Comparison of the performance of three ion mobility spectrometers for measurement of biogenic amines

35
Comparison of the performance of three ion 1 mobility spectrometers for measurement of 2 biogenic amines 3 Zeev Karpas a* , Ana V. Guamán b,c , Antonio Pardo b , Santiago Marco b,c 4 a 3QBD, Arad, Israel, and Chemistry Department, Nuclear research Center, Negev, Beer- 5 Sheva 84190, Israel 6 b Departament d’Electrònica, Universitat de Barcelona, Martí i Franqués 1, 08028- 7 Barcelona (Spain). 8 c Artificial Olfaction Lab, Institute for Bioengineering of Catalonia, Baldiri i Rexach, 4-8, 9 08028-Barcelona (Spain) 10 A. V. Guaman: a[email protected] 11 A. Pardo: [email protected] 12 S. Marco: [email protected] 13 *[email protected] , fax: 972-8-6469718; phone: 972-50-6232140 14 Abstract 15 The performance of three different types of ion mobility spectrometer 16 (IMS) devices: GDA2 with a radioactive ion source (Airsense, 17 Germany), UV-IMS with a photo-ionization source (G.A.S. Germany) 18 and VG-Test with a corona discharge source (3QBD, Israel) was 19 studied. The gas-phase ion chemistry in the IMS devices affected the 20 species formed and their measured reduced mobility values. The 21 sensitivity and limit of detection for trimethylamine (TMA), putrescine 22 and cadaverine were compared by continuous monitoring of a stream 23

Transcript of Comparison of the performance of three ion mobility spectrometers for measurement of biogenic amines

Comparison of the performance of three ion 1

mobility spectrometers for measurement of 2

biogenic amines 3

Zeev Karpasa*

, Ana V. Guamánb,c

, Antonio Pardob, Santiago Marco

b,c 4

a 3QBD, Arad, Israel, and Chemistry Department, Nuclear research Center, Negev, Beer-5

Sheva 84190, Israel 6

b Departament d’Electrònica, Universitat de Barcelona, Martí i Franqués 1, 08028-7

Barcelona (Spain). 8

c Artificial Olfaction Lab, Institute for Bioengineering of Catalonia, Baldiri i Rexach, 4-8, 9

08028-Barcelona (Spain) 10

A. V. Guaman: [email protected] 11

A. Pardo: [email protected] 12

S. Marco: [email protected] 13

*[email protected], fax: 972-8-6469718; phone: 972-50-6232140 14

Abstract 15

The performance of three different types of ion mobility spectrometer 16

(IMS) devices: GDA2 with a radioactive ion source (Airsense, 17

Germany), UV-IMS with a photo-ionization source (G.A.S. Germany) 18

and VG-Test with a corona discharge source (3QBD, Israel) was 19

studied. The gas-phase ion chemistry in the IMS devices affected the 20

species formed and their measured reduced mobility values. The 21

sensitivity and limit of detection for trimethylamine (TMA), putrescine 22

and cadaverine were compared by continuous monitoring of a stream 23

of air with a given concentration of the analyte and by measurement of 24

headspace vapors of TMA in a sealed vial. Preprocessing of the 25

mobility spectra and the effectiveness of multivariate curve resolution 26

techniques (MCR-LASSO) improved the accuracy of the 27

measurements by correcting baseline effects and adjusting for 28

variations in drift time as well as enhancing the signal to noise ratio 29

and deconvolution of the complex data matrix to their pure 30

components. The limit of detection for measurement of the biogenic 31

amines by the three IMS devices was between 0.1 and 1.2 ppm (for 32

TMA with the VG-Test and GDA, respectively) and between 0.2 and 33

0.7 ppm for putrescine and cadaverine with all three devices. 34

Considering the uncertainty in the LOD determination there is almost 35

no statistically significant difference between the three devices 36

although they differ in their operating temperature, ionization method, 37

drift tube design and dopant chemistry. This finding may have general 38

implications on the achievable performance of classic IMS devices. 39

Keywords: ion mobility spectrometry, comparison of performance, 40

biogenic amines, signal processing, sensitivity, vapor concentration 41

1. Introduction 42

The advent of modern ion mobility spectrometry (IMS) can be traced 43

back to the late 1960's and its appearance as an analytical technique 44

followed shortly after, as described in detail previously [1]. In the four 45

decades that followed, several vendors of analytical equipment have 46

marketed a variety of commercial products, using different ionization 47

sources, drift tube designs, operating temperatures and ranging in size 48

from pocket size to walk-in portals. These focused mainly on 49

applications such as the detection of explosives, narcotics and toxic 50

chemicals where the required response was simply "positive" or 51

"negative" for the presence of the target compound or analyte. 52

Surprisingly, only few direct comparisons of the performance of 53

different instruments are available, although reports of proton-54

attachment versus electron exchange ionization [2] and of high-field 55

ion mobility drift tubes and a FAIMS apparatus [3] were published. 56

Several studies attempted to compare the effect of the ionization 57

method on the performance of IMS instruments. Notable among those 58

are the reports of Borsdorf et al. [4-6] in which the gas phase ion 59

chemistry of isomeric hydrocarbons [4], terpenes [5] and substituted 60

toluene and aniline compounds [6] was compared when a radioactive 61

63Ni, a corona discharge (CD) and a photoionization (PI) ion source 62

was used. In the latter study, where three similar RAID1 (Bruker, 63

Germany) handheld IMS devices that differed only in their ion source, 64

the measured reduced mobility and relative abundance of the product 65

ions differed significantly among these instruments [6]. A few other 66

reports on the comparison of the performance of two types of IMS 67

instruments towards detection of odor signatures of smokeless gun 68

powders [7] and drugs [8,9] were also published. In many cases 69

vendors of IMS devices report the level of detection of their device for 70

a given set of compounds (usually belonging to one of the above 71

applications) allowing consumers to compare the instruments on the 72

basis of the manufacturers' claims [10]. 73

In the last few years, the interest in using IMS techniques has grown 74

considerably for several novel applications as recently reviewed [11], 75

particularly in the biological and medical fields, such as in the areas of 76

monitoring food safety and study of pathological conditions. Biogenic 77

amines are compounds that are present in each living cell and play an 78

important role in regulating the cell functions as described in a mini-79

review [12]. Thus, biogenic amines may serve as important markers 80

for diseases like bacterial vaginosis [13,14] and cancer [15] or food 81

spoilage [16-18]. The high proton affinity of amines in general, and 82

biogenic amines in particular, allows their measurement by stand-alone 83

IMS instruments without need for pre-separation and pre-concentration 84

that may be required for other applications where matrix effects could 85

deleteriously affect the measurements. Three important biogenic 86

amines were tested here: trimethylamine (TMA), putrescine (1,4-87

diaminobutane) and cadaverine (1,5-diminopentane). 88

In the present study, the performance of three different types of ion 89

mobility spectrometer (IMS) devices with regard to the measurement 90

of these three biogenic amines was compared. The sensitivity and limit 91

of detection for the three amines were determined by continuous 92

monitoring of a stream of air with a given concentration of the analyte. 93

For the volatile TMA measurement of headspace vapors in a sealed 94

vial to determine the response of the IMS instruments to the vapors 95

emanating from a given quantity of the biogenic amine deposited in the 96

vial was also tested. Advanced signal preprocessing, combined with 97

multivariate curve resolution techniques (MCR-LASSO) was carried 98

out to improve the accuracy of the spectra by extracting pure 99

components from complex data matrix [19,20]. 100

2. Experimental Section 101

2.1 Preparation of the samples and measurement methodology 102

Trimethylamine (purum, 45% in water), putrescine (99%, 1,4-103

diaminobutane), toluene (99.8%) and triethylphosphate (99.8%) were 104

purchased from Sigma-Aldrich and cadaverine (purum >97% 1,5-105

diminopentane) was acquired from Fluka. A stock solution of TMA 106

was prepared by weighing a sample of the analyte and dissolving it in a 107

weighed quantity of distilled water. Further dilute solutions were 108

prepared by mixing a measured volume of the stock solution with 109

distilled water. 110

A sample from each of the three amines was inserted in a permeation 111

tube that was placed in an oven equipped with three independently 112

controlled chambers (Owlstone OVG-4, UK) at the selected 113

temperature. The amount of the sample that emanated from the 114

permeation tube was determined by weighing the tube periodically. 115

However, it should be noted that the reaction of moisture in air with the 116

vapors of the amines may lead to deposition of reaction products on 117

external surfaces of the permeation tube leading to erroneous weight 118

results. Therefore the amount of sample emanating from the tube was 119

calibrated with pure, dry nitrogen. The air flowing through the oven 120

compartment was mixed with a stream of clean air to dilute the 121

concentration of the sample vapors. The rate of permeation depends on 122

the oven temperature so that combining the selected temperature with 123

the dilution factor was used to supply the analyte vapors according to 124

the desired concentration range. Ten different concentration with one 125

replicate were measured and the maximum concentrations (“zero split” 126

in the oven) of TMA (at 70ºC), putrescine (at 90ºC) and cadaverine (at 127

90ºC) in a carrier flow 400 mL min-1

of air were, 11.15, 16.21 and 8.49 128

ppm (by weight), respectively. Measurements were carried out to 129

determine the response of the three IMS instruments to a concentration 130

of analyte vapors in a stream of air. The airstream was introduced by 131

Teflon tubing to the inlet port of each device. The response was 132

measured and a limit of detection to vapors was derived from the 133

calibration curve. 134

For the headspace measurements, 50 µL of 15% KOH were placed in a 135

glass vial and then the selected amount of TMA in solution was added 136

and the vial was immediately sealed with an aluminum cap with a 137

PTFE/silicone septum. At the start of the measurement the septum was 138

pierced simultaneously by two syringe needles: one connected to a 139

stream of carrier (pure air) and the other to a Teflon 1/8" tube attached 140

to the inlet of the IMS. The carrier flow through the headspace vial 141

was 400 mL min-1

for the GDA2 and VG-Test and 100 mL min-1

for 142

the G.A.S. UV instrument. In the VG-Test the analyte vapors were 143

somewhat further diluted by the instrument's own carrier flow of 240 144

mL min-1

. 145

146

2.2 The Ion Mobility Spectrometers 147

The three IMS devices used in the present study were the handheld 148

GDA2 (Airsense, Germany [21]), the portable UV-IMS (G. A. S. 149

Germany [22]) and the desktop VG-Test (3QBD, Israel [23]). These 150

devices differ from one another in several aspects, as summarized in 151

Table 1. The most important features are the type of ion source, dopant 152

ion chemistry and operating temperature. Toluene was used as a 153

dopant in the UV-IMS instrument and the VG-Test contained a 154

permeation tube with triethylphosphate (TEP) as a dopant while the 155

GDA2 did not contain a dopant and thus the ionization processes are 156

based on so called "water chemistry". The drift tube temperature in the 157

GDA2 was 40-45ºC, in the VG-Test was 90ºC while the G.A.S. 158

operated at ambient temperature (about 26ºC). These differences led 159

to some disparities in the types of product ions found from each analyte 160

as described in the following section. 161

162

2.3 Safety considerations 163

According to the MSDS trimethylamine, putrescine, cadaverine and 164

lutidine are quite similar from the viewpoint of safety. They are 165

hazardous in case of skin or eye contact and of inhalation. They are 166

corrosive to skin and eyes on contact and liquid or spray mist may 167

produce tissue damage particularly on mucous membranes of eyes, 168

mouth and respiratory tract. Skin contact may also cause burns. 169

Inhalation of the spray mist may lead to severe irritation of respiratory 170

tract, characterized by coughing, choking, or shortness of breath. They 171

should be handled in a chemical hood. Potassium hydroxide is 172

corrosive and may cause serious burns. It is harmful by ingestion, 173

inhalation and when in contact with skin. If the solid or solution comes 174

into contact with the eyes, serious eye damage may result. 175

176

Table 1. Comparison of the main design and operating parameters of 177

the three IMS devices used in the present study. 178

GDA2 G. A. S.

UV-IMS

VG-Test

Type Handheld Portable Desktop

Ion source 63

Ni

100 MBq

UV

Lamp

10.6 eV

Corona Discharge

Standard

inlet

Gas or

vapors

Gas or

vapors

Swab

Drift Tube

temperature

[ºC]

40-45 Ambient 90

Dopant Water

chemistry

Toluene Triethylphosphate

Standard

flow of

sample [mL

min-1

]

400 100 240 + pump

suction

Drift Gas

flow

[mL min-1

]

200 150 200

Shutter

Grid Type

Bradbury

-Nielson

Bradbury

-Nielson

Tyndall-Powell

Opening

Time

[sec]

200 500 200

Drift

Length [cm]

6.29 12 6.4

Pressure (P) Ambient Ambient Ambient

Inlet Type Membrane Open

System

Open System

Electric

field

[V cm-1

]

289 320 280

179

180

181

2.4 Spectral Processing for the mobility spectra 182

The spectral processing for the mobility spectra used in this work 183

consisted of three main steps. (i) Signal preprocessing. (ii) Multivariate 184

signal analysis using a bilinear decomposition with non-negativite 185

constraints followed by quantitative calibration. (iii) The calculation of 186

the limit of detection. 187

The preprocessing of the mobility spectra (i) is needed to improve the 188

accuracy of the posterior data analysis and includes baseline correction, 189

peak alignment and noise filtering. The baseline from each spectrum 190

was corrected by fitting and subtracting a fourth order polynomial 191

using parts of the spectrum where no peaks were identified. Noise 192

reduction was performed using second order Savitzky-Golay [24] filter 193

with a 15 points sliding window. Finally, misalignment of each 194

spectrum was corrected with shift in x-axis (drift time) taking the 195

position of the reactant ion peaks (RIP) – the water peak for the GDA2 196

and the TEP peak for VG-TEST - as reference. This preprocessing 197

procedure was applied independently spectrum by spectrum. 198

However, a special preprocessing technique was applied to the UV-199

IMS spectra because the signal showed interfering noise from other 200

laboratory equipment. Unfortunately the noise was at quite a low 201

frequency (113 Hz arising from the vent pump of the chemical hood) 202

and in-band, precluding the use of frequency selective filters. Thus, a 203

filtering procedure based on principal component analysis (PCA) was 204

attempted. PCA is a technique widely used in image preprocessing to 205

reduce noise and improve signal to noise ratio [25,26]. This is a good 206

solution in those cases where the noise is approximately orthogonal to 207

the signal. Every measurement consisted of 60 spectra that were all 208

used for building the PCA model. Enough loadings were retained to 209

capture 95% of the total variance. Visual inspection of the loadings 210

was used to identify those dimensions that contained basically additive 211

noise. Subsequent spectra were projected to the subspace orthogonal to 212

the noise. At the end, a new data matrix was constructed using only the 213

loadings related to compound information. Assuming noise was 214

stationary during a single measurement; a dedicated PCA model was 215

used for every measurement. On the other hand, the alignment of 216

spectra was uniquely based on the location of the peak of interest 217

because the reactant ion peak (protonated toluene) in the UV-IMS 218

instrument was very broad so it could not be taken as a reference peak 219

for calibration of the drift time scale. 220

The procedure used for processing the ion mobility spectra (ii) was 221

described in some detail recently [20], so only a brief explanation will 222

be given here. The procedure was used for an optimal physical and 223

chemical interpretation of the system, and the analysis of the IMS 224

dataset was done both qualitatively and quantitatively. Multivariate 225

curve resolution with LASSO (MCR-LASSO) [27] was used to resolve 226

IMS spectra yielding a spectral profile and a concentration profile for 227

each species in the sample. Since the analytes were pure and the 228

concentrations rather low, the number of pure components was set at 229

two. One related to the analyte of interest (protonated monomer) and 230

the other related to the dopant (TEP in the VG-Test, Toluene in the 231

UV-IMS and the water chemistry reactant ion peak in the GDA). The 232

resultant calibration profile was used to build a calibration model. In 233

this context, two types of calibration models were performed. The first 234

model was a univariate model which was applied to the measurements 235

done at constant concentration using the volatile generator, and the 236

other was using polynomial-Partial Least Squares (poly-PLS) [28] in 237

the case of headspace measurement. The use of poly-PLS was applied 238

to the concentration profile (after MCR-LASSO) in which an 239

appreciable change in the evolution during the measurement is 240

observed, and at different concentrations the behaviour of the transient 241

changes. Additionally, both the polynomial order and the number of 242

latent variables were calculated using "leave-one-out" cross-validation. 243

Those models were built before calculating the limit of detection. 244

The calculation of the limit of detection (LOD) was based on the upper 245

and lower boundaries of the calibration model and not on blanks, 246

following IUPAC recommendations [29], 247

where q is the slope of the calibration curve, sy is the variance of the 248

data, c represents the concentration levels, t is the t-statistic value with 249

v degrees of freedom and α (0.05) is the confidence level. The v 250

degrees of freedom are estimated according the n number of calibration 251

points. 252

253

3. Results and Discussion 254

3.1 The effect of preprocessing the spectra 255

The signal to noise ratio (SNR) improved due to the preprocessing as 256

seen in the mobility spectra recorded for the VG-Test, GDA2 and UV-257

IMS in Figures 1a, 1b and 1c, respectively. 258

The three IMS devices were located very close to one another on the 259

laboratory bench but the UV-IMS seemed to be more sensitive to 260

external noise and vibrations than the others. This noise is stationary 261

within a measurement. UV-IMS spectra showed a correlated noise that 262

has been corrected with the preprocessing methodology explained 263

above. Although a good SNR has been obtained, the sensitivity to 264

external noise entails the development of tailored preprocessing 265

techniques. An additional consideration is the lack of a reference peak 266

in the UV-IMS because it does not allow accurate well defined peak 267

location and an appropriate alignment between samples (see Figure 1c). 268

However, adding a calibrant in all measurements with a peak located 269

far from the region of interest could be a good solution for the 270

alignment. 271

On the other hand, the preprocessing methodology proposed above for 272

both VG-Test and GDA can be used for off-line processing and the 273

enhanced SNR will improve the performance of both instruments. 274

(a)

(b)

(c)

Figure 1. The raw (left frame) and pre-processed (right frame) mobility 275

spectra of TMA with a concentration of ~5 ppm recorded for the (a) 276

VG-Test (b) GDA (c) UV-IMS. Note that the different reactant ions: 277

TEP, protonated water cluster and toluene, respectively. 278

3.2 The reduced mobility values 279

The reduced mobility values of the ions formed in TMA, putrescine 280

and cadaverine were determined relative to that of 2,4-lutidine [30]. 281

Due to the different operational conditions of the three IMS devices the 282

reduced mobility values appeared to differ significantly in some cases 283

(Table 2). In all cases the major ion produced from each of the 284

biogenic amines was the protonated monomer molecule. The core ions 285

are usually clustered with several water molecules and drift gas 286

molecules and the number of clustered molecules depends on the 287

conditions in the drift tube, mainly on the humidity and operating 288

temperature. Only in the GDA2 were pure proton bridged dimers 289

observed and when both putrescine and cadaverine were present a 290

mixed dimer was also formed. The ratio between the intensity of the 291

dimer and monomer peaks depended on the concentration of the 292

analyte vapors, as expected, but quantitative results were not 293

calculated. In the study of the gas phase ion chemistry of substituted 294

toluene and aniline compounds differences in the reduced mobility 295

values of the product were seen [6]. The formation of proton bridged 296

dimers was observed in the alkyl (methyl and ethyl) aniline derivatives 297

when a 63

Ni and PI ion sources were used but not with the CD source 298

[6]. As noted above, the three RAID1 instruments had similar drift 299

tubes and were operated under similar conditions and the only 300

difference was in the ion source. 301

The reduced mobility value for the putrescine protonated monomer 302

measured with the GDA2 instrument (1.94 cm2V

-1s

-1) differed 303

significantly from the value obtained with UV-IMS (1.99 cm2V

-1s

-1) 304

and the VG-Test device in the present study (2.02 cm2V

-1s

-1) and also 305

from the value of 2.04 cm2V

-1s

-1 reported in a recent work with a drift 306

tube temperature of 232C [31]. However, in studies carried out with 307

the VG-Test at 3QBD, Israel, measuring putrescine or cadaverine 308

vapors emanating from a sample that was placed on a cotton swab, an 309

additional peak for the monomers with reduced mobility values of 1.93 310

and 1.84 cm2V

-1s

-1, respectively, with slightly longer drift times were 311

observed and the relative intensity of the two species changed during 312

the measurement probably due to variations in the operational 313

conditions. Some differences between the three instruments were also 314

observed for the reduced mobility values of protonated monomers of 315

TMA and cadaverine, where TMA was quite close between the VG-316

Test and the UV-IMS and cadaverine was almost similar for the GDA2 317

and UV-IMS. 318

It should be noted that for a given analyte in a given instrument, the 319

drift time and reduced mobility did not change with concentration or 320

with the sample introduction method (constant concentration in an air 321

stream of vapors or headspace vapors). The differences observed in the 322

reduced mobility values were beyond the uncertainties involved in the 323

drift time measurements and mobility scale calibration procedures and 324

are thus indicative of different ion species formed in each IMS due to 325

the differences in operating conditions (ion source, temperature, 326

moisture, reactant ion chemistry and structural features of the drift 327

tube). Thus, discrepancies in reduced mobility values that have been 328

reported for different IMS devices are not necessarily the result of 329

erroneous measurements but rather a natural outcome of variations in 330

ionic species formed under different experimental conditions. This may 331

have major repercussions on the transferability of reduced mobility 332

values between different instruments and even for the same instrument 333

operated under different conditions. 334

335

336

Table 2. The core ions observed in trimethylamine [TMA], putrescine 337

[PUT] and cadaverine [CAD] and their reduced mobility values [cm2 338

V-1

s-1

], calculated relative to 2,4-lutidine [LUT]. All core ions shown 339

here are probably clustered with water and drift gas molecules. 340

Compound

Temperature

Ion Species GDA2

44ºC

G. A. S.

26ºC

VG-Test

90ºC

2,4-Lutidine [LUT]H+ 1.90 1.90 1.90

TMA [TMA]H+ 2.22 2.13 2.10

[TMA]2H+ 1.78 Not observed Not observed

Putrescine [PUT]H+ 1.94 1.99 2.02

[PUT]2H+ 1.46 Not observed Not observed

Cadaverine [CAD]H+ 1.81 1.82 1.87

[CAD]2H+ 1.36 Not observed Not observed

Mixed [CAD][PUT]H+ 1.41 Not observed Not observed

341

3.3 Vapor phase measurements: Calibration and limit of detection 342

3.3.1 Vapors in air 343

There are different ways to display the quantitative response of an IMS 344

to changing amounts of the analyte, whether it is in the vapor phase or 345

deposited on a substrate (filter paper, a glass vial or a swab). One could 346

plot the decrease in the reactant ion peak height or peak area, the 347

increase in the product ion, or ions, produced from the analyte or the 348

ratio of the product ion peak to the sum of reactant and product ions. 349

Table 3 summarizes the response of the GDA2, VG-Test and UV-IMS 350

to the concentrations of TMA, putrescine and cadaverine as predicted 351

according to the MCR-LASSO and PLS procedure described above, as 352

a function of the real concentration of the analyte vapors in air. 353

354

Table 3: The dependence of the response of the VG-Test, GDA2 and 355

UV-IMS on the concentration of trimethylamine, putrescine and 356

cadaverine. 357

Sample

Type

Calculation

VG-Test GDA2 G.A.S

RMSECV* R2 RMSECV R

2 RMSECV R

2

TMA Air (ppm) MCR 0.2 0.95 1.5 0.91 1 0.88

TMA

Headspace

(g)

MCR 0.6 0.94 1.1 0.99 1.1 0.92

Putrescine Air (ppm) MCR 0.3 0.96 0.1 0.95 0.7 0.96

Cadaverine Air (ppm) MCR 0.2 0.97 0.09 0.94 0.1 0.97

Root-mean-square error of cross-validation 358

The sensitivity of the VG-Test is quite similar for the three analytes, 359

but the GDA2 and UV-IMS show a different sensitivity towards TMA 360

compared with cadaverine and putrescine. 361

Table 4 shows a summary of the limits of detection obtained for the 362

three IMS instruments for the three biogenic amines that were 363

measured here. The procedure to determine the limit of detection, 364

based on MCR-LASSO, was given in detail previously [19,20] so only 365

a brief description was presented above. 366

It is well known that the ultimate constraint for limit of detection 367

measurements is the noise level (and signal to noise ratio). For this 368

reason noise filtering and signal enhancement (as provided by MCR-369

LASSO) are key elements for improving the LOD performance for 370

different analytical instruments. In the present study, it should be noted 371

that there were differences in the appearance of mobility spectra and 372

response characteristics of the three instruments. The UV-IMS spectra 373

nominally had the best SNR but this was offset by the lower 374

reproducibility of the device for a given concentration of the analyte. 375

The LOD measurements with the GDA suffered from the long 376

stabilization and clearance times between samples that affected the 377

determination of the blank levels. This could be due to the fact that this 378

device had a membrane in the inlet (Table 1). 379

380

381

Table 4. The limit of detection calculated on the basis of MCR-382

LASSO for vapors of trimethylamine, putrescine and cadaverine in air 383

for the GDA2, UV-IMS and VG-Test ion mobility spectrometers. Air 384

density was taken as 1.16 g L-1

at 300K and 1 Bar. Also shown is the 385

limit of detection for TMA in headspace vapors emanating from a 386

sample deposited in a 20 mL headspace vial. 387

Compound Sample type VG-Test GDA2 G.A.S

TMA Vapor in air (ppm)

nmole/L

0.1±0.2

1.5±2.9

1.2±1.5

11.8±14.7

1.0±1.0

8.5±8.5

Headspace vapors (µg)

nmole

0.9±0.6

15.3±8.8

1.9±1.2

32.2±20.3

0.7±1.1

11.9±18.6

Putrescine Vapor in air (ppm)

nmole/L

0.5±0.3

7.3±4.4

0.7±0.1

6.9±1.0

0.7±0.7

5.9±5.9

Cadaverine Vapor in air (ppm)

nmole/L

0.4±0.2

5.8±2.9

0.2±0.1

2.0±1.0

0.4±0.1

3.4±0.8

388

The LOD showed some dependence on the analyte and on the type of 389

IMS device used, but the range between the lowest LOD (0.1 ppm of 390

TMA vapor with the VG-Test) and highest LOD (1.2 ppm of TMA 391

with the GDA) was quite narrow. The LOD for the diamines varied 392

between 0.2 ppm for cadaverine with the GDA to 0.7 ppm for 393

putrescine with the UV-IMS instrument. The LOD for TMA deposited 394

in a headspace vial ranged from 0.7 to 1.9 g for the UV-IMS and 395

GDA, respectively. The fact that the performance of the three devices 396

was quite similar is really surprising considering that they differ so 397

much from each other in their operating temperature, ionization source, 398

dopant chemistry and drift tube design. If one takes into consideration 399

the uncertainty in the LOD determination there appears to be almost no 400

statistically significant difference in the sensitivity between these three 401

instruments. This finding may have general implications as to the 402

possible limit of detection that can be achieved with classic IMS drift 403

tubes (without pre-concentration or separation). 404

3.3.2 Headspace vapors from deposited sample 405

The response of the three IMS devices to TMA vapors in a headspace 406

vial was similar in principle though different in detail. First a rapid 407

increase within a few seconds in the intensity of the analyte signal, 408

concomitant with the decrease in the reactant ion peak, and then a 409

gradual decrease in the analyte signal and increase in the reactant ion 410

peak as the headspace vapors begin to clear out. This is clearly seen in 411

Figure 2 that depicts the analyte (TMA) and reactant ion peaks in the 412

VG-Test and GDA2 as a function of time. These changes are due to the 413

fact that the flow of the carrier through the headspace vial dilutes the 414

analyte vapors and the rate of dilution depends on the flow and volume 415

of the vial. 416

Theoretically, with perfect mixing and transport, a carrier flow of 400 417

mL min-1

(6.67 mL s-1

) with a 20 mL vial volume should dilute the 418

analyte vapors as shown in the top and middle of Figure 3 for the VG-419

Test and GDA2, respectively. The bottom trace of Figure 3 was 420

recorded for the UV-IMS at a flow of 100 mL min-1

. The graph was 421

adjusted to allow for the 5 seconds delay until the maximum signal 422

intensity of the analyte is obtained. Evidently, the clearing time is 423

longer due to imperfect mixing and transport of analyte vapors from 424

the sample vial into the drift tube. 425

426

427

Figure 2. Top frame: The signal intensity of the analyte (TMA) and reactant ion (TEP) 428

peaks in a headspace vial as a function of time. Bottom frame: The signal intensity of 429

the analyte (TMA) and reactant ion (RIP) peaks in a headspace vial as a function of 430

time. 431

432

433

434

435

436

437

438

439

440

441

442

443

444

445

446

447

448

449

450

451

452

453

454

Figure 3: The theoretical (with perfect mixing) dilution of the TMA 455

headspace analyte vapors for a carrier flow of 400 mL min-1

(6.67 mL 456

s-1

) and a 20 mL vial volume for the VG-Test device (top) and the 457

GDA (middle). The response of the UV-IMS and theoretical dilution 458

for a flow rate of 100 mL min-1

is also shown (bottom trace). For the 459

actual measurement the graph (based on Figure 2) was adjusted to 460

allow for the first 5 seconds until the maximum signal intensity of the 461

analyte is obtained. 462

Figure 4 shows the response of the VG-Test to different amounts of 463

TMA that were placed in a headspace vial as a function of time for the 464

VG-Test (top) and the GDA (bottom). The plots of the ratio of the 465

analyte peak [(TMA/(TMA+TEP)] in the VG-Test and 466

[(TMA/(TMA+RIP)] in the GDA2 show that the ratio and clearance 467

time increases with the amount of TMA deposited in the vial, as 468

expected. The calibration curve and limit of detection calculated 469

according to the MCR-LASSO procedure are shown in Tables 3 and 4. 470

In this case the limit of detection for TMA headspace vapors with the 471

GDA2 is slightly inferior to the LOD calculated for the VG-Test and 472

UV-IMS but all values are in the range of 0.7-1.9 µg (6.9-21.6 nmole) 473

of TMA deposited in the headspace vial. 474

Measurement of headspace vapors of putrescine produced a different 475

response in all three IMS devices. The putrescine signal was not 476

observed initially, even when the headspace vial was heated in the 477

homemade oven to 100ºC for several minutes, after about five minutes 478

the intensity of the putrescine peak started to increase and continued to 479

do so long after, theoretically, all the vapors in the headspace vial 480

should have been transported to the IMS. The intensity continued to 481

increase even after the vial was removed, indicating that vapors 482

remained in glass vial, the transfer line and the instrument itself. This 483

phenomenon made it quite impossible to create a calibration plot for 484

putrescine similar to that made for TMA headspace vapors. 485

486

487

488

489

490

491

492

493

494

495

Figure 4: The response to different amounts of TMA (in µg) that were 496

placed in a headspace vial as a function of time for the VG-Test (top) 497

and the GDA (bottom). Note that the ratio [(TMA/(TMA+TEP)] (top) 498

and [(TMA/(TMA+RIP)] (bottom) and the clearance time increase with 499

the amount of TMA deposited in the vial. 500

501

502

4. Conclusions 503

The variation in the reduced mobility values determined with the three 504

IMS instruments is indicative of the fact that there are different ion 505

species formed in each IMS due to the differences in operating 506

conditions (ion source, temperature, moisture level, reactant ion 507

chemistry and structural features of the drift tube). This could be 508

attributed to variation in the degree of clustering that is affected by the 509

different operating temperatures and the humidity in the three devices 510

and to the nature and characteristics of the core ion. As mentioned 511

above, this may affect the transferability of reduced mobility values 512

between different instruments (especially when operated at low drift 513

tube temperatures and uncontrolled humidity) and even for the same 514

instrument operated under different conditions. A case in point is the 515

study where different product ions were found in three similar IMS 516

devices, operated with identical parameters except the type of ion 517

source used [6]. Quantitative measurements to determine the limit of 518

detection were not reported in that study [6]. The practical conclusion 519

is that while a database of reduced mobility values can serve as a 520

guideline for identification of ion species, the analyte should be 521

measured and calibrated under the same operational conditions with the 522

same IMS instrument that will serve as an analytical tool. 523

The difference between the theoretical dilution of the analyte vapors 524

and the measured signal intensity in the headspace experiment shows 525

that the clearing time is longer, due to imperfect mixing and delayed 526

transport of analyte vapors from the sample vial into the drift tube. The 527

measurement of headspace vapors of putrescine showed that transport 528

of vapors from the heated headspace vial was much less efficient than 529

the transport of the volatile TMA vapors. This was established by the 530

fact that the putrescine peak increased slowly after a considerable delay 531

and continued to do so long after all the vapors in the headspace vial 532

should have been transported to the IMS theoretically. This 533

phenomenon made it quite impossible to create a calibration plot for 534

putrescine similar to that made for TMA headspace vapors. 535

Building calibration curves with raw spectra or noise filtered spectra 536

can lead to very different results. Correct preprocessing of spectra and 537

applying the appropriate multivariate signal processing is essential 538

before establishing an appropriate comparison, since this study shows 539

that different instruments have different noise levels. Those differences 540

may be partially due to different conditions regarding grounding, 541

shielding and cabling and obviously different electronic filtering and 542

amplification signal chains in the instrument electronics as well 543

external mechanical interfering noise. We believe that proper digital 544

filtering can recover the inherent noise limits of the different IMS 545

technologies. Surprisingly, the three devices showed quite similar 546

limits of detection for the three analytes although they differ so much 547

in their operating temperature, ionization source, dopant chemistry and 548

drift tube design. Considering the uncertainty in the LOD 549

determinations there appears to be almost no statistically significant 550

difference between these three instruments. This finding may have 551

general implications as to the possible limit of detection that can be 552

achieved with classic IMS drift tubes (without pre-concentration or 553

separation). 554

555

References 556

1. G.A. Eiceman, Z. Karpas, Ion mobility spectrometry – Second Edition, 557

Chapter 1, CRC Press, Boca-Raton, FL, 2005. 558

2. R. R. Kunz, W. F. Dinatale, P. Becotte-Haigh, P. Int. J. Mass 559

Spectrom. 226 (2003) 379–395. 560

3. L.A. Viehland, R. Guevremont, R.W. Purves, D.A. Barnett, Int. J. 561

Mass Spectrom. 197 (2000) 123–130. 562

4. H. Borsdorf, J.A. Stone, G.A. Eiceman, Int. J. Mass Spectrom. 246 563

(2005) 19-28. 564

5. H. Borsdorf, M. Rudolph, Int. J. Mass Spectrom. 208 (2001) 67-72. 565

6. H. Borsdorf, K. Neitsch, G.A. Eiceman, J.A. Stone, Talanta 78 (2009) 566

1464-1475. 567

7. M. Joshi, Y. Delgado, P. Guerra, H. Lai, J.R. Almirall, Forens. Sci. Int. 568

188 (2009) 112-118. 569

8. C.W. Su, K. Babcock, S. Rigdon, Int. J. Ion Mobil. Spectrom. 1 570

(1998)15-27. 571

9. S.S. Choi, et al. Bull. Kor. Chem. Soc. 31 (2010) 2382-2385. 572

10. K. Cottingham, Anal. Chem, (2003) (Oct. 1), 435A-439A. 573

11. S. Armenta, M. Alcala, M. Blanco, Anal. Chim. Acta 703 (2011) 114-574

123. 575

12. K. Igarashi, K. Kashiwagi, Biochem. Biophys. Res. Comm. 271 (2000) 576

559-564. 577

13. J.D. Sobel, Z. Karpas, A. Lorber, Eur. J. Obstet. Gynecol. Reprod. 578

Biol. 163 (2012) 81–84. 579

14. Z. Karpas, W. Chaim, R. Gdalevsky, B. Tilman, A. Lorber, Anal Chim 580

Acta 474 (2002) 115–123 581

15. U. Bachrach, Amino Acids 26 (2004) 307–309. 582

16. G.M. Bota, P.B. Harrington, Talanta 68 (2006) 629–635. 583

17. Z. Karpas, B. Tilman, R. Gdalevsky, A. Lorber, Anal Chim Acta 463 584

(2002) 155–163. 585

18. C. Y. Chong, F. Abu Bakar, A. R. Russly, B. Jamilah, N. A. Mahyudin, 586

Intern. Food Res. J. 18 (2011) 867-876. 587

19. Z. Karpas, A.V. Guaman, D. Calvo, A. Pardo, S. Marco, Talanta 93 (2012) 588

200-205. 589

20. V. Pomareda, A.V. Guamán, M. Mohammadnejad, D. Calvo, A. Pardo, 590

S. Marco, Chemeometrics Intel. Lab. Syst. (2012), 591

DOI: 10.1016/j.chemolab.2012.06.002. 592

21. www.airsense.com/en/products/gda-2/ (accessed October 25, 2012) 593

22. www.gas-dortmund.de/Products/UV-IMS/1.336.html (accessed 594

October 25, 2012) 595

23. www.3qbd.com/English/ (accessed October 25, 2012) 596

24. A. Savitzsky, M.J.E. Golay, Anal. Chem. 36 (1964) 1627–1639. 597

25. P. Razifar, H.H. Muhammed, F.Engbrant, P. Svensson,J.Olsson,E. 598

Bengtsson, B Langstrom, M. Bergstrom, The Open Neuroimaging J. 3 599

(2009) 1-16. 600

26. M. Statheropoulos, A. Pappa, P. Karamertzanis, H.L.C Meuzelaar, 601

Anal. Chim. Acta, 401(1999) 35-43 602

27. V. Pomareda, D. Calvo, A. Pardo, S. Marco, Chemometrics Intel. Lab. 603

Syst. 104 (2010) 318-332. 604

28. S. Wold, N. Kettaneh-Wold, B. Skageberg, Chemometrics Intel. Lab. 605

Syst., 7 (1989) 53-65. 606

29. J. Mocak, A. M. Bond, S. Mitchell, S. Scollary, Pure & Appl. Chem, 607

69 (1997) 297-328. 608

30. G.A. Eiceman, E.B. Nazarov, J.A. Stone, Anal. Chim. Acta 493 (2003) 609

185-194. 610

31. S. Armenta, M. Blanco, Analyst (2012) doi:10.1039/C2AN35965K. 611

612

613

Legend of figures 614

Figure 1. The raw (left frame) and pre-processed (right frame) mobility 615

spectra of TMA with a concentration of ~5 ppm recorded for the (a) 616

VG-Test (b) GDA (c) UV-IMS. Note that the different reactant ions: 617

TEP, protonated water cluster and toluene, respectively. 618

Figure 2. Top frame: The signal intensity of the analyte (TMA) and 619

reactant ion (TEP) peaks in a headspace vial as a function of time. 620

Bottom frame: The signal intensity of the analyte (TMA) and reactant 621

ion (RIP) peaks in a headspace vial as a function of time. 622

Figure 3: The theoretical (with perfect mixing) dilution of the TMA 623

headspace analyte vapors for a carrier flow of 400 mL min-1

(6.67 mL 624

s-1

) and a 20 mL vial volume for the VG-Test device (top) and the 625

GDA (middle). The response of the UV-IMS and theoretical dilution 626

for a flow rate of 100 mL min-1

is also shown (bottom trace). For the 627

actual measurement the graph (based on Figure 2) was adjusted to 628

allow for the first 5 seconds until the maximum signal intensity of the 629

analyte is obtained. 630

Figure 4: The response to different amounts of TMA (in µg) that were 631

placed in a headspace vial as a function of time for the VG-Test (top) 632

and the GDA (bottom). Note that the ratio [(TMA/(TMA+TEP)] (top) 633

and [(TMA/(TMA+RIP)] (bottom) and the clearance time increase with 634

the amount of TMA deposited in the vial. 635

636

List of Tables 637

Table 1. Comparison of the main design and operating parameters of 638

the three IMS devices used in the present study. 639

Table 2. The core ions observed in trimethylamine [TMA], putrescine 640

[PUT] and cadaverine [CAD] and their reduced mobility values [cm2 641

V-1

s-1

], calculated relative to 2,4-lutidine [LUT]. All core ions shown 642

here are probably clustered with water and drift gas molecules. 643

Table 3: The dependence of the response of the VG-Test, GDA2 and 644

UV-IMS on the concentration of trimethylamine, putrescine and 645

cadaverine. 646

Table 4. The limit of detection calculated on the basis of MCR-647

LASSO for vapors of trimethylamine, putrescine and cadaverine in air 648

for the GDA2, UV-IMS and VG-Test ion mobility spectrometers. Air 649

density was taken as 1.16 g L-1

at 300K and 1 Bar. Also shown is the 650

limit of detection for TMA in headspace vapors emanating from a 651

sample deposited in a 20 mL headspace vial. 652

653