2.1.

Sensor Configuration

Figure 1 shows the schematic view of the development HOXY optical oxygen sensor. The highly luminescent molecule tris(4,7-diphenyl-1,10-phenanthroline)- $\text{ruthenium}\left(\mathrm{II}\right){\mathrm{Cl}}_2$ , represented as $\mathrm{Ru}{\left(\mathrm{dpp}\right)}_3{\mathrm{Cl}}_2$ , is employed as the probe dye. It has a relatively high quantum yield of luminescence (0.30, dimensionless) and long luminescence lifetime (natural $\text{lifetime}=5.34\mu \mathrm{s}$ ). The lifetime of the dye in the sensor element was not measured. The dye is encapsulated in a gas permeable, but ion impermeable silicone layer, and incorporated to the optical probe (a glass capillary) to form the sensing element. A blue LED centered at $470\mathrm{nm}$ is used as the excitation light source. Upon excitation, $\mathrm{Ru}{\left(\mathrm{dpp}\right)}_3{\mathrm{Cl}}_2$ emits at $615\mathrm{nm}$ . In the presence of oxygen, the fluorescence is reversibly and quantitatively quenched. A photodiode, which is covered with a sheet of $575\text{‐}\mathrm{nm}$ long-pass optical filter, serves as the fluorescence detector (Fig. 1). There is significant emission from the LED at the detection wavelengths, but the offset is expected to be linear with LED output and was subtracted out during the calibration process. Additional details are provided in Ref. 29.

Fig. 1

End-on schematic view of the optical oxygen sensor. The LED fully illuminates the capillary tube; fluorescence is detected at $90\mathrm{deg}$ using a photodiode.

2.2.

Chemicals and Sensing Element

Commercially available $\mathrm{Ru}{\left(\mathrm{dpp}\right)}_3{\mathrm{Cl}}_2$ (GFS Chemicals) was recrystallized once in ethyl alcohol (Aldrich, HPLC grade) and dried at $95°\mathrm{C}$ . Milli-Q $\left(18\mathrm{M}\Omega \right)$ purified water was used for all purposes. To prepare the sensing element, $3.8\mathrm{mg}$ $\mathrm{Ru}{\left(\mathrm{dpp}\right)}_3{\mathrm{Cl}}_2$ , $4.0\mathrm{g}$ Superflex Clear RTV Silicone Adhesive Sealant (Loctite, distributed by Lab Safety Supply, Inc.), $0.48\mathrm{g}$ $\mathrm{Ti}{\mathrm{O}}_2$ (Aldrich), and $9\mathrm{mL}$ dichloromethane (Aldrich, 99.9%, HPLC grade) were weighed into a $20\text{‐}\mathrm{mL}$ precleaned amber glass vial, tightly closed, and mixed well with a shaker. The black shielding solution was prepared by mixing $0.42\mathrm{g}$ graphite (Aldrich), $4.2\mathrm{g}$ 3145 RTV silicone sealant (Dow Corning, MIL-A-46146), and $10\mathrm{mL}$ Toluene (GFS Chemicals, 99.7%) in a $20\text{‐}\mathrm{mL}$ precleaned amber glass vial. The vial was tightly closed and mixed well with a shaker. A transparent glass capillary tube ( $3\mathrm{mm}$ OD, $1.6\mathrm{mm}$ ID) was cut into $2.6\text{‐}\mathrm{cm}$ -length pieces. They were first cleaned with Alconox detergent solution by boiling for $15\min$ , and rinsed with water. The glass capillary tubes were then further cleaned by sonicating them at $80°\mathrm{C}$ in 80:20 (w∕w) ${\mathrm{H}}_2\mathrm{O}$ ∕ethanolamine (Aldrich, $99+\%$ ) for $15\min$ , rinsed with water for several times, and then dried at $95°\mathrm{C}$ .

The sensing membrane was constructed inside the glass capillary tube by flowing dry air through the capillary for $5\min$ . The fluorescent dye solution was then pushed through the capillary tube using a syringe. Dry air was then forced through the capillary for $5\min$ to dry the solution and form a thin, uniform layer. Air was forced through the capillary using a small air pump. A special mounting jig was fashioned to hold the sensor vertical while air flowed downward through the capillary. The second layer of black shielding membrane was formed in a similar manner. The finished sensing elements were placed in a covered dish for $1.5\mathrm{h}$ to allow further solidification. The sensors were then boiled in water for $20\min$ to reinforce the sensor layer adhesion to the capillary. Sensing elements were stored away from light for $24\mathrm{h}$ to cure before use. The sensor was subsequently sterilized by autoclaving at $110°\mathrm{C}$ for $15\min$ .

2.3.

Sensing Element Morphology

In order to characterize the surface of the Ru(II) complex film in the capillary, scanning electron microscope (SEM) images were obtained for the sensing element and shielding layers. Figure 2(a) shows the sensing element film coated into the glass capillary, without the shielding layer. The capillaries were mechanically fractured and placed into a SEM in environmental (wet) imaging mode. The imaging chamber pressure was approximately $1.5\text{Torr}$ . The sensor film was measured to be $33.7\pm 0.4\mu \mathrm{m}$ , and is shown to be uniform over the field of view. Figure 2(b) shows a magnification of some of the dome-like features seen in Fig. 2(a). The identity of the features is not known, but the overall surface morphology of the film is smooth. SEM images of the black shielding layer were identical to the sensing element, and are not shown. Since the two layers are primarily composed of the same material (silicone), the film thickness of the shielding layer could not be accurately measured.

Fig. 2

(a) SEM image showing cross section of sensing film (scale bar is $100\mu \mathrm{m}$ ). (b) Magnified image of sensing film and its surface features (scale bar $10\mu \mathrm{m}$ ).

2.4.

Sensor Hardware and Electronic Circuit

The entire sensor was built in a shielded aluminum box with a dimension of $2.3\times 2.4\times 2.5\mathrm{in}$ . The box is divided into two separable parts—a base plate and top cover plate. The main electronic board, photodiode, LED supporter, and cable adaptor are housed in the base plate. The blue LED is secured to the base plate by two small bolts. The LED is placed perpendicular to the sensing element and at the same height (Fig. 1). Two rubber seals, one between the base plate and the top plate and another between the cover and the base plate, are used to prevent moisture from entering the sensor. Ambient light did not enter the sensor once the system was closed. The advantage of designing the sensing element plate separated from the base plate is that the sensor can be sterilized independently of the sensor electronics.

Circuits for the $\pm 9\text{‐}\mathrm{V}$ dc power supply, signal amplifier, blue LED power supply, and LED pulse control are built into the main PCB board. A connector interfaces the main board and the photodiode board. The $+5\text{‐}\mathrm{V}$ dc from a National Instruments data acquisition board (NIDAQ PCI 6052E) is converted to $+12\text{‐}\mathrm{V}$ dc by a dc-dc converter (Pico5A12D); the $\pm 12\text{‐}\mathrm{V}$ power is further stabilized by two voltage regulators (78L09 and 79L09) to achieve $\pm 9\text{‐}\mathrm{V}$ dc, which supplied the power for the photodiode (with built-in preamplifier), amplifier, and blue LED (Lumex).

A Hamamatsu S5591 photodiode with built-in preamplifier is used for the light detector. A feedback resistor external to the photodiode case controls the gain for the photodiode. An external capacitor is used as a filter. All three components, photodiode, resistor, and capacitor, are placed in an independent board, and the $\pm 9\text{‐}\mathrm{V}$ power source and signal output are connected to the main board. The amplification circuit uses an optional amplifier (AD548JN), with a $500\text{‐}\mathrm{k}\Omega$ trim potentiometer for gain control. The voltage output is connected to a NI DAQ board for data acquisition. A blue LED is used as the excitation light source for the fluorescent sensing element. Its output (on∕off) and the pulse width are controlled via an electronic switch (MAX323CPA) through a digital I∕O channel by the software.

2.5.

Software

For long-term DO monitoring using an optical oxygen sensor based on dynamic fluorescence quenching, bleaching of the fluorescent dye due to long-term continuous irradiation can be a problem. Therefore, software written in LabVIEW (National Instruments, version 6i, Austin, TX) was developed to run the HOXY sensor in pulse mode to solve that problem. The advantage of using the LED in pulsed mode is that the photodegradation that occurs during continuous operation is avoided. For example, if we plan to use the sensor to continuously measure DO for $30\text{days}$ (that is $720\mathrm{h}$ , or $\mathrm{43,200}\min$ ), at a sampling rate of $1\text{sample}∕\min$ , and the LED pulse width is $100\mathrm{ms}$ (meaning that each minute the blue LED will be turned on for only $100\mathrm{ms}$ ), the total irradiation time in $30\text{days}$ is only $4320\mathrm{s}$ , or $72\min$ . Bleaching of dye in $72\min$ is insignificant and will not influence the results.

The HOXY software combines the oxygen sensor control and bioreactor control, forming a virtual instrument (VI). The software has a simple graphical interface that allows control over the various experiment parameters, such as blue LED pulse∕continuous mode, LED pulse width, sampling rate, ADC (analog∕digital converter) scan rate, number of scans, manual infusion∕auto infusion and its schedule, and calibration equation. Data for up to four different channels (sensors) can be simultaneously acquired and logged. This software has been successfully used to operate the HOXY oxygen sensor in long-term stability tests and continuous measurement of media DO in bioreactors supporting cell cultures lasting for $180\text{days}$ . The major functions of the HOXY software are data collection and display, LED pulse control (frequency $0.017\mathrm{Hz}$ and pulse width $100\mathrm{ms}$ ), conversion of photodiode voltage to oxygen concentration via a calibration function, and control cell culture infusion and perfusion.

3.1.

Oxygen Sensor Response

Figure 3(a) shows the response of the HOXY sensor when nitrogen and air are alternately flown through the sensor. The sensor output signal is 240% higher in nitrogen than in air, indicating that the HOXY sensitivity is adequate for the range of DO concentrations observed in cell culture media. Figure 3(b) shows a similar response curve in PBS solution with the DO alternated between $0\text{to}160\mathrm{mm}$ Hg. The typical oxygen concentration range in mammalian cell cultures observed in our laboratory is $0\text{to}100\mathrm{mm}$ Hg. In solution, the response time $\tau$ (time required to reach 90% of the steady state) from $0\mathrm{mm}$ Hg oxygen level to $160\mathrm{mm}$ Hg oxygen level is about $42\mathrm{s}$ , and from $160\mathrm{mm}$ Hg to a low level is about $390\mathrm{s}$ . The response times are given the time scale for mammalian cell consumption of oxygen in perfused bioreactor systems.

Fig. 3

(a) Oxygen sensor signal-time trace in nitrogen∕air and (b) signal-time trace in PBS solution with the DO alternated between $0\text{to}160\mathrm{mm}$ Hg.

3.2.

Calibration of the Oxygen Sensor

Calibration of the HOXY oxygen sensor in PBS solution was performed by periodically increasing the oxygen ratio of the mixture of air and nitrogen that continuously purged the solution flowing through the sensor. The dissolved oxygen concentrations of each step were measured by a Ciba-Corning blood gas analyzer (BGA Model 248). The calibration curve fitted by a biexponential equation is shown in Fig. 4 . The calibration equation for the HOXY oxygen sensor is given by the following:

Fig. 4

Oxygen sensor calibration curve.

3.3.

Oxygen Sensor Noise

Figure 5 shows the data of the sensor output in PBS solution at $37°\mathrm{C}$ during a period of $4\mathrm{h}$ . According to the signal fluctuation, random noise $\left(N\right)$ can be determined to be about $3\mathrm{mV}$ . The noise fluctuation is attributed to the LED light source and photodiode detector noise. The drift (noise) is $2\mathrm{mm}$ Hg, with an accuracy (measured versus a blood gas analyzer) of $5\mathrm{mm}$ Hg. The resolution of the sensor $\left(2N\right)$ is $0.15\mathrm{mm}$ Hg (at $50\mathrm{mm}$ Hg oxygen) and $0.375\mathrm{mm}$ Hg (at $130\mathrm{mm}$ Hg oxygen), respectively.

Fig. 5

HOXY oxygen sensor noise level measured in PBS in incubator at $37°\mathrm{C}$ .

3.4.

Long-term Stability Test of the Sensor

A HOXY oxygen sensor was employed to continuously measure the gaseous oxygen alternately in air or nitrogen at room temperature for a total period of $20\text{days}$ , and the data is shown in Fig. 6(a) . The drifting of the sensor voltage output in $20\text{days}$ is only $-2\%$ , indicating that bleaching of the dye in this period is not significant.

Fig. 6

(a) Long-term stability test of HOXY oxygen sensor in air∕nitrogen and (b) in air∕nitrogen-saturated PBS solution.

Figure 6(b) shows the result of a similar long-term continuous test of the same oxygen sensor in PBS solution for a total period of $15\text{days}$ ; the PBS solution was alternatively saturated with air or nitrogen at room temperature. Drifting of the sensor voltage output in all $15\text{days}$ is not detectable, meaning that leaching of the fluorescent dye is insignificant.