Canvas Complexity | Conserving Canvas

Introduction

The woven structure of canvas, although modified by the
subsequent layers of the painting, still influences the
mechanical and hydromechanical behavior of this complex
composite. To highlight the issues of complexity, this paper
brings together and reevaluates collaborative research
conducted over twenty-five years that addresses these issues
in the context of the structural conservation of canvas
paintings. It discusses the methodologies of research related
to experimental work, practical conservation implementation,
and the modeling of canvas and its composite properties. The
key findings, with examples, and the methods used to obtain
useful data and practical insights are given, along with the
references for the experimental details and results.

Recent research in the fields of polymer mechanics,
fiber-reinforced composites, and smart materials that are
relevant to conserving canvases of the past (and future) is
highlighted, as are directions for future research—pure and
applied—that will aid in the structural conservation of
paintings on canvas.

Physical Structure

Types of Canvas

In the context of artists canvas and lining fabric, most
fabrics are made from natural materials. For painted cloth,
silk, calico, cotton, and linen have been the most common
fabrics both geographically and historically. However, other
natural materials—for instance, bark cloth (Lennard, Tamura, and Nesbitt 2017)—are part of rich traditions of painting, and we are only
beginning to understand their behavior as painting supports
(Smith, Holmes-Smith, and Lennard 2019).

The choice of canvas by artists is often made for pragmatic
reasons, including local production, availability, expense,
and size. Important features of these natural materials are
their flexibility, texture, and absorbency (see “Moisture
Response” below), which make them suitable for the application
of paint. For fabric supports, some of these features are a
product of the preparation, but most originate from the fiber
type, yarn structure, and weave structure (flexibility and
texture). It is reasonable to suggest that in all but the most
prestigious commissions the artist had little control, or even
interest, in these factors, but instead innately, or through
treatises and word of mouth, thought of a canvas as either
suitable or not suitable for the execution of a particular
painting.

The discussion below focuses on cotton and linen but
acknowledges the use of silk, both in traditional non-Western
and Western art and in contemporary contexts. In a few cases,
it has been possible to trace the evidence of artists choice
through letters, colormen archives, and anecdotes (Johnson et al. 2010), but rarely has the production of the canvas been under
their control. Historically, cloth was a commodity for
household furnishings, military and naval outfitting, and
ecclesiastic and court events. It is with the advent of
industrial weaving that different weights of cloth became more
accessible through artists suppliers, and the small-batch
production of artists canvas started to occur. However, there
is mounting evidence through technical examination and
conservation records that, prior to industrialization, artists
had access to different types and weaves, and in some cases
clearly chose some over others (Heydenreich et al. 2008;
Seccaroni 2012). Sources, including the National Portrait Gallery’s list of
British artists suppliers (1650–1950),
individual websites, and publications that include research
into the trade in artists materials (Kirby, Nash, and Cannon 2010), provide invaluable information from which to build up a
better understanding of the context in which artists choose
their materials.

Contemporary artists make both conscious and pragmatic choices
to use best-quality artists canvas (usually fine plain-weave
linen or cheaper cotton duck, which is also available in wider
widths), sacking (jute, usually plain weave), bank money bags
(cotton and linen, plain or twill weave), and mattress ticking
(an old favorite, usually linen with a herringbone twill
weave), as well as older, traditional materials used in a
contemporary context, such as bark cloth (Schneider 2021).

Artists use canvas in a variety of ways: unstretched, sewn,
pierced, or formed into three dimensions. Canvas is also woven
from polyester specifically for art and conservation. This is
manufactured/supplied by companies such as Fredrix in the
United States, Haywards in the United Kingdom, and Lascaux in
Germany.
In the performing arts, there is more painting on plastic
gauzes and projection cloths made from woven and nonwoven
synthetics, including polyester, nylon, and Kevlar. The range
of fabrics used and available is best appreciated by visiting
the website of J. D. McDougall, a company that has supplied
fabrics for over a hundred years.

Types of Weave

A woven fabric made from a single type of fiber (e.g., flax)
is a composite influenced by the way it is grown or
synthesized and then processed into yarn ready for weaving.
The various levels of hierarchy in the structure influence, to
differing degrees, the overall fabric behavior. The yarn type,
diameter, flatness, and stiffness affect flexibility and
moisture response. The density of fiber yarns and type of
weave influence yarn mobility, stiffness, drape, fracture
toughness, and permeability (Young and Jardine 2012). The fabric permeability affects moisture response,
consolidation, impregnation, and lining treatments (Young and Ackroyd 2001).

Additionally, surface or impregnating coatings (e.g., size,
paint, and consolidants) affect tensile, shear, bending
properties, yarn mobility, and fracture toughness. For
traditional Western paintings, we have a good experimental and
empirical understanding of the complex composite of woven
fabric and coatings. Research into the behavior of
double-sided painted cloth such as trade union banners (fig. 15.1) (Smith, Thompson, and Hermens 2016;
Sanchez Villavicencio, Young, and Thompson 2022) or the behavior of twill and more complex weaves, however,
is in its infancy (see “Reconstructing Weaves” below).
Certainly, more research is required to determine the dominant
contributing factors in the mechanical and hydromechanical
behavior of canvas when new materials and techniques are used
in contemporary art or used in mixed media and with new
synthetic artists materials such as water-mixable oil paint,
inks, or acrylic-primed spun polyester canvas (see
“Reconstructions for the Contemporary Use of Canvas” below).

Prediction of painted canvas properties from experimental
mechanical and chemical data is the first step. Testing on
single-constituent yarns, free films, single layers, and/or
simple multilayers provides invaluable information that
relates to both the present and past and to changes in
properties caused by natural aging and conservation
treatments. However, these are relatively simple interactions
and therefore only partially representative. Conservation
treatments themselves are a valuable source of “live”
empirical data of the real complex interactions, but they are
not easily reproducible. Nevertheless, many years of
experience add up to a powerful knowledge base of possible
outcomes.

Analytical or computational studies for fabric/canvas behavior
have developed models that can account for one level of
hierarchy—fiber level, yarn level, weave pattern level—or with
continuous, homogeneous layers with simplified superposition
of layers. Ideally, however, to model fabric permeability, for
example, it is necessary to consider both microporosity (fiber
spacing) and macroporosity (yarn spacing) (Zeng et al. 2014). Models based on geomechanical structures (soil and rock)
have been suggested as a useful approach to studying
consolidation (Michalski 2008). Certainly, many of the phenomena—such as crack networks,
diffusion, and complex interacting layers—are common to both
fields. Therefore, applying such models would be beneficial to
canvas conservation studies.

Similarly, to predict biaxial behavior of woven fabric, the
model of the weave needs to consider friction at the yarn
crossover. These more complex multilevel models are yet to be
fully developed (Aliabadi 2015). Past attempts to model the complex structure of canvas
paintings and fabrics have used a finite element analysis
(FEA) approach (Guanzhi et al. 2017;
Mecklenburg, McCormick-Goodhart, and Tumosa 1994). By using analytical models developed for tensioned fabric
structures, the problems normally encountered with accurately
modeling complex curved surfaces are mitigated. Bicubic-spline
models developed originally for architectural applications
(Brew and Lewis 2007) are the most representative way to model a woven stretched
canvas on a stretcher (Young 2013; see “Strain Distribution” below). However, the interaction
of the uppermost layers can be successfully modeled with FEA,
taking into account their viscoelastic properties (Tantideeravit et al. 2013).

Dealing with Complexity

In what ways is it possible to complement the existing
research? One approach is epidemiological studies of
collections and promotion of documentation protocols that
include canvas weave pattern and count, fiber type, and
duty/colorman stamps. This information not only helps in the
identification of the date and source of the canvas but also
is invaluable for understanding the artist’s intent and the
provenance of artists materials (Johnson et al. 2010;
Murillo-Fuentes and Alba 2018). It is also important for conservation, as the weave
structure still influences the overall response of a painting,
whatever its age. Combined with consistent materials
characterization, such information may allow further insights
to be gained into the mechanical and hydromechanical behavior
of real paintings.

While it is not always possible to be historically accurate at
every level of the canvas structure, weave, painting, and
lining reconstructions allow for repeatability and for endless
combinations to be experimentally tested and trends in
behavior to be established (Daly Hartin et al. 2011; also see “Moisture Response of Linings” and “Reconstructing
Weaves” below). Lined paintings, as well as modern and
contemporary use of “canvas”—wherein if a material exists, an
artist will use it, and unconventionally—require an
understanding of geometry, construction, and potentially the
properties of many different materials. Reconstructions play a
crucial role in understanding these elements.

Strain Distribution

Several factors make the strain distribution within canvas
complex: it is a woven rather than a continuous, homogeneous
layer; the strain response of the yarns is usually different
in the weft and warp; and factors including friction, density,
and weave pattern influence the strain distribution.
Additionally, on a strainer or stretcher, the attachments and
the stretcher construction (corners) induce uneven loading.
While the stress distribution that this loading creates cannot
be directly measured, the strain distribution can be measured
by techniques including electronic speckle pattern
interferometry (ESPI), digital image correlation (DIC), and
photogrammetry. ESPI can be used to obtain accurate
quantitative measurements of strain distributions of primed
canvas on a stretcher that replicate a real painting
configuration. Biaxial tensile properties of a painting and
its constituents can be obtained by mechanical testing. By
combining biaxial tensile testing with two-dimensional strain
mapping, however, it is possible to gain an understanding of
the composite behavior of a stretched canvas and the forces to
which it is subjected.

The biaxial restraint of the canvas alters the strain
distribution around the tacks or staples, becoming
progressively more complex toward the corners. At the macro
level, the strain patterns induced by the attachments are
similar, with closer spacing resulting in more even strain
distribution. If the attachments pass through preprimed
canvas, there is reduced local cusping because of the greater
resistance of uncracked primed canvas to distortion in the
bias direction of the canvas. The restraint imposed by tight
corner folds reduces the high load that would be imposed on
the attachments near the corners if a loosely folded corner
was keyed out. Nevertheless, shrinkage of the canvas or keying
out will lead to significant strain concentrations in these
areas. The strain irregularities become significantly less if
the canvas is attached on the rear face of the stretcher
rather than the side. Staples are effective attachments until
the canvas between the legs begins to slip; tears may then
occur because the staple leg creates very high strain
concentrations. Tacks appear to be as effective in restraining
the canvas and less likely to cause tears (Young and Hibberd 2000).

Bicubic-spline modeling (validated by ESPI) was the first
computational model of a painting to incorporate the
stretcher, staples, corner folds, and frictional forces. The
inclusion in the model of the measured coefficient of friction
of 0.63 for a pine stretcher bar showed that areas of high
strain move outward toward the edges of the stretcher.
Figures 15.2 and
15.3 compare the measured strain
obtained using ESPI with the modeled strain for the same
canvas properties and loading conditions. The close
correlation of the two distributions in terms of overall
magnitude and specific features is very good. This gives a
high level of confidence in using a bicubic-spline model to
predict modes of failure and improve upon the present methods
of tensioning canvas by simulating the strains induced in
canvas under different conditions (Brew, Lewis, and Young 2016).

ESPI has also been a useful tool for evaluating structural
conservation treatments, for instance, tear mending. A
painting will have high strains near the tear and strain
concentrations at the ends of the tear, which are sites for
potential propagation of the tear. If a tear is close to a
corner or a tack/staple, this situation will be exacerbated
because of the nonuniform loading. Implicit in restoring the
mechanical integrity of a painting is the requirement to
reestablish a uniform strain field across the painting—or at
least one whose average strain is commensurate (Young 2003). This can be seen for the Heiber (thread-by-thread)
tear-mend strain map shown in
figure 15.4 for an acrylic-primed
canvas under 50 N biaxial tension. Redistribution, reduction,
or eradication of strain concentrations (one color in the map)
implies a uniform strain field. Both the patch (fig. 15.5) and the Heiber mend demonstrate that this can be achieved
to some degree.

In both cases, residual strain concentrations are present. For
the Heiber—or any equivalent tear mend—reducing these
concentrations below these levels is very hard to achieve by
visual inspection alone. A “perfect” mend would eliminate
strain concentrations around the original fracture site,
preventing potential propagation of the tear. Similarly, any
patch should have high fracture toughness and minimal
stiffness. The patch strain map (see
fig. 15.5) shows that the level of
strain concentrations has been reduced, but small
discontinuities in strain occur at the edges of the patch.

Patches, which impart additional flexural stiffness with the
aim of keeping the tear flat, are likely to result in an area
of lower strain across the patch, but also larger
discontinuities. The adhesive and type of adhesive interface
will be the major factors in determining whether a tear mend
is strong enough to withstand “normal” stress distributions
within the canvas. The onset of failure will be evident as an
increase in strain concentrations while loading.

Apart from assessing which techniques are mechanically most
successful, future research needs to look at how the materials
used in the subsequent layers of fill and retouching alter the
balance of forces in and around the tear. This includes
looking at interfacial tensions built up by the drying of
fills and coatings (Daly Hartin et al. 2011), as well as the ability of the filled mend to withstand
fatigue caused by cycling of temperature and relative humidity
(RH) (Young 2013).

Strip-lining is another structural treatment perceived as
minimally invasive that aims to reinstate, as far as possible,
the structural integrity of the painting. The various
configurations and methods to prevent a sharp change in
stiffness at the edge of a strip in the picture plane have
been evaluated by ESPI. For example, strip-lining with Beva
371 and polyester sailcloth (00169, manufactured solely to
order by Richard Hayward & Co., United Kingdom)—with
pinked edges to prevent a hard transition—actually results in
strain concentrations within the picture plane at the point of
the pinked triangle of polyester (Brew, Lewis, and Young 2016). As expected, under the same loading conditions a feathered
edge transition results in lower strain concentrations.
However, if too much adhesive is used, the stiffness of the
adhesive (even Beva 371) dominates, and a strain concentration
along the edge of the feathering occurs.

ESPI is a very sensitive technique and does not always work
well for real paintings, especially those with glossy
varnishes or that are in situ where exterior vibrations occur.
For the application of strip-linings and other structural
conservation treatments, DIC with a relatively inexpensive,
fast-frame-rate, high-resolution camera can produce good
results. Such systems are standard in engineering and offer a
complementary technique for nondestructive evaluation of
structural conservation treatments.

Moisture Response

The literature on the moisture response of canvas paintings
and the key findings are covered in On Canvas (Hackney 2020). For emphasis, some specific points from some of the
published literature follow.

Moisture Response of Original Supports

Most of the data in the literature on moisture response relate
to uniaxial testing. The results, however, can be misleading,
as uniaxially the canvas is unconstrained in one direction,
and this is not representative of the stresses that build up
under biaxial constraint on a stretcher. Nonetheless, careful
experimental design and interpretation can mitigate this
difference. One of the most valuable resources for measuring
moisture response has come from deaccessioned paintings and
from nineteenth-century primed loose-linings.
Figure 15.6 shows the typical load
response in the weft and warp direction for a primed
loose-lining produced by Roberson colormen (Carlyle, Young, and Jardine 2008). The tension in the two directions drops until an inversion
occurs at 70% RH, where the tension starts to rise as RH
increases.

This pattern occurs for many oil-primed canvases
(loose-linings), as was demonstrated by Hedley under uniaxial
tension (Hedley 1988). The initial drop in tension is attributed to the size
layer becoming softer as it absorbs moisture until it reaches
a gelatinous state. Simultaneously, the fibers in the canvas
are absorbing moisture. At some point, the swollen fibers
cause the canvas to contract, and because it is tacked in
place the tension rises. Typically, the tension in the weft
direction increases significantly more than the warp because
it has less crimp. The major influences on where this
inversion occurs are weave density and the glue-size
application. Inversions have been measured from between 65%
and 85% RH for nineteenth-century English commercially
glue-sized oil-primed canvases and glue-sized new canvas.
Figure 15.7 shows the response of
another Roberson primed loose-lining with the inversion at 65%
RH (Carlyle, Young, and Jardine 2008;
Carr et al. 2003).

Moisture Response of Linings

A database of moisture response for archival canvas,
reconstructions, and new types of canvas is useful in deciding
on a moisture treatment (how long and at how much moisture),
especially when the complexity of two canvases is involved, as
is the case with lined paintings. As a first approximation,
one can think of the problem as a superposition of two
canvases, and hence two moisture responses that induce
expansion and contraction, leading to stresses within the
canvases and induced stress distributions in all the layers.
This can be modeled with appropriate boundary conditions using
FEA and/or analytical models—if the properties of each layer
are known. However, there is often added complexity because
the lining adhesive (especially traditional glue-paste and
wax) has impregnated the original canvas, cracks, and
interfaces of the painting. Nevertheless, it is possible to
see trends in behavior when there is sufficient archival
material to characterize degradation and mechanical properties
(Young and Ackroyd 2001).

Moisture Response of Modern Materials

The use of synthetic/modern materials in painted textiles can
be traced to the patents for textile coatings of the
nineteenth century (Young 2012). Research into their properties began to have a direct
influence of practice with the work of Hedley and Hackney in
the 1980s (Hackney 2020). Artists’ use of new materials, plus the conservator’s
desire to find a suitable lining canvas, means we need to
continue to characterize a wide selection of natural and
synthetic fabrics. It is insufficient to characterize only the
type of canvas (linen or polyester), because even for
synthetic fabrics the yarn processing, fabric weave density,
and coatings (e.g., fire retardants) influence moisture
response (Young and Jardine 2012). More disconcerting for those choosing materials for
treatments is the generic naming of canvases; for example, “12
oz Belgian linen” is a name, not a description; it does not
even mean it comes from Belgium. This is confusing and
misleading if one assumes certain properties are dependent on
the material’s source country. Similarly, cotton duck data
produced over twenty years ago (Young 1996) will generally be valid today, but changes to the
source—and therefore the twist of the yarn, sizing, tension
during weaving, and subsequent regulatory coatings and
processes—can change the hydromechanical response.

Of course, this has always been the case. For instance, in the
eighteenth century, weavers in the east of Scotland branded
their linen fabrics “Osnaburg” (or Osnabrigg) an imitation of
Osnaburgh (also known as Osnabrück) (Young 2012). Similarly, Lascaux P360 polyester (a linen look-alike) has
changed properties since it was introduced (Young and Jardine 2012). Ideally, one should characterize (or at least empirically
test) each new batch if one cannot guarantee its response.

Reconstructing Weaves

Reconstructions of painted textiles in general—starting at the
yarn level and proceeding through weaving, stretching,
preparation layers, and subsequent artists
materials/techniques relevant to the work—are invaluable for
understanding how the manufacturing process, subsequent
preparations, and materials influence the aesthetic,
kinesthetic, and physiochemical behavior of the painting.
Fraught with uncertainty as to how authentic or historically
accurate they may be, reconstructions still allow one to
explore which properties have the biggest influence on
behavior by trying different variations and through repeated
testing.

Reconstruction of canvas weaves is a relatively new approach.
It has come about in part from a greater awareness of and
interest in the weave’s significance in the provenance,
interpretation, and conservation of a painting. For example,
during the conservation treatment at the Cleveland Museum of
Art of The Crucifixion of Saint Andrew (1606–7) by
Caravaggio, it was possible to retrieve some information about
the original canvas from the X-radiograph, even though it had
been lined with a plain-weave canvas.
This painting may have been cut down, which would mean less
cusping would be visible. Inferences drawn from the cusping
had been extrapolated from simple plain-weave canvases, rather
than the complex weave of Caravaggio’s original canvas.

While the clarity of the original weave was hard to discern, a
trained eye (in this case, that of Dr. Dan Coughlan, curator
and master weaver at Paisley Museum, Scotland) identified the
weave as a huckaback, which is a plain weave with a floating
warp. Hence, he was able to set up the weaving pattern for a
traditional four-frame hand loom. Linen and jute yarn samples
were sourced from Jos Vanneste, a Belgian linen company. The
closest match to a yarn fragment from the painting was a linen
yarn, which was then woven into a canvas by a local Scottish
weaver. Empirical testing found that the canvas was more
stable on the bias than an equivalent plain weave without a
floating warp, and it developed less cusping when tacked onto
a stretcher.

The canvas was also prepared with a traditional double ground
layer, and, interestingly, drying cracks within these layers
were of a very similar size and pattern (Brunton 2018). A more systematic and controlled set of tests is now being
performed on these samples using the biaxial tensile tester in
the Conservation Research Laboratory at the Kelvin Centre,
Glasgow University. Such reconstructions are useful not only
for understanding the complexity of canvas but also for
possible use as fabrics for structural treatments. Having
control over the weaving process allows for a bespoke pattern
(see
Loermans’s poster
in these proceedings) and, combined with testing, the ability
to tune its properties.

Reconstructions for the Contemporary Use of Canvas

The value of reconstructions for conservation/preventive
issues related to contemporary works as well as older works
cannot be underestimated. Either by working directly with the
artists, gallery, artists studio, fabricators, or documented
interviews, much can be gleaned, if not always volunteered,
that enable the use of the same materials and technique in the
reconstruction. This approach has been used in cleaning (Barker and Ormsby 2015;
Krueger 2017;
Diamond et al. 2019), very effectively for tear mending (Piotrowska and Amann 2009), and assessment of structural stability and preventive
treatments (Griffin, Young, and Hale 2014).

One example is a series of acrylic preprimed linen and
polyester canvases on which acrylic paint and screen-print ink
had been applied either directly or through a screen. These
works were unstretched on arrival for a display of the
artist’s work at a gallery in London. Several features were
considered undesirable: a general buckling of the canvas with
sharp cupping in certain areas, cracking in the upper paint
layers, and some buckling remaining in the works, once
stretched, with an overall lack of tension. By matching the
canvases as closely as possible and applying the inks used by
the artist, it was possible to conduct a series of tests on
the biaxial tensile tester to re-create the phenomenon and to
understand its cause.
This allowed recommendation of an appropriate treatment and
possible ways to mitigate the effects in the future. It was
found that the liquid phase of inks caused local shrinkage of
linen. In the synthetic canvases, the liquid phase interacted
with the acrylic priming, softening it, unlocking the woven
polyester yarns, and allowing distortions to occur.

Possibly more challenging to conservators of the future is the
embedding of materials within canvas (Nahum, McGuirk, and Watson 2019) and haptics (Bianchi 2016).
The testing of contemporary materials that an artist might use
as a “canvas” or in a sculpture/installation or that might be
a suitable alternative to traditional lining materials should
be an ongoing, proactive area of research within conservation.
While attempts have been made to work with manufacturers to
produce materials to our specifications, commercial production
of bespoke lining fabric is not viable. A proactive approach
is required to understand textiles manufactured for other
industries and to specify/design fabrics, as well as an
investment in our profession if we are to drive research and
production.

Future directions for the conservation of canvas could
include:

• Designing, fabricating, and assessing canvas for artists
  and conservators

• Collecting more data on material properties to make
  available through master classes and online resources,
  including how to interpret the data

• Devising better simple, studio-based evaluation tests
  before and during treatment or when using new materials

Future Options: Complex Canvas Complexity—Good Candidates for
Artists Canvas and Lining Fabrics

The largest drivers of the development of new materials come
from the aerospace, military, and apparel industries. The fact
that woven fabrics are part of contemporary composite
engineering materials attests to their success in increasing
the fracture toughness of structures. Fracture toughness
prevents cracks (tears) from propagating: in degraded canvas,
it is the brittleness of the yarns due to chain scission and
increased crystallinity of the cellulose (for flax and linen)
that substantially reduces the fracture toughness. Cotton duck
has a much lower fracture toughness even when new because of
its short staple length. Interestingly, natural materials,
including linen, are still part of active research into
improving the service life of structural composites, as they
have properties yet to be fully replicated by other methods
(Pandian and Jailani 2019).

Nonwoven fabrics have been used by both artists and
conservators, typically as part of collage pieces or, in
conservation applications, as interleaves in linings. Such
fabrics can be made to be homogeneous and heterogeneous, so it
would seem their relative lack of use is due to the lack of
texture, drape (ability to bend and form shapes, e.g., around
a stretcher bar corner), and “responsiveness” when painting.
However, the low absorbency of many synthetic materials may be
considered a good property for linings.

State-of-the-art fabric structures use surface modifications
to give the desired properties, such as hydrophobicity using
vapor deposition on fabric (Xu et al. 2019). Filled yarns have been developed to increase stiffness and
fracture toughness (Gilchrist, Svensson, and Shishoo 1998). Triaxial weaves, which have two sets of warps at 60
degrees to the weft, give increased stability and stiffness on
the bias and better fracture toughness (Wang et al. 2018).

Embedded sensors in fabrics have been the subject of
experimentation for over twenty years and offer the
possibility of in situ monitoring of canvas properties:
moisture content via electrical resistance and induced strain
via fiber optics (Zawadzki et al. 2012). The problem from a conservation/preventive point of view
is that embedded optical devices considerably stiffen the
fabric. However, with the present development of haptics and
wearable sensors in the military, sports, and gaming
industries (Muhammad Sayem et al. 2020), it is likely that the technology will evolve to enable
development of embedded sensors for structural conservation
and the monitoring of canvas complexity. Shape memory sensors
may also offer an additional option. For example,
self-regulating structures that respond to environmental
conditions could be woven into fabric (Ibrahim et al. 2010)—and at the least they are bound to be part of future
artworks.

Driven by the need to reduce our carbon footprint,
energy-harvesting fabrics are being developed that convert
ambient energy into electrical energy. These include
dye-sensitized solar cells fabricated into functionalized
yarns and made into films that can be spray-coated onto
textiles (Torah et al. 2018), as well as screen-printable polymer film and polymer
fibers that can harvest mechanical energy from textiles. Maybe
the canvas paintings of the future could provide their own
“active” microclimate.

Acknowledgments

I would like to thank Tim Green, retired paintings
conservator, Tate; Dean Yoder, senior conservator, Cleveland
Museum of Art; Dr. Dan Coughlan, curator and master weaver,
Paisley Museum; Jean Mabon, for her expertise and for weaving
the Cleveland canvas reconstructions; Jennifer Brunton,
technical art history student, Glasgow University; Janey
Zagreb (London); and my colleagues at the Kelvin Centre,
University of Glasgow.

Notes