Thermogravimetry-Mass Spectrometry for Carbon
Nanotube Detection in Complex Mixtures
The MIT Faculty has made this article openly available. Please share
how this access benefits you. Your story matters.
Citation
Plata, Desirée L., Christopher M. Reddy, and Philip M.
Gschwend. 2012. "Thermogravimetry–Mass Spectrometry for
Carbon Nanotube Detection in Complex Mixtures."
Environmental Science & Technology 46 (22): 12254–12261.
As Published
http://dx.doi.org/10.1021/es203198x
Publisher
American Chemical Society
Version
Author's final manuscript
Accessed
Wed May 25 22:05:24 EDT 2016
Citable Link
http://hdl.handle.net/1721.1/78935
Terms of Use
Creative Commons Attribution-Noncommercial-Share Alike 3.0
Detailed Terms
http://creativecommons.org/licenses/by-nc-sa/3.0/
1
2
Thermogravimetry - mass spectrometry for
3
carbon nanotube detection in complex mixtures
4
5
6
7
8
9
10
11
12
13
14
15
16
Desirée L. Plata*1,2,3, Christopher M. Reddy1, and Philip M. Gschwend2
1
Marine Chemistry and Geochemistry, Woods Hole Oceanographic Institution, Woods
Hole, Massachusetts 02543
2
Civil and Environmental Engineering, Massachusetts Institute of Technology,
Cambridge, Massachusetts 02139
3
Civil and Environmental Engineering, Duke University, Durham, North Carolina, 27708
*Corresponding author: desiree.plata@duke.edu
17
18
ABSTRACT
19
In spite of the growth of the carbon nanotube (CNT) industry, there are no established
20
analytical methods with which to detect or quantify CNTs in environmental matrices.
21
Given that CNTs have relatively high thermal stabilities, we investigated the use of
22
thermal techniques to isolate and quantify single wall carbon nanotubes (SWCNTs). Test
23
materials included ten types of commercial SWCNTs, representative biological
24
macromolecules (bovine serum albumin and methylcellulose), soot, natural coastal
25
sediments, and SWCNT-amended sediments. Different SWCNTs exhibited widely
26
diverse degradation temperatures, and thermal analytical methods may require SWCNT-
27
type specific parameters. To improve quantification capabilities, evolved gases were
28
monitored by mass spectrometry. SWCNTs produced diagnostic ion ratios, reflective of
29
their high carbon and low hydrogen and oxygen contents. Current detection limits are
30
roughly 4 ugSWCNT per sample (e.g., 100 ugSWCNT g-1sediment and 40 mg sample), controlled
31
by interfering ions associated with the instrument’s non-airtight design. While future
32
modifications could improve this limitation, the current method is sufficient for
33
quantifying SWCNTs in laboratories and industrial sites where SWCNTs are handled.
34
Furthermore, the method shows promise to distinguish between incidental (e.g., soot) and
35
engineered (e.g., SWCNTs) nanoparticles, which is not possible with current state-of-the-
36
art techniques.
37
38
INTRODUCTION
39
Carbon nanotubes (CNTs) have generated an impressive list of potential and
40
demonstrated applications1-3, and the CNT industry is growing at a remarkable rate,
41
doubling nearly every two years4. Historically, no industrially important chemical has
42
been synthesized, used, and disposed of without some release to the environment. Thus,
43
environmental levels of CNTs will likely increase with growth of the industry. Present
44
CNT analytical techniques are limited to electron microscopy (EM) and spectroscopic
45
analyses of small (mg), relatively pure samples. Some attempts have been made to
46
measure CNTs in complex matrices5, but further work remains before recoveries
47
accurately reflect environmental concentrations. As a result, there is no way to monitor
48
the release of industrially synthesized CNTs or to gauge their transport and ultimate
49
biological exposure levels.
50
Successful isolation and detection of CNTs from environmental matrices (e.g., air,
51
water, and sediment) will rely on unique chemical or physical properties of the CNTs that
52
are distinct from other ubiquitous carbon forms. CNTs are cylinders of sp2-hybridized
53
carbon atoms arranged in repeating, aromatic rings. While variations in synthetic
54
approaches and post-fabrication handling result in a diverse class of materials6, the
55
majority of CNTs will exist as agglomerated, hydrophobic particles7. A physically and
56
chemically similar particle that is ubiquitous in the environment is black carbon (BC),
57
which can be produced from forest fires or as a by-product from incomplete combustion
58
of fossil fuels (i.e., natural or incidental nanoparticles, respectively)8-11.
59
Methods to measure BC in environmental samples have been under development
60
for approximately 30 years8, and a recent laboratory inter-comparison12 revealed that
61
thermal oxidation techniques (e.g., chemothermal oxidation at 375oC13) successfully
62
isolate BC soot from other BC materials14, such as wood char. Thermal methods (e.g.,
63
thermogravimetric analysis or TGA) are also used to determine the purity of commercial
64
carbon nanotubes, which typically contain several forms of carbon (e.g., single-walled
65
CNTs (SWCNTs), multi-walled CNTs (MWCNTs), “multi-shell” carbon15, fullerenes,
66
and amorphous carbon), as well as residual metal catalysts, and there is promise that
67
similar techniques may be used to isolate CNTs in environmental samples.
68
To establish the feasibility and limitations of such analyses for the quantification
69
of CNTs in the environment, we determined the thermal stability of several commercially
70
available SWCNT powders, two representative biological macromolecular materials (i.e.,
71
bovine serum albumin (BSA) and methylcellulose), and representative environmental
72
mixtures (e.g., urban dust, diesel particulate matter, marine sediments, and SWCNT-
73
amended sediments). To improve detection, online mass spectrometry (MS) was
74
employed to monitor gases evolved during temperature-programmed oxidations (TPOs)
75
and hydrogen-assisted thermal degradations (HATDs). Using unique mass loss
76
temperatures and evolved gases, we evaluated the promise of TGA-MS as a tool to
77
separate and quantify incidental and engineered organic matter.
78
79
80
EXPERIMENTAL METHODS
Samples. SWCNT powders were purchased from multiple manufacturers and
81
characterized by Plata et al.6 (Table 1). Non-SWCNT test materials included potentially
82
interfering materials: fullerenes (C60, Materials and Electrochemical Research
83
Corporation), MWCNTs (Nanocyl), and bovine serum albumin, as well a non-interfering
84
compound (methylcellulose). Un-amended environmental matrices and mixtures
85
included a sediment sample collected from Dorchester Bay, Massachusetts16, urban dust
86
(National Institute of Standards and Technology Standard Reference Material (NIST
87
SRM) 1649a), diesel particulate matter (NIST SRM 1650), and chestnut wood char12.
88
SWCNT-amended sediments were prepared by thoroughly mixing SouthWest
89
Nanotechnologies Purified (SWP) SWCNTs to pre-sieved, ground, and homogenized
90
Dorchester Bay sediment (DBS). Less-concentrated sediment samples were prepared by
91
serial dilutions with un-amended DBS.
92
Several materials that were pre-combusted and expected to be free of organic
93
carbon were also analyzed. These included ground (to a fine powder) Ottawa quartz
94
sand, quartz fiber filters (QFFs), and platinum sample pans.
95
Thermogravimetric analysis-mass spectrometry. TPOs and HATDs were
96
performed using a TA Instruments Q50 TGA interfaced to a Pfeiffer ThermoStar
97
quadrupole mass spectrometer (MS) via a heated capillary transfer line (200oC, 2 mL
98
min-1 sampling rate). Sample sizes were 3-3.3 mg for relatively pure samples (i.e., non-
99
sedimentary material) and 40-44 mg for environmental matrices (e.g., sediments and
100
101
soils). Masses were known to 0.001 mg.
TPOs were conducted in dry air (100 mL min-1) with a thermal program (80 oC,
102
10 min hold, 5oC min-1 to 900oC, 60 min hold). HATDs were conducted in ultra high
103
purity hydrogen (10 mL min-1) with a thermal program (80oC, 60 min hold, 5oC min-1 to
104
900oC, 60 min hold). Ion currents m/z 2, 4, 12, 14, 16, 18, and 20-79 were scanned in
105
60-s sequence. Between runs, platinum sample pans were washed with 25 % (v/v)
106
HNO3, rinsed with copious 18 M water, and combusted at 900oC in air for 1 hr.
107
The proposed method bears similarity to thermal-optical methods (i.e., thermal
108
optical transmittance/reflectance (TOT/R)) for distinguishing elemental carbon and
109
organic carbon (EC/OC), especially with regard to the use of sequential reactive
110
atmospheres (e.g., helium followed by a helium/oxygen mixture). In EC/OC, the
111
formation of higher thermal stability material from lower thermal stability material
112
during the analysis (i.e., “charring”) is acknowledged, accounted for optically, and
113
subsequently used to correct the data. In our method, we explicitly seek to avoid charring
114
altogether, as CNTs confound the optical correction technique17, and it will result in
115
interferences that introduce significant error.
116
Correction for TPOs: To account for instrumental drift between sample analyses,
117
ion ratios (X sample /Y sample, e.g., 18sample/44sample) were normalized to a standard run using
118
the ratio of the average ion current m/z 28 in the sample (28sample) to the average ion
119
current m/z 28 in the standard (28standard) (Eqn 1). The ion m/z 28 was selected as an
120
indicator of N2, which is the principle component of the dry air carrier gas that was
121
delivered at a constant rate.
122
123
18

18
  28

sample
sample
sample








44



 sample corrected 44sample  28standard 
(1)
125
126
127

Correction
for HATDs: Our particular TGA-MS was not designed to be airtight,
128
and small air leaks in the system resulted in interfering background signals due to N2, O2,
129
and Ar (m/z 28, 32, 40, respectively). The N2 and O2 interferences were not improved by
130
placing the TGA-MS in a sealed Ar-filled chamber. Ethene (m/z 28) is an important
131
mass formed by carbon nanotube (CNT) reaction with H2, and to lower interference from
132
N2, the ion current m/z 28 for each sample was background subtracted. The correction
133
was performed by subtracting the average ion current m/z 28 during a background
134
segment (no sample degradation) of each analysis (28background), normalized for signal
135
drift using ion current for Ar (m/z 40), from the ion current m/z 28 during the rest of the
136
thermal program (28sample) (see Eqn 2).
137
138
140

141
142
143
 40

sample
28sample,subtracted  28sample  28background 

40

 background 
(2)

Ion ratios were then normalized as described by Eqn 1.
As with any mathematical treatment, error could be introduced by the
144
manipulations detailed in equations 1 and 2. For equation 1, if the 28sample and 28standard
145
signals have relative errors smaller than or equal to the relative errors on 18sample and
146
44sample (which they typically do), then the increased error is small (e.g., 7% to 7.2%
147
error following correction). For equation 2, if 28sample is large relative to 28background, and
148
40sample and 40background have similar error magnitudes, then the error introduced by the
149
subtraction step is incrementally small compared to the uncorrected error (e.g., a typical
150
sample might go from 5% to 6% error following the correction step).
151
152
153
RESULTS AND DISCUSSION
When subjected to increasing temperatures in a reactive atmosphere,
154
carbonaceous materials degrade at characteristic temperatures14,18that are reflective of
155
their structures and potentially other factors, such as the presence of residual catalytic
156
metals mixed with the material. Each mass loss event (i.e., a peak in a differentiated
157
mass loss plot or thermogram) may then, presumably, be attributed to a distinct phase in
158
the sample.
159
TPO of SWCNTs. SWCNTs exhibited a diverse range of oxidative temperatures
160
from 375 to 540oC (Figure 1). All mass losses were due to the oxidation of carbonaceous
161
material (as confirmed by MS and discussed later), except in the case of the volatilization
162
of molybdenum oxides (around 690oC) from the SWP sample.
163
The majority of SWCNT powders produced two oxidation peaks: (1) a lower
164
thermal stability signal that is conventionally presumed to be due to the oxidation of
165
amorphous carbon and (2) a higher thermal stability signal that is conventionally
166
presumed to be due to the oxidation of SWCNTs. In some cases (e.g., SES), contaminant
167
phases were not amorphous carbon, but instead, are higher thermal stability material such
168
as MWCNTs (see Supporting Information (SI) for representative MWCNT stability,
169
which has a peak oxidation around 600oC) or multi-shell carbon (i.e., soot). In other
170
cases (e.g., Nano Am., MER, CNI, and ManX), only one phase was readily apparent in
171
the thermogram, and this could be the result of either very pure SWCNTs or the co-
172
oxidation of multiple phases. For example, a tested C60 fullerene sample oxidized around
173
525oC (see SI), and, in theory, would have been thermally indiscernible from all of the
174
Fe, Co, and Mo containing tubes (i.e., SWP, NanoAm, SES), as well as the Fe-containing
175
ManX sample.
176
Several factors can influence the apparent thermal stability of a material,
177
including the heating rate, oxygen supply, and oxygen’s access to the carbon surfaces
178
(e.g., surface area and material packing24,25). The temperature program, sample size, and
179
bulk oxygen supply were well controlled in these experiments. Thus, the observed
180
differences in SWCNT oxidation temperatures must be due to some intrinsic property (or
181
properties) of the SWCNTs. Other researchers15,21 have noted that the SWCNT oxidation
182
temperature is inversely related to the total metal content, suggesting that metals may
183
catalyze the oxidation. Some have suggested that this catalysis results from the formation
184
of metal oxides that generate concentrated microenvironments of oxygen in close
185
proximity to the SWCNT lattice22. However, it is also possible that different catalytic
186
metals may produce SWCNTs with distinct properties (e.g., strain due to chirality or the
187
average surface area of SWCNTs) that influence the oxidation temperatures of the
188
nanotube products. While our data showed no significant correlation between the metal
189
content and the oxidation temperature, clearly, the type of metal mixture was important to
190
the oxidation temperature of the SWCNTs (Figure 2; n=10, see SI for metal-specific
191
data). For example, all Ni/Y catalyzed SWCNTs had maximum thermal oxidation
192
temperatures around 420oC (413 – 428oC, excluding the CSI sample, which has a
193
complex thermogram), whereas Fe/Co/Mo catalyzed samples exhibited much higher
194
thermal stabilities (between 510 and 540oC). Furthermore, in spite of a large spread in Ni
195
content (range of 10 wt %) of Ni/Y catalyzed samples, the oxidation temperatures were
196
fairly consistent between samples. To our knowledge, there has been no systematic study
197
of metal catalyzed oxidation of recalcitrant carbon forms. However, NiO is known to be
198
an efficient oxidation catalyst23, and others have shown that different transition metals
199
have distinct effects on the degradation of some materials24,25.
200
Unfortunately, the crystalloid forms of the metals in these CNT samples were
201
unknown and are not typically determined by manufacturers. The metal content is likely
202
too low to obtain informative x-ray diffraction data. X-ray photoelectron spectroscopy
203
would yield information on the metals present in the top several nm of the sample, yet
204
carbonaceous coatings obscure many of the metals in SWCNT powders. As an exercise,
205
we assumed the metals were present as typical metal oxides6 and calculated the total
206
possible metal oxide content to determine if there was a correlation between oxide
207
content and oxidation temperature; there was no such correlation (SI).
208
The wide distribution in SWCNT thermal stability across catalyst types
209
demonstrates that no single thermal oxidation technique will quantify all SWCNTs.
210
Knowing the type of SWCNT likely to be present in a field or lab sample (e.g., Ni- vs.
211
Co- catalyzed), one could tailor the analytical method to isolate and quantify the desired
212
variety of SWCNTs.
213
214
215
Here, we will focus on developing techniques to quantify the SWP nanotubes,
which have been selected as a standard material for SWCNTs by NIST.
TPO of natural organic materials. Thermal isolation approaches require minimal
216
overlap in the thermal stability of SWCNTs and other natural materials. Several
217
representative “natural” materials were chosen, including samples that contain no BC or
218
SWCNT (i.e., a carbohydrate, methylcellulose, and a protein, BSA) and samples that
219
contain different types of BC (i.e., wood char and diesel particulate matter) (Figure 3a).
220
Methylcellulose had a low thermal stability and narrow thermal response window,
221
perhaps consistent with its relatively oxidized chemical character, and did not interfere
222
with SWP analysis. In contrast, the heteropolymer BSA had a broad oxidation range with
223
some mass loss that interfered with SWP. Molecules with high nitrogen contents, such as
224
proteins, have a tendency to char, forming higher stability materials during the thermal
225
treatment. Visible charring was evident following BSA oxidation, and environmental
226
matrices with high protein contents (e.g., cells or tissue) will likely suffer from significant
227
interference17. These may be overcome with tandem analyses (e.g., TGA-MS) that
228
enable the distinction between evolved gases due to char-derived BC and native BC in
229
the sample (e.g., water evolved from char but not soot).
230
Natural BC in a sample matrix may exist in several forms, including wood char or
231
BC soot. Wood char had a lower thermal stability than SWP (Figure 3a), perhaps due to
232
greater oxygen access resulting from the porous structure of char. In spite of the lower
233
thermal stability of wood char, there was still a slight interference with the oxidation
234
onset of the SWCNTs. BC soot, on the other hand, was thermally distinct from SWP
235
largely decomposing at temperatures >600oC (Figure 3a). The higher stability of the
236
former is likely due to the multi-layer structure of aromatic sheets present in BC soot
237
spheres26, which limits oxygen access to the dense interior of the particles and increases
238
the ultimate oxidation temperature. In contrast, SWCNTs are made up of hollow
239
cylinders that have higher surface-area-to-carbon ratios. This structural difference
240
between BC and SWCNTs may provide an opportunity to distinguish the two materials
241
using thermal techniques.
242
While it makes intuitive sense, experiments should be undertaken to prove that
243
differences in oxygen access to carbon influence relative thermal stabilities of materials.
244
For example, comparative experiments with variable carbon geometries (i.e., all sp2-
245
hybridized, aromatic sheets, but differing surface area to total carbon ratios; e.g.,
246
graphene, fullerenes, SWCNT, MWCNT, soot) but constant moles of carbon would
247
indicate relative oxidation kinetics (e.g., by monitoring evolution of reaction products).
248
(Nanomaterials would have to be metal-free to avoid confounding effects that result from
249
metal contamination). Reaction rates could be related to the oxygen-access parameters
250
(e.g., surface area as measured by nitrogen absorption) to confirm the existence of a
251
causative relationship. Alternatively, a single material could be exposed in different
252
geometries (e.g., constant mass, variable surface access or crucible shapes) or thicknesses
253
(e.g., variable mass, same area). Such investigations would contribute to many fields that
254
invoke thermal analysis, especially for carbonaceous materials.
255
TPO of complex environmental matrices. In more complex environmental
256
matrices, full resolution of the diverse carbon phases may not be possible using thermal
257
methods alone. Aerosol concentrates of urban dust had significant mass loss throughout
258
the thermogram to 600oC, co-oxidizing with most SWCNTs (Figure 3b). These
259
interferences likely arose from the presence of recalcitrant sooty or fly ash materials
260
derived from incomplete combustion processes in urban environments, such as diesel
261
exhaust or industrial emissions. Sediment collected from Dorchester Bay, MA had a
262
thermal distribution consistent with the presence of fewer combustion-derived, high-
263
thermal stability materials (Figure 3c). As a result, there was less interference with SWP
264
SWCNTs. Nevertheless, the co-oxidizing phases were not negligible, and an additional
265
dimension of information was sought to improve detection limits of SWCNTs in
266
environmental mixtures.
267
TPO-MS: Unique ion ratios from SWCNTs. Mass spectral analysis of the gases
268
evolved during the thermal program revealed important masses associated with each
269
compound class (see SI for spectra). When dry air was used as a reactive gas, all
270
oxidized test materials yielded an expected combustion product, CO2 (m/z 44). Carbon
271
monoxide (CO, m/z 28) is another common combustion product, but its evolution was
272
obscured by an interfering signal from N2 (also m/z 28), the principle component of dry
273
air. (Using a mixture of helium and oxygen (80% He, 20% O2) as a reactive gas later
274
alleviated this interference). In addition to the typical carbon-derived combustion
275
products, the oxidation of methylcellulose and BSA formed H2O (m/z 18) and either NO
276
and/or CH2O (m/z 30). This is consistent with the appreciable H, O, and N contents of
277
these materials. In contrast, oxidation of BC soot and SWCNTs produced no measurable
278
H2O, NO, or CH2O. Historically, geochemists have taken advantage of these differences
279
in chemical structure to identify distinct materials in complex mixtures. For example, BC
280
soot, with very few functional groups, has O/C and H/C ratios below 0.3, whereas most
281
chars have O/C ratios from 0.3-0.5 and H/C ratios from 0.3-0.7526. Thus, the two closely
282
related materials can be distinguished with elemental analysis, and it is possible that
283
SWCNT and BC may be identified by analogous means.
284
The variability in H-to-C and O-to-C ratios can be represented in a van Krevelen
285
diagram (H/C vs. O/C ratios). These were originally employed for the study of coals,
286
where low H/C and O/C materials were considered BC or soot-like. Using a similar
287
concept, we monitored depletions in H, O, and N relative to C (i.e., m/z 18 (due to H2O)
288
and 30 (due to CH2O or NO)) relative to m/z 44 (due to CO2)) and sought to use those as
289
an indicator of SWCNT abundance. As distinct materials oxidized at different
290
temperatures, ion ratios varied throughout the course of the thermal treatment. Thus, ion
291
ratios were calculated at the expected temperature of maximum SWCNT oxidation (here,
292
525oC). Natural sediments were amended with variable amounts of SWP SWCNTs.
293
SWCNT powders have very low H, O, and N, contents, and as a result, they had low
294
18/44 (i.e., H2O/CO2) and 30/44 (i.e., CH2O or NO/CO2) ion ratios relative to unamended
295
sediments (Figure 4). Increased SWCNT contents in amended sediments corresponded to
296
reduced 18/44 and 30/44 ion ratios, and these were detectable at concentrations as low as
297
30 ugSWCNT gsediment-1. While these detection limits are sufficient for many laboratory-
298
based experiments and indoor air concentrates where SWCNTs are being handled or
299
produced (50 ug m-3)27, they may not be sensitive enough to measure SWCNTs in
300
environmental samples (0-1200 ng gdry sediment-1)28,29. The current lower bound on the
301
detection of SWCNTs is limited by baseline CO2 (i.e., high m/z 44) derived from co-
302
oxidizing natural sedimentary organic matter and CO2 leaks.
303
As CO2 is a generic product of organic carbon oxidation, it will likely present a
304
barrier to the detection of a trace contaminant in any sample matrix. Therefore, we
305
investigated the use of an alternate reactive gas, H2, to potentially produce SWCNT-
306
specific masses.
307
HATD of SWCNTs, natural organic materials, and complex environmental
308
matrices. Heating SWCNTs in pure hydrogen, rather than an oxygen-containing
309
atmosphere, increased the maximum degradation temperature of all SWCNTs (Figure 5).
310
In some cases, the thermal program did not fully degrade the SWCNTs or associated
311
contaminants (ManX, CLex, SES), in spite of a prolonged exposure (60 min) to H2 at
312
900oC.
313
In addition to shifts in the thermal stability, there was increased resolution of
314
previously co-oxidizing phases in some samples (e.g., MER and NanoAm.). Phases that
315
were not observed in TPOs became apparent in HATDs, and this suggests potential value
316
in a type of multi-dimensional analysis (e.g., HATD and subsequent TPO). Additionally,
317
the increased resolution afforded by HATD demonstrates that the use of TPO alone is not
318
sufficient to determine SWCNT purity. TPO is often used exclusively to determine the
319
quantity of carbonaceous impurities in SWCNTs 21,22,30, where low thermal stability
320
peaks are considered amorphous carbon and the mass remaining at the end of the TPO is
321
considered to be metal oxide. While the approach appears to be valid in some cases (e.g.,
322
for the CNI sample), TPO does not necessarily separate all carbon-based SWCNT co-
323
products.
324
In addition to increased phase resolution in SWCNT powders, HATD reduced
325
interferences from natural materials (Figure 6). Methylcellulose and BSA’s principle
326
mass losses occurred below 350 and 450oC, respectively, well below SWCNT
327
degradation temperatures. However, BSA charred into a high thermal stability material
328
(as it did in TPO), and 20% of the initial mass remained after the HATD treatment.
329
Diesel particulate matter survived prolonged HATD treatment at 900oC (Figure 6a). The
330
small mass loss peak below 300oC in the diesel soot sample was attributed to labile
331
organic compounds associated with the BC particles and accounted for 19% of the total
332
mass of the sample (NIST SRM 1650b Certificate of Analysis reports 20.2 ± 0.4 % total
333
extractable materials). SWP SWCNTs degraded at lower temperatures than BC soot,
334
between 750 and 900oC. Thus, SWCNTs and BC soot might be separable via thermal
335
methods. While the effective range of this method (i.e., fraction of SWCNT vs. BC)
336
remains to be determined, the approach could provide a route to differentiate engineered
337
nanoparticles from natural or incidental nanoparticles, which has been challenging to
338
date.
339
This is promising for specific applications (e.g., laboratory air), but thermal
340
interferences in complex environmental mixtures persisted. In the case of urban dust,
341
mass losses overlapped with those of SWP SWCNTs to a greater extent than under TPO.
342
Coastal sediments had almost identical co-degradation with SWP SWCNTs during both
343
TPO and HATD. To overcome this limitation, diagnostic masses from SWCNT
344
degradation were sought to identify the materials in natural environments.
345
HATD-MS: Unique ion ratios from SWCNTs. The mass spectra of evolved
346
gases during HATD showed that all test materials produced ethene (m/z 28) (SI).
347
Methylcellulose and BSA also produced H2O (m/z 18), reflecting the enriched oxygen
348
content of these materials relative to SWCNTs, which produced no detectable H2O. Test
349
materials produced few other diagnostic ions. However, SWCNTs degraded at rates that
350
were sufficient to draw down the signal of the reactant gas, H2 (m/z 2). This
351
consumption of H2 and depletion of H2O was detectable in SWCNT-amended sediments
352
down to (at least) 100 ugSWCNT g-1sediment (Figure 7; i.e., both ratios 2/28 and 18/28 are
353
depleted in the SWCNT-amended sediment relative to pure sediment). This is at least ten
354
times too high to access the expected sedimentary concentrations of 0-1200ng g sediment -1
355
28,29
356
.
Readily accessible improvements. The detection limit in the HATDs was set by
357
either (1) co-degradation of sedimentary phases or (2) interference with the mass 28
358
signal due to an air leak in the TGA-MS, which was not designed to be airtight (see SI for
359
supporting data). In addition to designing a leak-free instrument, these limitations could
360
be overcome in at least three accessible ways.
361
First, enhanced mass spectral resolution would allow one to distinguish between
362
N2 and ethene using their mass defects. Ethene’s exact mass is 28.05 amu, whereas N2’s
363
exact mass is 28.01 amu. Many mass spectrometers readily distinguish these two ions,
364
but the mass spectrometers that are typically sold for TGA applications (i.e., residual gas
365
analyzers (RGA)) have mass resolutions of only about 1 amu! Clearly, the resolving
366
power of this RGA MS is insufficient to distinguish N2 and C2H4. In addition, the RGA
367
has detection limits on the order of micrograms, whereas other mass spectrometers have
368
sensitivities in the picogram range (especially for non-fragmenting molecules).
369
Second, evolved gases could be cryogenically collected or otherwise captured and
370
then chromatographically separated and quantified via traditional gas chromatographic
371
methods, including flame ionization detection or improved mass spectral analysis.
372
Detection limits in the “offline” systems orders-of-magnitude improvements over RGA
373
and could offer additional discerning information (e.g., isotope ratios). Also, it is
374
important to note that RGAs have limited sampling rates (2 mL min-1). With typical
375
TGA reactive gas flows between 10 and 100 mL min-1, one is only sampling 2-20% of
376
the gases evolved during the TGA furnace; the rest of the signal is simply flushed away.
377
With cyrogenic focusing, all or most of the signal due to SWCNTs could be quantified,
378
rather than being flushed away as it is in online RGA MS.
379
Lastly, existing software for deconvoluting mass loss and or mass spectral
380
information (e.g., such as those used in Rock-Eval14) could potentially enable
381
computational peak separation where thermal separation is not possible, especially where
382
the evolved gases are unique.
383
Were the issues completely resolved, given the current pan geometry (40
384
mgsediment) and a better mass spectrometer (with 0.1 to 1 ng sensitivity), one could
385
measure 2.5- 25 ngSWCNT g sediment -1 easily, well within the range of expected
386
environmental concentrations28,29.
387
Current applications of TGA-MS to detect SWCNTs in the environment.
388
Without additional modification, this method might successfully distinguish between
389
natural and engineered nanoparticles (i.e., BC soot and SWCNTs). Until now, this has
390
been a significant challenge, and it presents an impasse to quantification of engineered
391
nanoparticle release from industrial manufacturing sites. For example, several
392
groups27,31,32 have relied on light-scattering techniques to measure the release of
393
nanoparticles during CNT production or post-fabrication handling. In all cases, 30-50 nm
394
particles were detected, but it was unclear whether these were BC soot or CNTs or both.
395
Occasionally, EM is used to distinguish the two materials, but this method is neither
396
quantitative nor comprehensive. In addition, recent standard methods for measuring
397
carbon nanotubes by the National Institute of Standards and Health (NIOSH) recommend
398
the use of breathing zone air samplers (ca. 15 L min-1) to collect particles on a filter,
399
followed by elemental analysis33. Blanks collected in low exposure areas are meant to
400
provide a background measure of incidental or natural nanoparticles. However,
401
processes used to produce CNTs can co-produce BC soot34. Thus, carbon abundance
402
near CNT-handling and production sources will reflect both engineered nanoparticles and
403
incidental particles and the NIOSH method will not give a true representation of
404
production-derived engineering nanoparticles. In the case of CNTs and BC soot, their
405
mechanisms of toxicity are quite different, and distinguishing the two will be important.
406
To surmount this limitation, particles could be collected and subjected to HATD,
407
where SWCNTs would be degraded and quantified, while BC soot would be preserved
408
for subsequent quantification by TPO. Thus, thermal analysis under suitable gas
409
mixtures provides a low-cost, high-throughput method for the isolation and quantification
410
of carbon phases that are otherwise difficult to distinguish.
411
412
ACKNOWLEDGMENTS
413
We thank the Martin and Chesonis Families for support via MIT and the MIT Institute
414
for Soldier Nanotechnologies.
415
416
SUPPORTING INFORMATION AVAILABLE Metal reduction potentials, metal vs.
417
oxidation regression, mass spectra from TPO and HATD, m/z 28 source identification.
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
REFERENCES
[1]
de Heer, W.A. Nanotubes and the pursuit of applications. MRS Bulletin April
2004, 281-285.
[2]
Ball, P. Roll up for the revolution. Nature. 2001, 414, 142-144.
[3]
Shi Kam, N.W.; O’Connell, M.; Widsdon, J.A.; Dai, H. Carbon nanotubes as
multi-functional biological transporters and near-infrared agents for selective
cancer cell destruction. PNAS. 2005, 102(33), 11600-11605.
[4]
Thayer, A.M. Carbon nanotubes by the metric ton. Chemical & Engineering News
2007, 85 (46), 29-35.
[5]
Sobek, A.; Bucheli, T.D. Testing the resistance of single-and multi-walled carbon
nanotubes to chemothermal oxidation used to isolate soots from environmental
samples. Environ. Pollution 2009, 157, 1065-1071.
[6]
Plata, D.L., Gschwend, P.M., Reddy, C.M. Industrially synthesized single walled
carbon nanotubes: compositional data for users, environmental risk assessments,
and source apportionment. Nanotechnology 2008, 19, 185706.
[7]
Girifalco, L.A.; Hodak, M.; Lee, R.S. Carbon nanotubes, buckyballs, ropes, and a
universal graphitic potential. Phys. Rev. B 2000, 62, 12104-13110.
[8]
Goldberg, E.D. Black carbon in the environment: Properties and Distribution.
John Wiley & Sons, New York, NY. 1985.
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
[9]
Park, R.J.; Jacob, D.J.; Chin, M.; Martin, R.V. Sources of carbonaceous aerosols
over the United States and implications for natural visibility. J. Geophysical
Research 2003, 108 (D12), 4355.
[10]
Schmidt, M.W.I.; Noack, A.G. Black carbon in soils and sediments: Analysis,
distribution, implications, and current challenges. Global Biogeochemical Cycles
2000, 14 (3), 777-793.
[11]
Flores-Cervantes, D.X.; Plata, D.L.; MacFarlane, J.K; Reddy, C.M.; Gschwend,
P.M. Black carbon in marine particulate organic carbon: Inputs and cycling of
highly recalcitrant organic carbon in the Gulf of Maine. Marine Chemistry 2009,
113, 172-181.
[12]
Hammes, K.; et 32 co-authors. Comparison of quantification methods to measure
fire-derived (black/elemental) carbon in soils and sediments using reference
materials from soil, water, sediment and the atmosphere. Global Biogeochemical
Cycles 2007, 21, GB3016.
[13]
Gustafsson, Ö.; Haghseta, F.; Chan, C.; MacFarlane, J.; Gschwend, P.M.
Quantification of the dilute sedimentary soot phase: implications for PAH
speciation and bioavailability. Environ. Sci. Technol. 1997, 31, 203-209.
[14]
Poot, A.; Quik, J.T.K.; Veld, H.; Koelmans, A.A. Quantification methods of black
carbon: Comparison of Rock-Eval analysis with traditional methods. J.
Chromatography A 2009, 1216, 613-622.
[15]
Harutyunyan, A.R.; Pradhan, B.K.l Chang, J.; Chen, G.; Eklund, P.C. Purification
of single-wall carbon nanotubes by selective microwave heating of catalyst
particles. J. Phys. Chem. B 2002, 106, 8671-8675.
[16]
Accardi-dey, A.M.; Gschwend, P.M. Assessing the combined roles of natural
organic matter and black carbon as sorbents in sediments. Environ. Sci. Technol.
2002, 36, 21-29.
[17]
Doudrick, K.D.; Herckes, P.; Westerhoff, P.K. Quantification of carbon
nanotubes in environmental matrices using programmed thermal analysis.
Environ. Sci. Technol. Submitted 2011, planned publication with this issue.
[18]
Haines, P.J. Principles of Thermal Analysis and Calorimetry. The Royal Society
of Chemistry, Cambridge, UK. 2002.
[19]
Flores-Cervantes, D.X. Black carbon in the Gulf of Maine: New insights into
inputs and cycling of combustion-derived organic carbon. MIT Thesis. 2008.
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
[20]
López-Fonseca, R.; Landa, I.; Elizundia, U.; Gutiérrez-Ortiz, M.A., GonzálezVelasco, J.R. Thermokinetic modeling for the combustion of carbonaceous
particulate matter. Combustion and Flame 2006, 144, 398-406.
[21]
Chiang, I.W.; Brinson, B.E.; Smalley, R.E.; Margrave, J.L.; Hauge, R.E.
Purification and characterization of single-wall carbon nanotubes. J. Phys. Chem.
B 2001, 105, 1157-1161.
[22]
Herrera, J.E.; Resasco, D.E. In situ TPO/Raman to characterize single-walled
carbon nanotubes. Chem. Phys. Lett. 2003, 376, 302-309.
[23]
Merritt, D.A.; Hayes, J.M.; Des Marais, D.J. Carbon isotopic analysis of
atmospheric methane by isotope-ratio-monitoring gas chromatography-mass
spectrometry. J. Geophys. Res. 1995, 100 (D1), 1317-1326.
[24]
Bridgewater, J.D.; Lim, J.; Vachet, R.W. Transition metal-peptide binding studied
by metal-catalyzed oxidation reactions and mass spectrometry. Anal. Chem. 2006,
78 (7), 2432-2438.
[25]
Alexander, B.; Park, R.J.; Jacob, D.J.; Gong, S. Transition metal-catalyzed
oxidation of atmospheric sulfur: Global implications for the sulfur budget. J.
Geophys. Res. 2009, 114, D02309 (13 pp).
[26]
Preston, C.M.; Schmidt, M.W.I. Black (pyrogenic) carbon: a synthesis of current
knowledge and uncertainties with special consideration of boreal regions.
Biogeosciences 2006, 3, 397-420.
[27]
Maynard, A.D.; Baron, P.A.; Foley, M.; Shvedova, A.A., Kisin, E.R.; Castranova,
V. Exposure to carbon nanotube material: Aerosol release during the handling of
unrefined single-walled carbon nanotube material. J. Toxicol. Environ. Health A
2004, 67, 87-107.
[28]
Plata, D.L. Carbon nanotube synthesis and detection: Limiting the environmental
impact of novel technologies. Thesis (Ph.D.) Joint Program in Oceanography/
Applied Ocean Science and Engineering (Massachusetts Institute of Technology,
Dept. of Civil and Environmental Engineering; and the Woods Hole
Oceanographic Institution). 2009.
[29]
Koelmans, A.A.; Nowack, B.; Wiesner, M.R. Comparison of manufacturered and
black carbon nanoparticle concentration in aquatic sediments. Environmental
Pollution 2009, 157, 1110-1116.
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
[30]
SouthWest Nano Technologies, Product Data Sheet,
http://www.swnano.com/tech/sg.php. Accessed 15 Jan 2009.
[31]
Yeganeh, B.; Kull, C.M.; Hull, M.S.; Marr, L.C. Characterization of airborne
particles during production of carbonaceous nanomaterials. Environ. Sci. Technol.
2008, 42 (12), 4600-4606.
[32]
Bello, D.; Hart, A.J.; Ahn, K.; Hallock, M.; Yamamoto, N.l Garcia, E.J.; Wardle,
B.L.; Ellenbecker, M.J. Particle exposure levels during growth and subsequent
handling aligned carbon nanotube films. Carbon 2008, 46, 974-981.
[33]
NIOSH Current Intelligence Bulletin, Occupational Exposure to Carbon
Nanotubes and Nanofibers. December 2010.
[34]
Plata, D.L.; Hart, A.J.; Reddy, C.M.; Gschwend, P.M. Early evaluation of
environmental impacts of carbon nanotube synthesis by catalytic chemical vapor
deposition. Environ. Sci. Technol. 2009, 43 (21), 8367-8343.
551
552
Table 1. Single-walled carbon nanotubes used in this study
SWCNT supplier namea
Symbolb
Metal contentc (wt %)
SES Research
SES
0.47 Fe, 2.8 Co, 0.69 Mo
Carbon Solutions Inc.
CSI
22.4 Ni, 6.0 Y
CarboLex, Inc.
CLex
24.3 Ni, 5.0 Y
Materials & Electrochemical
MER
3.53 Ni, 10.5 Co
Research Corp.
Helix Material Solutions
Helix
15.3 Ni, 5.3 Y
d
Manufacturer X
ManX
1.21 Fe, 0.08 Mo
Nanocs Inc.
Nanocs
20.3 Ni, 4.2 Y
Nanostructured & Amorphous
NanoAm
0.24 Fe, 2.87 Co, 1.00 Mo
Materials
SouthWest NanoTechnologies
SWP
0.07 Fe, 4.1 Co, 6.9 Mo
Inc. (SWeNT, Inc.)
Carbon Nanotechnologies Inc.
CNI
22 Fe
a
Companies that manufactured single-walled carbon nanotubes (SWCNTs) for
commercial sale in the United States during September 2005 and March 2006 are
listed. bSymbols used in the tables, figures, and text of this paper do not represent
notations employed by the manufacturers. cAs determined in Plata et al.6
d
Manufacturer closed on April 26, 2006 and requested that we protect the identity of
the business name.
553
554
555
556
557
558
559
560
561
562
563
564
565
566
Figure 1. Differentiated oxidation thermograms of SWCNT powders. Each SWCNT
sample is catalyzed with either (a) Ni and Y, (b) Fe, Co, and Mo, or (c) Fe and Mo (Man
X), Fe (CNI), or Ni and Co (MER). Multiple mass loss peaks are due to the oxidation of
different phases, which are all carbonaceous (except for the highest thermal stability peak
in the SWP sample, which is due to the loss of molybdenum oxides). For thermograms
with two major peaks, the lower stability peak was attributed to amorphous carbon and
the higher thermal stability peak was attributed to SWCNTs, except in the case of SES.
567
568
569
570
571
Figure 2. Temperature of maximum SWCNT mass loss rate as a function of total metal
and type of metal. The metal content of each SWCNT powder is detailed in Table 1.
572
573
574
575
576
577
578
579
580
581
582
583
Figure 3. Differentiated oxidation thermograms of representative materials that are
likely to be found in the environment. (a) Relatively pure test materials exhibited a wide
range of oxidation temperatures, some of which overlapped with SWCNTs. (b, c)
Environmental matrices, such as urban aerosol concentrates and sediments, had broad
mass loss peaks that would obscure some, but not all, SWCNT oxidation temperature
ranges. Note that mass losses at low temperatures are partially due to water loss from
associated minerals. Thermograms for C60 and MWCNTs are available in the Supporting
Information.
584
585
586
587
588
589
590
591
592
593
594
Figure 4. Diagnostic ion ratios observed in thermal oxidations of SWCNT-amended
coastal sediments. Increasing amounts (weight percent indicated in figure legend) of
SWCNTs were added to sediment collected from Dorchester Bay, MA. (a) When
oxidized, pure SWCNTs yielded very little mass fragment 18 (due to H2O) and mass
fragment 30 (due to NO or CH2O), and depletions in the 18/44 and 30/44 ion ratios were
observed with increasing SWCNT content of the sediments. (b) The lower bound on
SWCNT detection was set by co-oxidizing carbonaceous material, which generated CO2
and increased the baseline signal of ion current m/z 44.
595
596
597
598
599
600
601
602
603
604
605
606
Figure 5. Differentiated thermograms of SWCNT powders subjected to hydrogenassisted thermal degradation. Each SWCNT sample is catalyzed with either (a) Ni and Y,
(b) Fe, Co, and Mo, or (c) Fe and Mo (Man X), Fe (CNI), or Ni and Co (MER). Multiple
mass loss peaks are due to the degradation of different phases, which are all
carbonaceous. Some phases that were not observed in the TPOs are apparent during
HATDs (e.g., MER, NanoAm). After 60 min of thermal decomposition in H2 at 900oC,
some material in the SWCNT powders was not fully degraded.
607
608
609
610
611
612
613
614
615
616
617
618
Figure 6. Differentiated hydrogen-assisted thermograms of representative materials that
are likely to be found in the environment. (a) Relatively pure test materials exhibited a
wide range of degradation temperatures. Natural materials did not overlap with the mass
loss of SWCNTs. (b, c) Environmental matrices, such as urban aerosol concentrates and
sediments, had broad mass loss peaks. Note that some mass losses are partially due to
mineral dehydration at low temperature. The sediment sample exhibited reduced
interference with SWCNT degradation ranges in HATD. Thermogram for MWCNTs are
available in the Supporting Information.
619
620
621
622
623
624
625
626
627
628
629
630
631
Figure 7. Diagnostic ion ratios observed in hydrogen-assisted thermal degradation of
SWCNT-amended coastal sediments. Increasing amounts (weight percent indicated in
figure legend) of SWCNTs were added to sediment collected from Dorchester Bay, MA.
(a) When subjected to hydrogen-assisted thermal degradation, pure SWCNTs yielded
very little mass fragment 18 (due to H2O) and consumed mass fragment 2 (due to H2),
and depletions in the 18/28 and 30/28 ion ratios were observed with increasing SWCNT
content of the sediments. (b) The lower bound on the detection of SWCNTs was set by
co-degrading organic material in the sediments, which generated ethene and increased the
baseline signal of ion current m/z 28. Note, that the system was not designed to be air
tight, and air leak corrections were made using equation 2 to account for m/z 28 signal
due to leaked N2.