Content uploaded by Irena Kralj Cigić
Author content
All content in this area was uploaded by Irena Kralj Cigić on Nov 22, 2017
Content may be subject to copyright.
Material Degradomics: On the Smell of Old Books
Matija Strlicˇ,*
,†
Jacob Thomas,
‡
Tanja Trafela,
§
Linda Cse´ falvayova´,
†
Irena Kralj Cigic´,
§
Jana Kolar,
|
and May Cassar
†
Centre for Sustainable Heritage, The Bartlett School of Graduate Studies, University College London, Gower Street
(Torrington Place site), London, United Kingdom WCIE 6BT, Tate, Millbank, London, United Kingdom SW1P 4RG,
Faculty of Chemistry and Chemical Technology, University of Ljubljana, Asˇkercˇeva 5, Ljubljana, Slovenia SI-1000, and
Morana RTD d.o.o., Oslica 1b, Ivancˇna Gorica, Slovenia SI-1295
We successfully transferred and applied -omics concepts
to the study of material degradation, in particular historic
paper. The main volatile degradation products of paper,
constituting the particular “smell of old books”, were
determined using headspace analysis after a 24 h pre-
degradation procedure. Using supervised and unsuper-
vised methods of multivariate data analysis, we were able
to quantitatively correlate volatile degradation products
with properties important for the preservation of historic
paper: rosin, lignin and carbonyl group content, degree
of polymerization of cellulose, and paper acidity. On the
basis of volatile degradic footprinting, we identified deg-
radation markers for rosin and lignin in paper and
investigated their effect on degradation. Apart from the
known volatile paper degradation products acetic acid and
furfural, we also put forward a number of other com-
pounds of potential interest, most notably lipid peroxida-
tion products. The nondestructive approach can be used
for rapid identification of degraded historic objects on the
basis of the volatile degradation products emitted by
degrading paper.
The aroma of an old book is familiar to every user of a
traditional library. A combination of grassy notes with a tang of
acids and a hint of vanilla over an underlying mustiness, this
unmistakable smell is as much part of the book as its contents. It
is a result of the several hundred identified volatile and semiv-
olatile organic compounds (VOCs) off-gassing from paper and
the object in general.
1,2
The particular blend of compounds is a
result of a network of degradation pathways and is dependent on
the original composition of the object including paper substrate,
applied media, and binding (the sum of its “biography” including
possible interventions and its past and present local environment).
Recently, there has been an increased interest in the “smell
of old books” and in VOCs emitted from historic paper in
general.
2-5
The food and pharmaceuticals packaging industries
have studied the transfer of taint and odor from paper and
cardboard packaging during shipping and storage, extensively.
6,7
Other studies have developed VOC screening methods to identify
the fraction of recycled pulp in board stock,
8,9
and identification
of volatiles is now an almost routine analytical challenge. However,
heritage institutions (libraries, archives, and museums) are
interested in quantitative VOC analysis as a rapid diagnostic tool
for the degradation and condition of their collections as well as
evaluation of conservation treatments and materials analysis. Can
we identify degraded books by their smell or even extract
information about why the books degraded?
This question is especially interesting as heritage objects
present particular problems for analysis either due to their
uniqueness or due to diverging histories. Another factor is that it
is often impossible to sample. This necessitates nondestructive/
noninvasive methods, and headspace sampling coupled to GC/
MS is especially appropriate.
The uniqueness and complexity, as well as limitations of
destructive sampling, have limited much of heritage research to
single object technical/material studies or to the degradation of
constituent materials and/or simple surrogate objects. Yet models
based on simple systems fall short of describing the complexity
of heritage objects, and single object studies do not answer larger
questions about classes of objects. The complexity of heritage
objects is to an extent comparable to the complexity of living
organisms. “-Omics” type methodologies have been developed for
the study of living organisms,
10
could these methodologies be
transferred to the study of heritage objects?
In this regard, the concepts of metabolomics
11-14
are most
applicable to the purposes of heritage science. Metabolism, as a
process in living organisms that have the power to adapt and
renew, could be seen in parallel with degradation if one treats
* To whom correspondence should be addressed. E-mail: m.strlic@ucl.ac.uk.
†
University College London.
‡
Tate.
§
University of Ljubljana.
|
Morana RTD d.o.o.
(1) Donetzhuber, A; Johansson, B.; Johansson, K.; Lovgren, M.; Sarin, E. Nord.
Pulp Pap. Res. J. 1999,14 (1), 48–60
.
(2) Lattuati-Derieux, A.; Bonnassies-Termes, S.; Lave´drine, B. J. Chromatogr.,
A2004,1026, 9–18
.
(3) Strlicˇ, M.; Kralj Cigic´, I.; Kolar, J.; de Bruin, G.; Pihlar, B. Sensors 2007,
7, 3136–3145
.
(4) Lattuati-Derieux, A.; Bonnassies-Termes, S.; Lave´drine, B. J. Cult. Herit.
2006,7, 123–133
.
(5) Gibson, L. T.; Robertson, C. Advances in Paper Conservation Research
Conference, March 23-24, 2009, British Library: London, 2009, pp 40-
42.
(6) Isinay, E.; Yu¨zay, S.; Pack, S. Technol. Sci. 2007,20 (2), 99–112
.
(7) Forsgren, G.; Frisell, H.; Ericsson, B. Nord. Pulp Pap. Res. J. 1999,14
(1), 5–16
.
(8) Ziegleder, G. Packag. Technol. Sci. 2001,14 (4), 131–136
.
(9) Asensio, E.; Nerin, C. Packag. Technol. Sci. 2009,22 (6), 311–322
.
(10) Goodacre, R.; Roberts, L.; Ellis, D.; Thorogood, D.; Reader, S.; Ougham,
H.; King, I. Metabolomics 2007,3(4), 489–501
.
Anal. Chem. 2009, 81, 8617–8622
10.1021/ac9016049 CCC: $40.75 2009 American Chemical Society 8617Analytical Chemistry, Vol. 81, No. 20, October 15, 2009
Published on Web 09/17/2009
material objects as pseudo-organisms. Immediate parallels can be
drawn between the degradome (the sum of the products of various
degradation processes) and the metabolome of an organism.
Through similarities with metabolomics,
15
we propose to define
a new field of material degradomics (and related terms, Table 1).
In the case of VOC analysis, the technique of footprinting
11,15
is
especially applicable, as degradic footprinting samples the local
environment of the object rather than the object itself. Other
parallels can be sought between the genotype of an organism and
historic recipes/procedures used to make a historic material/
object. The phenotype of an organism, which results from
expression of its genes and environmental factors, can be seen
as analogous to the measurable or observable properties of a
historic object. In this way, quasi-genotype-phenotype mapping
experiments could be performed for heritage objects, either
through systematically altering recipes (suppression, overexpres-
sion, knockout, etc.) and/or through experiments with altered
environmental sequences to determine how these change the
appearance and/or permanence of historically informed replicas.
The -omics approach, i.e., the data-driven holistic and inductive
approach to experimentation replacing the hypothesis-driven,
reductionist cycle of knowledge, has not been explored in relation
to material characterization yet. The need for large sample sets
(see, e.g., ASTM E1655-05) to generate a statistically significant
sample space is undoubtedly a drawback of this approach. Material
degradomics differs from other -omics-type research in that it is
not possible to clone an object or grow new ones in a controlled
environment. In relation to historic objects, it is often possible to
collect expendable ones, and if the research question is related
to variations in composition, it is possible to combinatorially
prepare historically informed samples to study how small varia-
tions in composition affect the degradome.
In this work, we present a pilot study using well-characterized
historic paper samples. The most abundant VOCs emitted from
the samples were quantitatively determined using a footprinting
method, and the resulting data were used to identify degradation
markers for several types of paper commonly found in archival
and library collections. The identified markers were then used to
discriminate between papers of different stability. We investigated
the properties important for the preservation of historic paper:
ash, rosin, protein, lignin and carbonyl group content, degree of
polymerization of cellulose, and paper acidity. While rosin and
protein content provide information on the production technology,
it is well-known that lignin-containing and particularly acidic papers
are particularly unstable, which leads to a rapid decrease of the
degree of polymerization and oxidation, resulting in a high content
of carbonyl groups. In the present study, we were able to identify
volatile markers, which could be applied to studies of actual books
or documents to be used for monitoring degradation of a collection
in real time, based on the off-gassed VOCs.
MATERIALS AND METHODS
Samples. We investigated 72 well-characterized 19th and 20th
century historic papers from the SurveNIR reference sample
collection.
16
The samples were selected on the basis of their
composition to cover the span of important variability, i.e., the
range of measured properties (Table 2). The sample set consisted
of surface (gelatin) sized and rosin sized paper; bleached pulp,
groundwood, and rag fiber containing papers; and coated and
uncoated papers, representing the most important technologies
of paper production. They were chemically characterized for lignin
content and reducing carbonyl group content (using UV-vis
spectrophotometry), rosin content (using LC-MS/MS), ash con-
tent (using dry ashing and gravimmetry), pH (using potentiom-
etry), degree of polymerization (DP, using viscometry) and protein
content (using HPLC). All methods have been described in a
previous publication in detail.
17
Not all samples were analyzed
for all properties: determination of DP using viscometry is not
possible for papers with a high content of lignin while samples
with rosin content were not analyzed for protein content, as papers
were sized either using rosin or gelatin.
To characterize VOC emissions, we used a small amount of
paper sample (10 mg) to avoid saturation of the solid phase
(11) Kell, D. B.; Brown, M.; Davey, H. M.; Dunn, W. B.; Spasic, I.; Oliver, S. G.
Nat. Rev. Microbiol. 2005,3(7), 557–565
.
(12) Brown, M.; Dunn, W. B.; Ellis, D. I.; Goodacre, R.; Handl, J.; Knowles, J. D.;
O’Hagan, S.; Spasic´ , I.; Kell, D. B. Metabolomics 2005,1(1), 39–51
.
(13) Hollywood, K.; Brison, D. R.; Goodacre, R. Proteomics 2006,6(17), 4716–
4723
.
(14) Goodacre, R.; Vaidyanathan, S.; Dunn, W. B.; Harrigan, G. G.; Kell, D. B.
Trends Biotechnol. 2004,22 (5), 245–252
.
(15) Mapelli, V.; Olsson, L.; Nielsen, J. Trends Biotechnol. 2008,26 (9), 490–
497
.
(16) SurveNIR project webpage, www.science4heritage.org/survenir/ (accessed
April 9, 2009).
(17) Trafela, T.; Strlicˇ, M.; Kolar, J.; Lichtblau, D. A.; Anders, M.; Pucko
Mencigar, D.; Pihlar, B. Anal. Chem. 2007,79, 6319–6323
.
Table 1. Definitions of Material Degradomics Terms
material degradomics the analysis, both quantitative and
qualitative, of the degradation
products produced by a material,
their dynamics, composition,
interactions, and variations due to
composition, degradation processes,
use, and environment
degradome the sum of the products of the
various degradation reactions and
their interactions
degradic profiling quantitative analysis of a set of
degradation products in a selected
degradation pathway or a specific
class of compounds for a particular
object
degradic fingerprinting unbiased, global screening to classify
samples based on patterns of
degradation products or
“fingerprints” that change in
response to degradation, use,
storage conditions, or conservation
interventions
degradic footprinting fingerprinting analysis of degradation
products in the microenvironment
surrounding the material
Table 2. PLS Cross-Validation Data
cross validation
measured property range RMSECV R2
rosin 0-7 mg/g 1.2 mg/g 0.463
lignin 0-350 mg/g 75 mg/g 0.400
carbonyl group content 0-0.11 µmol/g 0.02 µmol/g 0.380
degree of polymerization 0-3000 580 0.352
pH 3.5-8.5 1.2 0.251
8618 Analytical Chemistry, Vol. 81, No. 20, October 15, 2009
microextraction (SPME) fiber, which could lead to distortion of
the chemometric results. The sample was inserted into a 20 mL
headspace vial and predegraded at 80 °C for 24 h. Subsequently,
the resulting VOCs were extracted for1hatroom temperature,
using SPME fibers (Supelco, Bellefonte, PA) with a divinylben-
zene/carboxen/polydimethylsiloxane (DVB/CAR/PDMS) station-
ary phase and thickness of 50/30 µm, and the VOCs were then
subjected to GC/MS analysis. As it is well-known and also recently
demonstrated in the case of VOCs from paper,
4
the affinity of the
SPME fibers toward various volatiles depends on their composi-
tion. For this reason, we opted for the most general one, i.e., DVB/
CAR/PDMS.
GC/MS. An Agilent Technologies 7890A gas chromatograph,
coupled to an Agilent Technologies 5975C quadrupole mass
spectrometer equipped with a Gerstel cooled injection system CIS
4 was used at 250 °C. A 60 m Restek RTX-20 column, I.D. 0.25
mm and 1 µm stationary phase thickness (Restek, Bellefonte, PA)
was used. The mobile phase used was helium (99.999%, Messer,
Frankfurt, Germany) at a flow of ∼0.9 mL min
-1. We used the
following oven temperature program: 1 min at 40 °C, followed
by heating to 280 °C at the rate of 10.0 °C min-1, after which
the temperature was constant for 40 min.
Ionization was performed using standard EI mode applying
70 eV at 230 °C. The interface was heated to 270 °C, and the
quadrupole mass analyzer was heated to 150 °C. Manual integra-
tion of chromatographic peaks was performed in order to obtain
best quality data. Peak identification was performed on the basis
of comparison of mass spectra with the NIST library, and
quantification was performed using peak areas in SIM mode. The
15 most abundant VOCs present in all chromatograms were
selected for further analyses: acetic acid, benzaldehyde, 2,3-
butanedione, butanol, decanal, 2,3-dihydrofuran, 2-ethylhexanol,
furfural, hexadecane, hexanal, methoxyphenyloxime, nonanal,
octanal, pentanal, and undecane.
Data Analysis. The peak areas were extracted, and a data
table containing all peak areas and chemical information on the
samples (pH, carbonyl group content, lignin content, protein
content, degree of polymerization) was constructed.
Using partial least-squares regression (PLS), multivariate
models that relate the VOC peak areas (measured variables) to
the chemical parameters (quantifiable properties) of the samples
were developed and validated using a leave-one-out cross-validation
procedure. The optimal number of volatile compounds used for
multivariate analyses was determined on the basis of the quality
of PLS calibrations. Martens’ uncertainty test was used to calculate
the uncertainty of regression coefficients and to identify and
eliminate variables (volatile compounds) that did not contain any
relevant information.
18,19
For each model, the RMSECV (root
mean square error of cross validation) was calculated and
expressed in the same units as the original values in order to
estimate the prediction capabilities of the models.
20
As a method of unsupervised classification, principal compo-
nent analysis (PCA) was used to summarize the main variations
in variables and project them onto a few new principal components
(PC). In order to prevent spurious correlations from being
interpreted as meaningful information, leave-one-out cross valida-
tion was used to assess the model complexity. PLS and PCA
multivariate calibrations were performed using Unscrambler v.9.7
(CAMO, Trondheim, Norway).
RESULTS AND DISCUSSION
Depending on the production technology and storage environ-
ment, paper can be a particularly durable organic material, its
lifetime being measured in millennia.
21
However, most paper
produced between 1850 and 1990 is likely not to survive more
than a century or two due to the inherent acidity autocatalyzing
its degradation. Cellulose is the most important structural element
of paper, and it is well-known that the rate of its degradation
depends on its immediate macromolecular and general environ-
ment. Historic paper can be an extremely diverse material: a
number of production techniques and raw materials have been
used throughout history.
22
Most of the research on paper
degradation has so far focused on only a handful of different paper
types. However, in order to holistically understand paper degrada-
tion, we have to take into account all parameters of paper
composition and production.
The availability of a large number of samples, preferably
historic samples, is a prerequisite for the degradomics approach.
We used a well-characterized historic paper sample collection
characterized for various properties: degree of polymerization of
cellulose, lignin content, gelatin content, pH, aluminum content,
ash content, fiber composition, rosin content, and reducing group
content.
16
PLS is of particular interest (Figure 1), as the loading weights
express how the information in each measured variable, i.e., VOC,
relates to the variation in the quantifiable property (Y), i.e., the
chemical or physical property in question. The loading weights
are normalized so that their weight can be interpreted as well as
their direction. The measured variables with bigger loading
weights are more important for the prediction of Y and, thus,
represent suitable volatile degradation markers. A causal relation-
ship (degradation pathway) should link the degradation marker
with the property Y. In Table 2, the PLS calibration and validation
data are provided, and in Figure 1, an example of the obtained
PLS calibration and validation is given for rosin content in historic
paper. The calibrations are of a relatively low quality due to the
averaging effect of PLS; however, the predictions are still valuable
as it is possible to extract qualitative data from the loading weight
graphs, which enable us to interpret the effect that individual paper
components have on its stability (Figure 1).
The loading weight graph for rosin content is particularly
interesting. Rosin is 90% a mixture of related diterpene acids,
mostly abietic and dehydroabietic acid, the other 10% being
represented by “neutral substances”, which are dominated by fatty
acid esters and hydrocarbons formed from sterols.
23
The purpose
of the addition of rosin was to make paper fibers hydrophobic in
order to make sheets writable. Rosin sizing (as opposed to gelatin
(18) Martens, H.; Martens, M. Food Qual. Prefer. 2000,11, 5–16
.
(19) Martens, H.; Hoy, M.; Westad, F.; Folkenberg, B.; Martens, M. Chemom.
Intell. Lab. Syst. 2001,58, 151–170
.
(20) Martens, H.; Naes, T. Multivariate calibration; John Wiley and Sons: New
York, 1989.
(21) Strlicˇ , M.; Kolar, J. Ageing and stabilisation of paper; National and University
Library: Ljubljana, 2005.
(22) Hunter, D. Papermaking: History and Technique of an Ancient Craft; Dover
Publishing: New York, 1987.
(23) Roberts, J. C. The Chemistry of Paper; The Royal Society of Chemistry:
Cambridge, U.K., 1996; pp 126-127.
8619Analytical Chemistry, Vol. 81, No. 20, October 15, 2009
sizing used in earlier papers) was used mainly after ∼1850, and
aluminum sulfate, used to precipitate rosin on fibers, is thought
to be the source of acidity leading to low stability of rosin-sized
papers. The volatile degradation compounds significant for predic-
tion of rosin are various aldehydes, ketones, and 2-ethylhexanol.
Aliphatic aldehydes and ketones are known to form during
autoxidation of fatty acids
24
and are indicative of oxidative stress.
25
The main PC axis is PC2 and would seem to express polarity/
nonpolarity.
Lignin is a complex aromatic cross-linked 3D polymer, and its
content is of major interest in preservation of historic paper-based
materials, as it has a pronounced effect on their stability.
26,27
It is
well-known that furfural is a product of cellulose and hemicellulose
degradation
28
in a series of reactions from acid-catalyzed hydroly-
sis of glycosidic bonds to dehydration of the resulting simple
sugars. A higher acidity content of paper results in more abundant
furfural formation, as recently shown.
3
Despite the number of
VOCs off-gassing from lignin-containing food packaging materials,
1,29
the loading weight plot is dominated by acetic acid, hexanal, and
also furfural, which demonstrate the high impact lignin has on
the overall hydrolytic and thermooxidative instability of cellulose,
in darkness.
(24) Spiteller, G.; Kern, W.; Spiteller, P. J Chromatogr., A 1999,843, 29–98
.
(25) Dalle-Donne, I.; Rossi, R.; Giustarini, D.; Milzani, A.; Colombo, R. Clin. Chim.
Acta 2003,329, 23–38
.
(26) Be´gin, P.; Deschaˆtelets, S.; Grattan, D.; Gurnagul, N.; Iraci, J.; Kaminska,
E.; Woods, D.; Zou, X. Restaurator 1998,19 (3), 135–154
.
(27) Jin, F.; Cao, J.; Zhou, Z.; Moriya, T.; Enomoto, H. Chem. Lett. 2004,33
(7), 910–911
.
(28) Ziegleder, G. Packag. Technol. Sci. 1998,11 (5), 231–239
.
(29) Nevell, T. P.; Zeronian, S. H. Cellulose Chemistry and Its Applications; Ellis
Horwood: Chichester, 1985.
Figure 1. (Top left) PLS cross validation for rosin content in paper on the basis of six volatile degradation products, with loading weights (top
right). Other plots represent loading weight graphs for PLS calibrations for lignin and carbonyl group contents, degree of polymerization, and
pH, as indicated.
8620 Analytical Chemistry, Vol. 81, No. 20, October 15, 2009
The loading weight plot for PLS of carbonyl content shows
that an elevated content is a consequence of several factors: not
only of degradation of the cellulosic components (furfural, acetic
acid) but also of the rosin components (2-ethylhexanol and
hexadecane). The content of rosin can, thus, be correlated with
cellulose degradation, probably due to higher acidity, and there-
fore cross-correlated with the production of furfural, which is itself
both a degradation product and a reducing aldehyde compound.
It appears, however, that rosin content has a strong effect on
accumulation of carbonyl groups.
DP is a parameter associated with the quality of cellulosic fibers
and the extent of their degradation. However, considering that
the initial DP values could have been very different, it is
remarkable that a PLS correlation was obtained at all. Four
compounds are negatively correlated with PC1, which represents
the DP: acetic acid and furfural, which are cellulose/lignin
degradation products, but also 2-ethylhexanol associated with rosin
content. It is, thus, confirmed that rosin has a negative effect on
the stability of paper.
Acidity of paper, measured as pH of its aqueous extract,
30
is
both a measure of the accumulation of acids, because of degrada-
tion, and the result of acids introduced into the material during
sizing. As it is the most important endogenous factor affecting
the stability of historic paper, it is of interest if there are any volatile
degradation products associated with it in order to provide the
basis for a nondestructive method of determination. Recently, a
study has shown
3
that furfural emission can be directly correlated
with paper acidity. Despite the less satisfactory PLS correlation,
the loading weights show that acetic acid and furfural, but also
compounds associated with rosin content (2-ethylhexanol and
hexadecane), are associated with low pH. Acetic acid emission
was previously not associated with pH directly,
3
probably because
of its abundance in the headspace above degrading paper, which
may lead to saturation of the SPME absorbent. In this work, the
sample size was, therefore, adjusted, in order to be able to
semiquantitatively evaluate the emissions of acetic acid, as well.
Not all measured properties were correlated with VOC emis-
sions: for ash content and protein content, no meaningful PLS
calibrations were obtained.
Historic paper is often classified according to the fibers used
in papermaking: rag (usually linen, hemp, and cotton rags)
containing paper, bleached pulp containing paper, and ground-
wood containing paper. The distinction between the groups is not
very clear, especially because mixtures of raw materials were often
used. Groundwood containing papers have a higher amount of
lignin since the fibers have not been delignified. The vast majority
of rag papers are hand-sized using gelatin, which is why protein
content can be used to discriminate between rag papers and other
classes. In Figure 2, the papers group into the three categories
based on the quantifiable properties very well. The most valuable
information can be extracted from the positions along the PC1
axis, which covers 99% of the total information: PC1 is dominated
by lignin content and PC2 is dominated by ash and protein content.
It is also evident that rag papers generally have higher pH than
other papers used in the study, as pH is positively correlated with
protein content. Carbonyl group content, which is associated with
(30) Strlicˇ , M.; Kolar, J.; Kocˇar, D.; Drnovsˇ ek, T.; S
ˇelih, V. S.; Susicˇ, R.; Pihlar.,
B. e-Preserv. Sci. 2004,1, 35–47
.
Figure 2. PCA (left) based on quantifiable paper properties: ash, rosin, protein, carbonyl group, and lignin contents, degree of polymerization,
and pH (above) and volatile degradation products (below). Right: correlation loadings.
8621Analytical Chemistry, Vol. 81, No. 20, October 15, 2009
oxidative degradation of organic paper components, is positively
correlated with lignin and rosin content and negatively correlated
with protein content. DP could not be used in this part of the
study, as it cannot be determined for high lignin containing papers
due to their insolubility in the solvent used for viscometry. Ash
content is negatively correlated with protein content since fillers
(which contribute to ash content) are mainly associated with
modern papermaking. Rosin content also plays a role in the
separation of rag papers from other papers, and it is anticorrelated
with protein content, which is understandable, as papers were
either rosin or gelatin sized. Aluminum, interestingly, does not
play a major role, which is contrary to the popular belief that
aluminum can be associated with paper acidity: it may have been
immediately after papers had been sized, but not after decades of
natural degradation.
Using VOCs, a different classification of the samples is
achieved, which is of high interest. PC1 is dominated by acetic
acid, a general degradation marker. PC2 is dominated by a variety
of aldehydes and 2-ethylhexanol, which are associated with rosin
content. Furfural, as a general degradation marker, affects only
PC1. It appears that rag papers dominate a cluster of stable papers,
while the rest fan out according to rosin and lignin content. Such
separation could be the consequence of the fact that all the studied
papers degrade according to the same mechanisms: acid catalyzed
hydrolysis and oxidation simultaneously. Such classification is of
considerably higher interest than the one on the basis of
quantifiable properties, because it could lead to development of a
method for classification of differently stable papers on the basis
of VOC emission.
Using PLS, we have, thus, shown that 11 out of the 15 major
volatile degradation products can be associated with a specific
paper component and most are degradation markers. The PCA
classification based on VOCs reveals further information on the
difference in the effect of rosin and lignin on paper degradation,
which should be studied in more detail. There also seems to be
a significant difference between the amount and identity of
volatiles emitted by historic degraded papers (as discussed in our
study) and contemporary nondegraded paper samples (as dis-
cussed in the literature). We are focusing further research on
increased sample throughput with suitable data pretreatment to
increase the value and amount of extractable information based
on the principles of degradomics.
While a destructive approach has been required in the
development phase, we believe that this technique eventually can
be used on individual paper documents or to survey large
collections in a wholly nondestructive manner. Capture of VOCs
in a sampling enclosure or in standing library stacks followed by
degradomics data analysis could eventually be used to evaluate
the type, present status of, and the future likely stability of paper-
based historic documents.
CONCLUSIONS
We have successfully transferred the -omics concepts to the
field of study of degradation of materials, paper in particular.
Analyses of volatile degradation compounds formed during the
degradation of a number of real historic samples were performed,
and peak areas of the major compounds were compared with
quantifiable paper properties using multivariate data analyses, i.e.,
partial least-squares and principal component analyses. This
represents the first holistic and inductive approach to experimen-
tation augmenting the overwhelming body of hypothesis-driven
research in heritage science.
The approach allows us to understand the close relations
between various paper components affecting paper degradation
(rosin, lignin), the consequences thereof (content of carbonyl
groups), and properties indicative of material condition (cellulose
degree of polymerization, pH). We have shown that
(1) volatile degradic footprinting is a key tool to advance our
understanding of the pathways leading to production of VOCs
during paper degradation
(2) the composition of papers can be deducted on the basis of
volatile degradation products
(3) VOCs, which generate the smell of books, provide informa-
tion on paper condition and stability
Despite the drawbacks associated with the need for numerous
samples, which can be prohibitive, our understanding of complex
reaction systems leading to degradation of valuable heritage
materials can be greatly increased using the proposed material
degradomics approach. However, the approach is generally useful
for all materials where degradation is an economic concern, e.g.,
polymers, food, and pharmaceuticals. A systematic variation of
material composition (through production of historically informed
replicas in the case of heritage objects) and degradation environ-
ments (under the influence of light, heat, oxygen) will not only
inform stabilization strategies (e.g., using antioxidants, modified
environments, or cool storage) but also could lead to optimization
of production methods of materials themselves.
ACKNOWLEDGMENT
The authors acknowledge financial support from the Slovenian
Research Agency, the National Archives of The Netherlands
(project PaperVOC), and the Slovenian Research Agency (Pro-
gramme P1-0153). Use of the SurveNIR Historic Paper Collection
(www.science4heritage.org/survenir) is gratefully acknowledged.
Received for review July 19, 2009. Accepted September 1,
2009.
AC9016049
8622 Analytical Chemistry, Vol. 81, No. 20, October 15, 2009
