Modified stainless steel surfaces targeted to reduce fouling ...

Journal of Food Engineering 80 (2007) 1176–1187www.elsevier.com/locate/jfoodengModified stainless steel surfaces targeted to reduce fouling –Evaluation of fouling by milk componentsRoxane Rosmaninho a, *, Olga Santos b , Tommy Nylander c , Marie Paulsson b ,Morgane Beuf d , Thierry Benezech d , Stergios Yiantsios e , Nikolaos Andritsos e ,Anastasios Karabelas e , Gerhard Rizzo f , Hans Müller-Steinhagen f , Luis F. Melo aa LEPAE, Department of Chemical Engineering, Faculty of Engineering,University of Porto, Rua Dr. Roberto Frias, s/n, 4200-465 Porto, Portugalb Department of Food Engineering, University of Lund, P.O. Box 124, S-221 00 Lund, Swedenc Department of Physical Chemistry II, University of Lund, P.O. Box 124, S-221 00 Lund, Swedend INRA Laboratoire de Genie des Procedes et Technologie Alimentaires, 369 Rue Jules Guesde, BP 39, F-59651 Villeneuve d’Ascq Cedex, Francee Chemical Process Engineering Research Institute – CERTH, 6 km Charilaou – Thermi Road, P.O. Box 361, GR-57001 Thermi, Thessaloniki, Greecef Institute for Thermodynamics and Thermal Engineering, University of Stuttgart, Pfaffenwaldring 6, 70550 Stuttgart, GermanyReceived 16 December 2005; received in revised form 8 September 2006; accepted 9 September 2006Available online 13 November 2006AbstractSeveral stainless steel based surfaces with different properties were evaluated according to their fouling behaviour for different dairyproducts under different conditions. Surface properties were obtained by the following modification techniques: SiF þ 3 , MoS2þ 2and TiCion implantation; diamond-like carbon (DLC) sputtering; DLC, DLC–Si–O and SiO x , plasma enhanced chemical vapor Deposition(PECVD); autocatalytic Ni–P–PTFE and silica coating. Aqueous solutions that simulate milk (SMUF – simulated milk ultrafiltratefor the mineral components, b-lactoglobulin for the protein components and FMF – fouling model fluid for complex milk systems) wereused to study the fouling behaviour during pasteurisation. Bacteriological deposition studies were also performed with two heat resistantstrains of Bacillus. The experiments were carried out at laboratory scale for the evaluation of calcium phosphate and protein deposition,and at pilot scale for adhesion of bacteria and deposits from complex milk systems.In all cases, the fouling behaviour was affected by the surface material, although in different ways for the deposition or the cleaningphases. For the non-microbiological deposits (calcium phosphate, whey protein and FMF milk-based product), the Ni–P–PTFE surfacewas the most promising one, since it generally promoted less deposit build up and, in all cases, was the easiest to clean. On the otherhand, for bacterial adhesion, the most suitable surface was the ion implanted (TiC) surface, which also showed less spores after the cleaningprocess.Ó 2006 Elsevier Ltd. All rights reserved.Keywords: Modified stainless steel; Fouling; Calcium phosphate; b-Lactoglobulin; Bacillus subtilis; Bacillus cereus1. Introduction* Corresponding author. Tel.: + 351 22 5081685; fax: + 351 22 5081449.E-mail address: roxane@fe.up.pt (R. Rosmaninho).The formation of unwanted deposits on the surfaces ofheat exchangers, known as fouling, is an inherent problemin most industrial processes. Environmental and energyconservation demands have reinforced its socio-economicrelevance in the last 10–15 years, but the underlying phenomenaare in many instances not yet well understoodand tools for mitigating its effects are not yet satisfactory.Well established chemical and mechanical fouling mitigationstrategies are available for some industrial cases. Mostof them necessitate stopping the heat transfer process forcleaning and the cost is often high in terms of productionlosses and labour. Modification of the fluid compositionby introducing chemical additives can be effective but thisis not possible in many industrial applications such as inthe dairy industry (Müller-Steinhagen, 1998). For these0260-8774/$ - see front matter Ó 2006 Elsevier Ltd. All rights reserved.doi:10.1016/j.jfoodeng.2006.09.008

R. Rosmaninho et al. / Journal of Food Engineering 80 (2007) 1176–1187 1177cases, attention must be paid to other factors known toaffect milk fouling. There are many significant physicochemicalfactors related to both the composition of milkand the properties of the heat exchanger surface thatshould be considered (Changani, Belmar-Beiny, & Fryer,1997). As has been suggested in recent years (Forster &Bohnet, 1999), a deeper knowledge on the relation betweenfouling material and the properties of the fouled surfaces isessential in order to delay or minimize fouling as well as todetermine optimal cleaning conditions for a given process.The aim of the present work is, therefore, to investigate thefouling behaviour of several milk components, milk relatedmicroorganisms and dairy products on different stainlesssteel surfaces previously characterized according to theirsurface properties (Santos et al., 2004).Milk is a complex biological fluid composed of severalcomponents including whey proteins and calcium phosphate,which are involved in the fouling process throughinteracting mechanisms. The whey proteins, particularlyb-lactoglobulin, foul after denaturation and aggregation,while calcium phosphate fouls as a result of local supersaturation(Changani et al., 1997). Fouling starts as soon as adairy product is brought into contact with a stainless steelsurface albeit at a molecular level and invisible to the nakedeye. It is believed that fouling starts with whey proteinadhesion at room temperature. Upon heating, the wheyproteins start to unfold and expose a free S–OH group,changing into an activated state which allows attachmentto the protein layer initially formed (Visser & Jeurnink,1997). This first protein layer, however, is affected by thedeposition surface since protein adsorption onto solid surfacesis the result of several interactions occurring betweenthe protein, the surface and the solvent with other solutespresent in the system (Haynes & Norde, 1994). These interactionsare significantly influenced by the surface properties,more precisely its surface energy. In previous studieson the adsorption of protein on surfaces with different surfaceenergies, no clear quantitative relationship was foundbetween the amount of protein deposit and the surfaceenergy (Addesso & Lund, 1997; Janocha et al., 2001; Yoon& Lund, 1994). Another major contribution to fouling inmilk processing is the presence of calcium and phosphateions. Upon heating of milk, part of these ions will tendto precipitate as a calcium phosphate salt (owing to theinverse solubility of calcium phosphate with temperature)which ultimately will form a mineral deposit on the stainlesssteel surface in a typical crystallization fouling process(Bott, 1995; Krause, 1993). Even when the deposits werepredominantly proteinaceous, a calcium phosphate layerwas found next to the surface due to migration of mineralions (Changani et al., 1997). Analysis of deposits formedon heat exchangers after heat processing showed them tobe a mixture of several calcium phosphate forms like dicalciumphosphate dihydrate (DCPD) and octacalcium phosphate(OCP) which at prolonged heating are eventuallytransferred to the least soluble and more stable calciumphosphate complex hydroxyapatite (HAP) (Visser & Jeurnink,1997). The crystallization process of calcium phosphateis, therefore, a very complex problem since severalforms can be formed at the same time depending on ambientconditions like the pH of solution, the temperature andthe presence of other milk components. To control the lasttype of interference, a simplified milk system is normallyused in fouling experiments. The most widely used mineralfouling solution, also used in the present work, is calledsimulated milk ultrafiltrate (SMUF) which was firstdescribed by Jenness and Koops (1962). This solution hasbeen used on a variety of fouling studies, from calciumphosphate crystallization (Andritsos, Yiantsios, & Karabelas,2002) to the fouling caused by whey permeates (Morison& Tie, 2002) with results comparable to reality. Inaddition to the above-mentioned bulk properties, crystallizationfouling is also known to be dependent on the depositionsurface characteristics (Bott, 1995). Consequentlysome studies were made on the effect of surface energyon the type and amount of crystallization deposit formed.Forster and Bohnet (1999) found a relation between surfaceenergy and the induction period of calcium sulphatefouling, indicating the advantage of using low energy surfaceson heating equipments. Later on, these authors (Forster& Bohnet, 1999) proved that the analysis of theinterfacial energy between surface and calcium sulphatedeposit can be used as a tool to predict the optimum surfaceenergy to mitigate fouling. The effect of surface energyon the nucleation and deposit development of calciumphosphate was also studied by Liu, Wu, Sethuraman,and Nancollas (1997) and by Wu, Zhuang, and Nancollas(1997). In both cases, it was found that nucleation anddeposit development was higher for surfaces having highvalues of surface energy. A different opinion was howeverpresented by Zhao and Müller-Steinhagen (2001) basedon a theoretical approach on the adhesion forces of calciumsulphate on modified heat transfer surfaces.Whey protein and calcium phosphate fouling are twodifferent processes that follow different kinetics. Althoughthey may interact in real milk processes, there is still a needto know more about the surface effect on each one of thesesprocesses. Consequently, the build up of minerals and proteinsin a fouling layer may be studied independently.An additional approach to dairy heat exchanger equipmentfouling, which has also been the subject of severalstudies (e.g. Boulangé-Petermann, 1996; Melo & Bott,1997), is to focus on bacterial adhesion on stainless steelbecause of its high importance related to food safety reasons.The species studied in the present work were Bacilluscereus and Bacillus subtilis which are known to develop infoods during storage and have often been responsible forfood borne diseases (Luby, Jones, Dowda, Kramer, &Horan, 1993). They have also been chosen for their abilityto adhere to various materials (Faille, Lebret, Gavini, &Maingonnat, 1997), for their resistance to heat and chemicalsin poultry meat (Faille et al., 1997) and in milk productprocessing lines (Faille, Fontaine, & Bénézech, 2001) sincethey are able to survive heat treatment and a cleaning-in-

1178 R. Rosmaninho et al. / Journal of Food Engineering 80 (2007) 1176–1187place (CIP) procedure. A relation between cell adhesionand surface energy was also investigated, but no consistenttheory could yet be found. Tsibouklis et al. (2000) defendedthe use of low energy surfaces to inhibit bacterial adhesion,but on the other hand Zhao (2004) presented the existenceof an optimum value of surface energy for which bacterialadhesion is minimized of around 30 m N/m.The different stainless steel based surfaces used for thefouling experiments were characterized according to theirsurface energy, chemical composition, roughness, andtopography and the results were recently published (Santoset al., 2004). The challenge of this part of the work wastherefore to find some relation between the fouling problemscaused by different components of milk and the surfaceenergy properties of the material, in order to find theoptimal stainless steel surface to be used in the constructionof heat exchanger equipment for the dairy industry.2. Materials and methods2.1. Surfaces and test fluidsThis work covers the fouling behaviour of different milkcomponents and milk products on different stainless steelbased surfaces. The description of both the different surfacesand the modification techniques used to obtain them,and the detailed description of the different foulant productsare presented below.2.1.1. Stainless steel surfacesThe stainless steel surfaces tested in this work were 316with 2R (cold rolled, bright annealed) finish, received froma major European stainless steel manufacturer. Some stainlesssteel samples were modified by the following surfacemodification techniques: direct and turbulent ion implantation(with SiF þ 3and MoS 2þ2ions); sputtering (thin diamond-likecarbon (DLC) films); plasma chemical vapordeposition (CVD) (thin films of DLC, SiO x , and DLC–Si–O) and an autocatalytic Ni–P–PTFE coating. All surfacemodifications were produced at or provided by theUniversity of Stuttgart (Germany), with the exception ofthe silica coating, which was produced by CERTH, Universityof Thessaloniki (Greece). More detailed informationconcerning the modifications can be found in Santoset al. (2004).2.1.2. Surface modification techniques2.1.2.1. Direct and turbulent ion implantation. In the direction implantation (or ion beam implantation) process, thesurfaces were bombarded with highly accelerated SiF þ 3 ionswith an average energy of 200 keV, achieving an ion concentrationof 5 · 10 16 ions/cm 2 . In this case, the ions weredirectly implanted into the metal material, forming a surfacealloy of approximately 0.2 lm in depth. MoS 2 ionswere implanted in the stainless steel surfaces by TurbulentIon Implantation. In this technique, the atoms impactingthe surface of the substrate had such a high energy thatthey could penetrate into the interior of the material, sincemolecules hitting the surface could push already depositedmolecules further inside the substrate, leading to implantationdepths of up to 10 lm and resulting into modified surfacelayers of 2–5 lm thickness.2.1.2.2. Sputtering and plasma CVD technique. Sputtercoating as well as plasma CVD – or plasma enhancedCVD – coating are plasma processes, which take place atsub-atmospheric pressures. Both processes belong to thethin film technology where the thickness of the coating isusually less then 2 lm. The advantages of the sputter coatingsare the hardness and the good adherence of the filmsproduced while the main disadvantages of plasma CVDcoatings are their reduced density and weaker adherenceto the substrate material, although they are of great interestfor industrial use where often the easy handling is crucial.The DLC sputter coating used in this work was a multilayercoating with a TiN sublayer, followed by a TiC filmand finally a DLC coating. The DLC–Si–O surface wasproduced by Plasma CVD using C x H x as precursor forthe DLC film and Hexamethyldisiloxan (HMDSO) forthe addition of silicon and oxygen. In addition, a plasmaCVD SiO x film was produced using HMDSO as precursor.2.1.2.3. Autocatalytic Ni–P–PTFE coating. The Ni–P–PTFE coating was produced by an autocatalytic platingprocess (Nasser, 2000). In this process the stainless steelsurface went through five steps: (1) an alkaline cleaningbath; (2) a pickling process; (3) activation by the galvanicdeposition of nickel and (4) an autocatalytic reaction inwhich a Ni–P plating is deposited. After the Ni–P platinghad reached a certain thickness, PTFE particles were addedto the coating bath, to be incorporated in the Ni–P matrix.In order to be effective in reducing fouling, the outer surfaceof the Ni–P–PTFE plating should contain a high percentageof homogenously distributed PTFE particles. Upto now, the maximum achievable amount of PTFE in theouter surface is 20%.2.1.2.4. Silica coating. Silica coated (SiO x ) surfaces wereprepared by a sol–gel process, where metal alkoxide compoundsare typically used as raw ingredients. The processstarted with the hydrolysis of such compounds in the presenceof water and condensation to form M–O–M bonds(‘‘M’’ being elements like Si, Ti, Zr...). After the hydrolysisand condensation have proceeded to a certain degree (solformation), a coating could be applied on the substratesby, for example, dip- or spin-coating; dip-coating, underwell controlled conditions, was used in the present study.The coating structure was then consolidated by high temperatureannealing in order to oxidize and remove anyresidual organic molecules. Sols were prepared frommethyl-triethoxysilane (MTES): H 2 O (molar ratio 1:3) solutionsin ethanol at various dilutions (1:10–1:4), andannealed at temperatures in the range of 200–500 °C. Underthese conditions, crack-free coatings could be obtained.

R. Rosmaninho et al. / Journal of Food Engineering 80 (2007) 1176–1187 1179In addition to the previous mentioned surfaces, a commerciallyavailable coated surface (Xylan) was also usedfor bacterial adhesion tests.2.1.3. Calcium phosphate solutionCalcium phosphate deposition was performed usingSimulated Milk Ultra Filtrate (SMUF), which is an aqueoussolution that simulates the mineral composition ofmilk. It was prepared with deionised water and appropriateamounts of the reagents as suggested by Jenness andKoops (1962). The resulting SMUF solution had a pHvalue between 6.1 and 6.3 and was adjusted to the desiredpH value by addition of KOH, NaOH or HCl. The finalcomposition of SMUF is presented in Table 1.2.1.4. b-Lactoglobulin solutionFor batch deposition, a pure b-lactoglobulin (b-Lg)solution (MDFbeta-1) provided by MD Foods (MDF,Decon Laboratories, Sussex, UK) containing 70 ± 1.1%(w/w) of b-Lg was used. The solution preparation methodand its gross chemical composition were described by Holtet al. (1999). The protein was dissolved in phosphate-bufferedsaline (PBS) (pH 6.0; 0.01 M phosphate; 0.15 MNaCl, ionic strength 0.17 M) in order to achieve a concentrationof 1 g/l. The water used was distilled and passedthrough a Milli-Q water purification system (MilliporeCorp., Bedford, MA, USA).2.1.5. Bacterial suspensionTwo Bacillus strains isolated from dairy processing lineswere used throughout this study, B. cereus CUETM 98/4and B. subtilis CUETM 98/7. The B. cereus CUETM 98/4 spores were found to be highly hydrophobic, while theB. subtilis CUETM 98/7 spores were hydrophilic asdescribed recently by Faille and Bénézech (2002). Sporeswere produced as previously described (Faille et al., 1997)and suspended in hot milk (10 7 spores/ml). The results wereevaluated by analysis of variance (General Linear Model)followed by the Student–Newman–Keuls mean comparisontest (SAS 8.01, Garry, USA).2.1.6. FMFFouling model fluid (FMF) was designed to mimic adairy dessert cream composed of whey proteins (2% w/w;Table 1Final composition of SMUF solutionsReagentsConcentration in solution (mM)KH 2 PO 4 11.61K 3 citrateÆH 2 O a 3.70Na 3 citrateÆ2H 2 O a 6.09K 2 SO 4 1.03K 2 CO 3 2.17KCl 8.05CaCl 2 Æ2H 2 O 8.98MgCl 2 Æ6H 2 O 3.21a Citrate ¼ C 6 H 8 O 3 .Armor Proteins, St Brice en Cogles, France), whole milkpowder (11% w/w; Isigny-sainte-mère, France), sucrose(8%; Beghin Say, France) and Xanthan gum (0.2% w/w;Degussa, Texturant Systems). The xanthan gum, a dairyproduct thickener, was chosen to modify and to adjustthe viscosity of the fluid. The apparent viscosity was measuredon samples taken before and after the heater, using aviscosimeter (Rheomat 30, Contraves). The shear thinningbehaviour of the fluid could be described by the power lawmodel (k = 0.45 Pa s n ; n = 0.39, for a temperature of 40°).Because of its similarity to milk composition, milk thermophysicalproperties (thermal conductivity and heat capacity)were used in the calculations.2.2. Stainless steel cleaning proceduresAll stainless steel samples were cleaned with a commercialalkaline detergent RBS35 (NFT 72151-72190) beforeeach experiment. The detergent was a mixture of non-ionicand anionic surfactants which were easily removed bywater and was obtained from RBS Chemical Products,Brussels, Belgium. The cleaning procedure used was as follows:(1) samples were immersed in a 2.0% w/v detergentsolution in distilled water at 65° for 5 min; (2) rinsed withdistilled water at 65° for 5 min and (3) rinsed with distilledwater at 20 °C.For the bacterial adhesion experiments, part of theprocedure was slightly changed in order to include a disinfectionphase: (1) samples were immersed in a 2.0% w/v detergent solution in distilled water at 65° for 5 min;(2) rinsed with distilled water at 65° for 5 min; (3) disinfectedby immersion in Oxygal (Europro, France) 1% insoftened water at room temperature and finally (4) rinsedwith water. After each biofouling experiment, a standardCIP procedure was followed: (1) rinsing for 5 min withsoftened water at room temperature; (2) 10 min inNaOH 0.5% at 60 °C and (3) 5 min rinsing with softenedwater at room temperature. Temperature and flow-ratewere kept under control and the mean velocity of thefluid was 7.15 · 10 1 m/s (Re = 3950) corresponding toturbulent flow regime. Four repetitions were carried outfor each test. Such a mild procedure was chosen in orderto better evaluate any differences in the ease of bacteriaremoval since the quantification of the hygienic statuswas based on the amount of spores remaining on thesurfaces after a complete cycle of fouling andcleaning.In the pilot plant experiments using Plate HeatExchangers (PHE), a simple CIP procedure was carriedout, using NaOH 0.5% at 60 °C for 15 min prior and afterthe rinsing steps. As the fouling layer was mainly composedof proteins, only alkaline cleaning was used. The detergentmean velocity was 0.427 m/s (Re = 3460), resulting in amean residence time of the NaOH solution of around 4 s,while the rinsing mean velocity was 0.712 m/s(Re = 4580). Residual fouling could be observed on theplates after cleaning due to the mild conditions chosen.

1180 R. Rosmaninho et al. / Journal of Food Engineering 80 (2007) 1176–11872.3. Laboratory tests with calcium phosphate deposition2.3.1. Batch deposition system2.3.1.1. Rotating disk apparatus. The fouling experiments ofcalcium phosphate under batch conditions were performedin a rotating disk apparatus (Rosmaninho & Melo, 2006)which allowed the study of the deposition process on eachsurface under controlled hydrodynamics and temperature.The deposition surface was attached to a heated coneand immersed in the deposition solution which was containedin a cylindrical temperature controlled vessel. Thesurface/cone set was allowed to rotate in the solution, theReynolds number of the system being controlled by itsrotating speed. Similar devices using the same concept havebeen recently used by other authors on the assessment ofboth fouling and cleaning processes for milk componentswith good comparison with published results on foulingand cleaning (Morison & Thorpe, 2002).2.3.1.2. Experimental procedure. The calcium phosphatebatch deposition experiments included two steps: (1) thedeposition of calcium phosphate from SMUF and (2) theremoval/detachment of the fouling deposits under definedhydrodynamic conditions in order to determine their consistencyby measuring the amount of deposit massremoved. Each surface was tested for two or three timesto both deposition and removal.For the deposition step, SMUF was heated to thedesired temperature (44 °C and 70 °C) and the oil in thecone was heated to a temperature which avoided significanttemperature differences between the bulk solution and thesurface of the deposition sample. The temperature of theoil was chosen in order to keep the surface’s temperatureat approximately 1 a C above the temperature of the SMUFand never higher. The cone with the stainless steel samplewas then immersed in the solution and started to rotateat a speed of 150 rpm (Re = 46000). The depositionoccurred during 120 min, keeping a constant rotationspeed. After that period, the sample was removed fromsolution and air dried. The amount of deposit formedwas evaluated by weighing the sample before and afterthe deposition procedure using an analytical balance(AND GR-200).The detachment experiments were carried out using thedeposits formed in the previous step. The process startedby replacing the SMUF solution by distilled water (pH6.8 and 44°) where the cone with the attached fouled samplewas immersed. The set-up was then started to rotate atincreasing speeds of 165, 180, 210, 240, 270 and 300 rpmfor periods of 5 min at each rotating speed. Afterwards,the sample was removed, dried and weighed in order toassess the total amount of deposit detached from thesurface.2.3.2. Continuous flow depositionCalcium phosphate deposition in a cylindrical pipetest section was studied at temperatures of 60 °C and70° in a once-through flow of supersaturated SMUFsolutions. The continuous flow deposition experimentswere carried out at isothermal conditions without anywall heating. Initially, deionised water was heated tothe desired temperature by passing through an electricheater. The SMUF solutions were then prepared by addingto the flowing water stream the appropriate amountsof the reagents.Details concerning the experimental set-up, the tubulartest section and the special specimen (semi-annular coupons)on which deposits developed can be found elsewhere(Andritsos & Karabelas, 1991; Andritsos et al., 2002).Unmodified stainless steel and Ni–P–PTFE coupons wereused. Evaluation of the modified surfaces was carried outin another test section placed at the downstream end, havinga 10 · 10 mm 2 free cross-sectional area. Modified andunmodified surfaces of 10 · 40 mm 2 size were glued on specialtrapezoidal-shaped stainless steel coupons and insertedinto the test section forming a channel of square crosssection.In the pH range of the experiments (between 5.8 and 7.0)the solution was supersaturated with respect to almost allpossible calcium phosphate crystalline phases as deducedby calculating the saturation ratio for each phase.2.4. Laboratory tests with protein deposition2.4.1. Batch deposition system2.4.1.1. Ellipsometer with cuvette. The adsorption of a pureb-Lg solution, under batch conditions, on the differentstainless steel surfaces was followed in a cuvette by in situnull ellipsometry (Santos et al., 2004) at25°C. With thistechnique the adsorption and desorption of proteins onsolid surfaces can be continuously monitored. A modifiedautomated Rudolph thin film ellipsometer type 43 603-200E (Rudolph Research, Fairfield, NJ, USA) was used,with a polarized He–Ne laser beam of wavelength 632.8nm incident at an angle of 68° in relation to a plane normalto the surface. A detailed description of the apparatus isgiven in Nylander (1987). A 7 ml trapezoid cuvette madeof fused quartz (Hellma, Germany) was used, whichallowed the light to pass perpendicularly to the cuvettewalls.2.4.1.2. Experimental procedure. The cuvette was filled with4.5 ml of buffer solution. The dry stainless steel sample wasfixed in the sample holder and lowered to approximately1 mm above the magnetic stirrer (325 rpm) present at thebottom of the cuvette. When the ellipsometric parametersfrom the surface were stable, 0.5 ml of b-Lg stock solutionwere injected into the cuvette resulting in a final proteinconcentration of 0.1 g/l. After recording the adsorptionfor 15 min, the b-Lg solution was replaced by PBS at a flowrate of 20 ml/min for 5 min. The measurements were continuedfor another 10 min under stirring. The mass of proteinadhered to the sample per surface area was calculatedaccording to Cuypers et al. (1983).

R. Rosmaninho et al. / Journal of Food Engineering 80 (2007) 1176–1187 11812.5. Laboratory tests with bacterial adhesion2.5.1. Experimental procedureIn order to obtain stainless steel surfaces covered by aconditioning film similar to the one found in industrialenvironments, all coupons were subjected to 12 cycles ofmilk fouling and cleaning. The fouling/cleaning cycle procedurewas previously described by Jullien (2001) and comprisedthe following steps: (1) 1 h fouling in reconstitutedwhole milk (UCL, Isigny Sainte Mère, France) at roomtemperature (fluid velocity 2.23 m/s, Re 1.32 · 10 4 ), (2)5 min rinse with softened water at room temperature (fluidvelocity 0.669 m/s, Re = 3950), (3) cleaning in NaOH 0.5%at 60 °C during 15 min (fluid velocity 1.34 m/s, Re = 7 900)and finally (4) 5 min rinse with softened water at room temperature(fluid velocity 0.669 m/s, Re = 3 950).The coupons were inserted in the test pipe (Fig. 1). Fivecoupons were placed into each pipe which had a hydraulicdiameter of 1.23 · 10 2 m. Fouling experiments underdynamic conditions were performed using the Bacillusspore suspension (10 7 UFC/ml) in hot whole milk(60 °C). With these experimental conditions, the intentionwas to simulate the adhesion process, after heat treatmentin the pipe equipment, inducing a large number of adheringspores.The test pipes were part of an experimental loop whichwas composed of a Millipore peristaltic pump (flow rateranging from 0 to 50 l/h), a surge tank to limit pulse flowrate variations, a vessel containing the fouling suspensionwhich was heated by a hot water bath and 5 test pipes (rectangularsection 11 · 14 mm). Fouling experiments ran for1 h, at a mean flow velocity of 0.107 m/s (Re = 632) whichcorresponded to the laminar regime. During the foulingprocedure, no significant spore germination could beobserved. After the fouling experiments, the loop was dismantledand the number of spores on each coupon wasdetermined in order to quantify the amount of adheringcells. Four repetitions were carried out for each conditiontested.2.6. Pilot-plant tests with fouling model fluid2.6.1. Experimental procedureFouling experiments with FMF were carried out in apilot plant consisting of: two stirred tanks, one for waterand the other for the fluid solution; two test plate heat10 mm15 mm17 mmCouponFlow225 mmFig. 1. Test pipes containing the different stainless steel coupons (leftfigure: cross section of pipe inlet).exchangers (PHE1 and PHE2) connected to two heatingunits and another heat exchanger working as cooler(PHE3). The pre-heating (PHE1) and the heating (PHE2)sections were composed of 10 and 15 plates (V2 type) withan effective heat exchange area of 0.160 m 2 and 0.260 m 2 ,respectively. Flow arrangements in the heating part werechosen in order to avoid flow maldistribution. The flowarrangement in the pre-heater and in the heater was countercurrent mono-channel per pass. To cover a wide temperaturerange seven channels were used. The pump usedfor the milk circuit was a hygienic volumetric pump (MoineauH series, PCM, France), Q max = 1.52 · 10 4 m 3 /s,P max = 9 bar. For the hot water circuit a centrifugal pumpwas used (Sterling Fluid systems SIHI, France, CEHA1203, Q max =3· 10 4 m 3 /s, P max = 5 bar). All heatexchangers were made by Alfa Laval Vicarb. The plateswere 0.256 m in length and 0.078 m in width with a hydraulicdiameter of 0.005 m (Beuf et al., 2003).The FMF had a pH of 7 and was stored in a stirred tankat 15 °C, to prevent bacteria proliferation prior to theexperiments. Subsequently, the fluid was heated from 15to 60 °C with circulating hot water at 63 °C inPHE1andthen heated to 102 °C in PHE2. The FMFs mean velocitywas fixed at a constant value of 0.142 m/s (Re = 120, i.e.transitional flow regime), which would result in a mean residencetime in PHE2 of 12.5 s. The hot water mean velocitywas kept at 0.712 m/s (Re = 7470, i.e. turbulent flowregime). FMF was discarded after the cooling process inPHE3. To allow the comparison between the different surfaces,three channels (channels 4, 5 and 6) of PHE2 weremade up of coated stainless steel pates, while the othersconsisted of the original plates. However in the case ofion implanted SiF þ 3and sputtered DLC surfaces only onechannel was modified.After each fouling cycle, PHE2 was dismantled and allplates were dried and weighed in order to determine thedeposit mass formed on each plate. After each cleaningcycle the same procedure was applied. The experimentswere performed in triplicate.3. Results and discussions3.1. Laboratory tests with calcium phosphate deposition3.1.1. Batch deposition systemDeposition of calcium phosphate was performed at twodifferent temperatures: 44 °C, which corresponds to thebeginning of the formation of calcium phosphate particlesin the bulk as detected by the increase in turbidity of thesolution, and 70 °C, to obtain the fouling behaviour at temperaturessimilar to the pasteurisation ones. The two chosentemperatures also allow the evaluation of newlyformed particle aggregates (44 °C) and mature calciumphosphate aggregates (70 °C). As shown in Fig. 2, theamount of deposit formed on each surface for both temperaturesdepended on the surface characteristics and wasalways higher for higher deposition temperatures. For both

1182 R. Rosmaninho et al. / Journal of Food Engineering 80 (2007) 1176–11872 )formed(g /mdeposit9.08.07.06.05.04.03.02.01.00.070ºC44ºC2R Ni SiOx SiF silica DLCSiO MoS2 DLCFig. 2. Amount of calcium phosphate deposit formed on the differentsurfaces at 44 and 70 °C.temperatures, more deposit was found on SiO x and ionimplanted (SiF þ 3and MoS 2 ) surfaces, while less depositwas observed on Ni–P–PTFE surfaces.The deposits were very similar for all the surfaces andconsisted of submicron spheroids of approximately0.1 lm for lower temperatures and of approximately0.4 lm for higher temperatures, forming tightly packedclusters.Removal/detachment experiments were performed oneach deposit under its temperature of formation. The totalpercentage of deposit mass removed from each surface,under the same hydrodynamic and thermal conditions, ispresented in Fig. 3. Different percentages of mass removalwere found from surface to surface, being higher for thesurfaces where the deposited mass was smaller, as for Ni–P–PTFE surfaces. For each surface, a dependence of massremoval with temperature was found although no systematicmathematical relationship could be concluded.3.1.2. Continuous flow deposition systemUnder continuous flow conditions, the deposited masstended to increase with time, usually with a macroscopicallyobserved induction period in the range of 5–20 min.The curve of deposited mass versus pH exhibited a ‘‘bell’’shape for both temperatures, similar to that observed inthe deposition of other sparingly soluble species (PbS,CdS, FeS x ) in the same type of experiments (e.g. Andritsos& Karabelas, 1991). At 60 °C, deposits started to beformed at pH approximately 6.0, the deposition rate thenincreased sharply with a slight increase in pH. Maximumdeposition rate occurred at pH of approximately 6.3, asshown in Fig. 4.Above this pH the deposited mass decreased, since bulkprecipitation (as evidenced from the increased solution turbidityin Fig. 4) diminished supersaturation and tended toreduce the driving force for wall precipitation. It is likelythat part of the deposits for pH P 6.3 is due to particulatedeposition. At 70 °C, the maximum deposition rate wasclearly higher, it occurred at a slightly lower pH, but theoverall deposition pattern was similar.At the maximum deposition rate, the following stepswere observed in the deposit development: initially, a thinand well adhering layer was formed rather rapidly (i.e. inless than 5 min) covering almost entirely the substrate. Furthergrowth occurred mostly in the form of ‘‘overgrowths’’(Fig. 5), developing rather randomly on the covered substrate.More overgrowths appeared at the lower part ofthe pipe inner surface, and their concentration seemed todecrease towards the upper part, suggesting a slight influenceof gravity.The overgrowths consisted of submicron spheroids ofsizes less than 0.1 lm, forming tightly packed clusters.These deposits could not be washed off from the stainlesssteel coupons by simple rinsing with water, but they couldD eposited mass (mg/dm2 )1801601401201008060402005.6 5.8 6 6.2 6.4 6.6 6.8 7 7.2 7.4pH0.61009080706050403020100% removal2R70ºC44ºCNi SiOx SiF silica DLCSiO MoS2 DLCFig. 3. Percentage of calcium phosphate removed for each surface duringthe removal experiments.absorbanceLight0.50.40.30.20.10.05.6 5.8 6 6.2 6.4 6.6 6.8 7 7.2 7.4pHFig. 4. (a) Deposited mass for a 60 min run time versus pH at T =60°Cand flow velocity of 0.32 m/s (Re = 8800). (b) Light absorbance of theSMUF solution as a function of pH at T =60°C.

R. Rosmaninho et al. / Journal of Food Engineering 80 (2007) 1176–1187 1183Fig. 5. (a–c) SEM micrographs of the mostly amorphous deposits formed from SMUF at pH = 6.3 and 60 °C (flow velocity 0.32 m/s) at various run timeson 2R surfaces; (d) shows the deposits formed on a DLC substrate.be removed rather easily by mechanical means after beingdried, leaving the adhering initial layer on the substrate.The deposited mass on the Ni–P–PTFE coupons did notdiffer appreciably from that on 2R stainless steel coupons,but the initial layer did not adhere as strongly as on theuntreated metal surfaces. At 60 °C, an apparent lack ofcrystalline order was evident from XRD analyses, althougha broad peak was observed at approx. 32°, characteristic ofa poorly crystallized HAP phase. This peak was intensifiedat 70 °C.Experimental results on the influence of the substrate(typical results are shown in Fig. 6) indicated that the substratematerial did not cause any discernible difference withregard to deposited mass and deposit morphology, suggestingthat the interfacial properties did not control the scaleformation process. There were some minor differences inthe deposited mass, especially on the top coupon, but theyoccurred regardless of the coupon material. This behaviourmay be explained by considering the formation of an initialcompact and adhered layer due to wall crystallization, afterwhich the nature of the substrate was unlikely to affect thesubsequent deposit development since the deposition mechanismmay also include particulate fouling from bulk precipitation,rather than pure surface crystallisation The onlydifference that could be observed among the various couponswas that the probability to have ‘‘cracks’’ (as observedunder the microscope after drying) in the initial coherentlayer was greater in most of the modified/treated surfacesthan in unmodified 2R ones. The above observations seemto be in agreement with those of Britten, Green, Boulet,and Paquin (1988) who found that non-metallic surfacesdid not affect the amount of deposits upon milk heating at100° but affected the strength of adhesion of those deposits.2 )(mg/dmDeposit2 )(mg/dmDeposit250200150100500806040200sidebottomsideIon-imlp 2Rsilica DLCsidebottomsideBased on the facts previously presented, it was possibleto conclude that the surface properties played differentroles on the deposition process of calcium phosphate whenconsidering batch or continuous flow deposition. Regardingthe deposition process, different amounts of calciumphosphate deposits were formed on the modified surfacesunder batch conditions (more calcium phosphate depositwas formed on SiO x , SiF þ 3and MoS 2 surfaces while lesstoptopsilica 2R Ion-impl. Ni-P-PTFEFig. 6. Mass of deposits formed from SMUF on modified and unmodifiedsurfaces: (a) T =60°C, flow velocity 0.32 m/s, pH = 6.4; (b) T =70°C,flow velocity 0.32 m/s, pH = 6.0.

1184 R. Rosmaninho et al. / Journal of Food Engineering 80 (2007) 1176–1187deposit was formed on Ni–P–PTFE, silica and DLC–Si–Osurfaces), while in the continuous flow experiments no significantdifference in the amount of deposit was observed.Probably, this was due to the fact that the degree of supersaturationwas kept constant in the continuous flow experimentswhile it decreased during the batch tests. Therefore,a much higher amount of calcium phosphate deposit wasobtained in the continuous flow test, which could havemasked the effect of the original surface.Regarding the detachment properties, both in the batchand continuous flow systems the deposits formed on Ni–P–PTFE surfaces were more easily removed (cleaned). Therefore,the effect of surface properties was noticeable whenstudying thinner layers (as in the batch tests) and finalcleaning efficiencies.3.2. Protein deposition3.2.1. Batch deposition systemAdsorption of b-Lg in the cuvette was followed at 25 °Cand the adsorbed and desorbed amount determined foreach surface (Fig. 7).The amount adsorbed was not significantly differentamong the surfaces, whereas protein desorbability wasdependent on the surface. As in the case of the mineralsolution (SMUF), a higher amount of protein depositcould be removed from the Ni–P–PTFE surface than fromthe others, indicating that protein desorbability was higherfor Ni–P–PTFE.3.3. Laboratory tests with bacterial adhesionSpore adhesion was found to be slightly influenced by thesurface materials (Fig. 8) with the number of adheringspores ranging from 8.59 · 10 3 to 1.43 · 10 4 UFC/cm 2 forB. cereus (p-value = 0.036) and from 6.34 · 10 2 to1.11 · 10 3 UFC/cm 2 for B. subtilis (p-value = 0.039). Thesedifferences do not depend on the position of the samplessince it was observed that the place of the sample insidethe test rig had no effect on the number of residual B. cereusorB. subtilis spores (p-value of 0.29 and 0.74 respectively).Fig. 8. Initial adhering Bacillus cereusin hot milk.or Bacillus subtilis (h) sporesAccording to the Student–Newman–Keuls test (Alphalevel = 0.05) the five materials could be grouped into twogroups depending on their hygienic status. Only materialsexhibiting non-common letters could be considered to besignificantly different in terms of hygiene (Table 2).The mean comparison test was unable to discriminatethe initial B. cereus and B. subtilisadhesion level betweenthe control stainless steel and the other modified materials.Only Xylan and the TiC ion implanted surfaces were significantlydifferent with respect to the level of contaminationfor B. cereus, as indicated by the letters A and B. The formershowed a higher initial adhesion and the latter a lowerinitial adhesion of B. cereus spores. No significant differenceswere found in the case of B. subtilis.The number of residual adhering spores found aftercleaning was between 2.15 · 10 1 and 5.43 · 10 3 UFC/cm 2for B. cereus and between 2.96 and 1.78 · 10 1 UFC/cm 2for B. subtilis (Fig. 9).Table 2Statistical grouping of the surfaces according to their hygienic status(Student–Newmann–Keuls test, Alpha level = 0.05)Materials Statistical groupingB. cereus adherent spores B. subtilis adherent spores2R (control) BA AMoS 2 BA ATiC B ANi–P–PTFE BA AXylan Ò A A2 )(mg/m2.52Before rinseAfter rinseAdsorbed amount1.510.50316 2RSiF SiOx DLC Ni-P-PTFEFig. 7. Adsorbed amount of pure b-Lg solution, after 10 min, on thesurfaces tested before and after rinse with PBS. b-Lg was adsorbed from a0.1 g/l solution in PBS at pH 6 and at 25 °C.Fig. 9. Residual adhering spores of Bacillus cereussubtilis (h) after cleaning in place (CIP) procedure.and Bacillus

R. Rosmaninho et al. / Journal of Food Engineering 80 (2007) 1176–1187 1185The analysis of variance indicated that 98% of the variabilityin the number of residual B. cereus spores could beexplained by the different surface properties (p-value

1186 R. Rosmaninho et al. / Journal of Food Engineering 80 (2007) 1176–1187(2) From a food contamination point of view, based onthe data from the adhesion of B. cereus spores, themost suitable surface appeared to be the ionimplanted (TiC) surface, which also showed lessspores after the cleaning process. Looking only atthe number of adhered spores, the Ni–P–PTFE andthe Xylan surfaces did not seem to be as good.However, the relevance of the biological versus non-biologicalfouling behaviour should be assessed. In the heatingprocess (pasteurization), the main fraction of the deposit isthe non-biological one, since the thermal conditions are farfrom optimal for the survival of bacteria (note that this isnot applicable to the cooling section of the heat exchangers).Therefore, it is proposed that the choice of the idealheating surface in milk pasteurisation should take intoaccount primarily the non-biological data.In such a case, the Ni–P–PTFE coating seems to be themost appropriate surface treatment to minimize milk foulingeffects. This conclusion is supported by the fact that thecharacteristics provided by the Ni–P–PTFE surface modificationmethod are systematically different from the othermethods, as shown by the authors in a previous paper(Santos et al., 2004): (i) Ni–P–PTFE was the most hydrophobicsurface of them all, the others being essentiallyhydrophilic; (ii) the apolar component (London–van derWaals) of the surface energy of Ni–P–PTFE had the lowestvalue among all surfaces, the others showing higher butsimilar values among them; (iii) the electron donor component(polar) of the surface energy had undoubtedly thelowest value for the Ni–P–PTFE material (0.1 m N/magainst 48.2 m N/m for the unmodified 2R surface). In thatprevious work, it was also noted that the Ni–P–PTFE coatingwas the only modification method that clearly changedthe topography of the unmodified steel surfaces, by maskingthe grain boundaries at the surface, which is due to thefact that this type of coating creates the thickest layer onthe stainless steel surface (10 lm, against 0.2–2.5 lm forall the other surfaces). Even so, the thickness of the Ni–P–PTFE layer is small enough not to have a significanteffect on the thermal resistance of the heat transfer wall.AcknowledgementsThis paper is dedicated to the memory of Dr. Hans Visser,the initiator and co-ordinator of this work within theMODSTEEL project.Financial support was obtained from the EuropeanCommunity under the ‘‘Competitive and SustainableGrowth’’ Programme (MODSTEEL, Contract No.G5RD-CT-1999-00066, Project No. GRD1-1999-10856)and from SIMUMILK (project no. POCTI/EQU/47654/2002, funded by FCT, Portugal, and by European structuralfunds). The funding of R. Rosmaninho’s PhD grantby the Portuguese Science Foundation (FCT) is gratefullyacknowledged.ReferencesAddesso, A., & Lund, D. B. (1997). Influence of solid surface energy onprotein adsorption. Journal of Food Processing and Preservation, 21,319–333.Andritsos, N., & Karabelas, A. J. (1991). Crystallization and depositformation of lead sulfide from aqueous solutions I. Deposition rates.Journal of Colloids and Interface Science, 145, 158–169.Andritsos, N., Yiantsios, S. G., & Karabelas, A. J. (2002). Calciumphosphate scale formation from simulated milk ultrafiltrate solutions.Food and Bioproducts Processing, 80(C4), 223–230.Beuf, M., Rizzo, G., Leuliet, J. C., Müller-Steinhagen, H., Yiantsios, S., &Karabelas, A. (2003). Fouling and cleaning of modified stainless steelplate heat exchangers processing milk products. Heat exchangercleaning and fouling conference: fundamentals and applications, SantaFe, New Mexico, USA. Available from http://services.bepress.com/eci/heatexchanger/.Bott, T. R. (1995). Fouling of heat exchangers. Amsterdam: ElsevierScience.Boulangé-Petermann, L. (1996). Processes of bioadhesion on stainless steelsurfaces and cleanability: a review with special reference to the foodindustry. Biofouling, 10(4), 275–300.Britten, M., Green, M. L., Boulet, M., & Paquin, P. (1988). Depositformation on heated surfaces: effect of interface energetics. Journal ofDairy Research, 55, 551–562.Changani, S. D., Belmar-Beiny, M. T., & Fryer, P. J. (1997). Engineeringand chemical factors associated with fouling and cleaning in milkprocessing. Experimental Thermal and Fluid Science, 14(4), 392–406.Cuypers, P. A., Corsel, J. W. W., Jansen, M. P., Kop, J. M. M., Hermens,W. Th., & Hemker, H. C. (1983). The adsorption of prothrombin tophosphstidylserine multilayers quantitated by ellipsometry. The Journalof Biological Chemistry, 258(4), 2426–2431.Faille, C., & Bénézech, T. (2002). Adhesion of Bacillus spores andEscherichia coli cells to inert surfaces: role of surface hydrophobicity.Canadian Journal of Microbiology, 48(8), 728–738.Faille, C., Fontaine, F., & Bénézech, T. (2001). Potential occurrence ofadhering living Bacillus spores in milk product processing lines.Journal of Applied Microbiology, 90, 892–900.Faille, C., Lebret, V., Gavini, F., & Maingonnat, F. J. (1997). Injury andlethality of heat treatment of Bacillus cereus spores suspended in bufferand in poultry meat. Journal of Food Protection, 60, 544–547.Forster, M., & Bohnet, M. (1999). Influence of the interfacial free energycrystal/heat transfer surface on the induction period during fouling.International Journal of Thermal Sciences, 38(11), 944–954.Haynes, A., & Norde, W. (1994). Globular protein at solid/liquidinterfaces. Colloids and Surfaces B: Biointerfaces, 2(6), 517–566.Holt, C., McPhail, D., Nevison, I., Nylander, T., Otte, J., Ipsen, R. H.,et al. (1999). Apparent chemical composition of nine commercial orsemi-commercial whey protein concentrates, isolates and fractions.International Journal of Food Science and Technology, 34(5–6),543–556.Janocha, A., Hegemann, D., Oehr, C., Brunner, H., Rupp, F., & Geis-Gerstorfer, J. (2001). Adsorption of protein on plasma-polysiloxanelayers of different surface energies. Surface and Coatings Technology,142–144, 1051–1055.Jenness, R., & Koops, J. (1962). Preparation and properties of a saltsolution which simulates milk ultrafiltrate. The Netherlands Milk andDairy Journal, 16(3), 153–164.Jullien, C. (2001). Caractère hygiénique des matériaux utilisés dans les agroindustries-Contribution des caractères de surface des micro-organismes(spores de Bacillus) et des matériaux. Ph.D. thesis Université deCompiègne.Krause, S. (1993). Fouling of heat-transfer surfaces by crystallization andsedimentation. International Chemical Engineering, 33(3), 355–401.Liu, Y., Wu, W., Sethuraman, G., & Nancollas, G. H. (1997). Intergrowthof calcium phosphates: an interfacial energy approach. Journal ofCrystal Growth, 174(1–4), 386–392.

R. Rosmaninho et al. / Journal of Food Engineering 80 (2007) 1176–1187 1187Luby, S., Jones, J., Dowda, H., Kramer, J., & Horan, J. (1993). A largeoutbreak of gastro-enteritis caused by diarrheal toxin producingBacillus cereus. Journal of Infectious Diseases, 167, 1452–1455.Melo, L. F., & Bott, T. R. (1997). Biofouling in water systems.Experimental Thermal and Fluid Science, 14(4), 375–381.Morison, K. R., & Thorpe, R. J. (2002). Spinning disc cleaning ofskimmed milk and whey protein deposits. Food and BioproductsProcessing, 80(C4), 319–325.Morison, K. R., & Tie, S. H. (2002). The development and investigation ofa model milk mineral fouling solution. Food and Bioproducts Processing,80(C4), 326–331.Müller-Steinhagen, H. (1998). Mitigation of process heat exchangerfouling: An integral approach. Chemical Engineering Research &Design, 76(A2), 97–107.Nasser, K. (2000). Galvanotechnik. Grundlagen, Verfahren, Praxis,München: Carl Hauser.Nylander, T. (1987). Proteins at the metal/water interface- adsorption andsolution behaviour. Ph.D. Dissertation, Department of Food Technology,Lund University, Lund, Sweden.Rosmaninho, R., & Melo, L. F. (2006). Calcium phosphate depositionfrom simulated milk ultrafiltrate on different stainless steel-basedsurfaces. International Dairy Journal, 16, 81–87.Santos, O., Nylander, T., Rosmaninho, R., Rizzo, G., Yiantsios, S.,Andritsos, N., et al. (2004). Modified stainless steel surfaces targetedto reduce fouling – surface characterization. Journal of Food Engineering,64(1), 63–79.Tsibouklis, J., Stone, M., Thorpe, A. A., Graham, P., Nevell, T. G., &Ewen, R. J. (2000). Inhibiting bacterial adhesion onto surfaces: thenon-stick coating approach. International Journal of Adhesion andAdhesives, 20(2), 91–96.Visser, J., & Jeurnink, T. J. M. (1997). Fouling of heat exchangers in thedairy industry. Experimental Thermal and Fluid Science, 14(4),407–424.Wu, W. J., Zhuang, H. Z., & Nancollas, G. H. (1997). Heterogeneousnucleation of calcium phosphates on solid surfaces in aqueoussolution. Journal of Biomedical Materials Research, 35(1), 93–99.Yoon, J., & Lund, D. B. (1994). Magnetic treatment of milk and surfacetreatment of plate heat exchangers: effects on milk fouling. Journal ofFood Science, 59(5), 964–980.Zhao, Q. (2004). Effect of surface free energy of graded NI-P-PTFEcoatings on bacterial adhesion. Surface and Coatings Technology,185(2–3), 199–204.Zhao, Q., & Müller-Steinhagen, M. (2001). Intermolecular and adhesionforces of deposits on modified heat transfer surfaces. In H. Muller-Steinhagen, M. Reza Malayeri, & A. P. Watkinson (Eds.), Proceedingsinternational conference on heat exchanger fouling – Fundamentalapplications and technical solutions (pp. 41–46). Germany: PublicoPublications.