J Radioanal Nucl Chem (2017) 312:461–470 DOI 10.1007/s10967-017-5252-8
UV–Vis spectroscopy with chemometric data treatment: an option for on-line control in nuclear industry Dmitry Kirsanov1,2
•
Alisa Rudnitskaya3 • Andrey Legin1,2 • Vasily Babain2,4
Received: 3 March 2017 / Published online: 22 April 2017 Ó Akade´miai Kiado´, Budapest, Hungary 2017
Abstract Chemometrics can be very useful for the classical field of UV–Vis determination of metals in aqueous solutions. A conventional approach consisting of using selective bands in a univariate mode is often not applicable to the real-world samples from e.g. hydrometallurgical processes, because of overlapping signals, light scattering on foreign particles, gas bubble formation, etc. And this is where chemometrics can do a good job. This paper overviews certain contributions to the field of multivariate data processing of UV–Vis spectra for seemingly simple case of metal detection in aqueous solutions. Special attention is given to applications in nuclear technology field. Keywords UV–Vis spectroscopy Chemometrics Metals On-line control Process analytical technology Nuclear technology
& Dmitry Kirsanov
[email protected] 1
Institute of Chemistry, St. Petersburg State University, Universitetskaya nab. 7/9, St. Petersburg, Russia 199034
2
Laboratory of Artificial Sensory Systems, ITMO University, Kronverkskiy pr. 49, St. Petersburg, Russia 197101
3
CESAM and Chemistry Department, Aveiro University, Campus Universita´rio De Santiago, 3810-193 Aveiro, Portugal
4
Three Arc MiningInc., Stary Tolmachevsky per. 5, Moscow, Russia 115184
Introduction Origins of chemometrics are usually placed in the 1960s when computing became largely available for the scientific community. Accessibility of the computers stimulated the development of theoretical chemistry involving complex calculation, e.g. molecular modelling or quantitative structure–activity relationship, as well as applications of statistics to analytical chemistry in particular to the treatment of complex instrumental signals. Thus, the advent of analytical instruments generating for each sample a large number of measurements, which were often collinear and non-selective, was another catalyzer for the introduction of chemometrics. The term ‘‘chemometrics’’, which came to designate a combination of statistics, computationally intensive multivariate methods and experimental design applied to analytical chemistry, was coined by Wold in 1972 [1]. Since then various methods and applications of chemometrics have been developed. Multivariate calibration, structure–activity modelling, pattern recognition, classification, discriminant analysis, and multivariate process modelling and monitoring were identified as areas, where chemometrics has been the most successful [2]. Further evidence of chemometrics success is its widespread use in the industry as it was introduced for process monitoring in all kinds of manufacturing processes, from petrochemical to pharmaceutical and food. Though numerous chemometric methods and algorithms have been developed in the course of the years, Principal Component Analysis (PCA) and Partial Least Square regression (PLSR) remain the most popular. One of the most striking examples of usefulness of chemometrics is Near-Infrared (NIR) spectroscopy. NIR spectroscopy was first introduced in the 1950s but was scarcely used till 80s. It was deemed a ‘‘sleeper among
123
462
spectroscopic techniques’’ due to its presumably high potential but very limited practical applications [3]. Indeed, NIR spectra typically contain broad and overlapping bands influenced by various physical, chemical and structural parameters. NIR spectra are, thus, information-rich but non-selective, and univariate approach taking into account a single band at a time is not applicable. On the other hand multivariate analysis techniques allowed extracting analytical information from NIR spectra and opened up new horizons for its application as a process control tool in industry [4–6]. Other spectroscopic techniques such as inductively coupled plasma-optical emission/atomic absorption spectroscopy (ICP-OES/AAS) [7], ICP-mass spectrometry (ICP-MS) [8], laser-induced breakdown spectroscopy (LIBS) [9] and Fourier Transform mid-infrared spectroscopy (FT-MIR) [10] also benefited from the use of chemometrics. UV–Vis spectrophotometry is considered as a simple technique useful for analysis and characterization of individual compounds but with limited applications in the field of complex media analysis. The analysis of multicomponent samples by UV–Vis is hindered by the lack of sensitivity and selectivity due to overlapping bands of the components, thus preventing the use of this method for online control of industrial processes. The application of chemometrics to the resolution of overlapping spectra of complex mixtures in order to increase the potential of UV– Vis spectroscopy seems to be a natural next step [11]. The purpose of this review is to demonstrate how the use of chemometrics to the processing of UV–Vis spectra may advance applications of this method to the practical tasks. We will discuss briefly the use of UV–Vis spectrometry for metal quantification in aqueous solutions and then we will make a special focus on nuclear technology related applications, since it is there where one needs fast and reliable answer about content of various metals in on-line mode. Various organic substances can be quantified from UV–Vis spectra with the help of chemometrics as well, however we will not focus on these applications. In no way this review is claiming to embrace all the literature available on the topic; we will rather be concentrated on certain selected publications, which we believe reflect the current situation and trends in the field.
Detection of metals with chromogenic ligands with UV–Vis spectroscopy and chemometrics In spite of moderate sensitivity, UV–Vis spectroscopy is still a very widely applied technique. Direct metal analysis with UV–Vis spectroscopy is possible and generally would be suitable for on-line implementation in technological process control, however its application is only limited to
123
J Radioanal Nucl Chem (2017) 312:461–470
colored metal solutions, so the number of potential analytes is quite small. Another concern regarding direct UV–Vis quantification is that detection limits are rather high in this case. Traditional approach to analysis of metals with UV–Vis relies on the use of chromogenic ligands, which upon the binding reaction with target analyte yield intense colored complexes. Though this procedure allows for significant decrease of detection limits of the method, it is normally effective only when determination of single metal is required. Chromogenic ligands are usually not highly selective and spectra that they form with different metals are often highly overlapping, which leads to the decrease of analysis precision or even make it altogether impossible. One of the ways of dealing with it is to perform separation and/or masking prior to determination. This approach is quite laborious, which diminish UV–Vis attractiveness as simple and fast technique, while also invalidate it as a potential method for on or in-line analysis. As an alternative, chemometrics can be used to circumvent insufficient selectivity and carry on multicomponent analysis without sample preparation. A very nice example from one of the first demonstrations of the potential of chemometrics in this field is reported in the work of Otto and Wegscheider [12]. The authors demonstrated that simultaneous quantitative analysis of iron, cobalt, nickel, copper and palladium in mixtures at low concentrations (down to 10-6 M) using visible spectroscopy of diethyldithiocarbamate chelate complexes of these metals is possible. Multilinear regression (MLR) was compared with PLSR (projection on latent structures) and the latter one displayed superior performance ensuring relative errors below 6%. There is also a quote by authors we would like to reproduce here, since it is very appropriate and consistent with the scope of this review: ‘‘The successful simultaneous determination of five metal ions by means of their DEDTC (diethyldithiocarbamate) complexes raises the question of whether there are some reagents out of hundreds proposed for spectrophotometric single component analysis that are more suitable for SMA (spectrophotometric multicomponent analysis) of metals than is DEDTC. In this context the existing one-wavelength based photometric methods have to be revisited with respect to their performance in the whole spectral range.’’ Indeed, with popularization of chemometrics, promoted by appearance of several tutorial reviews on multivariate calibration (see e.g. [13, 14]) numerous studies emerged revisiting the UV–Vis methods and their capabilities in the new ‘‘multivariate’’ context. The example of classical study in this research direction is described in the work [15], where 2-(5-bromo-2-pyridylazo)-5-diethylaminophenol (Br-PADAP) was applied as a coloring agent for simultaneous determination of Fe, Cu, Zn, Co and Ni in aqueous media and for quantification of
J Radioanal Nucl Chem (2017) 312:461–470
Fe, Cu and Zn in blood serum as well. Spectrophotometric data in visible range were processed with PLSR. The authors did intensive research on establishing optimal parameters for analysis and studied several possible interferences. Another typical study of this kind was reported in Ref. [16] where the authors used MLR to process visible spectra in the range 390–480 nm for simultaneous determination of cadmium, copper and zinc. Colored complexes of the metals were obtained using 1,5-bis(di-2-pyridylmethylene)thiocarbonohydrazide. Estimated linear determination ranges were 0.1–1.7 lg/ml for Cd, 0.1–1.3 lg/ml for Cu and 0.2–1.2 lg/ml for Zn. Strong interference from Co(II), Ni(II), Fe(III), Hg(II) and Hg(I) ions was observed since these metals also form colored complexes with the ligand. As a demonstration, diverse spiked samples and reference materials were analyzed. Later the same authors reported on simultaneous determination of iron, cobalt, nickel and copper using the same chelating agent but applying principal component regression (PCR), PLS1 and PLS2 for processing of the visible spectra (390–510 nm) [17].
Analysis of metals with UV–Vis characteristic absorption spectra and chemometrics Distinct characteristic bands of the electronic transitions in some metals allow for reliable UV–Vis quantitative concentration assessment even without dyes. In this case, univariate approach to UV–Vis spectra often fails in complex mixtures due to broad overlapping peaks of components. Once again, performance of UV–Vis spectroscopy can be significantly improved by chemometric data processing. One of the first examples of such combination was reported in 1987 [18]. In this study eight rare earth elements (Pr, Nd, Sm, Eu, Dy, Ho, Er and Tm) where simultaneously determined by spectrophotometry in 350–850 nm wavelength range with repetitive spectral subtraction method (RSSM). The results were compared with those from derivative spectra and from correction factor methods and were shown to have higher precision. Besides, the application of the suggested method to quantification of these metals in monazite and xenotime ores was presented. The study of Carey and Wangen [19] describes application of PLSR to simultaneous quantification of plutonium and nitric acid using visible absorption spectra (500–880 nm). Although the determination of Pu in this range is quite straightforward, the analysis of nitric acid is more challenging since it has only a minor contribution to the spectra of the mixture. Both analytes were successfully quantified in concentration ranges 1.99–29.9 g/L of Pu and 0.44–3.08 mol/L of HNO3. The same authors also reported on the application of PCA and
463
nearest neighbor classifier to evaluate the effects of fluoride and oxalate on the Pu spectra [20]. Simultaneous determination of Ce, Pr and Nd using PLSR is described in Ref. [21]. The spectra of mixtures were recorded in 290–800 nm range. Although the experimental design of model mixtures was not optimal (lanthanide concentrations were varied only over two–three fixed levels), the resulted models allowed for determination of concentrations in commercial rare-earth product with significantly greater precision than the conventional univariate calibration method. The authors also tried to evaluate samarium concentration but in the real sample it was too low and out of calibration range. In the works of Meinrath quantitative spectroscopic speciation of various hydrolysed U(VI) solutions has been demonstrated using factor analysis [22, 23] The same author with co-workers successfully applied computer-assisted target factor analysis combined with Monte Carlo simulations and resampling techniques (i.e. threshold bootstrap methods) to the speciation of lanthanides in solutions employing UV–Vis spectroscopy data. Interactions of Nd(III) with various ligands including polyoxometalates [24], nicotinic acid N-oxide, derivatives of carboxylated pyridine N-oxide and Arsenazo III have been studied [25–27]. Interaction of hexavalent uranium with pyridine carboxylic acid N-oxide derivatives was also studied [28]. Aforementioned works specifically focused on application of the resampling techniques to the error propagation and uncertainty estimation of the measurements in the situations when classical error progression approximations are not applicable, of which use of multivariate methods for resolution of the overlapping UV–Vis spectra was selected as a case study. The report [29] addresses the problem of overlapping peaks of europium, terbium and yttrium in the presence of perchloric acid. This overlapping seriously hinders individual analysis of rare-earth elements in mixtures. The authors proposed the approach based on Kalman filter and absorption coefficient matrix obtained from MLR for simultaneous quantification of all three metals in the complex sample of phosphor powder. Somewhat similar idea was exploited in the work [30] where the authors studied the data set of UV–Vis spectra (180–800 nm) for four metal ions (Fe, Ni, Co, Cu) in aqueous solutions.. Calibration models in order to predict content of the metals in test solutions were obtained using MLR and K-matrix approach. The latter allowed obtaining better results in metal quantification while also providing information of real spectral features. These results are however come with a caveat that K matrix approach works well only in relatively simple cases when spectra obey Beer’s law and the level of noise and unidentified components is negligible.
123
464
Using PCA and target transformation factor analysis (TTFA) resolution of the visible spectra (300–500 nm) obtained in solutions of modified calixarene ligand and rare earth ions Eu3?, Dy3?, and Tb3? was achieved [31]. The stoichiometry of the complexes was established and corresponding stability constants were calculated. The authors of the work [32] reported on feasibility study of simultaneous determination of several lanthanides: Ce, Pr and Nd using 200–1000 nm spectral measurements. In order to provide for quantitative assessment of concentrations PLSR and correlation constrained MCR-ALS were employed. While it was possible to determine Nd and Pr, the quantification of Ce posed certain problem due to strong overlapping of Ce band with HNO3 band under experimental conditions. The significant part of the applications in this section is devoted to the analysis of lanthanides and actinides. This is easy to explain taking into account that these elements have direct relation to nuclear fuel cycle where the problem of development of fast and reliable methods suitable for on-line control is really urgent.
Towards on-line control with UV–Vis spectrometry There is a great interest in on-line determination of metals coming from the nuclear technology field. Timely, preferably on-line, monitoring of the industrial process and detection of deviations from the normal operating conditions is of paramount importance for the nuclear technology and, in particular, reprocessing of the spent nuclear fuel (SNF). Excellent review of analytical problems related to the PUREX process, was done in two works [33, 34]. Similarly to other industrial processes, continuous monitoring provides possibility to enhance performance and minimize losses by timely correcting deviating processes. Currently the whole new paradigm is evolving to address these issues (process analytical technology—PAT) [35]. In the case of radiochemical processes, such monitoring would also serve to critically improve safety due to better accountability of fissile materials. Currently, process control at the spent nuclear fuel processing facilities is principally done in the laboratory using inductively coupled plasma mass spectrometry [8] and thermal ionization mass spectrometry [36] with prior sample preparation using isotope dilution [37] or separation using high-performance liquid or ion chromatography [38]. While these techniques are highly accurate and selective, they are resource intensive and have extended analysis times which may result in the time lags of about 15–20 h between sampling and final result—obviously, too long. Such prolonged analysis times are particularly unacceptable in the light of demands of
123
J Radioanal Nucl Chem (2017) 312:461–470
non-proliferation policy, which require information about plutonium concentration as soon as possible. Peculiar analytical problems in nuclear chemistry are caused by radioactivity of analysed solutions, which can sometimes exceed 1010 BkL-1. Consequently, analytical tasks such as sampling and sample handling, which are very simple and trivial with non-radioactive solutions, become extremely complicated and time consuming. It is worth to note that direct contact of operator with radioactive samples have to be minimized or, ideally, excluded. Thus, the urgent need of the on-line nondestructive monitoring of radiochemical processes is indisputable. Separation technologies currently employed for reprocessing of spent nuclear fuel, including the PUREX and UREX? processes, are based on liquid–liquid extraction. Uranium, plutonium and neptunium are extracted from the aqueous solutions of the spent fuel using organic solvents, while fission products remain in the aqueous phase [39]. The next step consists in extraction of minor actinides from raffinates of PUREX-process [40, 41]. The distribution of each element between aqueous and organic phases and, consequently, efficacy of separation and purification is governed by the major process variables such as acid concentration, organic ligand concentration, reduction potential and temperature. Element distribution should remain within established limits, corresponding to the normal operating conditions (NOC), ensuring stable process functioning. On-line measurements of the solutions’ composition would allow maintaining process parameters within NOC and timely correct the deviations. Primary analytes of interest to be monitored during SNF reprocessing are metals, mainly actinides and lanthanides, and nitric acid. Instrumentation to be usable for the on-line monitoring of the processing of the spent nuclear fuel must be robust, require little maintenance and tolerate harsh process conditions such as high radioactivity levels and high acid concentrations. Nondestructive optical techniques, primarily UV–Vis, Raman and IR spectroscopy comply with requirements enumerated above and are convenient tools for metal quantification in solutions. Often mentioned drawback of UV–Vis spectroscopy is that characteristic absorption allows only for rather high detection limits, but spent nuclear fuel processing deals with rather high concentrations so it is not an issue. Detection limit of direct UV–Vis spectroscopy for actinides is about 0.1 g/L, while concentrations of uranium in technological solutions varies in the range of 1–300 g/L, plutonium of 0.1–20 g/L and nitric acid of 0.1–3 mol/L. Such concentrations of uranium and plutonium are high enough to be detected by UV–Vis spectroscopy. Requirement of very low detection limits for the analytical instruments currently employed for spent nuclear fuel analysis is dictated by the necessity to dilute
J Radioanal Nucl Chem (2017) 312:461–470
samples due to their high radioactivity. Dilution of thousand to ten thousand times or even more are typically done, otherwise handling of the samples is too dangerous. Use of on-line or in-line techniques such as UV–Vis spectroscopy eliminates this problem. Applications of UV–Vis spectroscopy for the process control in nuclear technology will be discussed in more details below. Application of UV–Vis spectroscopy in nuclear technology. Univariate approach: detection using single wavelength Possibility of implementation of on-line control of the continuous treatment process using spectroscopic techniques also has been intensively studied since 1970s [42]. However, the first attempt to apply UV–Vis spectroscopy for the quantification of rare earth metals in the presence of the nitric acid dates back to 1941 [43]. In this work transmittance spectra of lanthanum, cerium, praseodymium, neodymium, samarium, europium and gadolinium in the solutions of nitric acid were measured in the range from 350 to 1000 nm. All metals but gadolinium had absorption bands in the studied spectral range, which were partly overlapping. Simultaneous determination of Pr, Nd and Sm in the mixed solutions was possible using three absorption bands and parameters characterizing molar absorptivity of these metals for accounting for the mutual interferences. Initially gamma- spectrometry and X-ray fluorescence methods were proposed for organization of on-line control of liquid spent nuclear fuel processing [44]. Later on, several applications of the optical methods such as Raman [45, 46], UV–Vis/NIR [47] and UV–Vis [48, 49] spectroscopy to the detection of inorganic and organic compounds including actinides have been reported. Microelectro-mechanical system (MEMS) has been developed and tested for the spectrophotometric detection of neodymium in the concentration range from 18 to 230 g/L at the constant background of 3 M nitric acid [47]. Spectrophotometric determination of uranium in the presence of varying concentration of nitric acid required use of two wavelengths to account for the dependence of the uranyl molar absorptivity on the nitrate concentration [48]. Use of two bands at 426 and 416 nm allowed for determination of both uranyl in the concentration range from 20 to 200 g/L and nitrate in the concentration range from 3 to 5 M with relative standard deviations of 5 and 15%, respectively. Detailed study of the effect of nitrate on uranyl speciation resulting in the significant changes in the U(VI) visible spectra with changing nitrate concentrations was reported in. Alterations of uranyl UV–Vis spectra in the concentration ranges from 0.01 to 1.26 M of uranium and from 0.01 to 8 M of nitric acid were studied, however, no
465
attempt to quantify uranium and nitrate was made. Similar research was presented in another work [50]. The potential of UV–Vis spectroscopy can be quite limited in comparison with other techniques, especially when resulting spectra are treated in traditional univariate way. UV–Vis spectroscopy and Time Resolved Laser induced Fluorescence Spectroscopy (TRLFC) have been compared in Ref. [51] to determine the uranium concentration in nitric acid solutions. The authors concluded that both methods are capable of achieving this goal though at different concentration levels: linearity range was ca. from 1 mM to 0.2 M for UV–Vis and from 0.09 to 6 mM for TRFLC. They also noted that TRLFC was superior in that it could qualitatively detect uranium presence in the solutions in the absence of calibration. This conclusion, however, was made without using any chemometric methods for spectra treatment. Further limitations of univariate approach were evidenced [52, 53]. Optic fiber aided method was proposed in the work [52] for determination of ruthenium in the solutions mimicking aqueous streams of nuclear reprocessing. While nitric acid and other metals as zirconium and strontium did not interfere with analysis, negative influence of uranium was observed. Thus, uranium has to be separated from solution prior to ruthenium determination, which renders this approach inapplicable to on-line analysis. Direct detection of U, Nd and Pd in 3 M nitric acid solutions by reflection absorption spectrophotometry was discussed in Ref. [53]. Unfortunately, the authors did not account for the influence of nitric acid concentration on the metals’ spectra. It is, however, well known that bands in uranium spectrum are sensitive to nitric acid concentration. It was therefore becoming obvious early on, that use of single wavelength for quantification of the constituents of spent nuclear fuel is not feasible due to the interferences. Besides influence of nitrate on the uranyl speciation and overlapping spectra of rare earth metals discussed above, such effects as redox reactions of actinides [54, 55] and formation of cation–cation complexes [56] affect speciation of the metals and their UV–Vis spectra. Selectivity can be improved by prior separation of the metals in the mixtures or by use of ligands capable of forming colored complexes with metals such as e.g. arsenazo. It should be noted, though, that addition of reagents to the sample is hardly compatible with on-(in-)line measurements. Alternative approach can consist in the use of chemometrics instead of ligands for the resolution of overlapping optical spectra. Recently reported application of the UV–Vis spectroscopy together with chemometric data treatment to the analysis of the nuclear technological solutions are discussed below. Early attempts to analyse actinides and nitric acid in aqueous (and organic) solution in on-line mode were taken
123
466
in the 80-th, when fiber optic technic became available [57]. For example, report about uranium detection in aqueous and organic phases by UV–Vis spectroscopy in extraction experiments in hot cells was presented in 1986 [58]. However the accuracy of results was quite poor, because mathematical treatment of spectra was not employed. Multivariate approach: detection using whole spectra and chemometrics First attempt to combine UV–Vis spectroscopy with chemometrics for analysis of actinides dates back to 1988. In the studies [59, 60] UV–Vis spectroscopy with optic fiber and PLSR for data treatment was tested for simultaneous determination of uranium and nitrate. PLSR modelling was used for the analysis of the model solutions of uranium (VI) [61] and plutonium(III)/(IV) [20] in nitric acid. Calibration models for determination of both uranium(VI) and nitrate were made using absorption spectra in the spectral range from 380 to 500 nm. Concentration ranges typical for the PUREX process were selected and were of 1–30 gL-1 for uranium and 0.02–3 molL-1 for nitric acid. Prediction of uranium(VI) and nitrate was possible with Root Mean Square Errors (RMSE) of 0.2 gL-1 and 0.07 molL-1, respectively. Similar approach was applied to the simultaneous quantification of Pu(III) and nitric acid [58]. Furthermore, influence of different concentrations of nitric acid, oxalate and fluoride on visible spectra of Pu(IV) have been studied using PCA and Nearest Neighbor classification. Interest to the applications of chemometric techniques to the processing of the UV–Vis spectra of aqueous solutions of spent nuclear fuel has resurfaced recently and a series of reports on this topic have been published [62, 63]. UV–Vis spectroscopy in combination with PLSR was applied to the quantification of Np and Pu in the model solutions simulating feeds obtained from the PUREX process [64]. All model solutions contained fixed background of uranyl (50 g/L) and nitric acid (2 M), while concentration of Np and Pu changed in the range from 0.05 to 1.55 g/L. Mean Relative Errors (MREs) of determination of neptunium and plutonium in the tests were 7.1 and 4.4%, respectively. This work was continued in the model solutions of more complex composition and employing both UV–Vis and IR spectra for the actinides’ quantification [65]. The background solution of aqueous phase contained all typical fission products in their common concentration. Determination of uranium, neptunium and plutonium both in aqueous and organic phases was performed using optical spectra and PLSR. The content of all key components can be determined with mean relative errors not exceeding 10%.
123
J Radioanal Nucl Chem (2017) 312:461–470
Analysis of uranium, nitric acid and nitrous acid was studied in Ref. [66] using spectra acquired in the range from 300 to 500 nm and MLR for calibration. Quantification of all three analytes was possible not only in aqueous but also U(VI)–nitric acid–30% TBP/kerosene system. The authors of this study point out the advantages of on-line monitoring by spectroscopy: ‘‘No need for radiochemical separation process; absence of radioactive liquid-waste from the analytical process; simplification of the analytical process and decrease of radiation levels of the analyze and reliability of measurement results as the form of samples will not be changed by radiochemical separation.’’ Raman and Vis–NIR spectroscopy have been applied to the detection of uranyl nitrate, plutonium in both IV and V oxidation states, neptunium(V) and neodymium in the spent nuclear fuel [61]. Aqueous solutions of the commercial fuel ATM-109 and ATM-105 and model feed solutions were prepared on varying background of nitric acid and extracted using dodecane solution of tributylphosphate. Spectra of resulting aqueous and organic phases were recorded. Model feed solutions were employed for identification of the characteristic bands of each of the solution constituents. Raman spectra were used of quantification of uranyl and nitrate while Vis–NIR spectra for determination of total Pu, Np(V) and Nd(III). Results were found to be in agreement with ICP-MS measurements. In the following work, possibility of determination of Pu(IV), Pu(VI), Np(V), the Np(V)-U(VI) cation–cation complex, and Nd(III) in fuel solutions using Vis–NIR spectra was demonstrated [67]. Similar tasks can be addressed also using Vis–NIR and Raman spectroscopy. For example, the analysis of aqueous solutions containing different concentrations of nitric acid, sodium nitrate and neodymium at varying temperatures was studied in Ref. [68]. A model based on the PLSR and spectroscopic data was developed and its capability to follow up changes of the chemical composition of the process stream was demonstrated. While both Raman and NIR spectra allowed quantification of NO3- and H? concentrations, only NIR was usable for determination of Nd(III). In the reports [69, 70] the method was suggested for strong acid solution monitoring using Raman spectroscopy. A step further has been undertaken in the work [71] bringing application of spectroscopy closer to the real process control in nuclear technology. Measurements were made in the aqueous and organic phases in the countercurrent solvent extraction testing set-up equipped with optical Vis–NIR and Raman probes, which were installed inside centrifugal contactors. A series of solutions, sequentially increasing nitric acid and uranyl nitrate concentrations, were fed into the centrifugal contactor flow loop system and spectra were collected upon each concentrations change. Calibration models were produced
J Radioanal Nucl Chem (2017) 312:461–470
467
Fig. 1 a Raw optical spectra recorded during the reduction process in Pu(IV) solution. Color corresponds to process time: red—start (t = 0 min), blue—finish (t = 5 h). b Example of restored concentration profiles reduction of Pu and Np. (Color figure online)
using spectra measured in the model solutions. Combination of Raman spectroscopy techniques with chemometrics allowed to perform online and real time monitoring of changes in the concentrations of uranyl nitrate, nitric acid and total nitrate in the concentration range from 0 to 0.6 M, 0 to 6 M and 0 to 7.2 M, respectively. In the first tests involving online measurements in the laboratory equipment, which were reported in these works, simple solutions containing only nitric acid and uranyl nitrate were used. Vis–NIR probe was also installed in the testing equipment and spectra of other spent nuclear fuel compounds in simulated solutions and respective calibration models have been established. Thus, applications of both Raman and Vis–NIR spectroscopy for online monitoring of spent nuclear fuel processing is the next logical step. Unfortunately, in the works discussed above data treatment is described very briefly and schematically. Apart from mentioning what methods were used, very little information is provided. Such important issues as spectra pre-treatment, temporary drift, validation of calibration models, were not addressed at all.
Another research was aimed at on-line determination of actinides (Np and Pu) during their reduction at certain PUREX stage [72] The experimental solutions were closely mimicking the composition of real industrial media and spectral measurements (184–1000 nm range) were taken during the reduction of metals with hydrazine, which leads to gas bubbles formation and disturbs the spectra significantly. Figure 1 a shows the typical view of the spectra obtained during Pu(IV) to Pu(III) reduction and one can see baseline curvature, strong baseline drift, excessive noise and out of scale spectrometer readings. To make these data suitable for further processing the authors applied asymmetrical least squares procedure. Consecutive application of MCR-ALS allowed for getting the concentration profiles of reduced and oxidized actinides during the process (Fig. 1 b). Reprinted from Chemometrics and Intelligent Laboratory Systems, vol. 146, B. Debus, D. Kirsanov, C. Ruckebusch, M. Agafonova-Moroz, V. Babain, A. Lumpov, A. Legin, pp. 241–249, Copyright [72] with permission from Elsevier.
123
468
A quite challenging problem of simultaneous quantification of cerium and nitric acid in aqueous solutions was addressed in Ref. [73]. The complexity of this case is due to the fact, that in the studied wavelength range (200–1000 nm) cerium has only one broad absorbance band which is overlapping with nitric acid signal and thus Lambert–Beer law is not suitable for direct quantification of the components. This issue invalidates the use of conventional PLSR and initial attempt of cerium quantification failed [32], so the authors have suggested a suitable alternative based on Non-Linear Multivariate Curve Resolution (NL-MCR), which yielded accurate determination of cerium and nitric acid concentrations in new samples.
Conclusion UV–Vis spectroscopy is a very simple, accessible and convenient analytical instrument for quantitative detection of metals. This can be done in two different ways: using chromogenic ligands and using characteristic absorption of an analyte. In the context of process control in nuclear industry the latter option appears to be very attractive, since it makes UV–Vis spectroscopy fully applicable for on-line monitoring purposes. With the development of optic fiber spectroscopic probes the benefits of UV–Vis for nuclear industry are even more pronounced as fiber probes allows for remote measurements and radiation exposure minimization. UV–Vis spectroscopy can be criticized for low sensitivity and low selectivity, however when dealing with real technological solutions the first drawback is not relevant due to sufficient concentration of most of analytes in technological streams and the second drawback caused by overlapping of spectral signatures of different components can be elegantly circumvented by the application of modern chemometric tools (the latter however requires the fulfillment of certain conditions, formulated by Manne [74]). The literature survey shows the distinct growing interest to this combination of UV– Vis and chemometrics especially in the nuclear technology research domain. While at the moment most of the applications were performed in controlled laboratory environment we see no essential limitations for implementation of this approach at industrial sites. It is important to note that most of the studies reported so far applied only very basic chemometric procedures like PCA and PLSR and even with these simple tools the results obtained are quite convincing. With transition to industrial sites and application of more sophisticated and tailored multivariate methods UV–Vis can definitely become one more success story of chemometrics along with NIR spectroscopy.
123
J Radioanal Nucl Chem (2017) 312:461–470 Acknowledgement This work was partially financially supported by Government of Russian Federation (Grant 074-U01).
References 1. Wold S (1972) Spline functions, a new tool in data-analysis. Kem Tidskr 3:34–37 2. Wold S, Sjo¨stro¨m M (1998) Chemometrics, present and future success. Chemom Intel Lab Syst 44:3–14 3. Wetzel D (1983) Near-infrared reflectance analysis. Anal Chem 55:1165A–1176A 4. Blanco M, Villarroya I (2002) NIR spectroscopy: a rapid-response analytical tool. Trends Anal Chem 21:240–250 5. Roggo Y, Chalus P, Maurer L, Lema-Martinez C, Edmond A, Jent N (2007) A review of near infrared spectroscopy and chemometrics in pharmaceutical technologies. J Pharmaceut Biomed 44:683–700 6. Jimare´ Benito M, Bosch Ojeda C, Sanchez Rojas F (2008) Process analytical chemistry: applications of near infrared spectrometry in environmental and food analysis—an overview. Appl Spectrosc Rev 43:452–484 7. Andrade-Garda JM (ed) (2009) Basic chemometric techniques in atomic spectroscopy. RSC Publishing, London 8. Henry R, Koller D, Liezers M, Farmer OT III, Barinaga C, Koppenaal D, Wacker J (2001) New advances in inductively coupled plasma–mass spectrometry (ICP-MS) for routine measurements in the nuclear industry. J Radioanal Nucl Chem 249:103–108 9. Pathak AK, Kumar R, Singh VK, Agrawal R, Rai S, Kumar Rai A (2012) Basic chemometric techniques in atomic spectroscopy. Appl Spectrosc Rev 47:14–40 10. Downey G (1998) Food and food ingredient authentication by mid-infrared spectroscopy and chemometrics (review). Trends Anal Chem 17:418–424 11. Łobin´ski R, Marczenko Z (1992) Recent advances in ultravioletvisible spectrophotometry. Crit Rev Anal Chem 23:55–111 12. Otto M, Wegscheider W (1985) Spectrophotometric multicomponent analysis applied to trace metal determinations. Anal Chem 57:63–69 13. Bebee K, Kowalski B (1987) An introduction to multivariate calibration and analysis. Anal Chem 59:1007 14. Thomas E, Haaland D (1990) Comparison of multivariate calibration methods for quantitative spectral analysis. Anal Chem 62:1091–1099 15. Vitouchova´ M, Jancˇa´ˇr L, Sommer L (1992) Interaction of iron(II) and the simultaneous spectrophotometric determination of Fe, Cu, Zn, Co and Ni with 2-(5-bromo-2-pyridylazo)-5-diethylaminophenol. Fresenius J Anal Chem 343:274–279 16. Rodriguez A, de Torres A, Pavon J, Ojeda C (1993) Simultaneous spectrophotometric determination of cadmium, copper and zinc. Talanta 40:1861–1866 17. Rodriguez A, de Torres A, Pavon J, Ojeda C (1998) Simultaneous determination of iron, cobalt, nickel and copper by UV–visible spectrophotometry with multivariate calibration. Talanta 47:463–470 18. Iida Y (1987) Repetitive spectral subtraction method for the spectrophotometric determination of rare earth elements. Fresenius J Anal Chem 328:547–552 19. Carey W, Wangen L (1989) Spectrophotometric method for the analysis of plutonium and nitric acid using partial least-squares regression. Anal Chem 61:1667–1669 20. Carey W, Wangen L (1991) Determining chemical characteristics of plutonium solutions using visible spectrometry and
J Radioanal Nucl Chem (2017) 312:461–470
21.
22. 23. 24.
25.
26.
27.
28.
29.
30. 31.
32.
33. 34. 35.
36.
37.
38.
39. 40. 41.
multivariate chemometric methods. Chemom Intell Lab Syst 10:245–257 Peralta-Zamora P, Cornejo-Ponce L, Nagata N, Poppi R (1997) Chemometric alternatives for resolution of classical analytical problems. Spectrophotometric determination of lanthanide mixtures. Talanta 44:1815–1822 Meinrath G (1998) Chemometric analysis: uranium(VI) hydrolysis by UV–Vis spectroscopy. J Alloy Compounds 277:777–781 Meinrath G (1998) Direct spectroscopic speciation of schoepiteaqueous phase equilibria. J Radioanal Nucl Chem 232:179–188 Meinrath G, Lis S, But S, Elbanowski M (2001) Chemometric and statistical analysis of polyoxometalate interaction with lanthanide(III) ions. Talanta 55:371–386 Meinrath G, Lis S, Elbanowski M (2004) Spectroscopy, chemometrics and metrology—three aspects of lanthanide chemistry. J Alloy Compounds 380:413–417 Meinrath G, Lis S, Bo¨hme U (2006) Quantitative evaluation of Ln(III) pyridine N-oxide carboxylic acid spectra under chemometric and metrological aspects. J Alloy Compounds 412:962–969 Kaczmarek M, Meinrath G, Lis S, Kufelnicki A (2008) The interaction of arsenazo III with Nd(III)—a chemometric and metrological analysis. J Solut Chem 37:933–946 Lis S, Meinrath G, Glatty Z, Kubicki M (2010) Spectroscopic speciation and structural characterisation of uranyl(VI) interaction with pyridine carboxylic acid N-oxide derivatives. Inorg Chim Acta 363:3847–3855 Ni Y, Wu Y (1999) Spectrophotometric determination of europium, terbium and yttrium in a perchloric acid solution by the Kalman filter approach. Anal Sci 15:1123–1127 Haswell S, Walmsley A (1999) Chemometrics: the issues of measurement and modelling. Anal Chim Acta 400:399–412 Wang L, Wang X, Wang Y (2013) Structure of a piperidinemodified calix[4]arene derivative and spectral resolution of its interaction with rare earth metals with chemometric methods. Spectrochim Acta A 105:62–66 Rodionova O, Tikhomirova T, Pomerantsev A (2015) Spectrophotometric determination of Rare Earth Elements in aqueous nitric acid solutions for process control. Anal Chim Acta 869:59–67 Baumga¨rtner F, Ertel D (1980) The modern PUREX process and its analytical requirements. J Radioanal Chem 58(1–2):11–28 Ache H (1992) Analytical chemistry in nuclear technology. Fresenius J Anal Chem 343:852–862 Pomerantsev A, Rodionova O (2012) Process analytical technology: a critical view of the chemometricians. J Chemom 26:299–310 Richter S, Goldberg S (2003) Improved techniques for high accuracy isotope ratio measurements of nuclear materials using thermal ionization mass spectrometry. Int J Mass Spectrom 229:181–197 Chartier F, Aubert M, Pilier M (1999) Determination of Am and Cm in spent nuclear fuels by isotope dilution inductively coupled plasma mass spectrometry and isotope dilution thermal ionization mass spectrometry after separation by high-performance liquid chromatography. Fresenius J Anal Chem 364:320–327 Betti M (1997) Use of ion chromatography for the determination of fission products and actinides in nuclear applications. J Chromatogr A 789:369–379 Benedict M, Pigford T, Levi H (1981) Nuclear chemical engineering, 2nd edn. McGraw-Hill, New York, pp 457–564 Mathur J, Murali M, Nash K (2001) Actinide partitioning—a review. Solv Extr Ion Exch 19:357–390 Tachimori S, Morita Y (2009) Overview of solvent extraction chemistry for reprocessing. Ion exchange & solvent extraction: a series of advances, vol 19. CRC Press, Boca Raton, pp 1–63
469 42. Bostick D (1978) The simultaneous analysis of uranium and nitrate. ORNL/TM-6292, Oak Ridge National Laboratory Report. http://www.osti.gov/scitech/biblio/5080075, Accessed 25 Jan 2017 43. Rodden C (1941) Spectrophotometric determination of praseodymium, neodymium, and samarium. J Res Nat Bur Stand 26:557–570 44. Parus J, Kierzek J, Zoltowski T (1977) Online control of nuclear fuel reprocessing. Nukleonika 22:759–776 45. Madic C, Hobart D, Begun G (1983) Raman spectrometric studies of actinide(V) and actinide(VI) complexes in aqueous sodium-carbonate solution and of solid sodium actinide(V) carbonate compounds. Inorg Chem 22:1494–1503 46. Madic C, Begun G, Hobart D, Hahn R (1984) Raman spectroscopy of neptunyl and plutonyl ions in aqueous solution: neptunium(VI) and plutonium(VI) and disproportionation of plutonium(V). Inorg Chem 23:1914–1921 47. Colston B, Choppin G (2001) Evaluating the performance of a stopped-flow near-infrared spectrophotometer for studying fast kinetics of actinide reactions. J Radioanal Nucl Chem 251:21–26 48. Janssens-Maenhout G, Nucifora S (2007) Feasibility study of a microsystem to analyse radioactive solutions. Nucl Eng Design 237:1209–1219 49. Warburton J, Smith N, Czerwinski K (2010) Method for online process monitoring for use in solvent extraction and actinide separations. Sep Scie Technol 45:1763–1768 50. Lascola R, Livingston R, Sanders M, McCarty J, Dunning J (2002) Online spectrophotometric measurements of uranium and nitrate concentrations of process solutions for Savannah River Site’s H-Canyon. J Process Anal Chem 7:14–20 51. Smith N, Cerefice G, Czerwinski K (2013) Fluorescence and absorbance spectroscopy of the uranyl ion in nitric acid for process monitoring applications. J Radioanal Nucl Chem 295:1553–1560 52. Fujii T, Egusa S, Uehara A, Yamana H, Morita Y (2013) Quantitative analysis of neodymium, uranium, and palladium in nitric acid solution by reflection absorption spectrophotometry. J Radioanal Nucl Chem 295:2059–2062 53. Ganesh S, Velavendan P, Pandey N, Kamachi Mudali U, Natarajan R (2013) Direct spectrophotometric determination of ruthenium in aqueous streams of nuclear reprocessing. Radioanal Nucl Chem 295:2091–2094 54. Fukasawa T, Kawamura F (1991) Photochemical reactions of neptunium in nitric acid solution containing photocatalyst. J Nucl Scie Technol 28:27–32 55. Precek M, Paulenova A, Mincher B (2012) Reduction of Np(VI) in irradiated solutions of nitric acid. Proced Chem 7:51–58 56. Guillaume B, Hobart D, Bourges J (1981) Cation-cation complexes of pentavalent actinides 2. Spectrophotometric study of complexes of Am(V) with U022? and Np022? in aqueous perchlorate solution. J Inorg Nucl Chem 43:3295–3299 57. Boisde G, Perez J (1984) Remote Spectrometry With Optical Fibers, Ten Years Of Development And Prospects For On-Line Control. Proceedings SPIE 0514, 2nd international conference on optical fiber sensors: OFS’84, 227 (November 21, 1984); http:// dx.doi.org/10.1117/12.945088, Accessed 25 Jan 2017 58. Moser D, Klatt L (1986) Application of in-line photometer to solvent extraction process control, Control and instrumentation http://www.iaea.org/inis/collection/NCLCollectionStore/_Public/ 18/050/18050429.pdf, Accessed 25 Jan 2017 59. Van Hare D, O’Rourke P, Prather W (1988) Online fiber optic spectrophotometry (No. DP-MS-88-186; CONF-881143-1) Savannah River National Laboratory Report, Aiken, USA, http:// www.osti.gov/scitech/biblio/6710161, Accessed 25 Jan 2017 60. O’Rourke P, Van Hare D, Prather W (1992) On-line process control monitoring system. US Patent 5131746
123
470 61. Biirck J (1991) Spectrophotometric determination of uranium and nitric acid by applying partial least squares regression to uranium(VI) absorption spectra. Anal Chim Acta 254:159–165 62. Bryan S, Levitskaia T (2007) Monitoring and control of Purex radiochemical processes. Proceedings of international conference GLOBAL-2007, Boise, Idaho. http://toc.proceedings.com/ 02031webtoc.pdf, Accessed 25 Jan 2017 63. Bryan S, Levitskaia T, Casella A, Peterson J, Johnsen A, Lines A, Thomas E (2011) In: Nash KL, Lumetta GJ (eds) Advanced separation techniques for nuclear fuel reprocessing and radioactive waste treatment. Woodhead Publishing, Sawston 64. Kirsanov D, Babain V, Agafonova-Moroz M, Lumpov A, Legin A (2012) Combination of optical spectroscopy and chemometric techniques–a possible way for on-line monitoring of spent nuclear fuel (SNF) reprocessing. Radiochim Acta 100:185–188 65. Kirsanov D, Babain V, Agafonova-Moroz M, Lumpov A, Legin A (2013) Approach to on-line monitoring of PUREX process using chemometric processing of the optical spectral data. Radiochim Acta 101:149–154 66. Li L, Zhang H, Ye G (2013) Simultaneous spectrophotometric determination of uranium, nitric acid and nitrous acid by leastsquares method in PUREX process. J Radioanal Nucl Chem 295:325–330 67. Bryan S, Levitskaia T, Johnsen A, Orton C, Peterson J (2011) Spectroscopic monitoring of spent nuclear fuel reprocessing streams: an evaluation of spent fuel solutions via Raman, visible, and near-infrared spectroscopy. Radiochim Acta 99:563–572 68. Nee K, Bryan S, Levitskaia T, Nilsson M (2013) Spectroscopic and physicochemical measurements for on-line monitoring of
123
J Radioanal Nucl Chem (2017) 312:461–470
69.
70.
71.
72.
73.
74.
used nuclear fuel separation processes. Proceedings of international conference GLOBAL-2013, Salt Lake City, Utah, September 29–October 3, 2013, pp. 931–935, http://toc.proceed ings.com/21109webtoc.pdf, Accessed 25 Jan 2017 Casella A, Levitskaia T, Peterson J, Bryan S (2013) Water O–H stretching Raman signature for strong acid monitoring via multivariate analysis. Anal Chem 85:4120–4128 Casella A, Ahlers L, Campbell E, Levitskaia T, Peterson J, Smith F, Bryan S (2015) Development of online spectroscopic pH monitoring for nuclear fuel reprocessing plants: weak acid schemes. Anal Chem 87:5139–5147 Bryan S, Levitskaia T, Casella A, Peterson J (2013) Spectroscopic online monitoring for process control and safeguarding of radiochemical fuel reprocessing streams. In the proceedings of WM2013 conference, Phoenix, Arizona USA, February 24 – 28, 2013, http://www.wmsym.org/archives/2013/papers/13553.pdf, Accessed 25 Jan 2017 Debus B, Kirsanov D, Ruckebusch C, Agafonova-Moroz M, Babain V, Lumpov A, Legin A (2015) Restoring important process information from complex optical spectra with MCR-ALS: Case study of actinides reduction in spent nuclear fuel reprocessing. Chemom Intel Lab Syst 146:241–249 Rodionova O, Pomerantsev A (2016) Non-linear multivariate curve resolution applied to the spectrophotometric determination of cerium(III) in aqueous nitric acid solutions for process control. Anal Methods 8:435–440 Manne R (1995) On the resolution problem in hyphenated chromatography. Chemom Intel Lab Syst 27:89–94