Rapid determination of Cu, Fe, Mg, Mn and Zn in wood pulp by direct sample insertion-inductively coupled plasma-optical emission spectrometry using a pyrolytically coated graphite sample probe Michael E. Rybak,a Panos Hatsis,b Kevin Thurbideb and Eric D. Salina aDepartment of Chemistry, McGill University, 801 Sherbrooke St. W., Montre�al, Que�bec, Canada H3A 2K6 bPulp and Paper Research Centre, McGill University, 342 University St., Montre�al, Que�bec, Canada H3A 2A7 Received 2nd March 1999, Accepted 23rd August 1999 A rapid method for screening wood pulp samples by direct sample insertion-inductively coupled plasma-optical emission spectrometry (DSI-ICP-OES) is described.Solid wood pulp samples were introduced directly into an inductively coupled plasma, using a pyrolytically coated graphite DSI sample probe, after in situ chemical treatment with HCl and NaF. Drying and ashing steps were performed by inductively heating the sample probe in the ICP coil prior to plasma ignition.The analysis time of the method from sample acquisition to analysis was of the order of several minutes per sample, as compared to several hours when conventional dissolution methods are used. Agreement with reference values for wood pulp samples ranged from 3.4±16% (absolute) for high-concentration analytes (Mg, Mn) and 1.7±50% (absolute) for low-concentration ones (Cu, Fe, Zn) using external standards. Precision ranged from 6±50% RSD and was highly dependent on the element and pulp sample studied.Absolute detection limits for the method were of the range of 50±1000 pg, translating into relative detection limits of 20±400 ppb based on a 2.5 mg pulp sample. The merits of using DSI-ICP-OES for the direct analysis of wood pulps, and of using a pyrolytically coated graphite probe for this type of application, are discussed. Introduction The detrimental environmental ramiÆcations from bleaching wood pulps with chlorinated reagents have recently led to increased regulatory pressure to use a totally chlorine free (TCF) bleaching process in the pulp and paper industry.1 In TCF bleaching, hydrogen peroxide (H2O2), by means of its alkaline reacting species (HOO2) and decomposition intermediates (HO? and O2 2?), is used to delignify and brighten the pulp. The decomposition of H2O2 is integral to the deligniÆcation and bleaching processes, but it must be carefully controlled in order to accomplish TCF bleaching efÆciently.While the hydroperoxy anion (HOO2) is primarily responsible for the brightening of the pulp, the hydroxide (HO?) and superoxide (O2 2?) radicals account for much of its deligniÆcation. These radicals, however, only show marginal selectivity towards lignin over cellulose, and destruction of the cellulose results in a lower yield and a weaker pulp. To further complicate matters, certain transition metal species (e.g., MnO2, Mn2z, Cu2z and Fe2z) are known to accelerate H2O2 decomposition, whereas other species (e.g., Mg2 z, SiO3 22) will inhibit this acceleration. 2 The aforementioned metal species are commonly present in wood, and because of their inØuence on hydrogen peroxide degradation, there exists an ideal metal content proÆle for effective TCF bleaching.3 The metal proÆle of the pulp sample may be adjusted either by chelation of the metals with ethylenediaminetetraacetate (EDTA) or diethylenetriaminepentaacetate (DTPA), or by washing the pulp at a low pH (1.5± 3.0) followed by replenishment of the magnesium ion.4 With the metal content proÆle of the pulp having such a great inØuence on the TCF bleaching process, and the adjustment of this proÆle a common practice in the paper industry, there exists a need for a means by which the levels of metals present in the pulp can be determined with reasonable speed and accuracy.Magnesium and manganese, the two most important elements in terms of their inØuence on the TCF bleaching process, are found in relatively high concentration in pulps.Typical concentration ranges are 200±400 ppm for Mg and 50± 250 ppm for Mn in Canadian kraft (chemically treated) pulps. Other elements that occur in lower concentrations include Cu and Zn (0±10 ppm) and Fe (20±100 ppm). The desired metal proÆle for a pulp destined for TCF bleaching is such that the Mg level is maintained at least within the natural range expected, and that the concentration of the transition metals is reduced to its lowest level possible (of the order of 1 ppm or less).Considering the range of metal concentrations expected to occur naturally, and the thresholds for these metals deemed acceptable for TCF bleaching, the desired techniques for determining these analytes in wood pulp should have at least semi-quantitative capabilities, with quantitative results for the most inØuential elements being desirable. The current method used by the paper industry for the determination of these metals in pulp samples involves dry or wet ashing of the sample followed by hot-plate digestion in HCl, and subsequent analysis by Øame atomic absorption spectrometry (FAAS).5,6 The use of inductively coupled plasma-optical emission spectrometry (ICP-OES) in the above methods in place of FAAS is now a common practice in the pulp and paper industry, primarily because of the multielement capabilities of ICP-OES.Although they are commonplace for preparing solid samples for analysis, hot-plate digestions have several inherent disadvantages: volatile element losses; contamination of the sample from air, contact with the sample vessel, or reagents required for sample digestion; and unacceptably long sample dissolution times.An expeditious alternative to dissolution of the solid pulp sample would be the direct analysis of the solid itself. Many solid samples have been successfully determined by ICP-OES by taking advantage of thermal sample introduction techniques such as electrothermal vaporizaton (ETV)7,8 or direct sample insertion (DSI).9±12 In J.Anal. At. Spectrom., 1999, 14, 1715±1722 1715 This Journal is # The Royal Society of Chemistry 1999many of these cases solids were introduced into the ICP by ETV or DSI with minimal a priori sample treatment and any necessary sample treatment was performed in situ. Of these two techniques, the open design of the DSI lends itself most conveniently to rapid replacement of the sample holder, and the addition of both solid samples and liquid reagents.Consequently, it was decided that the direct analysis of pulp samples would be approached using this technique. In its most conventional conÆguration, DSI entails the axial elevation of a sample directly into the center channel of the annular plasma discharge by means of a sample carrying probe. The intrinsic beneÆts of DSI are obvious: 100% of the sample is introduced into the excitation source, and cup-shaped sample probes facilitate the introduction of various solids and liquids, as well as the addition of reagents for in situ chemical sample treatment.Physical sample treatment steps, such as drying and pyrolysis, can also be performed either by proximate positioning of the sample underneath the plasma, or by induction heating in the ICP load coil prior to ignition of the plasma.13 Graphite cup DSI, like ETV, is hampered by the formation of refractory carbides, which prove difÆcult to volatilize, by sample intercalation into the pores and interstices of the graphite, which results in poor reproducibility in the volatilization event, and by the susceptibility of graphite to chemical attack. Intercalation can be minimized and a resistance to chemical attack can be imparted to a graphite surface by depositing a highly ordered layer of pyrolytic graphite.14 Recently, a means of depositing a pyrolytic graphite coating on the interior of a graphite DSI cup in the ICP was developed.15 Promising improvements in signal reproducibility and sensitivity were observed, but the performance of the new coated probe had yet to be evaluated in terms of its resistance to chemical attack and usefulness for routine analysispe of this study was two-fold: to evaluate the use of DSI as an expeditious means of screening wood pulps for metals that inØuence the efÆciency of the TCF bleaching process; and to evaluate the performance of pyrolytically coated graphite DSI probes for performing routine analyses with extensive in situ chemical treatment. Experimental Pyrolytically coated graphite probes and DSI apparatus Hollow-stemmed, long undercut graphite cup sample probes were machined in-house from J@ high density graphite electrodes (S-8 HD, Bay Carbon, Bay City, MI, USA) on a benchtop lathe (Emco Compact 5, Emco Maier, Columbus, OH, USA) according to the dimensions indicated in Fig. 1(a). The interior of the cup portion of the DSI probe was then pyrolytically coated with graphite directly in the plasma15 by means of a vapor phase deposition procedure, depicted schematically in Fig. 1(b). In brief, a 10% (v/v) mixture of methane in argon was directed through the hollow stem of the probe toward the walls of the cup interior as the cup portion of the probe was positioned in a 2 kW argon plasma. The methane undergoes gas-phase pyrolysis, large aromatic molecules are generated by dehydrogenation, and collision of these macromolecules with the substrate results in a pyrolytic graphite deposit.Details of the experimental conditions used in the coating process appear in Table 1, and Æne points pertaining to the pyrolytic coating procedure have been previously described.13 A stepper motor controlled direct sample insertion device (DSID)13 was used to elevate the DSI sample probes axially into a 27.12 MHz inductively coupled plasma source with an automatching network (HFP-2500 and AMN-2500E respectively, Plasma Therm, Inc., St.Petersburg, FL, USA). Optical emission signals collected from the ICP were imaged to the entrance slit of a Rowland circle-type polychromator (Model 750, Thermo Jarrell Ash, Franklin, MA, USA) equipped with a galvanically driven quartz refractor plate in the incident light Fig. 1 Pyrolytically coated direct sample insertion (DSI) probe: (a) dimensions of probe used; (b) depiction of the pyrolytic coating process.Table 1 Experimental summary Pyrolytic coating process– Plasma forward power 2 kW ReØected power 0±8 W Plasma gas Øow rate 16 l min21 Auxiliary gas Øow rate 2 l min21 Coating gas Øow rate 500 ml min21 Coating time 20 min Insertion depth 0 mm ATOLCa Sampling and sample pretreatment– Sample mass 1±4 mg of dried pulp Chemical treatment 20 ml of conc. HCl 10 ml of 10% (m/v) NaF External standards 20 ml of mixed element standard: Cu: 0.25±1.25 ppm Fe: 1±5 ppm Mg: 25±125 ppm Mn: 10±50 ppm Zn: 0.5±2.5 ppm Inductive drying forward power 50 W (y150 �C) Inductive drying reØected power 8±10 W Drying time 90 s Inductive pyrolysis forward power 150 W (y550 �C) Inductive pyrolysis reØected power 50±55W Pyrolysis time 90 s Drying/pyrolysis probe position 0 mm ATOLC Direct sample insertion analysis– Plasma forward power 2 kW ReØected power 0±5 W Plasma gas Øow rate 16 l min21 Auxiliary gas Øow rate 1.8 l min21 Insertion depth 0 mm ATOLC Viewing height 20 m ATOLC Insertion time 30 s Exposure time 40 ms per position Number of exposures per traceb 300 Galvanometer settle time 3 ms Wavelengths monitored 324.8 nm (Cu I)c 259.9 nm (Fe II) 279.6 nm (Mg II) 293.3 nm (Mn II) 213.9 nm (Zn I) aATOLC: Above top of load coil.bIncludes both on-peak and off-peak exposures. cOrigin of emission lines are denoted as I (ground state) or II (singly ionized). 1716 J. Anal. At. Spectrom., 1999, 14, 1715±1722path for off-peak background correction and capable of highspeed signal processing suitable for transient signals16 (Trulogic Systems, Mississauga, ON, Canada).Signals were interpreted using Grams/32 (Galactic Industries, Salem, NH, USA). Samples, standards and reagents Wood pulp samples used throughout this study were obtained from the Pulp and Paper Research Institute of Canada (Paprican, Pointe-Claire, QC, Canada). The unavailability of a certiÆed reference material for wood pulp necessitated the use of pulp samples that had been analyzed using a standardized method (Canadian Pulp and Paper Association Standard Method G.34P with ICP-OES, analyses performed by Paprican) for this study.Two pulp samples representative of the most common types encountered in routine wood pulp analysis were selected: a kraft pulp (brownstock) [Fig. 2(a)], in which alkaline attack is used to chemically fragment the lignin molecules of the wood chips (by cooking the chips in a solution of NaOH and Na2S at approximately 175 �C), and a thermomechanical pulp (TMP) [Fig. 2(b)], which is generated by pressurized steam pretreatment of wood chips followed by mechanical shredding and deÆbering of the chips by means of a rotary-disk reÆner.17 These same samples are currently being developed and characterized by Paprican for use as industry standard reference materials. Mixed element standard solutions were prepared by serial dilution of both multi-element and single-element standards (High-Purity Standards, Charleston, SC, USA, and Fisher ScientiÆc, Nepean, ON, Canada, respectively) with 0.5% trace metal grade HNO3 (Instra-Analyzed, J.T. Baker, Phillipsburgh, NJ, USA) in distilled, deionized water (Milli-Q water system, Millipore Corp., Bedford, MA, USA). Trace metal grade HCl (Instra-Analyzed) and ACS grade NaF (Baker- Analyzed, J. T. Baker) were used for in situ treatment of the pulp samples. Procedure An accurately determined mass of dried wood pulp in the range of 1±4 mg was deposited directly into a graphite DSI cup on a 0.01 mg readable balance.The probe was then mounted on the DSID, and 20 ml of concentrated HCl and 10 ml of a 10% (m/v) solution of NaF were added to the cup [Fig. 3(a)]. The DSI cup was then positioned such that the top of the cup was even with the top of the highest turn of the ICP load coil, i.e., 0 mm above the top of the load coil (ATOLC). Forward power was then applied to the ICP load coil at settings of 50 W for 120 s to dry the sample and 150 W for 120 s to ash the sample by inductively heating the graphite DSI probe [Fig. 3(b)]. These power settings have been determined to correspond to probe temperatures of approximately 150 �C and 550 �C, respectively. Although more conservative drying and pyrolysis times of the order of 30±60 s would have been sufÆcient for each respective task, longer times were employed as a conservative approach. The sample probe was then retracted below the ICP coil and the plasma was ignited and set to a forward power of 2.0 kW [Fig. 3(c)]. The probe was then elevated by the stepper motor until it was just below the plasma discharge (220 mm ATOLC) and stopped there for 2 s prior to insertion so that the plasma could recover from disruption in the Ar Øow caused by the initial probe elevation. Finally, the probe was inserted into the plasma [Fig. 3(d)] for 30 s at a position of 0 mm ATOLC, retracted and cooled.Signal acquisition was started upon the arrival of the probe at the stabilization position. Details regarding other instrumental settings appear in Table 1. Unless explicitly stated otherwise, it can be assumed that the procedure described above was used throughout the study. External standardization was used to determine the concentration of the metals present in the pulp samples. Calibration curves were constructed from signals acquired using aqueous multi-element standards. The standard solutions were analyzed in a manner identical to that described above for the pulp samples, including drying and pyrolysis steps, except that 20 ml of standard were deposited in the sample probe in place of the pulp sample. Fig. 2 Pulp samples used in this study: (a) kraft pulp (brownstock); (b) thermomechanical pulp (TMP). Fig. 3 Schematic depiction of the DSI procedure: (a) 10 ml of 10% (m/v) NaF and 20 ml of conc. HCl are added to a pulp sample of known mass; (b) inductive drying (50W forward power) and pyrolysis (150 W) ated sample; (c) retraction of the sample probe and ignition of the 2.0 kW ICP; (d) insertion of the sample.J. Anal. At. Spectrom., 1999, 14, 1715±1722 1717Results and discussion Method development In developing the procedure used for the direct analysis of wood pulps by DSI, the wood pulp was at Ærst analyzed using no chemical pretreatment, and drying and pyrolysis steps were performed only out of necessity so as to prevent plasma overloading.Although a transient signal could be obtained by this simple procedure it was found to be highly irreproducible (when corrected for sample mass) and the signals would evolve over a relatively long period of time, often on the order of 5±15 s for most elements. Since HCl is used in the pulp and paper industry standard solution dissolution methods5,6 to treat the pulp sample after ashing, the addition of an aliquot of concentrated HCl was incorporated into the DSI procedure. No appreciable improvement in signal was observed when HCl was added to the DSI cup with the deposited pulp sample after the drying and ashing steps. When HCl addition preceded the desolvation and pyrolysis events, however, the time over which the signals would evolve was diminished, although the signals were still irregular in appearance and poor in terms of reproducibility.Sodium Øuoride, as well as other halidecontaining solids and gases, are known to act as halogenating agents, forming relatively volatile metal±halide compounds. Agents such as these have been used to improve the volatilization of analytes in DSI from graphite probes, especially for refractory carbide and oxide forming elements.18 In experiments in the present study in which NaF was added, improved analyte volatilization was observed when addition was incorporated prior to drying and ashing of the pulp sample.The best results in terms of signal appearance and reproducibility, however, were achieved when HCl and NaF were added together before sample drying and pyrolysis.Fig. 4 shows the signals obtained from a brownstock sample inserted using a pyrolytically coated DSI probe when 20 ml of concentrated HCl and 10 ml of 10% (m/v) NaF are added prior to drying and ashing. Most analytes were completely vaporized in less than 5 s, with extremely volatile elements such as Zn being completely volatilized in 2 s.Iron, however, was very difÆcult to volatilize, with analyte still being vaporized from the probe 25 s after insertion into the plasma (Fig. 4). This is not a fault of the analysis procedure, but rather a shortcoming of DSI as an analytical technique. The longer time needed to vaporize Fe results simply because the temperature at which Fe is volatilized is relatively high in comparison to the other elements studied, and the maximum temperature that the sample probe can attain upon insertion, close to 2000 �C, was insufÆcient for rapid volatilization. The maximum temperature that a DSI probe can reach upon insertion can be increased by using a mixed-gas plasma, such as an oxygen±argon plasma,19 but the use of such conditions will result in consumption of the sample probe. When solid samples are analyzed by DSI, the reproducible deposition of a given mass of solid sample for each assay is often not a practicality, thus necessitating the incorporation of an acceptable sample mass range for insertion analysis.In this study, a range of 1±4 mg was established as suitable for the mass of pulp that could be used in DSI. The rationale behind this directive takes the following points into consideration. Since sample mass was being determined on a balance readable to 0.01 mg, the use of samples less than 1 mg would greatly compromise the precision to which the pulp mass could be determined and thus degrade the overall precision of the technique.The metals of interest present in wood pulp occur often at concentrations of 1 ppm or higher, meaning that the detection limit of the technique used would have to be at least 1 ng absolute if a 1 mg pulp sample was used. This is well above the detection limits that have been previously demonstrated with the instrumentation used in this study.13 With at least 1 mg of pulp sample, transient signals were visually discernible from the background emission for all analytes of interest in both the brownstock (Fig. 4) and TMP samples. For the mass range of 1±4 mg, a linear response of signal area as a function of sample mass was observed for the analytes of interest. Fig. 5 shows this behavior for Mg and Mn, the highest-concentration analytes in both pulp samples. Although larger sample masses could be determined with greater precision, non-linearity in the analyte transient signal area as a function of pulp mass was observed at higher sample masses (w8 mg).Probe comparison Fig. 6 shows the signals obtained with both pyrolytically coated and uncoated DSI probes for a Øuffed softwood TMP sample. As expected, the pyrolytically coated probe yielded sharp transient signals with little to no multiple peaking as compared to the uncoated probe. Since the uncoated probe surface is thinner, highly irregular and more porous in comparison to the pyrolytically coated surface, it was more Fig. 4 Typical signals for a kraft pulp (brownstock) sample treated with 10 ml of 10% (m/v) NaF and 20 ml of concentrated HCl using a pyrolytically coated DSI probe.Sample mass is approximately 1.5 mg. Fig. 5 Analyte transient signal area as a function of pulp sample mass (brownstock) (z, Mn; 6, Mg). 1718 J. Anal. At. Spectrom., 1999, 14, 1715±1722susceptible to factors such as intercalation and preferential volatilization due to spatial temperature disparities, both of which lead to multiple peaking.Table 2 compares the signal reproducibility between a pyrolytically coated probe and an uncoated probe for the analysis of the same Øuffed softwood TMP. Surprisingly, when both probes were relatively new (less than 25 sample insertions) there was a marginal difference between the % RSD values of the integrated signals for most of the elements studied. This is in stark contrast to what had been observed with pyrolytically coated DSI probes previously.15 In explaining this, the origin of the primary analyte volatilization event upon sample insertion must be considered.Wood pulp is a highly polymerized, Æbrous material, and when a pulp sample was dried and pyrolyzed, the sample was reduced to a carbonaceous residue with a high surface area. After pyrolysis, the metals initially present in the pulp sample were most likely still in contact with this carbon residue and were vaporized from the surface upon insertion into the plasma. The result was an initial vaporization event that was inØuenced greatly by the sample residue, which was identical in both cases.While this may explain the similarities in signal reproducibility, vaporization from the carbon residue alone does not account for the marked difference in signal appearance in Fig. 6. The difference in signal appearance probably arises from secondary recondensation and revaporization events on the probe surface subsequent to vaporization from the carbonaceous pulp residue.Although recondensation±revaporization after the initial vaporization event will inØuence the appearance of the peak shape, the initial vaporization event will be the predominant inØuence in signal reproducibility. While the observed signal reproducibility was virtually identical for both pyrolytically coated and uncoated probes that were relatively new, the precision achieved with the latter degraded rapidly after 25±50 insertions to the point where the signal RSD was on the order of 100±200%.Visual inspection of the probe revealed that the cup portion of the probe was being oxidized and was disintegrating with repeated use. It is obvious that damage to and disintegration of the sample probe will degrade signal reproducibility due to sample losses, intercalation and the like, and that the highly crystalline, ordered, non-porous surface of the pyrolytically coated probe will demonstrate ameliorated resistance to oxidate attack. Less obvious was the source of the oxidative attack.Several potential sources of oxidation exist. Firstly, it has been documented that Na, as well as other alkali and alkaline earth elements, have a catalytic effect on the rate of oxidation from graphite.14 As an example, 20±40 ppm of Na, K, V or Cu has been shown to increase the rate of dry oxidation of graphite by upwards of 6-fold. Since O2 and CO2 both support the oxidation of graphite, and both were expelled from the pulp sample during the pyrolysis step, a catalytic oxidative effect from the Na deposited as NaF was probably occurring.Additionally, the reaction of Øuorine with graphite could be taking place, resulting in the loss of graphite from the probe due to the formation of various Øuorocarbon species, such as CF4 and C2F6, which are known to occur at temperatures less than 700 �C.20 Limit of detection (LOD) Determining the limit of detection (LOD) for the method was not trivial due to the unavailability of a true Æeld blank, i.e., a blank consisting of a matrix representative of the sample being analyzed (a wood pulp sample in this case).21 Although a blank was used in the external standardization curve, it was an aqueous sample, and the use of its standard deviation for the purposes of estimating the LOD would be inappropriate.A better approach would be the duplicate analysis of samples with analyte concentrations 10±30 times the expected detection limit, and then using the observed standard deviation in the LOD estimation.Owing to the limitations of the samples available, this approach was also not possible. As a compromise, the off-peak transient signals used for background correction were taken from the pulp samples analyzed, corrected for their respective offset from zero along the ordinate axis, and integrated. The standard deviation from these integrated signals was then incorporated into the LOD deÆnition as recommended by IUPAC (eqn. 1): cL à 3sB=m Ö1Ü where cL is the concentration LOD, sB is the standard deviation in the blank signal, and m is the slope of the calibration curve. The calculated LODs using this practice appear in Table 3. For the values reported in Table 3, it is Fig. 6 InØuence of the pyrolytic coating on analyte volatilization for a Øuffed softwood thermomechanical pulp (---, uncoated probe; – pyrolytically coated probe). Table 2 Signal reproducibility comparison (Øuffed softwood TMP) RSD (n~10) (%) Element Uncoated probe (v25 insertions) Pyrolytically coated probe Cu 23 23 Fe 11.5 13 Mg 6.3 6.0 Mn 14 4.1 Zn 10.3 10.8 J.Anal. At. Spectrom., 1999, 14, 1715±1722 1719important to mention that the off-peak signals were not corrected for sample mass, and that the standard deviations were pooled from the two different samples analyzed (n~10 for each sample of brownstock and TMP). This was done because the external calibration curve was assumed to be valid for all pulp samples irrespective of sample type or mass.By comparison these values are, with the exception of Mg, approximately one order of magnitude higher than those obtained when the standard deviation of the integrated aqueous blank signal was used in eqn. 1 (Mg was approximately 102 higher). Care should be taken in extracting a practical quantitation limit (PQL) from these LOD values (normally 3±10 times the method detection limit), as they are not based on the standard deviation of low-concentration samples.The PQL will as a result most likely be somewhat higher than expected. Precision and accuracy Table 4 presents a comparison of the precision and accuracy for the analysis of the two pulp samples studied using DSI (with external standards) and the industry standard method (CPPA Method G.34P). In terms of precision, the wet oxidation method almost always faired better than DSI. With aqueous solutions DSI has been shown capable of achieving precision that rivals that of solution nebulization. Percentage RSD values, typically of less than 5%, and in extraordinary cases less than 1%, have been realized.22 Irreproducibility in sample deposition, insertion and volatilization is the primary inØuence in the observed precision. However, when analyzing a solid directly by DSI the precision is often much worse, as the previously mentioned factors are now more inØuential, and new factors such as inter-sample heterogeneity and sample size variability further compound the imprecision of the technique.Evidence of the inØuence of sample homogeneity on precision can be seen when the % RSD values obtained for the TMP sample in Table 4 are compared with those obtained for a similar TMP sample that had been mechanically Øuffed (Table 3). Although low-concentration elements (Cu and Zn) show no appreciable improvement in precision, for highconcentration elements (Mg and Mn) the % RSD is reduced by approximately an order of magnitude when the sample is Øuffed.It is also important to note that the physical constraints of DSI in terms of sample size dictated the use of an extremely small sample (on the order of 1±4 mg). Although this mass of pulp was more than sufÆcient to generate detectable signals for the analytes of interest, slight errors in accurately determining such a minute sample mass can greatly inØuence the observed precision. When the DSI precision was compared between the two pulp types studied, the reproducibility of the brownstock was generally better than the TMP sample, with RSD values ranging from 7±20% versus 16±44%, respectively (Table 4).A similar trend in reproducibility was also observed with the wet oxidation values, suggesting that an inherent difference that exists between the two pulps plays a role. The difference is that in the kraft pulping process (brownstock), the lignin is dissolved away chemically, leaving cellulose and hemicellulose in the form of intact Æbers, whereas thermomechanical pulping (TMP) yields a distribution of shortened Æbers resultant from a mechanical shredding and deÆbering process.17 It should be noted, however, that the kraft process, in practice, also chemically degrades a certain amount of the cellulose and hemicellulose Æbers, whereas the thermomechanical pulping leaves most of the cellulose and hemicellulose intact.Metals in wood such as Ca, K, and Mg are often partially bound to the carboxyl groups present in the cellulose and hemicellulose, and heavy metals such as Fe and Mn are often chelated by wood constituents.23 The chemical damage from the conditions experienced in the kraft process makes the analytes easier to liberate from the ashed pulp in both the wet oxidation and DSI analyses, thus yielding the improved precision in the brownstock values.When the accuracy (using the CPPA Method G.34P values as a reference) was compared for the pulps, a trend opposite to that observed with the precision values was evident.For Mg and Mn, the two elements present in the highest concentrations, agreement with the TMP values was 5.0 and 4.4%, respectively, as compared to 15 and 16% for the brownstock sample (Table 4). The discrepancy in accuracy can be attributed in part to sampling of the pulp for analysis. The TMP sample, upon air drying, took the form of rather large, Æbrous pieces [Fig. 2(b)] that had to be physically separated into smaller pieces with plastic tweezers in order to Æt into the DSI sample probe, and this action probably assisted in homogenizing the TMP sample.By comparison, the brownstock sample [Fig. 2(a)] was not as Æbrous, but occurred in a variety of sizes, the smaller of which were suitable for deposition into the DSI cup. Sampling was consequently favored towards these smaller pieces, as the larger pieces of the brownstock pulp proved difÆcult to manually separate into smaller ones.The same trend was true of Fe, although the agreement between the values was somewhat poorer (29% versus 44% for TMP and brownstock, respectively). In both pulp samples, the value for Fe concentration as determined by DSI was always lower than at determined by the wet oxidation method, due most likely to the incomplete volatilization of Fe from the sample probe as described earlier (Fig. 4). The concentrations reported for Cu and Zn were quite low in comparison to the previous three elements, with Cu being close to the detection limit of the wet oxidation technique.Consequently, the % RSD was relatively high by comparison and no deÆnite correlation between sample and % RSD could be obtained. Notwithstanding Fe, the error in the determination of the elements does not appear to be biased low or high of the reference value. This is reinforced by the fact that the average error for all of the determinations is only 4% (Table 4).Standard additions were performed on both the brownstock and TMP samples, but no apparent beneÆt was seen in adopting this calibration approach for the pulp analysis. In addition to being more laborious in terms of the number of samples that had to be run, the obtained precision and accuracy were most often poorer than obtained with external standards. This is attributable to the way that the reproducibility of the pulp signals inØuences the determined analyte concentration differently in external standards and standard additions. With standard additions, the uncertainty in the slope and intercept of the calibration curve was considerably higher than that in the external standards curve, a consequence of the fact that the standard additions curve was generated from pulp samples, as opposed to liquid standards in the external calibration case.Furthermore, the curve itself is used in standard additions to determine the analyte concentration (by determining the point of intercept with the abscissa), whereas analyte concentration is Table 3 Limit of detection (LOD) for pulp analysis LOD Element Absolute/ng Relative (ppb)a Cu 0.052 21 Fe 0.94 375 Mg 6.7 2700 Mn 0.56 225 Zn 0.20 80 aBased on a 2.5mg pulp sample (median of mass range used in this study). 1720 J. Anal. At. Spectrom., 1999, 14, 1715±1722determined by interpolation of signals from a more precise curve in external standards. Conclusions The respective merits of rapidly analyzing wood pulps for various metals by DSI-ICP-OES and using a pyrolytically coated graphite cup DSI probe for this type of routine analysis have been demonstrated. The pulp and paper industry has a persistent need for on-line and extremely rapid off-line methods of monitoring various process parameters so as to prevent unnecessary delays in production and the production of offgrade products.24 The use of DSI for rapidly determining the trace metals proÆle of wood pulps prior to TCF bleaching proves excellent in fulÆlling this mandate.Considering the short times needed to dry, ash, and insert the sample into the plasma, a raw pulp sample analysis by DSI can be completed in 5 min from pulp sample procurement. The expeditious nature of the DSI analysis compares extremely well relative to the several hours needed to digest pulp using industry standard wet oxidation methods.5,6 Although precision and accuracy do suffer somewhat when a solid sample is analyzed directly by DSI, quantitative results were obtained for process-inØuential metals in wood pulp samples (Mg and Mn), and semi-quantitative results were realized for low-concentration, but still process-inØuential metals (Cu, Fe, and Zn).With the exception of the values obtained for Fe, there appears to be little bias in the error of the determination (average error of 4%). It is common for industry pulp samples to have typical metal concentrations that vary (or will be varied by means of chemical treatment) in concentration over several orders of magnitude.With this reality considered, the demonstrated accuracy and precision of DSI-ICP-OES appears more than adequate for the purposes of a rapid screening technique. It is important, however, to consider that these results are indeed preliminary, as they are based on the replicate analysis of only two, albeit well characterized, samples (one of each wood pulp type).The replicate analysis of more samples of each pulp type covering the expected concentration range for the analytes of interest would give a more comprehensive picture statistically in terms of the expected precision and accuracy of the technique. Precision can be improved by homogenizing the pulp sample prior to analysis, e.g., by mechanical ØufÆng (compare % RSD for Øuffed TMP in Table 2 with TMP in Table 4), and by using a larger sample. Although DSI and ETV have physical constraints that limit the size of sample that can be deposited, higher capacity thermal sample introduction techniques, such as induction heating vaporization (IHV),25 are quite capable of handling larger samples.The direct analysis of wood pulps using a technique such as IHV should be explored in the future. The pyrolytically coated graphite DSI probe demonstrated a greatly enhanced resistance to oxidative and chemical attack, resulting in a longer useful lifetime than an uncoated graphite probe.Although an improvement in precision was not observed for the pulp samples, greater signal sensitivity and shorter signal evolution times were observed when a pyrolytically coated probe was used. Acknowledgements The authors wish to thank the Analytical Services Division of the Pulp and Paper Research Institute of Canada for the generous provision of wood pulp samples and analyses for this study. For scholarship support, MER would like to gratefully acknowledge the Ænancial support of the Province of Que�bec through Fonds pour la Formation des Chercheurs et l'Aide a¡ la Recherche (FCAR), and PH gratefully thanks the Natural Sciences and Engineering Research Council of Canada (NSERC).Authors MER and EDS gratefully acknowledge funding from NSERC through an NSERC Operating Grant. References 1 B. Van Lierop, N. Liebergott and M. Faubert, J. Pulp Pap. Sci., 1994, 20, J193. 2 J. D. Sinkey and N. S. Thompson, Pap. Puu, 1974, 5, 473. 3 J. Prasakis, M. Sain and C. Daneault, TAPPI J., 1996, 79, 161. 4 J. Bouchard, H. M. Nugent and R. M. Berry, Preprints CPPA International Pulp Bleaching Conference, 1994, 33. 5 Standard Method G.34P, Canadian Pulp and Paper Association. 6 Standard Method T 266 om-88, Technical Association of the Pulp and Paper Industry (TAPPI). 7 D. R. Hull and G. Horlick, Spectrochim. Acta, Part B, 1984, 38, 843. 8 I. Atsuya, T. Itoh and T. Kurotaki, Spectrochim. Acta, Part B, 1991, 46, 103. 9 A. Lorber and Z. Goldbart, Analyst, 1985, 110, 155. 10 E. D. Salin, C. V. Monasterios and A. M. Jones, Anal. Chem., 1986, 58, 780. 11 W. E. Pettit and G. Horlick, Spectrochim. Acta, Part B, 1986, 41, 699. 12 G. Zaray, J. A. C. Broekaert and F. Leis, Spectrochim. Acta, Part B, 1988, 43, 241. 13 C. D. Skinner and E. D. Salin, J. Anal. At. Spectrom., 1997, 12, 725. 14 W. Huettner and C. Busche, Fresenius Z. Anal. Chem., 1986, 323, 674. 15 M. E. Rybak and E. D. Salin, J. Anal. At. Spectrom., 1998, 13, 707. 16 G. Le�ge¡re and P. Burgener, ICP Inf. Newsl., 1987, 13, 521. 17 G. A. Smook, Handbook for Pulp and Paper Technologists, Angus Wilde, Vancouver, 2nd edn., pp. 36±45. 18 V. Karanassios, M. Abdullah and G. Horlick, Spectrochim. Acta, Part B, 1990, 45, 119. Table 4 Determination of Cu, Fe, Mg, Mn, and Zn in pulp samples Direct sample insertion (n~7) CPPA method G.34P (Paprican) (n~3) Pulp sample Element Concentration (ppm) RSD (%) Concentration (ppm) RSD (%) Error (%) Brownstock Cu 1.27 11 0.84 7.39 51 Fe 10.7 20 19.4 12.5 245 Mg 593 13 515 1.9 15 Mn 93.2 6.7 80.3 0.99 16 Zn 11.8 20 11.6 1.9 1.7 TMP Cu 1.64 18 1.10 40 49 Fe 40 44 56.7 1.6 229 Mg 201 39 212 15 25.2 Mn 91 41 88 15 3.4 Zn 11.8 16 14.0 7.7 216 Average error 4.1 J. Anal. At. Spectrom., 1999, 14, 1715±1722 172119 X. R. Liu and G. Horlick, J. Anal. At. Spectrom., 1994, 9, 833. 20 W. Ru» dorff and G. Ru» dorff, Z. Anorg. Allg. Chem., 1947, 253, 281. 21 Analytical Methods Committee, Analyst, 1987, 112, 199. 22 R. L. A. Sing and E. D. Salin, Anal. Chem., 1989, 61, 163. 23 E. Sjo» stro»m, Wood Chemistry: Fundamentals and Applions, Academic Press, San Diego, CA, USA, 2nd edn., p. 107. 24 B. B. Sithole�, Anal. Chem., 1995, 67, 87R. 25 D. M. Goltz, C. D. Skinner and E. D. Salin, Spectrochim. Acta, Part B, 1998, 53, 1139. Paper 9/01694E 1722 J. Anal. At. Spectrom., 1999, 14, 1715±17