Chemical distinctions between Stradivaris maple and
modern tonewood
Hwan-Ching Taia,1,2, Guo-Chian Lia,1, Shing-Jong Huangb, Chang-Ruei Jhua, Jen-Hsuan Chunga, Bo Y. Wanga,
Chia-Shuo Hsua, Brigitte Brandmairc, Dai-Ting Chungd, Hao Ming Chena, and Jerry Chun Chung Chana
a
Department of Chemistry, National Taiwan University, Taipei 10617, Taiwan; bInstrumentation Center, National Taiwan University, Taipei 10617, Taiwan;
c
Private address, 86567 Hilgertshausen-Tandern, Germany; and dChimei Museum, Tainan 71755, Taiwan
Edited by Jerrold Meinwald, Cornell University, Ithaca, NY, and approved November 21, 2016 (received for review July 10, 2016)
Violins made by Antonio Stradivari are renowned for having been the
preferred instruments of many leading violinists for over two
centuries. There have been long-standing questions about whether
wood used by Stradivari possessed unique properties compared with
modern tonewood for violin making. Analyses of maple samples
removed from four Stradivari and a Guarneri instrument revealed
highly distinct organic and inorganic compositions compared with
modern maples. By solid-state 13C NMR spectroscopy, we observed
that about one-third of hemicellulose had decomposed after three
centuries, accompanied by signs of lignin oxidation. No apparent
changes in cellulose were detected by NMR and synchrotron X-ray
diffraction. By thermogravimetric analysis, historical maples exhibited
reduced equilibrium moisture content. In differential scanning calorim-
etry measurements, only maples from Stradivari violins, but not his
cellos, exhibited unusual thermooxidation patterns distinct from nat-
ural wood. Elemental analyses by inductively coupled plasma mass
spectrometry suggested that Stradivaris maples were treated with
complex mineral preservatives containing Al, Ca, Cu, Na, K, and Zn.
This type of chemical seasoning was an unusual practice, unknown to
later generations of violin makers. In their current state, maples in
Stradivari violins have very different chemical properties compared
with their modern counterparts, likely due to the combined effects
of aging, chemical treatments, and vibrations. These findings may in-
spire further chemical experimentation with tonewood processing for
instrument making in the 21st century.
|
Cremona stringed instrument | wood treatment | wood aging |
Italian violin
T he invention of the modern violin has generally been attrib-
uted to Andrea Amati, who initiated the golden age of violin
making in Cremona, Italy (15501750). The most famous Cre-
monese violin maker was Antonio Stradivari (AS, or Antonius
Stradivarius, 16441737), whose instruments are renowned for
their tonal qualities. Stradivaris success in violin building was
only matched by his neighbor, Giuseppe Guarneri del Ges
(DG, or Joseph Guarnerius, 16981744), and leading violinists
today still mostly prefer to play AS and DG instruments. It re-
mains a great mystery why functionally equivalent instruments
have not been successfully constructed to replace these heavily
repaired Cremonese antiques, despite rapid technological advances
over the last 200 y. For both historical and practical purposes, it is
important to understand what types of materials were used to
construct these famous violins.
Violin acoustics is mainly determined by vibrations of the front
and back plates, made of spruce and maple, respectively. It has
long been speculated that Stradivaris wood possessed unusual
properties not found in tonewood used by modern makers. Based
on visual and microscopic inspections, experts suggested that
wood varieties chosen by Stradivari have always been commer-
cially available (13). Wood densities in Cremonese violins and
modern tonewood also appeared to be similar, judging by com-
puted tomography measurements (4). There is little evidence to
suggest that the Cremonese tradition was lost due to diminished
www.pnas.org/cgi/doi/10.1073/pnas.1611253114
wood supply or deforestation, or that Stradivari preferred denser
woods that grew slowly during the Maunder Minimum (5).
Elemental analyses of AS and DG maple specimens conducted
by Nagyvary et al. (6) detected unusual minerals, which implicated
chemical treatments. They also reported severe degradation of the
lignocellulosehemicellulose matrix in violins through hydrolysis
and oxidation, which was attributed to chemical manipulation (7).
However, the conventional belief held by some violin restorers was
that Cremonese maple plates appeared stiffer or more elastic than
modern counterparts, generally assessed by tapping and listening
(8). This apparent paradox motivated us to analyze the maples of
five Cremonese instruments from three independent sources and
CHEMISTRY
compare their organic fiber composition, cellulose crystallinity,
moisture content, thermooxidative properties, and inorganic
elemental composition.
The maple samples analyzed in this study are listed in Table 1
(see SI Appendix, Table S1 for details). Four of our historical
specimens were removed from Cremonese instruments during
interior repairs of back plates. Three of them have been previously
analyzed by Nagyvary et al. (1717 AS violin; 1731 AS cello; 1741
DG violin) (6, 7), with the remaining materials transferred to us by
courtesy of J. Nagyvary. The other back plate sample (1707 AS
cello) was repaired at a different workshop (courtesy of Guy
Rabut). We also received the original neck of a 1725 AS violin
(courtesy of Chimei Museum). The neck was originally affixed to
the body with nails, which eventually rusted. It had to be modified
Significance
There have been numerous attempts to elucidate the secrets of
Stradivari violins, to explain why functional replacements have not
been reproduced over the past two centuries. Whether there are
systematic differences between Stradivari violins and later imita-
tions has been heatedly debated. Our analysis of Stradivaris ma-
ples from three independent sources showed reproducible
differences in chemical compositions compared with modern ma-
ples. Stradivaris use of mineral-treated maples belonged to a
forgotten tradition unknown to later violin makers. His maple also
appeared to be transformed by aging and vibration, resulting in a
unique composite material unavailable to modern makers. Modern
chemical analyses may, therefore, improve our understanding of
Stradivaris unique craft and inspire the development of novel
material approaches in instrument making.
Author contributions: H.-C.T., B.B., D.-T.C., H.M.C., and J.C.C.C. designed research; B.B. and
D.-T.C. provided historical consultation; G.-C.L., S.-J.H., C.-R.J., J.-H.C., B.Y.W., and C.-S.H.
performed research; G.-C.L., S.-J.H., C.-R.J., and J.-H.C. analyzed data; and H.-C.T. wrote
the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
Freely available online through the PNAS open access option.
1
H.-C.T. and G.-C.L. contributed equally to this work.
2
To whom correspondence should be addressed. Email: hctai@ntu.edu.tw.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.
1073/pnas.1611253114/-/DCSupplemental.
PNAS Early Edition | 1 of 6
Table 1. Maple samples analyzed in this study quantitative comparison was determined to be 1 ms, which was
Sample Description Origin also the conventional standard for studies on modern and ar-
chaeological woods (10, 11). Typical 13C{1H} CPMAS spectra
M1 Modern #1 European, flamed with contact time of 1 ms are shown in Fig. 1. The peak intensities
M2 Modern #2 European, flamed of these NMR spectra (SI Appendix, Figs. S12S14 and Table S2)
M3 Modern #3 European, flamed were subjected to principal component analysis (PCA), as shown
M4 Modern #4 European, plain in Fig. 2. We observed that historical maples were clearly sepa-
M5 Modern #5 European, plain rated from the modern controls along the first principal compo-
H1 1707 AS cello Back plate nent, which was primarily associated with the following peaks: 22,
H2 1717 AS violin Back plate, near the lower edge 56, 63, 72, 75, and 153 ppm. Differences along the first principal
H3 1725 AS violin neck Neck heel, original wood component (65% of the variance) were therefore strongly corre-
H4 1731 AS cello Back plate lated with hemicellulose changes, followed by lignin alterations.
H5 1741 DG violin Back plate, center region The second principal component further differentiated the his-
H6 Neck extension ca.1800 Heel extension on the 1725 neck torical neck samples from the historical back plate samples.
The rate of hemicellulose decomposition was estimated by the
hydrolysis of acetyl esters (22 ppm, acetyl CH3) and the chemical
by inserting new wood and reattached by gluing, which probably transformation of furanoses and pyranoses (C-2, C-3, and C-5 at
occurred around 1800 (2, 9). We cut into both the original and 75 ppm and C-6 at 63 ppm). After the subtraction of cellulose
extension parts to retrieve previously untouched wood from deep background signals, the rates of decomposition were found to be
locations (SI Appendix, Fig. S1). The extension piece (ca. 1800) similar at these three peaks (SI Appendix, Table S3). The average
provided a side-by-side aged control for Stradivaris neck material. signal reduction of hemicellulose was 36.9 3.9% (mean SE) in
The five modern controls were maple tonewoods recently pur- five Cremonese specimens compared with five modern controls (P <
chased in northern Italy. Our measurements revealed that Cre- 0.0001, Welchs t test with unequal variance). Modern and historical
monese maple clearly differed from modern tonewood with regard samples could be clearly differentiated by the extent of hemicellulose
to both organic and inorganic compositions, and the reasons for decomposition (Fig. 3A), providing experimental verification for the
such distinctions were further investigated and discussed. aged nature of our Cremonese samples. Assuming first-order ki-
netics, the half-life of hemicellulose decomposition was estimated
Results and Discussion around 409 56 y (mean SD) by regression analysis (SI Appendix,
Hemicellulose Decomposed in Historical Maples. The organic fiber Fig. S15). This was in reasonable agreement with previously pub-
compositions of historical and modern maples were analyzed lished measurements for dry-aged maple (358 y) (12), spruce (833
by 13C{1H} cross-polarization magic angle spinning (CPMAS) and 335 y) (12, 13), and fir (739 y, SI Appendix, Table S4) (14).
NMR spectroscopy. Historical maples exhibited reduced 13C
intensities at hemicellulose-related peaks (22, 62, 72, 75, and Lignin Became Oxidized in Historical Maples. In historical maples,
84 ppm) and lignin-related peaks (56 and 153 ppm) compared signal reductions at the methoxy (56 ppm) and aromatic carbons
with modern maples. These changes were evident across the (153 ppm) were indicative of lignin oxidation (SI Appendix, Fig.
spectra acquired at different CP contact times (0.2 ms to 10 ms, S14). There was only a weak correlation between lignin sig-
SI Appendix, Figs. S2S11), and the optimal contact time for nal changes at methoxy and aromatic carbons, suggesting that

Fig. 1. The 13C{1H} CPMAS spectra of maples de-

rived from modern control (M3) and three historical
specimens. The CP contact time was set to 1 ms. The
peaks were assigned to lignin, total cellulose (TC),
crystalline cellulose (CC), amorphous cellulose (AC),
hemicellulose (HC), and lignin.

2 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.1611253114 Tai et al.


Fig. 2. PCA score plot for the peak intensities of the CPMAS spectra mea-
sured in maple samples. The first principal component (PC1) can clearly
distinguish the modern and historical maples. Different colors represent
different sample origins (green for modern, red for historical back plates,
and blue for historical necks).
multiple oxidation mechanisms may have been involved. There
were two plausible oxidation mechanisms in Cremonese speci-
mens: oxidation by atmospheric oxygen and photooxidation.
Stradivaris own letter indicated the exposure of instruments to
sunlight (1), and further light exposures were inevitable over the
course of three centuries. Oxidation may cause lignin demethox-
ylation (15) and the formation of quinone groups (7, 16), resulting
in peak broadening due to increased structural heterogeneity, as
well as decreased CP signals due to hydrogen loss nearby. In
modern wood, the carbonyl peak at 173 ppm is mainly associated
with the acetyl groups in hemicellulose. However, the carbonyl
signal at 173 ppm in historical maples did not show a corre-
sponding decrease to the acetyl-CH3 signal at 22 ppm (SI Ap-
pendix, Fig. S14). This was probably due to the formation of
carbonyls and carboxylic acids from lignin oxidation (17), and
plausibly from the oxidation of carbohydrates as well.
Cellulose Remained Stable in Historical Maples. Cellulose microfibrils
in the wood contain individual rod-shaped crystalline domains sur-
rounded by amorphous regions. Because amorphous cellulose is less
stable than crystalline cellulose, the former may preferentially de-
grade over time, leading to higher crystalline/amorphous cellulose
ratios (18). However, this was not observed in Cremonese maples, as
we estimated the relative degree of cellulose crystallinity by the in-
tensity ratio of the NMR peaks at 89 and 105 ppm (crystalline cel-
lulose/total cellulose) (11, 19), which appeared to be unaltered in
Cremonese samples (Fig. 3A). We also measured the lengths and
widths of cellulose crystalline domains by wide-angle X-ray scattering
using synchrotron radiation, based on the (004) and (200) reflections,
respectively (SI Appendix, Figs. S16S18). Crystallite lengths and
widths, as well as d spacing, have also remained largely invariant in
historical maples (Fig. 3B and SI Appendix, Table S5). We observed
neither the evidence for the degradation of amorphous cellulose, nor
the interconversion between crystalline and amorphous cellulose, nor
the changes in crystallite size and density. Our findings with respect to
cellulose stability, mild lignin oxidation, and significant hemicellulose
decomposition in historical maples were in reasonable agreement
with the reported effects of dry aging on wood (8).
Quantitation of Cell Wall Composition by Different NMR Techniques.
Previous, Nagyvary et al. (7) have shown by 13C{1H} CPMAS
that hemicellulose and lignin signals were severely reduced in
Tai et al.
1717 AS violin and 1741 DG violin compared with modern
maples and 1731 AS cello, which was interpreted as evidence of
special chemical treatments applied to violins but not to cellos.
However, our 13C{1H} CPMAS data did not agree with their
previous findings. Our reanalysis of the same violin samples de-
tected only modest changes in hemicellulose and lignin, based on
extensive comparisons using different CP contact times (0.2 ms to
10 ms, SI Appendix, Fig. S11). We also applied the multiple-cross
polarization (multiCP) measurement method (SI Appendix, Fig.
S19 and Table S6), which has been recently proposed to improve
quantitation accuracy in 13C{1H} CPMAS NMR (20), to re-
confirm our observation. There were only minor spectral differ-
ences among the 1717 AS violin, 1725 AS violin neck, and 1731 AS
cello in multiCP measurements (SI Appendix, Fig. S19).
Two factors may have led to the discrepancy between our NMR
data and those of Nagyvary et al. (7). First, the maple shavings of
1717 and 1741 violins previously studied by Nagyvary et al. were
removed from interior areas that exhibited surface blackening,
possibly caused by surface heating. This discoloration was highly
unusual for Cremonese instruments, and we were informed by
J. Nagyvary that the darker surface portions have been used up for
analyses in the original reports (6, 7), whereas we only received
residual samples from deeper regions (>0.3 mm from the surface)
showing normal coloration consistent with the yellowing of natu-
CHEMISTRY
rally aged wood (SI Appendix, Fig. S20). Second, Nagyvary et al.
(7) used a fixed CP contact time of 5 ms for all of the NMR
measurements, where it was implicitly assumed that all carbon
signals would exhibit the same decay rate during the contact time.
By contrast, we carried out the CP experiments with different
contact times, and, indeed, we observed that there was a large
variation in signal decay rate for the carbon signals across different
samples (SI Appendix, Fig. S11 and Table S7). Consequently, the
spectra acquired with the contact time of 5 ms were not warranted
for quantitative analyses.
Thermal Analyses of Historical Maples. Interestingly, a peculiar dif-
ference was found between Stradivaris violin and cello samples
through differential scanning calorimetry (DSC) analyses (Fig. 4
and SI Appendix, Table S8). When gradually heated to 600 C under
air atmosphere, the thermograms of modern maples and two AS
cellos exhibited similar exothermic profiles with only two peaks. In
contrast, both 1717 AS violin and 1725 AS violin neck exhibited
three exothermic peaks. The unique DSC profile associated with
Stradivari violins has never been observed in natural wood, but only
in fungus-degraded wood (21) and simple mixtures of holocellulose
and lignin powders without mutual impregnation (22). It was noted
that when holocellulose and lignin were mixed together by copre-
cipitation to promote molecular adhesion, the thermogram
only showed two peaks, just like natural wood (22). Hence, the
Fig. 3. (A) Relative cellulose crystallinity plotted against relative hemi-
cellulose levels measured by 13C{1H} CPMAS. (B) Crystallite lengths and
widths estimated from X-ray diffraction patterns. Filled diamonds, modern
maples; open squares, historical specimens.
PNAS Early Edition | 3 of 6

Fig. 4. DSC thermograms of modern maple (M3) and five historical
specimens.
molecular adhesion between holocellulose and lignin appeared to
have been reduced in Stradivari violins, even compared with his
cellos from the same era. This may help explain why Nagyvary et al.
(7) observed obvious discrepancies between Stradivaris violins and
cellos by 13C{1H} CPMAS NMR, as changes in wood fiber ultra-
structure may affect the spin dynamics of the CP process.
By thermogravimetric analysis (TGA), we also observed a re-
duction of moisture content in Cremonese maples by about 25%
(Fig. 5 and SI Appendix, Table S9). Four modern maples exhibited
an average equilibrium moisture content (EMC) of 11.55 0.33%
(mean SE) at about 25 C and 58% relative humidity, in good
agreement with reported values for maple (23). The four Stradi-
vari maples exhibited 8.67 0.53% EMC, and the difference was
highly significant (P < 0.01, Welchs t test with unequal variance).
Reduced hygroscopicity in historical maples may be attributed to
the decomposition of hemicellulose (24, 25), which is more hy-
groscopic than cellulose and lignin (26).
Cremonese Maples Received Mineral Treatments. To understand
whether chemical treatments had possibly altered wood properties,
three historical samples (1731 cello, 1725 neck, and neck exten-
sion), with sufficient materials, were quantified for 25 elements
using inductively coupled plasma mass spectrometry (ICP-MS),
with the results listed in SI Appendix, Table S10. These included 20
metals and 5 metalloids/nonmetals (As, B, Ge, Sb, and P), but Cl, S,
and Si could not be included due to instrumental limitations. In the
PCA analysis of the quantified elements (Fig. 6), the neck extension
and modern controls were closely clustered and represented wood
receiving little or no mineral treatment, with total metal content
around 2,000 ppm. By contrast, the total metal content was elevated
to 9,000 ppm in the 1725 neck and 1731 cello, which clearly stood
out as being unnatural in the PCA diagram. These two Stradivari
instruments showed similar increases in K, Na, Ca, Cu, and Zn. The
AS neck contained elevated levels of Al, and the AS cello was
higher in B.
Nagyvary et al. (6) have previously shown in a qualitative
manner, by electron probe microanalyzer, that maples used by
Stradivari and Guarneri had altered mineral composition. Here
we provide a quantitative assessment of elemental changes by
ICP-MS measurements. The changes detected in our experiments
could neither be simply be attributed to surface applications of
potassium silicate (3) nor Pozzolana ash (27), which have been
previously proposed as wood sealers used by Stradivari but failed
4 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.1611253114
to be detected by recent studies (28, 29). Modest increases (tens of
parts per million) in fungicidal elements including Cu, Zn, and B
may have been introduced by either surface application or liquid
infusion, but much greater increases in Ca, Na, K, and Al could
only be explained by the latter. In the 18th century, mineral in-
fusion could have been achieved by extended soaking and/or sap
displacement procedures using gravity flow (30). Both soaking and
sap displacement could help remove sap nutrients that support
microorganism growth and help introduce various preservatives
such as alum, sodium chloride, or vitriols (sulfates of copper
and zinc).
Because minerals initially introduced into the wood could
have been washed out during later treatment steps, the residual
elements currently detected in Stradivaris maples are insuffi-
cient for ascertaining the original mineral formulations or ap-
plication procedures. There is also no available method to detect
past heat treatments such as boiling, steaming, baking, or fuming.
It is therefore plausible that chemical or physical treatments may
account for the different DSC profiles between Stradivaris vio-
lins and cellos, but different vibration frequencies may also be
critical. Even a few hours of vibration has been shown to reduce
the internal friction of wood, presumably by rearranging hydro-
gen bonds and polymer chains (31). Therefore, it is conceivable
that high-frequency vibrations could have gradually altered
the ultrastructure of wood fibers, acting in concert with hemi-
cellulose decomposition over time. The interactions among
mineral preservatives, age-dependent chemical changes, me-
chanical vibrations, and the composite nature of wood are extremely
complex, and the potential outcomes are not well understood.
Although chemical changes of wood generally have measurable
effects on its vibrational properties (8), the acoustic effects as-
sociated with the specific chemical alterations observed in this
study remain to be investigated.
Availability of Mineral-Treated Wood. Violin models designed by
Stradivari have been widely copied for over two centuries by
violin makers hoping to recreate similar playing qualities, but
gaining access to wood materials resembling Stradivaris maple
may be difficult. First, it is unclear if there are accelerated aging
methods that can faithfully mimic the effects of natural aging.
There have been numerous anecdotes about artificial treatments
in modern instruments that led to wood structural damages and
irreparable failures after just 10 y to 20 y of playing. Secondly,
wood infusion with mineral solutions was not a widespread
practice and it has long been obsolete (32). Because violin plates
are only a few millimeters thick, dimensional stability is an ex-
tremely important tonewood quality. Immersion of dried tone-
wood into liquids is likely to cause swelling damage and
compromise dimensional stability, and is therefore generally
Fig. 5. (A) TGA curves of modern and historical maples. (B) EMC values
derived from TGA data (**P < 0.01 by Welchs t test with unequal variance; n =
4 for each group).
Tai et al.

Fig. 6. PCA score plot for elemental compositions measured in maples by
ICP-MS. The 1731 cello (H4) was analyzed twice using two separate wood
fragments. Different colors represent different sample origins (green for
modern, red for historical back plates, and blue for historical necks).
avoided. Before the introduction of pressurization methods in
the 1800s, extensive mineral penetration by soaking or sap dis-
placement was probably only achievable with water-saturated
wood, before full drying took placelikely applied by wood
suppliers rather than violin makers.
To our knowledge, current supplies of commercial tonewood are
not mineral-infused, but just air-dried for 3 y to 10 y or even up to a
century to achieve ideal dimensional stability (8). Mineral infusion
was not detected in our modern controls and the neck extension
sample (Italy, ca. 1800), nor in the Jay viola (England, 1769) and
Gand & Bernardel violin (France, ca. 1845) previously analyzed by
Nagyvary et al. (6). Moreover, mineral treatment of wood has not
been mentioned in published treatises on violin making that con-
tain lutherie knowledge tracing back to the 18th and 19th centuries
(13, 33, 34). Building stringed instruments using mineral-treated
maples probably belonged to a long-forgotten tradition with few
practitioners to begin with. To this date, we have not identified the
use of mineral-infused wood outside Cremona, which may partly
explain why some restorers have noticed that worm damages were
relatively rare in Cremonese instruments.
Nevertheless, wood preservation using mineral treatment had a
long history in Italy and surrounding countries. Alum treatment of
wood was first recorded by ancient Romans for flame retardation
(35), and Paracelsus described, in the 16th century, the use of sal
gemma (alum) to enhance wood durability (36). Writings from
the 16th century about the positive effect of mineral treatments
on instrument acoustics have been recently rediscovered (37).
Chemical seasoning of wood using various minerals was still used
in the 20th century timber industry, but infrequently (32). With
our limited analytical data, it is difficult to elucidate the effects of
mineral treatments in Stradivari instruments. For instance, we do
not know if the pH was sufficiently acidic or basic during the
treatment to promote hemicellulose hydrolysis. It is also possible
that Al and Ca ions added to the wood may act as chemical cross-
linkersbeing chelated by hydroxyl, phenol, and carboxylate
groups on hemicellulose, lignin, and cellulose (38, 39)to com-
pensate for the natural degradation of hemicellulose over
time. Additional antique instruments from Cremona and
other Italian cities need to be analyzed to elucidate if there
were unique wood properties associated with Cremonese
makers or Stradivari himself.
Tai et al.
Conclusion
In this study, we provided a quantitative assessment of wood
chemical changes in antique musical instruments, using a com-
bination of five analytical methods: NMR, synchrotron X-ray,
DSC, TGA, and ICP-MS. In Cremonese maples, we observed
normal chemical changes associated with dry aging, including
hemicellulose decomposition, lignin oxidation, and reduced
EMC. However, we also observed an unnatural elevation of
certain inorganic elements, which could be attributed to the use
of mineral-treated maples by Cremonese makers.
From a conservation perspective, it appears that maples in
Stradivari cellos are reasonably preserved. The crystallinity of
cellulose has remained intact, as well as the adhesion between
holocellulose and lignin. The mechanical strength of aged wood
depends on the stability of crystalline cellulose as well as the in-
tegrity of the surrounding lignincarbohydrate matrix (25, 40, 41).
Hence, it is a concern that continued decomposition of hemi-
cellulose may eventually cause structural breakdown.
In Stradivari violins, we already observe evidence of reduced
adhesion between holocellulose and lignin. It is plausible that high-
frequency vibration accelerated the breakdown of cell wall com-
ponents, but whether wood treatment may have also played a role is
still unclear. It remains possible that some breakdown was initially
promoted by heat treatments or acidic/basic pH. However, it is also
CHEMISTRY
possible that increased divalent and trivalent metal ions could cross-
link carbohydrate and lignin chains to compensate for reduced mo-
lecular adhesion. Additional antique samples and analytical methods
are required to further elucidate the complex interactions between
aging, wood treatments, and long-term vibrations, which appear to
have jointly transformed the chemical properties of maples in Stra-
divari violins. Changes in spruce soundboards also warrant further
investigation. Our findings may inspire further chemical experimen-
tation with wood for 21st century instrument making, applicable not
just to violins but also to wooden instruments across different cul-
tures. Stradivaris maple may represent a singular case in the history
of wooden musical instruments, but its implications and impact may
be far-reaching.
Materials and Methods
Wood Samples. Historical maple samples were taken from these instruments:
1707 AS cello (H1), 1717 AS violin (H2), 1725 AS violin neck (H3), 1731 AS cello
(H4), and 1741 DG violin (H5), and neck extension ca. 1800 (H6). The modern
controls included five tonewood-grade maples of European origin (M1M5).
See SI Appendix, Table S1 for details.
NMR Spectroscopy. The 13C{1H} CPMAS NMR spectra were acquired at 13C and
1
H frequencies of 150.9 MHz and 600.1 MHz, respectively, on a Bruker
Avance III NMR spectrometer (14.1 T) equipped with a commercial 4-mm
probe. Wood samples (10 mg) were finely cut using scalpels and spun at
12.5 kHz with recycle delay of 2.0 s. During the CP contact, the 1H nutation
frequency was linearly ramped from 35 kHz to 50 kHz, and that of 13C was
optimized for the HartmannHahn matching condition. Proton decoupling
of 75 kHz was applied during the t2 acquisition.
The number of scans was typically 8,192 for measurements with the
contact time of 1 ms and 4,096 for experiments with variable contact times
(0.2 ms to 10 ms). Because of the limited resolution of the NMR spectra for
intact wood, we chose to analyze the peak intensities of the spectra without
performing spectral deconvolution. For the variable contact time experi-
ments, the peak intensities were analyzed by the following equation (42):


It = A1-exp-t=TCH exp -t T1 ,
where I(t) is the signal intensity at the contact time t, A is the intensity
factor, TCH characterizes the time constant of the polarization transfer
between 1 H and 13 C, and T1 is the 1H relaxation time in the spin-locking
radio frequency (RF) field. The curve fitting was performed in Origin2015
software (OriginLab).
For the multiCP measurements (20), the spinning frequencies were set at
12 kHz. Typical /2 pulse lengths of 4 s and 3.57 s were applied for the 13C
and 1H channels, respectively. Proton decoupling field strength of 75 kHz
was used during the acquisition. A total of nine CP blocks were implemented
PNAS Early Edition | 5 of 6
with the contact time of 1 ms and the RF amplitude ramping of 90 to 100%.
The recycle delay was 2 s, and the duration of the repolarization period tz
was 0.9 s. After the multiCP excitation, a Hahn echo of two rotor periods was
applied before the signal detection.
Wide-Angle X-Ray Scattering. The experiments were performed at the BL01C2
beamline of the National Synchrotron Radiation Research Center, in which the
ring was operated at 1.5 GeV energy with a typical current around 360 mA. A
thin slice of wood sample was fixed with 3M Magic Tape and placed on the
sample stage. Two pairs of slits and one collimator were set up to provide a
collimated beam with dimensions of 0.1 0.1 mm (H V) at the sample. The
wavelength of the incident X-rays was 1.033210 (12 keV), delivered from the
5-T Superconducting Wavelength Shifter and an Si (111) triangular crystal
monochromator. The scattering signal was collected with a Mar345 imaging
plate area detector, setting the exposure duration at 30 s. The diffraction
angles were calibrated according to Bragg positions of CeO2 (SRM 674b)
standards in desired geometry, and then GSAS II software (Argonne National
Laboratory) was used to obtain corresponding one-dimensional powder dif-
fraction profile with cake-type integration. The profile data were exported to
Origin2015 software for baseline correction and smoothing. The size of the
crystalline domain (L) was estimated using the Scherrer equation
L = K =B cos,
where K is the shape factor equal to 0.94 for wood cellulose (43), is the
X-ray wavelength, B is the full width at half maximum in 2 units, and is
the Bragg angle.
ICP-MS. About 20 mg to 50 mg of finely cut wood sample was weighed and
digested with 5 mL of 70% (wt/wt) nitric acid (trace metal grade) in a 50-mL glass
centrifuge tube precleaned with 20% (wt/wt) nitric acid. The digestion mixture
was heated in a 100 C water bath for 8 h, followed by dilution with deionized
water to 50 mL, and analyzed by Agilent 7500ce ICP-MS. High-purity standards
of individual elements (1,000 mg/L) were diluted in 1% nitric acid to
prepare standards for calibration curves. ICP-MS was operated at 1.5-kW RF
1. Hill WH, Hill AF, Hill AE (1902) Antonio Stradivari: His Life and Work (16441737)
(Dover, New York).
2. Pollens S (2010) Stradivari (Cambridge Univ Press, Cambridge, UK).
3. Sacconi SF (1979) The Secrets of Stradivari, trans Dipper A, Rivaroli C (Libreria del
Convegno, Cremona, Italy).
4. Stoel BC, Borman TM, de Jongh R (2012) Wood densitometry in 17th and 18th century
Dutch, German, Austrian and French violins, compared to classical Cremonese and
modern violins. PLoS One 7(10):e46629.
5. Burcklem L, Grissino-Mayer HD (2003) Stradivari, violins, tree rings, and the Maunder
Minimum: a hypothesis. Dendrochronologia 21(1):4145.
6. Nagyvary J, Guillemette RN, Spiegelman CH (2009) Mineral preservatives in the wood
of Stradivari and Guarneri. PLoS One 4(1):e4245.
7. Nagyvary J, DiVerdi JA, Owen NL, Tolley HD (2006) Wood used by Stradivari and
Guarneri. Nature 444(7119):565.
8. Bucur V (2006) Acoustics of Wood (Springer, Berlin), 2nd Ed.
9. Hart G (1909) The Violin: Its Famous Makers and Their Imitators (Dulau, London).
10. Bardet M, Foray MF, Maron S, Goncalves P, Tran QK (2004) Characterization of wood
components of Portuguese medieval dugout canoes with high-resolution solid-state
NMR. Carbohydr Polym 57(4):419424.
11. Gil AM, Pascoal Neto C (1999) Solid-state NMR studies of wood and other lignocel-
lulosic materials. Annu Rep NMR Spectrosc 37:75117.
12. Pishik I, Fefilon V, Burkovskaya V (2011) Chemical composition and chemical prop-
erties of new and old wood. Lesnoi J 14(6):8993.
13. Kranitz K (2014) Effect of natural aging on wood. Doctoral thesis (Eidgenssische
Technische Hochschule Zrich, Zurich).
14. Kacik F, Smira P, Kacikova D, Reinprecht L, Nasswettrova A (2014) Chemical changes in
fir wood from old buildings due to ageing. Cell Chem Technol 48(1-2):7988.
15. Paulsson M, Parkas J (2012) Review: Light-induced yellowing of lignocellulosic pulps
Mechanisms and preventive methods. BioResources 7(4):59956040.
16. Ganne-Chedeville C, Jaaskelainen A-S, Froidevaux J, Hughes M, Navi P (2012) Natural
and artificial ageing of spruce wood as observed by FTIR-ATR and UVRR spectroscopy.
Holzforschung 66(2):163170.
17. Evstigneeva EI, et al. (2015) Chemical structure and physicochemical properties of
oxidized hydrolysis lignin. Russ J Appl Chem 88(8):12951303.
18. Yonenobu H, Tsuchikawa S (2003) Near-infrared spectroscopic comparison of antique
and modern wood. Appl Spectrosc 57(11):14511453.
19. Wikberg H (2004) Advanced solid state NMR spectroscopic techniques in the study of
thermally modified wood. Doctoral thesis (Univ Helsinki, Helsinki).
20. Johnson RL, Schmidt-Rohr K (2014) Quantitative solid-state 13C NMR with signal en-
hancement by multiple cross polarization. J Magn Reson 239:4449.
6 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.1611253114
power and <5-W reflected power, with plasma/auxiliary/nebulizer flow rates
at 15/0.8/0.95 L/min.
Thermal Analysis. DSC and TGA experiments were conducted using a Mettler
TGA/DSC 1 instrument. Before analysis, finely cut wood samples were equilibrated
for 48 h in a sealed chamber with saturated sodium bromide solution, with room
temperature around 25 C and internal relative humidity around 58%. It only
took around 5 min to transfer 5 mg of wood from the chamber to an alumi-
num crucible with a pierced lid and to have it stabilized inside the instrument at
34 C. Subsequently, the sample was heated at 10 C/min under air atmosphere
and recorded from 35 C to 600 C. EMC was calculated as (W35 W150)/W150,
where W35 and W150 were weights recorded at 35 C and 150 C, respectively.
Color Photography. Small pieces of wood samples were placed on white filter
papers inside a desktop photo studio with external lighting using white
compact fluorescent lamps. A Takumar SMC 50 mm F1.4 lens (Ricoh) was
mounted in reverse using appropriate adapters on a 24.3-megapixel Sony alpha
ILCE-7 camera. The reversed connection resulted in a flange focal distance of
45.46 mm and ensured that all in-focus objects had the same magnification
ratio. We performed color calibration on wood images by photographing
QPcard 203 and using QPcalibration software ver. 1.99b, with final processing in
Adobe Lightroom ver. 5.7.1.
Statistical Tests. Welchs t test and linear regression were analyzed using
Graphpad Prism software ver. 5.0. PCA was performed with Origin2015
software by choosing the correlation matrix (standardized) option.
ACKNOWLEDGMENTS. We thank Joseph Nagyvary, Chimei Museum, and Guy
Rabut for providing historical wood samples, and Sandro Chiao and Boa-Tsang
Lee for providing modern maples. We thank Wen-Feng Chang of the Instrument
Center at National Tsing-Hua University (Hsinchu, Taiwan) for ICP-MS measure-
ments, and the National Synchrotron Radiation Research Center (Hsinchu,
Taiwan) for technical services. This work was supported by Grant 102R890927
from National Taiwan University, and Grants 103-2113-M-002-007-MY2 and 105-
2633-M-002-001 from Ministry of Science and Technology, Taiwan.
21. Reh U, Kraepelin G, Lamprecht I (1986) Use of differential scanning calorimetry for
structural analysis of fungally degraded wood. Appl Environ Microbiol 52(5):11011106.
22. Tsujiyama S, Miyamori A (2000) Assignment of DSC thermograms of wood and its
components. Thermochim Acta 351(1-2):177181.
23. Goulet M, Hernandez R (1991) Influence of moisture sorption on the strength of
sugar maple wood in tangential tension. Wood Fiber Sci 23(2):197206.
24. Kohara J, Okamoto H (1955) Studies of Japanese old timbers. Sci Rep Saikyo Univ 7
(1a):920.
25. Yokoyama M, et al. (2009) Mechanical characteristics of aged Hinoki wood from
Japanese historical buildings. C R Phys 10(7):601611.
26. Kollmann FFP, Cote WA, Jr (1968) Principles of Wood Science and Technology: I Solid
Wood (Springer, Berlin).
27. Barlow CY, Edwards PP, Millward GR, Raphael RA, Rubio DJ (1988) Wood treatment
used in Cremonese instruments. Nature 332(6162):313.
28. Echard JP, et al. (2010) The nature of the extraordinary finish of Stradivaris instru-
ments. Angew Chem Int Ed Engl 49(1):197201.
29. Brandmair B, Greiner PS (2010) Stradivari Varnish (B. Brandmair, Munich).
30. Parnell EA (1844) Applied Chemistry: In Manufactures, Arts, and Domestic Economy
(D. Appleton, New York).
31. Sobue N, Okayasu S (1992) Effects of continuous vibration on dynamic viscoelasticity
of wood. J Soc Mat Sci Jpn 41(461):164169.
32. Kollmann FFP, Kuenzi EW, Stamm AJ (1975) Principles of Wood Science and
Technology: II Wood Based Materials (Springer, Berlin).
33. Regazzi R (1986) The Manuscript on Violin Making by G. A. Marchi - Bologna 1786,
trans Sbarra N (A. Forni, Bologna, Italy).
34. Roy K (2006) The Violin: Its History and Making (Karl Roy, Barrington, NH).
35. Cornell TJ (2013) The Fragments of the Roman Historians (Oxford Univ Press, Oxford).
36. Waite AE (1894) The Hermetic and Alchemical Writings of Paracelsus (J. Elliott, London).
37. Gug R (1991) Salted soundboards and sweet sounds. Strad 102(1214):506511.
38. Dheu-Andries ML, Perez S (1983) Geometrical features of calciumcarbohydrate in-
teractions. Carbohydr Res 124(2):324332.
39. Nolte H, John S, Smidsrod O, Stokke BT (1992) Gelation of xanthan with trivalent
metal ions. Carbohydr Polym 18(4):243251.
40. Lionetto F, Del Sole R, Cannoletta D, Maffezzoli A (2012) Monitoring wood degra-
dation during weathering by cellulose crystallinity. Materials (Basel) 5(10):19101922.
41. Obataya E (2012) Effects of aging on the vibrational properties of wood. J Cult Herit
13(3 Suppl):S21S25.
42. Love GD, Snape CE, Jarvis MC (1992) Determination of the aromatic lignin content in
oak wood by quantitative solid-state 13C-NMR. Biopolymers 32(9):11871192.
43. Fernandes AN, et al. (2011) Nanostructure of cellulose microfibrils in spruce wood.
Proc Natl Acad Sci USA 108(47):E1195E1203.
Tai et al.
 Supporting Information Appendix

for

Chemical distinctions between Stradivaris maple and modern tonewood

Hwan-Ching Tai1, Guo-Chian Li1, Shing-Jong Huang2, Chang-Ruei Jhu1, Jen-Hsuan Chung1,

Bo Y. Wang1, Chia-Shuo Hsu1, Brigitte Brandmair3, Dai-Ting Chung4, Hao Ming Chen1,

Jerry Chun Chung Chan1

1
Department of Chemistry, National Taiwan University, Taipei 10617, Taiwan;
2
Instrumentation Center, National Taiwan University, Taipei 10617, Taiwan;
3
Private address, 86567 Hilgertshausen-Tandern, Germany;
4
Chimei Museum, Tainan 71755, Taiwan

These authors contributed equally to this work

1
 (I) Supplementary Tables

Table S1. Origins of maple samples analyzed in this study

Sampling depth
Sample Description Origin Provider
from surface

M1 modern #1 European [1] >2 mm Sandro Chiao

M2 modern #2 European [1] >2 mm Sandro Chiao

M3 modern #3 European [1] >2 mm Sandro Chiao

M4 modern #4 European [2] >2 mm Boa-Tsang Lee

M5 modern #5 European [2] >2 mm Boa-Tsang Lee

H1 1707 AS cello back plate [3] not available Guy Rabut

H2 1717 AS violin back plate, near the lower edge [4] >0.3 mm Joseph Nagyvary

H3 1725 AS neck neck heel, original wood [5] > 3 mm Chimei Museum

H4 1731 AS cello back plate [4] >0.3 mm Joseph Nagyvary

H5 1741 DG violin back plate, center [4] >0.3 mm Joseph Nagyvary

H6 neck extension c.1800 heel extension on the 1725 neck [6] > 3 mm Chimei Museum

Notes:
[1] Flamed tonewood from Sandro Chiaos workshop in Taipei, Taiwan. Purchased in Italy, circa 2005,
probably of Balkan origin. Maple species traditionally used by Italian violin makers included Acer
pseudoplatanus, Acer platanoides, and Acer campestre, the woods of which were practically
indistinguishable.
[2] Plain-figured tonewood from Boa-Tsang Lees workshop in Taipei, Taiwan. Purchased in Italy, circa
1995, probably of Italian origin.
[3] The sample was taken in the course of restoration in the 1980's and generously provided by Guy Rabut
(New York, NY).
[4] The sample was taken in the course of restoration and generously provided by Ren A. Morel (New
York, NY). Portions of the sample had been used for previous studies (1, 2), and the remaining
material was generously provided by Joseph Nagyvary.
[5] The original neck removed from the 1725 Stradivari Brancaccio violin, which was owned by the
Brancaccio family in Napoli during the 19th century. See Fig. S1 for sampling location.
[6] Wood inserted by an unidentified Italian restorer, circa 1800. See Fig. S1 for sampling location.

2
 Table S2. 13C{1H} CPMAS NMR peak intensities of modern and historical maples

13
C chemical shift (ppm)
Sample
22 56 63 65 72 75 84 89 105 137 148 153 173

modern M1 0.307 0.691 0.850 0.749 2.158 2.379 0.494 0.315 1.000 0.085 0.078 0.201 0.086

modern M2 0.326 0.639 0.780 0.700 2.030 2.278 0.479 0.333 1.000 0.098 0.089 0.197 0.086

modern M3 0.331 0.628 0.794 0.724 2.064 2.265 0.446 0.318 1.000 0.087 0.083 0.185 0.084

modern M4 0.261 0.622 0.759 0.717 2.005 2.177 0.446 0.350 1.000 0.082 0.084 0.167 0.076

modern M5 0.313 0.575 0.764 0.695 1.979 2.249 0.464 0.345 1.000 0.085 0.095 0.183 0.085
neck extension
0.209 0.482 0.635 0.649 1.809 1.828 0.429 0.371 1.000 0.081 0.096 0.119 0.118
c. 1800
AS neck 1725 0.154 0.455 0.547 0.607 1.740 1.825 0.430 0.391 1.000 0.081 0.094 0.154 0.083

AS violin 1717 0.190 0.485 0.608 0.682 1.886 1.857 0.388 0.355 1.000 0.070 0.090 0.129 0.062

DG violin 1741 0.166 0.427 0.622 0.696 1.868 1.801 0.341 0.315 1.000 0.050 0.091 0.139 0.060

AS cello 1707 0.236 0.510 0.653 0.714 1.914 1.886 0.417 0.377 1.000 0.084 0.085 0.156 0.066

AS cello 1731 0.241 0.478 0.643 0.688 1.872 1.878 0.394 0.316 1.000 0.056 0.085 0.140 0.080

Note: CP contact time = 1 ms; the peak intensities were normalized with reference to the peak at 105 ppm
(C-1, total cellulose)

3
Table S3. Peak intensities of hemicellulose extracted from 13C{1H} CPMAS spectra

Mean
22 ppm 63 ppm 75 ppm
Sample (hemicellulose
(acetyl CH3) (C-6) (C-2, 3, 5)
level)
modern M1 0.997 1.125 1.100 1.074

modern M2 1.061 0.980 1.008 1.016

modern M3 1.075 1.010 0.996 1.027

modern M4 0.848 0.937 0.915 0.900

modern M5 1.018 0.947 0.981 0.982

neck extension c. 1800 0.679 0.680 0.593 0.651

AS neck 1725 0.501 0.498 0.591 0.530

AS violin 1717 0.618 0.624 0.620 0.621

DG violin 1741 0.540 0.654 0.568 0.587

AS cello 1707 0.769 0.718 0.646 0.711

AS cello 1731 0.782 0.697 0.639 0.706

Note: From Table S2, the cellulose background of the peaks at 63 and 75 ppm were estimated to
be 0.306 and 1.184, respectively. These values were derived from the spin-locking cellulose
spectra of Kraft pulp reported by Liitia (3). Note that the peak at 22 ppm does not have cellulose
background. After subtracting the cellulose background, the mean of each NMR peak of the five
modern maples was used to scale the intensities of the corresponding NMR peaks of the historical
maples.

4
 Table S4. Linear regression analysis of hemicellulose decomposition rate

Age range Slope* Half-life Std. dev. of

Sample series R2 Data source
(year) (103 year1) (year) half-life

Tonewood
10308 1.693 0.8564 409 56 Table S3
maples

Maple 0100 1.934 0.9976 358 18 Pishik et al.#

Spruce 0700 0.8318 0.9605 833 76 Pishik et al.#

Spruce 0210 2.068 0.5947 335 160 Kranitz(4)

Fir 0389 0.9379 0.9773 739 56 Kacik et al.(5)

* Slope of log [hemicellulose/cellulose] vs. sample age

#
The original study by Pishik et al. was published in Russian (6), and the relevant data has been
summarized by Bucur (7).

5
 Table S5. Wide-angle X-ray scattering data of modern and historical maples

(200) (004)
Sample
domain domain
2 FWHM* d (nm) 2 FWHM* d (nm)
size (nm) size (nm)
modern M1 14.85 1.763 3.18 0.400 23.10 0.1840 30.87 0.258

modern M2 14.83 1.871 3.00 0.400 23.08 0.1687 33.67 0.258

modern M3 14.82 1.871 3.00 0.401 23.08 0.1687 33.67 0.258

modern M4 14.91 1.858 3.02 0.398 23.14 0.2116 26.84 0.258

modern M5 14.91 1.822 3.08 0.398 23.10 0.2024 28.06 0.258

AS cello 1707 14.94 1.822 3.08 0.397 23.17 0.1748 32.50 0.257

AS cello 1731 14.76 1.886 2.98 0.402 22.97 0.2024 28.06 0.259

AS violin 1717 14.82 2.116 2.65 0.400 23.00 0.2116 26.84 0.259

DG violin 1741 14.84 1.674 3.35 0.400 22.98 0.1840 30.86 0.259

AS neck 1725 15.12 1.757 3.19 0.393 23.12 0.1932 29.40 0.258
neck extension
14.99 1.803 3.11 0.396 23.13 0.2024 28.06 0.258
c.1800

* full-width-at-half-maxima in 2 units (degrees)

6
Table S6. Comparison of the peak intensities between the 13C{1H} CPMAS spectrum
with contact time of 1 ms and the 13C{1H} multiCP spectrum

Neck extension
Peak Modern M3 AS neck 1725 AS violin 1717 AS cello 1731
c. 1800

(ppm) 1 ms multiCP 1 ms multiCP 1 ms multiCP 1 ms multiCP 1 ms multiCP

22 0.331 0.253 0.209 0.172 0.154 0.140 0.190 0.149 0.241 0.184

56 0.628 0.512 0.482 0.411 0.455 0.459 0.485 0.444 0.478 0.473

63 0.794 0.633 0.635 0.567 0.547 0.534 0.608 0.585 0.643 0.601

65 0.724 0.702 0.649 0.639 0.607 0.632 0.682 0.743 0.688 0.686

72 2.064 1.931 1.809 1.810 1.740 1.740 1.886 1.841 1.872 1.777

75 2.265 1.933 1.828 1.871 1.825 1.825 1.857 1.757 1.878 1.691

84 0.446 0.521 0.429 0.448 0.430 0.467 0.388 0.489 0.394 0.436

89 0.318 0.384 0.371 0.357 0.391 0.381 0.355 0.367 0.316 0.362

105 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000

137 0.087 0.208 0.081 0.119 0.081 0.153 0.070 0.148 0.056 0.133

148 0.083 0.171 0.096 0.156 0.094 0.129 0.090 0.183 0.085 0.169

153 0.185 0.373 0.119 0.290 0.154 0.327 0.129 0.269 0.140 0.272

173 0.084 0.138 0.118 0.121 0.083 0.211 0.062 0.120 0.080 0.138

7
 Table S7. T1 data of modern and historical maples

13
C chemical shift (ppm)
Sample
22 56 75 84 89 105 153 173

T1 (ms) 9.54 9.34 6.87 6.61 7.40 7.33 16.93 17.12

modern M1
Std. dev. 0.71 0.64 0.20 0.22 0.36 0.26 2.91 6.12

T1 (ms) 6.21 6.57 4.02 3.48 3.89 4.47 6.58 7.29

modern M4
Std. dev. 0.42 0.39 0.22 0.29 0.27 0.16 0.51 1.02

T1 (ms) 6.91 6.12 4.19 4.10 4.79 4.86 7.79 6.33

modern M5
Std. dev. 0.65 0.35 0.33 0.42 0.45 0.20 0.95 1.82

T1 (ms) 6.47 6.42 4.27 3.84 4.37 4.86 6.74 8.30

AS cello 1707
Std. dev. 0.53 0.68 0.20 0.40 0.43 0.13 1.39 1.99

T1 (ms) 8.84 8.25 6.53 6.45 7.26 6.70 10.48 6.36

AS cello 1731
Std. dev. 1.11 0.32 0.22 0.45 0.59 0.17 1.31 1.04

T1 (ms) 10.72 8.71 7.28 7.52 8.30 7.40 10.32 18.49

AS violin 1717
Std. dev. 1.68 0.61 0.40 0.76 1.27 0.38 1.59 10.64

T1 (ms) 5.93 5.45 3.83 3.19 3.35 4.02 7.59 8.22

DG violin 1741
Std. dev. 1.13 0.34 0.12 0.40 0.30 0.08 1.94 1.91

T1 (ms) 9.08 8.30 6.48 7.06 7.44 7.11 9.79 4.80

AS neck 1725
Std. dev. 1.16 0.92 0.39 0.56 0.56 0.42 2.04 1.79

T1 (ms) 7.26 7.94 6.05 6.00 7.20 6.41 6.81 4.51

neck extension
c.1800
Std. dev. 1.14 0.65 0.29 0.64 0.59 0.12 1.95 1.89

8
 Table S8. DSC peak profiles of modern and historical maples

Early peak Intermediary Late peak

Sample Sample number
(oC) peak (oC) (oC)
modern M1 355 ND 476
modern M2 359 ND 486
modern M3 361 ND 491
modern M4 346 ND 462
modern M5 364 ND 493
#1 344 ND 486
neck extension
c. 1800 #2 344 ND 492
AS cello 1707 358 ND 481
#1 347 ND 479
AS cello 1731
#2 350 ND 490
left side #1 317 419 464

AS violin neck left side #2 317 424 471

1725 right side #1 314 412 463
right side #2 314 417 469
#1 314 409 465
AS violin 1717
#2 315 408 463

ND = not detected

9
 Table S9. Moisture content of modern and historical maples measured by TGA

Equilibration condition*
Weight at Weight at EMC
Sample
temperature relative 35 oC (mg) 150 oC (mg) (%)
(oC) humidity (%)

modern M1 25 58 6.0986 5.5147 10.6

modern M2 25 58 6.5295 5.8435 11.7

modern M3 25 58 5.6514 5.0547 11.8

modern M5 25 58 5.3106 4.7391 12.1

AS cello 1707 25 58 6.3026 5.8451 7.8

AS cello 1731 25 58 5.8127 5.3837 8.0

AS violin 1717 25 58 5.0166 4.6135 8.7

AS neck 1725 25 58 5.2546 4.7702 10.2

neck extension
25 58 5.1567 4.7421 8.7
c. 1800

* Equilibrated for 48 hours in a sealed chamber with saturated NaBr solution. The room temperature was
controlled between 2427 oC, which corresponded to 5758% relative humidity inside the chamber.

10
 Table S10. ICP-MS measurements of modern and historical maples

neck 1731 AS 1731 AS

modern modern modern modern modern 1725 AS Detection
Element Unit extension cello cello
M1 M2 M3 M4 M5 neck limit
c. 1800 sample 1* sample 2*

Al ppm 10.6 289.6 114.5 2.6 52.0 ND 547.2 29.8 32.0 0.127
As ppm ND ND ND ND ND ND ND ND ND 0.002
B ppm 4.3 3.5 3.3 8.5 8.2 6.3 6.0 31.4 32.4 0.157
Ba ppm 0.6 0.9 0.6 1.5 0.8 1.3 1.2 2.1 2.9 0.005
Ca ppm 420.0 395.9 357.0 460.6 455.5 404.2 1448.0 1621.0 1867.4 0.332
Cr ppm ND ND ND ND ND ND ND ND ND 0.013
Cu ppm 0.9 1.6 1.3 1.4 1.0 4.6 15.8 30.0 33.6 0.007
Fe ppm 3.8 ND ND 2.1 ND ND 553.7 #
ND ND 0.356
Ge ppm ND ND ND ND ND ND ND ND ND 0.002
Hg ppm ND ND ND ND ND ND ND ND ND 0.098
K ppm 838.9 623.9 593.8 1789.4 1045.9 1035.4 4181.0 3709.0 3984.0 1.514
Li ppm ND ND ND ND ND ND ND ND ND 0.017
Mg ppm 302.8 405.1 403.8 450.5 262.6 449.0 271.8 183.2 214.4 0.027
Mn ppm 3.7 2.2 1.8 5.4 3.4 1.5 5.4 3.2 4.9 0.001
Na ppm 24.5 45.5 45.3 35.1 27.6 216.2 2727.0 3292.0 3485.0 0.529
Ni ppm ND 0.2 ND 0.3 0.5 0.2 1.1 10.2 12.8 0.004
P ppm 69.4 110.2 23.3 124.4 131.9 126.2 30.3 123.2 141.6 0.679
Pb ppm 0.3 0.3 0.2 0.2 0.3 2.9 1.6 7.7 10.5 0.003
Sb ppm ND ND ND ND ND ND ND ND ND 0.004
Se ppm ND ND ND ND ND ND ND ND ND 0.098
Sn ppm 0.3 0.3 0.2 0.4 0.3 0.4 1.1 1.1 1.1 0.003
Sr ppm 3.1 0.9 0.8 2.8 0.9 8.3 8.8 4.2 4.8 0.003
Ti ppm 2.5 19.2 3.3 3.6 6.6 4.8 1.5 27.7 35.3 0.001
Zn ppm 4.3 8.2 6.1 4.7 4.9 55.9 36.4 44.7 42.0 0.030
Zr ppm ND ND ND ND ND ND ND ND ND 0.001

Total
ppm 1616 1794 1529 2761 1862 2185 9248 8966 9731
metal

Notes:
* Two different wood fragments from the 1731 cello were analyzed independently
#
Possible Fe contamination by rusted nails inside
ND = not detected

11
 (II) Supplementary Figure Legends

Figure S1. Photographs of the 1725 AS violin neck. (a) The original neck of the 1725 violin was
initially nailed to the body by Stradivari (black arrowhead indicates nail holes). (b) The open
arrowhead indicates where the inserted heel extension wood (c. 1800) was sampled for analysis.
Stradivaris original wood was sampled from both the left side and the right side (black arrowheads)
of the extension material, using a combination of mechanical drill and hand tools to retrieve wood
more than 3 mm beneath the surface.

Figure S2. 13C{1H} CPMAS NMR spectra of modern maple M1 with different contact times. The
experimental contact time was indicated for each trace.

Figure S3. 13C{1H} CPMAS NMR spectra of modern maple M4 with different contact times.

Figure S4. 13C{1H} CPMAS NMR spectra of modern maple M5 with different contact times.

Figure S5. 13C{1H} CPMAS NMR spectra of 1707 AS cello maple with different contact times.

Figure S6. 13C{1H} CPMAS NMR spectra of 1717 AS violin maple with different contact times.

Figure S7. 13C{1H} CPMAS NMR spectra of 1725 AS neck maple with different contact times.

Figure S8. 13C{1H} CPMAS NMR spectra of 1731 AS cello maple with different contact times.

Figure S9. 13C{1H} CPMAS NMR spectra of 1741 DG violin maple with different contact times.

Figure S10 13C{1H} CPMAS NMR spectra of neck extension (c. 1800) with different contact times.

Figure S11. Variation of 13C{1H} CPMAS peak intensities at different contact times. For each maple

12
sample, the peak intensities at different chemical shifts and different contact times from the same
sample were normalized with reference to the signal at 105 ppm with contact time equal to 1 ms.

Figure S12. 13C{1H} CPMAS NMR spectra of 5 modern and 6 historical maples. Acquired with 1
ms CP contact time and normalized to the total cellulose peak at 105 ppm.

Figure S13. 13C{1H} CPMAS NMR spectra of 5 modern and 6 historical maples, enlarged insets.
The spectra from Figure S12 are shown as enlarged insets for detailed comparisons. The peaks were
assigned to total cellulose (TC), crystalline cellulose (CC), amorphous cellulose (AC), hemicellulose
(HC), and lignin.

Fig. S14. Comparison of 13C{1H} CPMAS peak intensities of the spectra of modern and historical
maples. For brevity, only the mean of 5 modern maples are shown (Table S2), where the error bar
represents 95% confidence interval of the mean. Values of the historical maples were normalized
with reference to the mean of the modern maples.

Figure S15. Kinetics of hemicellulose decomposition in historical maples. Linear regression analysis
based on the mean hemicellulose levels in Table S3.

Figure S16. Wide-angle X-ray scattering patterns of selected maple specimens.

Figure S17. Diffraction peaks of (004) obtained for modern and historical maple samples.

Figure S18. Diffraction peaks of (200) obtained for modern and historical maple samples.

Figure S19. 13C{1H} multiCP spectra of maples from modern control (M3) and three historical
specimens.

Figure S20. Color comparison between modern and historical maple specimens.

13
 (III) References

1. Nagyvary J, DiVerdi JA, Owen NL, & Tolley HD (2006) Wood used by Stradivari and
Guarneri. Nature 444:565.
2. Nagyvary J, Guillemette RN, & Spiegelman CH (2009) Mineral preservatives in the wood of
Stradivari and Guarneri. PloS one 4:e4245.
3. Liitia T (2002) Application of Modern NMR Spectroscopic Techniques to Structural Studies
of Wood and Pulp Components. Doctoral thesis (University of Helsinki, Helsinky, Finland).
4. Kranitz K (2014) Effect of natural aging on wood. Doctoral thesis (Eidgenossische
Technische Hochschule Zurich, Zurich, Switzerland).
5. Kacik F, Smira P, Kacikova D, Reinprecht L, & Nasswettrova A (2014) Chemical changes in
fir wood from old buildings due to ageing. Cellulose Chem. Technol. 48:79-88.
6. Pishik I, Fefilon V, & Burkovskaya V (2011) Chemical Composition and Chemical Properties
of New and Old Wood. Lesnoi J. 14:89-93.
7. Bucur V (2006) Acoustics of Wood (Springer, Berlin, Germany) 2nd Ed.

14
Figure S1

a b
nail holes

heel
extension
original
original
wood
wood
Figure S2
10.0 ms
7.5 ms
5.0 ms
3.0 ms
2.0 ms
1.5 ms
1.0 ms
0.75 ms
0.5 ms
0.2 ms
200 180 160 140 120 100 80 60 40 20 0
13
C chemical shift (ppm)
Figure S3
10.0 ms
7.5 ms
5.0 ms
3.0 ms
2.0 ms
1.5 ms
1.0 ms
0.75 ms
0.5 ms
0.2 ms
200 180 160 140 120 100 80 60 40 20 0
13
C chemical shift (ppm)
Figure S4
10.0 ms
7.5 ms
5.0 ms
3.0 ms
2.0 ms
1.5 ms
1.0 ms
0.75 ms
0.5 ms
0.2 ms
200 180 160 140 120 100 80 60 40 20 0
13
C chemical shift (ppm)
Figure S5
10.0 ms
7.5 ms
5.0 ms
3.0 ms
2.0 ms
1.5 ms
1.0 ms
0.75 ms
0.5 ms
0.2 ms
200 180 160 140 120 100 80 60 40 20 0
13
C chemical shift (ppm)
Figure S6
10.0 ms
7.5 ms
5.0 ms
3.0 ms
2.0 ms
1.5 ms
1.0 ms
0.75 ms
0.5 ms
0.2 ms
200 180 160 140 120 100 80 60 40 20 0
13
C chemical shift (ppm)
Figure S7
10.0 ms
7.5 ms
5.0 ms
3.0 ms
2.0 ms
1.5 ms
1.0 ms
0.75 ms
0.5 ms
0.2 ms
200 180 160 140 120 100 80 60 40 20 0
13
C chemical shift (ppm)
Figure S8
10.0 ms
7.5 ms
5.0 ms
3.0 ms
2.0 ms
1.5 ms
1.0 ms
0.75 ms
0.5 ms
0.2 ms
200 180 160 140 120 100 80 60 40 20 0
13
C chemical shift (ppm)
Figure S9
10.0 ms
7.5 ms
5.0 ms
3.0 ms
2.0 ms
1.5 ms
1.0 ms
0.75 ms
0.5 ms
0.2 ms
200 180 160 140 120 100 80 60 40 20 0
13
C chemical shift (ppm)
Figure S10
10.0 ms
7.5 ms
5.0 ms
3.0 ms
2.0 ms
1.5 ms
1.0 ms
0.75 ms
0.5 ms
0.2 ms
200 180 160 140 120 100 80 60 40 20 0
13
C chemical shift (ppm)
Figure S11

0.8

signal intensity at 56 ppm

signal intensity at 22 ppm

0.3
0.6

0.2
0.4

0.1
0.2

0.0 0.0
0.25 0.5 1 2 4 8 0.25 0.5 1 2 4 8
contact time (ms) contact time (ms)
3
signal intensity at 75 ppm

signal intensity at 84 ppm

0.6

2
0.4

1
0.2

0 0.0
0.25 0.5 1 2 4 8 0.25 0.5 1 2 4 8
contact time (ms) contact time (ms)
0.5 1.5
signal intensity at 105 ppm
signal intensity at 89 ppm

0.4

1.0
0.3

0.2
0.5

0.1

0.0 0.0
0.25 0.5 1 2 4 8 0.25 0.5 1 2 4 8
contact time (ms) contact time (ms)
0.30 0.16
signal intensity at 153 ppm

signal intensity at 173 ppm

0.25
0.12
0.20

0.15 0.08

0.10
0.04
0.05

0.00 0.00
0.25 0.5 1 2 4 8 0.25 0.5 1 2 4 8
contact time (ms) contact time (ms)

AS cello 1707 AS violin 1717 AS neck 1725

AS cello 1731 DG violin 1741 neck extension
modern M1 modern M4 modern M5
 Figure S12

modern M1
modern M2
modern M3
modern M4
modern M5
neck extension
AS neck 1725
AS cello 1707
AS cello 1731
AS violin 1717
DG violin 1741
intensity

200 180 160 140 120 100 80 60 40 20 0

13
C chemical shift (ppm)
Figure S13
AC+HC
HC
CC lignin

30 28 26 24 22 20 18 16 14 12 10 70 68 66 64 62 60 58 56 54 52 50
13 13
C chemical shift (ppm) C chemical shift (ppm)

TC+HC AC+HC

CC

80 78 76 74 72 70 94 92 90 88 86 84 82 80
13 13
C chemical shift (ppm) C chemical shift (ppm)

lignin modern M1
modern M2
modern M3
modern M4
lignin
modern M5
neck extension
AS neck 1725
AS cello 1707
AS cello 1731

160 158 156 154 152 150 148 146 144 142 140 AS violin 1717
13
C chemical shift (ppm) DG violin 1741
 Figure S14

modern (mean) DG violin 1741 AS violin 1717 AS cello 1731

AS cello 1707 AS neck 1725 neck extension
1.6

1.4
normalized intensity

1.2

1.0

0.8

0.6

0.4

0.2

0.0
173 153 148 137 105 89 84 75 72 65 63 56 22
13
C chemical shift (ppm)
Figure S15

0.2
log (hemicellulose level)

0.1

-0.1

-0.2

-0.3 H4 H1

-0.4 Y 2= -1.693E-03 X+1.56E-02

R = 0.8564 H2
-0.5 H6

-0.6 H5
H3
-0.7
0 50 100 150 200 250 300 350

Year
Figure S16

modern maple M3

AS neck 1725
Figure S17

maple M1
maple M2
maple M3
(004) maple M4
maple M5
neck extension
AS neck 1725
intensity

AS cello 1707
AS cello 1731
AS violin 1717
DG violin 1741

22 23 24

2 (degree)
Figure S18

maple M1
maple M2
maple M3 (200)
maple M4
maple M5
neck extension
AS neck 1725
intensity

AS cello 1707
AS cello 1731
AS violin 1717
DG violin 1741

8 10 12 14 16 18
2 (degree)
 Figure S19

modern maple
AS neck 1725
AS violin 1717
AS cello 1731
intensity

180 170 160 150 140 130 120 30 25 20 15

200 180 160 140 120 100 80 60 40 20 0

13
C chemical shift (ppm)
Figure S20
modern maple

modern M1 modern M2 modern M3

5 mm

modern M4 modern M5

historical maple

AS cello 1707 AS cello 1731 AS violin 1717

5 mm
AS neck, 1725 DG violin 1741 neck extension
c. 1800