Document [original]

Optimization of smartphone-based evaluation of tube-type

passive samplers using atmospheric nitrogen dioxide determination

as an example

Shi Chenglu, He Xinyi, Kanokwan Kiwfo, Andreas Held and Wolfgang Frenzel*

Technische Universität Berlin, Institute of Environmental Technology,

Environmental Chemistry and Air Research

Str. d. 17. Juni 135, D-10623, Berlin, Germany

This work is licensed under a Creative Commons Attribution 4.0 International License (CC BY

4.0), http://creativecommons.org/licenses/by/4.0

Available via the institutional repository of Technische Universität Berlin:

http://dx.doi.org/10.14279/depositonce-20617

Abstract

Key parameters affecting smartphone-based color evaluation of tube-type passive samplers for

determination of nitrogen dioxide (NO2) have been thoroughly optimized. Photographic

parameters investigated were conditions of illumination and distance between the smartphone

and the colored objects. Other parameters optimized refer to the measurement of the color depth

of the nitrite assay (nitrite being the reaction product formed when NO2 is sampled with

triethanolamine (TEA) as the sorptive medium). Photos were directly (in-situ) taken of the

colored liquids contained in the cap of the passive sampler tube. RGB-values of the photos were

retrieved using ImageJ®. Calibration was done using liquid standards prepared and processed in

the same way as the sample solutions. As a result of the optimization high sensitivity of nitrite

determination with a limit of detection (LOD) of 12 ng, a working range of 30-600 ng, and good

precision (<8% RSD) was obtained. For 14 days sampling and using Palmes-type passive

samplers, the working range and LOD for atmospheric NO2 are 10-120 µg m-3 and 5 µg m-3,

respectively. The experience gained during the investigation and optimization of relevant

experimental parameters was exemplarily put into practice in small measurement campaigns

determining ambient NO2 in the city of Berlin (Germany).

*Corresponding author. E-mail: [email protected]

Keywords: passive sampling; NO2 measurement; ambient air; smartphone-based color

evaluation; photographing at ambient light; photo box omitted; freehand photographing;

within sampling-tube (in-situ) detection

Introduction

Passive (also termed diffusive) sampling techniques for gaseous air pollutants have

attracted considerable attention in the past decades [1-6]. They have been extensively

employed in numerous research projects and for nitrogen dioxide (NO2) determination

in ambient air they are a widely used and accepted complementary method in several

national and regional networks [7-10]. Passive sampling of NO2 in ambient air has also

been implemented e.g., into Official Standard Methods of the European Union [11].

The tube-type passive samplers introduced by Palmes et al. [12] in the 1970s are

probably the most widely employed configuration in ambient air monitoring. What

makes this sampler construction particularly valuable is its simple construction using

inexpensive materials and the fact that the atmospheric concentration can be calculated

from the amount of gaseous compound or a reaction product thereof, thus omitting gas-

phase calibration [3,4,8,9]. In order to collect the analyte gas the bottom of the passive

sampler (generally a removable cap) contains a solid adsorber material or an absorber

liquid immobilized on the surface of a support material. A typical analytical procedure

for evaluation of passive sampler tubes encompasses, after gas collection, the extraction

of the collected analyte or a reaction product from the adsorber or the support material

followed by transfer of the extract into a vessel before determination is performed using

an appropriate analytical determination method. Quite often photometric methods are

applied [4,5,7]. In this case a sometimes-reported alternative to the extraction of the

analyte and derivatization in separate vessels is to provide the color forming reagent(s)

directly into the cap of the passive sampler and only after completion of the reaction to

transfer the colored solution into a cuvette for photometric determination [12-14].

Hence, most of the solutions handling steps remain. We have not found any study or

publication where photometric determination has been performed in-situ within a tube-

type passive sampler. Obviously, this is hindered since transmission measurements

cannot easily be accomplished with the given geometry of caps of passive samplers. On

the contrary, reflectometric detection is the preferred method whenever the colored

reaction product of a photometric assay is present in a flat geometry as is for instance

the case of paper-based analytical devices, lateral flow (immuno) assays and test strip

evaluation [15-18]. It is also often applied to the analysis of liquids contained in micro-

titer plates and as thin layers distributed on solid reflective surfaces (spot tests) [19,20].

Instruments for reflectometric detection are either commercially available or purpose

made reflectometers and photometric plate readers. Several of the instruments provide

detection of (only) a selected wavelength required for the specific photometric assay,

while others permit multi-wavelength detection or scanning options with high spectral

resolution. Reflectometer configurations encompass bench-top instruments, modular

systems and hand-hold portable devices. Differences among the available instruments

also exist between systems where the colored samples are placed within the instrument

(free-space reflectometers) and those with light-guiding (optical) fibers offering high

geometric flexibility [21,22]. Yet, another application of reflectometric detection is the

evaluation of colored spots in thin-layer chromatography; this is often termed

densitometry [23,24]. The common instrumentation for this application are so-called

densitometers which are commercially available in a variety of configurations [25].

Hence, considering the idea of the present work of in-situ detection of the colored

reaction product within a passive sampler tube (or more precisely in a cap containing

the absorber immobilized on a supplementary material) reflectometric detection

appeared to be the optical method of choice and was therefore explored.

The explosive development in the broad field of opto-electronics during the past

decades has revolutionized (not only) optical detection with formerly unimaginable

capabilities. One important area of digital color and color depth measurements for

analytical chemical purposes is the evaluation of photometric assays using digital

scanner and smartphone-based imaging. Both are the subject matter of several reviews

[26-31]. The evaluation of the huge number of published papers has evidenced that

measurement of absorbance/transmission and reflection are about equally often applied.

When photos are taken from colored objects for further evaluation, it is the light

reflected from the illuminated surface (i.e., the remission value being the equivalent to

transmission in absorbance measurements) that is detected. Thus, this type of

measurements can be meaningfully categorized as reflectometry. However, the

scientifically appropriate term “reflectometry” for many of the color imaging appli-

cations using digital scanners or smartphone photographing is rarely used in the relevant

literature. The appropriate term for quantification of remission is the reflection density

(formerly termed optical density) [32-34]. Instead, in most papers on digital color

imaging and quantification the reflection density has been termed intensity (e.g.,

[26,27]), effective intensity [35] or it is sometimes - very unfavorably - named

“effective absorbance” [36,37]. The term color depth is sometimes used to qualitatively

describe the color impression by bare eye observation [31].

More recently, the use of smartphones for color detection in the evaluation of passive

sampling of gases has been presented in quite a number of papers [38]. However, the

term passive sampling is only rarely used in the title (and often also not throughout the

entire text) of relevant publications. A closer inspection of all papers where gas

sampling was done passively by molecular diffusion followed by smartphone-based

color detection and evaluation revealed that there are severe differences regarding the

design of the sampling devices and the way how color detection and evaluation has

been performed. In a recent comprehensive review [38] we have drawn the attention to

the fact that a conceptual difference exists between passive samplers that are suitable for

gas concentration measurements and those providing information about the time-

weighted deposition flux of the gaseous compounds of interest. Regarding the former

configuration the determination of O3 [39] and NO2 [40] are the sole applications.

In a paper of Cerrato-Alvarez et al. [39] the combination of passive sampling of ozone

using commercially available Ogawa passive samplers followed by digital analysis of

images taken with the camera of a smartphone is reported. The analytical signal used

was the degree of decolorization of the blue color of indigotrisulfonate (ITS) deposited

on cellulose filters upon the reaction with ozone. After sampling the filters were

removed from the sampler bodies and placed in a photobox before images of the filter

pads were taken and evaluated by digital analysis. Souza et al. [40] reported the use

passive samplers for NO2 determination that were purpose-made from conical Falcon

tubes used in chemical laboratories cut to the desired size. Gas collection was made

using a mixture of triethanolamine (TEA), ethylene glycol and acetone as the sorbent

immobilized on Whatman cellulose filters. For determination of the nitrite formed

during adsorption of NO2 the filter papers were removed from the sampler and - in a

rather complicated procedure - a gel containing the Griess-Ilosvay reagent was

distributed across the collecting surface with a plastic ruler whereafter the filters were

placed in a photobox and pictures taken with a smartphone.

In order to simplify the analytical procedure of evaluating passive samplers we have

investigated possibilities avoiding the removal of the absorber from the passive

samplers for subsequent color detection. Instead, in-situ detection was employed using

either visual color comparison after addition of the color forming reagent within the

sampler [41,42] or the application of smartphone-based detection [42,43]. For principle

investigations we have so far focused our work on the determination of NO2. This is due

to the fact that NO2 is one of the most problematic gaseous pollutants [44-48] and in our

working group a long-lasting experience exists with passive sampling of NO2 (e.g., [49-

53]). Hence, we are familiar with suitable passive sampler configurations and the

chemical assay for determination of nitrite being the reaction product of NO2 with

triethanolamine (TEA); TEA serving as the common trapping compound for NO2

[7,8,11,12].

In the present paper, the thorough optimization of an in-situ smartphone-based

evaluation procedure of the colored azo dye solution of the Griess-Ilosvay assay [54,55]

formed within the caps of a tube-type passive sampler is described. Since avoiding the

use of a photo box (applied in the majority of papers referring to smartphone-based

color detection (e.g., [26,28,31] and references therein) for photographing the colored

objects was intended, photographic parameters such as conditions of illumination and

distance between the smartphone and the colored objects were investigated. Additional

parameters studied refer to the evaluation of the color depth of the reaction product

formed in the nitrite assay within the sampler caps. RGB values were eventually

retrieved from the photographs using the ImageJ® software [28-31,56,57]. Conside-

rable efforts of our current work related to smartphone-based evaluation of passive

samplers have been devoted to finding a suitable and reliable way of calibration [43].

This will also be briefly discussed in this paper. Application of the optimized method to

the determination of NO2 in ambient air in Berlin will be presented and results

compared with a common photometric evaluation procedure of passive samplers

established at our institute for many years [49-53].

Experimental

Reagents and solutions

All reagents used were of analytical grade quality. Deionized water was used for

preparing solutions. A nitrite stock solution (10 g L-1 NO2-) was prepared by dissolving

1.500 g NaNO2 in 100 mL of water. This solution was further diluted to obtain nitrite

standards for spike experiments and calibration purposes (see below). The Griess-

Ilosvay reaction [54,55] was used for derivatization of nitrite. The fundamental

chemical assay is part of an official standard method for active sampling of NO2 which

goes back to a still valid procedure published by Saltzman as early as 1954 [58,59].

Therefore, quite often in the context of NO2 determinations this type of reaction has

been named Saltzman or Griess-Saltzman method.

Of the many variants of the underlying assay published in connection with passive

sampling of NO2 (e.g., [7,8,10,11,60]) we basically adhered to the receipt used by

Palmes et al. [12] in their early paper on photometric evaluation of passive samplers.

One component of the derivatization reaction is sulfanilamide (SA). The SA reagent

solution was prepared by weighing 10 g SA into a 500 mL volumetric flask, adding 25

mL 85% o-phosphoric acid and about 350 mL water. After complete dissolution the

flask was filled up with water to the mark. The second reagent for the derivatization

reaction is N-[1-Naphthyl]-ethylenediamine dihydrochloride (NEDD). For preparing

this solution 140 mg NEDD were weighed into a 100 mL volumetric flask containing

about 80 mL water. After dissolution, the flask was filled up with water to the mark and

the solution filled into a brown glass bottle. The SA and NEED solutions were stored in

a refrigerator at ~ 6°C. To serve as a color reagent for determination of nitrite these two

solutions (further on termed combination reagent) were mixed 10+1 (SA+NEDD) on

the day of use and not stored longer than ~ 8 h before nitrite determinations were

performed. This was done because the shelf life of the combination reagent is relatively

low being indirectly recognizable by an increase of the reagent blank value with time.

The TEA absorber solution used as a trapping agent for NO2 collection was prepared by

weighing 20 g of TEA (Merck, Darmstadt) into a beaker and adding 80 g of water.

Thorough mixing is required due to the high viscosity of TEA. A small volume of this

20% (w/w) TEA solution was directly applied with a micropipette to the absorber pads

when passive samplers were prepared (see below). Care was taken to open the bottle

with TEA solution only for a minimum of time when the liquid was withdrawn for

preparation of the passive samplers since the liquid surface of the TEA solution acts

already as a passive sampling medium for NO2 in an uncontrolled manner leading to

elevated blank values.

Construction and preparation of passive samplers

The tube-type passive samplers employed in this work were of the Palmes design

[2,7,12,60]. Figure 1(a)-(c) show views of different passive samplers and photographs

of the assembled samplers fixed in a purpose made holder and mounted in a shelter at

the sampling site, respectively. Polymethacrylate tubes of 7.0, 1.0 and 1.5 cm lengths,

inner and outer diameter, respectively, were used for all samplers. Cellulose pads of

13.0 mm diameter were punched out from Whatman filter paper 41 and used as support

for the TEA absorber solution as an alternative to the stainless-steel grids used by

Palmes et al. [12] and in many other publications and reports [7-9,11]. Three different

types (Types 1-3) of flexible polymer caps were used (see Figure 1(a)) to accept the

cellulose pads. An additional one served as closure of the passive samplers during

transportation and storage. Type 1 caps are yellow polyethylene caps of 15 mm i.d. and

an inner height of 20 mm (protective caps for pipes GPN-210, Pöppelmann, Lohne,

Germany [61]).

Figure 1. (a) Exploded view of the tube-type passive samplers with three different caps for

insertion of the absorber pads employed in the present work. See text for further details. (b) 3D

printed holder and four passive samplers inserted in the holder. (c) Final configuration of

passive samplers fixed in a plastic shelter displayed at the sampling sites. One sampler tube of

four remains closed and serves as field blank.

They have been used for application of the photometric standard procedure established

in our laboratory and in a preliminary test of smartphone-based evaluation (see below).

Type 2 caps are opaque white elastic silicon caps (HKIT fastener, German distributor

[62]) with 14.7 mm i.d. and an inner height of 15 mm. They were used in few initial

experiments when smartphone-based evaluation was employed, yet later exchanged for

Type 3 caps. Caps of Type 3 were cut from Type 2 caps to a smaller inner height of

8 mm in the university workshop and applied in the majority of the investigations using

smartphone-based evaluation presented in the present work. Slightly conical yellow

caps (protective caps for pipes GPN-610, Pöppelmann, Lohne, Germany, see Figure

1(a)) with 16 mm i.d. at half height were used as closures for all passive samplers.

In order to prepare the passive samplers for at-site NO2 collection a single pad is placed

in the respective caps and 20 µL of TEA-solution pipetted onto this pad. The 20 µL

volume evenly distributes in the porous cellulose pads due to capillary forces and no

drops are formed at the surface. Then the polymethacrylate tubes are pressed thoroughly

into the caps so that the pads are fixed tightly at the bottom. The closing caps are finally

fixed on top of the tubes. Four sampling tubes were hooked in a 3D-printed holder (see

Figure 1(b)) and the holder was fixed in a plastic shelter of about 500 mL capacity

serving as a protection device from heavy rain and as a windbreak (see Figure 1(c)).

The samplers are then ready for transportation to and deployment at the sampling sites.

Apparatus for photometric detection and smartphone-based evaluation

Photometric measurements were used in the present paper only as a reference for the

determination of nitrite (see below). To this end a common laboratory instrument

(Shimadzu, Model UV-1201) was employed with a wavelength set to 540 nm, i.e., the

maximum of the absorption band of the pink azo compound formed in the Griess-

Ilosvay reaction. Absorbance measurements were made in a 1 cm path-length cuvette

against pure water. The smartphones used for taking all photographs were Apple

iPhones (Model 12mini and model XR). In previous work [43] it was shown that they

yield identical color response when used for the present application. The photos were

stored in JPEG format on a laptop and uploaded into ImageJ® for recording of RGB

values. Details of the entire color evaluation procedure are given in the supplementary

material (see at the end of the manuscript).

A home-made photo box was employed for some of the measurements. It was a cube of

23.3 cm x 20.6 cm x 31.5 cm width, depth, and height, respectively. The front side of

the photo box was left open to have easy access for placing the objects to be

photographed on the bottom of the cube. While the photos were taken the front was

closed with an opaque black curtain to prevent disturbances by ambient light. The

removable cover plate of the cube was made with two side-by-side cutouts, one for the

light source (Key Light mini (20LAD9901), Corsair Memory Inc, Taiwan) and the other

for taking pictures with the smartphone.

Standard photometric procedure for nitrite determination of extracts from

passive samplers

For extraction of nitrite from the absorber pads exactly 2 mL water is pipetted into the

yellow caps of Type 1. The solutions are repeatedly aspirated and dispensed with a

pipette to achieve complete extraction and homogeneous distribution of nitrite in the

extract volume. Then 1 mL is taken off and transferred into a small glass vial followed

by the addition of 1 mL of the combination reagent (see Reagents and Solutions).

Following the reagent addition, the solution is mixed and after a waiting time of 15-20

min for completion of the color reaction part of the solution is filled into a 1-cm path-

length cuvette and the absorbance measured at 540 nm against pure water. For

calibration purposes nitrite standards were prepared and further treated likewise to the

extracts. The results obtained with the standard procedure were used as reference for

comparison with the proposed optimized smartphone-based evaluation procedure with

respect to sensitivity, precision, and limit of detection (LOD). The standard procedure

was also employed in ambient air NO2 measurements and served likewise as the in-

house reference method.

Preparation of test solutions for optimization of smartphone-based

evaluation

For parameter optimization of smartphone-based evaluation the caps Type 1-3 intended

for use in passive samplers were repurposed to a kind of mini-cups for the reaction

between nitrite standards and the combination reagent. To this end the absorber pads

were inserted into the (now named) mini-cups and a certain volume of nitrite standard

solutions of different concentrations (termed nitrite spike in the following) pipetted onto

the pads. Then, a defined volume of the combination reagent was added and to achieve

a homogeneous mixture the liquid volume repeatedly (4-5 times) aspirated and

dispensed by means of the pipette. After completion of the reaction the colored

solutions formed were eventually photographed using the built-in camera of the

smartphone. All photos taken were saved in JPEG format and imported into ImageJ®

for retrieval of RGB values. Only the green color channel was used for evaluation of

color depth since green is the complementary color to red-violet and gives the largest

change of reflectance values with analyte concentration. Further details of color

measurements are outlined in the supplementary material. Calibration of the

smartphone-based evaluation procedure and analysis of the real ambient air samples

was done using the optimized experimental conditions (vide infra).

Procedure of NO2 measurement in ambient air

Passive samplers for real sample collection were prepared in the laboratory as described

above. The plastic shelters containing four samplers each were transferred to the desired

sampling sites. Two shelters were mounted at each site in suitable locations (e.g., lamp

posts, canopies, balustrades of balconies) at a height of 1.5-2.5 m and a minimum

distance from the walls of buildings of 1 m to warrant freely diffusive analyte transport

from all directions. One of these shelters was equipped with the previously employed

and proven passive samplers with yellow PE caps [50-53], the other one with passive

samplers fitted with the cut-to-shorter-length white opaque silicon caps. After the

allotted sampling period the shelters were collected, transferred to the laboratory, and

analysed within the next two days latest. The evaluation procedure for samplers with

yellow caps has been described above. For smartphone-based evaluation of the white

opaque caps the optimized experimental conditions outlined below were used.

Results and Discussion

In the following sections the experiments conducted for optimization of parameters

affecting smartphone-based evaluation are described and results obtained presented.

One part of the optimization refers to parameters relevant to taking the photos, i.e., the

color of the caps of the passive samplers, the distance and orientation of the objects to

be photographed in relation to the camera, and the illumination conditions. Another

parameter optimized is the volume of the colored liquid reaction product contained in

the sampler caps. The optimized conditions have been used for the final calibration of

smartphone-based nitrite determination and eventually applied to the determination of

NO2 in ambient air using passive sampling. Results of nitrite calibration using the

established photometric standard procedure are also briefly presented since this

procedure has been used for comparative reasons with respect to analytical

specifications (reproducibility, sensitivity and LOD) and of real sample analysis. In

order to take photos of the caps they were placed on a PVC supporting block. The

smartphone was generally held freehand in a parallel position right above the objects to

be photographed. In a few instances the smartphone was mounted on a tripod to permit

precise setting of the distances between the smartphone and the objects. The center of

the image (visible on the screen of the smartphone camera as a small circle) was

adjusted to be above the middle of the objects.

Influence of the color of the caps of the passive collectors

In previous work of our group on passive sampling with Palmes-type sampler tubes we

have used low cost and readily available yellow caps intentionally made as protective

end caps for installation tubing [61]. When these caps were employed in an

investigation about in-situ color detection by visual color comparison [41,42], the color

impression (hue) was totally different from the red-violet color of the nitrite assay; it

appeared red-orange instead (see Figure 2). Therefore, in that study opaque white caps

had finally been used. In order to investigate the influence of the color of the caps in the

evaluation of the colored adsorber pads using smartphone photographing and retrieval

of RGB values the yellow polyethylene caps and white opaque silicon caps of almost

the same dimensions have been compared. In both cases, absorber pads were inserted

into the caps and 30 µL of a 5 mg L-1 nitrite standard solution (corresponding to 0.15

µg nitrite) or pure water were added with a micropipette. After the solutions were

distributed within the absorber pads by capillary forces, 1 mL of the combination

reagent was added to all caps. Repeated aspiration and draining of the solution with the

micropipette ensured that the nitrite standard and the blank solutions became well

mixed with the color reagent. After a reaction time of approx. 15-20 min, joint photos of

all caps were taken with the smartphone under room light illuminated by fluorescence

tubes mounted at the ceiling of the laboratory. In the photo shown in Figure 2 the

respective RGB values are also given for the four different caps. Comparison of data of

yellow and white caps shows that the reflectance values of the individual color channels

are considerably different. For the blank solutions the reflectance is (expectedly) higher

when white caps are used. The reflectance values of the green channel for the standard

solution indicate a larger difference with respect to the blank for the white cap, too.

When the yellow caps are used, the very low reflectance of the blue channel (because

blue is the complementary color to yellow) is noticeable. Small differences of the red

channel and no difference of the blue color channel exist between the blank and the

standard solutions.

Figure 2. Photo of blank (left) and standard solutions (right) of azo dye formed in the Griess-

Ilosvay reaction in caps of different color. In both cases absorber pads were inserted into the

caps prior to application of the respective solutions. The RGB (red-green-blue) reflection values

of the respective solutions retrieved are given at the left and right edge of the photo.

As a result of these findings in all further measurements the white opaque silicon caps

have been employed. For evaluation of the reflection density only the green color

channel was used further on.

Study on blank values

For the determination of blank values in the smartphone-based evaluation, three pairs of

blank solutions pipetted into the caps of the passive samplers with inserted pads were

photographed together and the RGB values retrieved. The images are shown in

Figure 3. The measurements took place in the laboratory without any further protective

device (no clean air box). The photos were taken at room light from a height of about 50

cm, once immediately after application of the solutions (Figure 3(a)) and once after a

waiting time of about 2 hours (Figure 3(b)). The second image was taken in order to

observe possible changes of the reflectance due to color ageing or an increase of the

reagent blank due to collection of ambient NO2 from the laboratory atmosphere. To two

of the caps only 250 µL of pure water was pipetted, thus representing the zero value and

reference point for maximum remission. To two further caps 250 µL of the combination

color forming reagent was added representing the reagent blank. To yet two other caps

20 µL of TEA and 250 µL of the combination reagent were added. These caps represent

the passive sampler’s preparation blank values.

Figure 3. Photos taken of blank solutions (a) directly after pipetting the solutions into the cap

containing absorber pads and (b) after a waiting time of about 2 h. The numbers between the

caps refer to the reflectance values (average values of the two duplicate solutions positioned

above each other) of the green color channel. For further explanations see text.

The reflectance values of the green color channel of the blank solutions retrieved

immediately after preparation of the sampling caps (Figure 3(a)) do not differ. Also, the

reflectance value of the pure water blank is the same after two hours waiting time.

However, the blank values of the cups containing the reagent solutions with and without

TEA added (last two caps in the two rows) are significantly lower, i.e., the reflectance is

lower, and this is caused by collection of NO2 from the laboratory atmosphere and start

of the colorization reaction. Blank values can be minimized by preparing the passive

samplers in a glovebox [49,50]. For reasons of practicability, we have omitted that in

the current work and have closed the passive samplers immediately after pipetting the

TEA solution onto the cellulose absorber pads. After return of the passive samplers

from the sampling site and removal of the caps containing the pads, the further steps

were conducted as fast as possible. This kept the blank values at a low level. For control

of field blank values one passive sampler of each set of four samplers was left closed

during the sampling period and the value of it’s green color channel served as the field

blank.

Influence of the composition and volume of the color-forming reagent

Based on our previous experience [49-53] the Griess-Ilosvay reagent composition was

the same as that described under Reagents and Solutions for application of the

photometric standard procedure. It is similar to that used by Palmes et al. in their

pioneering paper [12] and in many subsequently published papers by different authors or

air monitoring authorities. To account for the fact that in the smartphone-based

evaluation procedure (as described below) the reagent solution is not (or only

insignificantly) diluted by the sample solution the combination reagent has been diluted

1+1 with water prior to use. Tests with a higher dilution (up to 5-fold) of the

combination reagent had no influence on the final color depth but the time for full color

development increased slightly. Higher reagent concentrations were not investigated

since this causes an unnecessary consumption of reagent. Therefore, the 1+1 dilution of

the combination reagent with water was used throughout all smartphone-based

evaluation measurements presented below.

To investigate the influence of the reagent volume added to the caps (acting as a mini-

vessel for the chemical reaction) on the RGB values the following experiment was

carried out using white opaque silicon caps. A volume of either 30 µL pure water or

30 µL of a 5 mg L-1 nitrite standard solution (corresponding to 0.15 µg NO2-) was added

to 5 caps each using a micropipette. After the solutions were evenly distributed on the

surface of the absorber pads, different volumes of the combination reagent in the range

30-1000 µL were added to the caps. Lower reagent volumes were not tested since they

were almost completely sucked off by the porous cellulose absorber pads. Volumes

higher than 1000 µL exceed the capacity of the caps. Repeated aspiration and draining

of the reagent solution with the micropipette ensured that the nitrite standard and the

color reagent were well mixed. After a reaction time of about 15-20 min, a photo of the

caps was taken with the smartphone indoors with light from the fluorescence tubes

mounted on the ceiling of the laboratory (see Figure 4). The values of the green color

channel retrieved of the colored adsorber pads for the different reagent volumes added

are also given in Figure 4.

Visual inspection of the six colored solutions shown in Figure 4 indicates that for 30, 50

and 100 µL reagent addition the color distribution across the surface of the pad is not

homogeneous. This suggests an uneven supply of color forming reagent resulting from

insufficient reagent distribution by capillary forces. The reflectance values for reagent

addition volumes in the range 250 to 1000 µL of the green color channel given beneath

the colored solutions shown in Figure 4 do not differ significantly. Considering that the

nitrite concentration is decreasing in proportion to increasing reagent volume this result

demonstrates that the lower concentration is compensated by the longer path length of

the reflected light within the sample solution (as is the case according to Lambert-Beer

law in absorbance measurements). In conclusion, it is apparent that the volume of

reagent solution added to the caps below 250 µL is unsuitable and volumes above

250 µL give similar reflectance values. Since higher volumes are regarded an

unnecessary consumption of reagent solution, 250 µL of reagent were added in all

further investigations.

Figure 4. Photo of caps with absorber pads to which solutions of different volumes (i.e., 30, 50,

100, 250, 500, and 1000 µL from left to right) of the combined color forming reagent have been

added. Prior to reagent addition the pads were spiked with 30 µl of a 5 mg L-1 nitrite standard

solution. The numbers beneath the caps are the respective reflectance values of the green color

channel.

Influence of using the zoom option of the smartphone and the distance and

orientation of the objects to be photographed in relation to the camera

When taking photos of the caps with the smartphone, the preserved image sections

depend on the selected zoom factor of the smartphone and the distance between the

smartphone and the caps. In the experiments performed we have changed the distance

between 10 and 50 cm (adjusted by means of a tripod keeping the smartphone in a

defined position perpendicular over the caps containing the colored solutions) and

adjusted the zoom factor so that all cups are within the image including a small part of

the supporting PVC block. In Figure 5 and Figure 6 images as obtained at various

distances are shown of five silicon cups with 15 mm inner height containing 250 µL of

colored solutions of identical composition. The absorber pads were spiked with 30 µL

of a 5 mg L-1 nitrite solution followed by the addition of 250 µL of combined reagent.

After a reaction time of about 15 min the photos were taken indoors with illumination

by fluorescence tubes mounted on the ceiling of the laboratory.

The mean reflectance values given in Figure 5 of the green color channel of the five

cups containing nitrite standard solutions obtained for different distances differ

significantly. This is expected due to by the different amount of reflected light collected

for different distances between the camera lens and the cups containing the colored

absorber pads. It must also be taken into consideration that the camera may cast a

shadow on the cups (partially depending on the distance) so that at given room light the

colored pads are illuminated differently. However, the standard deviation for the five

colored solutions is about the same for all distances. A closer look at the photographs

shown in Figure 5 evidences an obvious distortion of the images of the circular caps

positioned at the outer ends of the respective rows at distances of 10 and 20 cm; we

have termed this the dome effect. The dome effect is connected with the focal length of

the camera lens and is increasingly pronounced with decreasing distance between the

objects and the camera. This makes it more difficult to select the region of interest for

retrieval of RGB values. Another observation was that at low distances the shadow of

the smartphone camera and the photographer cast onto the colored solutions which

resulted in lower and less reproducible light reflection due to reduced luminosity.

Therefore, 50 cm distance was used in all further evaluations as a reasonable

compromise between practicability and absence of the dome effect. It is worth

mentioning that the distance of 50 cm was not always precisely kept since in many

cases the photos were taken freehand. However, the variation of the distance of 50 cm

was typically below ± 5 cm.

Figure 5. Photos of five caps taken at various distances between the camera and the objects.

The caps were equipped with absorber pads and spiked with 30 µl of 5 mg L-1 nitrite solutions

(corresponding to 0.15 µg NO2-) to which subsequently 250 µL of combined reagent solution

have been added. The photos were taken after about 15-20 min reaction time. The numbers on

the right side of the rows of caps are the respective mean reflection values of the green color

channel and the standard deviations.

Figure 6. Photo of five caps

arranged in a circular manner

taken at various distances bet-

ween the camera and the objects.

The caps were equipped with

absorber pads and spiked with 30

µL of 5 mg L-1 NO2- solutions

(corresponding to 0.15 µg NO2-)

to which subsequently 250 µL of

combined reagent solution has

been added. The photos were

taken after about 15-20 min

reaction time. The numbers on the

right side of the caps are the

respective mean reflection values

of the green color channel and the

standard deviations.

In an additional experiment the cups were placed on the supporting block in a circular

manner (see Figure 6). The distance between the cups and the camera was again varied

between 10 and 50 cm. The photos were also made indoors with illumination of the

fluorescence tubes mounted on the ceiling of the laboratory. Though the distortion of

the images of the caps at low distances was less pronounced compared with in-line

arrangement, 50 cm remained the favorable distance. The respective mean values of the

green color channel for the cups filled with identical solutions given in Figure 5 and

Figure 6 are reasonably reproducible in all cases (RSD ranging from 2.1-5.2 %).

Obviously, at 50 cm distance there is no preference for arrangement of the cups either

in-line or in a circular manner. In further measurements we have generally positioned

the cups in a line, or in several lines one above the other.

Influence of illumination

When we initially viewed the colored solutions in the caps at different illumination

conditions with bare eyes the color impression was different with respect to color depth

and, to a lesser extent, with respect to the hue. Depending on the orientation of the cups

relative to the incidence of the light a shadow was cast sometimes onto the solutions

inside the caps (see e.g., images of Figure 5 taken at 10 and 25 cm distance). This was

caused by the edge of the cups and could be prevented when the light source was

arranged perpendicularly in relation to the caps. Since this is not always easy to

accomplish, we have reduced the inner height of the silicon caps from 15 mm to 8 mm.

This is the minimum height to allow sufficient reagent volume to be added to the caps

(generally 250 µL was used) and still ensures that the attached caps sit firmly on the

polymethacrylate tubes of the passive samplers during transportation and exposure at

the sampling sites.

The importance of illumination conditions in smartphone-based evaluation of colored

solutions of solid surfaces has been emphasized again and again in several reviews and

original papers. Different solutions to this problem have been proposed [26,30,63-68].

One frequent way to avoid the changing impact of illumination when smartphone-based

imaging is applied is the protection of the objects from ambient light using a photo box

[26-31]. This has the obvious advantage that illumination conditions can be precisely

controlled and will not change in repetitive measurements performed for instance on

another day or another location. It may also offer the possibility to take images of

calibration solutions or colored pads independently of those of the samples, i.e., once a

calibration has been performed many samples can be subsequently analyzed. However,

picture taking in a photo box is impractical outside a laboratory in the field. Another

means to overcome the influence of variable ambient light is the use of the flashlight of

the smartphone [63]. The idea behind is to crossfade the ambient light so that the flash

becomes the dominant light source solely responsible for sample illumination. Our trials

with flashlight illumination were not very convincing [43]. When the photos taken were

inspected often a single small bright spot or several small bright spots were visible on

the colored liquid surface resulting from the refraction and specular reflection of the

flashlight (see Figure 7(b)). This makes the selection of an undisturbed area for color

retrieval difficult. Yet other approaches to overcome the problem of color and color

depth dependence on illumination conditions (i.e., brightness and color temperature of

the light source) is to apply illumination compensation algorithms [64,65], and/or

machine or deep learning classification [66-68]. Because of the complex calibration

strategy and sophisticated evaluation of data involved this way appeared not attractive

for our application. Another frequently employed strategy to compensate for the color

dependency on lighting conditions during the acquisition procedure of the colored

solutions or solid surfaces is to use color charts or references prepared in the same ways

as the samples [26,69-72]. In such cases it is common to take photos of the reference(s)

and the colored objects/samples simultaneously. This has often been done in placing the

objects in a photo box but photographing at ambient light has also been applied. Our

attempts to prepare a color chart for evaluation of the passives samples were so far not

crowed with success [43]. Therefore, we favored the use of reference standards for

calibration, and this was maintained throughout the current work.

To determine the influence of different ambient lighting conditions on the results of

evaluating RGB values of the colored solutions photos were taken (i) under room light

with fluorescence tubes mounted on the ceiling of the laboratory, (ii) indoors close to

the window without additional artificial light (i.e., the only light was from outdoors

shining through the windows), (iii) indoors without artificial light away from the

window in a shaded area of the room, (iv) with the flashlight of the smartphone camera,

and (v) with the caps placed in the photo box. As aforesaid, the use of a photo box and

smartphone flashlight was only applied for comparison. All photographs were made of

three sets of five cups each containing the same solutions arranged in a line (see Figure

7). The three sets arranged above each other in a line refer to blank solution, and two

standard solutions, respectively. The two standards solution were obtained by spiking

the absorber pads with 30 µL 5 mg L-1 (central line) and 20 mg L-1 (lower line) nitrite

solutions (corresponding to 0.15 and 0.6 µg NO2-) followed by the addition of 250 µL

of combined color-forming reagent solution. The distance between the cups and the

smartphone camera was kept constant at 50 cm (zoom factor ~ 4).

The reflectance values of the green color channel data obtained clearly show significant

differences under the various illumination conditions. A look at the reflectance values of

the blank solutions gave the highest value when the photo was taken close to the

window with no light on (the values were retrieved in the lighter part of the image of

Figure 7(d)). This is somewhat surprising since according to the visual impression the

luminous intensity was highest for illumination was flashlight. The reflectance values of

the green color channel for the standard solutions followed a slightly different order

compared with the blank solution. With respect to the application of smartphone-based

color evaluation the differences of reflectance values obtained at various light

conditions do not necessarily pose a serious problem. It is clear, however, that quan-

tification of samples at a given illumination requires the joint/parallel photo-graphing of

sample solutions together with suitable references. The shadows cast on about half the

area of the image shown in Figure 7(d) are unfavorable regarding the selection of an

appropriate region of interest and this kind of illumination was therefore not any longer

considered. In further evaluations illumination in a shadowed area with the fluorescence

light off was employed.

Calibration of smartphone-based evaluation under optimized experimental

conditions

For quantification of measurements calibration is indispensable. The standards used for

calibration should resemble the samples to be analyzed as much as possible. In passive

sampling of gases in most instances gas-phase calibration is not required [1,3,11,12]. It

is sufficient to quantify the amount of the compound collected (i.e., nitrite in the case of

Figure 7. Photos of blank and standard solutions taken under different illumination condition.

(a) photo box with LED illumination, (b) with the flashlight of the smartphone camera, (c) room

light with fluorescence tubes mounted on the ceiling of the laboratory, (d) close to the window

without additional artificial light (i.e., the only light was from outdoors shining through the

windows), and (e) indoors without artificial light away from the window in a shaded area of the

room. The three rows of five caps are all equipped with absorber pads. The first row is the blank

solution, the second and third row have been spiked with 0.15 and 0.6 µg NO2-, respectively.

The photos were taken after about 15-20 min reaction time. The numbers beside the rows of

caps are the respective mean reflectance values of the green color channel and standard

deviations.

sampling NO2 with TEA as absorber solution) and with the known uptake rate of the

passive sampler (obtained from sampler geometry and diffusion coefficient of the

analyte in air) and the exposure time the gas concentration can be calculated using

Fick’s law of diffusion (see supplementary material). Our procedure for calibration of

nitrite collected within the caps of the passive samplers pretreated with TEA was as

follows: Cellulose absorber pads were placed in seven of the shortened (cut to 8 mm

inner height) white opaque silicon caps and 20 µL TEA solution added. After about 10 s

for distribution of the TEA 30 µL of nitrite concentrations of variable concentration was

added. For initial evaluation of the calibration curve seven nitrite standard spikes in the

range 1-20 mg L-1 onto the pads (corresponding to a nitrite mass of 0.03-0.60 µg) and a

blank solution of 30 µL pure water were used. The seven caps were arranged in a line.

After another about 10 s exactly 250 µL of the combined color forming reagent solution

was pipetted into all caps. To ensure good contact between the nitrite spikes and the

color reagent the liquid volume was aspirated and dispensed 4-5 times in repetition with

the aid of a micropipette. After a reaction time of 15-30 min a photo was taken of all

caps at once. The smartphone camera was positioned perpen-dicular over the line of

caps at a distance of 50 cm between the smartphone and the objects. Photos were made

in the laboratory away from the window in a shaded area with no artificial light (no

flash, no fluorescence lights on).

Due to the expected exponential decay of diffuse reflectance with increasing

concentration (see above) the reflectance values of the green channel were plotted

against absolute nitrite mass (in µg), and an exponential function fitted to the data. In

Figure 8 the resulting plot and corresponding regression data are shown.

Figure 8. Calibration plot of smartphone-based evaluation of the azo-dye formed upon reaction

of nitrite with the Griess-Ilosvay reagent in the range 0.03-0.6 µg NO2-. See text for further

explanations.

Precision of five repetitive measurements over the entire concentration range was < 8 %

RSD. The LOD of the method (3 σ criterion) is 0.45 mg L-1 NO2- corresponding to an

absolute mass of 13.5 ng nitrite (values based on 250 µL reagent addition).

Considering that for real sample analysis calibration standards need to be prepared

whenever a set of samples is to be evaluated a lower number of standards would be

desirable to reduce the workload. Therefore, a further evaluation of the calibration data

was done reducing the number of standards for calculation of the curve from seven to

four (three intermediate standards of 0.075, 0.225 and 0.45 µg nitrite were left away).

The resulting regression data (y = 197.1 e-1.73x, R2 = 0.9956) are almost identical to that

of the 7-point calibration so that a 4-point calibration is regarded sufficient. Repetition

of the same calibration experiments under other illumination conditions (photo box and

indoors with artificial light from the fluorescence lamp mounted at the ceiling of the

lab) have shown, that reflectance values of the green color channel are (expectedly)

different and decay rates also differ slightly but the coefficients of determination were

better 0.998 in all cased[43]. It is worth mentioning that calibrations performed on other

days at apparently the same experimental conditions gave similar results, but slight

systematic deviations of the regression function were generally observed. However, the

calibration curves for standard solutions in the same concentration range could likewise

be well fitted by an exponential function with good coefficients of determination (R2 >

0.99) (see Figure 9).

Figure 9. Calibration plots obtained on different days for the same set of four standards at

apparently the same illumination conditions, (i.e., without artificial light away from the window

in a shaded area of the laboratory). The six curves marked in different colors (some overlap so

that they cannot be distinguished) are the results of calibrations performed on six randomly

selected days within the experimental time period between May and August 2023. See text for

further explanations.

Calibration of nitrite determination using the photometric standard

procedure

Our long-lasting experience with photometric evaluation of passive samplers has

resulted in an in-house standard procedure [49-53]. It has shown to be in good

agreement with the official standard gas-phase chemiluminescence method [73] for NO2

measurements of ambient air. Therefore, it was also employed in the present paper for

comparative reasons with the proposed smartphone-based evaluation in real sample

analysis.

The nitrite calibration using the “Standard Procedure” (see above) and photometric

detection was made in the concentration range 0.05-2 mg L-1. Regression analysis of the

linear function of absorbance versus nitrite concentration (in mg L-1) after reagent blank

corrections gave y = 0.517x + 0.0006 with a coefficient of determination of R2 =1.0000.

The slope and the abscissa of the calibration curve are consistent within 2% deviation

with results obtained in the past about 12 years of experience when the same analytical

procedure was applied [49-53] indicating high accuracy of the calibration. The relative

standard deviations of repetitive measurements (n=5) of the same standard solutions

were typically <3% over the entire working range. The limit of detection (LOD)

calculated (using the 3σ criterion) is 0.014 mg L-1 nitrite. Considering the extraction

procedure of the absorber pads of the passive samplers with 2 ml water this value

corresponds to a mass detection limit of 28 ng (not considering the contribution of blank

values associated with the sampling step; see below). For reasons of quality control,

before analysis of a set of passive samplers was started an initial test was made with a

single standard solution of 1 mg L-1 NO2- and a blank solution. For the standard and

blank solutions control values for absorbance of 0.520 ± 0.005 and max. 0.003, respec-

tively, were set. In case these values were not obtained all glassware and components of

the passive samplers were rinsed thoroughly with water again and the reagent solutions

were freshly prepared.

Results of NO2 measurements in ambient air

Prior to small measurement campaigns of passive sampling of NO2 at different sites in

Berlin an investigation of real sample analysis was made at the atmospheric test site of

our department. The intention was to determine the precision and LOD which can be

obtained for NO2 measurements. To this end three sets of shelters with four samplers

each were hung side by side on a truss. From six of the total twelve samplers the caps

were removed and NO2 was sampled for a period of 14 days. The six other samplers

were kept closed during the sampling period serving as field blanks. After sampling the

passive samplers were returned to the laboratory and evaluated using the smartphone-

based evaluation method under the optimized experimental conditions. The precision

(given as relative standard deviation) of each of the six passive samplers was <4.5 %.

Based on the standard deviation of the blank values and the nitrite calibration plot the

limit of detection (LOD) was calculated to be 0.85 µg nitrite. Transformation to

atmospheric concentrations for 14 days sampling using Fick's law of diffusion (see

ESM) the gas-phase LOD is about 5 µg m-3. It is worth noting that prolongation of the

sampling period which was expected providing lower LODs does not show a significant

decrease due to the increasingly important contribution of the blank (for 4 weeks

sampling the LOD was about 4 µg m-3).

Two small measurement campaigns of passive sampling of NO2 were conducted at ten

in-town sites of Berlin (Germany). Two in-town sites selected were at a street with

relatively high traffic volume (ca. 60000 per 24 h), the seven others in side-streets

within Berlin (in-town background). At each of the sites two shelters with four passive

samplers each were mounted. As before, one set was equipped with passive samplers

with yellow PE caps and the other set was equipped with samplers with white silicon

caps. Of the four samplers of one kind was kept closed during the sampling period and

results obtained served a field blank. After the allotted period of exposure, the samplers

were returned to the laboratory and nitrite determined using both smartphone-based

evaluation (of samplers with white silicon caps) and application of the photometric

standard procedure (of samplers with yellow caps). Figure 10 shows exemplarily repre-

sentative photos taken of the colored solutions obtained after addition of the color-

forming reagent to the white opaque passive sampler caps (Figure 10(a)) and of the

color developed when the standard evaluation procedure has been applied (Figure

10(b)). In Figure 10(a) the five cups in the upper row are the blank and four nitrite

standards and the second row are the field blank and three samplers from ambient air

measurements. In Figure 10(b) the nine solutions (from left to right) refer to blank, four

standards, the field blank and three samples from ambient air measurements.

Figure 10. (a) Photo of pads of standards and samples treated with combined reagent, (b) photo

of standards and extracts of samples prepared for photometric determination. The samples were

taken in parallel using the passive samplers described under “Experimental” for the two

different evaluation procedures. See text for further explanations.

Irrespective of the evaluation methods employed for determination of nitrite, the NO2

concentrations in ambient air were calculated using Fick’s law of diffusion (no gas-

phase calibration performed). This is typically done in passive sampling methodology

with tube-type samplers and has been approved in numerous publications [2,3,5,12]. A

brief outline of the calculation procedure of atmospheric NO2 concentrations from

nitrite determination in the collection caps is presented in the ESM. The results of

respective NO2 concentrations in the two campaigns of the two procedures calculated

are presented in Table1.

Table 1. Compilation of results from ambient air NO2-measurements in two campaigns

in the city of Berlin (Germany). Data obtained by application of the smartphone-based

evaluation and the in-house standard photometric method are juxtaposed. For

comparative reasons the mean values obtained by the Berlin governmental air pollution

network (BLUME) [74] at comparable site characteristics for the same period are also

displayed. For further explanations see text.

Sampling site NO2 concentration (µg m-3)

Smartphone Photometry BLUME

Campaign 1

(May 22nd – June 5th)

1 - Dauerwaldweg 1 9.3 ± 1.1 7.8 ± 0.1

2 - Helene-Jacobs-Str. 30 11.9 ± 2.8 9.9 ± 0.6

3 - Abram-Joffe-Str. 18 9.5 ± 0.6 8.9 ± 0.6

4 - Stefan-Zweig-Str. 7 9.7 ± 0.7 8.7 ± 0.2

5 - Alfred-Jung-Str. 14 9.8 ± 1.7 9.8 ± 0.7

6 - Otto-Suhr-Allee 62 11.5 ± 1.3 11.5 ± 0.3

Mean value 10.3 ± 1.1 9.4 ± 1.3 9.4 ± 1.5

Campaign 2

(July 13th – July 24th)

7 - Hauptstr. 18 32.6 ± 1.3 30.4

8 - Hauptstr. 59 35.8 ± 1.8 30.2

Mean value 34.2 ± 2.3 30.3 ± 0.1 21.6 ± 5.8

1 - Dauerwaldweg 1 11.0 ± 0.3 10.5

9 - Eisenzahn Str. 15 15.8 ± 0.8 13.1

10 - Münchener Str. 21 14.2 ± 0.3 11.6

Mean value 13.7 ± 2.4 11.7 ± 1.3 11.8 ± 2.8

Reasonable agreement is obvious between the established photometric procedure and

the smartphone-based evaluation developed for the in-town background stations (No. 1-

6, 9 and 10). However, distinct differences exist between our results (using the smart-

phone-based approach and photometric evaluation) and the BLUME data for strongly

traffic influenced sites (No. 7 and 8) [74]. The considerably higher NO2 concentrations

found in our measurements might be due to the specific characteristics of the sites

chosen by us. Further measurements are in progress (also at the same sites where

official measurements of BLUME are made) to clarify the discrepancies.

Conclusions and outlook

The proposed smartphone-based in-situ evaluation procedure for passive samplers for

NO2 has distinct advantages over common photometric measurements employed in the

majority of applications published so far. The addition of the color forming reagent

directly onto the absorber pad where nitrite has been formed (resulting from trapping

NO2 with TEA) is the only liquid handling step. This considerably simplifies the overall

analytical procedure and mitigates the risk of errors. The analytical specifications of the

proposed method with respect to LOD and working range are similar to the photometric

evaluation based on aqueous extraction of nitrite from the absorber pads followed by

derivatization reaction in a separate vessel. However, precision of the proposed new

method is slightly worse. Application of the two evaluation methods was made in a

preliminary measurement campaign in the city of Berlin evidencing reasonable

agreement. From an instrumental point of view, taking photos of the absorber pads with

a common smartphone (nowadays almost always on standby for other purposes) in

combination with freeware for retrieval of the RGB values is readily done. Expensive

photometric instruments for absorbance evaluation are not any longer required.

From the results of smartphone-based evaluation of passive samplers obtained in our

investigation some general conclusions can be drawn that apply likewise to applications

whenever colored liquid films or colored surfaces result from the analytical procedure.

Understandably, in reflectometric detection including smartphone photographing from

above the background color beneath the liquid and underlying the colored solid support

plays a role and a white reflecting surface is generally of advantage. When the liquid

film is present in a container, the container should also preferentially be white opaque.

The height of the liquid film is not a very important factor with respect to sensitivity.

However, a too thin film (i.e., too low liquid volume in the container) should be avoided

since even distribution of the liquid on the underlying surface is not always warranted;

when porous support materials are used a too low liquid volume might be completely

sucked off and color evaluation may become erratic. The distance between the colored

objects and the smartphone has significant influence on the signal since illuminance of

the objects varies. The zoom factor of the camera should be adjusted in a way that the

sections photographed of the set of objects remains about the same in repetitive

measurements. The dome effect observed in our application when the distance between

the camera and the objects is short might be of relevance also in other studies when

many objects are photographed at the same time which are positioned in a longer row,

i.e., the distance between the objects at the edges are far away from each other. The

most important factor of smartphone-based color evaluation is connected to the

illumination conditions. Huge differences of reflectance values for colored liquid films

of the same dye concentration occur attributable to variation of the luminosity of the

ambient light. The use of a photo box or the flashlight of the camera were omitted in the

present work since the former detracts from at-site (in-the-field) applications and the

latter has created isolated bright reflection spots on the surface of the colored solutions

which make the selection of a suitable region for color evaluation more difficult.

Whenever photos are taken without a photo box for proper quantification of results it is

mandatory to jointly photograph samples and a suitable reference. In this case

reflectance values might be (or are) different for different experimental conditions, but

this likewise happens to standard and sample solutions so that reliable results can still

be achieved.

The features of the entire method of passive sampling followed by in-situ smartphone-

based evaluation can be summarized as follows: (i) the tube-type passive samplers

employed in our work are an established configuration for sampling of many gases. The

individual parts of the passive samplers can be purchased at very low cost and the

assembly does not require any tools; (ii) the addition of a color forming-reagent to the

absorber surface of the passive samplers is the only liquid handling step required; (iii)

smartphone photographing of the colored liquid contained in the collection caps of the

passive samplers is easily done and evaluation of the RGB values of the colored spots

using freeware ImageJ® can be learned within a few hours.

Calculation of the analyte concentration in the liquid-phase (in the case of sampling

NO2 the analyte is nitrite) requires a preliminary calibration with suitable standards. In

the proposed method nitrite solutions have been employed which were treated in the

same way as the samples. However, irrespective of the calibration procedure applied it

is mandatory to take joint photos of standards and samples to account for the impact of

the distance between the smartphone camera and the colored objects and different

illumination conditions on the reflectance values. The retrieval of RGB values of the

color spots is then made from one image. However, the liquid-phase calibration requires

availability to a laboratory and experience with preparation of solutions. Currently (after

so far less successful attempts [43]) we are continuing exploring further possibilities

preparing color cards as the reference for quantification. First results are quite

promising so that the preparation of liquid standards might be avoided simplifying the

overall evaluation procedure. Another recently explored simplification of the color

development is to use a spray reagent which is a common way to visualize analytes in

thin-layer chromatography [75,76]. This in mind the way could be paved for true at-site

evaluation of passive samplers also accessible to unskilled persons. The application of

the developed method in citizen science initiatives [53,77,78] could provide to the

participants insights into color evaluation methodology and opportunities to monitor

their local air quality.

Credit authorship contribution statement

Chenglu Shi: Investigation. Xinyi He: Investigation. Kanokwan Kiwfo: Writing –

review & editing. Andreas Held: Review & editing. Wolfgang Frenzel:

Conceptualization, Supervision, Writing – original draft

Data availability statement

The data presented in this study are available from the corresponding author on

reasonable request.

Acknowledgements

Thanks are due to Max Zeidler supporting the construction of the photo box and

performing a lot of additional mechanical work. We also like to thank Sven Klemer for

continuous technical assistance in the experimental work. K. Kiwfo appreciates support

by a Georg Forster Research Fellowship of the Alexander von Humboldt Foundation

(Ref 3.5 - THA - 1227646 - GF-P).

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal

relationships that could have appeared to influence the work reported in this paper.

References

[1] J. Namiesnik, T. Gorecki, E. Kozlowski, L. Torres, M. Mathieu, Passive dosimeters

- An approach to atmospheric pollutants analysis, Sci. Total Environ. 38 (1988) 225-

258. DOI: 10.1016/0048-9697(84)90219-5

[2] R.H. Brown, The use of diffusive samplers for monitoring of ambient air (Technical

Report), Pure & Appl. Chem. 65 (1993) 1859-1874. DOI: 10.1351/pac199365081859

[3] R.H. Brown, Monitoring the ambient environment with diffusive samplers: Theory

and practical considerations, J. Environ. Monit. 2 (2000) 1-9. DOI: 10.1039/A906404D

[4] U.S. Environmental Protection Agency (EPA), Passive samplers for investigations

of air quality: Method description, implementation, and comparison to alternative

sampling methods. https://clu-in.org/download/issues/vi/VI-passive-samplers-600-R-

14-434.pdf (accessed 10 March 2024).

[5] J.J. Pienaar, J.P. Beukes, P.G. van Zyl, C.M.B. Lehmann, J. Aherne, Chapter 2 –

Passive diffusion sampling devices for monitoring ambient air concentrations, In:

Monitoring of Air Pollutants - Sampling, Sample Preparation and Analytical

Techniques, Forbes, P.B.C. (Ed.), Vol. 70, Wilson & Wilson’s Comprehensive

Analytical Chemistry, Elsevier, Amsterdam, 2015, pp. 13-52.

[6] World Meteorological Organization - Global Atmospheric Watch (GAW), Report on

passive samplers for atmospheric chemistry measurements and their role in GAW,

Collection(s) and Series: WMO/TD- No. 829; GAW Report- No. 122.

https://library.wmo.int/doc_num.php?explnum_id=9671 (accessed 10 March 2024).

[7] D. Buzica, M. Gerboles, L. Amantini, Laboratory and field inter-comparison of NO2

diffusive samplers, European Commission; Institute for Environment and Sustainability,

Emissions and Health Unit.

https://publications.jrc.ec.europa.eu/repository/handle/JRC26285 (accessed 10 March

2024).

[8] T. Hafkenscheid, A. Fromage-Mariette, E. Goelen, M. Hangartner, U. Pfeffer, H.

Plaisance, F. de Santis, K. Saunders, W. Swaans, Y.S. Tang, J. Targa, C. van Hoek, M.

Gerboles, Review of the application of diffusive samplers in the European Union for the

monitoring of nitrogen dioxide in ambient air; European Commission Joint Research

Centre Institute for Environment and Sustainability, Luxembourg, 2009. ISBN 978-92-

79-12052-7

[9] C. Varshney, A.P. Singh, Passive samplers for NOx monitoring: A critical review,

Environmentalist 23 (2003) 127-136. DOI: 10.1023/A:1024883620408

[10] K. Stevenson, T. Bush, D. Mooney, Five years of nitrogen dioxide measurements

with diffusion tube samplers at over 1000 sites in the UK, Atmos. Environ. 35 (2001)

281-287. DOI: 10.1016/S1352-2310(00)00171-0

[11] European Standards, CSN EN 16339, Ambient air - Method for the determination

of the concentration of nitrogen dioxide by diffusive sampling,

https://www.en-standard.eu/csn-en-16339-ambient-air-method-for-the-determination-

of-the-concentration-of-nitrogen-dioxide-by-diffusive-sampling/ (accessed 10 March

2024).

[12] E.D. Palmes, A.F. Gunnison, J. Dimattio, C. Tomczyk, Personal sampler for

nitrogen dioxide, Am. Ind. Hyg. Assoc. J. 37 (1976) 570-577. DOI:

10.1080/0002889768507522

[13] C. Kirby, M. Fox, J. Waterhouse, Reliability of nitrogen dioxide passive diffusion

tubes for ambient measurements: in-situ properties of the triethanolamine absorbent, J.

Environ. Mon. 2 (2000) 307-312. DOI: 10.1039/B003124K

[14] B.C. Cardoff, J. Hodgeson, Passive sampler for ambient levels of nitrogen dioxide,

Anal. Chem. 55 (1983) 2083-2085. DOI: 10.1021/ac00263a017

[15] E.W. Nery, L.T. Kubota, L.T., Sensing approaches on paper-based devices: A

review, Anal. Bioanal. Chem. 405 (2013) 7573-7595. DOI: 10.1007/s00216-013-6911-4

[16] T. Akyazi, L. Basabe-Desmonts, F. Benito-Lopez, Review on microfluidic paper-

based analytical devices towards commercialisation, Anal. Chim. Acta 1001 (2017) 1-

17. DOI: 10.1016/j.aca.2017.11.010

[17] H.A. Silva-Neto, L.R. Sousa, W.K.T. Coltro, Chapter 4 - Colorimetric paper-based

analytical devices. In: Paper-based analytical devices for chemical analysis and

diagnostics, W.R. de Araujo; T.L.R.C. Paixão (Eds.), Elsevier, 2022, 59-79. ISBN

9780128205341. DOI: 10.1016/B978-0-12-820534-1.00009-8

[18] G.G. Morbioli, T. Mazzu-Nascimento, A.M. Stockton, E. Carrilho, Technical

aspects and challenges of colorimetric detection with microfluidic paper based

analytical devices (µPADs) - A review, Anal. Chim. Acta 970 (2017) 1-22. DOI:

10.1016/j.aca.2017.03.037

[19] J. Yang, K. Wang, H. Xu, W. Yan, Q, Jin, D. Cui, Detection platforms for point-of-

care testing based on colorimetric, luminescent and magnetic assay: A review, Talanta

202 (2019) 96-110. DOI: 10.1016/j.talanta.2019.04.054

[20] A.E. Urusov, A.V. Zherdev, B.B. Dzantiev, Towards lateral flow quantitative

assays: Detection approaches, Biosensors 9 (2019) 89. DOI: 10.3390/bios9030089

[21] Ocean Optics, Reflectance Measurement. Ocean Insight. www.oceaninsight.com.

https://www.oceaninsight.com/knowledge-hub/measurement-techniques/reflectance/

(accessed 10 March 2024).

[22] R. Paschotta, Reflectometers, rp-photonics encyclopedia. https://www.rp-

photonics.com/reflectometers.html (accessed 10 March 2024).

[23] S. Ebel, P. Post, Theory of reflectance measurements in quantitative evaluation in

TLC and HPTLC, J. Sep. Sci. 4 (1981) 337-342. DOI: d10.1002/jhrc.1240040706

[24] C. Cimpoiu, S. Hodisan,S. Quantitative thin-layer chromatography analysis by

photodensiometry, Rev. Anal. Chem. 21 (2002) 55-75. DOI:

10.1515/REVAC.2002.21.1.55

[25] P. Bernard-Savary, C.F. Poole, C.F. Instrument platforms for thin-layer

chromatography, J. Chrom. 1421 (2015) 184-202. DOI: 10.1016/j.chroma.2015.08.002

[26] L. Capitán-Vallvey, N. López-Ruiz, A. Martínez-Olmos, M.M. Erenas, A.J. Palma,

Recent developments in computer vision-based analytical chemistry: A tutorial review,

Anal. Chim. Acta 23 (2015) 23-56. DOI: 10.1016/j.aca.2015.10.009

[27] M. Rezazadeh, S. Seidi, M. Lid, S. Pedersen-Bjergaard, Y. Yamini, Y. The modern

role of smartphones in analytical chemistry, TrAC Trends Anal. Chem. 118 (2019) 548-

555. DOI: 10.1016/j.trac.2019.06.019

[28] G.M. Fernandes, W.R. Silva, D.N. Barreto, R.S. Lamarca, P.C.F.L. Gomes, J.F. da

Silveira Petruci, A.D. Batista, Novel approaches for colorimetric measurements in

analytical chemistry - A review, Anal. Chim. Acta 1135 (2020) 187-203. DOI:

10.1016/j.aca.2020.07.030

[29] Smartphone-based detecion devices, Ch. Hussain, C.M. Ed., Emerging Trends in

Analytical Chemistry, Elsevier, 2021, https://www.elsevier.com/books/smartphone-

based-detection-devices/hussain/978-0-12-823696-3 (accessed 10 March 2024).

[30] Y.Fan, J. Li, Y. Guo, L. Xie, G. Zhang, Digital image colorimetry on smartphone

for chemical analysis: A review. Measurement 171 (2021) 108829. DOI:

10.1016/j.measurement.2020.108829

[31] M. Pohanka, Small camera as a handheld colorimetric tool in the analytical

chemistry, Chem. papers 71 (2017) 1553-1561. DOI: 10.1007/s11696-017-0166-z

[32] H. Weik, Reflection density, In: Computer Science and Communications

Dictionary. Springer, Boston, MA, 2017.

[33] Reflection density – ISO definition, Photography and graphic technology - Density

measurements - Part 4: Geometric conditions for reflection density, ISO 5-4:2009(en).

https://www.iso.org/obp/ui/#iso:std:iso:5:-4:ed-3:v1:en (accessed 10 March 2024).

[34] R.W. Frei, Diffuse reflectance spectroscopy; Applications, Standards, and

Calibration (With Special Reference to Chromatography), J. Res. Natl. Bur. Stand.

Phys. Chem. 80 (1976) 551-565. DOI: 10.6028/jres.080a.055

[35] M.L. Firdaus, W. Alwi, F. Trinoveldi, I. Rahayu, L. Rahmidar, K. Warsito,

Determination of chromium and iron using digital image-based colorimetry, Procedia

Environ. Sci. 20 (2014) 298-304. DOI: 10.1016/j.proenv.2014.03.037

[36] Z. Geng, Y. Miao, G. Zhang, X. Liang, Colorimetric biosensor based on

smartphone: State-of-art, Sens. Actuators 349 (2023) 114056. DOI:

10.1016/j.sna.2022.114056

[37] J.F. Bergua, R. Álvarez-Diduk, A. Idili, C. Parolo, M. Maymó, L. Hu, A. Merkoçi,

Low-cost, user-friendly, all-integrated smartphone-based microplate reader for optical-

based biological and chemical analyses, Anal. Chem. 94 (2022) 1271-1285. DOI:

10.1021/acs.analchem.1c04491

[38] K. Kiwfo, K. Grudpan, A. Held, W. Frenzel, Smartphone-based color evaluation of

passive samplers for gases: A review, Atmosphere 15 (2024) 451. DOI:

10.3390/atmos15040451

[39] M. Cerrato-Alvarez, S. Frutos-Puerto, C. Miró-Rodríguez, E. Pinilla-Gil, Measure-

ment of tropospheric ozone by digital image analysis of indigotrisulfonate-impregnated

passive sampling pads using a smartphone camera, Microchem. J. 154 (2020) 104535.

DOI: 10.1016/j.microc.2019.1045

[40] P.A. de França Souza, J.L.A. Neto, A.A. Cardoso, A simple technique based on

digital images for determination of nitrogen dioxide in ambient air, Water, Air, Soil

Pollut. 232 (2021) 72. DOI: 10.1007/s11270-021-05031-4

[41] S. Adhikari, S.C. Conrad, A. Held, W. Frenzel, Simplified procedures for

evaluation of passive samplers for nitrogen dioxide, Talanta Open 5 (2022) 100096.

DOI: 10.1016/j.talo.2022.100096

[42] C. Shi, Smartphone-basierte und kolorimetrische Auswertung von Passivsammlern

für die Bestimmung von Stickstoffdioxid, Bachelor Thesis, Technische Universität

Berlin, Germany, 2023.

[43] X. He, Beiträge zur Smartphone-basierten Auswertung von Passivsammlern für

NO2 und Anwendung der Methode zur großflächigen Ermittlung der NO2-Belastung in

Berlin, Master Thesis, Technische Universität Berlin, Germany, 2023.

[44] U. Latza, S. Gerdes, X. Baur, Effects of nitrogen dioxide on human health:

Systematic review of experimental and epidemiological studies conducted between

2002 and 2006, Int. J. Hyg. Environ. Health 212 (2009) 271-287. DOI:

10.1016/j.ijheh.2008.06.003

[45] S. Depayras, T. Kondakova, H.J. Heipieper, M.G.J. Feuilloley, N. Orange, C.

Duclairoir-Poc, The hidden face of nitrogen oxides species: From toxic effects to

potential cure? Emerging Pollutants - Some Strategies for the Quality Preservation of

Our Environment, InTech, Jul. 2018, 04. DOI: 10.5772/intechopen.75822

[46] R.W. Atkinson, B.K. Butland, H.R. Anderson, R.L. Maynard, Long-term

concentrations of nitrogen dioxide and mortality - A meta-analysis of cohort studies,

Epidemiology 29 (2018) 460–472. DOI: 10.1097/EDE.0000000000000847

[47] P.J. Crutzen, The role of NO and NO2 in the chemistry of the troposphere and

stratosphere, Ann. Rev. Earth Planet. Sci. 7 (1979) 443-472. DOI:

10.1146/annurev.ea.07.050179.002303

[48] United States Environmental Protection Agency, Nitrogen dioxide (NO2) pollution,

https://www.epa.gov/no2-pollution (accessed 10 March 2024).

[49] A. Hapka, Beiträge zur Passiven Probenahme von Stickstoffdioxid für

Immissionsmessungen, Diploma Thesis, Technische Universität Berlin, Germany, 2003.

[50] S. Eckardt, Untersuchungen zur Passiven Probenahme für Immissionsmessungen,

Master Thesis, Technische Universität Berlin, Germany, 2014.

[51] G. Tumurbaatar, Vergleichende Untersuchungen zur NO2-Bestimmung in

Außenluft und in Innenräumen mittels Passiver Probenahme, Master Thesis, Technische

Universität Berlin, Germany, 2016.

[52] A. Wuttke, Optimierung, Vereinfachung und Miniaturisierung photometrischer und

kolorimetrischer Auswertemethoden von Passivsammlern zur Bestimmung von

Stickstoffdioxid, Bachelor Thesis, Technische Universität Berlin, Germany, 2019.

[53] A. Höhne, R.A.A. Schulte, M. Kulicke, T.-T. Huynh, M. Telgmann, W. Frenzel, A.

Held, Assessing the spatial distribution of NO2 and influencing factors in urban areas -

passive sampling in a citizen science project in Berlin, Atmosphere 14 (2023) 360. DOI:

10.3390/atmos14020360

[54] V.M. Ivanov, The 125th anniversary of the Griess reagent, J. Anal. Chem. 59

(2004) 1002-1005. DOI: 10.1023/B:JANC.0000043920.77446.d7

[55] L. Varadi, M. Breedon, F.F. Chen, A. Trinchi, I.S. Cole, G. Wei, Evaluation of

novel Griess-reagent candidates for nitrite sensing in aqueous media identified via

molecular fingerprint searching, RCS Adv. 9 (2019) 3994. DOI: 10.1039/c8ra07656a

[56] Wikipedia, 2022, RGB color spaces,

https://en.wikipedia.org/wiki/RGB_color_spaces (accessed 10 March 2024).

[57] ImageJ - Imaging Processing and Analysis in Jawa. https://imagej.nih.gov/ij/

(accessed 10 March 2024).

[58] B.E. Saltzman, Colorimetric microdetermination of nitrogen dioxide in the

atmosphere, Anal. Chem. 26 (1954) 1949-1955. DOI: 10.1021/ac60096a025

[59] International Organization for Standardization, Ambient air - Determination of

mass concentration of nitrogen dioxide - Modified Griess-Saltzman method, ISO

6768:1998. https://www.iso.org/standard/13262.html (accessed 10 March 2024).

[60] J. Targa, A. Loader, Diffusion tubes for ambient NO2 monitoring: Practical

guidance for laboratories and users, AEA Energy & Environment 2008

https://laqm.defra.gov.uk/documents/0802141004_NO2_WG_PracticalGuidance_Issue

1a.pdf (accessed 10 March 2024).

[61] Kunststoff-Schutzlösungen für alle Fälle von Pöppelmann KAPSTO®,

Pöppelmann, Germany. https://www.poeppelmann.com/de/unternehmen (accessed 8

March 2024).

[62] HKIT-fastener, German distributor. Silicon Gummi Kappe Weiß Rund,

https://www.ebay.de/str/hkitfastener (accessed 8 March 2024).

[63] T. Kong, J.B. You, B. Zhang, B. Nguyen, F. Tarlan, K. Jarvi, D. Sinton,

Accessory-free quantitative smartphone imaging of colorimetric paper-based assays,

Lab Chip 11 (2019) 1991-1999. DOI: 10.1039/C9LC00165D

[64] M.E. Ferreira, J. Tirapu-Azpiroz, D.V.L.M. Marçal, A.F. da Silva, R.L. Ohta, M.B.

Steiner, Illumination compensation algorithm for colorimetric detection of microfluidic

paper-based devices with a smartphone, Proc. SPIE 11651, Optical Diagnostics and

Sensing XXI: Toward Point-of-Care Diagnostics, 2021. DOI: 10.1117/12.2578441

[65] L. Shen, J.A. Hagen, I. Papautsky, Point-of-care colorimetric detection with a

smartphone, Lab. Chip 12 (2012) 4240-4243. DOI: 10.1039/C2LC40741H

[66] S. Duan, T. Cai, J. Zhu, X. Yang, E.G. Lim, K. Huang, K. Hoettges, Q. Zhang, H.

Fu, Q. Guo, X. Liu, Z. Yang, P. Song, Deep learning-assisted ultra-accurate smartphone

testing of paper-based colorimetric ELISA assays, Anal. Chim. Acta 1248 (2023)

340868. DOI: 10.1016/j.aca.2023.340868

[67] R. Thakur, P. Maheshwari, S.K. Datta, S.K. Dubey, Smartphone-based, automated

detection of urine albumin using deep learning approach, Measurement 194 (2022)

110948. DOI: 10.1016/j.measurement.2022.110948

[68] V. Duan, T. Isık, V. Kılıç, N. Horzum, A field-deployable water quality monitoring

with machine learning-based smartphone colorimetry, Anal. Methods 14 (2022) 3458.

DOI: 10.1039/D2AY00785A

[69] E. Kehoe, R.L. Penn, Introducing colorimetric analysis with camera phones and

digital cameras: An activity for high school or general chemistry, J. Chem. Edu. 90

(2013) 1191-1195. DOI: 10.1021/ed300567p

[70] M.L. Firdaus, A. Aprian, N. Meileza, M. Hitsmi, R. Elvia, L. Rahmidar, R.

Khaydarov, Smartphone coupled with paper-based colorimetric device for sensitive and

portable mercury ion sensing. Chemosensors 7 (2019) 25. DOI:

10.3390/chemosensors7020025

[71] K. Kesonkan, C. Yeerum, K. Kiwfo, K. Grudpan, M. Vongboot, Green

downscaling of solvent extractive determination employing coconut oil as natural

solvent with smartphone colorimetric detection: Demonstrating the concept via Cu(II)

assay using 1,5-Diphenylcarbazide, Molecules 27 (2022) 8622. DOI:

10.3390/molecules27238622

[72] S. Wirojsaengthong, W. Aeungmaitrepirom, F. Unob, S. Fuangswasdi, P.

Varanusupakul, K. Mueangdech, T. Treetos, P. Puthongkham, Portable quantitative

chemical analysis: Digital image colorimetric detection of Fe(III) with curcumin paper,

J. Chem. Educ. 100 (2023) 3604. DOI: 10.1021/acs.jchemed.3c00483

[73] CEN-EN 14211:2005, Ambient air quality - Standard method for the measurement

of the concentration of nitrogen dioxide and nitrogen monoxide by chemiluminescence.

https://standards.iteh.ai/catalog/standards/cen/181f3d8d-a64c-46fb-b90a-

7affd10c7c0b/en-14211-2005 (accessed 10 March 2024).

[74] Senatsverwaltung für Umwelt, Verkehr und Klimaschutz, Berliner Luftgüte-

messnetz - BLUME.

https://luftdaten.berlin.de/pollution/no2?stationgroup=traffic&period=1m&timespan=cu

stom&sta-rt%5Bdate%5D=01.08.2020&end%5Bdate%5D=30.09.2020 (accessed 10

March, 2024).

[75] D. Waldi, Spray reagents for thin-layer chromatography. In: Thin-Layer Chromato-

graphy, E. Stahl (Ed.), Springer, Berlin, Heidelberg, 1965.

[76] H. Jork, W. Funk, W. Fischer, H. Wimmer, Thin-layer chromatography: Reagents

and detection methods, Vol.1, Part 1, VCH Publishers, California, 1990. ISBN

0895738767

[77] S. De Craemer, J. Vercauteren, F. Fierens, W. Lefebvre, F.J.R. Meysman, Using

large-scale NO2 data from citizen science for air-quality compliance and policy support,

Environ. Sci. Technol. 54 (2020) 11070. DOI: 10.1021/acs.est.0c02436

[78] J. Perelló, A. Cigarini, J. Vicens, I. Bonhoure, D. Rojas-Rueda, M.J.

Nieuwenhuijsen, M. Cirach, C. Daher, J. Targa, A. Ripoll,Large-scale citizen science

provides high-resolution nitrogen dioxide values and health impact while enhancing

community knowledge and collective action, Sci. Total Environ. 789 (2021) 147750.

DOI: 10.1016/j.scitotenv.2021.147750

Supplementary material to MS

Optimization of smartphone-based evaluation of tube-type passive samplers using

atmospheric nitrogen dioxide determination as an example

Image analysis employing the RGB color space [S1] has been successfully used for

evaluation of colored reaction products of chemical assays of a huge number of analytes

[S2-S7]. In the context of the present MS on passive sampling of NO2, smartphone-

based color detection has been applied to the determination of nitrite being the reaction

product of gaseous NO2 and triethanolamine [S8-S11]. The sequence of steps from

photographing of the colored reaction product to the final retrieval and quantification of

the green channel of the RGB values is described in the first part (A) of this ESM. In

part (B) the color analysis procedure and the way of presentation of results are outlined.

Part (C) explains the calculation of gas-phase NO2 based on Fick’s law of diffusion

from quantification of the collected mass of nitrite.

In part (D) specific results are presented as obtained of some of the investigations

performed during the optimization, the optimization of the calibration procedure and in

real sample analysis

(A) Sequence of image analysis

Photos were taken using Apple iPhones (model 12mini and model XR) at various

illumination conditions indoors in a research laboratory. Care was taken to place the

caps of the passive samplers to be photographed so that no (or only minimal) shadow

was cast from the edge of the caps onto the adsorber pads.

The distance of the smartphone camera to the colored objects was - after optimization of

this parameter - kept at 50 cm. The zoom option of the smartphone (zoom factor ~ 4.0)

was used to select an image section that was only slightly larger than the total area of

the PCV block where the caps of passive samplers (containing the colored solutions)

were placed. The photos acquired by the smartphone were stored in JPEG format.

Although direct RGB reading can be done using the smartphone, the open-source

platform ImageJ® was applied for processing and analyzing the photos stored in JPEG

format. A detailed description of ImageJ and the respective features of this programme

can be found on the internet [S12]. The ImageJ® software was installed on a standard

laptop. After uploading a photo, the regions of interest (ROIs) were selected by visual

inspection. Since the color distribution across an individual absorber pad is highly

homogeneous, a circular area covering about 90% of the central part of the pads was

selected. This is exemplarily shown in Figure S1 for five caps containing colored azo

dye solutions of different concentration.

Figure S1: The inner dashed circles show the selection of the region of interest used for

evaluation of the reflectance values of the green color channel. See text for further explanation.

Although the reflectance values of all three color channels are recorded by ImageJ®

only the green channel was used for evaluation. Green is the complementary color to the

red-violet color of the azo-dye product of the Griess-Ilosvay derivatization reaction and

evaluation of the green channel provides the highest difference of reflectometric

response between the dye solutions and a blank.

For the purpose of parameter optimization, calibration, and real sample analysis the

respective solutions contained in the sampling cap of the passive samplers have always

been photographed together. This avoids deviations and misinterpretation of results due

to different light conditions and variations of the distance between the smartphone

camera and the objects (details are outlines in the body of the MS). RGB values of the

colored spots and the zero-reference (blank) solutions are retrieved from the “group

photos” and the data exported to MS-Excel for statistical evaluation (mean values,

standard deviations, regression analysis).

(B) Color analysis procedure and presentation of results

The color analysis is performed by freeware software ImageJ® in terms of RGB values

of 8-bit resolution. Accordingly, the 256-bit color scale is from 255,255,255 for white,

and 0,0,0 for black [S1,S3,S7,S13]. The relevant maximum and minimum values of

reflection in the concrete situation of evaluation of passives sampler are the remission of

the pure absorber pads within the sampler caps and a totally color-saturated pad

resulting from the nitrite assay of a very high concentration. These are so to say the

cornerstones of the scale.

The values of the cornerstones, however, change depending on illumination conditions

and the distance between the camera and the pads. In addition, differences have also

been reported in several publications when different cell phones have been used under

otherwise identical conditions [S5,S7,S14]. In our work we present all data in the form

of RGB values as obtained from the evaluation procedure of the images, i.e., no

attempts were made to use e.g., a truly white surface or the blank values for referencing

the color values. However, we have refrained from that since this allows to better

compare the results of the optimization studies, in particular the mentioned influence of

illumination and distance between camera and objects. The use of the two different

smartphones for photographing applied in our work gave identical results.

The fundamental process of passive sampling of gases is the diffusion-controlled

trapping of the respective gas on a suitable solid or liquid surface. Assuming that the

concentration gradient within the tube of a Palmes-type sampler is linearly decreasing

from the opening exposed to the atmosphere to the surface of the absorber, the mass

transport of a gas to a suitable sorbent can be calculated using Fick’s 1st law of diffusion

[S8,S15-S18].

J = dm/(A dt) = - D dc/dl Eq.1

J diffusive mass flux

m mass of diffusing analyte

t time

D Diffusion coefficient of the respective analyte gas in air

A area through which analytes pass by diffusive transport

dc concentration gradient of target analyte along the sampler tube

dl length of the diffusion path

Instead of the commonly used flux of quantity in mol per time, it is exchanged for mass

flux (e.g., in µg s-1) here. This is done because the typical units for gas-phase concen-

trations are µg m-3 rather than mol/m3. Assuming further that the absorber solution

(triethanolamine in the present case) is an ideal sink for NO2 (i.e., the NO2 concen-

tration is zero at the surface of the absorbing material) Eq.1 can be rewritten in the

integrated form for NO2 as the example gas:

J = m/t = (-D A cNO2) / l Eq.2

with cNO2 being the atmospheric NO2 concentration outside of the sampler. Due to the

1:1 stoichiometric relation between NO2 and NO2-, the mass of NO2 diffusing to the

absorber surface can be substituted for the mass of nitrite collected. Rearrangement of

Eq.2 then leads to the final formula for calculation of the average gas-phase NO2

concentration in ambient air during the sampling duration for a tube-type passive

sampler of given dimensions:

210

tDA

−

[µg m-3] Eq.3

cNO2 atmospheric analyte gas concentration (µg m-3)

mNO2- mass of nitrite collected (µg)

l length of the sampler tube (cm)

A cross sectional area of the sampling tube (cm2)

t sampling duration (s)

D diffusion coefficient of NO2 in air (cm2 s-1)

The factor 106 results from the conversion of cm3 to m3.

The quotient (A D)/l is defined as the uptake rate (UR) of a passive sampler [S10,S16-

S18] and has the dimension volume per time (e.g., ml min-1). For a particular passive

sampler geometry and a specific gas, the UR is constant at a given temperature. Hence,

Eq.3 can be simplified to:

210

tUR

cNO

−

[µg m-3] Eq.4

Using the length (7.0 cm) and inner diameter (1.0 cm) of the passive sampler tubes

employed in the present paper and a diffusion coefficient of NO2 of 0.154 cm2 s-1

(accepted value for 20°C [S9-S12, S19] the uptake rate for NO2 amounts to 1.04 mL

min-1 (or 0.0173 cm3 s-1). This value was used in the present paper for calculation of

atmospheric NO2 concentrations; corrections for temperature and pressure were not

applied.

References - Supplementary material

S1 RGB color model, Wikipedia, the free encyclopedia.

https://en.wikipedia.org/wiki/RGB_color_model, accessed August 2023

S2 Smartphone-based detection devices: Emerging trends in analytical techniques,

C.M. Hussain (Ed.), Elsevier, 2021. ISBN: 9780128236963

S3 M. Pohanka, Chem. Pap. 71, 1553 (2017). doi.org/10.1007/s11696-017-0166-z

S4 O.V. Monogarova, K.V. Oskolok and V.V. Apyari, J. Anal. Chem. 73, 1076 (2018).

doi.org/10.1134/S1061934818110060

S5 V. Kiliç, N. Horzum and M.E. Solmaz, From sophisticated analysis to colorimetric

determination: Smartphone spectrometers and colorimetry, In: Color detection, L.-W.

Zeng, S.-L, Cao (Eds.) Intech Open, 2020. doi.org/10.5772/intechopen.82227

S6 G.M. Fernandes, W.R. Silva, D.N. Barreto, R.S. Lamarca, P.C.F. Lima Gomes, J.F.

da Silveira Petruci and A.D. Batista, Anal. Chim. Acta 1135, 187 (2020).

doi.org/10.1016/j.aca.2020.07.030.

S7 Y. Fan, J. Li, Y. Guo, L. Xie and G. Zhang, Measurement 171, 108829 (2021).

doi.org/10.1016/j.measurement.2020.108829

S8 E.D. Palmes, A.F. Gunnison, J. Di Mattio and C. Tomczyk, Am. Ind. Hyg. Ass. J.

37, 570 (1976). doi.10.1080/0002889768507522

S9 D. Buzica, M. Gerboles and L. Amantini, Laboratory and field inter-comparison of

NO2 diffusive samplers, European Commission; Institute for Environment and

Sustainability, Emissions and Health Unit, 2003.

https://pubs.rsc.org/en/content/articlehtml/2006/em/b511271k

S10 J. Targa and A. Loader, Diffusion tubes for ambient NO2 monitoring: Practical

guidance for laboratories and users, AEA Energy and Environment, EA/ENV/R/2504 –

Issue 1a, 2008.

https://laqm.defra.gov.uk/documents/0802141004_NO2_WG_PracticalGuidance_Issue

1a.pdf

S11 European Standards, CSN EN 16339, Ambient air - Method for the determination

of the concentration of nitrogen dioxide by diffusive sampling, 2013.

https://www.en-standard.eu/csn-en-16339-ambient-air-method-for-the-determination-

of-the-concentration-of-nitrogen-dioxide-by-diffusive-sampling/ (accessed 23 August

2023)

S12 ImageJ - Imaging Processing and Analysis in Jawa.

https://imagej.nih.gov/ij/ (accessed, 23 August 2023)

S13 G. Bakker, Color Coding and the RGB Color Space, 2014

https://geraldbakker.nl/psnumbers/rgb-explained.html

S14 M. Nixon, F. Outlaw and T.S. Leung, Accurate device-independent colorimetric

measurements using smartphones. PLoS ONE 15(3): e0230561.

doi.org/10.1371/journal.pone.0230561

S15 A. Fick, Annalen der Physik (in German), 94, 59 (1855.

doi:10.1002/andp.18551700105

S16 H. Brown, Pure & Appl. Chem. 65, 1859 (1993). doi:10.1351/pac199365081859

S17 M. Harper and C.J. Purnell, Am. Ind. Hyg. Assoc. J. 48, 214 (1987).

doi.org/10.1080/15298668791384652

S18 T. Hafkenscheid, A. Fromage-Mariette, E. Goelen, M. Hangartner, U. Pfeffer, H.

Plaisance, F. de Santis, K. Saunders, W. Swaans, Y.S. Tang, J. Targa, C. van Hoek and

M. Gerboles, Review of the application of diffusive samplers in the European Union for

the monitoring of nitrogen dioxide in ambient air, European Commission Joint Research

Centre Institute for Environment and Sustainability, 2009. ISBN 978-92-79-12052-7

S19 W.J. Massman, Atmos. Environ. 32, 1111 (1998). doi.org/10.1016/S1352-

2310(97)00391-9