Article

pubs.acs.org/ac

Feasibility of Terahertz Time-Domain Spectroscopy to Detect

Tetracyclines Hydrochloride in Infant Milk Powder
Jianyuan Qin, Lijuan Xie,* and Yibin Ying
College of Biosystems Engineering and Food Science, Zhejiang University, Hangzhou 310058, PR China
*
S Supporting Information

ABSTRACT: We report the use of terahertz time-domain spectroscopy (THz-TDS) to detect tetracyclines hydrochloride
(TCsH) in infant milk powder for the rst time. Four kinds of TCsH exhibited their unique spectral features in the region of
0.31.8 THz. The main spectral features of these TCsH were still detectable when mixed with infant milk powder with
concentrations at 1%50%, even in the ternary mixtures. The results from chemometrics analysis showed that qualitative and
quantitative detection of TCsH in infant milk powder could be successfully achieved. The residual predictive deviation (RPD)
values of all these TCsH models were all higher than 2, indicating these models were considered good and could be used in
screening purposes. The RPD values of TCH, DTCH, and CTCH models were higher than 3, which were considered excellent
for prediction purposes. These preliminary results indicated that THz-TDS combined with chemometrics analysis was suitable
for detecting the presence of TCsH residues in a food matrix.

T etracyclines (TCs) is the generic name of a family of

antibiotics with a broad antibacterial spectrum against
both Gram-positive and Gram-negative bacteria, including the
species Spirochete, Actinomyces, and Mycoplasma.1 These
antibiotics are used routinely in veterinary medicine for
prevention and control of disease and also as a kind of feed
additive to promote animal growth.2 Among them, tetracycline
(TC), oxytetracycline (OTC), doxycycline (DTC), and
chlortetracycline (CTC) are the most widely used and
commercially available. These TCs have similar molecular
structures, and their basic structure is four fused six-membered
rings.3 The overuse of these antibiotics in dairy and the failure
to follow good veterinary practices can lead to unsafe residue
levels in various milk or meat products, which will have
potential adverse eects on human health. In fact, the
occurrence of TCs residues in milk or meat products is a
cause of concern to consumers worldwide because of possible
toxic or allergic reactions or antimicrobial resistance.4 Now
large amounts of milk or meat products are consumed all over
the world, thus good quality control programs are of great
signicance to guarantee food safety for the consumers.
To ensure human food safety, several detection methods
have also been successfully proposed for determination of TCs
in various food matrixes. The most widespread used methods
are liquid chromatography operated under reverse phase mode5
and combined with various detection schemes68 because of
their superior sensitivity and selectivity. However, these
methods are time-consuming and expensive and require tedious
sample preparation procedures.9 Other methods based on
immunology assays including immunochemistry10 and bio-
sensors11 are time-saving, sensitive, and highly selective, but
they are very dicult to establish.12 In recent years,
spectroscopic methods1315 are also preliminarily used in the
elds of TCs residue detection, mainly because of their
nondestruction and generally simple sample pretreatment.
However, these traditional spectroscopic methods have their
own limitations. For example, near-infrared spectroscopy
(NIR) is usually suitable for qualitative analysis rather than
quantitative analysis due to its complex spectra for intra-
molecular overtone and combination bands of chemicals.16
Terahertz spectroscopy, being described from 0.1 to 10 THz
(30 m3 mm),17 is a type of electromagnetic radiation that
sandwiches between microwave and infrared regions. Com-
Received: August 21, 2014
Accepted: October 27, 2014
Published: October 27, 2014

2014 American Chemical Society 11750 dx.doi.org/10.1021/ac503212q | Anal. Chem. 2014, 86, 1175011757
Analytical Chemistry
pared to other spectroscopy, terahertz spectroscopy is
considered safe both for sample and operator due to its low-
energy and nonionizing. More importantly, terahertz spectros-
copy has advantages in molecule analysis since molecular
vibrations and rotations lie totally or partially in this region.18
Terahertz time-domain spectroscopy (THz-TDS),19,20 with the
ability to measure terahertz electric eld coherently and thus
provide time-resolved amplitude and phase information that are
related to the absorptive and refractive properties of the sample,
is the most commonly used terahertz spectroscopic techni-
que.17 Over the last 10 years, the unique characteristics17,19 of
terahertz spectroscopy combined with the advances in terahertz
system have led to the THz-TDS technique being used for an
increasing variety.2123
THz-TDS together with chemometrics analysis is increas-
ingly used for qualitative and quantitative detection.21,24
However, as far as we know, the research on antibiotics
detection using THz-TDS is quite limited. Redo-Sanchez et al.
reported the specic ngerprints of eight antibiotics in the
region of 0.12.0 THz. The main spectral features of two
antibiotics were still detectable when mixed with three food
matrixes at 50%.25 In 2013, Massaouti et al. applied THz-TDS
for the detection and discrimination of three antibiotics
residues in honey in the region of 0.56.0 THz, revealing
that antibiotic residues were traceable in highly absorptive
honey.26 Tetracyclines hydrochloride (TCsH) are cheap, stable,
easy to be absorbed, and easy to decompose into TCs.
Therefore, they are widely used in veterinary medicine instead
of TCs. However, until very recently, no qualitative and
quantitative detection of TCsH in a food matrix has been
reported. In this paper, we study the spectral features of TCsH
and quantify their concentrations in infant milk powder for the
rst time. Four kinds of TCsH were chosen as research
subjects. The results showed that THz-TDS combined with
chemometrics analysis was a powerful tool for detecting the
presence of TCsH residues in a food matrix.
EXPERIMENTAL SECTION
Chemicals and Materials. TCsH samples, including
tetracycline hydrochloride (TCH, >98.5%), oxytetracycline
hydrochloride (OTCH, >95105%), doxycycline hydrochlor-
ide (DTCH, >88% 94%), and chlortetracycline hydro-
chloride (CTCH, >97%) were purchased from Sangon Biotech
(Shanghai, China). Infant milk powder which mainly contains
fresh milk and some other nutrients such as rened vegetable
oil, saccharides, whey protein, olene acids, etc., was purchased
from Beingmate (Hangzhou, China). All these chemicals were
of USP grade and used without further purication.
Sample Preparation. The details of sample preparation
method were described in the Supporting Information.
Spectral Measurements. All THz-TDS measurements
were collected using a Z3-XL THz-TDS system. The details of
Z3-XL system, experimental conditions, and spectra collection
method were described in the Supporting Information.
Parameters Extraction. The echo pulse which was created
by an etalon artifact corresponding to multiple reections
between the two surfaces27 or the air/pellet interfaces of the
sample in the optical path28 usually occurred in the time-
domain spectrum of a thin sample tablet. Therefore, it must be
removed prior to obtain the frequency-domain spectrum. After
cutting the echo pulse in the time-domain spectrum, a
conventional fast Fourier transform (FFT) was applied to
11751
Article
obtain the spectral distribution of the terahertz pulse in the
frequency-domain17
E( ) = A() ei() = dt E(t ) eit
(1)
where, A() and () were the amplitude and phase of
electric eld, respectively. E(t) was the time-domain waveform.
The absorbance () could be extracted by comparing the
sample spectrum with the reference spectrum,
A
= ln S
AR (2)
where AR and AS were the amplitude of the reference and
sample signals, respectively.
Chemometrics Analysis. It is usually dicult to directly
perform qualitative and quantitative analysis because of the
complexity and high dimension of terahertz spectral data. In
this case, chemometrics analysis is needed to extract spectral
features and investigate the correlation between the spectra and
sample characteristics. Multivariate analysis techniques, includ-
ing principal component analysis (PCA), partial least-squares-
discriminant analysis (PLS-DA), and partial least-squares
regression (PLSR), which have proved to be eective in
many applications29,30 were used in the present study for
detection of TCsH in infant milk powder. The details of these
three multivariate analysis techniques and their application in
our present work were described in the Supporting
Information.
RESULTS AND DISCUSSION
Pure Compounds. The time-domain spectra of reference
and TCsH were plotted in Figure S1a (Supporting
Information). Every absorbance spectrum was the average of
three replicates (three pellet samples), i.e., nine spectra. As
mentioned above, the echo pulse appeared in the spectrum
after 15.8 ps (as shown by the dashed line). Therefore, the
time-domain signal must be cut before 15.8 ps. In the present
work, the eective time window was chosen as 015.8 ps. The
FFT was performed to get the frequency-domain spectra, and
they were plotted in Figure S1b (Supporting Information). It
could be found that TCsH had absorbance above 0.3 THz, and
their amplitudes were attenuated to almost zero above 2.0 THz.
The maximum absorbance (max) that could be measured
reliably with a given frequency-dependent dynamic range was
calculated according to the method developed by Jepsen et
al.,31 as shown in Figure S1c (Supporting Information).
Comparing between the absorbance of TCsH and the upper
limit of the detectable absorbance, max, it was clear that the
absorbance measurement was valid for frequencies below 1.8
THz (as shown by the dashed line). Thus, the eective spectral
width used throughout the paper was 0.31.8 THz. Figure 1
presents the absorbance spectra of TCsH and infant milk
powder. Note that these TCsH, though with similar molecular
structures, exhibited their own distinctive terahertz spectral
features. Specically, TCH (black line) exhibited a strong
absorption peak at 1.40 THz, a relatively weak absorption peak
at 0.79 THz, and a weaker absorption peak at 1.60 THz.
DTCH (blue line) showed two relatively distinct absorption
peaks at 1.20 and 1.57 THz and a relatively weak absorption
peak at 0.76 THz. CTCH (magenta line) had a strong
absorption peak at 1.34 THz, a relatively weak absorption peak
at 0.76 THz, and weaker absorption peaks at 1.00 and 1.62
THz. In contrast, OTCH (red line) demonstrated a broad
dx.doi.org/10.1021/ac503212q | Anal. Chem. 2014, 86, 1175011757
Analytical Chemistry
Figure 1. (a) Absorbance spectra of dierent kinds of TCsH and
infant milk powder. The black line represents TCH, red line represents
OTCH, blue line represents DTCH, magenta line represents CTCH,
and olive line represents infant milk powder.
absorption peak centered at 1.3 THz and very weak
absorption peaks at 1.48, 1.58, and 1.70 THz which were
Figure 2. Absorbance spectra of the binary mixtures containing dierent kinds of TCsH in infant milk powder: (a) TCH, (b) OTCH, (c) DTCH,
and (d) CTCH.
11752
Article
dicult to identify. Comparing the absorbance spectra of these
TCsH in the region of 0.31.8 THz, we found that TCH,
DTCH, and CTCH showed their own distinct absorption
peaks, while OTCH had broad and very weak absorption peaks.
These unique absorption peaks of TCsH were likely due to
their dierent crystalline structures and diverse vibrational
motions.32 When mixed with a food matrix, their absorption
peaks would be slightly dierent due to destructions of their
crystalline structure and the interactions between TCsH and
the matrix. On the basis of their spectral ngerprints, dierent
kinds of TCsH could be easily discriminated, even in a food
matrix. Infant milk powder (olive line) had a broad terahertz
absorbance prole, and no obvious absorption peaks were
observed, as shown in Figure 1. This result agreed well with the
spectral features of milk powder reported before,25 which also
indicated that the spectral feature of infant milk powder was
dominated by its main ingredient (fresh milk), while the spectra
features of other nutrients3335 were ill identied due to their
lower content than that of fresh milk.
Binary Mixtures. In this section, the measurements of
binary mixtures were performed to investigate the possibility of
THz-TDS for qualitative and quantitative detection of TCsH in
infant milk powder. The absorbance spectra of binary mixtures
were plotted in Figure 2. Every absorbance spectrum was the
dx.doi.org/10.1021/ac503212q | Anal. Chem. 2014, 86, 1175011757
Analytical Chemistry Article

Figure 3. Two-dimensional (2D) PCs score plots using the rst two PCs for TCsH groups: (a) TCH group and OTCH group, (b) TCH and
DTCH groups, (c) TCH and CTCH groups, (d) OTCH and DTCH groups, (e) OTCH and CTCH groups, and (f) DTCH and CTCH groups.

average of three replicates (three pellet samples), i.e., nine
spectra. Figure 2a depicts the absorbance spectra of binary
mixtures containing TCH in infant milk powder. Because of the
strong absorbance of infant milk powder, the spectral features
of TCH with concentration below 3% were ill identied.
However, as concentration increasing to 5%, the absorption
peaks at 0.79, 1.40, and 1.60 THz could be observed, especially
11753
the strong peak at 1.40 THz. The inset in Figure 2a shows that
the absorbance spectrum of each concentration could be well
distinguished from each other, and the absorbance increased
with the concentration increasing. Figure 2b shows the
absorbance spectra of binary mixtures containing OTCH in
infant milk powder. As we could observed in the absorbance
spectrum of OTCH, there were only broad and very weak
dx.doi.org/10.1021/ac503212q | Anal. Chem. 2014, 86, 1175011757
Analytical Chemistry
absorption peaks. The enlarged views of any frequency bands
showed that these absorbance spectra crossed together and no
clear correlation between the absorbance and concentration
was observed (data not shown), except for the region of 1.34
1.41 THz, as shown in the inset in Figure 2b. In the region of
1.341.41 THz, the absorbance spectrum of each concen-
tration could be distinguished from each other, and the
absorbance increased with the concentration increasing. The
absorbance spectra of binary mixtures containing DTCH in
infant milk powder were plotted in Figure 2c. The absorbance
spectra with concentration above 3% exhibited ngerprint
features at 0.76, 1.20, and 1.57 THz. It was interesting to note
that there was a relatively strong absorption peak near 1.37
THz, and it had no correlation with the change of
concentration. At this moment, the origin of this peak was
still not very clear. Previous work showed that terahertz spectra
were sensitive to the vibrational motions of the molecular
structure and its interactions with the surrounding solid
medium.32 The interaction of DTCH and infant milk powder
existed in their mixture, which would lead to a change in
terahertz signal of the sample. Calculating the refractive indices
of these samples, a corresponding change in refractive indices of
these samples was observed (data not shown). Therefore, we
suggested that the appearance of absorption peak at 1.37 THz
might be due to the interaction of particles between DTCH
and infant milk powder, and the corresponding change in the
refractive index of the mixture with DTCH in the infant milk
powder. Additionally, the absorbance spectrum with concen-
tration at 1% did not show any absorption peaks, but it could
still be identied from the 0% one (infant milk powder).
However, the enlarged views of any frequency bands showed
that these absorbance spectra crossed together and no clear
correlation between the absorbance and concentration was
observed (data not shown). Therefore, we concluded that it
was dicult to distinguish the absorbance spectra of binary
mixture with DTCH at dierent concentrations in infant milk
powder and there was no clear correlation between the
absorbance and concentration at any frequency bands in the
region of 0.31.8 THz. The main spectral features of CTCH
were still identiable when mixed with infant milk powder, as
shown in Figure 2d. The spectral features of samples with
concentrations ranging from 1% to 10% were dominated by the
strong absorbance of infant milk powder, and no obvious
absorption peaks were observed. However, as concentration
increasing to 15%, the strong peaks at 1.34 THz, and the weak
peaks at 0.76, 1.00, and 1.62 THz appeared. In addition, as
shown in the inset in Figure 2d, the absorbance spectrum of
each concentration could be distinguished from each other.
Above 10%, the absorbance increased with the concentration
increasing. While below 10%, there was no obvious regularity
between the absorbance and concentration, which might be due
to the relatively weak spectral features of CTCH and the strong
absorbance of infant milk powder.
Comparing the absorbance spectra of TCsH in infant milk
powder in the region of 0.31.8 THz, it indicated that all
TCsH were still identiable when mixed with infant milk
powder. For lower concentrations, the spectral features were
often dominated by the strong absorbance of infant milk
powder, thus the absorption peaks could not be identied
clearly. However, the absorbance spectra with lower concen-
trations could still be separated from one another according to
their dierent absorbance. This preliminary study demonstrated
11754
Article
that we could achieve qualitative detection of TCsH in a food
matrix with strong absorbance.
For a preliminary study, PCA was performed to classify
binary mixtures containing dierent kinds of TCsH in infant
milk powder using the spectral data as in Figure 2, that was,
every absorbance spectrum was the average of three replicates
(three pellet samples), i.e., nine spectra. According to the result
from outlier detection, two outliers were found, 5% CTCH and
10% CTCH, and they were removed from the sample set. The
loading spectra for the rst three PCs plotted in Figure S2a
(Supporting Information) were analyzed to investigate the bias
for spectral classication between TCsH groups. PC1 explained
73.31% of the total variances and the highest loadings were
found at 1.40 and 1.70 THz mainly related to the spectral
features of TCH and OTCH, respectively. PC2 accounted for
13.07% of the variation, and the highest loadings were found at
0.54, 1.20, and 1.57 THz mainly related to the spectral features
of DTCH, 1.40 THz mainly related to the spectral features of
TCH. PC3 accounted for 7.60% of the variation, and the
highest loadings were found at 0.54 and 1.57 THz mainly
related to the spectral features of DTCH, 0.79 and 1.40 THz
mainly related to the spectral features of TCH, and 1.62 THz
mainly related to the spectral features of CTCH. All these
loadings were associated with the spectral features of TCsH
that helped to explain the bias for the classication of TCsH
groups. Figure S2b in the Supporting Information shows the
three-dimensional (3D) PCs score plots of all TCsH groups
using the rst three PCs which accounted for 93.98% of the
total variation. To get a more clear observation, the two-
dimensional (2D) PCs score plots of every two TCsH groups
using the rst two PCs vectors were plotted in Figure 3af and
analyzed in detail. Complete separations were observed
between TCH and OTCH groups (Figure 3a) and TCH and
DTCH groups (Figure 3b). One reason for the clear separation
between TCH and OTCH groups could be that TCH exhibited
strong spectral features while OTCH demonstrated broad and
very weak spectral features. Such a clear separation between
TCH and DTCH groups, however, might be due to the
completely dierent positions of their unique absorption peaks.
In Figure 3c,e,f, there were not complete but only general
separations between every two groups. Samples with
concentrations at 1% and 3% from CTCH group could not
be accurately classied, which might be due to the close
positions of absorption peaks of TCH and CTCH, the
relatively weak spectral features of DTCH, and hardly any
spectral features of CTCH samples with concentrations less
than 3%. The relatively poor separation was observed between
OTCH and DTCH groups with a few samples overlapped
(Figure 3d). Samples with concentrations less than 10% could
not be clearly classied, and we attributed this to the relatively
weak spectral features of DTCH and the broad and very weak
spectral features of OTCH.
PLS-DA was also applied to quantitatively classify TCsH
groups using the spectral data as in Figure 2; that was, every
absorbance spectrum was the average of three replicates (three
pellet samples), i.e., nine spectra. Table 1 shows the statistics
classication for TCsH groups using PLS-DA models. The
results indicated that the classication accuracies of the PLS-DA
model for TCsH groups ranged from 83.33 to 100%. The
model of TCH and DTCH groups performed the best, and the
classication accuracy was 100%. While the model of OTCH
and DTCH groups showed the lowest classication accuracy
(83.33%). The results from PLS-DA models were in
dx.doi.org/10.1021/ac503212q | Anal. Chem. 2014, 86, 1175011757
Analytical Chemistry Article

Table 1. Statistics Classication for TCsH Groups Using PLSR models were plotted in Figure 4, and their corresponding
PLS-DA Models descriptive statistics were summarized in Table 2. Good
descriptive statistics were obtained for all TCsH, with higher
groups correctly classied in calibration (%)
correlation coecient (rc > 0.95 and rcv > 0.93) and lower root-
TCH and OTCH 91.67 mean-square error (RMSEC < 4.8%, RMSECV < 5.9%,
TCH and DTCH 100
RMSEP < 6.0%). CTCH model performed the best with the
TCH and CTCH 86.36
highest RPD value (RPD = 4.67), rc = 0.9933, rcv = 0.9857,
OTCH and DTCH 83.33
RMSEC = 1.94%, RMSECV = 2.82%, and RMSEP = 3.63%.
OTCH and CTCH 90.90
The DTCH model came second, and the TCH model was
DTCH and CTCH 90.90
third. The OTCH model came in last with the lowest value of
RPD (RPD = 2.83), rc = 0.9596, rcv = 0.9529, RMSEC = 4.71%,
accordance with those from PCA, indicating the possibility of RMSECV = 5.09%, and RMSEP = 5.99%. The relatively low
classifying binary mixtures containing dierent kinds of TCsH RPD value of OTCH model might be due to the broad and
in infant milk powder using THz-TDS together with chemo- very weak spectral features of OTCH. However, the RPD
metrics analysis. values of all these TCsH models were all higher than 2,
In order to investigate the applicability of THz-TDS for indicating these models were considered good and could be
quantitative determination of TCsH in infant milk powder, we used in screening purposes.36 The RPD values of TCH,
built PLSR models for predicting the concentrations of TCH, DTCH, and CTCH models were both higher than 3, which
OTCH, DTCH, and CTCH in infant milk powder, were considered excellent for prediction purposes.37 Despite
respectively. Every absorbance spectrum was the average of the limited sensitivity of this technique, these results highlight
three measurements for one pellet sample, i.e., three spectra. In the potential of THz-TDS combined with chemometrics
each model, 39 spectra were used as a calibration set and 13 analysis to be used as a rapid, in-line technique. A terahertz
spectra were used as a prediction set. According to the result spectrometer with higher signal-to-noise ratio (SNR), an
from outlier detection, no outliers were found. The results of optimum modeling approach, or signal amplication technol-

Figure 4. Correlation statistics between the actual values and calculated values of dierent kinds of TCsH in infant milk powder using PLSR: (a)
TCH, (b) OTCH, (c) DTCH, and (d) CTCH.

11755 dx.doi.org/10.1021/ac503212q | Anal. Chem. 2014, 86, 1175011757

Analytical Chemistry
Table 2. Descriptive Statistics of the Multivariate Calibration Models Built by PLSR
model rc RMSECa rcv
TCH 0.9649 4.40 0.9520
OTCH 0.9596 4.71 0.9529
DTCH 0.9630 4.51 0.9383
CTCH 0.9933 1.94 0.9857
a
Unit used, %.
ogy (such as using metamaterial as a sensor38) will help to
reduce the detection limit in food quality control.
Ternary Mixtures. In order to detect TCsH multiresidues,
we investigated ternary mixtures containing two dierent kinds
of TCsH in infant milk powder. TCH and DTCH were chosen
as samples since they had more notable absorption peaks, and
the positions of these peaks were far from each other. The
resulting absorbance spectra were plotted in Figure 5. All these
Figure 5. Absorbance spectra of the ternary mixtures containing TCH
and DTCH in infant milk powder.
absorbance spectra showed spectral features of both TCH and
DTCH (as shown by the dashed arrows). The absorption peak
near 1.40 THz was derived from TCH, while absorption peaks
near 1.20 and 1.57 THz were come from DTCH. However, the
absorption peak near 0.77 THz was considered as a
combination of 0.79 THz from TCH and 0.76 THz from
DTCH. Additionally, each concentration could be distin-
guished from each other very well, and the positions of
absorption peaks slightly shifted to higher frequency as the
concentration increasing (as shown by the blue arrow). The
reason for this shift was still not very clear. However, the work
by Balan et al.39 and Burnett et al.40 provided us with
theoretical and experimental foundation to explain the peaks
shift to some extent. They adopted an eective medium theory
to describe the absorption properties of powder particle
embedded in a matrix and found that particle shape had a
dramatic inuence on the calculated spectra with some
vibrational modes shifting. In our work, the compounds were
ionic when embedded in infant milk powder. The summed
spectra would not be the simple linear additivity of the
compounds and matrix but peak shift might be observed in
some case, as could be approximated by the eective medium
approximation. Thus, we initially contributed this slight shift to
the vibrational modes resulting from the change of particle
shape after the mixing. These initial results implied that THz-
11756
Article
RMSECVa RMSEPa RPD factors
5.13 5.59 3.04 2
5.09 5.99 2.83 2
5.81 5.10 3.33 3
2.82 3.63 4.67 6
TDS could be used to detect TCsH multiresidues in infant milk
powder.
CONCLUSIONS
The applicability of THz-TDS combined with chemometrics
analysis to detect TCsH in infant milk powder is present in this
study. The results provide spectral features of TCsH in the
region of 0.31.8 THz for the rst time. TCH, DTCH, and
CTCH show their own distinct absorption peaks, while OTCH
has broad and very weak absorption peaks. On the basis of their
spectral ngerprints, dierent kinds of TCsH could be easily
discriminated. The results from PCA and PLS-DA models of
binary mixtures containing dierent kinds of TCsH in infant
milk powder show that the dierences of each mixture are
obvious enough to generally or even completely separate one
from another, except for OTCH and DTCH groups. With the
help of PLSR, the concentration of TCsH in infant milk
powder is successfully quantied. The RPD values of all these
TCsH models were all higher than 2, indicating these models
were considered good and could be used in screening
purposes.36 The RPD values of TCH, DTCH, and CTCH
models were higher than 3, which were considered excellent for
prediction purposes.37
Despite the limited sensitivity of the current technique, the
current work shows its ability to detect TCsH residues in infant
milk powder. Moreover, given that terahertz radiation could
penetrate commonly used packaging materials, these results
open up the feasibility of this technique to detect the presence
of antibiotics in sealed packaged foods. However, the work
reported here is just a preliminary study, and further study will
be required to improve the precision and robustness of the
prediction model. The development of novel terahertz
spectrometers with higher SNR as well as the optimization of
chemometrics analysis for spectral information will realistically
improve the detection capability in the next years. Therefore,
we expect that THz-TDS will improve its detection limit in the
near future to the point of being suitable for commercial
applications in food quality control.
ASSOCIATED CONTENT
* Supporting Information
S
Additional information as noted in text. This material is
available free of charge via the Internet at http://pubs.acs.org.
AUTHOR INFORMATION
Corresponding Author
*Phone: 0086-571-88982820. Fax: 0086-571-88982820. E-mail:
ljxie@zju.edu.cn.
Notes
The authors declare no competing nancial interest.
dx.doi.org/10.1021/ac503212q | Anal. Chem. 2014, 86, 1175011757
Analytical Chemistry

Article

ACKNOWLEDGMENTS (34) Markelz, A.; Roitberg, A.; Heilweil, E. Chem. Phys. Lett. 2000,
320, 4248.
The authors gratefully acknowledge the nancial support (35) Upadhya, P.; Shen, Y.; Davies, A.; Linfield, E. Vib. Spectrosc.
provided by the National Natural Science Foundation of 2004, 35, 139143.
China (Grant No. 31471410). (36) Chang, C.-W.; Laird, D. A.; Mausbach, M. J.; Hurburgh, C. R.

Soil. Sci. Soc. Am. J. 2001, 65, 480490.

REFERENCES (37) Cozzolino, D.; Kwiatkowski, M.; Parker, M.; Cynkar, W.;
Dambergs, R.; Gishen, M.; Herderich, M. Anal. Chim. Acta 2004, 513,
(1) Michalova, E.; Novotna, P.; Schlegelova, J. Vet. Med.-Czech. 2004, 7380.
49, 79100. (38) Wu, X.; Quan, B.; Pan, X.; Xu, X.; Lu, X.; Gu, C.; Wang, L.
(2) Sczesny, S.; Nau, H.; Hamscher, G. J. Agric. Food Chem. 2003, 51, Biosens. Bioelectron. 2012, 42, 626631.
697703. (39) Balan, E.; Saitta, A. M.; Mauri, F.; Calas, G. Am. Mineral. 2001,
(3) Zhou, J.; Xue, X.; Li, Y.; Zhang, J.; Chen, F.; Wu, L.; Chen, L.; 86, 13211330.
Zhao, J. Food Chem. 2009, 115, 10741080. (40) Burnett, A. D.; Kendrick, J.; Russell, C.; Christensen, J.;
(4) van den Bogaard, A. E.; Stobberingh, E. E. Int. J. Antimicrob. Cunningham, J. E.; Pearson, A. R.; Linfield, E. H.; Davies, A. G. Anal.
Agents 2000, 14, 327335. Chem. 2013, 85, 79267934.
(5) Oka, H.; Ito, Y.; Matsumoto, H. J. Chromatogr., A 2000, 882,
109133.
(6) Fritz, J. W.; Zuo, Y. Food Chem. 2007, 105, 12971301.
(7) Schneider, M. J.; Braden, S. E.; Reyes-Herrera, I.; Donoghue, D. J.
J. Chromatogr., B 2007, 846, 813.
(8) Blasco, C.; Corcia, A. D.; Pico, Y. Food Chem. 2009, 116, 1005
1012.
(9) Moreno-Bondi, M. C.; Marazuela, M. D.; Herranz, S.; Rodriguez,
E. Anal. Bioanal. Chem. 2009, 395, 921946.
(10) Conzuelo, F.; Campuzano, S.; Gamella, M.; Pinacho, D. G.;
Reviejo, A. J.; Marco, M. P.; Pingarron, J. M. Biosens. Bioelectron. 2013,
50, 100105.
(11) Rebe Raz, S.; Bremer, M. G.; Haasnoot, W.; Norde, W. Anal.
Chem. 2009, 81, 77437749.
(12) Chafer-Pericas, C.; Maquieira, A .; Puchades, R. TrAc, Trend.
Anal. Chem. 2010, 29, 10381049.
(13) Sivakesava, S.; Irudayaraj, J. J. Dairy Sci. 2002, 85, 487493.
(14) Leypold, C. F.; Reiher, M.; Brehm, G.; Schmitt, M. O.;
Schneider, S.; Matousek, P.; Towrie, M. Phys. Chem. Chem. Phys. 2003,
5, 11491157.
(15) Rodrguez, N.; Real, B. D.; Cruz Ortiz, M.; Sarabia, L. A.;
Herrero, A. Anal. Chim. Acta 2009, 632, 4251.
(16) Auer, M. E.; Griesser, U. J.; Sawatzki, J. J. Mol. Struct. 2003,
661662, 307317.
(17) Zhang, X. C.; Xu, J. Introduction to THz Wave Photonics;
Springer: New York, 2009.
(18) Pickwell, E.; Wallace, V. J. Phys. D Appl. Phys. 2006, 39, R301
R310.
(19) Ferguson, B.; Zhang, X. C. Nat. Mater. 2002, 1, 2633.
(20) Baxter, J. B.; Guglietta, G. W. Anal. Chem. 2011, 83, 43424368.
(21) Qin, J.; Ying, Y.; Xie, L. Appl. Spectrosc. Rev. 2013, 48, 439457.
(22) Joyce, H. J.; Docherty, C. J.; Gao, Q.; Tan, H. H.; Jagadish, C.;
Lloyd-Hughes, J.; Herz, L. M.; Johnston, M. B. Nanotechnology 2013,
24, 214006.
(23) Song, H.-J.; Ajito, K.; Muramoto, Y.; Wakatsuki, A.; Nagatsuma,
T.; Kukutsu, N. Electron. Lett. 2012, 48, 953954.
(24) Shen, Y. C. Int. J. Pharmaceut. 2011, 417, 4860.
(25) Redo-Sanchez, A.; Salvatella, G.; Galceran, R.; Roldos, E.;
Garca-Reguero, J. A.; Castellari, M.; Tejada, J. Analyst 2011, 136,
17331738.
(26) Massaouti, M.; Daskalaki, C.; Gorodetsky, A.; Koulouklidis, A.
D.; Tzortzakis, S. Appl. Spectrosc. 2013, 67, 12641269.
(27) Ueno, Y.; Rungsawang, R.; Tomita, I.; Ajito, K. Anal. Chem.
2006, 78, 54245428.
(28) Kim, J.; Chung, H.; Arnold, M. A.; Namkung, H. Anal. Chem.
2013, 85, 36743681.
(29) Cozzolino, D.; Smyth, H. E.; Gishen, M. J. Agric. Food Chem.
2003, 51, 77037708.
(30) Alexandrakis, D.; Downey, G.; Scannell, A. G. M. J. Agric. Food
Chem. 2008, 56, 34313437.
(31) Jepsen, P. U.; Fischer, B. M. Opt. Lett. 2005, 30, 2931.
(32) Wu, H.; Heilweil, E. J.; Hussain, A. S.; Khan, M. A. Int. J. Pharm.
2007, 343, 148158.
(33) Li, J. Appl. Spectrosc. 2010, 64, 231234.

11757 dx.doi.org/10.1021/ac503212q | Anal. Chem. 2014, 86, 1175011757