1 Introduction

Phosphates are an essential nutrient for all forms of life (where they are used in biological energy production) and find extensive uses in fertilizers and industrial process [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15]. Unfortunately, they are also a significant pollutant in surface water [10], where high levels of phosphates lead to eutrophication—acceleration of algae growth—that causes reduction or even elimination of oxygen in the water and the production of toxins and harmful gases [10]. This lack of oxygen can kill fish and the algae can be hazardous to humans, pets, and animals/livestock [1, 5,6,7, 10]. Given the health and environmental hazards associated with phosphate contamination in water there has been much work devoted to developing techniques to measure phosphate concentrations in water, with the ideal technique being simple, easy to use, weather resistant, and low-cost. Note that depending on the application and environmental standard, target limits-of-detection for elemental phosphorus in water range from 0.25 mg/l(\(\sim\)8 \(\mu\)M) to 0.01 mg/l(\(\sim\)33 nM) [16].

One of the simplest techniques proposed for phosphate detection is fluorescent sensor materials which react to the presence of phosphates in water resulting in changes to their fluorescence properties. These changes can be quantified in-lab using known phosphate solutions to produce a calibration curve. Once this calibration curve is known, a water monitoring agency can collect a water sample from a water source and determine its phosphate concentration.

Many fluorescent sensors have been developed over the past decade for the purpose of phosphate detection with varying degrees of success, including: Tb-H\(_3\)cpboda[17], Eu-doped BUC-14 nanocrystals[18], UiO-66-NH\(_2\)[19], Eu-BTB[20], Fe-complex[21], Tb@Zn metal organic framework (MOF)[22], Tb(H\(_2\)O)(BTB)[23], Tb-MOF[24], Zr-UiO-66-N\(_2\)H\(_3\)[25], Al-MOF[26], Eu-MOF[27], Eu-ciprofloxacin[28], ZnO quantum dots and MOF[29], and Zn-MOF[30]. While many of these sensors have good sensitivity and selectivity for phosphates, all of them are metal-based complexes requiring complicated and expensive synthesis. This complication and cost hinders their wide-spread use.

In this study, we take an alternative approach, where we test a widely available, off-the-shelf phosphor, Eu(acac)\(_3\) and test its phosphate sensitivity. We find that while its sensitivity is not as high as some of the purpose-made sensors, it is still a reasonable sensor for some applications. Additionally, we characterize the sensor’s performance as a function of sensor concentration and find that its quenching coefficient is inversely proportional to sensor concentration, which results in the limit-of-detection being linear in sensor concentration. These observations are important for real world application of fluorescence-based phosphate sensors and to our knowledge have never been reported in the literature before.

2 Methods

2.1 Sample preparation

The sample solutions used in this study were prepared from two stock solutions: (1) was prepared by dissolving Eu(acac)\(_3\) in methanol (MeOH) at a concentration of 1000 mg/L (2226 \(\mu\)M), while (2) was prepared by dissolving NaH\(_2\)PO\(_4\cdot 2\)H\(_2\)O (Sigma Aldrich \(\ge\)99.0%) in DI water at a concentration of 1000 mg/L (8335 \(\mu\)M). Note that DI water is used to ensure that there are no contaminates in the water to begin with. To obtain lower concentration solutions we diluted the stock solutions using an appropriate amount of either MeOH for solution (1) or DI water for solution (2). Once we had a MeOH and water solution with the correct concentrations, we then combined them in equal volumes to perform measurements. Note that in the results below we report the concentrations in the individual solutions, such that the final concentration is actually half of what is reported. This choice is based on the fact that in a real-world application we will want to determine the phosphate concentration in a water sample and not the concentration in the combined MeOH/water solution.

2.2 Fluorescence spectroscopy

To perform fluorescence spectroscopy on the solutions we used a fiber-coupled ns DPSS UV laser (Photonics Industries DCH-355-5, 355 nm, 10 ns, 1 kHz), an Ocean Insight Square One Cuvette holder, and a fiber-coupled spectrometer consisting of a Princeton Instruments SpectraPro 2500i monochromator (0.5 m, 600 g/mm, 150 \(\mu\)m slit width, 0.47 nm resolution) and attached PI-Max 4 ICCD camera. We measured five separate spectra for each sample (with a shake in between each spectra) and then computed their average spectrum and standard deviation. We additionally calculated their correlation coefficients relative to the reference sample (i.e., the sample with no phosphate) using Pearson’s correlation coefficient.

3 Results and discussions

3.1 Effect of PO\(_4\) on fluorescence

We begin by first considering the effect of phosphate concentration on EA’s fluorescence spectrum for a sensor concentration of 2226 \(\mu\)M in MeOH and seven different phosphate concentrations in water. Note that the concentrations reported below are for the starting solutions, which are cut in half after mixing the EA/MeOH and PO\(_4\)/Water. Figure 1a shows the background subtracted fluorescence spectra and Fig. 2b shows the same spectra normalized to their peak intensities. From Fig. 1a we find that as the phosphate concentration increases the fluorescence intensity decreases due to phosphate-induced quenching. Additionally, from Fig. 2b, we observe that the spectral features remain consistent up to 833 \(\mu\)M but then change for our next concentration tested (4170 \(\mu\)M). Namely, we find that the \({}^5D_0\rightarrow {}^7F_2\) transition broadens and the \({}^5D_0\rightarrow {}^7F_1\) transition increases in intensity. Note that this transition occurs as the PO\(_4\) concentration increases above the EA concentration and that the resulting spectra look similar to the fluorescence spectra of EuPO\(_4\) [31, 32]. This observation suggests that at high phosphate concentrations some of the acac ligands are replaced by PO\(_4\) anions to either form EuPO\(_4\) or a mixed complex with both acac and PO\(_4\). Note that we do not observe any precipitation of crystals from the solution, which suggests that the product of this reaction is still soluble for the concentrations tested.

Fig. 1
figure 1

Baseline subtracted fluorescence spectra a and normalized spectra b from Eu(acac)\(_3\) in MeOH/water solution for different concentrations of phosphate in water

Full size image

To quantify the effects of PO\(_4\) on EA’s fluorescence, we compute the peak intensity ratio \(I_0/I(C)\) (where \(I_0\) is the peak intensity for the zero-phosphate concentration solution) and the spectral correlation coefficient (Pearson’s correlation coefficient for each spectrum relative to the zero-phosphate concentration’s spectrum) for each phosphate concentration C. Figure 2a shows the resulting peak intensity ratios and Fig. 2b shows the resulting in correlation coefficients with fits to a Hill equation added as a guide for the eye. Additionally, we perform linear fits for each ratio using a phosphate range less than the EA concentration and plot those in the inset of Fig. 2a. From both Fig. 2a, b we find that the intensity ratio’s and correlation coefficient’s phosphate-dependence changes as the phosphor’s concentration is modified.

Fig. 2
figure 2

Peak intensity ratio (\(I_0/I\)) a and spectral correlation coefficient b as a function of phosphate concentration with fits to a Hill equation as a guide for the eye. Inset a: Linear fits of ratios at low phosphate concentrations

Full size image

Namely, we find that the peak intensity ratio (Fig. 2a) increases linearly for phosphate concentrations less than the EA concentration, with the slope increasing with decreasing EA concentration. However, as the phosphate concentration increases above the EA concentration, the ratio curves behave nonlinearly. This is demonstrated by comparing the linear extrapolations (dashed curves) to the data which increases sublinearly for higher concentrations and appear to be increasing towards an asymptotic value. This asymptotic value is found to depend on the EA concentration, with larger EA concentrations producing greater asymptotic intensity ratios.

Turning to the spectral correlation coefficient (Fig. 2b) we find that while the intensity of the spectra change at low phosphate concentrations, the spectral shapes are essentially unchanged (the correlation coefficients are \(\approx\) 1). However, when the phosphate concentration becomes greater than the EA concentration the spectral correlation decreases to an asymptotic value which depends on the EA concentration. To get a better sense of these spectral changes we plot the normalized spectra for each EA concentration for a phosphate concentration of 8335 \(\mu\)M in Fig. 3. From Fig. 3 we find that the peak structures of the \({}^5D_0\rightarrow {}^7F_2\) and \({}^5D_0\rightarrow {}^7F_1\) transitions are similar across EA concentrations, but their relative intensity changes, with the \({}^5D_0\rightarrow {}^7F_1\) transition increasing in relative intensity (compared to the \({}^5D_0\rightarrow {}^7F_2\) transition) as the EA concentration increases.

Fig. 3
figure 3

Fluorescence spectra for different EA concentrations at a PO\(_4\) concentration of 8335 \(\mu\)M

Full size image

Based on the results from Figs. 2a, b, and 3 we hypothesize that for phosphate concentrations less than the EA concentration, the phosphates participate in collisional quenching, which results in decreased fluorescence intensity but limited effect on the spectral shape. This result occurs as collisional quenching involves a quencher (i.e., a phosphate anion) colliding with an excited phosphor molecule, resulting in the molecule returning to the ground state without emitting light [33]. As this process does not change the structure of the phosphor its spectral shape remains the same, but the total intensity from the solution decreases as some quantity of excited phosphor molecules now decay non-radiatively. On the other hand, as the concentration continues to increase, interactions between the phosphate anions and EA molecules occur that result in the formation of new complexes where the PO\(_4\) anions associate with the Eu\(^{3+}\) ions. This change in the complex structure results in both a decrease in fluorescence intensity and changes to the spectral shape.

3.2 PO\(_4\) limit of detection

Having determined the intensity ratio as a function of PO\(_4\) concentration we next determine EA’s limit of detection (LOD). Standard practice to calculate the LOD is to fit the low concentration ratio data (where the ratio is still linear in concentration) to a Stern-Volmer equation:

$$\begin{aligned} \frac{I_0}{I}=1+K_{SV}C, \end{aligned}$$
(1)

where C is the phosphate concentration and \(K_{SV}\) is a quenching coefficient. Note that while we use a linear fit to determine the LOD, this does not fully determine the calibration curve of the sensor as the ratio is nonlinear at higher phosphate concentrations. In actual practice it is necessary to have a nonlinear calibration curve to determine phosphate concentration in field samples. We are only fitting the low-concentration data to a line to determine the LOD.

Once the quenching coefficient is determined the LOD is calculated as LOD \(=3\sigma /K_{SV}\), where \(\sigma\) is the uncertainty in the reference intensity ratio found by measuring the reference sample’s intensity five times. Table 1 tabulates the quenching coefficient and LOD for each phosphor concentration and Fig. 4 plots the LOD as a function of EA concentration.

Table 1 Quenching coefficients and LOD for Eu(acac)\(_3\) at different concentrations
Full size table
Fig. 4
figure 4

LOD as a function of Eu(acac)\(_3\) concentration with a linear fit

Full size image

From Table 1 and Fig. 4 we find—for the range of EA concentrations tested—that the quenching constant is inversely proportional to EA concentration \(\rho\),

$$\begin{aligned} K_{SV}=\frac{\kappa }{\rho }, \end{aligned}$$
(2)

where the concentration-independent sensitivity \(\kappa\) = 1.558± 0.012. This inverse relationship between the quenching constant and the EA concentration results in the LOD increasing linearly with EA concentration with a slope of 0.0606 ± 0.0034.

3.3 Comparison to the literature

Having determined EA’s quenching constant, LOD, and \(\kappa\) for three different sensor concentrations, we now compare its performance to other phosphate-detecting sensors found in the literature. However, this quantitative comparison proves challenging as most current papers only report LOD (ignoring the effects of sensor concentration), don’t report uncertainties in their calculations, often fail to report their quenching constants, and rarely report the molar concentrations of sensors used. These deficits make it very difficult to make quantitative comparisons between papers. Note that comparing LOD between papers—which is typically done in the literature—is not a good comparison of sensor performance as LOD depends not only on the properties of the sensor, but also on the sensor concentration and the precision of the spectroscopic system used. Additionally, the LOD does not characterize the uncertainties associated with a given phosphor analysis system nor the range over which the phosphor is sensitive to phosphates.

We propose that if other fluorescene-based phosphor sensors quenching coefficients obey Eq. 2 the better comparison of sensors is the concentration-independent sensitivity \(\kappa\), which is fundamentally related to the interaction of the sensor and phosphate ion and is independent of sensor concentration and instrumentation. However, the assumption that other sensors obey Eq. 2 cannot be proven based on the current literature as the vast majority of studies only use a single phosphor concentration.

With these difficulties and assumptions in mind, we now make qualitative comparisons of different sensors from the literature in Table 2. Note that sensor concentrations marked with an asterisk denote concentrations that were calculated for this paper using the limited information presented in the literature. Therefore these concentrations should be taken with a grain of salt as they were not reported directly in the original papers.

Table 2 Comparison of phosphate detection performance for different sensors reported in the literature
Full size table

From Table 2 we find that EA’s LOD for the 556 \(\mu\)M and 2226 \(\mu\)M are on the high end of values reported, while the 44 \(\mu\)M concentration’s LOD is near the median LOD. Additionally, we find that EA’s concentration-independent sensitivity \(\kappa\) is on the low end as compared to the estimated values from the literature. This lower sensitivity is not surprising as EA was chosen as a low cost off-the-shelf sensor, rather than specifically designed for phosphate detection. While EA is not the most sensitive sensor, it is still sensitive enough for some applications which only require \(\mu\)M sensitivity such as streams that do not flow into lakes or reservoirs [16].

Note that Table 2 also lists one other off-the-shelf sensor material(s): Rhodamine 6 g and Molybdate. This combination of materials are roughly 3.69\(\times\) more sensitive than EA (based on the \(\kappa\) we estimated from Ref. [34]) and is also lower cost than the purpose-made sensors. However, this combination is more complicated to implement, requiring two solutions made using five different chemicals: Rhodamine 6 g, p-octylphenoxypolyethoxyethanol, ammonium heptamolybdate, hydrochloric acid, and water [34], while our approach only requires EA and methanol. Additionally, Rhodamine 6 g is significantly less photostable than EA for pumping with green light. To demonstrate this we consider each molecules photodegradation number B, which represents the number of photons a molecule can absorb before it is damaged. For Rhodamine 6 g, B is on the order of 10\(^5\)-10\(^6\) photons/molecule [35,36,37,38], while EA’s estimated photodegradation number is on the order of 10\(^{11}\) photons/molecule [39]. Therefore, EA will require replacement in a phosphate-sensing system significantly less often than Rhodamine 6 g; which will reduce the lifetime cost associated with replacing the phosphor in the phosphate-sensing system.

4 Conclusions

We have tested the off-the-shelf phosphor Eu(acac)\(_3\) (EA) as a fluorescent phosphate sensor in water. We find that at phosphate concentrations less than the phosphor concentration the fluorescence spectra decrease in intensity, but their spectral shapes are maintained. However, once the phosphate concentration exceeds the phosphors concentration the spectral shape begins changing. We hypothesize that this change is due to the formation of a new Eu complex that includes the PO\(_4\) ion. In the future, we plan on exploring this effect further by performing fluorescence lifetime studies of these solutions and attempting to precipitate out these new complexes for further characterization (e.g., FTIR, NMR, XRD). Additionally, these future studies will consider the selectivity of EA by testing other possible contaminates such as CO\(_3\), SO\(_4\), NO\(_3\), Cr\(_2\)O\(_7\), CrO\(_4\), F, Cl, etc.). Note that we have performed initial measurements for CO\(_3\), SO\(_4\), NO\(_3\) at high contaminate concentrations and find that EA’s fluorescence is relatively unchanged by SO\(_4\) and NO\(_4\), while CO\(_3\) is found to actually increase the fluorescence intensity.

With regard to EA’s performance as a phosphate sensor we find that it has a concentration-independent sensitivity of 1.558 ± 0.012, which results in a limit-of-detection (LOD) of 3.39 ± 0.68 \(\mu\)M at a sensor concentration of 44 μM. This LOD is found to be near the median of other fluorescent phosphate sensors, while its concentration-independent sensitivity is found to be lower than other values estimated from the literature. We note that significant inconsistencies in reporting and limited experimental details in the literature makes quantitative comparisons difficult.

Finally, we characterized EA’s performance as a function of sensor concentration and found that the quenching coefficient is inversely proportional to the sensor concentration, while the LOD is linear in concentration. To the best of our knowledge, these relationships have not been reported previously, with most papers ignoring the effect of sensor concentration. This is an oversight that should be remedied in future studies on fluorescent phosphate sensors as these results demonstrate that the current metric of LOD for sensors is poorly chosen. Furthermore, they fundamentally relate to the application of these sensors in real-world scenarios. For instance, since the LOD appears to depend linearly on sensor concentration, it is desirable to minimize the concentration used. However, decreasing the concentration also decreases the fluorescence signal and therefore there needs to be a balance between enough concentration to obtain a good fluorescence signal, while also minimizing the LOD. Additionally, as the quenching coefficient—which is used for calibration—depends on sensor concentration, care must be taken to have well-known concentrations to properly calibrate the response of the fluorescent phosphate sensor. Failure to account for this concentration dependence could lead to inaccurate results in field applications.

5 Disclaimer

This report was prepared as an account of work sponsored by an agency of the United States Government. Neither the United States Government nor any agency thereof, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States Government or any agency thereof. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof.