2.1.

Materials

Optical fibers including SMF and PCF ($10.1/125\mu \text{m}$ LMA-10) are purchased from Corning, (New York) and NKT Photonics, (Birkerød, Denmark), respectively.

Anhydrous methanol, sodium hydroxide granules, silver nitrate, 2-hydroxyethyl methacrylate (HEMA), 2-(dimethylamino)ethyl methacrylate (DMAEM), and ethylene glycol dimethacrylate (EGDMA) are purchased from Alfa Aesar, (Massachusetts) and glucose, tin dichloride 3-(trimethoxysilyl)propyl methacrylate (TMSPM) and $N,N,N^{\prime },N^{\prime }$-tetramethylethylenediamine (TEMED), inhibitor removers and ammonium persulfate (APS) are procured from Sigma, (St. Louis, Missouri). Sulfuric acid (95% to 97%) and hydrogen peroxide (30%) are from Merck, (New Jersey) and VWR, (Pennsylvania), respectively.

All reagents are of analytical grade and solutions are prepared using ultrapure water with resistivity 18.2 MΩ cm from Milli-Q system.

2.2.

Methods

2.2.6.

Experimental setup

Broadband light source (AS3223-BA2) with emission wavelength from 1530 to 1605 nm is used to interrogate the fiber sensor with the reflected light detected by an optical spectrum analyzer (AQ6370, Yokogawa, Tokyo, Japan) via an optical circulator as outlined in Fig. 1. The reflection spectra are recorded and further processed using MATLAB. pH response is characterized by immersing the pH sensor in phosphate buffer saline spiked with aqueous hydrochloric acid or sodium hydroxide.

Fig. 1

Schematic of the experimental setup. The LMA-10 core and cladding modes are represented by yellow and green arrows, respectively.

3.1.

Single-Pass Versus Double-Pass Fiber Interferometer

The proposed sensor improves on the single-pass Mach–Zehnder interferometer previously reported.^7,8 The fundamental optical mode guided by LMA-10 is a quasi-Gaussian mode with mode field diameter 8.8 μm (@1064 nm) which makes it confined within the optical fiber’s core and is thus termed the core mode. The collapsed air holes at the splice joint represent a region of uniform refractive index where the light guidance is lost. As a result, the guided mode of the SMF diffracts and broadens as it transverses the collapsed region to LMA-10. The extent of broadening can be estimated from Gaussian beam approximation¹² and the broadened beam excites fundamental mode and higher-order modes of LMA-10. These higher-order modes suffer from high optical losses and have significant energies propagating within the structured cladding of the LMA-10 and are commonly termed as cladding modes.

A second collapsed region serves to recombine the LMA-10’s core and cladding modes into the core mode of the ensuing SMF. Since the core and cladding modes have different propagation constants, they recombine to produce an interference pattern that is commonly described by a two-beam interference (between a core and cladding mode) model:⁷

Since the cladding modes propagate by reflection off the fiber cladding-exterior interface, their effective refractive indices are affected by the exterior refractive index while the core mode, due to its confinement within the fiber core, remains relatively insensitive to the exterior. As a result, a change in exterior refractive index alters the phase difference between the core and cladding modes leading to a phase shift in $I_{\text{interference}}$ that may be characterized by the amount of shifting of the interference dips.

A common drawback of the Mach–Zehnder interferometer is the inverse relationship between the interferometer length and the free spectral range (FSR) of the interference pattern.¹³ A longer interferometer, while able to give higher sensitivity through a larger optical path length difference, decreases the FSR, thus limiting dynamic range. We seek to mitigate this problem by coating a silver mirror at the interferometer end so that the light propagates through the interferometer twice. This also turns the interferometer into a probe device that is more convenient for sensing application. Figure 2 shows the interference pattern for a double-pass and single-pass Mach–Zehnder interferometer with the inset monitoring the spectral position of an interference dip with change in external refractive index. Although there is no significant improvement in sensitivity, larger fringe contrast is observed in the interferogram of the double-pass interferometer. Peak-to-peak values normalized with respect to the FSR were 0.863 and 0.444 for double-pass and single-pass interferometer, respectively, hence the quality factor, $Q$, of the interferometer is improved.¹⁴

Fig. 2

Spectra of double-pass and single-pass interferometers. (inset) Monitoring of spectra dip shifts with changes in external refractive indices.

3.2.

Coating pH Responsive Hydrogel

Optical fibers are first silanized with TMSPM to modify the surface with vinyl group for chemical attachment with the pH responsive hydrogel. The hydrogel is formed by in-situ free radical initiated polymerization process from pregel mixture of the constituent monomers, cross–linkers, and initiators. The relative amount of monomers, cross-linkers, and initiator influences the morphology and structure of the hydrogel which in turns affects its swelling/deswelling characteristics. A major challenge lies in the reproducible coating of a thin uniform hydrogel film on the optical fiber. A thin film is desirable since the swelling and deswelling are, in part, a diffusion driven process and thicker film would degrade responsiveness of the sensor. Surface initiated polymerization that immobilizes initiator on the surface to fabricate ultrathin hydrogel films (5 to 100 nm) have been reported¹⁵ but is cumbersome to implement on optical fibers as it requires additional chemical modification steps. By carrying out the polymerization process in custom-made PDMS channels, a uniform hydrogel film several micrometers in thickness surrounding the optical fiber can be formed. The thickness of the hydrogel coating extends beyond the penetration depth of the evanescent field of cladding modes (several hundreds of nanometers); hence, the sensor is sensitive to refractive index changes of the hydrogel coating but remains relatively insensitive to thickness change in the hydrogel coating.

3.3.

pH Response in Spiked Phosphate Buffered Saline

The interference spectra of the pH probe while immersed in phosphate buffered saline (PBS) of varying pH are shown in Fig. 3 with the corresponding shift in interference dip positions. The dip positions are obtained using a MATLAB script that locates dips based on the first derivative of spectra data. It is observed that the spectral dip consistently shifts to longer wavelength when the pH is increased from 5 to 9.

Fig. 3

(a) Interferogram at varying pH in phosphate buffered saline. (b) Corresponding spectral dip position.

This behavior is attributed to the swelling/deswelling behavior of the hydrogel in response to pH changes. The hydrogel, having a higher refractive index than the solution, increases the refractive index at the fiber surface when it deswells. This increases the effective index of the cladding modes, causing the interference pattern to shift toward longer wavelength. Similarly, a decrease in refractive index at the fiber surface occurs when the hydrogel swells, resulting in the shift of interference pattern toward shorter wavelength.

The pH-dependent swelling and deswelling of the hydrogel are studied at the bulk level by monitoring mass and refractive index change of a hydrogel slab. The hydrogel slab is polymerized from 1 ml of pregel solution with composition outlined in Sec. 2.2.4. The solution is dispensed into a 50-mm diameter Petri dish and APS/TEMED is added to initiate polymerization in a vacuum chamber. The hydrogel thus formed is immersed in deionized water for 2 days and cut into a rectangular slab. The mass of the hydrogel slab is measured using a microquartz balance (Mx5, Mettler Toledo, Ohio) after immersing in PBS of the studied pH for 24 h and drying with a lint free cloth. Refractive indices are measured using with an Abbe refractometer (ATAGO, Tokyo, Japan NAR-4T Solid) using glycerol (Sigma, St. Louis, Missouri, RI 1.4695) as contact liquid. Dry mass of the hydrogel was measured after heating at 40°C in a vacuum chamber overnight.

The mass swell ratio, defined as the ratio of the swollen hydrogel slab over its dry mass, is 1.78 and 1.55 in pH 5 and pH 9 buffer solutions, respectively. This represents a 15% mass difference between the mass of the hydrogel under acidic and alkaline conditions. The increased mass in acidic medium corresponds to swelling of the hydrogel network leading to the increased water content. The increased water content reduces the refractive index of the hydrogel as water has a lower refractive index (RI 1.33) than the polymer [RI of poly(2-hydroxyethyl methacrylate) is approximately 1.51]. The refractive index of the hydrogel in pH 5 is 1.39 and increases to 1.42 when the pH is increased to 9. This is still lower than the refractive index of material for the optical fiber (silica, $\mathrm{RI}\sim 1.4457$), enabling propagation of cladding modes even when the hydrogel is collapsed at high pH.

This swelling/deswelling behavior of the hydrogels is described previously by thermodynamic theories which treat swollen gel as a polymer solution of an analogous linear polymer in a solvent.^16,17 Polymer-solvent interactions, governed by thermodynamic variables, determine the mixing tendency between polymer and solvent responsible for driving solvent into the gel, thus exerting an expansion force that is opposed by an elastic retractive force exerted by polymer chains between the cross-links. An equilibrium swelling state is reached when expansion force balances the retractive force.

pH responsive gel incorporates ionizable acidic or basic groups that alter their swelling behavior. The ionization state of these groups is affected by local pH and changes the charge density within the gel. Here, DMAEM is copolymerized with HEMA with the former contributing basic amino groups to the gel network. Under acidic conditions, protonation of the amino groups within the gel leads to an increased charge density. Electrostatic repulsion between the charged groups causes the expansion of the gel network. In addition, diffusion of counter ions from the solution to balance the internal charges also increases the osmotic pressure within the gel and contributes to gel swelling. The equilibrium swelling of pH responsive gel has been modeled by various groups and found to depend on charge of the ionic monomer, $\mathrm{p}{K}_{\mathrm{a}}$ of the ionizable group, mole ratio of ionizable monomer, degree of cross-linking, and nature of surrounding solvent.¹⁸

The graph in Fig. 3 assumes a typical sigmoidal titration curve with an apparent $\mathrm{p}{K}_{\mathrm{a}}$ value fitted to be 7.31. This is approximately 1.1 pH units lower than the $\mathrm{p}{K}_{\mathrm{a}}$ of DMAEM in solution¹⁹ as more acidic condition is required to protonate DMAEM in the polymeric form owning to unfavorable electrostatic interactions.²⁰ The apparent $\mathrm{p}{K}_{\mathrm{a}}$ is close to physiological pH of pH 7.4, making it apt for physiological measurements due to its high sensitivity around the physiological pH.

A linear regression line can be used to fit the pH range: pH 6.75 to 8.25 with $r^2$ value of 0.986. The pH sensitivity within this linear range is $1.71\mathrm{nm}/\mathrm{pH}$, which is higher than reported sensors using pH sensitive polyelectrolyte multilayers.^21,22 The limit of detection based on ratio of sensitivity over $3\sigma$, where $\sigma$ is the standard deviation of five readings, is pH 0.004.

3.4.

Responsiveness of the pH Sensor

The responsiveness of the proposed pH sensor is tested by allowing the sensor to equilibrate in spiked PBS of pH 6.75 before changing to a spiked PBS of pH 7.75 (Fig. 4). The solution is changed to pH 6.75 spiked PBS following equilibrium at the higher pH.

Fig. 4

Step response of fiber sensor from pH 6.75 to pH 7.75.

The time taken to reach 90% response, ${\tau }_{90}$, is measured to be 55 s for change to higher pH and 79 s for change to lower pH. This responsiveness is slightly improved over other reported hydrogel based pH sensor²³ and is in part due to the robustness of an optical based sensing platform. The diffusion of protons through the hydrogel matrix plays an important role in the kinetics of gel swelling/deswelling,²⁴ thus thinner gel with larger mesh size are thought to be more responsive. The responsiveness of our sensor could be further increased by decreasing the gel thickness. Nevertheless, the responsiveness of the sensor is sufficient to investigate most biological processes.

3.5.

Hysteresis of the pH Sensor

The pH sensor is subjected to cyclic changes in pH to test for its repeatability which is essential for practical usage (Fig. 5). The reversible swelling and deswelling of HEMA-co-DMAEM hydrogel ensures minimal hysteresis with largest reported error of 0.8%.

Fig. 5

Sensor pH response under cyclic pH change.

3.6.

Stability of the pH Sensor

The stability of the sensor is tested over an 8-h period. Figure 6 shows the sensor performance with minimal fluctuations in PBS of varying pH. The maximum deviation is 0.04 nm corresponding to pH 0.0234 units. Therefore, the proposed sensor has minimal drifting and do not require frequent recalibration that is common for conventional glass electrode pH meter, making the sensor suitable for continuous pH monitoring.

Fig. 6

Stability of pH sensor.

3.7.

Application of the pH Sensor for Cell Culture Monitoring

Extracellular pH is among the most important parameter influencing cell health and growth and is thus closely monitored in cell culture experiments. The unique demands for continuous monitoring over time under a sterile enclosed environment in typical cell cultures render the conventional glass electrode pH meter inadequate owning to its bulkiness and need for regular calibration and maintenance.

The proposed pH sensor poses several advantages like its miniature size. The use of HEMA as the backbone of the pH responsive gel ensures good biocompatibility, resistance to degradation and allows the sensor to undergo heat sterilization without sustaining damage.

The fiber pH sensor is calibrated in spiked culture media prior to measurement. Figure 7 illustrates the pH change in the culture media measured by our fiber pH sensor and conventional pH meter, highlighting the close proximity between readings from both devices with an average 0.465% error. The media pH is noted to drop drastically from Day 3 corresponding to a sharp increase in cell number before stabilizing at around pH 6.8 from Day 5. Cell death rate is below 10% throughout the duration of the experiment.

Fig. 7

(a) Calibration of pH sensor in Dulbecco’s modified Eagle medium. Fitting of linear regression line between pH 6.50 and pH 7.75 with $r^2$ value of 0.9762. (b) Monitoring of cell culture pH using fiber pH sensor and glass electrode pH meter. Inset: Monitoring of cell proliferation.

The data agrees with reported pH 6.5 to 6.7 in in-vivo studies of tumor microenvironment.³ The acidic microenvironment that has become characteristic of tumor is the result of increased metabolic activities and a shift in cellular metabolism in response to hypoxia or in part, to confer competitive advantage over normal cells that are intolerant to acidity.²⁵ The fall in pH is also part of a malignant progression and favors upregulation of proteolytic enzymes that drives metastasis.²⁶ Clearly, pH is vital in cancer studies and our proposed sensor can be useful in-vitro studies to understand cancer or evaluate pH targeted cancer therapies.