Anal Bioanal Chem (2002) 374 : 561–565
DOI 10.1007/s00216-002-1503-8
O R I G I N A L PA P E R
Janina Ohnsmann · Guillermo Quintás ·
Salvador Garrigues · Miguel de la Guardia
Determination of caffeine in tea samples
by Fourier transform infrared spectrometry
Received: 30 April 2002 / Revised: 17 June 2002 / Accepted: 24 June 2002 / Published online: 10 September 2002
© Springer-Verlag 2002
Abstract A sustainable and environmentally friendly pro-
cedure has been developed for the FTIR determination of
caffeine in tea leaf samples. The method is based on the
extraction with ammonia and CHCl3 and direct determi-
nation of caffeine on the chloroform extracts using peak
height absorbance measurements at 1658.5 cm–1 and ex-
ternal calibration. The method provides a sensitivity of
0.2142 absorbance units mg–1 mL and a limit of detection
of 1 mg L–1, corresponding to 0.002% m/m caffeine in tea
leaves. As compared with a reference procedure, based on
UV absorbance measurement at 276 nm after low pres-
sure column chromatography, the developed procedure re-
duces the consume of CHCl3 by a factor of 10, that of
NH4OH by a factor of 20 and avoids the use of diethyl
ether and Celite. The time required to do the analysis of a
sample is 15 minutes as compared with the 6 hours for the
reference one.
Keywords Caffeine · Tea · FTIR · Solvent extraction
Introduction
Caffeine is the most important alkaloid present in tea and
thus, the determination of this stimulatory effect alkaloid
is an important factor in the quality control of tea.
Caffeine concentrations vary from one to another type
of tea with typical concentration levels of 3.2% (m/m) for
black tea and concentrations varying from 1 to 3% (m/m)
in green tea [1].
Caffeine has been determined in drinks by different an-
alytical techniques such as UV-VIS spectrophotometry [2,
Janina Ohnsmann is on leave from the University of Mainz
(Germany)
J. Ohnsmann · G. Quintás · S. Garrigues (✉) · M. de la Guardia
Department of Analytical Chemistry, University of Valencia,
Edifici Jeroni Muñoz, 50th Dr. Moliner,
46100 Burjassot, Valencia, Spain
e-mail: salvador.garrigues@uv.es
3, 4] or amperometry [5] and most frequently by using
separation methods like high performance liquid chro-
matography (HPLC) [6, 7, 10], ion chromatography [11],
capillary electrophoresis [12, 13, 14], as well as high per-
formance thin layer chromatography (HPTLC) [15, 16,
17] and micellar capillary electrophoresis [18, 19, 20].
The official method for caffeine determination in tea
leaves [21] is based on low-pressure column chromatog-
raphy and spectrophotometric determination at 276 nm.
This method involves a sample preparation time of more
than 6 hours and the use of 200 mL diethyl ether and
50 mL CHCl3 for each sample. So it is a costly and te-
dious methodology, making the development of a fast,
simple and less contaminant procedure to determine caf-
feine in tea necessary.
Caffeine has been determined in other beverages like
coke [22] and coffee samples [23, 24] by the use of FTIR
spectrometry, because this technique offers an excellent
sensitivity and selectivity for this kind of analysis being
possible to do a practically direct determination after a
simple solid phase [22] or a solvent extraction [23, 24].
The main object of this work was the development of
an FTIR method for caffeine determination in tea samples
in order to reduce the time and the solvent volume re-
quired as much as possible for this determination.
Experimental
Apparatus and reagents
A Magna FTIR 750 system from Nicolet (Madison, USA)
equipped with a temperature-stabilised DGTS detector with a
Ge/KBr beamsplitter and precise digital signal processing (DSP)
was employed for all the FTIR measurements, with a nominal res-
olution of 4 cm–1, accumulating 25 scans and using a micro flow-
through cell with ZnSe windows and a pathlength of 0.457 mm.
A stopped-flow system was employed for sample introduction
inside the IR cell with a Gilson Minipuls P2 (Villiers-le-Bel,
France) peristaltic pump. For sample transport, Viton (Isoversinic)
3 mm o.d. and 1 mm i.d. pump tubing and PTFE 0.8 mm i.d. con-
necting tubes were used. For processing the FTIR absorbance data,
the 2.1 version of the Omnic software of Nicolet Corp. was em-
ployed.
562
A Hewlett-Packard (Palo Alto, USA) 8452A UV-VIS DAD
system was employed for the UV-Vis determination of caffeine.
All the reagents employed in this study were of analytical
grade. Chloroform, stabilised with 150 ppm of amylene, was ob-
tained from Scharlau (Barcelona, Spain), ammonium hydroxide
30% from Panreac (Barcelona, Spain), and caffeine from Fluka
(Buchs, Switzerland). The tea samples were obtained from the
Spanish market.
Reference procedure
A 1-g ground sample was accurately weighed inside a 100-mL
beaker, 50 mL NH4OH (4 mol/L) was added and the mixture
warmed on a boiling H2O bath for 2 min. After cooling, transferred
quantitatively to a 100-mL volumetric flask and diluted to the vol-
ume with H2O. Then, 5-mL aliquots of the turbid solution were
carefully mixed with 6 g Celite 545 and the mixture was placed in-
side a 25×250 mm tube column to obtain a homogenous and com-
pact layer above a glass wool plug and a mixture of 3 g Celite 545
and 2 mL NaOH 2 mol/L. This basic column was mounted above
the acid column prepared from well mixed 3 mL 2 M H2SO4 and
3 g Celite 545 placed between glass wool layers inside a 25×
250 mm tube. A 150-mL aliquot of H2O-saturated ether was se-
quentially passed through the basic and acidic columns and dis-
carded. A 50-mL aliquot of H2O-saturated ether was then passed
through the acidic column and discarded. A 48-mL aliquot of H2O-
saturated CHCl3 was also passed through the acidic column. The
first portions of the eluent were then passed through the basic col-
umn and the caffeine solution recovered in a 50-mL volumetric
flask. The column eluate was diluted to the mark with H2O-satu-
rated CHCl3 and the absorbance read at 276 nm against an H2O-
saturated CHCl3 blank, using caffeine CHCl3 solutions as cali-
brants.
Recommended Procedure
determination of caffeine peak height values at 1658.5 cm–1, cor-
rected using a baseline at 1800 cm–1, obtained for samples were
interpolated in a calibration line obtained from caffeine calibrant
solutions directly prepared in CHCl3 and measured under the same
conditions as the sample extracts and using a background of
CHCl3.
Results and Discussion
FTIR spectra of caffeine and tea extracts
Figure 1 shows the FTIR spectra of the caffeine calibrant
solution in chloroform and a tea extract. Both spectra are
very similar and have a characteristic absorption that can
be used for the determination of caffeine the main differ-
ence being the presence at 1730 cm–1 of an overlapped
band probably due to the absorption of other extracted
components present in tea samples. The carbonyl bands at
1704.8 cm–1 and 1658.5 cm–1 are the most intense bands,
but additionally there are others present at 1554.4 cm–1,
1415.5 cm–1 and 1361 cm–1. So that could be useful for
the analysis of CHCl3 extracts obtained from tea leaves.
In Figure 2, the spectra of three different kinds of tea
present the shoulder at 1730 cm–1 but demonstrate that is
possible to identify the main characteristic bands of caf-
feine in every tea sample extracted with CHCl3. So differ-
ent studies were carried out in order to find the appropri-
ate conditions for the quantitative extraction of caffeine
and its determination in the simplest way possible .

A 250-mg sample of a blended tea was accurately weighed in a
glass tube, 5 mL NH4OH 2 mol/L was added and then the mixture
was mechanically shaken for 2 min. Then 5 mL CHCl3 were added
and the mixture mechanically shaken for an additional 2 min. Af-
ter that, the mixture was centrifuged for 10 min.
The chloroform solution was introduced into a microflow-
through IR cell using a peristaltic pump and the FTIR spectra ob-
tained between 4000 and 600 cm–1 at 4 cm–1 resolution and accu-
mulating 25 scans for spectra. The background spectrum was es-
tablished from the cell filled with a blank extract solution. For the
Study of the extraction conditions
Caffeine extraction from an aqueous solution to a chloro-
form phase is a well known quantitative process. For this
reason, in the determination of caffeine in tea samples, the
main problem is to guarantee the extraction of the caffeine
from the solid to an aqueous phase. Previous studies [23,
24] proved that caffeine can be easily extracted by using

Fig. 1 FTIR spectra of a caffeine

standard of 1.5 mg mL–1 (—), a sam- 
ple extract (--) and a blank solution 
(-·-·). Every spectrum was obtained
for 25 cumulated scans at a resolution
of 4 cm–1. For the sample extract the 
spectrum was obtained using a blank
solution as background, the caffeine
standard was measured in the same  

conditions as the sample extract but

$EVRUEDQFH

using a background of CHCl3







 


   
:DYHQXPEHUV FP
 563
Fig. 2 FTIR spectra of a Jasmine, *UHHQ WHD H[WUDFW 
Green and Ceylon tea samples ex-
tracted in CHCl3 






&H\ODQ WHD H[WUDFW 



$EVRUEDQFH


 


-DVVPLQ WHD H[WUDFW 





 

     
:DYHQXPEHUV FP

alkaline solutions such as ammonia. So the shaking mode, Table 1 Calibration lines obtained for FTIR determination of caf-
shaking time and the volume and concentration of feine using different bands
NH4OH were studied in order to obtain the best conditions Band Calibration line R2 b RSDc LODd
for a quantitative extraction of caffeine from tea leaves. (Baseline) % (mg L–1)
Jasmine tea samples and peak height measurements at
1704.8 cm–1 y =–0.0017+0.1333·Ca 0.9984 0.3% 1.5
1658.5 cm–1, corrected using a baseline at 1800 cm–1, (1800 cm–1)
were employed in order to search for the best extraction 1704.8 cm–1 y =–0.0005+0.1053·C 0.9995 0.3% 2.2
conditions. (1683 cm–1)
Two different shaking modes, mechanical and ultra- 1658.5 cm–1 y =–0.0016+0.2142·C 0.9993 0.1% 1
sonic, were assayed. For both modes, shaking times from (1800 cm–1)
1 to 5 minutes were evaluated. From the results obtained 1554.4 cm–1 y =–0.0001+0.0361·C 0.961 1.3% 4.8
it can be concluded that there is no significant (at p=0.05) (1800 cm–1)
difference between sonication and mechanical shaking, 1415.5 cm–1 y =–0.0009+0.0177·C 0.8995 5.1% 2.1
the simplest method being the latter procedure. Using the (1800 cm–1)
mechanical mode, caffeine was quantitatively recovered Notes: Calibration lines were established in the concentration
after a 2-min shaking time. range from 0 to 4 mg mL–1 caffeine using 5 standard solutions;
The increase in NH4OH concentration from 0.25 to aC: Concentration of caffeine in mg mL;
bR2 : Regression coefficient;
4 mol/L does not influence the caffeine extraction. So a cRSD: Relative standard deviation for five independent measure-
NH4OH 2 mol/L concentration was selected in order to ments of a sample containing 1.45 mg mL–1 caffeine;
guarantee a precise and robust procedure. dLOD Limit of detection established for a probability level of
The effect of the volume of the NH4OH was evaluated 99.6% (k=3).
from 5 to 10 mL, it was found that 5 mL was enough to
obtain a quantitative recovery of caffeine from tea. The
use of reduced volumes of NH4OH and CHCl3 permitted matrix interference during the extraction step (see Table 1).
us to carry out all the different steps of the extraction, In order to choose between the criteria, the height ratio
such as solid-liquid extraction with ammonia, liquid-liq- based on the most intense bands was calculated for the
uid extraction with chloroform and phase separation calibration solution, sample extracts and spiked samples.
through centrifugation. In fact, a constant proportion between bands considered in
all types of solutions indicates a good comparability be-
tween natural sample extracts, external standards and
Selection of the appropriate bands standard addition solutions.
for FTIR determination of caffeine Table 2 shows the peak height ratio between the three
bands considered working with a fixed baseline criterion
Figure 1 shows that at least two bands, namely 1704.8 and (at 1800 cm–1) and with corrected values at 1883 cm–1 for
1658.5 cm–1, could be used for the quantitative determina- the carbonyl bands at 1704.8 and 1658.5 cm–1.
tion of caffeine in tea extracts. For these bands, peak As it can be seen, the bands at 1704 and 1658.5 cm–1
height measurements were considered. When applying are those which provided the best sensitivity and the best
two alternative baseline criteria for the carbonyl band at comparability when considering pure CHCl3 solutions,
1704,8 cm–1, it is possible to compensate for an eventual extracts obtained from natural samples and those found
564

Table 2 Peak height ratio through bands presented by caffeine at 1704.8 cm–1, 1658.5 cm–1 and 1554.4 cm–1 in different types of sam-
ple extracts and standards
H1704/H1658± sa H1704/H1554± sa H1658/H1554± sa H1704/H1658± sb H1704/H1554± sc H1658/H1554± sd
Standards 0.632± 0.002 3.8±0.3 6.1±0.4 0.565±0.004 3.0±0.1 5.4±0.2
Ceylon Tea 0.634±0.002 3.61±0.05 5.69±0.07 0.568±0.005 2.73±0.07 4.81±0.09
Jasmine Tea 0.632±0.002 3.73±0.09 5.9±0.1 0.579±0.006 3.0±0.1 5.2±0.2
Green Tea 0.630±0.002 3.41±0.05 5.42±0.08 0.558±0.003 2.53±0.05 4.54±0.07
Spiked Ceylon Tea 0.633±0.002 3.9±0.1 6.1±0.1 0.573±0.002 3.0±0.1 5.3±0.2
Spiked Jasmine Tea 0.634±0.001 3.77±0.03 5.94±0.06 0.574±0.002 2.93±0.03 5.10±0.06
aUsing a common baseline established at 1800 cm–1 for all the bands.
b,c,dConsidering
a baseline correction at 1883 cm–1for the bands at 1704.8 and 1658.2 cm–1, and corrected at 1800 cm–1 for the band at
1554.4 cm–1.

Fig. 3 FTIR spectra from the residue

remaining after evaporation onto a
BaF2 plate of 100 µL of caffeine stan-  -DVPLQH WHD
dard solution and the chloroform ex-

tracts of tea samples. The spectra have
been shifted to avoid band overlap- 
ping &H\ODQ WHD
$EVRUEDQFH





 &DIIHLQH VWDQGDUG

*UHHQ WHD






    

:DYHQXPEHUV FP

for spiked samples, thus indicating that a method based on Table 3 Results obtained for the determination of caffeine in tea
the aforementioned bands could be a robust alternative for samples by different procedures
caffeine determination in tea leaves. However, the band at Sample Caffeine % (m/m)
1704.8 cm–1 could be affected by that at 1730 cm–1 which
appears in some sample extracts and, because of that peak FTIR determination Reference
height measurements at 1658.5 cm–1 established with a method
LLEa AOACb
baseline at 1800 cm–1, seems the best alternative for FTIR preparation preparation
determination of caffeine in tea.
On the other hand, Figure 3 shows the FTIR spectra of Ceylon Tea 2.67±0.03 2.59±0.02 2.1±0.4
the solid residue obtained after evaporation, onto a BaF2 Green Tea 2.70±0.02 2.21±0.02 2.1±0.2
plate, of a few microliters of the chloroform extracts of Jasmine Tea 2.29±0.04 2.19±0.07 1.9±0.1
different tea samples considered and a calibrant solution aLiquid-liquidextraction;
of caffeine. On comparison with the caffeine standard, bFTIR determination after the official AOAC preparation method.
one can see differences in the relative intensity of the Data reported are the average of 3 independent determination ± the
band at 1704.8 cm–1 especially for Jasmine tea samples; corresponding standard deviation.
which confirms the selection of the band at 1658.5 cm–1
for quantification of caffeine in the organic extracts of tea
samples. In fact, it seems that the results from the reference
method based on UV detection at 276 nm are a little bit
low and that the simple liquid-liquid extraction and FTIR
Analysis of natural samples of tea measurement gives errors a little too high compared with
FTIR determinations carried out after the AOAC prepara-
As can be seen in Table 3, caffeine concentrations found tion procedure.
by FTIR in three types of tea samples are comparable to In order to verify the accuracy of the developed proce-
those found in the literature and data found are of the dure, recovery experiments were carried out by spiking
same order to those obtained by the official method. Ceylon and Jasmine tea samples with amounts of caffeine
 565
Table 4 Recovery of caffeine from tea samples Acknowledgements. The authors acknowledge the financial sup-
port of the Generalitat Valenciana Project GV01–249. G. Quintás
Sample Caffeine Caffeine Recovery Mean acknowledge the grant provided by the Laboratorio de Higiene
added (mg) found (mg) (%)a recovery (%) Laboral y Ambiental of the Universitat de Valencia and J. Ohns-
mann acknowledge the Erasmus grant of the European Union.
Ceylon Tea 2.50 2.62 104.7±0.2 98±6
5.00 4.75 94.9±0.2
7.35 6.95 94.6±0.2
References
Jasmine Tea 2.51 2.51 100.2±0.6 97±3
4.92 4.82 98.0±0.1 1. Fernandez PL, Martin HJ, Gonzalez AG, Pablos F (2000) Ana-
7.33 6.84 93.3±0.3 lyst 125:421–425.
2. Capeda A, Paseiro P, Simal J, Rodríguez JL (1990) Alimenta-
aValues correspond to the average recovery ± the standard devia- ria 211:23–27.
tion of 3 independent determinations 3. Capeda A, Paseiro P, Simal J, Rodríguez JL (1991) An Broma-
tol 42:241–249.
4. Lau O, Luk S, Cheng O, Chiu TPY (1992) Analyst 117:
777–783.
ranging from 2.5 to 7.35 mg. Data in Table 4 show that 5. Meyer A, Ngiruwonsanga T, Henze G (1996) Fresenius J Anal
average recovery percentages vary from 97 to 98% thus Chem 356:284–287.
indicating the absence of systematic errors of the FTIR 6. Muhtadi F, El-Hawary S, Hifnawy MS (1990) J Liq Chro-
matogr 13:1013–1028.
measurements carried out after solvent extraction of sam- 7. Arlorio M, Coisson JD, Martelli A (2000) Chromatographia
ples with ammonia and CHCl3. 52:579–583.
8. Rojo-de Camargo MC, Teledol MCF (1999) J Sci Agric 79:
1861–1864.
9. Shi YL, Han JT, Liu QP, Cheng JK (1996) Fenxi Shiyanshi
Analytical performance of the developed procedure 15:15–18.
10. Gennaro MC, Abrigo C, (1992) Fresenius J Anal Chem 343:
The methodology developed for caffeine determination in 523–525.
tea leaves by FTIR measurement of CHCl3 extracts ob- 11. Chen QC, Mou SF, Hou XP, Ni ZM (1998) Anal Chim Acta
tained after extraction of the solid samples with ammonia 371:287–296.
12. Hurst WJ, Martin RA (1993) Analusis 21:389–391.
gives a sensitivity of 0.2142 absorbance units per mg 13. Jin WR, Yu DQ, Dong Q, Ye XY (2000) J Chromatogr Science
mL–1 caffeine with a limit of detection of 1 mg mL–1, cor- 38:11–15.
responding to a 0.002% m/m concentration in the solid 14. Swinney KA, Bornhop DJ (1999) J Microcolumn Sep 11:
sample and a limit of quantification of 3.4 mg L–1 thus of 596–604.
15. Sarin RK, Sharma GP, Vrshney KM, Rasool SN (1998) J Chro-
the order of 0.007% m/m caffeine in tea leaves (taking matogr A 822:332–335.
into account the extraction of 250 mg sample with a final 16. Vovk I, Golc-Wondra A, Prosek M (1997) J Planar Chro-
volume of 5 mL CHCl3). matogr Mod TLC 10:416–419.
On comparing the solvent extraction FTIR procedure 17. Walker JC, Zaugg SE, Walker EB (1997) J Chromatogr A 781:
481–485.
with the reference one using low pressure chromatogra- 18. Zhao Y, Lunte CE (1997) J Chromatogr Biomed Appl 688:
phy and UV determination, the consumption of CHCl3 265–274.
was reduced by a factor of 10 and that of NH4OH by a 19. Sun BG, Hou PL, Lin YF (1999) Fenxi Kexue Xuebao 15:
factor of 20, thus providing a cheaper alternative. The use 124–127.
of diethyl ether is also avoided and as well as that of 20. Thompson CO, Trenerry VC, Kemmery B (1995) J Chro-
matogr A 694:507–514.
Celite, thus providing an economic sustainable procedure 21. Herrlich K (1990) Association of Official Analytical Chemists,
which drastically reduces the environmental impact. In Official Methods of Analysis, Arlington, 769–781.
addition, the time required for each determination is ap- 22. Daghbouche Y, Garrigues S, Vidal MT, de la Guardia M
proximately 15 minutes, compared with 6 hours needed (1997) Anal Chem 69:1086–1091.
23. Bouhsain Z, Garrigues JM, Garrigues S, de la Guardia M
for the reference procedure. So, the new procedure not (1999) Vibr Spectroscopy 21:143–150.
only increases the productivity of the laboratory but also 24. Garrigues JM, Bouhsain Z, Garrigues S, de la Guardia M
reduces the cost of each determination. (2000) Fresenius J Anal Chem 366:319–322.