Elisabeth
Wilhelm
a,
Kaustubh
Deshpande
a,
Frederik
Kotz
a,
Dieter
Schild
b,
Nico
Keller
a,
Stefan
Heissler
c,
Kai
Sachsenheimer
a,
Kerstin
Länge
a,
Christiane
Neumann
a and
Bastian. E.
Rapp
*a
aKarlsruhe Institute of Technology (KIT), Institute of Microstructure Technology (IMT), Hermann-von-Helmholtz-Platz 1, 76344 Eggenstein-Leopoldshafen, Germany. E-mail: bastian.rapp@kit.edu; Fax: +49 (0)721 608 26667; Tel: +49 (0)721 608 28981
bKarlsruhe Institute of Technology (KIT), Institute for Nuclear Waste Disposal (INE), Hermann-von-Helmholtz-Platz 1, 76344 Eggenstein-Leopoldshafen, Germany
cKarlsruhe Institute of Technology (KIT), Institute of Functional Interfaces (IFG), Hermann-von-Helmholtz-Platz 1, 76344 Eggenstein-Leopoldshafen, Germany
First published on 9th February 2015
In this article we introduce and compare three techniques for low-cost and rapid bonding of stereolithographically structured epoxy components to polydimethylsiloxane (PDMS). In short, we first create a polysiloxane layer on the epoxy surface via silane surface coupling and polymerization. Afterwards, the modified epoxy surface can be bonded to a PDMS component at room temperature using a handheld corona discharger, which is a commonly used low-cost technique for bonding two PDMS components. Using these methods bonds of desirable strength can be generated within half an hour. Depending on the epoxy resin, we found it necessary to modify the silanization procedure. Therefore, we provide a total of three different silanization techniques that allow bonding of a wide variety of stereolithographically structurable epoxy resins. The first technique is a UV-light induced silanization process which couples a silane that contains an epoxy-ring ((3-glycidoxypropyl)trimethoxysilane (GPTMS)). For surfaces that cannot be modified with this silane we use dimethoxydimethylsilane (DMDMS). This silane can either be coupled to the surface by a sol–gel process or UV-light induced polymerisation. The sol–gel process which is a heat induced surface modification technique results in high bond strengths. Because of the heat which triggers the sol–gel process, this technique is limited to epoxy polymers with high glass transition temperatures. For the majority of stereolithographically structured epoxy resins which typically have glass transition temperatures of around 60 °C the light-induced bonding technique is preferable. For all three techniques we performed DIN EN-conform tensile testing demonstrating maximum bond strengths of up to 350 kPa which is comparable with bond strengths reported for PDMS-to-PDMS bonds. For all bond methods, long-term stability as well as hydrolytic stability was assessed.
Rigid microfluidic channels can be built by various means.5 For prototyping, stereolithography (STL) has proven to be a convenient technique. The main advantage of this manufacturing technology is that it allows one-step manufacturing of complex 3D structures. Other commonly used microfabrication techniques, such as lithography, are limited to 2.5D structures. The shapes of these structures can only be defined in x- and y-direction. The height of the structure (z-direction) is always the same. Near-3D structures can be realized by stacking several of these structures. However, this somewhat restricts the design possibilities.5 Channels with circular cross-sections or inclined connections between several layers are challenging to manufacture. In STL these structures can be fabricated easily. In STL a light beam is used to locally polymerize a photosensitive resin. Almost all commercially available photosensitive resins are based on an epoxy polymer. In order to fabricate a component by means of STL a structure defined by a digital CAD file is split into several layers a process referred to as slicing. Using the digital structure information from these slices the light beam writes a pattern into the liquid resin. Wherever the surface is exposed the resin hardens. The process is repeated for each layer. Each layer is manufactured directly on top of the last layer resulting in a near-3D approximation of the digital 3D structure. After this process the rigid structure can be removed from the resin. The liquid resin that remained in the cavities of the component has to be washed out. This process enables the creation of almost all 3D structures that can be created using a CAD system. In addition STL allows fast prototyping. Especially in research and development this technique significantly reduces the time required for transferring a microfluidic chip concept to a testable physical structure.6 However, in addition to rigid channel structures for guiding pressure, membrane-based actuators require at least one flexible wall which can bulge. Even though multimaterial STL has been introduced recently,7 the fabrication technique is still limited to rigid polymers. Therefore STL can only be used effectively for designing complex active microfluidic systems if a suitable bonding strategy for bonding flexible PDMS membranes to stereolithographically manufactured rigid epoxy components is available. Methods for bonding other materials have been described in literature. The most commonly used technique includes thermal bonding, which is suitable for all thermoplastic polymers.8 However, PDMS is not a thermoplastic material, so thermal bonding is not possible without using an intermediate layer of uncured PDMS.9 Intermediate layers have also been proposed for gluing PDMS to other substrates.10 In microfluidics gluing is a challenging process since the glue, i.e. intermediate layers of uncured PDMS, driven by capillary forces, tends to spread into the small structures during the bonding process, resulting in clogged channels. Other bonding techniques that are based on plasma-induced bonding11 have also been proposed. Despite the fact that plasma-induced bonding is a well-established fabrication process in microfluidics, it is a vacuum process which can only be carried out if suitable equipment is available and thus limits the experimental throughput as the processes are usually time-consuming. To overcome these drawbacks, surface activation using a handheld corona discharger has been described in literature as a low cost method for PDMS-to-PDMS bonding.12 As the technique can be carried out at ambient conditions it can be performed in nearly every lab. Although the process is highly suitable for bonding PDMS components it fails to bond PDMS to most other plastic surfaces directly. In order to solve this problem, several bonding methods using functional silanes have been proposed recently:
• Chemical gluing13 is a technique which uses two silanization steps. The first step generates amino groups on the PDMS surface using (3-aminopropyl)triethoxysilane (APTES). In the second silanization step, epoxy groups are created on the second substrate using (3-glycidoxypropyl)trimethoxysilane (GPTMS).14 Conformal contact of the prepared surfaces leads to an amine–epoxy bond. This technique has been used for bonding PDMS to glass,14 polymethylmethacrylate (PMMA), polycarbonate (PC), polyimide (PI), polyethylene terephthalate (PET),15 cyclic olefin copolymer (COC)16 and to PDMS.17
• Silanization using APTES18 imitates the surface characteristics of PDMS on non-PDMS polymer surfaces, such as, e.g., PMMA,19 PC,20 COC, polystyrene (PS),21 polyethersulfone (PES), PET,22 thio/lene-based epoxies,23 polypropylene (PP), high density polyethylene (HDPE), and Teflon,24 and on metal surfaces, such as gold, platinum, cooper, aluminium, and iron.24 The APTES molecule can be coupled to surfaces in different ways, depending on the surface and the protocol used for silanization.25 Even though most papers did not investigate the chemical coupling mechanism it seems that the ethoxy function groups of the molecule form a layer that has properties similar to PDMS. By activating both the PDMS and the coated surfaces with plasma the two materials can be irreversibly bonded.
• Silanization with tetraethoxysilane (TEOS) was proposed for bonding thermoplastic polymers to PDMS.26 The silane is coupled to the surface using a sol–gel process. This technique is especially suited for all applications that cannot tolerate remaining functional groups on the surface of the substrates. For example, chemical gluing leaves unreacted epoxy and amine groups on the substrate surfaces which may interfere with subsequent experiments and therefore need to be blocked.
However, all of these methods either require high temperature above 50 °C or immerse the chip material at some point in a reactive solution, such as GPTMS dissolved in anhydrous ethanol14 or APTES dissolved in water.18
Another technique that has been proposed recently uses poly[dimethylsiloxane-co-(3-aminopropyl)methylsiloxane] to create a bondable surface on non-PDMS polymers, such as PC, PET, polyvinylchloride (PVC) and polyimide (PI). The surface modification process of this technique can be carried out at room temperature and does not include any solvent treatment. However, to complete the bonding process the samples are exposed to a temperature of about 80 °C for one hour after plasma activation.27
We found all of these techniques unsuitable for STL epoxy components, as high temperature treatment results in channel deformation due to the low glass transition temperatures of most STL structurable epoxies, and many (mostly polar) solvents will cause significant swelling of the chip material.
To overcome these problems we recently introduced a UV-light based silanization technique using GPTMS for bonding stereolithographically structured epoxies, e.g., Accura 60 to PDMS.4 During attempts to apply this bonding technique to other stereolithographically structured epoxies that are often used in microfluidics we found that the ability of GPTMS to couple to these surfaces varies widely. This is most likely due to the various admixtures often blended into stereolithography resins to optimize the optical and mechanical behaviour of the cured components. These admixtures result in varying chemical compositions of the resulting surfaces which necessitates optimizations to our original protocol.
In this paper we describe three strategies to couple silanes to the surfaces of stereolithographically structured epoxy polymers. We are confident that most commercially available stereolithographically structured epoxy resins can be bonded to PDMS using one of these methods. As in our previous paper,4 we optimized the methods such that a handheld corona discharger is sufficient for bonding once the appropriate silane layer is in place and polymerized to a polysiloxane layer. This allows corona-induced activation and subsequent bonding to a PDMS layer as commonly employed in microfluidics. Among the three techniques, one is based on a sol–gel process involving heat treatment and is therefore limited to epoxy resins with glass transition temperatures of 70 °C or higher. The other two techniques use a UV-light induced polymerisation to form the polysiloxane layer. None of the latter methods involve solvent treatment or tempering steps. Since the whole process is carried out at room temperature, it is especially suitable for silanization of epoxy resins with lower glass transition temperatures. By applying these techniques to three different commonly used epoxy resins (Accura 60, Watershed XC 11122 and ProtoTherm 12110) we demonstrate that the strategy used for coupling a silane to the surface layer highly influences the bond strengths achievable. Choosing the adequate silanization technique maximum bond strengths of 687 kPa, 349 kPa and 310 kPa were obtained in DIN EN-conform tensile testing for Accura 60, Watershed XC 11122 and DSM 12120 HT, respectively. Furthermore, we demonstrate hydrolytic stability of the obtained bonds as well as long-term stability during cyclic loading with pressurized air for millions of actuation cycles.
However, plasma-induced bonding can be facilitated by first introducing a polysiloxane layer on these surfaces. As stated, this is carried out by coupling and crosslinking silanes to the surface. The crucial task is to identify the ideal silane for a specific surface and the best coupling strategy. When it comes to STL epoxy resins this is especially difficult. Most commercially available STL epoxy resins contain admixtures. These admixtures are used for tailoring mechanical or optical properties, swelling behaviour in aqueous solutions or the glass transition temperature of the resin. Due to these admixtures STL epoxy components can have material properties that are different from the ones of pure epoxy polymers.
The first silane we tested was GPTMS which contains an epoxy group.4 However, we found this protocol insufficient for other commonly used STL resins, e.g., Watershed XC 11122 and Somos DSM 12120 HT. Thus we introduced a second silane, i.e., dimethoxydimethylsilane (DMDMS), which has two methoxy groups. However, we found that none of these silanes is suitable for all three epoxies studied. Thus the ideal silane for a specific resin has to be identified before bonding. In addition the bonding protocol of the surface modification procedure needs to be adapted. All silanization processes described in this paper rely on an acidic attack at a functional group of the silane to be coupled. This attack can either be triggered by using a photo acid generator (PAG) resulting in light-induced polymerisation or by using a sol–gel process with admixed acids. The bond strengths achieved with UV-light induced polymerisation and sol–gel induced polymerisation are different, even though the same silane was coupled to the same surface. Suitable silanization protocols were developed which allow introducing the necessary polysiloxane layers. Subsequently, corona-based activation and bonding to PDMS substrates can be carried out.
For preparing the samples Accura 60, Watershed XC 11122 (both kindly provided by 3D Systems GmbH, Germany) and Somos DSM 12120 HT (structured by Proform AG, Switzerland) were used as typical STL resins which have been used in microfluidics before.29 With this material selection we cover all STL resins that are recommended for microfluidic applications by well-established manufacturers. Elastosil RT 601, which was used for preparing the molds for the samples, and Elastosil M 4600, used for fabrication of the PDMS-membranes, were purchased from Wacker Chemie AG (Germany). PMMA substrates (10 × 10 cm2) used as substrates for spin coating of the PDMS membranes were fabricated by the machine shop in-house. Unfilled epoxy resin 1122A used as material for the body of the test specimens for tensile testing was purchased from RS Components (United Kingdom). Water dyed with red ink was used to test the hydrolytic stability of the samples.
The angle between sample surface and analyser was set to 45°. For analysis of pure epoxy materials cut surfaces of the samples were prepared by using a cleaned scalpel. The surfaces of coated samples were analysed directly. Survey scans were recorded with an X-ray source power of 30 W and pass energy of 187.85 eV. Narrow scans of the elemental lines were recorded at 23.5 eV pass energy.
All spectra were charge referenced to the O 1s at 532.0 eV. Data analysis was performed using ULVAC-PHI MultiPak program, version 9.4.
The specimens for these tests were produced in a similar fashion to the ones for tensile testing. A rectangular PDMS mold was used to create a thin layer of the specific cured STL epoxy resin, namely Accura 60 and Watershed XC 11122, before filling the mold completely with epoxy resin 1122A. After demolding the two-layer component was split into five smaller pieces using a saw. Two holes (diameter: 1 mm) were drilled into each of the fragments forming a channel structure with a 90°-turn (see Fig. 1c). The hole was then sealed from the top (i.e., the surface which consisted of the respective STL epoxy resin) with a flexible PDMS membrane using the respective bonding method. The channel was then connected to a pressurised air-line and the pressure set to 17.2 kPa via a manually controllable manometer. Using a computer controllable valve (type Angar scientific 2312, Angar scientific inc., USA) this pressure was applied cyclically to the sample at a frequency of 10 Hz. After every 30 hours (which corresponds to about one million load cycles) the pressure was doubled to 34.5 kPa. This pressure was reported as effective pressure for pneumatically controlled push-down valves made by soft lithography.30 After another 30 hours the pressure was doubled again resulting in a step-wise increasing load profile (see Fig. 1d). The next step, 68.9 kPa, was chosen because this pressure is used in pressure driven flow focusing.31 During the whole time the membranes were checked for delamination, rupture or sealing failures.
In order to do so microfluidic chips containing one channel (width: 250 μm, height: 250 μm, length: 45 mm) manufactured from Accura 60, Watershed XC 11122 and DSM 12120 HT, respectively, were manufactured using STL by Proform AG (Switzerland).
These channels were bonded to a PDMS membrane using either the GPTMS or DMDMS photoinduced bonding method. The samples were connected to a peristaltic pump (type MC-MS/CA-4/8, purchased from Ismatec, Germany) using polytetrafluoroethylene (PTFE) tubes (see Fig. 1b). The samples were then probed with water. The flow rate within the microfluidic channel was 170 μl min−1. During the whole experiment the samples were examined for leakage.
A simple method for investigating the effectiveness of surface modifications is based on measurement of the advancing surface contact angle. This method is often used in literature in order to determine whether a specific silane was successfully coupled to a surface. Table 1 summarizes the results of water contact angle measurements carried out on cured Accura 60, Watershed XC 11122 and DSM 12120 HT surfaces. The uncoated cured epoxy surface shows contact angles of around 70° for water. After creation of the siloxane layer a significant increase in the advancing contact angle can be observed on all epoxy surfaces, indicating that the surface has been modified successfully.
Investigated epoxy material | Pure epoxy surface | Epoxy modified with GPTMS using UV light | Epoxy modified with DMDMS using UV light | Epoxy modified with DMDMS using sol–gel method |
---|---|---|---|---|
a Purely absorbed siloxane layer which proved to be unusable for bonding. | ||||
Accura 60 | 71.5° ± 3.25° | 84.6° ± 1.85° | 103.4°a ± 3.17° | 81.8°a ± 0.92° |
Watershed XC 11122 | 66.2° ± 1.46° | 96.06 ± 0.78a | 90.01° ± 12.28° | 77.97° ± 3.52° |
DSM 12120 HT | 70.2° ± 2.73° | 87.5°a ± 2.28° | 106.0° ± 5.34° | 82.3° ± 2.58° |
However, successful creation of a siloxane layer is only the first step of the bonding process. We found contact angle measurement not sufficient to predict whether or not the surface was suitable for creating bonds of sufficient strengths. This is due to the fact that the creation of the siloxane layer may have been in fact successful but the linkage of this layer to the surface has been insufficient. The explanation for this behaviour is a result of the ability of the silane to crosslink, even if it is unable to couple to the surface chemically. This results in a purely adsorbed siloxane layer created on top of the respective cured epoxy surface. Depending on the silane used this adsorbed siloxane layer may have the same surface properties as covalently correctly coupled siloxane layers and therefore shows identical contact angles (see Table 1). We therefore characterized all substrates using pull-off tests. During these test we found that siloxane layers made from DMDMS are not able to efficiently link to Accura 60. On the other hand siloxane layers created from GPTMS are not effectively linked to Watershed XC 11122 and DSM 12120 HT. From the last two materials the membranes could be peeled off easily by hand. As depicted in Table 1 these combinations showed the same shift in the contact angle as the ones that could be used for efficient bonds.
The contact angles of the pure surfaces also cannot be used to predict the silane which should be used for bonding. As depicted in Table 1 Accura 60 and DSM 12120 HT exploit almost the same contact angle even though the silanes that form sufficient bonds on these surfaces are not the same. This result could be substantiated by contact angle measurements with solvents of different polarities on uncoated epoxy components for which the experimental data can be found in the ESI.†
Sample | C | O | F | Si | Cl |
---|---|---|---|---|---|
Watershed XC 11122 pure | 80.6 | 18.7 | 0.4 | 0.3 | |
Watershed XC 11122 + DMDMS UV | 48.5 | 25.6 | 25.9 | ||
Watershed XC 11122 + DMDMS sol gel | 49.9 | 25.7 | 24.4 | ||
DSM 12120 HT pure | 71.7 | 22.90 | 5.4 | ||
DSM 12120 HT +DMDMS UV | 49.9 | 26.8 | 23.3 | ||
DSM 12120 HT + DMDMS sol gel | 49.6 | 27.3 | 23.1 | ||
Accura 60 pure | 73.6 | 24.3 | 2.1 | ||
Accura 60 + GPTMS UV | 56.4 | 32.0 | 0.4 | 11.2 |
Narrow XPS scans of O 1s, C 1s, Si 2p, and valence band of uncoated and coated epoxy samples were compared with the literature reference for PDMS.33Fig. 2 exemplarily depicts the Narrow XPS scans of DSM 12120 HT. It can be seen that the coated surfaces are almost identical to the reference spectrum whereas the pure epoxy sample shows significant variances in all four scans.
![]() | ||
Fig. 2 Survey (a) and narrow XPS scans of O 1s (b), C 1s (c), Si 2p (d), and valence band (e) of pure and coated DSM 12120 HT samples. As depicted the spectra of the coated samples are almost identical with the spectrum of pure PDMS which has been taken from the literature.33 The curves of pure DSM 12120 HT, which were obtained using the same method show significant differences compared to the PDMS reference. The spectra of the other epoxy materials and their coated surfaces can be found in the ESI.† These measurements show that the created layers resemble PDMS very closely chemically. |
The structure of uncoated and coated surfaces was investigated using SEM. The images obtained from Accura 60 surfaces are depicted in Fig. 3 (pictures of Watershed XC 11122 and Somos DSM 12120 HT can be found in the ESI†). The pure sample (Fig. 3a) shows the typical striped structure of a stereolithographically structured substrates. This structure is generated during the manufacturing process when the polymer is polymerized layer by layer with a typical layer thickness of 50–100 μm to form the component. The reasons for the large standard deviation we see in the results of the contact angle measurement results from this microstructure. Even after surface modification the structure can be clearly seen on the SEM images. In addition to the striped structure the coated samples also exploit a sporadic unevenness which is due to the formation of polysiloxane agglomerated during the coating process (see Fig. 3c). The lateral dimensions of these agglomerates vary between 3 μm and 35 μm.
![]() | ||
Fig. 3 SEM images of Accura 60 surfaces before and after coating. a) The image of a pure Accura 60 surface shows the stripped structure that is commonly found on all stereolithographically structured components. The lines indicate the individual layers that have been polymerized one after another during the manufacturing process. b) This structure can also be seen on the coated device. c) A close up at the coated surface shows that the coating introduces an additional unevenness. This might be due to the fact that the surplus silane agglomerates on the surface. Similar results have been obtained for the other materials. The pictures of these experiments can be found in the ESI.† |
Epoxy resin | Accura 60 | Somos DSM 12120 HT | Watershed XC 11122 | ||
---|---|---|---|---|---|
Silane + bonding method | GPTMS + UV | DMDMS + UV | DMDMS + sol–gel | DMDMS + UV | DMDMS + sol–gel |
Average bond strength [kPa] | 356.9 ± 211.0 | 67.2 ± 38.8 | 156.8 ± 104.8 | 33.7 ± 19.7 | 193.1 ± 110.5 |
Maximum bond strength [kPa] | 686.5 | 132.3 | 310.4 | 69.3 | 349.3 |
Minimum bond strength [kPa] | 156.8 | 30.2 | 77.4 | 15.5 | 54.9 |
However, these values have not been determined using the same test setup. Thus they cannot be directly compared. Our setup eliminates the influence of shear forces completely whereas the setup used in the reference referred to a mixed loading case.28 For information about the behaviour of the respective bonds in mixed load cases see section “Long-term stability of bonded PDMS/epoxy compounds under cyclic loading”.
We found average bond strength of ~360 kPa for Accura 60 samples bonded using a siloxane layer created from GPTMS. This value is higher than reported for pure PDMS/PDMS bonds. It is important to note that without creation of the siloxane layer Accura 60 cannot be bonded to PDMS at all. The bonds on siloxane layer created from DMDMS via the sol–gel-method on DSM 12120 HT and Watershed XC 11122 reached average bond strengths of ~160 kPa and ~210 kPa, respectively. Again, without the siloxane layer bonding of these substrates is not possible at all. Using the photoinduced siloxane layer from DMDMS we obtained bond strengths of ~70 kPa and ~30 kPa for DSM 12120 HT and Watershed XC 11122, respectively. In general, bonds on siloxane layers created from DMDMS via the photoinduced technique were not as strong as the bonds that were formed on the same epoxy substrate using the sol–gel process. However for most STL epoxies the photoinduced bonding is preferable, because it is a mild process that requires no solvent or heat treatment. As discussed, deformation or permanent damage may result from these treatments, especially given the fact that the sol–gel process requires a mixture of water, ethanol and hydrochloric acid. If strongly crosslinked epoxy resins with high glass transition temperatures are used, the sol–gel process is a good alternative for creating stronger bonds. The big variance between the maximum and minimum values that can be seen for all sample types has been reported to be a typical phenomenon for plasma- and corona-induced bonding methods.28 As the actual bonding between the siloxane layer and the PDMS substrates was based on corona activation we expected comparable variations in bond strengths.
In summary, all bonding techniques yielded, if applied using the correct silane on the respective epoxy surface, bonds of strength and durability which exceed by far the demands of typical microfluidic applications, such as pneumatically controlled valving or pressure driven flow focusing. They all endured more than two million duty cycles which amounts in a total duration of 60 hours at 10 Hz. Again, it has to be noted that none of the material combinations tested can be bonded without the siloxane layer.
Using DIN EN-conform tensile testing we could proof bond strengths of ~357 kPa for bonds on photoinduced created siloxane layers from GPTMS on Accura 60, ~67 kPa for bonds on photoinduced created siloxane layers from DMDMS on DSM 12120 HT and ~193 kPa for bonds on siloxane layers created on Watershed XC 11122 using DMDMS via the sol–gel technique, respectively. These values compare well with literature values of 290 kPa reported for PDMS/PDMS bonds and outperform these values significantly for some material combinations.28 The best bonding techniques for the respective epoxy surface were subjected to cyclic load testing in order to assess their applicability for membrane actuators. During tests each sample endured at least 2.3 million duty cycles. All but one sample failed due to membrane rupture and not due to bond failure, even at elevated pressures. This indicates that the weak point in these experiments was not the bond but the 300 μm thick PDMS membranes that were used. For the newly reported creation of siloxane layers via photoinduced surface modification methods we also demonstrated hydrolytic stability. In these experiments all but two bonds were leak-tight for more than 840 hours in operation.
In summary, the techniques presented in this paper enable fast and easy bonding STL epoxy components to PDMS. By conducting various characterizing experiments, including DIN EN-conform tensile testing, we could show that the bonds on siloxane layers created with either the described photoinduced or sol–gel based techniques can be used in membrane actuators with millions of load cycles sustained at elevated pressures as well as in microfluidic chips where they are exposed to aqueous solutions.
This work has partly been funded by the German Federal Ministry of Education and Research (BMBF), research grant 16SV5775.
Footnote |
† Electronic supplementary information (ESI) available: Additional XPS measurements, ATR-IR spectra, SEM images of coated and uncoated surfaces and contact angle measurements of solvents with different polarities on uncoated epoxies. See DOI: 10.1039/c4lc01440e |
This journal is © The Royal Society of Chemistry 2015 |