Equilibrium Measurements of the NH3-CO2-H2O System ...publications.lib.chalmers.se/records/fulltext/250668/local_250668.pdf · ResearchArticle Equilibrium Measurements of the NH 3-CO

Research ArticleEquilibrium Measurements of the NH3-CO2-H2OSystem Speciation Based on Raman Spectroscopy andMultivariate Modeling

Maths Halstensen1 Henrik Jilvero2 Wathsala N Jinadasa1 and Klaus-J Jens1

1Applied Chemometrics Research Group (ACRG) University College of Southeast Norway Porsgrunn Norway2Department of Energy and Environment Chalmers University of Technology 412 96 Goteborg Sweden

Correspondence should be addressed to Maths Halstensen mathshalstensenusnno

Received 29 December 2016 Revised 14 March 2017 Accepted 4 April 2017 Published 11 May 2017

Academic Editor Adriano A Gomes

Copyright copy 2017 Maths Halstensen et al This is an open access article distributed under the Creative Commons AttributionLicense which permits unrestricted use distribution and reproduction in any medium provided the original work is properlycited

Liquid speciation is important for reliable process design and optimization of gas-liquid absorption process Liquid-phasespeciation methods are currently available although they involve tedious and time-consuming laboratory work Ramanspectroscopy is well suited for in situ monitoring of aqueous chemical reactions Here we report on the development of a methodfor speciation of the CO2-NH3-H2O equilibrium using Raman spectroscopy and PLS-RmodelingThe quantificationmethodologypresented here offers a novel approach to provide rapid and reliable predictions of the carbon distribution of the CO2-NH3-H2Osystem which may be used for process control and optimization Validation of the reported speciation method which is based onindependent knownNH3-CO2-H2Osolutions shows estimated prediction uncertainties for carbonate bicarbonate and carbamateof 645mmolkg H2O 3439mmolkg H2O and 1009mmolkg H2O respectively

1 Introduction

The rapid increase in the level of CO2 in the earthrsquos atmo-sphere is recognized as the single most important environ-mental challenge facing our global society [1] All climatechange mitigation plans rely on carbon dioxide capture andstorage as a near-term ldquoimmediate responserdquo technology [2]Despite the various global CO2 capture research and devel-opment initiatives the well-established gas-liquid absorptionprocess is expected to be the technology of choice forearly large-scale deployment [3] with the chilled ammoniaprocess (CAP) [4] being one of the currently demonstratedtechnologies

Out of several postcombustion techniques available tocapture CO2 from coal power plants amine solutions havebeen commonly tested and usedThe disadvantages of aminetechnology are that it requires large amount of energy in thestripping process and has thermal and oxidative degradationand corrosion problems [5] Ammonia technology is an alter-native to overcome these drawbacks This process requires

less energy for stripping and the heat of reaction is muchlower than amine processThere is lessmaintenance cost thanamine process as there are no degradation or corrosion issuesCO2 reacted ammonia can be used to produce fertilizerThe process of CO2 capture by ammonia can be twofolddepending on the temperature of CO2 absorption If theabsorption is performed under 2ndash10∘C which is called thechilled ammonia process there can be precipitations ofammonium carbonate compounds while if the absorption isincreased to 25ndash40∘C the precipitation problem is eliminated[6]

If chemical speciation data could be generated concomi-tant with the determination of the physical and chemicalsolvent properties the development and accuracy of thethermodynamic process model would be greatly facilitatedImprovement and optimization of commercial processesthat target needed cost reductions require access to rig-orous thermodynamic models that build on liquid-phasespeciation data Such data are currently acquired throughtedious and time-consuming laboratory work The standard

HindawiJournal of ChemistryVolume 2017 Article ID 7590506 13 pageshttpsdoiorg10115520177590506

2 Journal of Chemistry

wet chemistry titration to determine liquid phase is themost popular method This method takes time and errorscan be propagated during sampling chemical preparationweighing and titration [7] Various analytical techniques canbe used to determine the species present in CO2 reactedammonia solution 13C nuclear magnetic resonance (NMR)spectroscopy has been used [8] to determine ionic species ina range of ammonia concentration from 069 to 895molLand CO2 loading (total CO2 molesinitial solvent moles)concentrations from 033 to 072 Fourier transform infraredspectroscopy (FT-IR) and X-ray diffraction (XRD) wereused to qualitatively distinguish ammonium bicarbonatefrom ammonium carbonate and ammonium carbamatewhile CHN elemental analysis and near-infrared (NIR) spec-troscopywere used to quantify ammoniumbicarbonate basedon multivariate regression methods as reported in [9] Wenand Brooker [10] Zhao et al [11] andKim et al [12] suggestedmethods to determine carbon species in CO2-NH3-H2Osystem based on factor analysis where they assumed that thecorresponding Raman intensities were directly related to theconcentrations of species

This work reports on a method that combines Ramanspectroscopy and partial least-squares regression (PLS-R) forin situ solvent speciation in the chilled ammonia processand which does not rely on calibration by an independentanalysis method [13] for example NMR It describes thedevelopment validation and application of the methodfor determination of the liquid phase composition of theNH3-CO2-H2O system while comparisons of the speciationresults obtained here and those from established CO2-NH3-H2O thermodynamic equilibrium models are described in aseparate publication [14]

In the thermodynamic modeling of the CO2-NH3-H2Osystem a vapor-liquid equilibrium (VLE) is assumed to existfor water ammonia and CO2 The liquid phase can bedescribed by reactions (1)ndash(5) The chemical composition ofaqueous ammonia solution is described by

NH3 +H2O 999448999471 NH4+ +OHminus (1)

The reactions of dissolved CO2 in aqueous ammonia solutionare given by reactions (2)ndash(5) Within the present systemCO2 is bound as the anion species of carbonate [CO3

2minus]bicarbonate [HCO3

minus] and carbamate [NH2COOminus]

CO2(aq) +H2O 999448999471 H+ +HCO3minus (2)

HCO3minus 999448999471 H+ + CO32minus (3)

CO2(aq) + 2NH3 999448999471 NH4+ + NH2COOminus (4)

NH4+ +NH2COOminus +H2O999448999471 NH4+ +HCO3

minus +NH3 (5)

During CO2 absorption by the aqueous NH3 solutionreactions (1)ndash(5) will equilibrate according to the CO2 con-centration pressure and temperature the carbon distribu-tion in the solvent is given by the concentrations of theanions in reactions (2)ndash(5) In addition to these reactions

and depending on the reaction conditions various othercompounds may precipitate such as ammonium bicarbonate(NH4HCO3) ammonium carbonate [(NH4)2CO3] ammo-nium carbamate (NH4NH2CO2) and ammonium sesquicar-bonate [(NH4)2CO3sdotNH4HCO3] Even though the precipita-tion is promoted by low temperature the high water contentin the solvent reduces much of this possibility

Spectroscopy Raman spectroscopy in particular is wellknown for in situ monitoring of the chemical reactionsof aqueous solutions [16 17] The water molecule showsonly weak Raman scattering hence Raman spectroscopy haspotential advantage over IR spectroscopy [18] for aqueousphase analysis such as required for the aqueous chilledammonia CO2 capture solvent The Raman band envelopesof aqueous solutions of ammonium carbonate ammoniumbicarbonate and ammonium carbamate have been identifiedand analyzed [10] and form the basis for the previouslyreported speciation studies of the CO2-NH3-H2O system [1112] However evaluation of the Raman spectra in these pre-vious studies was based on univariate single-band analysisThe use of superior multivariate partial least-squares regres-sion (PLS-R) [19] methods which exploit the multivariateinformation in the spectra has to the best of our knowledgenot been reported previously Furthermore the methoddeveloped in the present study is not limited to laboratoryapplications but can also be used to monitor continuouslya reactive process which presents an opportunity for itsimplementation as an on-line process analytical technology(PAT) [20] in carbon capture plants for optimization andefficient operation

2 Materials and Methods

The development of the Raman spectroscopy-based methodfor speciation of the CO2-NH3-H2O system is based on PLS-R analysis of a series of samples of known composition Theprocedure involves

(1) preparation of aqueous solutions that containknown concentrations of [CO3

2minus] [HCO3minus] and

[NH2COOminus] one set of solutions was prepared forPLS-R model calibration and a second independentset of solutions was used for validation

(2) determination of the Raman spectrum of each samplesolution

(3) preprocessing of the Raman spectra by cropping eachspectrum to cover the range 450ndash2300 cmminus1 withsubsequent elimination of the inconsistently varyingbaseline from spectrum-to-spectrum and centeringof each wavelength by subtraction of the mean

(4) calibration of the PLS-R models for the anionicspecies [CO3

2minus] [HCO3minus] and [NH2COOminus] using

the Raman spectra of the solutions prepared formodel calibration in the previous step (Step (3))

(5) validation of the PLS-R models using the Ramanspectra of the solutions prepared formodel validationin Step (3)

Journal of Chemistry 3

x-variables

Sam

ples

n

x11 x12

x21 x22

middot middot middot

middot middot middot middot middot middot


middot middot middotmiddot middot middotxn1 xn2

p

W

P

T

X Y

X = TPT + E

Y = UQT + F

y11 y1q

y21 y2q

yn1 ynq

UQ

qy-variables

Figure 1 119883-119884 data and models in the PLS-R modeling approach [15] The parameters that emanate from PLS-R calibration include 119875 (119883loadings)119882 (119883 loading weights) 119879 (119883 scores) 119880 (119884 scores) and 119876 (119884 loadings) 119864 and 119865 are the residuals for119883 and 119884 respectively

The procedures required to achieve reliable speciation usingthe proposed method are described in the following sections

21 Raman Spectroscopy The Raman phenomenon is basedon quantized vibrational changes that are associated withelectromagnetic radiation absorption An important advan-tage of Raman spectroscopy over infrared spectroscopy isthat the water molecule shows very weak Raman scatteringwhereas in infrared spectroscopy the water molecule showsstrong absorption across an important part of the infraredspectrum

The Raman instrument used in the investigationsreported in the present paper was the RXN2 portablemultichannel Raman spectrometer (Kaiser Optical SystemsInc) Four fiber optic probes can be connected and utilizedthrough an automatic sequential scanning system that isintegrated into the instrument The specifications of theRXN2 Raman spectrometer are listed in Table 1

Raman spectra of aqueous solutions can be acquiredusing either a noncontact probe optic whereby the samplesolutions are not in direct contact with the probe optic oran immersion probe optic whereby the sample solution is indirect contact with the probe optic The Raman spectra wereacquired using a short-focus (200120583m) sapphire windowHastelloy immersion probe (Kaiser Optical Systems Inc)

22 Partial Least-Squares Regression Modeling PLS-R is anempirical data-driven modeling approach that requires bothrepresentative input data (119883) and output data (119884) A detaileddescription of PLS-R and validation can be found in literature[15 19 21] In this study PLS-R is used in combination withRaman spectroscopy The 119883 matrix contains Raman spectrathat represent different concentrations and the 119884 vectorcontains known reference concentrations from the sample

Table 1 Specifications of the RXN2 Raman spectrometer

Name DescriptionExcitation laserwavelength (nm) 785

Spectral range (cmminus1) 100ndash3425Spectral resolution(cmminus1) 4

Operating temperaturerange (∘C) 15ndash30

Number of channels 4Laser type Invictus NIR diode laser

Spectrograph f18 Holographic imagingspectrograph

Grating Holographic transmission gratingDetector TE cooled 1024 CCD DetectorMultichannel scanning 4-Channel sequential operationCal-Check Automatic analyzer monitor

Auto-Cal Automated calibration of axis and laserwavelength

Immersion probe optic 200 120583m (shortfocus)Hastelloysapphire window

preparation step All models reported in this article arevalidated based on independent test data (test set validation)[21]

The concentration range spanned by 119883 and 119884 shouldreflect the concentrations to be predicted The overall aimof PLS-R is to model simultaneously the multivariate inputdata (119883) and the output response (119884) (Figure 1) PLS-R avoidsmany of the problems associated with the traditionalmultiple


linear regression (MLR) and principal component regression(PCR) [15] methods PLS-R is advantageous in cases wherethe 119883 matrix contains collinear data Colinearity is in mostcases unavoidable in spectroscopy as there are significantlymore variables (waveshifts) than samples (observations)When the number of variables is significantly higher than thenumber of samples colinearity is guaranteedWhen collineardata are used MLR fails or becomes unstable AlthoughPCR can deal with colinearity it has other issues relatedto the way in which the 119883-matrix is decomposed withoututilizing information on 119884 often leading to models with ahigher number of components than the PLS-R case Anotheradvantage of applying PLS-R is that the feature parametersthat emerge from themodel calibration stage that is119883 scores(119879) 119883 loadings (119875) 119883 loading weights (119882) and 119884 scores(119880) can be plotted and interpreted to support the calibrationprocedure (see Figure 1)

An important aspect of PLS-R modeling is model vali-dation [21] which is important to determine the complexityof the model The correct complexity of a PLS-R model isdefined as the optimal number of PLS components which canonly be determined by proper validation of the model usingan independent dataset that is acquired and used exclusivelyfor this purpose To determine the optimal number ofcomponents in a PLS-R model a criterion based on the so-called root mean square error of prediction (RMSEP) [15] iscalculated and minimized RMSEP is based on predictions of119884 for example concentrations from the validation datasetUsing several models with different numbers of componentsthe predictions are compared to the reference values of 119884in the RMSEP calculation The model that has the optimalnumber of components is defined as the one that ends upwiththe lowest RMSEP valueTheRMSEP values are derived using(6)

RMSEP = radicsum119868119894=1 (ypredicted minus yreference)2119868 (6)

where ypredicted is the predicted value from the PLS-R modelwhich is compared to the reference value yreference The sumof the squared prediction errors is divided by 119868 which is thenumber of samples in the validation dataset

23 Preparation of the Calibration and Validation SamplesThe ranges of concentrations expressed in moles of eachanion per kilogram of H2O (molality) of the calibration andvalidation set were predetermined to reflect the concentra-tions expected in samples that would in the future be sub-jected to the developed speciation method The calibrationand validation sets covered the same concentration rangeAnalytical grade chemicals and Milli-Q water (182MΩsdotcm)were used to prepare the samples The ammonia solu-tion (25wt) was supplied by Merck KGaA (DarmstadtGermany) Sodium hydrogen bicarbonate (997) sodiumcarbonate (999) and ammonium carbamate (98) weresupplied by Sigma-Aldrich (Steinheim Germany) All chem-icals were used as received All solutions were prepared gravi-metrically using a Mettler Toledo balance (plusmn01mg) Forty

solutions of each of Na2CO3 NaHCO3 and NH4NH2CO2spanning the concentration ranges of 0ndash07molkg H2O0ndash096molkg H2O and 0ndash256molkg H2O respectivelywere prepared for calibration and validation purposes

The selection of concentration range for calibration andvalidation range is important especially when the modelis used for analyzing future samples Three factors wereconsidered during this selection which are the solubility ofthe chemicals expected species concentrations and Ramaninstrument performance As reported in [22] solubility ofNa2CO3 is 07 g at 0∘C and 125 g at 10∘C per kg H2ONaHCO3 solubility in water is 069 g at 0∘C and 0815 g at10∘C per kg H2O Ammonium carbamate is freely soluble inwater Holmes et al [8] report a comparison of equilibriummeasurements of ionic system in CO2-NH3-H2O systemsfor different initial concentrations of CO2 loading Thecomparison is based on his experimental work on 13C NMRmeasurements with three thermodynamic models of Pitzermodel [23] NRTL model [24] and TIDES model [25] Thiscomparison gives an indication of expected species concen-tration for a given CO2 loading and NH3 concentration Forthe demonstration of the proposed method in this studydifferent samples prepared using 5wt ammonia in the CO2loading range from 0 to 06mol CO2mol NH3 were usedand the expected species concentration reasonably falls in thecalibration and validation range based on the reported workby Holmes et al [8] The limitation of the Raman instrumentwas also considered when selecting the concentration rangeAn overview of the sample solutions used for the calibrationand validation of the PLS-R models including the respectiveconcentrations can be found in Appendix

24 Acquisition of Raman Spectra The Raman spectra weremeasured with the Kaiser RXN2 Raman spectrometer usinga laser power of 400mW and a total exposure time of 60seconds with six scans of 10 seconds each being applied toachieve a good signal-to-noise ratio Tomaintain a consistenttemperature in the spectrometer the instrument was stabi-lized for 30 minutes before each measurement series Theshort-focus immersion optic was fitted onto the fiber opticprobe head and cleaned with acetone The immersion probewas then positioned vertically using a stand with the opticalwindow facing down A glass container that contained thesample solution was positioned under the immersion opticwhich was then carefully immersed in the solution The tipof the optic was positioned in the center of the solutionapproximately 20mm from the bottom of the glass containerThe sample and probe optic were protected from externallight sources (such as fluorescent light) using aluminum foilThe Raman spectrum was obtained by initiating a scan inthe instrument software In the intervals between samplemeasurements the probe was cleaned in acetone to avoidcross-contamination of sample solutions

25 Preprocessing of Raman Spectra Figure 2 presents anexample of the preprocessing of the Raman spectra (cal-ibration and validation samples for bicarbonate) Duringpreprocessing all the Raman spectra were cropped so as tocover the range of 450ndash2300 cmminus1 since wavelengths outside


100 650 1200 1750 2300 2850 33250

21

Inte

nsity

(arb

itrar

y times105

unit)

Waveshift (cmminus1)

(a)

450 750 1050 1350 1650 1950 23000

525

Inte

nsity

(arb

itrar

y times104

unit)


(b)

Inte

nsity

(arb

itrar

y

450 750 1050 1350 1650 1950 23000

100005000

unit)


(c)

Inte

nsity

(arb

itrar

y

450 750 1050 1350 1650 1950 2300minus5000

50000


unit)

(d)

Figure 2 Preprocessing steps of the calibration and validation Raman spectra of bicarbonate (a) Raw spectra for the range of 100ndash3325 cmminus1(b) Spectra cropped to cover the range of 450ndash2300 cmminus1 (c) Spectra baseline corrected using theWhittaker filter (lambda = 100 119875 = 0001)(d) Mean-centered spectra for PLS-R model calibration

this range were not useful for the PLS-R analysis Figures2(a) and 2(b) show the bicarbonate spectra before and aftercropping respectively In this range the Raman spectra showan inconsistent baseline which is particularly evident at thehighest Raman shifts (above 2000 cmminus1) Raman spectra withvariable baselines often generate regression models that havea higher number of components than is necessary sincethe model also needs to model the baseline drift To ensurea less complex model a baseline correction is performedto decrease or remove the baseline drift in each Ramanspectrum Several baseline correction methods are available[26] Spectroscopic data from a Raman spectrometer canbe decomposed into three parts (1) the analytical signal(2) the baseline and (3) noise Noise was not a problemin the present study whereas the baseline required someattention The goal of all the baseline correction methods isto estimate the baseline in order to remove itTheMATLAB2012b software (MathWorks Inc) in combination with PLSToolbox 731 (Eigenvector Research Inc) was used to find asuitable baseline correction in the present case The lowestRMSEP was gained using the Whittaker filtering method[27] included in PLS Toolbox 731 which preserves theoriginal shape of the signal part of the Raman spectrumTheWhittaker filter was applied with the following parameterslambda = 100 and 119875 = 0001 The lambda parameterdefines how much curvature is allowed while the 119875 param-eter holds information about asymmetry in the spectraFigure 2(c) shows the Raman spectra of bicarbonate after

baseline correction using the Whittaker filter Finally thespectra were centered by subtracting the mean from eachvariable (waveshift) in the Raman spectra Centering [19] isapplied prior to PLS-R calibration to avoid the need for anadditional PLS-component to describe the mean of the datawhichwould entail amore complexmodel Figure 2(d) showsthe data after mean-centering To summarize the spectrawere cropped to lie within the range of 450ndash2300 cmminus1the baseline was corrected using the Whittaker filter withlambda value of 100 and 119875 value of 0001 and the spectrawere centered on the mean prior to calibration of the PLS-R models of the three species (carbonate bicarbonate andcarbamate)

26 PLS-R Modeling of Carbonate Bicarbonate and Carba-mate Individual models for carbonate bicarbonate andcarbamate were calibrated based on the obtained Ramanspectra of aqueous solutions that contained known concen-trations in the ranges of 0ndash07molkg 0ndash096molkg and0ndash256molkg respectively Since the aim was to developPLS-R models that could be used for speciation of a realCO2-NH3-H2O system a selection of variables (waveshifts)to be included in each model was carried out The waveshiftsto be included in the model of each respective anion werechosen based on backward selection whereby only thewavelengths related to ammonia and the other two CO2anion species were omitted In themodel for the prediction ofcarbonate the Ramanwavelengths associated with ammonia


Table 2 Vibrational assignments of (NH4)2CO3 NH4HCO3 andH2NCOONH4 aqueous solutions and omitted frequencies in the respectivePLS-R models Adapted from the data of Wen and Brooker [10]

Frequency [cmminus1] Assignment Omitted frequencies for the PLS-R anion modelCarbonate Bicarbonate Carbamate

3390 Antisymmetric N-H stretch of NH3 X X X3310 Symmetric N-H stretch of NH3 X X X3220 Fermi resonance with N-H symmetry stretch of NH3 X X X3050 Symmetric N-H stretch of NH4

+ X X X2850 A combination of fundamentals of NH4

+ X X X1690 NH2 deformation of NH4

+ X X X1645 Antisymmetric deformation of NH3 X1630 C-O antisymmetric stretch of HCO3

minus X1550 Antisymmetric CO2 stretch of H2NCOOminus X1436 1380 CO antisymmetric stretch of CO3

2minus X X X X X X1430 Antisymmetric NH2 deformation of NH4

+ X X X1405 Symmetric CO2 stretch of H2NCOOminus X X X1360 CO symmetric stretch of HCO3

minus X X1302 C-OH bend of HCO3

minus X X X1120 CN stretch of H2NCOOminus X X X1065 CO symmetric stretch of CO3

2minus X X1034 NH2 wag of H2NCOOminus X X1017 C-OH stretch of HCO3

minus X X X680 CO2 antisymmetric deformation of CO3

2minus X X X640 (OH)-CO bend of HCO3

minus X X X570 Torsion about CO2 skeleton of H2NCOOminus X X X

bicarbonate and carbamate were omitted The wavelengthranges used in the modeling of bicarbonate analysis wereselected based on the same principle while the carbamatemodel was based on a more limited range of wavelengthsSince PLS-R modeling of carbamate is more challengingthan modeling of the other two species forward selectionwas used to define the wavelength ranges in this modelFine-tuning of the wavelength selection was made based onprediction results in which the model with a combination ofvariables that resulted in the lowest prediction uncertaintywas used The carbonate bicarbonate and carbamatemodels were based on the respective frequency ranges of[450ndash520 750ndash950 1062ndash1100 1140ndash1200 1520ndash1650 and1760ndash2300] cmminus1 [450ndash520 750ndash950 1140ndash1200 1350ndash13901520ndash1650 and 1760ndash2300] cmminus1 and [1033ndash1043] cmminus1Table 2 lists the frequencies omitted from each PLS-R anionmodel

The PLS-R models for carbonate bicarbonate and car-bamate were all based on 20 calibration spectra in additionto the 20 independent spectra that were used exclusively forthe validation Figure 3 shows the spectra used to calibrateand validate the carbonate bicarbonate and carbamate pre-diction models

Dissolution of carbonate in water will lead to the follow-ing equilibrium state as given in reaction (7)

CO32minus +H2O 999448999471 OHminus +HCO3

minus (7)

However given the detection limit of the Raman spec-trometer only the carbonate band was observed Thereforethis reaction was neglected in the carbonate PLS-R modeldevelopment However in the cases of bicarbonate disso-lution reaction (8) and for carbamate dissolution reactions(7)ndash(13) were observed


CO32minus(aq) +NH4+(aq) 999448999471 HCO3

minus(aq) +NH3(aq) (9)

NH3 +H2O 999448999471 NH4+ +OHminus (10)

NH4+ 999448999471 H+ +NH3 (11)

HCO3minus(aq) +NH3(aq) 999448999471 H2NCOOminus(aq) +H2O (12)

H2NCOOminus(aq) +H2O 999448999471 CO32minus(aq) +NH4+(aq) (13)

Thus the quantitative PLS-R models were developed accord-ing to the following steps

(1) Calibrate and validate the carbonate model(2) Predict carbonate in the bicarbonate calibration and

validation datasets and correct the bicarbonate refer-ences accordingly

(3) Calibrate and validate the bicarbonate model basedon the corrected dataset obtained in Step (2)


450 750 1050 1350 1650 1950 23000

21

43

times104


Inte

nsity

(arb

itrar

y un

it)

(a)

450 750 1050 1350

1390

1650 1950 23000

5000

10000

Inte

nsity

(arb

itrar

y un

it)


(b)

450 750 1050 1350 1650 1950 23000

1

2

3times104


Inte

nsity

(arb

itrar

y un

it)

(c)

Figure 3 Raman spectra used in the PLS-R model calibration and validation (a) Raman spectra of carbonate (b) Raman spectra ofbicarbonate (c) Raman spectra of carbamate

(4) Apply the carbonate model from Step (1) and bicar-bonate model from Step (3) to predict carbonateand bicarbonate in the carbamate calibration andvalidation datasets

(5) Correct the carbamate references in the calibrationand validation datasets based on the predictions ofcarbonate and bicarbonate made in Step (4)

(6) Calibrate and validate the carbamate model

All the PLS-R models developed in the present study weredeveloped using theUnscrambler X ver 103 software Finallythe speciation method was demonstrated based on realisticaqueous solutions that contained 5wt ammonia loadedwith CO2 Since reference concentrations for carbonatebicarbonate and carbamate are not available for thesedatasets the prediction results were assessed based on thecalculated prediction uncertainties provided by UnscramblerX ver 103

3 Results and Discussion

The calibration and validation results for each of the respec-tive PLS-Rmodels for carbonate bicarbonate and carbamateare presented in the respective subsections below Outlierswere detected in the validation and calibration sets forcarbonate using the 1199051 minus 1199061 scatter plots (data not shown)An outlier may result from an air bubble sticking to thesapphire window in the tip of the probe optic or from the

probe optic being inserted so far down into the solution thatthe measurement is influenced by the glass container

31 PLS-R Validation Results for Carbonate Bicarbonate andCarbamate Figures 4ndash6 show the validation results for thePLS-R models of carbonate bicarbonate and carbamaterespectively The results include plots of the scores (1199051 minus 1199052)regression coefficients (119861) residual validation variances andpredicted concentrations versus the reference concentrationsThe score plots are used to visualize how the calibrationspectra compare to the validation spectra The regressioncoefficients show the weight that each wavelength is assignedin the predictionThe residual validation variance plot showsthe size of the residual for models with an increasing numberof components The predicted versus measured plots showhow the predicted concentrations from the validation datasetin comparisonwith the references calculated from the samplepreparation stage Prediction performance is evaluated froman interpretation of the statistical parameters of merit whichinclude 1199032 RMSEP and the slope of the regression line

For the carbonate series one outlier identified accordingto the definition provided previously [19] was removedfrom the validation dataset No outliers were detected in thecalibration datasetThe outlier was not considered thereafterFigure 4 shows selected plots from the calibration and valida-tion of the PLS-R model for carbonate In bicarbonate seriesno outliers were detected in the calibration and validationdatasets Figure 5 shows selected plots from the calibration


minus5 0 5PLS component number 1

minus4000

minus2000

0

2000

4000

PLS

com

pone

nt n

umbe

r 2

Calibration samplesValidation samples

times104

Score plot (t1 minus t2)

(a)

400 600 800 1000 1200 1400 1600 1800 2000 2200 2400minus1

0

1

2

3

4

B (a

rbitr

ary

unit)

Regression coefficients (B)times10minus6


(b)

0 1 2 3 4 5 6PLS component number

0

001

002

003

004

005

Y-v

aria

nce

Residual validation variance

(c)

0 01 02 03 04 05 06 07Measured carbonate concentration (molkg)

0

02

04

06

Refe

renc

e car

bona

te

conc

entr

atio

n (m

olk

g)

Predicted versus reference

SlopeOffsetCorrelation (r)

09995RMSEPr2

098980001309998

00043

(d)

Figure 4 PLS-R model for carbonate (a) Score plot of PLS components 1 versus 2 showing the calibration samples and validation samples(b) Regression coefficients based on a one-component PLS-R model (c) The residual validation variance shows that one PLS-componentis optimal (d) Predicted carbonate concentrations based on the one-component PLS-R model versus the reference concentrations obtainedfrom the solute ion preparation stage

Table 3 PLS-R modeling of carbonate bicarbonate and carbamate

Model Carbonate model Bicarbonate model Carbamate model

Wavelength rangesincluded [cmminus1]

450ndash520 750ndash9501062ndash1100 1140ndash1200

1520ndash1650 and 1760ndash2300


1520ndash1650 and 1760ndash23001033ndash1043

Slope 09898 09636 102461199032 09995 09932 09968RMSEP [mmolkg] 43 241 499Number ofcomponents 1 1 1

Number of outliers 1 (in the validation set) 0 0

and validation of the PLS-R model for bicarbonate Forcarbamate series also no outliers were detected in the cali-bration or validation datasets Figure 6 shows selected plotsfrom the calibration and validation of the PLS-R model forcarbamate

The wavelength ranges used in the PLS-R models forthe predictions of carbonate bicarbonate and carbamateare presented in Table 3 Some variables (waveshifts) withregression coefficients close to zero were also included as

it was found that the prediction uncertainties were slightlyimproved when these wavelengths were included The valuesfor the slope 1199032 RMSEP the number of PLS components andoutliers in all three models are listed in Table 3

Multivariate analysis has been proven to overcome manychallenges in univariatemethod Univariatemethod is simpleand samples for calibration can be prepared using one chem-ical when there are no interferences from other constituentsThere are many instances when the property of interest


minus6000 minus4000 minus2000 0 2000 4000 6000PLS component number 1

minus4000

minus2000

0

2000

4000

PLS

com

pone

nt n

umbe

r 2Score plot (t1 minus t2)


(a)

0 500 1000 1500 2000 2500minus5

0

5

10

15

20

B (a

rbitr

ary

unit)



(b)


0

002

004

006

008

Y-v

aria

nce


(c)

Reference bicarbonate concentration (molkg)

0

02

04

06

08

1

Pred

icte

d bi

carb

onat

e co

ncen

trat

ion

(mol

kg)


0 02 04 06 08 1


09932RMSEPr2

096360019109966

00241

(d)

Figure 5 PLS-R model for bicarbonate (a) Score plot of PLS components 1 versus 2 showing the calibration samples and validation samples(b) Regression coefficients based on a one-component PLS-R model (c) The residual validation variance shows that one PLS-component isoptimal (d) Predicted bicarbonate concentrations based on the one-component PLS-R model versus the reference concentrations obtainedfrom the solution preparation stage

cannot be described by one peak According to Table 2vibrational modes assigned to carbonate bicarbonate andcarbamate fall closer to each other in the finger print areafrom 1000 to 1420 cmminus1 The NH3-CO2-H2O system consistsof several equilibrium reactions and the compositions ofcarbonate bicarbonate and carbamate are influenced by theconcentrations of each other It can be seen from the calibra-tion spectrum that there are completely visible independentnicely shaped peak assigned to each other however this maybe not true when it comes to the real CO2 loaded ammoniasystem Two scenarios of misuse of the multivariate regres-sion in spectroscopic applications have been explained byEsbensen et al [28] They are assigning individual peaks forregressionwhich are identified by the preprocessed data or byregression coefficientsThe regression coefficients are used tocalculate the response value from the X-measurements Thesize of the coefficients gives an indication of which variableshave an important impact on the response variables Assign-ing wavelengths selected during calibration development forregression must be done with caution because more than onewavelengths are associated with the functional group to somedegree [28] Many factors including scatter effect affect the

wavelength position and only by using a wavelength regioncan the robustness of the calibration model be increased

The score plots shown in Figures 4ndash6 reveal that thecalibration and validation datasets in all three cases span thesame score space which indicates similarity of the datasetsThe most important wavelengths are those with regres-sion coefficients that show the largest deviation from zeroThe regression coefficients listed in Figure 4 (carbonate)Figure 5 (bicarbonate) and Figure 6 (carbamate) show thatthe most important wavelength ranges are 1060ndash1070 cmminus11350ndash1380 cmminus1 and 1033ndash1043 cmminus1 respectively In all threemodels only a small contribution is gained fromwavelengthsoutside these ranges The residual validation variance plotshows that one-component PLS-R is sufficient for all threemodels The slope of the regression line is 096ndash102 and the1199032 is 00993ndash0999 which is close to the optimal value of 10The average prediction errors that is RMSEP values for car-bonate bicarbonate and carbamate were 43mmolkg H2O241mmolkg H2O and 499mmolkg H2O respectively

32 Demonstration of the Method The proposed speciationmethod was demonstrated using aqueous solutions of 5 wtammonia loaded with CO2 Since reference concentrations


minus6 minus4 minus2 0 2 4 6minus2

minus15minus1

minus050

051

152

PLS component number 1

PLS

com

pone

nt n

umbe

r 2


times104

times104


(a)

400 600 800 1000 1200 1400 1600 1800 2000 2200 2400minus1

0

1

2

3

4

5

B (a

rbitr

ary

unit)



(b)

0 1 2 3 4 5 60

00501

01502

02503

035

PLS component number


Y-v

aria

nce

(c)

0 1 20

02

02

04

04

06

06

08

08

112

12

14

14

16

16

18

18

2

Reference carbamate concentration (molkg)

Pred

icte

d ca

rbam

ate

conc

entr

atio

n (m

olk

g)



09936RMSEPr2

10246minus0006109968

00499

(d)

Figure 6 PLS-R model for carbamate (a) Score plot of PLS components 1 versus 2 showing the calibration samples (filled circles) andvalidation samples (open circles) (b) Regression coefficients based on a one-component PLS-R model (c) The residual validation varianceshows that one PLS-component is optimal (d) Predicted carbamate concentrations based on the one-component PLS-R model versus thereference concentrations obtained from the solution preparation stage

of carbonate bicarbonate and carbamate were not availablethe prediction performance level of each model was assessedbased on the calculated prediction uncertainties as definedpreviously [29] The demonstration dataset was based on14 solutions with different loadings (mole-ratio CO2NH3)of CO2 which were measured three times (42 samples intotal) The solvent CO2 loading is one of the primary processparameters in the operation of systems for the chemicalabsorption of CO2

Figure 7 shows the predicted concentrations of the mea-sured species The average prediction uncertainties for car-bonate bicarbonate and carbamate were 645mmolkg H2O3439mmolkg H2O and 1009mmolkg H2O respectivelyOverall the predictions of the sample solutions with corre-sponding uncertainties give a satisfactory outcome regardingspeciation of all the anions The last three predictions forcarbamate shown in Figure 7(c) reveal greater uncertaintiesthan those noted for the other predictions This is due toeither the precipitation of solids or the presence of a slightlyhigher concentration of ammonia in this sample While aprecipitate was not visible in this sample the conditionswere close to those for which precipitation is expected Thusthe influence of precipitation could not be ruled out Inthe accompanying report [14] the method presented here is

compared to the experimental data of the same samples withthe precipitation-titration method with good agreementIn the present study the difference observed between thedemonstration dataset and the carbamate calibration is thateven though all the same species are present the CO2 loadingdiffers (see reactions (12) and (13)) In the demonstrationdataset the amounts of ammonia and water are roughlyconstant and the amount of CO2 increases continuously withincreasing sample number In the demonstration dataset theCO2 loading is increased from 0 to 06 Thus the presentmethod can relate the CO2 loading to the liquid carbondistribution through reactions (2)ndash(5)

33 Comparison of the Model with Literature Three modelsdeveloped in this study have been used in the study ofVLE data of chilled ammonia system [14] This study showsthe model predictability which has been compared withtwo thermodynamic models of Darde et al [6] and Queand Chen [24] and experimental work carried out by Wenand Brooker [10] Holmes et al [8] Zhao et al [11] andAhn et al [30] The composition analysis performed byZhao et al [11] for CO2-NH3-H2O system is based onunivariate analysis of Raman measurements They recordresults for different initial ammonia concentrations in the


0 3 6 9 12 15 18 21 24 27 30 33 36 39 420

02

01

04

03

Sample number

Pred

icte

d

conc

entr

atio

n(m

olk

g)

carb

onat

e

(a)

0 3 6 9 12 15 18 21 24 27 30 33 36 39 42minus05

0

05

1

Sample number

Pred

icte

d

conc

entr

atio

n(m

olk

g)

bica

rbon

ate

(b)

0 3 6 9 12 15 18 21 24 27 30 33 36 39 420

1

15

05

2

Sample number

Pred

icte

d

conc

entr

atio

n(m

olk

g)

carb

amat

e

(c)

Figure 7 Predicted concentrations [molkg] of (a) carbonate (b) bicarbonate and (c) carbamate

range of 069molL to 210molL and CO2 concentrationrange from 018 to 067molL The work is related to lowconcentrations of ammonia The proposed method in thisstudy is independent of ammonia concentration and is basedon three calibration sets each with 20 measurements whichwere followed by validation using independent data setspanning in the calibration range for each species Zhaorsquosmethod includes calculating molar scattering intensity (J)of each carbon species by preparation of series of solutionsof sodium carbonate and sodium bicarbonate to calculateJ of carbonate and bicarbonate while J of carbamate wascalculated by carbon conservation balance Therefore themolar scattering intensity of carbamate was dependent onthose of other 2 components

4 Conclusion

A method for speciation of the CO2-NH3-H2O system isproposed The proposed method can be applied without theneed for additional analytical calibrationmethods Speciationis achieved based on a combination of Raman spectroscopyand multivariate PLS-R modeling wherein the so-called fullspectrum calibration method is applied to extract informa-tion from the entire spectrumThe concentrations of carbon-ate bicarbonate and carbamate were predicted with an aver-age prediction error (RMSEP) as being 43mmolkg H2O241mmolkg H2O and 499mmolkg H2O respectively

For the method demonstration case which lacked refer-ence concentrations the prediction uncertainties for car-bonate bicarbonate and carbamate were 645mmolkgH2O 3439mmolkg H2O and 1009mmolkg H2O respec-tively

Appendix

Concentrations of the Carbonate Bicarbonateand Carbamate Species in the Solutions Usedfor Model Calibration and Validation

See Table 4

Conflicts of Interest

The authors declare no conflicts of interest regarding thepublication of this research paper

Acknowledgments

The financial support provided by the Norwegian ResearchCouncil and the industrial partners in the KMB project (no199905S60) and the PhD scholarship financed by the PhDprogramme in Process Energy and Automation at UniversityCollege in Southeast Norway are gratefully acknowledged


Table 4 Overview of sample solutions reported as molkg H2O

Sample number Carbonate [molkg H2O] Bicarbonate [molkg H2O] Carbamate [molkg H2O]Calibration Validation Calibration Validation Calibration Validation(1) 000000 001974 000000 003441 000000 007923(2) 003840 005585 005671 005704 014939 014652(3) 007280 009137 010508 010159 029518 027757(4) 011073 012401 015180 015504 039826 040627(5) 014710 015777 020941 020138 053542 055145(6) 018544 019187 025647 025214 067098 067585(7) 022126 022658 030250 030090 080054 080525(8) 025856 026310 035346 035263 093483 094422(9) 029493 029814 040131 040289 108200 105871(10) 033113 033291 046320 045270 121012 121201(11) 036895 036560 049901 049946 133941 133618(12) 039913 040255 054737 054932 147757 147498(13) 044205 043683 059693 060006 161910 160257(14) 047772 047227 065191 065196 173472 173898(15) 051599 050607 070127 070248 187207 187232(16) 055162 054010 074870 075143 202088 201094(17) 059048 057554 079937 079809 212369 213626(18) 062521 061002 084721 084682 227083 228208(19) 066210 064656 090173 089758 241909 241723(20) 069948 068035 095353 094859 255658 244528

References

[1] S Solomon D Qin M Manning et al ldquoClimate Change 2007The Physical Science Basisrdquo Contribution of Working GroupI to the Fourth Assessment Report of the IntergovernmentalPanel on Climate Change Cambridge University Press Cam-bridge UK 2007

[2] IEA Technology Roadmap Carbon Capture and Storage Inter-national Energy Agency Paris France 2013

[3] G T Rochelle ldquoamine scrubbing for CO2 capturerdquo Science vol325 no 5948 pp 1652ndash1654 2009

[4] G d Koeijer Y O Enge C Thebault S Berg J Lindland andS J Overa ldquoEuropean CO2 test centre mongstadndashtesting ver-ification and demonstration of post-combustion technologiesrdquoEnergy Procedia vol 1 no 1 pp 1321ndash1326 2009

[5] O Lawal A Bello and R Idem ldquoThe role of methyl dietha-nolamine (MDEA) in preventing the oxidative degradationof CO2 loaded and concentrated aqueous monoethanolamine(MEA)-MDEA blends during CO2 absorption from flue gasesrdquoIndustrial and Engineering Chemistry Research vol 44 no 6 pp1874ndash1896 2005

[6] V Darde K Thomsen W J M van Well and E H StenbyldquoChilled ammonia process for CO2 capturerdquo InternationalJournal of Greenhouse Gas Control vol 4 no 2 pp 131ndash1362010

[7] J T Yeh H W Pennline K P Resnik and K Rygle AbsorptionAnd Regeneration Studies for CO2 Capture by Aqueous Ammo-nia Pittsburgh PA USA 2004

[8] P E Holmes M Naaz and B E Poling ldquoIon concentrationsin the CO2-NH3-H2O system from 13C NMR spectroscopyrdquoIndustrial and Engineering Chemistry Research vol 37 no 8 pp3281ndash3287 1998

[9] L Meng S Burris H Bui and W-P Pan ldquoDevelopment of ananalytical method for distinguishing ammonium bicarbonatefrom the products of an aqueous ammonia CO2 scrubberrdquoAnalytical Chemistry vol 77 no 18 pp 5947ndash5952 2005

[10] N Wen and M H Brooker ldquoAmmonium carbonate ammo-nium bicarbonate and ammonium carbamate equilibria araman studyrdquo Journal of Physical Chemistry vol 99 no 1 pp359ndash368 1995

[11] Q Zhao S Wang F Qin and C Chen ldquoComposition analysisof CO2-NH3-H2O system based on Raman spectrardquo Industrialand Engineering Chemistry Research vol 50 no 9 pp 5316ndash5325 2011

[12] Y J Kim J K YouWHHong K B Yi CHKo and J-NKimldquoCharacteristics of CO2 absorption into aqueous ammoniardquoSeparation Science and Technology vol 43 no 4 pp 766ndash7772008

[13] P A G L Samarakoon N H Andersen C Perinu and K-JJens ldquoEquilibria of MEA DEA and AMP with bicarbonate andcarbamate a Raman studyrdquo Energy Procedia vol 37 pp 2002ndash2010 2013

[14] H Jilvero K-J Jens F Normann et al ldquoEquilibrium measure-ments of the NH3-CO2-H2O system - measurement and eval-uation of vapor-liquid equilibrium data at low temperaturesrdquoFluid Phase Equilibria vol 385 pp 237ndash247 2015

[15] K H Esbensen D Guyot F Westad and L P HoumollerMultivariate Data Analysis in Practice CAMO Software OsloNorway 2010

[16] U Lichtfers andB Rumpf ldquoInfrared spectroscopic studies of thedetermination of species concentrations in aqueous solutionscontaining ammonia and carbon dioxiderdquo Chemie IngenieurTechnik vol 72 pp 1526ndash1530 2000


[17] U Lichtfers and B Rumpf ldquoAn infrared spectroscopic investi-gation on the species distribution in the system NH3 + CO2+ H2OrdquoThermodynamic Properties of Complex Fluid Mixturespp 92ndash119 2004

[18] J R Ferraro K Nakamoto and C W Brown IntroductoryRaman Spectroscopy Academic Press San Diego Calif USA2 edition 2008

[19] H Martens and T NaesMultivariate calibration John Wiley ampSons London UK 1989

[20] K A Bakeev Process analytical technology spectroscopic toolsand implementation strategies for the chemical and pharmaceu-tical industries John Wiley amp Sons 2010

[21] K H Esbensen and P Geladi ldquoPrinciples of proper validationuse and abuse of re-sampling for validationrdquo Journal of Chemo-metrics vol 24 no 3-4 pp 168ndash187 2010

[22] A Seidell Solubilities of inorganic and organic compounds acompilation of quantitative solubility data from the periodicalliterature D Van Nostrand Company New York NY USA 2edition 1919

[23] K S Pitzer ldquoThermodynamics of electrolytes I theoreticalbasis and general equationsrdquo Journal of Physical Chemistry vol77 no 2 pp 268ndash277 1973

[24] H Que and C-C Chen ldquoThermodynamic modeling of theNH3-CO2-H2O system with electrolyte NRTL modelrdquo Indus-trial and Engineering Chemistry Research vol 50 no 19 pp11406ndash11421 2011

[25] EM Pawlikowski J Newman and JM Prausnitz ldquoPhase equi-libria for aqueous solutions of ammonia and carbon dioxiderdquoIndustrial Engineering Chemistry Research Process Design andDevelopment vol 21 pp 764ndash770 1982

[26] K H Liland T Almoslashy and B-H Mevik ldquoOptimal choiceof baseline correction for multivariate calibration of spectrardquoApplied Spectroscopy vol 64 no 9 pp 1007ndash1016 2010

[27] P H C Eilers ldquoA perfect smootherrdquo Analytical Chemistry vol75 no 14 pp 3631ndash3636 2003

[28] K Esbensen P Geladi A Larsen and P Williams ldquoMyth-busters one can always assign a wavelengthwavelength regionspecific for onersquos applicationrdquoNIRNews vol 26 no 4 pp 15ndash172015

[29] M Hoslashy K Steen and H Martens ldquoReview of partial leastsquares regression prediction error in Unscramblerrdquo Chemo-metrics and Intelligent Laboratory Systems vol 44 no 1-2 pp123ndash133 1998

[30] C K Ahn H W Lee Y S Chang et al ldquoCharacterization ofammonia-based CO2 capture process using ion speciationrdquoInternational Journal of Greenhouse Gas Control vol 5 no 6pp 1606ndash1613 2011

Submit your manuscripts athttpswwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Inorganic ChemistryInternational Journal of

Hindawi Publishing Corporation httpwwwhindawicom Volume 201

International Journal ofInternational Journal ofPhotoenergy


Carbohydrate Chemistry

International Journal ofInternational Journal of


Journal of

Chemistry


Advances in

Physical Chemistry

Hindawi Publishing Corporationhttpwwwhindawicom

Analytical Methods in Chemistry

Journal of

Volume 2014

Bioinorganic Chemistry and ApplicationsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

SpectroscopyInternational Journal of


The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Medicinal ChemistryInternational Journal of


Chromatography Research International


Applied ChemistryJournal of



Theoretical ChemistryJournal of


Journal of

Spectroscopy

Analytical ChemistryInternational Journal of


Journal of


Quantum Chemistry


Organic Chemistry International

ElectrochemistryInternational Journal of



CatalystsJournal of


wet chemistry titration to determine liquid phase is themost popular method This method takes time and errorscan be propagated during sampling chemical preparationweighing and titration [7] Various analytical techniques canbe used to determine the species present in CO2 reactedammonia solution 13C nuclear magnetic resonance (NMR)spectroscopy has been used [8] to determine ionic species ina range of ammonia concentration from 069 to 895molLand CO2 loading (total CO2 molesinitial solvent moles)concentrations from 033 to 072 Fourier transform infraredspectroscopy (FT-IR) and X-ray diffraction (XRD) wereused to qualitatively distinguish ammonium bicarbonatefrom ammonium carbonate and ammonium carbamatewhile CHN elemental analysis and near-infrared (NIR) spec-troscopywere used to quantify ammoniumbicarbonate basedon multivariate regression methods as reported in [9] Wenand Brooker [10] Zhao et al [11] andKim et al [12] suggestedmethods to determine carbon species in CO2-NH3-H2Osystem based on factor analysis where they assumed that thecorresponding Raman intensities were directly related to theconcentrations of species

This work reports on a method that combines Ramanspectroscopy and partial least-squares regression (PLS-R) forin situ solvent speciation in the chilled ammonia processand which does not rely on calibration by an independentanalysis method [13] for example NMR It describes thedevelopment validation and application of the methodfor determination of the liquid phase composition of theNH3-CO2-H2O system while comparisons of the speciationresults obtained here and those from established CO2-NH3-H2O thermodynamic equilibrium models are described in aseparate publication [14]

In the thermodynamic modeling of the CO2-NH3-H2Osystem a vapor-liquid equilibrium (VLE) is assumed to existfor water ammonia and CO2 The liquid phase can bedescribed by reactions (1)ndash(5) The chemical composition ofaqueous ammonia solution is described by

NH3 +H2O 999448999471 NH4+ +OHminus (1)

The reactions of dissolved CO2 in aqueous ammonia solutionare given by reactions (2)ndash(5) Within the present systemCO2 is bound as the anion species of carbonate [CO3

2minus]bicarbonate [HCO3

minus] and carbamate [NH2COOminus]

CO2(aq) +H2O 999448999471 H+ +HCO3minus (2)


CO2(aq) + 2NH3 999448999471 NH4+ + NH2COOminus (4)

NH4+ +NH2COOminus +H2O999448999471 NH4+ +HCO3

minus +NH3 (5)

During CO2 absorption by the aqueous NH3 solutionreactions (1)ndash(5) will equilibrate according to the CO2 con-centration pressure and temperature the carbon distribu-tion in the solvent is given by the concentrations of theanions in reactions (2)ndash(5) In addition to these reactions

and depending on the reaction conditions various othercompounds may precipitate such as ammonium bicarbonate(NH4HCO3) ammonium carbonate [(NH4)2CO3] ammo-nium carbamate (NH4NH2CO2) and ammonium sesquicar-bonate [(NH4)2CO3sdotNH4HCO3] Even though the precipita-tion is promoted by low temperature the high water contentin the solvent reduces much of this possibility

Spectroscopy Raman spectroscopy in particular is wellknown for in situ monitoring of the chemical reactionsof aqueous solutions [16 17] The water molecule showsonly weak Raman scattering hence Raman spectroscopy haspotential advantage over IR spectroscopy [18] for aqueousphase analysis such as required for the aqueous chilledammonia CO2 capture solvent The Raman band envelopesof aqueous solutions of ammonium carbonate ammoniumbicarbonate and ammonium carbamate have been identifiedand analyzed [10] and form the basis for the previouslyreported speciation studies of the CO2-NH3-H2O system [1112] However evaluation of the Raman spectra in these pre-vious studies was based on univariate single-band analysisThe use of superior multivariate partial least-squares regres-sion (PLS-R) [19] methods which exploit the multivariateinformation in the spectra has to the best of our knowledgenot been reported previously Furthermore the methoddeveloped in the present study is not limited to laboratoryapplications but can also be used to monitor continuouslya reactive process which presents an opportunity for itsimplementation as an on-line process analytical technology(PAT) [20] in carbon capture plants for optimization andefficient operation

2 Materials and Methods

The development of the Raman spectroscopy-based methodfor speciation of the CO2-NH3-H2O system is based on PLS-R analysis of a series of samples of known composition Theprocedure involves

(1) preparation of aqueous solutions that containknown concentrations of [CO3

2minus] [HCO3minus] and

[NH2COOminus] one set of solutions was prepared forPLS-R model calibration and a second independentset of solutions was used for validation

(2) determination of the Raman spectrum of each samplesolution

(3) preprocessing of the Raman spectra by cropping eachspectrum to cover the range 450ndash2300 cmminus1 withsubsequent elimination of the inconsistently varyingbaseline from spectrum-to-spectrum and centeringof each wavelength by subtraction of the mean

(4) calibration of the PLS-R models for the anionicspecies [CO3

2minus] [HCO3minus] and [NH2COOminus] using

the Raman spectra of the solutions prepared formodel calibration in the previous step (Step (3))

(5) validation of the PLS-R models using the Ramanspectra of the solutions prepared formodel validationin Step (3)


x-variables

Sam

ples

n

x11 x12

x21 x22





p

W

P

T

X Y

X = TPT + E

Y = UQT + F

y11 y1q

y21 y2q

yn1 ynq

UQ

qy-variables





























100 650 1200 1750 2300 2850 33250

21

Inte

nsity

(arb

itrar

y times105

unit)


(a)

450 750 1050 1350 1650 1950 23000

525

Inte

nsity

(arb

itrar

y times104

unit)


(b)

Inte

nsity

(arb

itrar

y

450 750 1050 1350 1650 1950 23000

100005000

unit)


(c)

Inte

nsity

(arb

itrar

y

450 750 1050 1350 1650 1950 2300minus5000

50000


unit)

(d)

























minus (7)





NH3 +H2O 999448999471 NH4+ +OHminus (10)

NH4+ 999448999471 H+ +NH3 (11)








450 750 1050 1350 1650 1950 23000

21

43

times104


Inte

nsity

(arb

itrar

y un

it)

(a)

450 750 1050 1350

1390

1650 1950 23000

5000

10000

Inte

nsity

(arb

itrar

y un

it)


(b)

450 750 1050 1350 1650 1950 23000

1

2

3times104


Inte

nsity

(arb

itrar

y un

it)

(c)













minus4000

minus2000

0

2000

4000

PLS

com

pone

nt n

umbe

r 2


times104


(a)

400 600 800 1000 1200 1400 1600 1800 2000 2200 2400minus1

0

1

2

3

4

B (a

rbitr

ary

unit)



(b)


0

001

002

003

004

005

Y-v

aria

nce


(c)


0

02

04

06

Refe

renc

e car

bona

te

conc

entr

atio

n (m

olk

g)



09995RMSEPr2

098980001309998

00043

(d)

















minus4000

minus2000

0

2000

4000

PLS

com

pone

nt n

umbe



(a)

0 500 1000 1500 2000 2500minus5

0

5

10

15

20

B (a

rbitr

ary

unit)



(b)


0

002

004

006

008

Y-v

aria

nce


(c)


0

02

04

06

08

1

Pred

icte

d bi

carb

onat

e co

ncen

trat

ion

(mol

kg)


0 02 04 06 08 1


09932RMSEPr2

096360019109966

00241

(d)








minus15minus1

minus050

051

152


PLS

com

pone

nt n

umbe

r 2


times104

times104


(a)

400 600 800 1000 1200 1400 1600 1800 2000 2200 2400minus1

0

1

2

3

4

5

B (a

rbitr

ary

unit)



(b)

0 1 2 3 4 5 60

00501

01502

02503

035



Y-v

aria

nce

(c)

0 1 20

02

02

04

04

06

06

08

08

112

12

14

14

16

16

18

18

2


Pred

icte

d ca

rbam

ate

conc

entr

atio

n (m

olk

g)



09936RMSEPr2

10246minus0006109968

00499

(d)







0 3 6 9 12 15 18 21 24 27 30 33 36 39 420

02

01

04

03

Sample number

Pred

icte

d

conc

entr

atio

n(m

olk

g)

carb

onat

e

(a)

0 3 6 9 12 15 18 21 24 27 30 33 36 39 42minus05

0

05

1

Sample number

Pred

icte

d

conc

entr

atio

n(m

olk

g)

bica

rbon

ate

(b)

0 3 6 9 12 15 18 21 24 27 30 33 36 39 420

1

15

05

2

Sample number

Pred

icte

d

conc

entr

atio

n(m

olk

g)

carb

amat

e

(c)



4 Conclusion



Appendix


See Table 4



Acknowledgments





References









































Journal of

Chemistry


Advances in

Physical Chemistry



Journal of

Volume 2014














Journal of

Spectroscopy



Journal of


Quantum Chemistry






CatalystsJournal of


x-variables

Sam

ples

n

x11 x12

x21 x22





p

W

P

T

X Y

X = TPT + E

Y = UQT + F

y11 y1q

y21 y2q

yn1 ynq

UQ

qy-variables





























100 650 1200 1750 2300 2850 33250

21

Inte

nsity

(arb

itrar

y times105

unit)


(a)

450 750 1050 1350 1650 1950 23000

525

Inte

nsity

(arb

itrar

y times104

unit)


(b)

Inte

nsity

(arb

itrar

y

450 750 1050 1350 1650 1950 23000

100005000

unit)


(c)

Inte

nsity

(arb

itrar

y

450 750 1050 1350 1650 1950 2300minus5000

50000


unit)

(d)

























minus (7)





NH3 +H2O 999448999471 NH4+ +OHminus (10)

NH4+ 999448999471 H+ +NH3 (11)








450 750 1050 1350 1650 1950 23000

21

43

times104


Inte

nsity

(arb

itrar

y un

it)

(a)

450 750 1050 1350

1390

1650 1950 23000

5000

10000

Inte

nsity

(arb

itrar

y un

it)


(b)

450 750 1050 1350 1650 1950 23000

1

2

3times104


Inte

nsity

(arb

itrar

y un

it)

(c)













minus4000

minus2000

0

2000

4000

PLS

com

pone

nt n

umbe

r 2


times104


(a)

400 600 800 1000 1200 1400 1600 1800 2000 2200 2400minus1

0

1

2

3

4

B (a

rbitr

ary

unit)



(b)


0

001

002

003

004

005

Y-v

aria

nce


(c)


0

02

04

06

Refe

renc

e car

bona

te

conc

entr

atio

n (m

olk

g)



09995RMSEPr2

098980001309998

00043

(d)

















minus4000

minus2000

0

2000

4000

PLS

com

pone

nt n

umbe



(a)

0 500 1000 1500 2000 2500minus5

0

5

10

15

20

B (a

rbitr

ary

unit)



(b)


0

002

004

006

008

Y-v

aria

nce


(c)


0

02

04

06

08

1

Pred

icte

d bi

carb

onat

e co

ncen

trat

ion

(mol

kg)


0 02 04 06 08 1


09932RMSEPr2

096360019109966

00241

(d)








minus15minus1

minus050

051

152


PLS

com

pone

nt n

umbe

r 2


times104

times104


(a)

400 600 800 1000 1200 1400 1600 1800 2000 2200 2400minus1

0

1

2

3

4

5

B (a

rbitr

ary

unit)



(b)

0 1 2 3 4 5 60

00501

01502

02503

035



Y-v

aria

nce

(c)

0 1 20

02

02

04

04

06

06

08

08

112

12

14

14

16

16

18

18

2


Pred

icte

d ca

rbam

ate

conc

entr

atio

n (m

olk

g)



09936RMSEPr2

10246minus0006109968

00499

(d)







0 3 6 9 12 15 18 21 24 27 30 33 36 39 420

02

01

04

03

Sample number

Pred

icte

d

conc

entr

atio

n(m

olk

g)

carb

onat

e

(a)

0 3 6 9 12 15 18 21 24 27 30 33 36 39 42minus05

0

05

1

Sample number

Pred

icte

d

conc

entr

atio

n(m

olk

g)

bica

rbon

ate

(b)

0 3 6 9 12 15 18 21 24 27 30 33 36 39 420

1

15

05

2

Sample number

Pred

icte

d

conc

entr

atio

n(m

olk

g)

carb

amat

e

(c)



4 Conclusion



Appendix


See Table 4



Acknowledgments





References









































Journal of

Chemistry


Advances in

Physical Chemistry



Journal of

Volume 2014














Journal of

Spectroscopy



Journal of


Quantum Chemistry






CatalystsJournal of












100 650 1200 1750 2300 2850 33250

21

Inte

nsity

(arb

itrar

y times105

unit)


(a)

450 750 1050 1350 1650 1950 23000

525

Inte

nsity

(arb

itrar

y times104

unit)


(b)

Inte

nsity

(arb

itrar

y

450 750 1050 1350 1650 1950 23000

100005000

unit)


(c)

Inte

nsity

(arb

itrar

y

450 750 1050 1350 1650 1950 2300minus5000

50000


unit)

(d)

























minus (7)





NH3 +H2O 999448999471 NH4+ +OHminus (10)

NH4+ 999448999471 H+ +NH3 (11)








450 750 1050 1350 1650 1950 23000

21

43

times104


Inte

nsity

(arb

itrar

y un

it)

(a)

450 750 1050 1350

1390

1650 1950 23000

5000

10000

Inte

nsity

(arb

itrar

y un

it)


(b)

450 750 1050 1350 1650 1950 23000

1

2

3times104


Inte

nsity

(arb

itrar

y un

it)

(c)













minus4000

minus2000

0

2000

4000

PLS

com

pone

nt n

umbe

r 2


times104


(a)

400 600 800 1000 1200 1400 1600 1800 2000 2200 2400minus1

0

1

2

3

4

B (a

rbitr

ary

unit)



(b)


0

001

002

003

004

005

Y-v

aria

nce


(c)


0

02

04

06

Refe

renc

e car

bona

te

conc

entr

atio

n (m

olk

g)



09995RMSEPr2

098980001309998

00043

(d)

















minus4000

minus2000

0

2000

4000

PLS

com

pone

nt n

umbe



(a)

0 500 1000 1500 2000 2500minus5

0

5

10

15

20

B (a

rbitr

ary

unit)



(b)


0

002

004

006

008

Y-v

aria

nce


(c)


0

02

04

06

08

1

Pred

icte

d bi

carb

onat

e co

ncen

trat

ion

(mol

kg)


0 02 04 06 08 1


09932RMSEPr2

096360019109966

00241

(d)








minus15minus1

minus050

051

152


PLS

com

pone

nt n

umbe

r 2


times104

times104


(a)

400 600 800 1000 1200 1400 1600 1800 2000 2200 2400minus1

0

1

2

3

4

5

B (a

rbitr

ary

unit)



(b)

0 1 2 3 4 5 60

00501

01502

02503

035



Y-v

aria

nce

(c)

0 1 20

02

02

04

04

06

06

08

08

112

12

14

14

16

16

18

18

2


Pred

icte

d ca

rbam

ate

conc

entr

atio

n (m

olk

g)



09936RMSEPr2

10246minus0006109968

00499

(d)







0 3 6 9 12 15 18 21 24 27 30 33 36 39 420

02

01

04

03

Sample number

Pred

icte

d

conc

entr

atio

n(m

olk

g)

carb

onat

e

(a)

0 3 6 9 12 15 18 21 24 27 30 33 36 39 42minus05

0

05

1

Sample number

Pred

icte

d

conc

entr

atio

n(m

olk

g)

bica

rbon

ate

(b)

0 3 6 9 12 15 18 21 24 27 30 33 36 39 420

1

15

05

2

Sample number

Pred

icte

d

conc

entr

atio

n(m

olk

g)

carb

amat

e

(c)



4 Conclusion



Appendix


See Table 4



Acknowledgments





References









































Journal of

Chemistry


Advances in

Physical Chemistry



Journal of

Volume 2014














Journal of

Spectroscopy



Journal of


Quantum Chemistry






CatalystsJournal of


100 650 1200 1750 2300 2850 33250

21

Inte

nsity

(arb

itrar

y times105

unit)


(a)

450 750 1050 1350 1650 1950 23000

525

Inte

nsity

(arb

itrar

y times104

unit)


(b)

Inte

nsity

(arb

itrar

y

450 750 1050 1350 1650 1950 23000

100005000

unit)


(c)

Inte

nsity

(arb

itrar

y

450 750 1050 1350 1650 1950 2300minus5000

50000


unit)

(d)

























minus (7)





NH3 +H2O 999448999471 NH4+ +OHminus (10)

NH4+ 999448999471 H+ +NH3 (11)








450 750 1050 1350 1650 1950 23000

21

43

times104


Inte

nsity

(arb

itrar

y un

it)

(a)

450 750 1050 1350

1390

1650 1950 23000

5000

10000

Inte

nsity

(arb

itrar

y un

it)


(b)

450 750 1050 1350 1650 1950 23000

1

2

3times104


Inte

nsity

(arb

itrar

y un

it)

(c)













minus4000

minus2000

0

2000

4000

PLS

com

pone

nt n

umbe

r 2


times104


(a)

400 600 800 1000 1200 1400 1600 1800 2000 2200 2400minus1

0

1

2

3

4

B (a

rbitr

ary

unit)



(b)


0

001

002

003

004

005

Y-v

aria

nce


(c)


0

02

04

06

Refe

renc

e car

bona

te

conc

entr

atio

n (m

olk

g)



09995RMSEPr2

098980001309998

00043

(d)

















minus4000

minus2000

0

2000

4000

PLS

com

pone

nt n

umbe



(a)

0 500 1000 1500 2000 2500minus5

0

5

10

15

20

B (a

rbitr

ary

unit)



(b)


0

002

004

006

008

Y-v

aria

nce


(c)


0

02

04

06

08

1

Pred

icte

d bi

carb

onat

e co

ncen

trat

ion

(mol

kg)


0 02 04 06 08 1


09932RMSEPr2

096360019109966

00241

(d)








minus15minus1

minus050

051

152


PLS

com

pone

nt n

umbe

r 2


times104

times104


(a)

400 600 800 1000 1200 1400 1600 1800 2000 2200 2400minus1

0

1

2

3

4

5

B (a

rbitr

ary

unit)



(b)

0 1 2 3 4 5 60

00501

01502

02503

035



Y-v

aria

nce

(c)

0 1 20

02

02

04

04

06

06

08

08

112

12

14

14

16

16

18

18

2


Pred

icte

d ca

rbam

ate

conc

entr

atio

n (m

olk

g)



09936RMSEPr2

10246minus0006109968

00499

(d)







0 3 6 9 12 15 18 21 24 27 30 33 36 39 420

02

01

04

03

Sample number

Pred

icte

d

conc

entr

atio

n(m

olk

g)

carb

onat

e

(a)

0 3 6 9 12 15 18 21 24 27 30 33 36 39 42minus05

0

05

1

Sample number

Pred

icte

d

conc

entr

atio

n(m

olk

g)

bica

rbon

ate

(b)

0 3 6 9 12 15 18 21 24 27 30 33 36 39 420

1

15

05

2

Sample number

Pred

icte

d

conc

entr

atio

n(m

olk

g)

carb

amat

e

(c)



4 Conclusion



Appendix


See Table 4



Acknowledgments





References









































Journal of

Chemistry


Advances in

Physical Chemistry



Journal of

Volume 2014














Journal of

Spectroscopy



Journal of


Quantum Chemistry






CatalystsJournal of





















minus (7)





NH3 +H2O 999448999471 NH4+ +OHminus (10)

NH4+ 999448999471 H+ +NH3 (11)








450 750 1050 1350 1650 1950 23000

21

43

times104


Inte

nsity

(arb

itrar

y un

it)

(a)

450 750 1050 1350

1390

1650 1950 23000

5000

10000

Inte

nsity

(arb

itrar

y un

it)


(b)

450 750 1050 1350 1650 1950 23000

1

2

3times104


Inte

nsity

(arb

itrar

y un

it)

(c)













minus4000

minus2000

0

2000

4000

PLS

com

pone

nt n

umbe

r 2


times104


(a)

400 600 800 1000 1200 1400 1600 1800 2000 2200 2400minus1

0

1

2

3

4

B (a

rbitr

ary

unit)



(b)


0

001

002

003

004

005

Y-v

aria

nce


(c)


0

02

04

06

Refe

renc

e car

bona

te

conc

entr

atio

n (m

olk

g)



09995RMSEPr2

098980001309998

00043

(d)

















minus4000

minus2000

0

2000

4000

PLS

com

pone

nt n

umbe



(a)

0 500 1000 1500 2000 2500minus5

0

5

10

15

20

B (a

rbitr

ary

unit)



(b)


0

002

004

006

008

Y-v

aria

nce


(c)


0

02

04

06

08

1

Pred

icte

d bi

carb

onat

e co

ncen

trat

ion

(mol

kg)


0 02 04 06 08 1


09932RMSEPr2

096360019109966

00241

(d)








minus15minus1

minus050

051

152


PLS

com

pone

nt n

umbe

r 2


times104

times104


(a)

400 600 800 1000 1200 1400 1600 1800 2000 2200 2400minus1

0

1

2

3

4

5

B (a

rbitr

ary

unit)



(b)

0 1 2 3 4 5 60

00501

01502

02503

035



Y-v

aria

nce

(c)

0 1 20

02

02

04

04

06

06

08

08

112

12

14

14

16

16

18

18

2


Pred

icte

d ca

rbam

ate

conc

entr

atio

n (m

olk

g)



09936RMSEPr2

10246minus0006109968

00499

(d)







0 3 6 9 12 15 18 21 24 27 30 33 36 39 420

02

01

04

03

Sample number

Pred

icte

d

conc

entr

atio

n(m

olk

g)

carb

onat

e

(a)

0 3 6 9 12 15 18 21 24 27 30 33 36 39 42minus05

0

05

1

Sample number

Pred

icte

d

conc

entr

atio

n(m

olk

g)

bica

rbon

ate

(b)

0 3 6 9 12 15 18 21 24 27 30 33 36 39 420

1

15

05

2

Sample number

Pred

icte

d

conc

entr

atio

n(m

olk

g)

carb

amat

e

(c)



4 Conclusion



Appendix


See Table 4



Acknowledgments





References









































Journal of

Chemistry


Advances in

Physical Chemistry



Journal of

Volume 2014














Journal of

Spectroscopy



Journal of


Quantum Chemistry






CatalystsJournal of


450 750 1050 1350 1650 1950 23000

21

43

times104


Inte

nsity

(arb

itrar

y un

it)

(a)

450 750 1050 1350

1390

1650 1950 23000

5000

10000

Inte

nsity

(arb

itrar

y un

it)


(b)

450 750 1050 1350 1650 1950 23000

1

2

3times104


Inte

nsity

(arb

itrar

y un

it)

(c)













minus4000

minus2000

0

2000

4000

PLS

com

pone

nt n

umbe

r 2


times104


(a)

400 600 800 1000 1200 1400 1600 1800 2000 2200 2400minus1

0

1

2

3

4

B (a

rbitr

ary

unit)



(b)


0

001

002

003

004

005

Y-v

aria

nce


(c)


0

02

04

06

Refe

renc

e car

bona

te

conc

entr

atio

n (m

olk

g)



09995RMSEPr2

098980001309998

00043

(d)

















minus4000

minus2000

0

2000

4000

PLS

com

pone

nt n

umbe



(a)

0 500 1000 1500 2000 2500minus5

0

5

10

15

20

B (a

rbitr

ary

unit)



(b)


0

002

004

006

008

Y-v

aria

nce


(c)


0

02

04

06

08

1

Pred

icte

d bi

carb

onat

e co

ncen

trat

ion

(mol

kg)


0 02 04 06 08 1


09932RMSEPr2

096360019109966

00241

(d)








minus15minus1

minus050

051

152


PLS

com

pone

nt n

umbe

r 2


times104

times104


(a)

400 600 800 1000 1200 1400 1600 1800 2000 2200 2400minus1

0

1

2

3

4

5

B (a

rbitr

ary

unit)



(b)

0 1 2 3 4 5 60

00501

01502

02503

035



Y-v

aria

nce

(c)

0 1 20

02

02

04

04

06

06

08

08

112

12

14

14

16

16

18

18

2


Pred

icte

d ca

rbam

ate

conc

entr

atio

n (m

olk

g)



09936RMSEPr2

10246minus0006109968

00499

(d)







0 3 6 9 12 15 18 21 24 27 30 33 36 39 420

02

01

04

03

Sample number

Pred

icte

d

conc

entr

atio

n(m

olk

g)

carb

onat

e

(a)

0 3 6 9 12 15 18 21 24 27 30 33 36 39 42minus05

0

05

1

Sample number

Pred

icte

d

conc

entr

atio

n(m

olk

g)

bica

rbon

ate

(b)

0 3 6 9 12 15 18 21 24 27 30 33 36 39 420

1

15

05

2

Sample number

Pred

icte

d

conc

entr

atio

n(m

olk

g)

carb

amat

e

(c)



4 Conclusion



Appendix


See Table 4



Acknowledgments





References









































Journal of

Chemistry


Advances in

Physical Chemistry



Journal of

Volume 2014














Journal of

Spectroscopy



Journal of


Quantum Chemistry






CatalystsJournal of



minus4000

minus2000

0

2000

4000

PLS

com

pone

nt n

umbe

r 2


times104


(a)

400 600 800 1000 1200 1400 1600 1800 2000 2200 2400minus1

0

1

2

3

4

B (a

rbitr

ary

unit)



(b)


0

001

002

003

004

005

Y-v

aria

nce


(c)


0

02

04

06

Refe

renc

e car

bona

te

conc

entr

atio

n (m

olk

g)



09995RMSEPr2

098980001309998

00043

(d)

















minus4000

minus2000

0

2000

4000

PLS

com

pone

nt n

umbe



(a)

0 500 1000 1500 2000 2500minus5

0

5

10

15

20

B (a

rbitr

ary

unit)



(b)


0

002

004

006

008

Y-v

aria

nce


(c)


0

02

04

06

08

1

Pred

icte

d bi

carb

onat

e co

ncen

trat

ion

(mol

kg)


0 02 04 06 08 1


09932RMSEPr2

096360019109966

00241

(d)








minus15minus1

minus050

051

152


PLS

com

pone

nt n

umbe

r 2


times104

times104


(a)

400 600 800 1000 1200 1400 1600 1800 2000 2200 2400minus1

0

1

2

3

4

5

B (a

rbitr

ary

unit)



(b)

0 1 2 3 4 5 60

00501

01502

02503

035



Y-v

aria

nce

(c)

0 1 20

02

02

04

04

06

06

08

08

112

12

14

14

16

16

18

18

2


Pred

icte

d ca

rbam

ate

conc

entr

atio

n (m

olk

g)



09936RMSEPr2

10246minus0006109968

00499

(d)







0 3 6 9 12 15 18 21 24 27 30 33 36 39 420

02

01

04

03

Sample number

Pred

icte

d

conc

entr

atio

n(m

olk

g)

carb

onat

e

(a)

0 3 6 9 12 15 18 21 24 27 30 33 36 39 42minus05

0

05

1

Sample number

Pred

icte

d

conc

entr

atio

n(m

olk

g)

bica

rbon

ate

(b)

0 3 6 9 12 15 18 21 24 27 30 33 36 39 420

1

15

05

2

Sample number

Pred

icte

d

conc

entr

atio

n(m

olk

g)

carb

amat

e

(c)



4 Conclusion



Appendix


See Table 4



Acknowledgments





References









































Journal of

Chemistry


Advances in

Physical Chemistry



Journal of

Volume 2014














Journal of

Spectroscopy



Journal of


Quantum Chemistry






CatalystsJournal of



minus4000

minus2000

0

2000

4000

PLS

com

pone

nt n

umbe



(a)

0 500 1000 1500 2000 2500minus5

0

5

10

15

20

B (a

rbitr

ary

unit)



(b)


0

002

004

006

008

Y-v

aria

nce


(c)


0

02

04

06

08

1

Pred

icte

d bi

carb

onat

e co

ncen

trat

ion

(mol

kg)


0 02 04 06 08 1


09932RMSEPr2

096360019109966

00241

(d)








minus15minus1

minus050

051

152


PLS

com

pone

nt n

umbe

r 2


times104

times104


(a)

400 600 800 1000 1200 1400 1600 1800 2000 2200 2400minus1

0

1

2

3

4

5

B (a

rbitr

ary

unit)



(b)

0 1 2 3 4 5 60

00501

01502

02503

035



Y-v

aria

nce

(c)

0 1 20

02

02

04

04

06

06

08

08

112

12

14

14

16

16

18

18

2


Pred

icte

d ca

rbam

ate

conc

entr

atio

n (m

olk

g)



09936RMSEPr2

10246minus0006109968

00499

(d)







0 3 6 9 12 15 18 21 24 27 30 33 36 39 420

02

01

04

03

Sample number

Pred

icte

d

conc

entr

atio

n(m

olk

g)

carb

onat

e

(a)

0 3 6 9 12 15 18 21 24 27 30 33 36 39 42minus05

0

05

1

Sample number

Pred

icte

d

conc

entr

atio

n(m

olk

g)

bica

rbon

ate

(b)

0 3 6 9 12 15 18 21 24 27 30 33 36 39 420

1

15

05

2

Sample number

Pred

icte

d

conc

entr

atio

n(m

olk

g)

carb

amat

e

(c)



4 Conclusion



Appendix


See Table 4



Acknowledgments





References









































Journal of

Chemistry


Advances in

Physical Chemistry



Journal of

Volume 2014














Journal of

Spectroscopy



Journal of


Quantum Chemistry






CatalystsJournal of



minus15minus1

minus050

051

152


PLS

com

pone

nt n

umbe

r 2


times104

times104


(a)

400 600 800 1000 1200 1400 1600 1800 2000 2200 2400minus1

0

1

2

3

4

5

B (a

rbitr

ary

unit)



(b)

0 1 2 3 4 5 60

00501

01502

02503

035



Y-v

aria

nce

(c)

0 1 20

02

02

04

04

06

06

08

08

112

12

14

14

16

16

18

18

2


Pred

icte

d ca

rbam

ate

conc

entr

atio

n (m

olk

g)



09936RMSEPr2

10246minus0006109968

00499

(d)







0 3 6 9 12 15 18 21 24 27 30 33 36 39 420

02

01

04

03

Sample number

Pred

icte

d

conc

entr

atio

n(m

olk

g)

carb

onat

e

(a)

0 3 6 9 12 15 18 21 24 27 30 33 36 39 42minus05

0

05

1

Sample number

Pred

icte

d

conc

entr

atio

n(m

olk

g)

bica

rbon

ate

(b)

0 3 6 9 12 15 18 21 24 27 30 33 36 39 420

1

15

05

2

Sample number

Pred

icte

d

conc

entr

atio

n(m

olk

g)

carb

amat

e

(c)



4 Conclusion



Appendix


See Table 4



Acknowledgments





References









































Journal of

Chemistry


Advances in

Physical Chemistry



Journal of

Volume 2014














Journal of

Spectroscopy



Journal of


Quantum Chemistry






CatalystsJournal of


0 3 6 9 12 15 18 21 24 27 30 33 36 39 420

02

01

04

03

Sample number

Pred

icte

d

conc

entr

atio

n(m

olk

g)

carb

onat

e

(a)

0 3 6 9 12 15 18 21 24 27 30 33 36 39 42minus05

0

05

1

Sample number

Pred

icte

d

conc

entr

atio

n(m

olk

g)

bica

rbon

ate

(b)

0 3 6 9 12 15 18 21 24 27 30 33 36 39 420

1

15

05

2

Sample number

Pred

icte

d

conc

entr

atio

n(m

olk

g)

carb

amat

e

(c)



4 Conclusion



Appendix


See Table 4



Acknowledgments





References









































Journal of

Chemistry


Advances in

Physical Chemistry



Journal of

Volume 2014














Journal of

Spectroscopy



Journal of


Quantum Chemistry






CatalystsJournal of




References









































Journal of

Chemistry


Advances in

Physical Chemistry



Journal of

Volume 2014














Journal of

Spectroscopy



Journal of


Quantum Chemistry






CatalystsJournal of

























Journal of

Chemistry


Advances in

Physical Chemistry



Journal of

Volume 2014














Journal of

Spectroscopy



Journal of


Quantum Chemistry






CatalystsJournal of










Journal of

Chemistry


Advances in

Physical Chemistry



Journal of

Volume 2014














Journal of

Spectroscopy



Journal of


Quantum Chemistry






CatalystsJournal of

Documents

Equilibrium Measurements of the NH3-CO2-H2O System ...publications.lib.chalmers.se/records/fulltext/250668/local_250668.pdf · ResearchArticle Equilibrium Measurements of the NH 3-CO