objective: Objective Functions

Description Usage Arguments Details References See Also Examples

Description

Calculate statistical and thermodynamic quantities for activities of species. These functions can be specified as objectives in revisit and findit in order to identify optimal chemical conditions.

Usage

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
  SD(a1)
  CV(a1)
  shannon(a1)
  DGmix(loga1)
  qqr(loga1)
  logact(loga1, loga2)
  spearman(loga1, loga2)
  pearson(loga1, loga2)
  RMSD(loga1, loga2)
  CVRMSD(loga1, loga2)
  DDGmix(loga1, loga2)
  DGinf(a1, a2)
  DGtr(loga1, loga2, Astar)

Arguments

a1

numeric matrix, chemical activities of species

loga1

numeric matrix, logarithms of activity

loga2

numeric, reference values of logarithms of activity

a2

numeric, reference values of activity

Astar

numeric, reference values of chemical affinity

Details

The value in a1 or loga1 is a matrix of chemical activities or logarithms of activity with a column for each species, and a row for each chemical condition. Except for calculations of the Shannon entropy, all logarithmic bases (including in the equations below) are decimal.

SD, CV and shannon calculate the standard deviation, coefficient of variation, and Shannon entropy for the values in each row of a1. The Shannon entropy is calculated from the fractional abundances: H = sum(-p * log(p)) (natural logarithm), where p=a1/sum(a1).

DGmix calculates the Gibbs energy/2.303RT of ideal mixing from pure components corresponding to one molal (unit activity) solutions: DGmix/2.303RT = sum(a1 * loga1) (cf. Eq. 7.20 of Anderson, 2005).

qqr calculates the correlation coefficient on a quantile-quantile (Q-Q) plot (see qqnorm) for each row of loga1, giving some indication of the resemblance of the chemical activities to a log-normal distribution.

logact returns the logarithm of activity of a single species identified by index in loga2 (which of the species in the system).

spearman, pearson, RMSD and CVRMSD calculate Spearman's rank correlation coefficient, the Pearson correlation coefficient, the root mean sqaured deviation (RMSD) and the coefficient of variation of the RMSD between each row of loga1 and the values in loga2. The CVRMSD is computed as the RMSD divided by the mean of the values in loga1.

DDGmix calculates the difference in Gibbs energy/2.303RT of ideal mixing between the assemblages with logarithms of activity loga1 and loga2.

DGinf calculates the difference in Gibbs energy/2.303RT attributed to relative informatic entropy between an initial assemblage with activities a2 and final assemblage(s) with activities with activities in each row of a1. The equation used is DGinf/2.303RT = sum(p2 * (logp2 - logp1)), where p1 and p2 are the proportions, i.e. p1 = a1 / sum(a1) and p2 = a2 / sum(a2). This equation has the form of the Kullback-Leibler divergence, sometimes known as relative entropy (Ludovisi and Taticchi, 2006). In specific cases (systems where formulas of species are normalized by the balancing coefficients), the values of DGinf and DGtr are equal.

DGtr calculates the change in Gibbs energy/2.303RT of a system in which species with initial logarithms of activitiy (loga1) are transformed to the same species with different final logarithms of activity (loga2) at constant temperature, pressure and chemical potentials of basis species. It is calculated as the sum over species of (G2-G1) where G1/RT = -a1*Astar + a1*loga1 - a1 + a constant (where a1 is 10^loga1), likewise for G2, and where Astar is the starved affinity, that is the affinity of the reaction to form one mole of the species at unit activity from the basis species in their defined activities. The equation used arises from integrating dG = -A/dxi = -A/dn where xi is the reaction progress variable, dn/dxi = 1 is the reaction coefficient on the species, and A = Astar - 2.303RTloga is the chemical affinity (Dick and Shock, 2013).

Each objective function has an attribute (see attributes and structure) named optimum that takes the value of minimal (SD, CV, RMSD, CVRMSD, DGmix, DDGmix, DGtr) or maximal (logact, shannon, qqr, spearman, pearson).

References

Anderson, G. M. (2005) Thermodynamics of Natural Systems, 2nd ed., Cambridge University Press, 648 p. http://www.worldcat.org/oclc/474880901

Dick, J. M. and Shock, E. L. (2013) A metastable equilibrium model for the relative abundance of microbial phyla in a hot spring. PLoS ONE 8, e72395. http://dx.doi.org/10.1371/journal.pone.0072395

Ludovisi, A. and Taticchi, M. I. (2006) Investigating beta diversity by Kullback-Leibler information measures. Ecological Modelling 192, 299–313. http://dx.doi.org/10.1016/j.ecolmodel.2005.05.022

See Also

revisit, findit

Examples

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
## a made-up system: 4 species, 1 condition
loga1 <- t(-4:-1)
loga2 <- loga1 + 1
stopifnot(qqr(loga1) < 1)
stopifnot(RMSD(loga1, loga1) == 0)
stopifnot(RMSD(loga1, loga2) == 1)
stopifnot(CVRMSD(loga1, loga2) == -0.4) # 1 / mean(-4:-1)
stopifnot(spearman(loga1, loga2) == 1)
stopifnot(spearman(loga1, rev(loga2)) == -1)
# less statistical, more thermodynamical...
stopifnot(all.equal(DGmix(loga1), -0.1234)) # as expected for decimal logarithms
stopifnot(all.equal(DDGmix(loga1, loga2), 0.0004))

## transforming an equilibrium assemblage of n-alkanes
basis(c("CH4", "H2"), c("gas", "gas"))
species(c("methane", "ethane", "propane", "n-butane"), "liq")
# calculate equilibrium assemblages over a range of logaH2
a <- affinity(H2=c(-10, -5, 101), exceed.Ttr=TRUE)
e <- equilibrate(a)
# take a reference equilibrium distribution at logfH2 = -7.5
loga1 <- list2array(e$loga.equil)[51, ]
Astar <- list2array(e$Astar)[51, ]
# equilibrium at any other logfH2 is not equilibrium at logfH2 = -7.5
DGtr.out <- DDGmix.out <- numeric()
for(i in 1:length(a$vals[[1]])) {
  loga2 <- list2array(e$loga.equil)[i, ]
  DGtr.out <- c(DGtr.out, DGtr(t(loga1), loga2, t(Astar)))
  DDGmix.out <- c(DDGmix.out, DDGmix(t(loga1), loga2))
}
# all(DGtr >= 0) is TRUE
stopifnot(all(DGtr.out >= 0))
# all(DDGmix >= 0) is FALSE
stopifnot(!all(DDGmix.out >= 0))
# a plot is also nice
thermo.plot.new(xlim=range(a$vals[[1]]), xlab=axis.label("H2"),
  ylim=range(DDGmix.out, DGtr.out), ylab="energy")
abline(h=0, lty=2)
abline(v=-7.5, lty=2)
text(-7.6, 2, "reference condition", srt=90)
lines(a$vals[[1]], DDGmix.out)
lines(a$vals[[1]], DGtr.out)
text(-6, 5.5, expr.property("DDGmix/2.303RT"))
text(-6, 2.3, expr.property("DGtr/2.303RT"))
title(main=paste("transformation between metastable equilibrium\n",
  "assemblages of n-alkanes"))
# take-home message: use DGtr to measure distance from equilibrium in 
# open-system transformations (constant T, P, chemical potentials of basis species)


Search within the CHNOSZ package
Search all R packages, documentation and source code

Questions? Problems? Suggestions? or email at ian@mutexlabs.com.

Please suggest features or report bugs with the GitHub issue tracker.

All documentation is copyright its authors; we didn't write any of that.