After water and tea, beer is the most widely consumed drink worldwide, and the top consumed alcoholic beverage. Beer is made up of 93% water and natural ingredients such as malted barley, other cereals, hops, and yeast. Many different varieties are made by producers ranging from multinational industries to small operators, regional craft breweries, and brewpubs. Each recipe and production method confers a distinctive character and flavor to the beer, and this quality differentiation of beers—as for many other foodstuffs—has become imperative in a market where consumers are ever more critical, demanding, and fragmented in their choices. It has been shown that the three main cues driving consumers' cognitive mechanisms are brand, country of origin, and quality attributes of the product.
Therefore, for purposes such as authentication and on-line production monitoring, it is useful to be able to distinguish the main attributes that characterize different types of beers. With this aim in mind, we designed an experiment1 to determine whether optical methods can be used to recognize different classes of beers and identify the main characteristic of beer, which is its alcohol content.
Today, optical methods—and especially spectroscopy—provide a modern and effective way to perform fast, non-destructive, and reagent-free analysis of substances. Spectroscopy represents a ‘green’ and sustainable approach to analysis because it allows samples to be measured ‘as-is,’ without need for chemical treatment. This means there is no problematic waste to discharge, and safer handling of samples.
Figure 1. Differentiating the main attributes that characterize beers has application in quality monitoring and preventing food fraud.
For our experiment, we used a collection of 86 beer samples, each of a different brand. Of these, 50 were beers produced in Belgium, while the remainder came from other countries such as Italy, Germany, Denmark, England, the Netherlands, Japan, Czech Republic, Cuba, and Mexico (see Figure 1). The samples covered the principal ‘styles’ of beers, with top-fermented (e.g., golden ale, dark ale, weiss), bottom-fermented (e.g., lager, doppelbock, Mexican), and spontaneously fermented (e.g., golden and cherry lambic) varieties, with different colors and alcoholic strengths ranging from 0.5 to 11% by volume.
Figure 2. The diffuse-light absorption spectroscopy setup using optical fiber technology. VIS: Visible. NIR: Near-IR.
We analyzed the entire collection using both conventional optical measurements and a custom instrument setup for diffuse-light absorption spectroscopy in the visible and near-IR bands, consisting of an integrating sphere and optical fiber spectrometers (see Figure 2). The advantage of this arrangement is that the measured absorption is independent of any scattering (due to suspended particles in the samples), and so unaffected by the natural turbidity of the beer, which can instead confuse traditional absorption spectroscopy. To complete the optical characterization of the beer collection, we also measured other physical parameters such as turbidity and refractive index using standard optical instruments. Essentially, the diffuse-light absorption spectroscopy provided information about color (the visible band) and alcohol content (the near-IR band),2 while the turbidimetry and refractometry yielded an indication of the cloudiness of the beer and its Brix value (a measure of the sugar content of an aqueous solution, commonly assessed with a refractometer to monitor the progress of fermentation).3
Figure 3. Diffuse-light absorption spectra of the beer collection. Left: Visible spectra. Right: First derivative near-IR spectra. I.S.: Integrating sphere.
From the scattering-free near-IR spectra (see Figure 3) we obtained a straightforward prediction of alcohol content by applying partial least squares (PLS) regression, a widely used technique for predicting quantitative variables in a multicomponent mixture.4 We first calibrated the PLS model using 48 randomly selected beers, and then tested a validation model on the 38 remaining beers. The results (see Figure 4) confirm that our novel optical setup can be used for continuous online monitoring of alcohol content during beer production, without being confounded by variations in turbidity.
Figure 4. Prediction of the alcohol content of beer: results of calibration (left) and validation (right) models.
Figure 5. Beer clustering according to class (left) and classification map distinguishing Belgian beers from all the others (right). PC: Principal component. LDA: Linear discriminant analysis.
We also combined the scattering-free visible and near-IR data together with the turbidity and refractive index values to create a data matrix, and processed this by principal component analysis (PCA), a standard method for data dimensionality reduction and exploratory analysis.5 The procedure successfully grouped the beers in the collection according to their fermentation method and color: see Figure 5 (left). Because these are two key traits characterizing individual varieties, this indicates that optical technologies can be used to differentiate among classes of beers. Finally, since half of the beers were typical Belgian beers, we also applied linear discriminant analysis6 (LDA)—a common classification technique—to the optical data, and found that it successfully distinguished the Belgian beers from all the others: see Figure 5 (right).
In summary, optical methods combined with multivariate data analysis proved effective in classifying beers according to their main distinguishing attributes, such as the alcohol content, the type of beer, and its country of origin. This suggests that optics offers a straightforward method for monitoring beer quality indicators during production, as well as for authentication in food fraud prevention. Following up on these findings, we are currently working on implementing a prototype for on-line analysis, capable also of monitoring nutraceutical indicators such as carbohydrates and vitamins during production.
This work was sponsored in part by the Belgian Science Policy Office-Interuniversity Attraction Poles (DWTC-IAP), the Research Foundation Flanders (FWO), the 6th Framework Programme (FP) European Network of Excellence on Micro-Optics (NEMO), the 7th FP European Network of Excellence for Biophotonics (Photonics 4 Life), the Methusalem and Hercules foundations, and the Research Council (OZR) of Vrije Universiteit Brussel.
Anna Grazia Mignani, Leonardo Ciaccheri
‘Nello Carrara’ Institute of Applied Physics (IFAC)
National Research Council (CNR)
Sesto Fiorentino, Italy
Anna Grazia Mignani holds an MSc in physics and a PhD in non-destructive testing from the University of Florence (UNIFI) in Italy. She joined CNR-IFAC in 1983, and is currently a senior scientist. Her research interests include fiber optic and micro-optic sensors, guided-wave components for sensing applications, and sensor networks.
Leonardo Ciaccheri holds an MSc in physics and PhD in non-destructive testing from UNIFI. In 2003 he joined the Fiber Optic Group at CNR-IFAC, and has been on the permanent staff since 2009. His research interests are in spectroscopy for fiber optic sensors and multivariate analysis.
Heidi Ottevaere, Hugo Thienpont
Brussels Photonics Team (B-PHOT)
Vrije Universiteit Brussel (VUB)
Heidi Ottevaere has an MSc in photonics and a PhD in applied sciences from VUB, where she has been a full professor since 2009. Her research efforts are in micro-photonics and micro-optics, especially plastic micro-optics based on deep proton writing, interferometric measurement techniques, microlens characterization, and biophotonics.
Hugo Thienpont is a full professor at VUB. He chairs the Applied Physics and Photonics Department and is director of its photonics research group, B-PHOT. Over the years he has worked on a wide variety of fundamental and applied research topics, most of them relating to the fields of micro-photonics and micro-optics.
1. A. G. Mignani, L. Ciaccheri, A. A. Mencaglia, H. Ottevaere, E. E. Samano-Baca, H. Thienpont, Optical measurements and pattern-recognition techniques for identifying the characteristics of beer and distinguishing Belgian beers, Sens. Actuators B
, 2012. doi:10.1016/j.snb.2012.10.029
2. B. R. Buchanan, D. E. Honigs, C. J. Lee, W. Roth, Detection of ethanol in wines using optical-fiber measurements and near-infrared analysis, Appl. Spectrosc. 42, p. 1106-1111, 1988.
3. P. Fisicaro, F. Durbiano, F. Giuffredi, Laboratory accreditation in Italy for reference solution production of sucrose in water, Accredit. Qual. Assur. 11, p. 336-342, 2006.
4. S. Wold, M. Sjöström, L. Erikkson, PLS-regression: a basic tool for chemometrics, Chemom. Intell. Lab. Syst. 58, p. 109-130, 2001.
5. J. E. Jackson, A User's Guide to Principal Components, Wiley, Hoboken, 2003.
6. Handbook of Chemometrics and Qualimetrics, Elsevier Science BV, Amsterdam, 1998. ch. 33