systematic examination of the signal area precision of a single quadrupole enhanced low mass option...

RAPID COMMUNICATIONS IN MASS SPECTROMETRY

Rapid Commun. Mass Spectrom. 2006; 20: 2419–2426

) DOI: 10.1002/rcm.2611
Published online in Wiley InterScience (www.interscience.wiley.com
Systematic examination of the signal area precision

of a single quadrupole enhanced low mass option

(ELMO) TSQTM mass spectrometer

Klaus Fischer*, Susanne Hoffler and Axel MeyerUniversity of Trier, Faculty VI – Geography and Geosciences, Department of Analytical and Ecological Chemistry,

Behringstr. 21, 54296 Trier, Germany

Received 7 April 2006; Revised 7 June 2006; Accepted 14 June 2006

*CorrespoGeographcal ChemE-mail: fi

To examine the precision of the signal area response of an enhanced lowmass option (ELMO)MSQTM

mass spectrometer, operated in the negative electrospray ionization (ESI) mode, extended tests were

performed, using flow injection analysis mass spectrometry (FIA-MS). Analytes were nitrate, nitrite,

malonic acid, and D,L-mandelic acid. Composition and concentration of injected samples, application of

an ASRS anion suppressor and of the cone wash unit, methanol addition to the FIA flowmedium, and

the voltage bias of the hexapole transfer lenswere test variables. Individual test cycles comprised up to

90 injections, processed within 20h. With a few exceptions the signal response tended to decline over

time leading to a loss of more than 80% of the initial signal area in extreme cases. A hexapole radio-

frequency (RF) voltage bias of S0.3V led to an overall low detector response and to high losses of

sensitivity over time. Other correlations between the insufficient signal reproducibility and FIA-MS

operating conditions could not be established. The test scheme gave hints how to localize the cause of

the mass spectrometer malfunction. The repetition of the test scheme after remedying the detected

electronic default demonstrated that relative standard deviations less than 5% can be achieved for a

sequence of 30 injections if methanol is added to the FIA flow medium and if a suppressor is used.

Based on these findings a recommendation is formulated to supplement current test schemes for

instrument performance verification by a detector response precision criterion. Copyright # 2006

John Wiley & Sons, Ltd.

Ion chromatography/mass spectrometry (IC/MS) parallels

in most of its technical design high-performance liquid

chromatography/mass spectrometry (HPLC/MS) but,

depending on the manufacturer, there are some specific

features owing to the fact that the mass spectrometer has to

cope with common aqueous IC eluents, containing less

volatile acids, bases, buffers, or ion-pairing reagents. For the

analysis of small inorganic ions or low molecular weight

organic compounds, high detection sensitivity in the low

mass range is required and this is usually not offered by

standard single quadrupole MS systems. For instance, the

Finnigan enhanced low mass option (ELMO) MSQTM

detector (Thermo Electron Corp., Waltham, MA, USA),

which is coupled with an IC unit by Dionex Corp.

(Sunnyvale, CA, USA) to design a completely configured

IC/MS system, is equipped with a cone wash device and

with an ELMO detector, comprising a 5-mm hexapole radio-

frequency (RF) transfer lens and a separate RF generator.

Currently, themain classes of organic compounds targeted

by single quadrupole IC/MS are low molecular weight

organic acids,1–5 carbohydrates,6,7 and synthetic chelating

agents.8–10 Important inorganic analytes are oxyhalogenides,

ndence to: K. Fischer, University of Trier, Faculty VI –y/Geosciences, Department of Analytical and Ecologi-istry, Behringstr. 21, 54296 Trier, [email protected]

e.g. perchlorate, bromate, and iodate,4,11–16 and simple

halogenides, e.g. iodide.17,18

In view of the wide use of single quadrupole mass

spectrometers as mass-selective detectors, such instruments

should be suited for the quantitative analysis of large sample

series on a routine level. In addition to other analytical

quality parameters, repeatability, reproducibility and robust-

ness are decisive performance criteria.

From the beginning of our systematic performance tests of

a newly installed IC/API-MS system operated in the

negative electrospray ionization (ESI) mode we were

confronted with an insufficient repeatability of quantitative

results and with a serious signal drift, mainly in the form of

losing sensitivity over time. To document the state of the

mass spectrometer and to uncover the cause of its

malfunction, extended repeatability and reproducibility tests

were performed, varying the relevant operation parameters.

The test results should enable us to decide whether the

problem is caused by inadequate operation or by instru-

mental shortcomings. Furthermore, the test scheme should

offer the versatility to be useable for the performance control

of different MS systems and it should be capable of

Copyright # 2006 John Wiley & Sons, Ltd.

2420 K. Fischer, S. Hoffler and A. Meyer

scrutinizing the effectiveness of technical modifications

under realistic analytical conditions. Conducting the test

schedule over more than half a year, several thousand pieces

of data were gathered.

The test scheme helped to identify the cause of the

malfunction of the MS device. After technically upgrading

the detector, which should solve the electronic problem, some

of the test series were repeated. The presentation of the most

important test results, characterizing the situation before and

after the technical upgrade, has the following aims:

(a) t

Cop

o share the gained practical experience and to assist

other users facing similar problems in trouble shooting;

(b) t
o show which of the tested methodological and instru-
mental parameters had a systematic effect on the detector

performance;

(c) t
o provide reference data on analytical repeatability and
reproducibility, achievable with an ELMO TSQTM detec-

tor in the negative ion mode under realistic conditions;

(d) t
o generate discussion on how to assess system perform-
ance data and how to define whether the instrumental

performance is in a ’normal’, acceptable range or not; and

(e) t
o initiate considerations concerning an extension of man-
ufacturers’ responsibility for product quality, i.e. integ-

ration of standardized reproducibility measurements in

mass spectrometer functional verification protocols.

EXPERIMENTAL

Materials and reagentsAll reagents and solvents were of analytical-reagent grade.

Water was purified by reverse osmosis and then passed

through a Membrapure unit (Astacus Analytical, Boden-

heim, Germany). The purified water was used as flow

medium for the flow injection analysis (FIA)-MS system.

The selected test compounds were: D,L-mandelic acid

(1-hydroxy-1-phenylacetic acid) (>99%; Merck, Darmstadt,

Germany), malonic acid (>99%; Fluka, Buchs, Switzerland),

and standard solutions of nitrate (1 g�L�1; Merck) and

nitrite (1 g�L�1; Merck), prepared from their sodium salts.

Several tests were performed with admixtures of methanol

(for liquid chromatography; Merck) to the flow medium.

InstrumentalThe experiments were designed as flow injection/ESI-MS

analysis. A continuous flow of water or of water/methanol

(3:1, v/v) mixtures was maintained bymeans of a Dionex GP

50 gradient pump. Flow rates between 0.15 and

0.38mL�min�1 were tested but the usual settings were 0.2

or 0.38mL�min�1. The aqueous samples were injected into

the flowmedium (injection volume: 10mL) with a Dionex AS

50 autosampler. The analyte concentrations weremainly 10.0

or 3.0mg�L�1, but several tests were run with concentrations

of 20.0 or 0.1mg�L�1. Two types of samples were measured:

mixtures of two or of all four compounds, equally

concentrated (’multiple component standard’), and single

compound standard solutions. To control the stability of the

samples, a Dionex CD 25 conductivity detector was coupled

in-line before the ESI inlet during several test series.

yright # 2006 John Wiley & Sons, Ltd.

Additionally, a 2mm ASRS-Ultra suppressor (Dionex),

operated in the external water mode, was inserted before

the conductivity cell at specific test subsets. By using two

appropriate PEEK capillaries (length: 20 ft; diameter: 0.005 and

0.01 inch), a back pressure of at least 1250 psi was maintained.

A Finnigan Surveyor ELMO MSQTM mass spectrometer

with API interface, operated in the negative ESI mode, was

used. System control and data processing were handled by

the Dionex Chromeleon 6.40 software. Nitrogen, supplied by

a nitrogen generator, served as sheath and nebulizing gas,

maintaining a nitrogen pressure of 80 psi. The ESI probe

temperaturewas 4508Cor 5508C, cone voltagewas�50V and

ESI needle voltage was �3.0 kV or �3.5 kV. In several tests a

Dionex AXP-MS pump was used to wash the entrance cone

of the mass spectrometer with water or water/methanol

(3:1, v/v). The flow rate of the washing liquid was

0.04mL�min�1. The analytes were monitored in the se-

lected ion monitoring (SIM) mode observing the following

m/z values: 46 nitrite, 62 nitrate, 103 [M–H]� malonic acid,

and 151 [M–H]� mandelic acid. A full-scan mass spectrum

up to m/z 500 of malonic acid showed no peak from the

malonate dianion, m/z 51. Signals indicating the formation

of sodium malonate ion pairs or of ion clusters were also

not present. Signal areas (counts � min�1) were used as

quantitativemeasure. Themass spanwas 0.5 or 1.0 and dwell

times of 0.25, 0.3 or 0.5 s were selected. Alternatively,

negative ion full-scan mass spectra were recorded over the

range m/z 50–500 at a scan time of 0.5 s. The run time per

injection ranged from 7 to 10min. The initial voltage of the

hexapole RF lens bias was �0.3V. This value was raised to

�0.6V as a result of the related test series and finally adjusted

to �1.0V after the technical upgrade of the detector. During

the last minute of every run, the ESI voltage was converted to

positive polarity to avoid an accumulation of negative

charges in the ion focusing region of the mass spectrometer.

Data treatment

Mean values (x), standard deviations (SD), coefficients of

variation (% RSD) and linear regression functions with

regard to possible correlations between signal areas and MS

run time were calculated for all test sequences. Due to the

time-dependent bias of the data series the reliability of the

corresponding means did not increase with the number of

measurements as is expected where errors are randomly

distributed. Therefore, data were recognized as outliers and

rejected only if they could be traced back to improper

laboratory operations or instrument conditions.

In this paper the term ’repeatability’ applies to test schemes

conducted with replicate injections of one standard solution

exclusively. Otherwise the term ’reproducibility’ is used.

RESULTS AND DISCUSSION

Initial situationThe first test series was conducted with the aim of creating a

data matrix which would offer a first insight into the mass

specrometer detector performance, allowing for a rough

estimation of data trends and effect sizes. Therefore, a four-

compoundmixture containing the analytes in concentrations


DOI: 10.1002/rcm

Table 1. Repeatability of the mass spectrometer response

(SIM signal area) for four analytes, examined with a suite of

four injection sequences (test series 1)

Analyte

Subset #1–20 Complete series (#1–53)

x %RSD %CiSa x %RSD %CiSa

Nitrite 1975 7.4 þ5.7 1690 18.1 �41.9Nitrate 86839 5.2 �5.8 74616 15.8 �39.6Malonic acid 3438 23.8 �44.1 3987 26.5 þ80.6Mandelic acid 19065 17.6 �31.1 18811 11.8 �8.4

aChange in sensitivity: detector response for the last injection relatedto the first injection (both values derived from the linear regressionfunction).Analyte concentration of the multiple component standard: nitriteand nitrate, 10.0mg�L�1; mandelic acid, 10.5mg�L�1; malonic acid,10.9mg�L�1Total run time 17h; run time for subset #1–20: 3 h 20min.Time interval between every injection sequence 3h.

Table 2. Reproducibility of the nitrate SIM signal with and

without cone wash (test series 2)

Injection sequence

Without cone wash With cone wash

x % RSD %CiSa x %RSD %CiSa

#1–20 (single standard) 723 23.0 þ9.3 1087 13.1 �0#41–65 (mix) 453 25.9 þ24.8 563 17.7 �12.7Total of single injections 710 21.3 �1.9 966 22.5 �40.4Total of mix injections 455 27.3 þ16.8 515 34.7 �43.7Total series 573 32.7 �26.8 723 41.4 �67.4

aChange in sensitivity: detector response for the last injection relatedto the first injection (both values derived from the linear regressionfunction).Nitrate concentration, 0.1mg�L�1For further details, see text.

Table 3. Reproducibility of the mandelic acid SIM signal with

and without cone wash (test series 2)

Injection sequence

Without cone wash With cone wash

x % RSD %CiSa x %RSD %CiSa

#1–20 (single standard) 99.4 11.6 �12.8 147.8 14.3 þ4.8#21–45 (mix) 67.3 16.6 �11.1 89.4 18.7 �33.6Total of single injections 80.9 35.3 �66.7 120.4 35.7 �61.0Total of mix injections 59.1 28.6 �50.0 83.6 22.2 �43.1Total series 69.2 25.3 �71.8 100.6 36.7 �69.9

aChange in sensitivity: detector response for the last injection relatedto the first injection (both values derived from the linear regressionfunction).Mandelic acid concentration, 0.1mg�L�1For further details, see text.

Examination of the signal area precision of an ELMO TSQTM 2421

between 10.0 and 10.9mg�L�1 was injected 53 times. The

series was subdivided into four injection sequences. The first

sequence comprised 20 consecutive injections. Each of the

following three sequences included 11 injections. Between

every sequence a 3-h interval of continuous FIA-MS

operation without sample introduction was inserted. The

total run time was 17 h.

For the first injection sequence, the relative standard

deviation (RSD) and the change in the detector response

were relatively small for the inorganic ions (Table 1). The

variation of the response data was considerably higher for

the organic acids which showed a loss of 30% ormore of their

respective initial signal areas. Simultaneously, the signal of

the conductivity detector remained almost constant. Its RSD

was less than 2.1% without following any trend.

The heterogeneity of the sensitivity trend established for

the whole test series was remarkably higher than for the first

injection sequence. Whereas the detection sensitivity

decreased almost identically for the inorganic ions by

approximately 40%, it remained nearly constant formandelic

acid and increased by 80% for malonic acid. The divergence

of the sensitivity trends makes it clear that the observed

variations in the mass spectrometer performance cannot be

explained by a simple drift or saturation effect.

Also to be considered are the reasons for the tremendous

differences in the analyte specific detection sensitivities, most

pronounced in the case of nitrite and nitrate. Calculated on

the basis of the response means obtained from the first

injection sequence and taking account of the different masses

of the ions, the detection sensitivity for nitrate is 59 times that

for nitrite. Although the absolute number is not significant,

the order of magnitude was reproduced by many compara-

tive measurements. One reason for the low detection

sensitivity for the nitrite ion might be its tendency to

disproportionate into nitrogen oxide and nitric acid under

acidic conditions (transformation of the nitrite ion into

nitrous acid). This disproportionation is favored by increases

in temperature and of the nitrite ion concentration – both

processes accompany the electrospray formation. Indeed

full-scan measurements of the nitrite standard solutions

showed that small nitrate signals were always present in the

mass spectrum.


The following measurements (test series 2) were intended

to ascertain the influence of the cone wash device on data

precision. Trends in the detector response should be

compared for single substance and mix injection of nitrate

and mandelic acid. The analyte concentrations were reduced

to 0.1mg�L�1. The single compound standards were injected

30 times, forming three sequences, comprising 20 and 5

injections. Arranged into three intervals, comprising 25 and 5

injections, 35 measurements were made from the multiple

component standard. The whole experiment was performed

twice, first without cone wash, afterwards with such a wash.

Some general trends can be extracted from the test series

(Tables 2 and 3). First, the detector response for the single

compound standard was higher than for the mixed one. For

instance, the mean detector response for nitrate, injected

together with mandelic acid, was 53% of the value

established for the pure nitrate standard, measured at active

conewash (Table 2). Secondly, the operation of the conewash

elevated the detection sensitivity by 10% to 50% but exerted

no clear effect on data variation.

The complete mandelic acid data set is illustrated in Fig. 1.

Despite the high short-time variation of the data, the general

trend is apparent.

A significant time trend of the mass spectrometer signal

was observed while working with smaller injection

sequences (test series 3). After 20 injections of each of the

four single component standards the loss of sensitivity


DOI: 10.1002/rcm

Figure 1. Repeatability of the mandelic acid SIM response with and without cone wash (test series 2).

The total series comprised 65 injections (30 injections of the single (s) component standard,

35 injections of the multiple (m) component standard). Mandelic acid concentration 0.1 mg�L�1.

Table 4. Repeatability of the SIM signal area for four ana-

lytes, injected with single and multiple component standards.

Situation before MS upgrade (test series 3)

Analyte

Single componentstandard (20 injections)

Multiple componentstandard (25 injections)

x %RSD %CiSa x %RSD %CiSa

Nitrite 9203 6.5 �9.3 2106 6.3 �14.4Nitrate 85908 8.5 �20.7 72551 6.3 �15.5Malonic acid 2151 14.2 �19.1 1035 27.6 �55.6Mandelic acid 27854 6.0 �12.9 13463 12.4 �30.5

aChange in sensitivity: detector response for the last injection relatedto the first injection (both values derived from the linear regressionfunction).Analyte concentrations: nitrite and nitrate, 10.0mg�L�1; mandelicacid, 10.5mg�L�1; malonic acid, 10.9mg�L�1. Run times: Everysequence of 20 injections: 3 h, the sequence of 25 injections: 3 h45min.


varied between 9.3 and 20.7% (Table 4). The run time per

standard solution was 3 h. The 25-fold injection of the four-

component standard resulted in a considerably higher

depression of the mass spectrometer response for the

organic acids. The %RSD values and the signal area

decrease of the organic analytes correspond very well with

the results of the analogously configured subset #1–20, test

series 1 (Table 1). Paralleling the results noticed with test

series 2, the jointly injected analytes gave a lower detector

response than the individually added ones. This difference

was highest for nitrite which can be interpreted as an

additional indication of the occurrence of nitrite-degrading

reactions.

To optimize spray formation and to enhance the ion

transfer efficiency, organic modifiers, e.g. methanol, iso-

propanol or acetonitrile, are often added to aqueous eluents

before they enter the ESI interface. To reduce the salt content

of typical IC eluents and to minimize ion pair formation and

ion clustering in the atmospheric pressure region, a


suppressor, functioning as an ion exchanger, is coupled

between the column and the mass spectrometer interface.

The influence of such analytical conditions on detection

sensitivity was checked with a set of four test series.

Therefore, a standard solution combining all four com-

pounds in concentrations of 3.0mg�L�1 was used. For each

test series 30 injections of the multiple component standard

were made within 8 h. Equally concentrated solutions of

nitrate and ofmandelic acidwere also analyzedwithin a total

run time of 10 h.

The results for nitrite and mandelic acid are depicted in

Figs. 2(A) (nitrite) and 2(B) (mandelic acid), respectively. The

loss of sensitivity spanned from 65.1% (test series without

methanol addition/without suppressor) to 50.5% (with

methanol/without suppressor). The highest initial detection

sensitivity for nitrite was achieved working without altera-

tion of the initial FIA-MS conditions. The in-line coupling of

the suppressor together with addition of methanol severely

depressed the nitrite signal. On several occasions the nitrite

signal was not discernible from the baseline noise, making

the determination of the signal repeatability impossible

under that condition. This finding, which is confirmed later

by measurements with the upgraded MS system, supports

the hypothesis that the low signal response for nitrite follows

from its disproportionation, combined with the low dis-

sociation degree of the nitrous acid at acidic pH. The

suppressor promotes this reaction by converting sodium

nitrite into nitrous acid. Addition of methanol did reduce the

detector response to a certain extent but also led to a more

stable detector response.

In the case of mandelic acid (Fig. 2(B)) the addition of

methanol, at least without suppressor, is advantageous for

maximum detection sensitivity. As with the other tests the

loss of signal intensity over time was high and amounted up

to about 75% of its initial value. In terms of detection

sensitivity, the parameter combination ’without suppressor/

without methanol’ was the least favorable one. When

comparing the two tests with addition of methanol, the

adverse effect of the suppressor operation is hard to explain.


DOI: 10.1002/rcm

Figure 2. Influence of methanol addition and suppressor operation on MS sensitivity and SIM signal

repeatability for nitrite (A) and mandelic acid (B). Multiple component standard (concentration:

3.0 mg�L�1 of each analyte), flow rate 0.2 mL�min�1, water/MeOH (3:1, v/v), Dionex Ultra suppressor,

external water mode.


Since this compound was already injected in its free acid

form, the suppressor should not exert any influence on the

degree of dissociation of the acid. A certain portion of the

acid molecules might have been transferred from the flow

medium into the suppressor regenerant by penetrating the

suppressor membrane. In contrast to this, the use of the

suppressor (without methanol) was advantageous to achieve

maximum sensitivities for nitrate and malonic acid.

Sequences of individual determinations of mandelic acid

and nitrate were processed after the mix measurements.

Their results are in good qualitative and, in some respects,

even quantitative agreement with former findings, indicat-

ing that they are caused more by systematic than by random

effects or circumstances.

The hexapole RF transfer lens, mounted instead of a

conventional quadrupole-like RF/dc prefilter before the mass

analyzer in the ELMO MSQTM, focuses the ions produced in

the ESI source and transmits them to the quadrupole analyzer.

Increasing the hexapole offset voltage increases the kinetic

energy of the ions and accelerates their transfer into the mass

analyzer. Therefore, the setting of the lens biasmight be crucial

for detection sensitivity generally and for the mass-to-charge

dependence of the MS response especially.


Indeed the RF voltage bias test demonstrated a consider-

able effect on MS sensitivity and signal repeatability

(Table 5). Except for malonic acid, the highest initial and

mean sensitivity was established with a lens bias of �0.8V.

The loss of sensitivity and the RSDs were lowest at �0.6V.

Under these conditions, between two-thirds and half of the

initial detection sensitivity remained until the last injection.

A second test series with the same lens bias including

60 consecutive injections resulted in a somewhat better

repeatability probably due to the smaller number of

injections. At a lens setting of �0.3V the detector response

declined extremely for all analytes during the measuring

period. For instance, the means (relative peak areas) of #1–3

and of #78–80, malonic acid, were 1706 and 233, respectively.

The decline in the mandelic acid signal was of the same

order of magnitude. The mean values of the first and last

three injectionswere 7174 and 1301, respectively, indicating a

loss of sensitivity of 82% on that calculation basis.

Considering the significant effect of the RF voltage bias on

the detector performance it is obvious that the proper

selection of this parameter helps to mitigate but not to solve

the problem. On the other hand the finding that a certain

correlation between the kinetic energy of the ions and the


DOI: 10.1002/rcm

Table 5. Influence of the hexapole RF transfer lens voltage bias on SIM signal repeatability

Analyte

RF voltage bias

�0.3V �0.6V �0.8V

x % RSD %CiSa x %RSD %CiSa x %RSD %CiSa

Nitrite 186 61.3 �92.2 337 16.5 �37.3 683 32.3 �59.1Nitrate 8635 52.3 �83.6 12738 14.9 �37.0 20338 30.4 �57.4Malonic acid 513 83.3 �99 4102 19.7 �46.7 2183 41.5 �74.5Mandelic acid 2464 70.6 �97.7 5349 22.1 �50.9 8314 36.9 �54.2

aChange in sensitivity: detector response for the last injection related to the first injection (both values derived from the linear regressionfunction).Eighty injections of the multiple component standard (3.0mg�L�1 of each analyte) per lens voltage. Test period per voltage: 17 h including seven1-h breaks between every sequence of 10 consecutive injections. The FIA-MS system operated continuously without a break.


drift of the detector response exists could have led to the

hypothesis that the cause of the instrument malfunction

might be localized in the hexapole itself or in the ion transfer

region between the hexapole and the quadrupole analyzer.

This intensive examination enabled experts fromDionex and

Thermo Electron to find that the problems were caused by an

insufficient grounding of a specific component of the

hexapole in close vicinity to the quadrupole analyzer. In

parallel with the accumulation of negative charges at this

point, repulsive electrostatic forces were exerted on nega-

tively charged ions thus preventing them from entering the

analyzer. As seen in our investigation this effect was not

simply correlated with specific parameter settings and

additional electronic feedback mechanisms might have

intensified or attenuated the primary effect.

Performance after system upgradeTo check the system performance after remedying the

electronic default and to differentiate between random

phenomena and systematic effects observed before the

system upgrade, several of the repeatability tests were

repeated under the same operating conditions as before,

except for the selection of an �1.0V hexapole RF lens bias.

Since data evaluation of reproducibilitymeasurements is still

in progress, the results of only two test series are presented

here that most clearly demonstrate differences in system

performance.

Table 6. Repetition of test series 3 (cf. Table 4) after MS

upgrade

Analyte

Single componentstandard (20 injections)

Multiple componentstandard (25 injections)

x %RSD %CoSa x %RSD %CoSa

Nitrite 2465 9.3 �10.4 597 13.1 þ15.8Nitrate 27603 8.7 þ9.9 24128 10.1 þ17.2Malonic acid 62268 8.3 �15.6 29218 10.3 þ10.9Mandelic acid 65069 6.6 þ8.5 22288 9.9 þ17.6

aChange in sensitivity: detector response for the last injection relatedto the first injection (both values derived from the linear regressionfunction).Conditions as described in Table 4


The results from the repetition of test series 3 are combined

in Table 6. The RSD values for individual test sequences

spanned between 6.6% and 13.1%. The standard deviations

are not always smaller than those achieved with the first

application of the test scheme, but the data set is more

homogeneous in so far as extreme variabilities did not occur.

Furthermore, the RSDs of the separate injections of the

analytes are generally smaller than the injections of the

multiple component solution. In contrast to the first set of test

series, the detector sensitivity tended to increase with time

for most analytes. Nevertheless, the change in the detection

sensitivity is smaller than 20%, indicating a significant

improvement in the detector performance. Paralleling the

initial situation the mean detector response for nitrate is at

least one order of magnitude greater than for nitrite and the

mean peak areas of the separately injected compounds are

two- to four-fold greater than the jointly injected analytes,

except for nitrate. This result underlines that the use of mass

spectrometry does not reduce the necessity to separate the

analytes as completely as possible before detection.

The repetition of the tests with suppressor operation and

methanol addition confirmed dramatic improvements in

signal repeatability. As illustrated in Fig. 3(A), the order of

the various analytical conditions, ranked according to the

average nitrite peak area, is the same as before, but

the detector response is essentially more stable, especially

for the parameter setting ’with methanol/without suppres-

sor’. The same inference can be drawn from the data for

mandelic acid (Fig. 3(B)). For the two test series where the

suppressor was operated, the relative changes in sensitivity

wereþ0.4% (without methanol) andþ2.7% (with methanol),

respectively, compared with �93.5% and �81.3%, for the

initial situation. The worst result (�38.9%) was achieved

without suppressor operation and without methanol

addition. The new situation qualitatively parallels the earlier

in as far as the highest detection sensitivity for mandelic acid

(and now also for nitrate) was achieved with the operating

conditions ’methanol addition/inactivated suppressor’.

From the complete data set some correlations between

operational conditions and MS performance are now more

obvious than before. With respect to signal reproducibility

the parameter combination ’without methanol/without

suppressor’ yielded the worst results. Here the sensitivity

losses ranged from 26.8% to 38.9%. Using the suppressor a


DOI: 10.1002/rcm

Figure 3. Repetition of the test scheme for the determination of the influence of methanol addition and

suppressor operation on MS response after MS upgrade. Conditions as in Fig. 2.


sensitivity drift did not occur or was less than 5% in 5 out of 9

test sequences. In three cases only the drift was between 10%

and 15%. With one exception the RSD values were below

10%. Except for nitrite the addition of methanol increased

detection sensitivity.

CONCLUSIONS

The present results allow us to draw the following

conclusions:

1. E

Co

xtended signal area repeatability and reproducibility

tests were able to unambiguously identify an electronic

default of an ELMO MSQTM detector and they helped to

localize the cause of its malfunction. The same test scheme

confirmed significant improvements in the detector per-

formance after the default had been remedied. After

having technically upgraded the detector clear corre-

lations between FIA-MS operating conditions and system

pyright # 2006 John Wiley & Sons, Ltd.

performance were found, e.g. enhancement of the detec-

tion sensitivity by methanol addition and increase in

signal repeatability by suppressor operation. Further-

more, the repetition of the test scheme proved that several

of the earlier observed effects, e.g. the relatively low

detection sensitivity for nitrite, are of a systematic rather

than of a random nature.

2. A
ccording to Slingsby and Schnute,19 internal standards
should be used when the mass spectrometer is used for

quantification in IC/ESI-MS methods. Without doubt

internal standards are valuable tools to correct for matrix

effects, e.g. ion suppression, and for short-time fluctu-

ations of other system properties, e.g. spray formation and

ion transmission efficiencies. There are several reasons

why their recommendation would not help to correct data

biased by a detector malfunction. First, the addition of one

internal standard assumes that all analytes are affected in

the same way and to the same extent by system perform-

ance variations. If this assumption is not valid, as in our


DOI: 10.1002/rcm

Co


case, a stable labeled standard has to be added for every

analyte and this can be quite expensive. Furthermore,

suitable standards are not commercially available for all

relevant substances. Secondly, internal standards cannot

compensate for a progressive deterioration of the signal-

to-noise ratio. Thirdly, in this specific context, the use of

internal standards can lead the user to adapt to the

problem instead of examining and solving it.

3. T
he instrument functional verification and installation
acceptance tests are based on the measurement of the

signal-to-noise ratio of a sample three times consecutively

injected within a short time span. Depending on the

ionization mode, different samples, settings, and

parameters for qualification are specified. A second qual-

ity control criterion is delivered by the autotune process,

which tunes and calibrates the mass accuracy in the ESI

mode within the whole nominal mass range. Both pro-

cedures are not capable of detecting sensitivity drifts of

the mass spectrometer. Therefore, we strongly recom-

mend extending the installation acceptance procedure

by a test protocol suited to control the precision of the

mass spectrometer response. We suggest that the instru-

ment performance standards guaranteed by the manufac-

turer should be supplemented by a precision criterion.

According to our experiences the FIA-MS technique is

suited for the proposed test scheme. The test protocol

should encompass a minimum number of 30 replicate

injections within a time span of 6 h at least. The statistical

treatment of the data should include a method, e.g.

regression analysis, able to identify or exclude a possible

correlation between data variation and time at a given

significance level.

NOTICE

We do not claim that any other ELMO MSQTM apparatus

necessarily has or did have the same critical properties as our

instrument before its technical upgrade. Furthermore, we do

not claim that the observed properties do apply or did apply

to types ofMSQTM detectors (e.g. with square quadrupole RF

pyright # 2006 John Wiley & Sons, Ltd.

lenses), to ionization modes, to operating conditions, or to

analytes other than those used in this study.

Our instrument was installed in July 2003. The electric

fault was detected in autumn 2004 and fixed in January 2005.

It is likely that ELMO MSQTM detectors installed since the

latter date have undergone a technical development to

prevent the described phenomena.

AcknowledgementsThe authors would like to thank Prof. W. Buchberger for

critical comments on the manuscript.

REFERENCES

1. Mascolo G, Lopez A, Detomaso A, Lovecchio G. J. Chroma-togr. A 2005; 1067: 191.

2. Gamoh K, Saitoh H, Wada H. Rapid Commun. Mass Spectrom.2003; 17: 685.

3. HelalehMIH, Tanaka K, TaodaH, HuW,Hasebe K, HaddadPR. J. Chromatogr. A 2002; 956: 201.

4. Ahrer W, Buchberger W. J. Chromatogr. A 1999; 854: 275.5. Johnson SK, Houk LL, Feng J, Johnson DC, Houk RS. Anal.

Chim. Acta 1997; 341: 205.6. Guignard C, Jouve L, Bogeat-Triboulot MB, Dreyer E, Haus-

man J-F, Hoffmann L. J. Chromatogr. A 2005; 1085: 137.7. Bruggink C, Maurer R, Herrmann H, Cavalli S, Hoefler F.

J. Chromatogr. A 2005; 1085: 104.8. Knepper TP, Werner A, Bogenschutz G. J. Chromatogr. A

2005; 1085: 240.9. Dodi A, Monnier V. J. Chromatogr. A 2004; 1032: 87.10. Bauer KH, Knepper TP, Maes A, Schatz V, Voihsel M.

J. Chromatogr. A 1999; 837: 117.11. Cavalli S, Polesello S, Valsecchi S. J. Chromatogr. A 2005; 1085:

42.12. Mathew J, Gandhi J, Hedrick J, Application Report Agilent

Technologies, 2004;13. Application Note 151, Dionex Corporation, 2003.14. Roehl R, Slingsby R, Avdalovic N, Jackson PE. J. Chromatogr.

A 2002; 956: 245.15. Slingsby R, Kiser R. Trends Anal. Chem. 2001; 20: 288.16. Seubert A, Schminke G, NowakM, Ahrer W, Buchberger W.

J. Chromatogr. A 2000; 884: 191.17. Buchberger W, Czizsek B, Hann S, Stingeder G. J. Anal. At.

Spectrom. 2003; 18: 512.18. Buchberger W, Ahrer W. J. Chromatogr. A 1999; 850: 99.19. Slingsby RW, Schnute WC. Poster presentation 17thInt. Ion

Chromatography Symp., 20–23. 09. 2004, Trier, Germany.


DOI: 10.1002/rcm

systematic examination of the signal area precision of a single quadrupole enhanced low mass option...

Documents