Mineral Spirits-Based Microemulsions: A Novel Cleaning System for Painted Surfaces

This paper reports further developments emerging from a collaboration between The Dow Chemical Company, Tate, and the Getty Conservation Institute which seeks to explore improved cleaning systems for unvarnished modern painted surfaces. Specifically, the present study describes three novel microemulsion systems based on water and mineral spirits, each formulated with different surfactants, either ionic or non-ionic. Of particular interest in the systems examined is their capacity to form thermodynamically stable water-in-oil (solvent-continuous) microemulsions which are clear, fluid, and simple to prepare. Phase diagrams are presented for each system type. Compared against more conventional aqueous and hydrocarbon solvent cleaning liquids, findings are reported of systematic evaluations of the performance of selected microemulsion formulations in cleaning artificially soiled reference paint films. Summaries are included of case study conservation treatments conducted at Tate in which the mineral spirits-based microemulsions formed part of the surface-cleaning treatment strategy.


. INTRODUCTION
From the cleaning point of view, conservators understand that modern artists' paint quite often can be "tricky" paint. Typically not protected by a coat of varnish, modern paintsacrylic, vinyl, oil, and so oncan bind dirt strongly and be very sensitive to many of the common cleaning agents in the conservator's armoury. The responses and sensitivities of particular kinds of modern artists' paint to cleaning treatments have received an increasing amount of research attention in recent years, as evidenced, for example, by the various studies on water-sensitive modern oils (Mills et al. ; Tempest et al. ; Silvester et al. ) and on polyvinyl acetate paints (Pereira et al. ). But the largest body of recent research into the cleaning of modern paints has focused on artists' acrylic emulsion (latex) paints and on understanding the risks associated with commonly used wet and dry cleaning systems on the surface and bulk properties (Ormsby and Learner ). Once known, steps can be taken by the conservator to minimize those risks in practice. The fact remains, however, that the possibilities for selecting effective liquid cleaning agents for acrylic and other tricky modern paints are often constrained by their intrinsic sensitivity to liquids. Generally speaking two main approaches tend to be adopted for cleaning with ungelled liquids. One is to harness the dirt-removing power of aqueous solutions, through addition, say, of surfactant and/or chelate/deflocculant, at the same time controlling pH and conductivity to restrict paint swelling, pigment pick-up, and other possibly unwanted effects such as surfactant extraction. The critical influences of solution pH and conductivity on acrylic paint sensitivity have been demonstrated by studies that have explored ways of controlling waterinduced swelling and the extraction of surfactant and other constituents from these paint films during aqueous cleaning (Wolbers et al. ; Dillon et al. ) The other approach is to take an inactive, nonpolar organic solvent and try to improve its cleaning performance, especially its ability to pick up and disperse particulate dirt. Typically with the latter approach, aliphatic hydrocarbon solvents like mineral (petroleum) spirits would be selected, but the range of non-polar solvent options has broadened recently by the introduction to conservation practice of linear and cyclic silicone solvents (methicones and cyclomethicones) (Stavroudis ). By virtue of their low swelling power and low solvency interaction with acrylic paint media, such non-polar solvents might appear to offer potential for low risk wet-cleaning systems, but their relatively poor cleaning efficacy sometimes limits practical effectiveness. The idea of trying to combine the respective benefits of aqueous and non-polar solvent systemsthat is, exploiting the high cleaning efficacy and adaptability of the aqueous environment and the relatively low risk (low swelling) offered by non-polar solventsis the root of continuing development work by our group on a class of materials known as microemulsions, particularly the water-in-oil (WiO) type in which the non-polar (solvent) phase is the continuous phase. A pre-existing WiO microemulsion product of The Dow Chemical Company, INVERT™ , featured in the early stages of our evaluations of possible new cleaning agents for artists' acrylic emulsion paints (Keefe et al. ; Ormsby et al. ), and while this product did show very good cleaning efficacy, practical trials demonstrated it was generally too aggressive on the type of paint under consideration for safe, controlled use. Accordingly, we started investigations into alternative microemulsion formulations that would be better tailored to the specific application of removing dirt from works of art created in acrylic latex paint, or similarly sensitive media.
For the The Dow Chemical Company, Tate, and the Getty Conservation Institute (Dow/Tate/GCI) collaboration, the focus has been on WiO microemulsions in which the "oil" component is a mineral spirits-type aliphatic hydrocarbon solvent. Keefe et al. () described the process by which an initial series of specific microemulsion formulations (hereafter called Series ) was determined with the aid of Dow's high throughput robotic laboratory systems. The Series  microemulsions comprised Shellsol D solvent as the continuous water-immiscible phase, a linear alkylbenzene sulfonate surfactant (LAS, sodium dodecylbenzene sulfonate), two co-surfactants -butanol and hexanol, and water. Since their first presentation in , the Series  microemulsions have been evaluated from the practical point of view through a combination of systematic laboratory testing (at Tate), field trials, and subjective assessment during professional development workshops organized by GCI ("Cleaning Acrylic Painted Surfaces" [CAPS]) Los Angeles , New York , London , Washington D.C. , Sydney , and Otaawa .
The various practical evaluations of the Series I microemulsions provided constructive feedback on desirable and undesirable characteristics which then guided the formulation of two further types of mineral spirits-based WiO microemulsion, hereafter called Series  and Series , selected formulations from which have also been evaluated in systematic cleaning trials. The Series  microemulsions are prepared from water, an aliphatic mineral spirits (Shellsol D), with the Dow product ECOSURF™ EH Surfactant [note ] as primary surfactant and -butanol and -hexanol as co-surfactants. The Series  microemulsions were developed with the aim of eliminating the alcohol co-surfactants of the Series  and Series  formulations, primarily in attempt to reduce the activity on paint. Series  microemulsions thus comprise just water, an aliphatic mineral spirits (Shellsol D), and a single anionic surfactant, sodium dioctylsulfosuccinate (DOSS, the surfactant ingredient of Aerosol OT products). Two different forms of sodium dioctylsulfosuccinate have been used in the preparation of Series -type microemulsion systems: Series a is based on an existing Dow commercial product TRITON™ GR-M Surfactant; Series b is prepared from laboratory grade pure sodium dioctylsulfosuccinate. This paper has several aims: to describe in detail the mineral spirits-based Series , , and  microemulsions, including presentation of their phase diagrams; to report on the systematic trials conducted at Tate on comparative cleaning performance in which the microemulsions were evaluated alongside various other more conventional aqueous and mineral spirit-based cleaning liquids; and to present some case study conservation treatments in which the microemulsions have been evaluated as part of a broader surface-cleaning treatment strategy. Although the new microemulsion systems described here have been developed primarily for removal of dirt from modern unvarnished works, they may also find useful applications in cleaning older works: one such example is presented among the case studies.

. MICROEMULSIONS
Microemulsions are special combinations of a polar liquid, usually water or a dilute aqueous solution of electrolyte, and an immiscible, hydrophobic liquid that are brought into thermodynamically stable co-existence by means of surfactant and, in some cases, co-surfactant (co-solvent), the combination of which is sufficient to reduce the interfacial tension between the two immiscible phases to ultra-low values (Klier et al. ). Microemulsions are optically clear because the size of the domains of dispersed phase is considerably smaller than the wavelength of visible light, typically - nm. Microemulsions form upon simple mixing of the components and do not require the high shear conditions generally used in the formulation of ordinary emulsions, and typically they are low viscosity fluids. Another feature that distinguishes microemulsions from regular emulsions is that the domains are dynamic in terms of size and shape and undergo continuous break-up and reconstitution: the very low interfacial tension allows thermal motions to spontaneously disperse the two immiscible phases. The dynamic nature of these systems allows effective cleaning performance on hydrophobic and hydrophilic components of soils while limiting the degree of contact with the aqueous phase of the cleaner. Considering microemulsions as cleaning agents, a useful property is that the aqueous phase can to some degree accommodate soluble salts and/or other ingredients that might usefully modify performance. Microemulsions are typically described as direct (oil-in-water [OiW] or water-continuous), reversed (WiO or oil-continuous), or bi-continuous. The submicroscopic structures of hypothetical OiW, WiO, and bi-continuous microemulsion systems are illustrated schematically in figs. a, b, c. These examples include a small molecule co-surfactant (typically an aliphatic alcohol or glycol ether) and a primary surfactant residing at the oil/water interface. The phase behavior of simple microemulsion systems comprising oil, water and surfactant can be studied and described with the aid of ternary phase diagram in which each corner of the diagram represents % concentration of the particular component. Where the microemulsion ingredients consist of more than three different substances, as in the case of our Series  and Series  types, the phase behavior may still be characterized by a "pseudo-ternary" phase diagram in which the total proportion of surfactant and co-surfactant substances is combined on a single axis.
Microemulsions find use in a broad range of industrial and technological applications including enhanced oil recovery, oil removal from contaminated groundwater sites, consumer and industrial cleaning formulations, metal working, catalysis, advanced ceramics processing media, production of nanostructured materials, dyeing, agrochemicals, cosmetics, foods, pharmaceuticals (including controlled delivery), coating formulations, and biotechnology (Klier ; Sharma et al. ). They are not entirely new to art conservation. Carretti et al. () described two different xylene-in-water microemulsions (also referred to as "nanocontainers") for different cleaning applications: the removal of acrylic coatings/consolidants and of black crusts from historic wall paintings. Carretti et al. () reported a further development in the use of OiW microemulsions for cleaning works of art in which the microemulsions were embedded in a polymeric network. Cleaning systems described as microemulsions have been reported recently by Brajer et al. () for the removal of acrylic resin (PARA-LOID™ B) from wallpaintings: these are water-rich systems consisting of alcohol ethoxylate surfactants and butanone (methyl ethyl ketone), with variants including also other oxygenated solvents. Microemulsion formulations, also intended for the removal of polymer coatings from wallpaintings, are described in Baglioni et al. (, -), which contains a comprehensive bibliography on this application.

. MICROEMULSION COMPOSITIONS AND PHASE DIAGRAMS
As already noted, the Dow/Tate/GCI collaboration has identified three types of microemulsion (Series -) that might serve as useful cleaning liquids for the removal of dirt from artist's acrylic paint surfaces. The latter two series evolved under influence of feedback from testing at Dow, Tate and from CAPS workshop participants, the essential goal being to achieve lower levels of cleaning activity than the Series  systems. Full phase diagrams have been prepared for each type of microemulsion to provide a compositional map showing the combinations of each set of ingredients that form stable WiO microemulsions, those that form OiW microemulsions, and those that have different structure (multi-phase; bicontinuous). The three microemulsion types were all formulated with the same proprietary aliphatic hydrocarbon mineral spirit, Shellsol D, which in selected cases could be substituted with a similar product, Shellsol D. The primary surfactants selected for the three microemulsion classes were chosen on the basis of several criteria: their general cleaning strength in adjacent fields of textile and hard surface cleaning; their cleaning performance in previous studies on soiled acrylic painted surfaces (Keefe et al. ); their environmental profile; and their known capacity to form stable WiO microemulsions. In addition to the primary surfactant, Series  and  microemulsions contain a blend of -butanol and -hexanol as co-surfactants, inclusion of which was necessary to obtain stable microemulsions. Series  and  microemulsions have been prepared with both deionized water and adjusted water (pH ., conductivity . mS/cm) which illustrates how the water phase might be modified to optimize the overall cleaning performance and minimize risks associated with the swelling and extraction of materials from these paints (Dillon et al. ).
Dow's high throughput robotic laboratory systems, as described in Keefe et al. (, ), were utilized to study combinations of ingredients that resulted in stable microemulsions in the Series  system. The Series  and Series  microemulsions have been prepared manually. The microemulsions are easily prepared by hand, as required: typically made up in small batches of  g total weight (or multiples of  g), amounts of each ingredient by weight are determined using a two decimal place digital balance, and mixing is achieved by simple manual shaking. For microemulsions from Series  and Series , the ingredients can be combined in any order; for Series  emulsions, the LAS surfactant must first be dissolved in the water phase. The recommended order of addition of ingredients for Series  is discussed in more detail below.
General observations on each of the Series ,  and  microemulsions are provided in table . The compositions of each individual microemulsion formulation are presented in tables , , and . Phase diagrams for each series are shown in figs. , , and . Broadly speaking, cleaning efficacy within each microemulsion series varies across the compositional space described by the phase diagram, with greater cleaning performance at higher proportions of water and surfactant. The cleaning activity of a given microemulsion is also governed by a number of other factors, including: the surfactant type and strength (hydrophile lipophile balance [HLB] number), the presence and type of co-surfactant(s), modifications to the water phase, etc. For the microemulsion series described here, the strength of surfactant increases in the order TRITON™ GR-M Surfactant < ECOSURF™ EH- Surfactant < LAS. In some cases, as briefly described above, the microemulsion systems can accommodate a degree of modification to the water phase, for example by setting a particular pH and conductivity condition, and/or by adding small quantities of a chelating agent, such as citrate, or another surfactant such as ECOSURF™ EH- which might allow for fine-tuning of the cleaning activity.
In connection with the practical application of cleaning painted surfaces, the phase diagrams provide a useful graphic framework for exploring possible cleaning options, helping the user to vary the cleaning performance to suit the needs of the particular object under treatment, for example by shifting to higher or lower cleaning activity, or to lower impact on the original paint, as the situation dictates. These fine adjustments may be done by preparation of new batches of microemulsion having the desired proportions, which is a simple matter of weighing out and mixing the necessary amounts of each ingredient. Alternatively, in some cases, depending on the system type and the position in the phase diagram, the cleaning action of a given formulation may be adjusted by "dilution"; that is by addition of small aliquots of mineral spirits or water in amounts appropriate to shift the relative proportions in one or other direction within the appropriate microemulsion territory of the phase diagram. A further option for adjusting microemulsion cleaning performance is mixing of two preparations: if, say, the user has two ready-made stable WiO preparations with quite different relative proportions of ingredients, mixtures of those preparations will provide intermediate combinations in a linearly additive fashion.
Since all of the microemulsion formulations presented here include non-volatile surfactant components, a final "clearance" step is required in the cleaning process to ensure that any residues of surfactant are fully removed from the paint surface. A general rule-of-thumb is that the solvent used for clearance should be the same as, or close in nature to, the continuous phase of the microemulsion.

.. SERIES  MICROEMULSIONS
The Series  microemulsion system comprises linear alkyl benzene sulfonate surfactant (LAS), -butanol, hexanol, Shellsol D solvent, and water. The phase diagram ( fig. ) shows the respective combinations that result in WiO microemulsions, OiW microemulsions and -phase systems. The data on which the phase diagram is based are summarized as table . Within this series the LAS/-butanol/-hexanol ratio was kept constant (respectively :: parts by weight) to allow presentation of a coherent (pseudo-) ternary phase diagram for this five component mixture. table  includes also data on four other water/LAS surfactant/ (-butanol + -hexanol)/mineral spirits OiW microemulsion formulations: Sample IDs ME- ME- and ME- were prepared with an LAS/butanol/hexanol ratio of .:.:.; ME- was prepared with an LAS/butanol/hexanol ratio of .:.:.. These very closely related systems are included here, as they featured in early systematic comparative cleaning evaluations carried out at Tate, but strictly they should not be plotted on the Series  phase diagram, due to the slightly different surfactant/ co-surfactant blend ratio. The Series  microemulsions were prepared by first dissolving the LAS powder in deionized water. Depending on the grade of LAS, this may require gentle heating, for example by placing the bottle under hot tap water, or onto a low-heat hotplate and shaking by hand. The -butanol and -hexanol are then added to the aqueous solution of LAS, followed by the Shellsol D solvent. The LAS surfactant has the highest surfactant activity of the three systems; thus, of the different systems explored, LAS provides the lowest surfactant levels necessary to form stable microemulsion structures. Lower surfactant levels are beneficial to minimize residual surfactant on the substrate after cleaning treatment and clearance. The LAS surfactant system allows formation both of stable OiW and WiO microemulsions structures depending on the ingredient proportions and the consequent position in the phase diagram. OiW microemulsions are formed when water content is above % w/w; WiO microemulsions are formed at water contents of % w/w and below. Microemulsions did not form when combined surfactant content was very low (see most of the formulations with % w/w combined surfactant) or very high (see the formulations with greater than about % w/w combined surfactant). The structure of the microemulsions as either OiW or WiO was determined by conductivity measurements (Kizilbash et al. ). Water-continuous (OiW) systems have significantly higher conductivity (of the order of ×) that of oilcontinuous (WiO) systems. The cleaning efficacy of Series  microemulsions is generally high due to the LAS surfactant utilized in the system. Furthermore, the water-continuous (OiW) preparations, such as ME-, evaluated in the Tate cleaning trials, are expected to have somewhat greater cleaning efficacy than the formulations producing WiO microemulsions. These systems are probably more suited for treatment of objects that have high soiling levels and more tenaciously adhered dirt, as noted for case study .. The WiO options may also have utility for works of art highly sensitive to wet cleaning with water because Water phase of ME- consists of  parts of a % w/w. solution of NaCl and  parts deionised water.
the contact time required for effective cleaning is shorter due to the high cleaning efficacy. As with all of the systems described here, cleaning treatments using the Series  microemulsions require a clearance step using the appropriate continuous phase solvent depending on the type of system, OiW or WiO; that is, Shellsol D or D solvent for the WiO systems, and pH-and conductivity-adjusted water for the OiW systems. Where possible, if the sensitivity of the paint allows, it may prove beneficial to clear with both types of liquid to minimize LAS surfactant residues. One challenge encountered in the development of the Series  microemulsions was that the ability to form comparable microemulsion structures was strongly influenced by the choice of particular LAS product, the variation in microemulsion-forming behavior seemingly being related to product purity. Four different sources of LAS were investigated in this study, as described in   .. SERIES  MICROEMULSIONS The Series  system is based on water, Shellsol D solvent, ECOSURF™ EH- Surfactant, and a co-surfactant that is again a combination of -butanol and -hexanol. As in the Series  formulations the ECOSURF™ EH- Surfactant/-butanol/-hexanol ratio was held constant (:.:.) to allow presentation of a ternary phase diagram for this five component mixture. The phase diagram ( fig. ) shows the combinations of these constituents that result respectively in stable WiO microemulsions and two-phase systems. This set of constituents does not support OiW microemulsions. The data on which the phase diagram is based are summarized as table . The Series  microemulsions are prepared by simply combining in a small jar or vial the measured amounts of water, surfactant, -butanol, -hexanol, and hydrocarbon solvent to give the particular WiO formulation of interest, and shaking the mixture by hand. As described earlier, Series  systems may be prepared with deionized water, water adjusted for pH and conductivity (e.g. pH ;  mS/cm) or water with added chelating agent. Shellsol D or Shellsol D may be used interchangeably as the aliphatic mineral spirits component. The Series  microemulsions may be suited for cleaning treatments that require intermediate cleaning efficacy because of fairly heavy or tenacious soiling, coupled with sensitive pigment or binder systems. The -butanol/-hexanol co-surfactant combination is needed in these systems to generate the stable microemulsion structures, and its presence may contribute to enhanced cleaning efficacy, but also adds some odor to these preparations.
Clearance of the Series  WiO microemulsions should be carried out with the continuous phase solvent, that is, Shellsol D or D.

.. SERIES  MICROEMULSIONS
The Series  microemulsion system is the simplest of the three described here and comprises just water, aliphatic mineral spirits (Shellsol D or D), and an anionic surfactant; no co-surfactants are required. In the first formulation trials of the Series  systems (hereafter called Series a) the anionic surfactant employed was the Dow product TRITON™ GR-M Surfactant (which is sodium dioctylsulfosuccinate, supplied as % active ingredient in "petroleum distillate" solvent [note ]. The phase diagram ( fig. ) shows the combinations of these ingredients that result in stable WiO microemulsions, bi-continuous, or -phase systems; OiW microemulsions are not formed with the set of ingredients used for Series a. The data from which the phase diagram was generated are summarized as table . In table  values for the quantities of TRITON™ GR-M Surfactant are given first as weight proportion as received and, in parentheses, the net weight proportion of surfactant discounting the volatile solvent component. In similar fashion to the Series  and  systems, these preparations are made by simple hand-mixing of measured amounts of each ingredient to give the particular WiO microemulsion formulation of interest. Series a systems have been prepared with both deionized water and adjusted water (pH ;  mS/cm), etc. interchangeably, likewise with Shellsol D or D solvent interchangeably. This microemulsion class may be suited for cleaning treatments that require gentle cleaning due to highly sensitive pigment or binder systems and/or mild soiling.
One constraint affecting preparation and use of the Series a microemulsions became apparent as a consequence of the GCI's CAPS training workshops held outside the USA: TRITON™ GR-M Surfactant is not available in other regions of the world; and since it contains a flammable solvent the product cannot easily be shipped internationally. Accordingly, there was some incentive to identify alternative dioctylsulfosuccinate surfactant products which would allow for an analog of TRITON™ GR-M Surfactant system to be prepared from scratch from locally available materials, and ideally for the aromatic hydrocarbon solvent components to be substantially reduced or eliminated altogether. After some trials, it was found that analogous Series -type WiO microemulsions could be prepared using near-pure, laboratory grade sodium dioctylsulfosuccinate product (Sigma-Aldrich Cat. No. ), pre-dissolved to % w/w in aliphatic mineral spirits (e.g. Shellsol D/D) to which .% w/w -propanol had been added to marginally increase polarity and surfactant solubility. All of the combinations of water/aliphatic mineral spirits/ surfactant that formed WiO microemulsions using TRITON™ GR-M Surfactant also comfortably formed stable microemulsions with the laboratory grade sodium dioctylsulfosuccinate product (% w/ w in [Shellsol D + .% w/w -propanol]). However, as illustrated in fig. , it was found that with the laboratory grade sodium dioctylsulfosuccinate Series -type preparations formed stable microemulsions (hereafter called Series b) over a considerably broader area of the formulation space compared to TRITON™ GR-M. It can be seen straight away that the area of stable microemulsions extends further towards the top apex of the phase diagram, providing more systems that are relatively water-rich and low in surfactant, characteristics which may allow greater flexibility in terms of tailoring for specific cleaning problems while also attending to concerns about clearance. The identification of a particular formulation as microemulsion was based on the observable characteristics of: homogeneous single phase; optically clear, isotropic; and stable over time. However, the type of microemulsion system formed by the respective Series b formulations (WiO, OiW, and bicontinuous) has not yet been determined. Conductivity measurements, which could in principle distinguish between WiO and OiW microemulsions, could not be reliably acquired from these samples with the equipment available. Formulations lying within the Series b microemulsion space tend to have noticeably higher viscosity as the relative proportions of water and surfactant increase.
The solvent suggested for clearance of The Series a and b microemulsions is Shellsol D (or D) with .% w/w added -propanol, which is the continuous phase for the Series b systems.
Given the anionic nature of the sodium dioctylsulfosuccinate Series -type preparations it might be expected that there is limited scope for adjustment of the chemistry of the water phase, for example, by addition of other ionic additives such as a chelate like triammonium citrate (TAC). Preliminary investigations with the Series b type preparations suggest this is indeed the case: inclusion in the water phase of TAC at the % w/w level prevents microemulsion formation over most of the compositional space. However, when the TAC was reduced to .% w/w, it was found that some microemulsions could still form (hereafter referred to as Series b:TAC), but with a reduced distribution within the phase diagram compared to just water as the aqueous phase ( fig. ). As with the regular Series b systems described above, the particular type of microemulsion system formed (WiO, OiW, and bicontinuous) has not yet been determined: we distinguish here only between systems that are demonstrably not microemulsions (multi-phase and/or optically anisotropic) and microemulsions (homogeneous single phase; clear and optically isotropic).

.. CLEANING SYSTEMS
Selected formulations from each of the three series of mineral spirits-based microemulsions described above were evaluated for cleaning performance in systematic manual cleaning trials conducted at Tate. It was not possible to include every possible microemulsion combination in these trials, hence those selected from within each series represented a range of cleaning efficacies (with respect to removing artificial dirt from acrylic paint films) and were highlighted as promising during initial trials. Selection within each series was also influenced by the desire to trial systems representing a range of water, solvent, and surfactant contents. Those used for the reported study rated highly with respect to cleaning efficacy while also having relatively low impact on paint films. In addition, some had been FIG. . Phase diagram for Series b microemulsions: Water/ sodium dioctylsulfosuccinate/Shellsol D solvent. Formulations were identified as microemulsions on the basis of being optically isotropic (i.e. clear, homogeneous), but WiO systems are not here distinguished from OiW systems. Formulations that were multi-phase and/or optically anisotropic were characterized as non-microemulsions. used on case study works of art (see section ), where they had been chosen after a full assessment of surfacecleaning options had been carried out.
For comparative purposes, the mineral spirits-based microemulsions were evaluated alongside other waterand mineral spirits-based cleaning systems. A number of these had been evaluated in previous work by our group (Keefe et  Modifications included adjusting aqueous-system pH levels to around  and setting ionic strength. All of the surfactants used in the systems tested (see table ) originate from the Dow/Tate/GCI collaboration and have been introduced elsewhere (Dorman ; Stavroudis ). Practical experience on works of art has also influenced some of the minor changes to these systems, for example reducing the quantity of surfactants and chelating agents used in simple aqueous mixtures. The final set of preparations tested (table ) represents the main classes of wet-cleaning systems currently being explored for unvarnished modern paint surfaces; including simple aqueous systems based on deionized water, simple aliphatic hydrocarbon solvent systems (based on Shellsol D, and the mineral spirits-based microemulsions.

.. SAMPLES FOR CLEANING TRIALS
The cleaning performance of the various liquid preparations was evaluated by manual swab cleaning of artificially soiled artists' acrylic paints and acrylic priming. The paints were prepared by casting onto  oz. acrylic primed cotton canvas (Russell and Chapple, UK) using an adjustable film caster (Sheen Instruments, UK), to a dry thickness of  ±  µm. The samples of priming consisted of commercially prepared acrylic primed canvas boards (Atlantis Art Materials, UK). The paint colors (titanium white [PW]; azo yellow [PY]) and paint brands (Liquitex Heavy Body Acrylics and Talens Rembrandt Acrylics) were selected to explore the cleaning effects of the preparations on two brands with known surfactant levels, with more or less vulnerable pigment types, and with different inherent gloss levels. Constituents of these paints have been detailed elsewhere (Ormsby et al. , ). After natural ageing in ambient, dark conditions for  months, the paint and priming samples were soiled with an artificial dirt, the composition and application method of which have also been described previously (Ormsby et al. , ). The cleaning tests were carried out around  years later. At the time of testing, the unsoiled control sections of the Talens paint samples (both colors) had substantial amounts of detectable migrated surfactant, and the Liquitex samples varied between trace (PW) and moderate (PY) amounts of migrated surfactant. The surface conductivity of the unsoiled control samples varied between . and . mS/cm; the priming samples were close to zero.

SYSTEM PERFORMANCE
To prepare for cleaning tests, the soiled area of each sample was divided into squares and numbered (across the two samples tested per color and brand), leaving one square on each sample as a soiled control, to complement the unsoiled control. Cleaning was carried out according to a specific procedure designed to approximate conservation techniques while maintaining an approach that facilitated direct comparison between cleaning systems, samples and paint brands (Ormsby and von Aderkas ). The procedure was slightly modified from that used in a previously reported study (Ormsby et al. ).
For each test, a ready-made cotton swab (Dynarex, USA) was dipped into the cleaning solution and rolled once on paper towel to remove excess liquid. The swab was then rolled with consistent light pressure across the soiled surface of the sample until one of the following stopping points was reached: the sample was considered clean (as judged by eye); pigment was removed (visible on the swab); swelling or other change was noted (e.g. paint surface tackiness, abrasion, burnishing); or  swab rolls had been applied (i.e. the sample could not be fully cleaned). Aqueous solutions were cleared with water at pH  FIG. . Partial phase diagram for Series b:TAC microemulsions: Water + .%w/w TAC/sodium dioctylsulfosuccinate/ Shellsol D solvent. and conductivity  mS/cm (adjusted using a volatile acid and alkali); mineral spirit-based solutions and microemulsions were cleared with Alcosol (= Shellsol) D solvent.
The results of the comparative cleaning tests were quantified using three different descriptors: () The number of swab rolls were recorded until one of the above defined stopping points was reached (where one swab roll was defined as one backward and forward roll). The number of swab rolls to stopping point is effectively an inverse indicator of "cleaning efficacy": a liquid with a low value for this descriptor reflects rapidity of action and vice versa. () The relative degree of cleaning (dirt removal) was rated, after both the cleaning and clearance steps were performed and the surface had dried, across a scale from  to . This was done by visually comparing the cleaned area to the control areas by eye, with a rating of  considered clean (i.e. as clean as the unsoiled control). This indicator reflects ultimate capacity for cleaning. () An overall rating was used to describe (on a scale from  to ) the relative magnitude of overall change to the paint surface, incorporating both the amounts of dirt removed and any perceived alteration to the paint (see Appendix).
Additional notes were recorded on other aspects of the cleaning process, such as the consistency of swabbing action, degree of control over the cleaning solution, surface wetting behavior and any clearance issues encountered.

.. SUMMARY OF CLEANING TEST RESULTS
... AVERAGE SWAB ROLL NUMBERS ACROSS ALL PAINT SAMPLES AND CLEANING SYSTEMS As shown in fig. , the average number of swab rolls required to clean the samples with the microemulsions was considerably fewer than most of the aqueous systems and all of the simple mineral spirits systems. The aqueous systems with added ECOSURF™ EH- Surfactant and/or TAC resulted in a useful reduction of the numbers of swab rolls required to reach the cleaning end point. However in general, the aqueous systems required higher numbers of rolls than the microemulsion systems (with the exception of the .% ECOSURF™ EH- Surfactant and .% TAC blend which equalled the Series  microemulsion results). The enhanced efficacy of the microemulsions may be beneficial when paint surfaces are fragile or heavily soiled and/or where there is a risk of driving soiling into the paint surface with prolonged contact.
The different series of the mineral spirits-based microemulsions showed slight differences in soiling removal efficacy. The slightly lower efficacy of Series  and Series  microemulsions reflects the deliberate design and modification of these systems outlined in section . The addition of water into ME- and ME- increased the efficacy of soil removal (due to the higher water content), however due to the error margins noted, it was not possible to assess whether the addition of TAC to the water phase of the Series  and  microemulsions had any beneficial effect. For the same reason it was not possible to draw conclusions about differences in cleaning efficacy (soiling removal) within each microemulsion series. Using the same samples and assessment criteria as described in section ., additional tests carried out at Tate with four different microemulsion formulations from Series  [table ] -ME- (high water/high surfactant); ME- (high water/low surfactant); ME- (low water/low surfactant); and ME- (low water/high surfactant)confirmed that preparations with higher water content cleaned more efficiently, and that options containing high surfactant levels also increased the rate of soiling removal. Figure  shows the rating scale from  to  where the cleaning systems were ranked in terms of cleaning efficacy based on a visual judgement of the cleaned area after drying. A value of  was considered as clean as the unsoiled control area of each sample. The data presented in fig.  indicates that the microemulsions (and some of the TAC and surfactant-containing adjusted aqueous systems) cleaned the samples to the greatest degree. It was also noted, however, that none of the cleaning systems achieved a rating of . For the microemulsions, it appeared that Series  resulted in a slightly more clean paint surface than either Series  or .

... ASSESSING CHANGE DURING AND AFTER CLEANING
Change to the test samples during and after cleaning could be attributed to two primary effects: the removal of dirt (to varying degrees), and a permanent, perceived, undesirable alteration to the paint. Possibilities for unwanted change or effect on the paint consisted of: pigment removal (i.e. color on the swab); swelling/ abrasion noted at the paint surface during cleaning; significant gloss change noted after cleaning; or issues connected with clearance of the non-volatile components of the cleaning system, which are currently being explored.
Using the classification scheme and the numerical - rating scale described in section . and Appendix , the respective effects of each cleaning system on the four test paints and the acrylic priming were tallied. The tally of observations provided a general idea of how often unwanted change occurred with each system (Ormsby and von Aderkas ).
It suffices here to make some general, qualitative observations on the occurrences of unwanted change to the test paints as a consequence of the various cleaning systems under test. The simple aqueous systems provided a useful benchmark in terms of activity on the paint: virtually no incidences of unwanted change occurred with any of the simple water solutions (D.I. water; pH/conductivity-adjusted water; water with either TAC or ECOSURF™ EH), with one exception in the case of water plus both TAC and ECOSURF™ EH at .%. The aqueous solutions containing TAC and ECOSURF™ EH- Surfactant were almost free of unwanted change (observed at the macro scale) while remaining competitive in terms of cleaning performance.
In similar fashion to the simple aqueous systems, there were few occurrences of unwanted change associated with the set of simple mineral spirit preparations: Shellsol D alone, and with added ECOSURF™ EH or TRITON™ GR-M surfactant. However, some dissatisfaction was expressed at the handling properties of these systems and the level of dirt removal they achieved. It was also noted that these systems were "pushing dirt around" and leaving a "haze". This haze may at least in part be associated with the disturbance of the migrated surfactant layer, causing some scattering of light (Kampasakali et al. ), in addition to any soil remaining on the paint surface.
Compared to the simple aqueous solutions and simple mineral spirit preparations, rather more instances of unwanted change to the paint films occurred with the new microemulsion systems, which presumably reflects the enhanced activity of these multicomponent formulations. The main cause of the change noted was pigment transfer from the PY samples. Exceptions to this general pattern included some of the Series  microemulsions (ME- and [ME- + TAC], that is, the lowest surfactant-containing option tested for Series ) which produced very few instances of unwanted change and offered high cleaning efficacy (see fig. ).
Overall, of the group of mineral spirits-based microemulsions, the Series  options caused slightly fewer occurrences of unwanted change. This may be due to the speed at which the systems removed soiling, thereby reducing mechanical action and contact with the paint surface. The Series  microemulsions had a lower cleaning efficacy than Series , and caused more unwanted change in the form of pigment transfer, gloss change and surface tackiness, which is presumably at least partially due to the slower cleaning action of this series. The Series  microemulsions had the lowest cleaning efficacy of the three series, but had a more controllable action and caused fewer incidences of unwanted change than Series . Within Series , ME- + TAC did not cause any unwanted change, while ME- resulted in unwanted change for only one sample. Conservators testing the three microemulsion series at Tate and in CAPS workshops also reported that the Series  systems tended to act too quickly and were not able to be controlled in a satisfactory way. This may be related to the fact that many of the initial microemulsions tested from Series  were primarily water-continuous (OiW) systems.

... SAMPLES OF ACRYLIC PRIMING
Samples of commercial acrylic priming were included in the systematic cleaning trials as they are inherently rougher than the artists' paint films and represent a more challenging cleaning situation. On the whole, the samples of acrylic priming were resistant to cleaning. None of the simple aqueous and mineral spiritbased solutions cleaned the samples to a significant degree (all reached the  swab roll stopping point) and had a degree of cleaning rating ranging from  to  (of a possible ). However, the mineral spirit microemulsions (particularly the Series  and Series  options) proved to be relatively effective on this sample type, requiring between - swab rolls to reach a stopping point, with an average degree of cleaning rating between  and . It was also noted that it was particularly difficult to judge unwanted change for these samples, and that due to the roughness and greater surface area inherent to these films, none of the cleaned areas resembled the white unsoiled control. Applying microemulsion systems using a brush, in gelled form or through tissue may be useful options to explore.

. TREATMENT CASE STUDIES
Since , the mineral spirits-based microemulsion series have been adopted, where appropriate, as part of the surface-cleaning system testing strategy for unvarnished painted works of art in Tate's collection. Although the cleaning systems were originally designed and modified for use on acrylic paint substrates, they have also been used on aged oil-and polyvinyl acetate (PVAc)-based painted surfaces, as briefly summarized below, thus demonstrating their applicability to a wider range painted surfaces. It is important to note: that the microemulsion systems were used in combination with more established surface-cleaning methods; that the treatments were evaluated subjectively, not fully scientifically; and that in one case (see .) the evaluation and treatment were not completed due to competing program demands.

.. EVA HESSE ADDENDUM (). TATE T
This modern sculpture consists of a gray, titanium white-based acrylic dispersion copolymer polyethylacrylate/methylmethacrylate paint layer with an additional unpigmented PVAc coating on the rope sections. Several attempts had been made to remove the unwanted soiling from this work in the past, to no avail. During a treatment preparation campaign in -, it was established that the paint on the upper portion of the work was neither water nor mineral spirits sensitive and that the bulk of the soiling could be removed using water adjusted to pH  and  mS/cm conductivity. However, the soiling on the coated rope sections did not respond to a range of adjusted aqueous or simple mineral spirit systems. In addition, the PVAc coating and white paint proved to be sensitive to polar solvents. In light of these constraints, the three series of mineral spirits-based microemulsions were tested on the rope sections. The Series  and  systems series proved effective at removing soiling while leaving the PVAc coating intact. After extensive testing however, the Series  microemulsion ME- was chosen as the most appropriate preparation to clean the coated rope surfaces, as it removed the bulk of the soiling while maintaining the integrity of the coating. The microemulsion tests were cleared with Shellsol D solvent and the treatment supplemented in some areas, with modified aqueous systems. As mentioned, due to time constraints, this treatment was not completed prior to loan, and may be revisited with a full scientific evaluation in the near future. Figure  is a detail of Addendum, showing some of the cleaning tests carried out on one of the rope sections from this work.
.. RICHARD SMITH, PIANO (). TATE T Richard Smith's Piano () is a three-dimensional painting constructed from cotton duck on plywood. It is painted in PVAc and possibly oil media. In recent preparations for display, it required treatment in-situ to reduce fingerprints and accumulated soiling resulting from inherent difficulties associated with the handling of this large work. Previous attempts at surface cleaning had been abandoned due to the sensitivity of the yellow paint layer. After brush vacuuming and dry cleaning with sponges, testing for the removal of stubborn fingermarks was carried out using a group of fifteen aqueous and solvent-based systems. To summarize, the soiling on the white passages was successfully reduced using an aqueous system consisting of .% v/v ECOSURF™ EH- Surfactant and .% w/v TAC, cleared with pH and conductivity-adjusted water. The scuffs and fingerprints on the water-sensitive yellow passages were successfully reduced using Series  microemulsion ME-. This system offered controlled, effective dirt removal with no yellow pigment transfer, and clearance using Shellsol D solvent posed minimal risk to this water-sensitive passage. The aim of the treatment was to reduce the fingermarks and scuffs so that they no longer drew the eye towards them; for this, the Series  microemulsion proved valuable by facilitating the incremental reduction of fingermarks without compromising the surrounding paint layer through creating over-cleaned areas and/or pigment transfer. Figure   Turner's Two Recumbent Nude Figures was conserved in - as preparation for display at Tate Britain. At that point in the collaboration the first series of mineral spirit microemulsions were tested alongside other cleaning systems to remove a very stubborn, aged, multi-layered soil deposit from an aged oil paint film. The painting was initially lightly dry cleaned with sponges, then a series of aqueous-system cleaning tests revealed that the paint layer was slightly water sensitive and prone to blanching. After partial cleaning was achieved using established systems, some of the early Series  microemulsions (specifically ME-, , , ) were tested and found to remove the soiling successfully, at a similar controllable efficacy, apart from the ME- (the highest water content variant) which worked more slowly, presumably due to the predominantly oily nature of the soiling layers. According to our current understanding of this series, all of the microemulsions tested on this painting were OiW compositions (table ). The background of the painting was cleaned with ME-,  and  using large swabs, the central figure was cleaned with ME- and  using smaller swabs and by working through tissue and/or removing the soil-containing microemulsion with FIG. . Detail of cleaning tests (small white areas towards the lower portion, close to the end of the rope) using ME- on Eva Hesse's Addendum. This treatment was not completed, as the work was required for loan. © Tate, . brushes. Shellsol D solvent was used to clear the systems, followed by deionized water which was applied to clear possible LAS surfactant residues. Figure  shows a detail of the dirt removal process from the damaged, heavily soiled, face and neck area of Two Recumbent Nude Figures.

. CONCLUSIONS AND FURTHER WORK
The microemulsion systems described here are intended to provide practising conservators with new tools and options for the surface cleaning of artists' acrylic paints and other water-and solvent-sensitive paint surfaces. The WiO microemulsion formulations, especially, offer possibilities for exploiting positive aspects of water-based cleaning systems (good pick-up and dispersion of soils; control over the conductivity and pH; use of adjuvants such as chelates, etc.) while limiting, to a degree, the risks associated with exposure to aqueous cleaners. The preparations reported here introduce some new surfactant products to conservation, but these microemulsions are intentionally formulated around familiar ingredients, principally mineral spirits and proprietary grades thereof. They are easy to prepare, requiring simple weighing out of the ingredients and minimal shaking by hand. In addition to showing the ranges of proportions that form WiO microemulsions, OiW microemulsions, or other types of mixture, the phase diagrams presented here also provide a graphic framework that can guide the user in fine adjustments of the ingredient proportions in pursuit of optimal cleaning performance.
The preparations under consideration here all involve non-volatile surfactants, and all the usual concerns about such substances apply, foremost of which are the matters of clearance and of the possible risks associated with any non-cleared residues. Scientific studies (including X-ray photoelectron spectroscopy, X-ray absorption near-edge spectroscopy), and Time-of-Flight Secondary Ion Mass Spectrometry are presently under way which seek to assess the effectiveness of the clearance procedures that have been suggested for each of the three microemulsion system types presented, the main principle of which is dilution rinsing with the solvent of the continuous phase. Findings of those studies will be reported in due course (Ormsby et al. in press). One hypothetical advantage of the WiO microemulsions over conventional aqueous cleaning systems is reduced levels of swelling of the paint during exposure to the cleaning liquid; further studies are also envisaged that will examine the validity of this hypothesis in comparative studies of the swelling action on artists' acrylic paints of these classes of microemulsion compared to common aqueous cleaning solutions.
The new microemulsion preparations described here will continue to be evaluated from the practical point of view by the project members, through continuing systematic cleaning trials on simulated soiled artists' acrylic paint, through testing in relation to actual conservation treatments, and through inclusion in further iterations of the GCI/Tate CAPS professional development workshops. All of these vehicles for practical testing have provided useful feedback that has guided the technical and scientific development work, and they remain important ancillary activities for our group. The different modes of practical testing have already highlighted some lines of enquiry for further research activity beyond the clearance and swelling issues just noted, for example: determination of the tolerance, especially, of the Series  and Series  microemulsion systems to adjustment of the conditions (in terms of pH, conductivity, and chelate content) of the water phase; and viability of non-ionic alternatives to sodium dioctylsulfosuccinate in the Series -type systems. Additional microemulsion systems prepared from non-ionic surfactants and without higher alcohol co-surfactant may offer advantages in terms of clearance, reduced paint sensitivity, and the possibilities for fine-tuning cleaning action. We hope that this preliminary report on mineral spirits-based microemulsions provides the foundation for further research on and practice with this novel and interesting class of cleaning liquid in which two very dissimilar substances hydrocarbon solvent and water -are helped through mediation to achieve a compatible, stable relationship. works of art. We also gratefully acknowledge the assistance of Barbara Cressman, Jim DeFelippis and Carol Volz at The Dow Chemical Company for the analytical characterization of the commercial LAS products. APPENDIX The overall cleaning system ratings used to describe the relative magnitude of overall change to the sample surfaces, incorporating both removed dirt and any perceived alteration to the paint (as judged by eye), were assigned according to the following criteria:

ACKNOWLEDGMENTS
This scheme effectively divides cleaning systems into two classes in terms of their tendency to cause perceptible changes in the paint or not. Cleaning systems rated with even numbers - produced no perceived alteration to the paint; and within this class higher number ratings reflect greater degree of cleaning. Similarly, cleaning systems rated with odd numbers - produced some perceived alteration to the paint; and, again, within this class higher number ratings reflect greater degree of cleaning.
Ratings  or  were adopted for effects of cleaning systems that did not fall clearly into the two classes just mentioned.

NOTES
 The ECOSURF™ EH Non-ionic Surfactants (EH, EH, and EH) are ethoxylated/propoxylated branched alcohol (-ethyl hexanol) products varying essentially in their degree of ethoxylation/ propoxylation and hence HLB value. The ECOSURF™ Surfactant products emerged from earlier studies by the Dow/Tate/GCI collaboration (see Ormsby et al. ) as possible alternatives to alkyl phenol ethoxylate (APEO) non-ionic surfactants like TRITON TM X-. For technical information on the ECOSURF™ Surfactants, see http://www.dow.com/ products.  Ion chromatography, phosphorous NMR, and conductivity measurements of the four LAS sources indicated that the ionic content of the Sigma-Aldrich grade of LAS was twice that of the other LAS sources (see Table ). This material also had more total phosphorous content coupled with a higher ratio of pyro-phosphate relative to ortho-phosphate. In general anionic surfactant-based microemulsions can be very sensitive to inorganic electrolytes because the electrolytes can screen electrostatic interactions between the anionic head groups making them less soluble in water (Kizilbash et al. ). We believe that the higher ionic content of the Sigma-Aldrich material is preventing the formation of -phase microemulsions.  The product's MSDS sheet describes the solvent composition as: Solvent naphtha (petroleum) light aromatic naphtha CAS# -- % Medium aliphatic solvent naphtha (petroleum) CAS# -- % Heavy aromatic naphtha CAS# -- % -ethylhexanol CAS# -- % The presence of a moderate proportion of aromatic naphtha in the petroleum distillate solvent of TRITON™ GR-M Surfactant gives the product some odor.