For transmittance measurements of pH dependent color changing effects of PhR and BTB, the light is directly passing through the sample. Therefore, a light source is placed underneath the sample and the light is collected by the RGB sensor patch. The idea was initially visualized in a CAD sketch as shown in Fig. , and later manufactured by rapid prototyping via micro-milling and 3-D printing. This approach leads to a very cost-effective measurement system, which in general depends only on the RGB sensor and a control board, e.g., an Arduino Uno board. The expenses for both of these parts is in total EUR 30. Further costs only arise depending on the framework used around these two parts. For this prototype, the overall costs were less than EUR 50.

The manufactured prototype is shown in Fig. . Within the figure a simple sketch is included, which illustrates the working principle and the adaption of this prototype in comparison to the common optode concepts shown in Fig. . As mentioned before, optodes usually contain modified fiber optics or membrane technologies and monochromatic filters. Within this prototype none of these techniques are required. Instead of changing the geometrical dimension of the micro-bioreactors to implement membranes or fiber optics, the pH indicators are already included in the nutrition media and thus can be utilized to measure the pH. Secondly, no additional monochromatic filter is necessary since the RGB unit has chromatic filters on top of the respective photodiodes.

The LED originally mounted close to the top of the RGB detection unit (Fig. a) was desoldered, incorporated underneath the sample holder as indicated in Fig. c and reconnected electrically with the RGB sensor via small wires. Data acquisition and the control of the light source was achieved by an Arduino^© micro-controller. The user interface and the data storage were created using a LabVIEW^© program.

The basic idea behind the colorimetric measurement with the RGB sensor is shown in Fig.  for the example of PhR. In acidic pH regions (< 6) PhR indicates the pH through a yellow color. Following chromatic rules, high signals of green, less of red and least of blue are expected. In near-neutral pH regions, PhR changes the color from yellow to orange because the amount of protonated PhR increases. Thus, the RGB unit should measure decreasing signals of green and increasing signals of red and blue. With even higher pH values of more than 7.3 the protonated form of PhR dominates and gives the solution a magenta shade with a tendency towards red. The signals detected should change to high amounts of red and blue and very low amounts of green. In consideration of the explained example, the transformed hue values for PhR and additionally for BTB are summarized in Table .

Different measurements with various concentrations of PhR and BTB were carried out. BTB was used as a control chemical for evaluation purposes. Furthermore, experiments containing nutrition media were carried out and will be presented in Sect. .

The following data show sigmoidal shaped functions, which is a result of the color change of indicators. This can be either related to changes of the pH value or to concentrations of specific analytes . In general, the data fitting can be achieved with different sigmoidal models such as the Boltzmann equation, the logistic equation or the Hill function. However, within this work the fittings were achieved with the dose-response equation utilized by programs like Origin or R. A modified dose-response equation containing a hill slope coefficient showed the best fitting results regarding the least root-square sum and the double standard deviation (2σ).

Two different models were used to describe the sigmoidal shaped function of PhR (Eq. ) and BTB (Eq. ).

HSig=Hmin+Hmax-Hmin1+10(pH-pKa)αHSig=Hmin+Hmax-Hmin1+10(pKa-pH)α Hmax and Hmin are the respective hue values for the top and bottom plateau, HSig is the measured hue value given for a specific pH value, α is used to describe the slope steepness (hill coefficient) and pKa is the logarithmic acid dissociation constant of the indicator used.

On the basis of this mathematical approach, the calibration data, and thus the pH value, depend on four different parameters, which have to be fitted first. Hmax and Hmin should be determined experimentally at very high and low pH values because they indicate the maximum of the protonated and deprotonated form of the indicator used. It is said that most common indicators reach their respective maximum and minimum at about pH = pKa±3 . Therefore, all experiments were carried out in a pH range from 4 to 10 to fulfill these requirements for both PhR and BTB.

Although the pKa is commonly said to be constant, it depends on temperature, ionic strength and the solvent dielectric constant . PhR changes its color from yellow to magenta/red in the pH range between 6.4 and 8.2 with a pKa=7.4–7.9 . The color change of BTB appears at the limits of pH in between approximately 5.8 and 7.6 with a pKa=7.1 . Taking this into account, the pKa value is one of two parameters needed to be fitted mathematically. The second parameter is the so-called hill slope coefficient α (often referred as H, , or d, ).

Considering the different approaches of pKa and α, these parameters have to be fitted within a calibration. Thus, the mathematical model is fitted on the experimental data based on the method of the least root-square sum without weighting as shown in Eq. (): Distance=∑YData-YCurve2,RSS=Distance, where RSS is the root-square sum, YData is the measured hue value and YCurve is the hue value calculated with the model . The fitting itself is done by the solver function of MS Excel using the GRG-solver algorithm. This algorithm changes the pKa and α values starting at pKa=7.1 and α=1 to minimize the RSS value. As long as the pKa is within a range of 7 to 8 and α is within a range between 0.85 and 1.5, the data were determined to be fitted well.

Regarding the data shown in Fig. , the measured color angle of PhR ranges from H∼40∘ at pH = 2 to H∼-7∘ at pH = 11. Therefore, PhR should be yellow with a slightly red shift in alkaline pH regions, whereas it should be red with a tendency to magenta in acidic pH regions. Observations with the naked eye estimated hue values between 60∘ (yellow) and -40∘ (magenta, slightly red shifted). Although these data display a discrepancy between estimated (shown in Table ) and measured signals, the fitting results show stable measurement results. However, in acidic pH regions the discrepancy is quiet significant. The hue value in basic pH regions implies a strong red color, whereas the observable color contains much more magenta. This effect can be explained with an existing effect of the LED on the measurement. Therefore, two different experiments were carried out to validate this effect. First the concentration of PhR was increased to achieve a higher interaction between the dye and the light. Secondly, the hue of the LED without any dye included was determined.

A PhR : SPPB ratio of 1 : 5 was chosen to be measured using the developed prototype. The results, in the form of hue values, are shown in Fig. . The included error bars indicate the double standard deviation of the distance between the measured hue and the fitted data point. Within these results a double standard deviation of 1.5∘ was achieved. In comparison to the data shown before, a higher accuracy was obtained with a higher concentration of the indicator. This is basically induced by the stronger color which enables a higher level of light–matter interaction. An increase in excited indicator molecules consequently results in more distinct hue values. Additionally, the influence of the LED is decreased, because reduced non-interacted light emitted by the LED shines through the sample. Although the Hmax value obtained is nearly the same, Hmin is different with Hmin=-19.02∘. This hue value still displays a discrepancy to the expected hue value of about -40∘, but shows a far better tendency towards magenta. Considering this information, a change of the concentration of PhR did not result in a change of the maximum hue value, but resulted in a change of the obtained minimum.

Different hue measurements of PhR imply a possible influence of the LED on these measurements. For evaluation purposes, the color angle of the used LED was obtained and compared with hue values measured in two different PhR concentrations. The results of this experiment are reported in Fig. . In fact, the hue value of the LED (H=34.33∘) and the hue values of PhR in very acidic pH regions (Hmax(1:10)=37.57∘ and Hmax(1:5)=38.04∘) are very similar to each other. Currently, the results seem to indicate that the difference within acidic pH regions is not significantly influenced by a higher concentration of PhR. In contrast to the acidic pH regions, the difference between the color of the LED and the color of PhR in alkaline pH regions (Hmin(1:10)=-7.83∘ and Hmin(1:5)=-19.02∘) is much higher, and the hue values of the two PhR concentrations vary from each other. PhR has an absorbance maximum in the wavelength region of yellow light between 550 and 580 nm. With increasing pH values, more yellow light of the LED is absorbed, which results in the color change from yellow to magenta. Due to this absorption characteristic of PhR, the difference of the hue value between the LED and PhR in acidic pH regions is not influenced by the concentration of the dye. However, the effect of the absorbance of yellow light is much higher within alkaline pH regions and therefore dependent on the concentration.

In addition to the effect caused by the color spectrum of the LED, the position of the LED with respect to the sample is said to have a high influence on the measurement results. A matrix of pictures of PhR and BTB (1 : 10) at pH 3 and pH 10 was made and is shown in Fig. . To achieve this, the top layer of the second prototype was dismounted. From the edge to the center of the chamber, the intensity of the LED becomes stronger. This is mainly caused by the semicircular angle of radiation of the LED, thus yielding different path lengths and reflection angles of the light, which result in a specific light cone. The nearly unaffected light of the LED takes up about one-third of the complete chamber. Despite that, the experimental results show stable and reproducible signal sequences.

For a better comparison of the effect of different concentrations of BTB in SPPB, the results for two different dilution ratios are reported within one diagram shown in Fig. . Both curves for H(1:10) and H(1:5) share almost the same lower limit of Hmin(1:10)=32.24∘ and Hmin(1:5)=29.93∘. Additionally, the maximum plateaus with Hmax(1:10)=221.51∘ and Hmax(1:5)=226.38∘ are also nearly the same. The main parts of the curves are mostly identical, although a small difference can be seen.

Conversely to the initial assumption and comparing these results with the results of different concentrations of PhR, no differences between the two used concentrations of BTB were obtained. One possible explanation could be that both of the utilized BTB concentrations already show the maximum of the obtainable signals. In conclusion, BTB can be diluted even further in comparison to PhR because even with a dilution ratio of 1 : 10 the solution is still saturated with BTB. Therefore, BTB is more suitable for pH monitoring in even smaller micro-reactors in comparison to PhR. Only very small deviations are detected for the two dilutions of 1 : 10 and 1 : 5 in Fig. . These minor deviations are due to the mechanical manufacturing of the chambers resulting in slightly different surfaces (flatness, roughness, micro-scratches etc.) and the subsequent different optical properties of the chambers (transparency, scattering etc.). In the case of industrial manufacturing routines, even such small differences may be eliminated resulting in very precise determinations for the pH values.

Results reported beforehand were aiming for a characterization of PhR and BTB in pH ranging from about 3 to 10. Nevertheless, the following section is more focused on a defined pH range from pH = 4 to pH = 7.5. During cell culture, it is very uncommon for the solution to become alkaline. In most cases, dying cells produce specific acids which change the pH value of the nutrition media towards lower pH. Taking this into account, the non-diluted DMEM nutrition media with a basis pH of about 7.5 is used as the starting point. By using 1 M HCl, the pH values were set to acidic pH ranges. But conversely to pure PhR and BTB, the nutrition media is a far more complex solution. Besides different salts, liquids and nutrients such as glucose, it contains a pH buffer system to stabilize the pH in certain ranges for a better cell growth. The DMEM used has a sodium bicarbonate buffer included. This buffer system leads to a rising bubble formation within the solution at low pH values. These bubbles are caused by CO2, which is produced during the following chemical reaction: NaHCO3+HCl→NaCl+CO2+H2O.

A good example for this chemical reaction can be found within the human digestive system. However, within very small working volumes of 100–150 µL, these bubbles can have a high influence on the measured hue value because of additional scattering effects and local changes of the dye concentration.

The hue data of four different chambers are reported in Fig. . The curve has to be read from the near-neutral pH value backwards to the acidic pH range. With each experimental step, more HCl was added and therefore, the pH value shifted from 7.5 to 4. It was not possible to obtain any kind of usable data if nutrition media and HCl were pipetted together within the 150 µL chambers. The CO2 bubbles developing within the chamber had a strong negative influence on the measurements. Therefore, the pH was set in a beaker glass containing the stock solution of the nutrition media by adding small amounts of HCl. The developing bubbles were stirred out within this beaker glass. Meanwhile, the pH value is controlled with a pH glass electrode. After specific pH values were reached, the solution was taken from the beaker glass into specific reactor mock-up chambers. Although most of the evolving gas was stirred out, it was not possible to release all of the bubbles. The remaining CO2 bubbles were expected to have a high influence. However, no significant additional scattering was detected.

Although just small amounts of PhR are included within the used DMEM, the reported data display a high accuracy and stability within the hue value. The color change of PhR can be detected as a function of the pH.