k)
c oss +E(k)
loss.
Thus he es ima ion s ep sol es he cons ained op imiza ion p oblem:
min
xi,Bi, Ec oss, Eloss X
k E(k)
es −X
i
xiB(k)
i−E(k)
c oss −E(k)
loss!2
,
subjec o he physically equi ed cons ain s:
xi≥0,X
i
xi= 1, Ebind, Ec oss, Eloss ≥0.
These cons ain s ensu e ha ECM chooses solu ions ha a e bo h ma hema ically con-
sis en and physically meaning ul. The me hod is he e o e no a heu is ic bu a s anda d
applica ion o in e se-p oblem heo y and cons ained leas -squa es es ima ion.
As a esul , e en when exac decomposi ion is impossible, ECM p oduces he closes
consis en es ima e pe mi ed by mac oscopic in o ma ion, achie ing heo e ical soundness
ac oss all pa ial-in o ma ion egimes.
5 Discussion
I is impo an o emphasize ha he es ima ion s eps in ECM a e no heu is ic. They a e
dic a ed by he ma hema ical s uc u e o he esidual equa ion and ollow s anda d p inci-
ples om linea algeb a, in e se p oblem heo y, and cons ained op imiza ion. Whene e
he numbe o in e nal unknowns exceeds he dimension o obse able ene gy measu emen s,
exac econs uc ion becomes ma hema ically impossible. In such egimes, ECM pe o ms
he heo e ically co ec ac ion: i selec s he physically admissible solu ion ha minimizes
6
esidual inconsis ency unde non-nega i i y and composi ional cons ain s. The e o e, all
es ima ion esul s a e g ounded in es ablished iden i iabili y heo y a he han empi ical
guessing.
ECM p o ides a comple e classi ica ion o wha can and canno be eco e ed om ene gy-
only measu emen s. Recons uc ion hinges no on p ecision,
bu on s uc u al iden i iabili y. E en pe ec measu emen canno e eal mo e in o ma-
ion han he s uc u e allows.
Case 0 and Case 1 pe mi exac o nea -exac eco e y. Case 2–4 allow only pa ial
eco e y, depending on dimensionali y.
6 Conclusion
ECM shows ha ene gy esiduals encode signi ican ly mo e in o ma ion han p e iously
ecognized. The amewo k iden i ies p ecise econs uc ion limi s, cla i ies when binding is
eco e able, and p o ides es ima ion me hods when i is no .
Fu u e ex ensions include: empe a u e-dependen binding, nonlinea c oss-in e ac ions,
and dynamic ( ime- esol ed) econs uc ion.
Appendix: Nume ical Examples
Example 1: Known Composi ion
Gi en: x1= 0.7, x2= 0.3, B1= 2.0, B2= 5.0, Eloss = 1.0.
Then
Ebind = 0.7(2.0) + 0.3(5.0) = 2.9,
E es = 2.9+1.0=3.9.
Example 2: Unknown Composi ion, Known n= 2
Expe imen s:
E(1)
es = 4.0, E(2)
es = 5.2.
F om induced a ia ions, linea sys em econs uc s x1B1and x2B2sepa a ely.
7
Example 3: Case 4 Es ima ion
Measu ed:
E es = 7.4.
Es ima ed:
b
Ec oss = 1.2,b
Eloss = 0.9.
Thus:
b
Ebind = 7.4−(1.2+0.9) = 5.3.
Re e ences
[1] R. B. Bi d, W. E. S ewa , and E. N. Ligh oo , T anspo Phenomena, 2nd ed., Wiley
(2007).
[2] P. A kins and J. de Paula, Physical Chemis y, 10 h ed., Ox o d Uni e si y P ess (2014).
[3] H. B. Callen, The modynamics and an In oduc ion o The mos a is ics, 2nd ed., Wiley
(1985).
[4] G. S ang, Linea Algeb a and I s Applica ions, 4 h ed., B ooks/Cole (2006).
[5] P. C. Hansen, Rank-De icien and Disc e e Ill-Posed P oblems, SIAM Monog aphs on
Ma hema ical Modeling and Compu a ion (1998).
[6] A. Ta an ola, In e se P oblem Theo y and Me hods o Model Pa ame e Es ima ion,
SIAM (2005).
[7] D. J. C. MacKay, In o ma ion Theo y, In e ence, and Lea ning Algo i hms, Camb idge
Uni e si y P ess (2003).
[8] R. A. Ho n and C. R. Johnson, Ma ix Analysis, Camb idge Uni e si y P ess (1985).
[9] J. Nocedal and S. W igh , Nume ical Op imiza ion, Sp inge (2006).
[10] S. Boyd and L. Vandenbe ghe, Con ex Op imiza ion, Camb idge Uni e si y P ess
(2004).
[11] E. K eyszig, Ad anced Enginee ing Ma hema ics, 10 h ed., Wiley (2011).
[12] A. V. Oppenheim and A. S. Willsky, Signals and Sys ems, P en ice Hall (1997).
8
[13] P. R. Be ing on and D. K. Robinson, Da a Reduc ion and E o Analysis o he Physical
Sciences, McG aw-Hill (2003).
9
