Special feature: perspective Received: 29 July 2013
Revised: 14 October 2013
Accepted: 15 October 2013
Published online in Wiley Online Library
(wileyonlinelibrary.com) DOI 10.1002/jms.3299
On-line process monitoring of coffee roasting by resonant laser ionisation time-of-flight mass spectrometry: bridging the gap from industrial batch roasting to flavour formation inside an individual coffee bean R. Hertz-Schünemann,a R. Dorfner,b C. Yeretzian,c T. Streibela and R. Zimmermanna,b* Resonance-enhanced multiphoton ionisation time-of-flight mass spectrometry (REMPI-TOFMS) enables the fast and sensitive on-line monitoring of volatile organic compounds (VOC) formed during coffee roasting. On the one hand, REMPI-TOFMS was applied to monitor roasting gases of an industrial roaster (1500 kg/h capacity), with the aim of determining the roast degree in real-time from the transient chemical signature of VOCs. On the other hand, a previously developed μ-probe sampling device was used to analyse roasting gases from individual coffee beans. The aim was to explore fundamental processes at the individual bean level and link these to phenomena at the batch level. The pioneering single-bean experiments were conducted in two configurations: (1) VOCs formed inside a bean were sampled in situ, i.e. via a drilled μ-hole, from the interior, using a μ-probe (inside). (2) VOCs were sampled on-line in close vicinity of a single coffee bean’s surface (outside). The focus was on VOCs originating from hydrolysis and pyrolytic degradation of chlorogenic acids, like feruloyl quinic acid and caffeoyl quinic acid. The single bean experiments revealed interesting phenomena. First, differences in time–intensity profiles between inside versus outside (time shift of maximum) were observed and tentatively linked to the permeability of the bean’s cell walls material. Second, sharp bursts of some VOCs were observed, while others did exhibit smooth release curves. It is believed that these reflect a direct observation of bean popping during roasting. Finally, discrimination between Coffea arabica and Coffea canephora was demonstrated based on high-mass volatile markers, exclusively present in spectra of Coffea arabica. Copyright © 2013 John Wiley & Sons, Ltd. Additional supporting information may be found in the online version of this article at the publisher’s web site. Keywords: coffee; coffee roasting; aroma; resonance enhanced multiphoto-ionisation/REMPI; process monitoring
Introduction
J. Mass Spectrom. 2013, 48, 1253–1265
* Correspondence to: Ralf Zimmermann, University of Rostock, Joint Mass Spectrometry Centre, Chair of Analytical Chemistry, Institute of Chemistry, D-18059 Rostock, Germany. E-mail: [email protected] a Joint Mass Spectrometry Centre, Chair of Analytical Chemistry, Institute of Chemistry, University of Rostock, D-18059 Rostock, Germany b Joint Mass Spectrometry Centre, Comprehensive Molecular Analytics (CMA), Helmholtz Zentrum München – German Research Centre for Environmental Health, D-85764 Neuherberg, Germany c Zurich University of Applied Sciences, Institute of Chemistry and Biological Chemistry, Einsiedlerstrasse 31, CH-8820 Wädenswil, Switzerland
Copyright © 2013 John Wiley & Sons, Ltd.
1253
Green coffee contains only small concentrations of volatile- and semi-volatile organic compounds (VOC/SVOC) and a weak hay-like flavour that does not have any resemblance to the flavour of roasted coffee. When coffee is roasted, complex chemical and physical processes occur, which result in the formation of browning pigments and flavour compounds. Roasted coffee is characterised by a relatively high content of VOC and SVOC which makes it one of the foodstuffs with the richest flavour content (up 0.1% VOC/SVOC per dry weight of roasted coffee). More than 850 VOC/SVOC compounds have been identified[1] in roasted coffee (mostly in trace levels) from which about 500 are formed by the predominant Strecker and Maillard reactions.[1] However, despite extensive research on coffee flavour formation,[1–5] the knowledge about these processes remains fragmentary, and the personal experience of the coffee roast master is indispensable when it comes to control of the flavour formation process during roasting. Research on the formation of VOC in coffee roast gases has traditionally relied on chromatographic techniques, most often gas chromatography–mass spectrometry (GC-MS)[6–8] but also
high performance liquid chromatography–mass spectrometry (HPLC-MS).[9–12] A continuous, time-resolved coverage of the roasting process, however, is difficult with GC-MS or HPLC-MS, and on-line real-time analytical methods are better suited to follow the typically fast dynamics of flavour formation during the roasting process. Thus, several studies used on-line spectroscopic methods such as infrared (IR),[13,14] RAMAN spectroscopy[15,16] or sensor arrays (electronics noses).[17,18] Mass spectrometry using
R. Hertz-Schünemann et al.
1254
hard electron ionisation can be utilised in combination with the concept of the multivariate analysis, taking the individual mass channels as input parameters. This ‘MS sensor’[19,20] as a ‘fingerprint’ approach could also be applied in quality control to objectify/complement sensory testing.[21] However, fragmentation upon ionisation and the relative complexity of the chemical signature limits the usefulness of the latter approaches. Another strategy is to use on-line mass spectrometric techniques in combination with soft ionisation, which creates less ionisation induced fragmentation and simultaneously higher selectivity. In particular, on-line photo ionisation mass spectrometric techniques (PI-MS[22–25]) and on-line chemical ionisation mass-spectrometric approaches (e.g. proton-transfer-reaction, PTR-MS)[26–29] have been applied so far. Additionally, selected ion flow tube mass spectrometry (SIFT-MS) was applied for on-line analysis of trace gas molecules in air/breath samples, in combination with chemical ionisation.[30] Photo ionisation mass spectrometry[23] and later PTR-TOFMS[31] were both used to predict the roast degree of coffee by on-line analysis of the roast gas. These methods can detect VOC/SVOC in complex gaseous matrices with or without reduced interferences from fragments and inorganic bulk gases such as CO2, CO, N2 and H2O and achieve high time resolution and sensitivity and – in some cases – high chemical selectivities. The most selective of the latter MS-based on-line methods is the laser-based resonanceenhanced multiphoton ionisation process (REMPI). It is characterised by a very high selectivity and unsurpassed sensitivity for aromatic compounds. Yet, depending on the wavelength, only few non-aromatic compounds are ionised. REMPI requires the absorption of at least two laser photons to reach the ionisation energy of a molecule. For efficient REMPI, the laser UV-wavelength needs to be in resonance with the UV-absorption bands of the target molecules. By using UV-laser wavelengths in the 230–270 nm spectral range, a large number of aromatic compounds are very efficiently ionised, while aliphatic molecular species are not detected.[32–34] In 1996, REMPI-MS was applied for the first time to the analysis of coffee volatiles.[25] The REMPI method was complemented by the single photon ionisation methods with VUV-light, which has a very different selectivity than REMPI. Depending on the photon energy, it ionises the majority of the organic compounds.[32] Early SPI (single photon ionisation) and REMPI-MS studies in combination with multivariate analysis focused on the formation mechanisms of volatile phenolic roast products[22] and linked different roast degrees to the time formation behaviour of selected key compounds.[23] Due to its high selectivity and sensitivity, REMPI-MS turned out to be well suited for the on-line determination of the degree of roasting. The current work addresses two complementary applications of REMPI mass spectrometry in coffee sciences. On the one hand, and on a more applied side, REMPI-TOFMS was used to monitor the roasting process of a large industrial roaster. For this, a compact REMPI-TOFMS system was connected to an industrial scale batch roaster with an hourly roasting capacity of 1.5 metric tons of green coffee. The time–intensity profiles of a series of VOC/SVOC, which are indicative of the evolution of the roast degree, were successfully recorded. On the other hand, addressing fundamental scientific questions, the chemistry and kinetics of VOC/SVOC during roasting of individual coffee beans were examined. In this context, it is notable that the integrity of the individual coffee bean plays a decisive role for the formation of coffee flavour compounds[1,35,36]; roasting of ground green coffee does not result
wileyonlinelibrary.com/journal/jms
in the formation of the typical coffee aroma but rather in a ‘hay-like’ flavour.[37] Based on a previous study using SPI-MS,[38] a so-called μ-probe sampling device was used for sampling gases (1) from the inside of a coffee bean and (2) from just outside the surface (in the close vicinity of a coffee bean’s surface). The sampling technology was initially developed for the analysis of combustion gases within the rod of a cigarette.[39,40] Briefly, the stainless steel sampling capillary, of the μ-probe (external diameter 0.4 mm/internal diameter 0.2 mm), was coupled via a heated transfer capillary to the REMPI-mass spectrometer. The complete system is heated, so that clogging of the capillary is avoided, permitting the transfer of compounds with higher m/z values without memory effects. In this way, the high mass and time resolution of REMPI-TOFMS was combined with a unique sampling method from the interior of a single coffee bean. This work focuses on the monitoring of selected phenolic compounds, whose formation during roasting can be attributed to two different proposed chlorogenic acid degradation pathways. The application of the μ-probe technology to coffee beans allowed the observation of novel and unexpected phenomena, pertaining to dynamic processes as well as to differences in the chemical signature between Coffea arabica and Coffea canephora. First, comparing time–intensity profiles for selected VOC from inside and outside the bean, a time shift was observed for the intensity maximum between both profiles. This was tentatively linked to the diffusion of VOCs from inside towards the outside and is believed to be related to the permeability of the coffee bean cell wall material. Second, we observed for selected VOCs intense and sharp spikes in the time–intensity profiles, while other VOCs showed smooth profiles. We interpret this as a direct manifestation of popping/cracking caused by high internal pressures within the bean cell structure, and the simultaneous abrupt release of VOCs. Finally, comparing static mass spectral profiles recorded from outside single beans, we discovered that it was possible to distinguish between the two most common coffee species Coffea arabica and Coffea canephora, based on characteristic yet currently un-identified high molecular mass VOCs that appear solely in the spectra of Coffea arabica. On a more long-term perspective, this study aims at linking observations on the individual bean level with bulk-roasting phenomena and hence bridge the gap from the single bean to the bulk (large batch).[41]
Experimental and methods On-line REMPI-TOFMS on an industrial 1.5 t/h roaster For the analysis of VOC in the roast gases of an industrial coffee roaster, a compact and rugged photo ionisation-TOFMS instrument, designed for measurements under industrial application, was used. Briefly the instrument was equipped with a compact 10 Hz fixed-frequency Nd:YAG laser (Minilight-II, Continuum Inc., Santa Clara, CA, USA) with harmonic generation of 355 nm and 266 nm laser pulses. REMPI is performed with 266 nm pulses, while the 355 nm pulses are used in conjunction with a homebuilt third harmonic generation (THG) gas cell for generation of 118 nm laser pulses for SPI. The compact linear reflectron TOFMS (custom build by Kaesdorf Instrumente für Forschung und Industrie, Munich, Germany and later modified) was mounted together with turbo-molecular vacuum pumps and electronics in a 19 inch rack. The TOFMS allows an on-line detection limit
Copyright © 2013 John Wiley & Sons, Ltd.
J. Mass Spectrom. 2013, 48, 1253–1265
Single coffee bean measurements with direct MS in the low ppb range for many aromatic compounds using REMPI in the here applied configuration. The sample introduction system consists of a heated, hollow stainless steel needle (1 mm), directing a deactivated fused-silica capillary of 200 μm inner Ø to the centre of the TOFMS ion source. The fused-silica capillary, which is connected via a heated transfer line hose to the sampling system, acts as a restriction step between the ion source (pressure: 10 5 Torr) and the sampling system that is operated at ambient pressure. The sample gas flow through the capillary is about 2 ml/min. At the open end of the capillary, an effusive molecular beam is formed. The REMPI ionisation zone is located directly underneath the capillary opening in the centre of the TOFMS ion source. The whole system is mounted in a dust shielded 19 inch rack. Details and specifications of the compact system setup can be found in the literature.[42] For the on-line analysis on the industrial large-scale batch roaster (R1500/1R Probat-Werke, Emmerich, Germany), a heated sampling probe (220 °C) was introduced into the roasting gas duct through a feed through from the front side of the roasting drum (Fig. 1). The tip of the sampling probe was connected air tight via a flange, due to the slightly reduced pressure (below external atmospheric pressure) in the roasting gas duct. Roasting gas (1.5 l/min) was extracted by the heated sampling probe. The heated transfer line with the sampling capillary of the TOFMS device was coupled to the sampling train and 2 ml/min extracted at right angle from the heated sampling probe and transferred to the laser mass spectrometer. Details of sampling train setup, which was originally developed and optimised (as in previous measurements) in trials on a small-scale sample roaster, are reported in the literature.[23] The industrial roaster had a capacity 1500 kg/h with a maximal batch size of 205 kg for roasting times of 10 – 13 min. The heating was performed by natural gas (~110 MJ fuel/100 kg green coffee beans). Note, that the roasting gases are partially recycled in the process. About 180 kg of a blend of different Coffea arabicas was roasted. Once the intended roasting degree is achieved (at about 610 s roasting time and 207 °C), the exothermic roasting process was quenched by spaying 20 l water inside drum. The VOC/SVOC during the complete roasting cycles were measured using SPI and REMPI ionisation, respectively. Here, only the REMPI data is discussed.
REMPI-TOFMS of single beans using μ-probe sampling For the analysis of VOCs from single beans, a high-end REMPITOFMS instrument coupled to a μ-probe sampling device was used (Fig. 2A). Only a brief explanation of the homebuilt REMPITOFMS analyser is given here, as the system has been described in detail elsewhere.[34,43] Nd-YAG laser pulses (Continuum, Santa Clara, CA, USA, 1064 nm) were used to generate radiation at 355 nm, by frequency tripling the laser pulse (repetition rate 10 Hz, pulse width 5 ns). A beam splitter guided about 90% of the 355 nm output to an optical parametric oscillator (OPO, VISIR 2 + SHG, GWU-Lasertechnik GmbH, Erftstadt, Germany) with a thermally stabilized ß-barium borate (ß-BBO) crystal to create visible laser pulses. The remaining 10% are used to pump a third harmonic generation (THG) gas cell for generation of 118 nm pulses for single photon ionisation (SPI, not used in this work). The OPO output is frequency doubled to generate UV pulses for REMPI. Here, a wavelength of 248 nm was chosen, because previous studies showed that this wavelength is demonstrably well suited for detection of aromatic compounds.[21] By using appropriate optical elements, the beam was focused underneath the inlet needle in the ion source of the mass spectrometer. The inlet needle is connected to the sampling line and generates an effusive molecular beam in the ion source. REMPI parameters were adjusted to mostly fragmentation-free conditions (laser flux density ~ 1–5 107 Wcm 2). Ions were extracted into the flight tube of a reflectron TOFMS (Kaesdorf Instrumente für Forschung und Industrie, Munich, Germany). Two PC cards (Acquiris, Agilent Technologies, Basel, Switzerland, 250 MHz, 1 GS/s, 128 kb) recorded the resulting mass spectra, with a linear intensity range covering five orders of magnitude.[44] The mass range was set from m/z 10 to 380. Ten REMPI mass spectra were recorded per second (10 Hz) and the data recording and processing were performed by a purpose-written LabView (National Instruments, Austin, Texas, USA)-based software programme. Three to four replicates were measured for each experiment. All measurements were base line corrected and normalised to the total ion current. Only for Figs 5 and 7 original, not averaged measurements were used. A principle component analysis (PCA, Unscrambler, CAMO Software AS, Oslo, Norway)
J. Mass Spectrom. 2013, 48, 1253–1265
Copyright © 2013 John Wiley & Sons, Ltd.
wileyonlinelibrary.com/journal/jms
1255
Figure 1. Photograph (left) and scheme (right) of the industrial large-scale batch coffee roaster (R1500/1R, Probat-Werke, Emmerich, Germany) with a hourly capacity of 1.5 metric tons of green coffee and the sampling procedure. The sampling point and roast gas circulation paths are indicated.
R. Hertz-Schünemann et al.
Figure 2. A) Scheme of the _-probe and the REMPI-TOFMS. B) Overview of the roast gas sampling setup for _-probe measurements inside or in the vicinity (outside) of a single coffee bean. C) Photograph of the sampling setup for a _-probe measurement of roast gases from inside a single coffee bean.
was performed on normalised data, whereby only those 30 variables/ compounds were selected and included into the analysis that had the highest impact on the predictive model. For the single bean experiments, a μ-probe sampling setup was used. The μ-probe setup has been described elsewhere,[45] and a brief description will be given here: The μ-probe (Fig. 2B/C) consists of a conically shaped heated aluminium base body which is coupled via a heated adapter to the heated transfer line. The heart of the μ-probe is a small stainless steel capillary (ID 0.2 mm/OD 0.4 mm), which is connected to the transfer capillary using a capillary union. The μ-probe capillary sticks out of the conically shaped base body by approximately 4 mm and is thereby indirectly heated up to 120 °C (heat conductance). Two different types of μ-probe sampling experiment were performed:
section of a drilled roasted bean with the μ-probe sampling channel (a) and the ‘cut’ (b) is given in Fig. 3A. Here, the general structure of a bean is visualised with the bean endosperm (c), residues of the silverskin (d) and the amorphous zone (e). In Fig. 3B, an enhanced image of the μ-probe sampling channel is depicted. The endosperm consists of cells, which are filled with nutritious organic material (e.g. organic acids, carbohydrates, amino acids and lipids). These substances are the precursor material for the generation of coffee roasting aroma, which are decomposed during the pyrolysis process.[38,46,47] After roasting, the cells are more or less empty.[38]
i) For measurements conducted to sample roasting gases from within individual coffee beans (Fig. 2B/C), Ø 1 mm holes were drilled approximately 5 mm deep into the coffee beans. Then, the capillary of the μ-probe was inserted inside the bean and the hole was sealed by inorganic glue based on zirconium oxide (Polytec PT GmbH, Waldbronn, Germany). The hole was closed throughout the complete measurement. ii) For measurements conducted to sample roasting gases at the surface of individual coffee beans, coffee beans were placed individually in a 2 ml glass flask, which then was connected to the μ-probe (Fig. 2B).
Monitoring the coffee roasting process gas from an industrial coffee roaster
A hot air gun was taken for simulating the roast process, using a mantle thermocouple for controlling the roast temperature (T = 250 °C) on the surface of the bean (case i) or inside the glass (case ii) during heating, respectively. For the experiments, green Coffea arabica beans from Bolivia and green Coffea canephora beans from India (both in organic quality) were used.
Coffee bean morphology
1256
Raster electron microscopy (REM) images for visualisation of the location and the dimension of μ-probe sampling channel inside the coffee bean were recorded at the Electron Microscopy Centre of the University of Rostock. (EMZUniRo) In Fig. 3, two selected REM images are presented. The dimension scale is given in the lower right corner of the respective figures. The cross
wileyonlinelibrary.com/journal/jms
Results and discussion
Due to the high selectivity and softness of the ionisation step, photo ionisation mass spectrometry is well suited for the monitoring of process gases.[33,48,49] New developments in mass spectrometry allow the setup of robust process monitoring devices, based on REMPI-TOFMS. While the usage of such sophisticated technologies may currently seem expensive for process control and monitoring in food processing, applications for high value products and processes, such as coffee, are a promising option (the market value of high quality raw Arabica is less than 1/10 of the value of the processed commodity). Once optimised applications are established, customised and cost-effective instruments are expected to bring costs down. The compact linear reflection REMPI-TOFMS uses the output of a small 10 Hz Nd:YAG laser (Minilight-II, Continuum Inc., Santa Clara, CA, USA)[42] for REMPI ionisation. Figure 4 (left) shows a typical REMPI mass spectrum obtained from the roasting off-gas at 623 s roasting time (begin of the water quenching process). The spectrum is averaged over 50 laser shot (5 s). The REMPI mass spectra obtained at the industrial roaster agree well with results reported in previous tests on small-scale sample roasters.[22,23] Note, that the ionisation laser wavelength of 266 nm enhances the phenolic compounds in the REMPI mass spectrum (m/z 94, m/z 108, m/z 110, m/z 124, m/z 150, m/z 164), with 4-vinylguaiacol (m/z 150) as base peak, due to a high ionisation cross section. In
Copyright © 2013 John Wiley & Sons, Ltd.
J. Mass Spectrom. 2013, 48, 1253–1265
Single coffee bean measurements with direct MS
Figure 3. A) Raster Electron Microscope (REM) photograph of a cross section through a roasted coffee bean. Legend: a - drilled channel for μ-probe sampling, b - ‘cut’, c - endosperm (bean), d - residues of the silverskin, e - amorphous zone[46] B) Raster Electron Microscope (REM) photograph of the drilled channel for -probe sampling in the coffee bean.
Figure 4. REMPI-TOFMS measurement on an R1500/1R industrial large-scale batch roaster (Probat-Werke, Emmerich, Germany). Left) REMPI-TOFMS mass spectrum (@ 266 nm) at the end of the roasting process, at 620 s (begin of water quench); m/z values are indicated. Mainly phenolic compounds can be assigned (see text). Right) Time–intensity profile of the mass traces 150 m/z (4-vinylguaiacol) and 124 m/z (guaiacol), normalised to the maximum intensity, respectively. The onset of formation of 4-vinylguaiacol is slightly earlier than for guaiacol.
J. Mass Spectrom. 2013, 48, 1253–1265
can achieve a robust prediction of the roast-degree also under such complex industrial conditions. Single bean roasting Besides monitoring the roasting process at the industrial scale, the outstanding sensitivity of REMPI-TOFMS makes it uniquely suited to address fundamental questions related to the thermal treatment of single coffee beans. Here, we will discuss first results on REMPI-TOFMS analysis of VOC/SVOC generation and release from individual coffee beans during roasting (Arabica as well as Robusta beans). The motivation to this research is twofold: first, the single bean is the smallest processing unit (the “atom”) of coffee roasting. Linking processes at the level of the elementary unit of roasting to phenomena at the batch scale, we expect to gradually better comprehend the basic relations governing industrial roasting. Second, it is know that the chemical reactions and in particular the formation of the coffee aroma compounds during roasting strongly depend on the micro-conditions and micro-structure of the intact coffee bean.[50,51] Shifting the focus from large-scale batch roasting (180 kg green coffee/roasting batch) to the single bean level (~150 mg/per roasting), we observed a series of surprising phenomena that either pertain to the time–intensity release profile from the single beans, or to differences in the mass spectra between Robusta and Arabica beans.
Copyright © 2013 John Wiley & Sons, Ltd.
wileyonlinelibrary.com/journal/jms
1257
addition to phenols, indole (m/z 117), caffeine (m/z 194) and a C3-amine (m/z 59) are detectable as well. The weak signals at m/z 18 and m/z 28 originate from residual electron ionisation (EI) which is due to photo-induced electrons, accelerated in the electrostatic extraction field and which lead to EI ionisation of water and nitrogen. In the right part of Fig. 4, the time–intensity profiles of phenolics guaiacol (m/z 124) and 4-vinylguaiacol (m/z 150) during the roasting process are shown. The shaded region indicates the quenching of the roasting process by injection/spaying of water. The depicted compounds have been identified as roasting degree markers in previous work, with 4-vinylguaiacol being a product of the decarboxylation of ferulic acid – formed and released rather early in the process. This is evident in the roasting gas measurement at the industrial roaster as well. However, the differences in the release profiles among different compounds are less pronounced in the off-gas of the industrial roaster, compared to previous laboratory roaster experiments on 300 g sample roaster.[23] This is due to recycling a large fraction of the hot roast gas within in the industrial roaster: only a fraction of the roast gases are replaced at each cycle, e.g. added at the gas burner and discharged via the exhaust duct. While such recirculation of the roast gas may slightly complicate the on-line monitoring of the roast degree based on time–intensity profiles of VOC/SVOC in the off-gas of the roaster, a multicomponent model using multivariate statistical methods is
R. Hertz-Schünemann et al. Arabica vs. Robusta: While investigating the roasting process of individual Robusta and Arabica coffee beans, by sampling VOC/SVOC either (1) inside or (2) outside of the single bean, we observed consistent differences in the mass spectra between the two species. Up to date, a fast differentiation between the most common coffee species, Coffea arabica and Coffea canephora (Robusta) remains a challenging analytical task. In Fig. 5, REMPI-TOFMS data sampled via the μ-probe from inside bean are shown for Arabica and Robusta. The three-dimensional plots (Figs 5 A and 5D) depict an overview on exemplary single bean roasting results for an Arabica (5A) and a Robusta (5D) bean, respectively. A larger number of such 3D overview plots, allowing assessing the
variability, are given in the supplemental material (Figs S1 and S2). It can also be seen from the supplemental material that the variability is higher for the measurements inside relative to outside the beans (Table 1). This probably reflects the fact that the process of drilling the sampling channel as well as the position the channel needs to be further examined and standardised. As there are different tissues and morphological structures in the coffee beans,[38] a systematic investigation of the influence of the sampling point and channel depth needs to be performed in future studies. However, in general, the profiles are dominated by the masses m/z 150 and m/z 194, which can be assigned to be predominately 4-vinylguaiacol (m/z 150) and caffeine. A striking and consistent
Figure 5. REMPI-TOFMS measurements of coffee roasting gases extracted from the interior of individual coffee beans (inside) A) 3D representation of VOC/ SVOC time intensity profiles sampled by the +-probe from inside a single Arabica coffee bean B) REMPI-TOF mass spectrum measured inside of single Arabica beans (average of 4 measurements), plotted on a linear intensity axis. C) REMPI-TOF mass spectrum measured inside of single Arabica beans (average of four measurements), plotted on a logarithmic intensity axis. D) 3D representation of VOC/SVOC time intensity profiles sampled by the +-probe from inside a single Robusta coffee bean. E) REMPI-TOF mass spectrum measured inside of single Robusta beans (average of four measurements), plotted on a linear intensity axis. F) REMPI-TOF mass spectrum measured inside of single Robusta beans (average of four measurements), plotted on a logarithmic intensity axis.
Table 1. Maximal standard deviation values (sd) obtained for the 20 s sequences of the time profiles (four replicate measurements, compare with Fig. 8) of Robusta measurements inside (RI) and outside (RO) the bean (the minimal sd is 0.0 in all cases) m/z
1258
sd RI sd RO
59
94
110
117
124
136
150
194
max ±
0.73
0.62
0.54
0.84
0.70
0.59
0.53
0.90
max ±
0.28
0.22
0.17
0.19
0.28
1.00
0.34
0.21
wileyonlinelibrary.com/journal/jms
Copyright © 2013 John Wiley & Sons, Ltd.
J. Mass Spectrom. 2013, 48, 1253–1265
Single coffee bean measurements with direct MS difference between measurements on Robusta and Arabica is the occurrence of masses above m/z 200 for Arabica bean (see Figs S1 and S3 in the supplemental material). In the case of Robusta, these higher mass traces are nearly absent (Note that there are also variations in the observed intensity of the higher mass signals (m/z > 200), but the signals are always significantly higher in Arabica). This is also visible in the averaged mass spectra of three single bean roasting measurements for Arabica and Robusta, respectively (Figs 5 B/C and E/F). In particular, the REMPI mass spectra with logarithmic intensity axis (Figs 5C and 5F) show these differences very clearly. Although many molecular ion masses, which appear below m/z 200 in photo ionisation mass spectrometry (REMPI and SPI), can be identified/ assigned,[22–25,38] very little is known about the compounds that lead to ion signals in REMPI-TOFMS at masses above m/z 200. In Table 2, the REMPI active compounds, which were previously assigned, are listed together with the observed concentration range from literature data. The previously assigned compounds in the case of REMPI at 266 nm or 248 nm are phenolic compounds from the decomposition of chlorogenic acids. The isobaric alkylated pyrazines make a smaller yet not negligible contribution to the mass spectral ion intensities, since
Table 2. Assignment of observed masses in the REMPI mass spectra based on published literature and previous studies.[22–25,38] The concentrations are given for roast and ground coffee in [ppm] [1,35] and for extract or brew in [μg/l].[54,55] m/z 59 94 96 99 110 117 124 126 132 136 138 146 150 152 162
178
192 194
Substances C3-alkylated amine Phenol, Methylpyrazine Furfural Methylthiazole 1,2-Benzenediole, Methylfurfural Indole Guaiacol, Methylbenzenediole Hydroxymethylfurfural, Benzenetriole, Methylbenzofuran Vinyl-1,2-benzenediol Ethylcatechol, Dihydroxybenzaldehyde Phenylbutenal Furanylpyrazine Vinylguaiacol Vanillin, Ethylguaiacol Dihydroxy cinnamaldehyde, (Furanylmethyl)5-methylfuran Anthracene, Phenanthrene, Difurfurylether Methylphenanthrene, Methylanthracene Caffeine
J. Mass Spectrom. 2013, 48, 1253–1265
0.5–2.0[ppm][1] 1.55–14[ppm][1] 6.6–80[ppm][1] 22.5–85[ppm][1] 0.4–6[ppm][1] 20–340[ppm][1] 96.6–160[ppm][1] 0.3–2[ppm][1] 1.6–95.5[ppm][1] 31–75[ppm][35] 10–35[ppm][35] 0.15–130[ppm][1] 0.01–0.044[ppm][1] 4.4–40[ppm][1] 9–106[ppm][1] 0.1–20[ppm][1] 0.1–1.5[ppm][1] 0.6–0.7[ppm][1] 6–177.7[ppm][1] 2[35]–100[ppm][1] 0.3–18.1[ppm][1] 5–12[ppm][1] identified[1] 0.009–0.9 [μg/l][54,55] 0.124–0.9 [μg/l][54,55] 0.6–1.4[ppm][1] 0.002–0.317 [μg/l][55] 0.002–0.221 [μg/l][55] 1.1–2.8 [% DM][1]
Copyright © 2013 John Wiley & Sons, Ltd.
wileyonlinelibrary.com/journal/jms
1259
DM = Dry Matter
c
heterocyclic nitrogen compounds such as pyridine and pyrazine are not efficiently ionised in a REMPI process, at the wavelengths applied here (266 nm and 248 nm). Larger nitrogen heterocyclic compounds such as caffeine (m/z 194) and indol (m/z 117) are clearly seen in the REMPI mass spectra as well. Currently, no assignment can be made for this m/z ion signal at higher masses that appear to be characteristic for Arabica. Furthermore, these features have, until now, only been observed in single bean experiments, using the high end instruments and the heated μ-probe sampling setup. To be REMPI active, these compounds need to be aromatic or at least highly conjugated. Therefore, only specific substance classes are potential candidates such as flavonoid or polyphenolic structures. Polyphenols are yet known to be more abundant in Robusta (i.e. due to the higher chlorogenic acid contend). Note that REMPI can be extremely sensitive for aromatic compounds, so the absolute concentration of the compounds might be very low. Nevertheless, the clear differences in the higher mass region may render a REMPI-TOFMS analysis a valuable mean for a fast discrimination of Arabica from Robusta. Differences in m/z values can also be originated from different locations and/or growth conditions and/or pre-processing technologies. More investigations on different varieties of the two species as well as on the influence of the farming region, agronomic practices and different postharvest processing techniques (e.g. wet and dry processing) are needed at this point. For further analysis of the difference between the REMPI mass spectra from Arabica and Robusta, a so-called difference spectrum was calculated: the averaged (four measurements each) and normalised mass spectrum of coffee Robusta were subtracted from the one generated from Arabica measurements, respectively. The difference mass spectrum is given in Fig. 6. Compounds (i.e. m/z values) being more abundant in Arabica beans show up as positive peaks while compounds with higher concentration in Robusta exhibit negative peaks. Note that the absolute concentration cannot be determined from this plot. The molecular assignment of the REMPI detectable m/z values is listed in Table 2: Based on literature data of coffee aroma constituents and their observed concentrations as well as on the physical and chemical properties of the respective molecules (ionisation energies, photo ionisation cross sections), the m/z values observed by photo ionisation mass spectrometry were assigned to chemical compounds.[1,35,52–55] The assigned compounds are well supported by the literature (predominantly by GC-MS). Especially for the higher m/z values, however, it is increasingly difficult to achieve an unambiguous assignment. The difference spectrum clearly depicts the distinguishing chemical signature between Robusta and Arabica. From literature, it is known that Robusta coffee contains more sulphur compounds,[56,57] more caffeine[58] and more chlorogenic acids. [57–62] The higher content of chlorogenic acids results in an increased formation of phenolic compounds such as vinylguaiacol[1,62] (m/z 150), 1.2-benzenediole (m/z 110), phenol (m/z 94)[1] and as well as vinyl-1,2-benzenediol (m/z 136). Also, higher amounts of indole (m/z 117) were observed in Robusta beans. Indole is formed by the degradation of phenylalanine predominately at higher roasting temperatures (230–260 °C).[1] The higher caffeine content of the Robusta varieties is reflected as well in the difference spectrum. In contrast, the difference spectrum (Fig. 6) shows higher intensities of many compounds for Arabica beans. At lower masses, the C3-alkylamine at m/z 59, which is formed by pyrolysis of sulphur containing amino acids[1] (mainly methionine and cystine[63]), is an outstanding example. Casal et al. found more free methionine (in trace) in Arabica than
R. Hertz-Schünemann et al.
Figure 6. Difference of averaged and normalised REMPI mass spectrum of Arabica minus Robusta. Spectra correspond each to the averaged VOC/ SVOC from four different beans, sampled from outside the coffee bean.
in Robusta coffee beans.[64] The presence of furfural (m/z 96)[60] is notable and results from the higher carbohydrate contend of Arabica coffees. Furfural is formed by decomposition of pentosane or glucose and also by oxidation of furfuryl alcohol,[1] whereas the furfurylalcohol is a product of the reaction of (deoxy)ribose or sucrose with cycstein/methionine.[65,66] Beyond that, Sanz et al. reported similar contents for methylthiazole (m/z 99) in Arabica and Robusta coffee,[56] which is generated via condensation of carbohydrate fragments and cysteamine (degradation product of cysteine).[67] Also, relatively higher signal intensities for m/z 162 and m/z 132 are likely due to furan-based carbohydrate decomposition products (e.g. 2-(2-furanylmethyl)-5-methyl-furan[1] - m/z 162, methylbenzofuran - m/z 132). Hence, much of the observed differences at lower masses between Arabica and Robusta can tentatively be rationalised, based on published literature on the composition of the Arabica and Robusta varieties. Interesting is the homologous series of m/z 178, m/z 192, m/z 206, m/z 220, m/z 234 which belongs to the alkylated phenanthrenes (i.e. the substance class of the alkylated polycyclic aromatic hydrocarbons). Alkylated phenanthrenes are well known to be formed in relatively high abundances in pyrolysis processes of resin-containing wood (i.e. coniferous wood such as spruce). The pyrolysis of abietic acid, the primary component of resin acid, results in retene (m/z 234) and other alkylated phenanthrenes. Retene is also considered a marker compounds for coniferous wood combustion processes.[68] However, as can be seen from Fig. 5, the absolute contribution of these compounds to the aromatic signature of the coffee roast gases is very low. The observed larger masses (m/z
Revised: 14 October 2013
Accepted: 15 October 2013
Published online in Wiley Online Library
(wileyonlinelibrary.com) DOI 10.1002/jms.3299
On-line process monitoring of coffee roasting by resonant laser ionisation time-of-flight mass spectrometry: bridging the gap from industrial batch roasting to flavour formation inside an individual coffee bean R. Hertz-Schünemann,a R. Dorfner,b C. Yeretzian,c T. Streibela and R. Zimmermanna,b* Resonance-enhanced multiphoton ionisation time-of-flight mass spectrometry (REMPI-TOFMS) enables the fast and sensitive on-line monitoring of volatile organic compounds (VOC) formed during coffee roasting. On the one hand, REMPI-TOFMS was applied to monitor roasting gases of an industrial roaster (1500 kg/h capacity), with the aim of determining the roast degree in real-time from the transient chemical signature of VOCs. On the other hand, a previously developed μ-probe sampling device was used to analyse roasting gases from individual coffee beans. The aim was to explore fundamental processes at the individual bean level and link these to phenomena at the batch level. The pioneering single-bean experiments were conducted in two configurations: (1) VOCs formed inside a bean were sampled in situ, i.e. via a drilled μ-hole, from the interior, using a μ-probe (inside). (2) VOCs were sampled on-line in close vicinity of a single coffee bean’s surface (outside). The focus was on VOCs originating from hydrolysis and pyrolytic degradation of chlorogenic acids, like feruloyl quinic acid and caffeoyl quinic acid. The single bean experiments revealed interesting phenomena. First, differences in time–intensity profiles between inside versus outside (time shift of maximum) were observed and tentatively linked to the permeability of the bean’s cell walls material. Second, sharp bursts of some VOCs were observed, while others did exhibit smooth release curves. It is believed that these reflect a direct observation of bean popping during roasting. Finally, discrimination between Coffea arabica and Coffea canephora was demonstrated based on high-mass volatile markers, exclusively present in spectra of Coffea arabica. Copyright © 2013 John Wiley & Sons, Ltd. Additional supporting information may be found in the online version of this article at the publisher’s web site. Keywords: coffee; coffee roasting; aroma; resonance enhanced multiphoto-ionisation/REMPI; process monitoring
Introduction
J. Mass Spectrom. 2013, 48, 1253–1265
* Correspondence to: Ralf Zimmermann, University of Rostock, Joint Mass Spectrometry Centre, Chair of Analytical Chemistry, Institute of Chemistry, D-18059 Rostock, Germany. E-mail: [email protected] a Joint Mass Spectrometry Centre, Chair of Analytical Chemistry, Institute of Chemistry, University of Rostock, D-18059 Rostock, Germany b Joint Mass Spectrometry Centre, Comprehensive Molecular Analytics (CMA), Helmholtz Zentrum München – German Research Centre for Environmental Health, D-85764 Neuherberg, Germany c Zurich University of Applied Sciences, Institute of Chemistry and Biological Chemistry, Einsiedlerstrasse 31, CH-8820 Wädenswil, Switzerland
Copyright © 2013 John Wiley & Sons, Ltd.
1253
Green coffee contains only small concentrations of volatile- and semi-volatile organic compounds (VOC/SVOC) and a weak hay-like flavour that does not have any resemblance to the flavour of roasted coffee. When coffee is roasted, complex chemical and physical processes occur, which result in the formation of browning pigments and flavour compounds. Roasted coffee is characterised by a relatively high content of VOC and SVOC which makes it one of the foodstuffs with the richest flavour content (up 0.1% VOC/SVOC per dry weight of roasted coffee). More than 850 VOC/SVOC compounds have been identified[1] in roasted coffee (mostly in trace levels) from which about 500 are formed by the predominant Strecker and Maillard reactions.[1] However, despite extensive research on coffee flavour formation,[1–5] the knowledge about these processes remains fragmentary, and the personal experience of the coffee roast master is indispensable when it comes to control of the flavour formation process during roasting. Research on the formation of VOC in coffee roast gases has traditionally relied on chromatographic techniques, most often gas chromatography–mass spectrometry (GC-MS)[6–8] but also
high performance liquid chromatography–mass spectrometry (HPLC-MS).[9–12] A continuous, time-resolved coverage of the roasting process, however, is difficult with GC-MS or HPLC-MS, and on-line real-time analytical methods are better suited to follow the typically fast dynamics of flavour formation during the roasting process. Thus, several studies used on-line spectroscopic methods such as infrared (IR),[13,14] RAMAN spectroscopy[15,16] or sensor arrays (electronics noses).[17,18] Mass spectrometry using
R. Hertz-Schünemann et al.
1254
hard electron ionisation can be utilised in combination with the concept of the multivariate analysis, taking the individual mass channels as input parameters. This ‘MS sensor’[19,20] as a ‘fingerprint’ approach could also be applied in quality control to objectify/complement sensory testing.[21] However, fragmentation upon ionisation and the relative complexity of the chemical signature limits the usefulness of the latter approaches. Another strategy is to use on-line mass spectrometric techniques in combination with soft ionisation, which creates less ionisation induced fragmentation and simultaneously higher selectivity. In particular, on-line photo ionisation mass spectrometric techniques (PI-MS[22–25]) and on-line chemical ionisation mass-spectrometric approaches (e.g. proton-transfer-reaction, PTR-MS)[26–29] have been applied so far. Additionally, selected ion flow tube mass spectrometry (SIFT-MS) was applied for on-line analysis of trace gas molecules in air/breath samples, in combination with chemical ionisation.[30] Photo ionisation mass spectrometry[23] and later PTR-TOFMS[31] were both used to predict the roast degree of coffee by on-line analysis of the roast gas. These methods can detect VOC/SVOC in complex gaseous matrices with or without reduced interferences from fragments and inorganic bulk gases such as CO2, CO, N2 and H2O and achieve high time resolution and sensitivity and – in some cases – high chemical selectivities. The most selective of the latter MS-based on-line methods is the laser-based resonanceenhanced multiphoton ionisation process (REMPI). It is characterised by a very high selectivity and unsurpassed sensitivity for aromatic compounds. Yet, depending on the wavelength, only few non-aromatic compounds are ionised. REMPI requires the absorption of at least two laser photons to reach the ionisation energy of a molecule. For efficient REMPI, the laser UV-wavelength needs to be in resonance with the UV-absorption bands of the target molecules. By using UV-laser wavelengths in the 230–270 nm spectral range, a large number of aromatic compounds are very efficiently ionised, while aliphatic molecular species are not detected.[32–34] In 1996, REMPI-MS was applied for the first time to the analysis of coffee volatiles.[25] The REMPI method was complemented by the single photon ionisation methods with VUV-light, which has a very different selectivity than REMPI. Depending on the photon energy, it ionises the majority of the organic compounds.[32] Early SPI (single photon ionisation) and REMPI-MS studies in combination with multivariate analysis focused on the formation mechanisms of volatile phenolic roast products[22] and linked different roast degrees to the time formation behaviour of selected key compounds.[23] Due to its high selectivity and sensitivity, REMPI-MS turned out to be well suited for the on-line determination of the degree of roasting. The current work addresses two complementary applications of REMPI mass spectrometry in coffee sciences. On the one hand, and on a more applied side, REMPI-TOFMS was used to monitor the roasting process of a large industrial roaster. For this, a compact REMPI-TOFMS system was connected to an industrial scale batch roaster with an hourly roasting capacity of 1.5 metric tons of green coffee. The time–intensity profiles of a series of VOC/SVOC, which are indicative of the evolution of the roast degree, were successfully recorded. On the other hand, addressing fundamental scientific questions, the chemistry and kinetics of VOC/SVOC during roasting of individual coffee beans were examined. In this context, it is notable that the integrity of the individual coffee bean plays a decisive role for the formation of coffee flavour compounds[1,35,36]; roasting of ground green coffee does not result
wileyonlinelibrary.com/journal/jms
in the formation of the typical coffee aroma but rather in a ‘hay-like’ flavour.[37] Based on a previous study using SPI-MS,[38] a so-called μ-probe sampling device was used for sampling gases (1) from the inside of a coffee bean and (2) from just outside the surface (in the close vicinity of a coffee bean’s surface). The sampling technology was initially developed for the analysis of combustion gases within the rod of a cigarette.[39,40] Briefly, the stainless steel sampling capillary, of the μ-probe (external diameter 0.4 mm/internal diameter 0.2 mm), was coupled via a heated transfer capillary to the REMPI-mass spectrometer. The complete system is heated, so that clogging of the capillary is avoided, permitting the transfer of compounds with higher m/z values without memory effects. In this way, the high mass and time resolution of REMPI-TOFMS was combined with a unique sampling method from the interior of a single coffee bean. This work focuses on the monitoring of selected phenolic compounds, whose formation during roasting can be attributed to two different proposed chlorogenic acid degradation pathways. The application of the μ-probe technology to coffee beans allowed the observation of novel and unexpected phenomena, pertaining to dynamic processes as well as to differences in the chemical signature between Coffea arabica and Coffea canephora. First, comparing time–intensity profiles for selected VOC from inside and outside the bean, a time shift was observed for the intensity maximum between both profiles. This was tentatively linked to the diffusion of VOCs from inside towards the outside and is believed to be related to the permeability of the coffee bean cell wall material. Second, we observed for selected VOCs intense and sharp spikes in the time–intensity profiles, while other VOCs showed smooth profiles. We interpret this as a direct manifestation of popping/cracking caused by high internal pressures within the bean cell structure, and the simultaneous abrupt release of VOCs. Finally, comparing static mass spectral profiles recorded from outside single beans, we discovered that it was possible to distinguish between the two most common coffee species Coffea arabica and Coffea canephora, based on characteristic yet currently un-identified high molecular mass VOCs that appear solely in the spectra of Coffea arabica. On a more long-term perspective, this study aims at linking observations on the individual bean level with bulk-roasting phenomena and hence bridge the gap from the single bean to the bulk (large batch).[41]
Experimental and methods On-line REMPI-TOFMS on an industrial 1.5 t/h roaster For the analysis of VOC in the roast gases of an industrial coffee roaster, a compact and rugged photo ionisation-TOFMS instrument, designed for measurements under industrial application, was used. Briefly the instrument was equipped with a compact 10 Hz fixed-frequency Nd:YAG laser (Minilight-II, Continuum Inc., Santa Clara, CA, USA) with harmonic generation of 355 nm and 266 nm laser pulses. REMPI is performed with 266 nm pulses, while the 355 nm pulses are used in conjunction with a homebuilt third harmonic generation (THG) gas cell for generation of 118 nm laser pulses for SPI. The compact linear reflectron TOFMS (custom build by Kaesdorf Instrumente für Forschung und Industrie, Munich, Germany and later modified) was mounted together with turbo-molecular vacuum pumps and electronics in a 19 inch rack. The TOFMS allows an on-line detection limit
Copyright © 2013 John Wiley & Sons, Ltd.
J. Mass Spectrom. 2013, 48, 1253–1265
Single coffee bean measurements with direct MS in the low ppb range for many aromatic compounds using REMPI in the here applied configuration. The sample introduction system consists of a heated, hollow stainless steel needle (1 mm), directing a deactivated fused-silica capillary of 200 μm inner Ø to the centre of the TOFMS ion source. The fused-silica capillary, which is connected via a heated transfer line hose to the sampling system, acts as a restriction step between the ion source (pressure: 10 5 Torr) and the sampling system that is operated at ambient pressure. The sample gas flow through the capillary is about 2 ml/min. At the open end of the capillary, an effusive molecular beam is formed. The REMPI ionisation zone is located directly underneath the capillary opening in the centre of the TOFMS ion source. The whole system is mounted in a dust shielded 19 inch rack. Details and specifications of the compact system setup can be found in the literature.[42] For the on-line analysis on the industrial large-scale batch roaster (R1500/1R Probat-Werke, Emmerich, Germany), a heated sampling probe (220 °C) was introduced into the roasting gas duct through a feed through from the front side of the roasting drum (Fig. 1). The tip of the sampling probe was connected air tight via a flange, due to the slightly reduced pressure (below external atmospheric pressure) in the roasting gas duct. Roasting gas (1.5 l/min) was extracted by the heated sampling probe. The heated transfer line with the sampling capillary of the TOFMS device was coupled to the sampling train and 2 ml/min extracted at right angle from the heated sampling probe and transferred to the laser mass spectrometer. Details of sampling train setup, which was originally developed and optimised (as in previous measurements) in trials on a small-scale sample roaster, are reported in the literature.[23] The industrial roaster had a capacity 1500 kg/h with a maximal batch size of 205 kg for roasting times of 10 – 13 min. The heating was performed by natural gas (~110 MJ fuel/100 kg green coffee beans). Note, that the roasting gases are partially recycled in the process. About 180 kg of a blend of different Coffea arabicas was roasted. Once the intended roasting degree is achieved (at about 610 s roasting time and 207 °C), the exothermic roasting process was quenched by spaying 20 l water inside drum. The VOC/SVOC during the complete roasting cycles were measured using SPI and REMPI ionisation, respectively. Here, only the REMPI data is discussed.
REMPI-TOFMS of single beans using μ-probe sampling For the analysis of VOCs from single beans, a high-end REMPITOFMS instrument coupled to a μ-probe sampling device was used (Fig. 2A). Only a brief explanation of the homebuilt REMPITOFMS analyser is given here, as the system has been described in detail elsewhere.[34,43] Nd-YAG laser pulses (Continuum, Santa Clara, CA, USA, 1064 nm) were used to generate radiation at 355 nm, by frequency tripling the laser pulse (repetition rate 10 Hz, pulse width 5 ns). A beam splitter guided about 90% of the 355 nm output to an optical parametric oscillator (OPO, VISIR 2 + SHG, GWU-Lasertechnik GmbH, Erftstadt, Germany) with a thermally stabilized ß-barium borate (ß-BBO) crystal to create visible laser pulses. The remaining 10% are used to pump a third harmonic generation (THG) gas cell for generation of 118 nm pulses for single photon ionisation (SPI, not used in this work). The OPO output is frequency doubled to generate UV pulses for REMPI. Here, a wavelength of 248 nm was chosen, because previous studies showed that this wavelength is demonstrably well suited for detection of aromatic compounds.[21] By using appropriate optical elements, the beam was focused underneath the inlet needle in the ion source of the mass spectrometer. The inlet needle is connected to the sampling line and generates an effusive molecular beam in the ion source. REMPI parameters were adjusted to mostly fragmentation-free conditions (laser flux density ~ 1–5 107 Wcm 2). Ions were extracted into the flight tube of a reflectron TOFMS (Kaesdorf Instrumente für Forschung und Industrie, Munich, Germany). Two PC cards (Acquiris, Agilent Technologies, Basel, Switzerland, 250 MHz, 1 GS/s, 128 kb) recorded the resulting mass spectra, with a linear intensity range covering five orders of magnitude.[44] The mass range was set from m/z 10 to 380. Ten REMPI mass spectra were recorded per second (10 Hz) and the data recording and processing were performed by a purpose-written LabView (National Instruments, Austin, Texas, USA)-based software programme. Three to four replicates were measured for each experiment. All measurements were base line corrected and normalised to the total ion current. Only for Figs 5 and 7 original, not averaged measurements were used. A principle component analysis (PCA, Unscrambler, CAMO Software AS, Oslo, Norway)
J. Mass Spectrom. 2013, 48, 1253–1265
Copyright © 2013 John Wiley & Sons, Ltd.
wileyonlinelibrary.com/journal/jms
1255
Figure 1. Photograph (left) and scheme (right) of the industrial large-scale batch coffee roaster (R1500/1R, Probat-Werke, Emmerich, Germany) with a hourly capacity of 1.5 metric tons of green coffee and the sampling procedure. The sampling point and roast gas circulation paths are indicated.
R. Hertz-Schünemann et al.
Figure 2. A) Scheme of the _-probe and the REMPI-TOFMS. B) Overview of the roast gas sampling setup for _-probe measurements inside or in the vicinity (outside) of a single coffee bean. C) Photograph of the sampling setup for a _-probe measurement of roast gases from inside a single coffee bean.
was performed on normalised data, whereby only those 30 variables/ compounds were selected and included into the analysis that had the highest impact on the predictive model. For the single bean experiments, a μ-probe sampling setup was used. The μ-probe setup has been described elsewhere,[45] and a brief description will be given here: The μ-probe (Fig. 2B/C) consists of a conically shaped heated aluminium base body which is coupled via a heated adapter to the heated transfer line. The heart of the μ-probe is a small stainless steel capillary (ID 0.2 mm/OD 0.4 mm), which is connected to the transfer capillary using a capillary union. The μ-probe capillary sticks out of the conically shaped base body by approximately 4 mm and is thereby indirectly heated up to 120 °C (heat conductance). Two different types of μ-probe sampling experiment were performed:
section of a drilled roasted bean with the μ-probe sampling channel (a) and the ‘cut’ (b) is given in Fig. 3A. Here, the general structure of a bean is visualised with the bean endosperm (c), residues of the silverskin (d) and the amorphous zone (e). In Fig. 3B, an enhanced image of the μ-probe sampling channel is depicted. The endosperm consists of cells, which are filled with nutritious organic material (e.g. organic acids, carbohydrates, amino acids and lipids). These substances are the precursor material for the generation of coffee roasting aroma, which are decomposed during the pyrolysis process.[38,46,47] After roasting, the cells are more or less empty.[38]
i) For measurements conducted to sample roasting gases from within individual coffee beans (Fig. 2B/C), Ø 1 mm holes were drilled approximately 5 mm deep into the coffee beans. Then, the capillary of the μ-probe was inserted inside the bean and the hole was sealed by inorganic glue based on zirconium oxide (Polytec PT GmbH, Waldbronn, Germany). The hole was closed throughout the complete measurement. ii) For measurements conducted to sample roasting gases at the surface of individual coffee beans, coffee beans were placed individually in a 2 ml glass flask, which then was connected to the μ-probe (Fig. 2B).
Monitoring the coffee roasting process gas from an industrial coffee roaster
A hot air gun was taken for simulating the roast process, using a mantle thermocouple for controlling the roast temperature (T = 250 °C) on the surface of the bean (case i) or inside the glass (case ii) during heating, respectively. For the experiments, green Coffea arabica beans from Bolivia and green Coffea canephora beans from India (both in organic quality) were used.
Coffee bean morphology
1256
Raster electron microscopy (REM) images for visualisation of the location and the dimension of μ-probe sampling channel inside the coffee bean were recorded at the Electron Microscopy Centre of the University of Rostock. (EMZUniRo) In Fig. 3, two selected REM images are presented. The dimension scale is given in the lower right corner of the respective figures. The cross
wileyonlinelibrary.com/journal/jms
Results and discussion
Due to the high selectivity and softness of the ionisation step, photo ionisation mass spectrometry is well suited for the monitoring of process gases.[33,48,49] New developments in mass spectrometry allow the setup of robust process monitoring devices, based on REMPI-TOFMS. While the usage of such sophisticated technologies may currently seem expensive for process control and monitoring in food processing, applications for high value products and processes, such as coffee, are a promising option (the market value of high quality raw Arabica is less than 1/10 of the value of the processed commodity). Once optimised applications are established, customised and cost-effective instruments are expected to bring costs down. The compact linear reflection REMPI-TOFMS uses the output of a small 10 Hz Nd:YAG laser (Minilight-II, Continuum Inc., Santa Clara, CA, USA)[42] for REMPI ionisation. Figure 4 (left) shows a typical REMPI mass spectrum obtained from the roasting off-gas at 623 s roasting time (begin of the water quenching process). The spectrum is averaged over 50 laser shot (5 s). The REMPI mass spectra obtained at the industrial roaster agree well with results reported in previous tests on small-scale sample roasters.[22,23] Note, that the ionisation laser wavelength of 266 nm enhances the phenolic compounds in the REMPI mass spectrum (m/z 94, m/z 108, m/z 110, m/z 124, m/z 150, m/z 164), with 4-vinylguaiacol (m/z 150) as base peak, due to a high ionisation cross section. In
Copyright © 2013 John Wiley & Sons, Ltd.
J. Mass Spectrom. 2013, 48, 1253–1265
Single coffee bean measurements with direct MS
Figure 3. A) Raster Electron Microscope (REM) photograph of a cross section through a roasted coffee bean. Legend: a - drilled channel for μ-probe sampling, b - ‘cut’, c - endosperm (bean), d - residues of the silverskin, e - amorphous zone[46] B) Raster Electron Microscope (REM) photograph of the drilled channel for -probe sampling in the coffee bean.
Figure 4. REMPI-TOFMS measurement on an R1500/1R industrial large-scale batch roaster (Probat-Werke, Emmerich, Germany). Left) REMPI-TOFMS mass spectrum (@ 266 nm) at the end of the roasting process, at 620 s (begin of water quench); m/z values are indicated. Mainly phenolic compounds can be assigned (see text). Right) Time–intensity profile of the mass traces 150 m/z (4-vinylguaiacol) and 124 m/z (guaiacol), normalised to the maximum intensity, respectively. The onset of formation of 4-vinylguaiacol is slightly earlier than for guaiacol.
J. Mass Spectrom. 2013, 48, 1253–1265
can achieve a robust prediction of the roast-degree also under such complex industrial conditions. Single bean roasting Besides monitoring the roasting process at the industrial scale, the outstanding sensitivity of REMPI-TOFMS makes it uniquely suited to address fundamental questions related to the thermal treatment of single coffee beans. Here, we will discuss first results on REMPI-TOFMS analysis of VOC/SVOC generation and release from individual coffee beans during roasting (Arabica as well as Robusta beans). The motivation to this research is twofold: first, the single bean is the smallest processing unit (the “atom”) of coffee roasting. Linking processes at the level of the elementary unit of roasting to phenomena at the batch scale, we expect to gradually better comprehend the basic relations governing industrial roasting. Second, it is know that the chemical reactions and in particular the formation of the coffee aroma compounds during roasting strongly depend on the micro-conditions and micro-structure of the intact coffee bean.[50,51] Shifting the focus from large-scale batch roasting (180 kg green coffee/roasting batch) to the single bean level (~150 mg/per roasting), we observed a series of surprising phenomena that either pertain to the time–intensity release profile from the single beans, or to differences in the mass spectra between Robusta and Arabica beans.
Copyright © 2013 John Wiley & Sons, Ltd.
wileyonlinelibrary.com/journal/jms
1257
addition to phenols, indole (m/z 117), caffeine (m/z 194) and a C3-amine (m/z 59) are detectable as well. The weak signals at m/z 18 and m/z 28 originate from residual electron ionisation (EI) which is due to photo-induced electrons, accelerated in the electrostatic extraction field and which lead to EI ionisation of water and nitrogen. In the right part of Fig. 4, the time–intensity profiles of phenolics guaiacol (m/z 124) and 4-vinylguaiacol (m/z 150) during the roasting process are shown. The shaded region indicates the quenching of the roasting process by injection/spaying of water. The depicted compounds have been identified as roasting degree markers in previous work, with 4-vinylguaiacol being a product of the decarboxylation of ferulic acid – formed and released rather early in the process. This is evident in the roasting gas measurement at the industrial roaster as well. However, the differences in the release profiles among different compounds are less pronounced in the off-gas of the industrial roaster, compared to previous laboratory roaster experiments on 300 g sample roaster.[23] This is due to recycling a large fraction of the hot roast gas within in the industrial roaster: only a fraction of the roast gases are replaced at each cycle, e.g. added at the gas burner and discharged via the exhaust duct. While such recirculation of the roast gas may slightly complicate the on-line monitoring of the roast degree based on time–intensity profiles of VOC/SVOC in the off-gas of the roaster, a multicomponent model using multivariate statistical methods is
R. Hertz-Schünemann et al. Arabica vs. Robusta: While investigating the roasting process of individual Robusta and Arabica coffee beans, by sampling VOC/SVOC either (1) inside or (2) outside of the single bean, we observed consistent differences in the mass spectra between the two species. Up to date, a fast differentiation between the most common coffee species, Coffea arabica and Coffea canephora (Robusta) remains a challenging analytical task. In Fig. 5, REMPI-TOFMS data sampled via the μ-probe from inside bean are shown for Arabica and Robusta. The three-dimensional plots (Figs 5 A and 5D) depict an overview on exemplary single bean roasting results for an Arabica (5A) and a Robusta (5D) bean, respectively. A larger number of such 3D overview plots, allowing assessing the
variability, are given in the supplemental material (Figs S1 and S2). It can also be seen from the supplemental material that the variability is higher for the measurements inside relative to outside the beans (Table 1). This probably reflects the fact that the process of drilling the sampling channel as well as the position the channel needs to be further examined and standardised. As there are different tissues and morphological structures in the coffee beans,[38] a systematic investigation of the influence of the sampling point and channel depth needs to be performed in future studies. However, in general, the profiles are dominated by the masses m/z 150 and m/z 194, which can be assigned to be predominately 4-vinylguaiacol (m/z 150) and caffeine. A striking and consistent
Figure 5. REMPI-TOFMS measurements of coffee roasting gases extracted from the interior of individual coffee beans (inside) A) 3D representation of VOC/ SVOC time intensity profiles sampled by the +-probe from inside a single Arabica coffee bean B) REMPI-TOF mass spectrum measured inside of single Arabica beans (average of 4 measurements), plotted on a linear intensity axis. C) REMPI-TOF mass spectrum measured inside of single Arabica beans (average of four measurements), plotted on a logarithmic intensity axis. D) 3D representation of VOC/SVOC time intensity profiles sampled by the +-probe from inside a single Robusta coffee bean. E) REMPI-TOF mass spectrum measured inside of single Robusta beans (average of four measurements), plotted on a linear intensity axis. F) REMPI-TOF mass spectrum measured inside of single Robusta beans (average of four measurements), plotted on a logarithmic intensity axis.
Table 1. Maximal standard deviation values (sd) obtained for the 20 s sequences of the time profiles (four replicate measurements, compare with Fig. 8) of Robusta measurements inside (RI) and outside (RO) the bean (the minimal sd is 0.0 in all cases) m/z
1258
sd RI sd RO
59
94
110
117
124
136
150
194
max ±
0.73
0.62
0.54
0.84
0.70
0.59
0.53
0.90
max ±
0.28
0.22
0.17
0.19
0.28
1.00
0.34
0.21
wileyonlinelibrary.com/journal/jms
Copyright © 2013 John Wiley & Sons, Ltd.
J. Mass Spectrom. 2013, 48, 1253–1265
Single coffee bean measurements with direct MS difference between measurements on Robusta and Arabica is the occurrence of masses above m/z 200 for Arabica bean (see Figs S1 and S3 in the supplemental material). In the case of Robusta, these higher mass traces are nearly absent (Note that there are also variations in the observed intensity of the higher mass signals (m/z > 200), but the signals are always significantly higher in Arabica). This is also visible in the averaged mass spectra of three single bean roasting measurements for Arabica and Robusta, respectively (Figs 5 B/C and E/F). In particular, the REMPI mass spectra with logarithmic intensity axis (Figs 5C and 5F) show these differences very clearly. Although many molecular ion masses, which appear below m/z 200 in photo ionisation mass spectrometry (REMPI and SPI), can be identified/ assigned,[22–25,38] very little is known about the compounds that lead to ion signals in REMPI-TOFMS at masses above m/z 200. In Table 2, the REMPI active compounds, which were previously assigned, are listed together with the observed concentration range from literature data. The previously assigned compounds in the case of REMPI at 266 nm or 248 nm are phenolic compounds from the decomposition of chlorogenic acids. The isobaric alkylated pyrazines make a smaller yet not negligible contribution to the mass spectral ion intensities, since
Table 2. Assignment of observed masses in the REMPI mass spectra based on published literature and previous studies.[22–25,38] The concentrations are given for roast and ground coffee in [ppm] [1,35] and for extract or brew in [μg/l].[54,55] m/z 59 94 96 99 110 117 124 126 132 136 138 146 150 152 162
178
192 194
Substances C3-alkylated amine Phenol, Methylpyrazine Furfural Methylthiazole 1,2-Benzenediole, Methylfurfural Indole Guaiacol, Methylbenzenediole Hydroxymethylfurfural, Benzenetriole, Methylbenzofuran Vinyl-1,2-benzenediol Ethylcatechol, Dihydroxybenzaldehyde Phenylbutenal Furanylpyrazine Vinylguaiacol Vanillin, Ethylguaiacol Dihydroxy cinnamaldehyde, (Furanylmethyl)5-methylfuran Anthracene, Phenanthrene, Difurfurylether Methylphenanthrene, Methylanthracene Caffeine
J. Mass Spectrom. 2013, 48, 1253–1265
0.5–2.0[ppm][1] 1.55–14[ppm][1] 6.6–80[ppm][1] 22.5–85[ppm][1] 0.4–6[ppm][1] 20–340[ppm][1] 96.6–160[ppm][1] 0.3–2[ppm][1] 1.6–95.5[ppm][1] 31–75[ppm][35] 10–35[ppm][35] 0.15–130[ppm][1] 0.01–0.044[ppm][1] 4.4–40[ppm][1] 9–106[ppm][1] 0.1–20[ppm][1] 0.1–1.5[ppm][1] 0.6–0.7[ppm][1] 6–177.7[ppm][1] 2[35]–100[ppm][1] 0.3–18.1[ppm][1] 5–12[ppm][1] identified[1] 0.009–0.9 [μg/l][54,55] 0.124–0.9 [μg/l][54,55] 0.6–1.4[ppm][1] 0.002–0.317 [μg/l][55] 0.002–0.221 [μg/l][55] 1.1–2.8 [% DM][1]
Copyright © 2013 John Wiley & Sons, Ltd.
wileyonlinelibrary.com/journal/jms
1259
DM = Dry Matter
c
heterocyclic nitrogen compounds such as pyridine and pyrazine are not efficiently ionised in a REMPI process, at the wavelengths applied here (266 nm and 248 nm). Larger nitrogen heterocyclic compounds such as caffeine (m/z 194) and indol (m/z 117) are clearly seen in the REMPI mass spectra as well. Currently, no assignment can be made for this m/z ion signal at higher masses that appear to be characteristic for Arabica. Furthermore, these features have, until now, only been observed in single bean experiments, using the high end instruments and the heated μ-probe sampling setup. To be REMPI active, these compounds need to be aromatic or at least highly conjugated. Therefore, only specific substance classes are potential candidates such as flavonoid or polyphenolic structures. Polyphenols are yet known to be more abundant in Robusta (i.e. due to the higher chlorogenic acid contend). Note that REMPI can be extremely sensitive for aromatic compounds, so the absolute concentration of the compounds might be very low. Nevertheless, the clear differences in the higher mass region may render a REMPI-TOFMS analysis a valuable mean for a fast discrimination of Arabica from Robusta. Differences in m/z values can also be originated from different locations and/or growth conditions and/or pre-processing technologies. More investigations on different varieties of the two species as well as on the influence of the farming region, agronomic practices and different postharvest processing techniques (e.g. wet and dry processing) are needed at this point. For further analysis of the difference between the REMPI mass spectra from Arabica and Robusta, a so-called difference spectrum was calculated: the averaged (four measurements each) and normalised mass spectrum of coffee Robusta were subtracted from the one generated from Arabica measurements, respectively. The difference mass spectrum is given in Fig. 6. Compounds (i.e. m/z values) being more abundant in Arabica beans show up as positive peaks while compounds with higher concentration in Robusta exhibit negative peaks. Note that the absolute concentration cannot be determined from this plot. The molecular assignment of the REMPI detectable m/z values is listed in Table 2: Based on literature data of coffee aroma constituents and their observed concentrations as well as on the physical and chemical properties of the respective molecules (ionisation energies, photo ionisation cross sections), the m/z values observed by photo ionisation mass spectrometry were assigned to chemical compounds.[1,35,52–55] The assigned compounds are well supported by the literature (predominantly by GC-MS). Especially for the higher m/z values, however, it is increasingly difficult to achieve an unambiguous assignment. The difference spectrum clearly depicts the distinguishing chemical signature between Robusta and Arabica. From literature, it is known that Robusta coffee contains more sulphur compounds,[56,57] more caffeine[58] and more chlorogenic acids. [57–62] The higher content of chlorogenic acids results in an increased formation of phenolic compounds such as vinylguaiacol[1,62] (m/z 150), 1.2-benzenediole (m/z 110), phenol (m/z 94)[1] and as well as vinyl-1,2-benzenediol (m/z 136). Also, higher amounts of indole (m/z 117) were observed in Robusta beans. Indole is formed by the degradation of phenylalanine predominately at higher roasting temperatures (230–260 °C).[1] The higher caffeine content of the Robusta varieties is reflected as well in the difference spectrum. In contrast, the difference spectrum (Fig. 6) shows higher intensities of many compounds for Arabica beans. At lower masses, the C3-alkylamine at m/z 59, which is formed by pyrolysis of sulphur containing amino acids[1] (mainly methionine and cystine[63]), is an outstanding example. Casal et al. found more free methionine (in trace) in Arabica than
R. Hertz-Schünemann et al.
Figure 6. Difference of averaged and normalised REMPI mass spectrum of Arabica minus Robusta. Spectra correspond each to the averaged VOC/ SVOC from four different beans, sampled from outside the coffee bean.
in Robusta coffee beans.[64] The presence of furfural (m/z 96)[60] is notable and results from the higher carbohydrate contend of Arabica coffees. Furfural is formed by decomposition of pentosane or glucose and also by oxidation of furfuryl alcohol,[1] whereas the furfurylalcohol is a product of the reaction of (deoxy)ribose or sucrose with cycstein/methionine.[65,66] Beyond that, Sanz et al. reported similar contents for methylthiazole (m/z 99) in Arabica and Robusta coffee,[56] which is generated via condensation of carbohydrate fragments and cysteamine (degradation product of cysteine).[67] Also, relatively higher signal intensities for m/z 162 and m/z 132 are likely due to furan-based carbohydrate decomposition products (e.g. 2-(2-furanylmethyl)-5-methyl-furan[1] - m/z 162, methylbenzofuran - m/z 132). Hence, much of the observed differences at lower masses between Arabica and Robusta can tentatively be rationalised, based on published literature on the composition of the Arabica and Robusta varieties. Interesting is the homologous series of m/z 178, m/z 192, m/z 206, m/z 220, m/z 234 which belongs to the alkylated phenanthrenes (i.e. the substance class of the alkylated polycyclic aromatic hydrocarbons). Alkylated phenanthrenes are well known to be formed in relatively high abundances in pyrolysis processes of resin-containing wood (i.e. coniferous wood such as spruce). The pyrolysis of abietic acid, the primary component of resin acid, results in retene (m/z 234) and other alkylated phenanthrenes. Retene is also considered a marker compounds for coniferous wood combustion processes.[68] However, as can be seen from Fig. 5, the absolute contribution of these compounds to the aromatic signature of the coffee roast gases is very low. The observed larger masses (m/z
![[Investigation on coffee and coffee-substitutes. XVII. Behaviour of polysaccharide-complexes of robusta-coffee during roasting].](/assets/img/hold.png)
