Abnormal Behavior in Stern–Volmer Luminescence QuenchingMeasurements via Apparent Lifetime Methods

SARAH J. PAYNE, J. N. DEMAS,* and B. A. DEGRAFFChemistry Department, University of Virginia, Charlottesville, Virginia 22904 (S.J.P., J.N.D.); and James Madison University, Harrisonburg,

Virginia 22807 (B.A.D.)

Luminescence lifetimes are widely used as an analysis tool. Since decays in

analytical systems are frequently complex decays rather than single

exponentials, apparent lifetime methods based on the rapid lifetime

determination (RLD) method or single frequency phase shift (SFPS)

measurements are frequently used to reduce cost and simplify data

analysis. It is demonstrated here that these methods can produce large

errors under the right conditions. Both methods can give unexpected and

uncharacteristic Stern-Volmer quenching plots (SVQPs) in two-compo-

nent systems. Behaviors include bimodal quenching curves as well as

‘‘anti-quenching’’ curves. These phenomena are exacerbated by small

fractions of long unquenched components.

Index Headings: Stern–Volmer quenching; Artifacts; Anti-quenching;

Luminescence.

INTRODUCTION

Luminescent sensors are widely used within the biological,1

industrial,2 and analytical3 fields. They are used to detectanalytes such as oxygen,4 pH,5 humidity,6 and glucose,1 aswell as various cations and anions.7 Using luminescent sensorsfor detection is attractive for several reasons, including lowcost, excellent stability, sensitivity, ease of use, selectivity, andreusability.8

Luminescent signals can be observed in two primary ways:emission intensities (I) and lifetimes (s). By using lifetimes orpeak intensities, information about quenching and kinetics maybe determined. We focus on quenchometric sensors wherebythe analyte quenches a luminophore. Examples include tris(4,7-diphenyl-l,10) ruthenium (II) perchlorate to determine oxygenconcentrations9 and a 1,4-diphenylethynyl-benzene chromo-phore with 18-crown-6 functionalities bound to the outerphenyl rings for the detection of lanthanide ions.10

Lifetime measurements have several advantages overintensity measurements. Intensities can vary with changes inthe excitation intensity, detector aging, optical paths, andsensor photodecomposition. These effects can necessitate useof an internal standard for accurate quantitative measurements.On the other hand, one benefit of using intensity measurementsis that one deals analytically with a single value, the emissionintensity, for each sample. Advantages of using lifetimes stemfrom the fact that they are intrinsic properties of moleculesunchanged by fluctuations in the excitation source or theconcentration of the luminophore. A drawback in lifetimesystems is that decay curves may not be well-defined singleexponential decays but complex multi-exponential decays withvariable contributions. This causes difficulties in fitting the dataand generating analytically useful calibration curves. Apparentlifetime methods can provide simple solutions for analytically

complex scenarios with no need for detailed analysis. Thesemethods are used in microscopy to look at processes withinbiological samples11 as well as for analytical experimentscarried out in relatively unstable systems such as flames.12

Quenchometric measurements are generally based on Stern–Volmer quenching, which relates the concentration of thequencher to a ratio of the unquenched and quenched lifetimes.For complex decays the apparent lifetimes are frequently used.The two common ways of evaluating apparent lifetimes are therapid lifetime determination (RLD) and the single frequencyphase shift (SFPS) methods. The RLD method was firstintroduced by Ashworth,13 and since then the errors14–16 andoptimum methods of analysis17 have been studied. SFPS is themost widely used technique and is now common influorescence lifetime imaging microscopy (FLIM).18–20

These apparent lifetime methods can be extremely usefulfor measurements where there is limited computationalcapacity and where rapid real-time response, with instrumen-tal simplicity and low cost, is essential. However, there canalso be pitfalls. It has generally been accepted that anydeviations from linear Stern–Volmer quenching plots(SVQPs) would still change monotonically and provide ameaningful calibration curve.21 Upon further investigation,we have found that this is not always the case and bizarreSVQPs can be obtained. Using oxygen quenching ofruthenium complexes we show that multicomponent systemsevaluated with apparent lifetime methods can produce ratherunusual behavior in SVQPs.

EXPERIMENTAL

Materials. Tris-(4,7-diphenyl-1,10-phenanthroline) rutheni-um (II) dichloride, [Ru(dpp)3]Cl2 was obtained and usedunaltered from GFS Chemicals, Inc. Tris-(2,20-bipyridine)ruthenium (II) dichloride [Ru(bpy)3]Cl2 was synthesized viastandard literature methods,22 but can also be obtained fromGFS Chemicals, Inc., as well as from other suppliers. Theruthenium nanoparticles (PD-Ru1 carboxylated, product#15910), 0.4% (w/w), were obtained from Active Motif(Carlsbad, CA). They have an average diameter of 40 nmand an emission maximum of 600 nm. The solvents were all ofreagent grade. Water was distilled.

Instrumentation. Lifetime measurements were obtained viaa homebuilt system. The excitation source was a 337 nmpulsed N2 laser, and stray red plasma excitation light wasremoved with a 1 cm 0.1 M aqueous CuSO4 solution filter,which cuts off above 600 nm. The emission was observedusing an RCA C7164S photomultiplier tube with a 1 cmsaturated aqueous NaNO2 filter (blocks below 400 nm) and amonochromator (Jobin Yvon 5/124V). The monochromatorwas set to 620 nm, near the emission maxima of all the samplesA digitizing oscilloscope (Tektronix TDS540) averaged 300

Received 11 December 2008; accepted 22 January 2009.* Author to whom correspondence should be sent. E-mail: demas@virginia.edu.

Volume 63, Number 4, 2009 APPLIED SPECTROSCOPY 4370003-7028/09/6304-0437$2.00/0

� 2009 Society for Applied Spectroscopy

decay curves before transferring the decay to a PC. The wholesystem was controlled with Labview 5.1. Absorbance mea-surements were obtained using a UV-Vis spectrometer (PerkinElmer Lambda 25).

Simulations. All simulations were performed in Mathcad14. The RLD method is illustrated in Fig. 1. The areas A0 andA1 under two contiguous portions of the decay each of width Dtare measured. The apparent RLD lifetime (sRLD) is given by

sRLD ¼Dt

lnðA0=A1Þð1Þ

where Dt is the time interval over which A0 and A1 areintegrated. Analytical integrals are used in the simulations. Forsingle exponential decays, sRLD is the actual lifetime, but formore complex decays it is an apparent lifetime.

Two-component systems were modeled using

IðtÞ ¼ f1e�t=s1 þ f2e�t=s2 ð2Þ

where I(t) is the intensity as a function of time, fi is thefractional pre-exponential contribution from each component,and si is the lifetime as a function of quencher concentration[Q] given by

si ¼s0i

1þ Ksvi½Q� ð3Þ

where s0 is the lifetime in absence of quencher, Ksv is theStern–Volmer quenching constant, and i denotes the compo-nent (i.e., 1 or 2). The RLD lifetimes as a function of quencherconcentration or pressure in this work were then determinedusing Eq. 1.

Single frequency phase shift apparent lifetimes are calculat-ed from23

U ¼ arctanN

D

� �ð4Þ

N ¼Xn

i¼1

fisin ui cos ui ð5Þ

D ¼Xn

i¼1

ficos2ui ð6Þ

ui ¼ arctanð2pFsiÞ ð7Þ

where F is the frequency of modulation and si is the individuallifetime of the component.

All data are expressed as Stern–Volmer quenching plots.Equation 8 gives the relationship between the apparent lifetime(sapp) and quencher concentration [O2].

s0app

sapp¼ 1þ Ksv½Q� ð8Þ

where s0app is the lifetime in the absence of quencher, sapp isthe lifetime at different quencher concentrations [Q], and Ksv isthe Stern–Volmer quenching constant. For single componentsystems exhibiting only dynamic quenching, the linear form isobeyed, but this is frequently not true for two-componentsystems. Since oxygen is always a dynamic quencher in thesesystems, we can ignore any static quenching. Further, thecounterions involved are all innocuous non-quenchers.

In the simulations, two-component decays were evaluated byboth RLD and SFPS with various pre-exponential factors andvalues of Ksv, s, Dt (RLD), or frequency (SFPS).

Synthetic Systems. To generate real data with controllableparameters, we collected decays from short- and long-livedruthenium complexes in water and glycerol under varyingdegrees of quenching. Synthetic decay curves were thenprepared by mathematically mixing appropriate amounts of thedifferent real sample curves. Samples were aqueousRu(bpy)3

2þ (1) purged with nitrogen, air, 40% oxygen, 60%oxygen, and 100% oxygen. Decay curves of glycerol solutionsof Ru(bpy)3

2þ (2) and Ru(dpp)32þ (3) were obtained in air. The

glycerol decay curves were used as our unquenched compo-nent. The decay for component 1 was then added to 2.

FIG. 1. Illustration of the area under the curve for the RLD method. The A’sare the areas under the curve.

FIG. 2. Illustration of how decay A and decay B are added together to give thedata for the mixed systems (AþB).

438 Volume 63, Number 4, 2009

Similarly, component 1 was added to 3 (see Fig. 2 where adecay A is added to decay B to form the decay AþB).

Mixed System. To observe behaviors in a single solution anappropriate amount of phosphorescent nanoparticles (s0 ¼ 6.2ls) was mixed with 1.0 mL of 4.2 3 10�6 M Ru(bpy)3

2þ inwater (s0¼614 ns) to give a pre-exponential ratio of 2:5 for thenanoparticles to the Ru(bpy)3

2þ solution. Three separate decaycurves were taken at five different oxygen concentrations (0%,21%, 40%, 60%, and 100%). To generate SVQPs, RLDlifetimes were determined at various Dt values for each decaycurve. For each Dt and oxygen concentration, the three sRLD

values were averaged and used to generate SVQPs. Thestandard deviations were typically better than 5–10%. Theparameters (s, Ksv, f) for the nanoparticles and the Ru(bpy)3

2þ

solution were entered into the model to observe how well themodel fit real data.

Evaluation of Lifetimes. The RLD lifetimes were calculatedfrom experimental decays using the trapezoidal approximation

Area ¼Xn

i¼0

ðtiþ1 � tiÞf ðtiÞ þ f ðtiþ1Þ

2

� �ð9Þ

where the summations are over each Dt interval. The highdensity of data points ensures that the decay does not changegreatly between points and any approximation would havebeen adequate. These sRLD values were then displayed inStern–Volmer plots. Properties of each solution were enteredinto the model to compare our calculations with the real data.

RESULTS AND DISCUSSION

Rapid Lifetime Determination Simulations. Testing anumber of different sets of parameters, it was found that non-ideal behavior required the longer-lived component to beunquenched or weakly quenched compared to the shorter-livedquenched component. Figure 3 illustrates typical simulateddata that were modeled using the following parameters:Component 1: s01 ¼ 10 ls, Ksv1 ¼ 0 atm�1; Component 2:s02 ¼ 5 ls, Ksv2¼ 10 atm�1 with a 1:1 pre-exponential ratio.

As Dt increases SVQPs change dramatically and exhibit veryunusual behavior. At smaller Dt values the SVQP more closelyresembles what we would expect for a single componentquenched system. With very long Dt values the SVQP beginsto resemble an unquenched system. Intermediate Dt values givebimodal as well as anti-quenching SVQPs. Anti-quenchingoccurs when the apparent lifetime increases with increasingquencher concentration.

Synthetic Rapid Lifetime Determination Systems. Syn-thetic systems were created to span a range of behavior tomimic the theoretical simulations. The lifetimes and Ksv valuesare given in Table I. In Fig. 4 we show the SVQP for thecombination of Ru(bpy)3

2þ in water and Ru(bpy)32þ in glycerol

data using RLD lifetimes. The symbols are the data points andthe lines are the theoretical fits with the parameters for Ksv, s,and f values from the different components. The theoretical andexperimental values agree well. Figure 5 shows the combinedRu(bpy)3

2þ in water and Ru(dpp)32þ in glycerol, which is

similar to the results shown in Fig. 4. The major difference isthat the combination of Ru(bpy)3

2þ in water and theRu(bpy)3

2þ in glycerol (Fig. 4) more readily gave bimodalquenching curves while the Ru(bpy)3

2þ in water andRu(dpp)3

2þ in glycerol combination (Fig. 5) gave significant

FIG. 3. Model of RLD data at varying values of Dt. Component 1: s01¼10 ls,Ksv1 ¼ 0 atm�1; Component 2: s02 ¼ 5 ls, Ksv2 ¼ 10 atm�1 with a 1:1 pre-exponential ratio. The numbers are the RLD Dt integration width for eachcurve.

TABLE I. Lifetime and Ksv values for ruthenium complexes.

s0 (ns) Ksv (atm�1)

Ru(bpy)32þ in H2O 414a 2.637

Ru(bpy)32þ in glycerol 945a 0.b

Ru(dpp)32þ in glycerol 5253a 0.b

a Values in air.b Unquenched data used that makes Ksv ¼ 0.

FIG. 4. Stern–Volmer plots of synthetic systems. Ru(bpy)32þ in water and

Ru(bpy)32þ in glycerol. Symbols are actual data points and lines are the

modeled data. The numbers are the RLD Dt integration width for each curve.

APPLIED SPECTROSCOPY 439

anti-quenching curves. Again, experiment and theory agreewell.

Rapid Lifetime Determination with Nanoparticles.Nanoparticles find wide use in analytical applications,24 buthere we are only using them for unquenched luminescences.Having demonstrated that the simulations show good agree-ment with actual data, two components in a single solutionwere observed: Ru(dpp)3

2þ nanoparticles, and Ru(bpy)32þ in

water with a pre-exponential ratio of 2:5. A representativeSVQP is given in Fig. 6. The SVQP displays the unusualbimodal and anti-quenching behavior as seen previously.However, there is very poor agreement with the simulateddata using the parameters for the systems. The trends are thesame as in the simulations, but the severity of the anomalies isreduced for the nanoparticle mixtures. A possible explanationis as follows. Theoretically higher percentages of theunquenched component lessens distortions, and the polymerused exhibited minimal oxygen quenching. Thus, if Ru(bpy)3

2þ

associates with the nanoparticles and is less quenched, thecomposition of the system is no longer that assumed in themodel and the deviations from theory could be accounted for.

To test whether Ru(bpy)32þwas binding to the nanoparticles,

the Ru(bpy)32þ nanoparticle mixture was centrifuged to remove

the nanoparticles. Figure 7 shows the absorbance spectrum ofthe supernatant liquid compared to the original Ru(bpy)3

2þ.Although there was still some nanoparticle scattering thatdistorts and offsets the spectrum, there was a 30% reduction inabsorbance at the Ru(bpy)3

2þ peak (454 nm), indicating that asignificant amount of Ru(bpy)3

2þ associated with the nanopar-ticles and was removed with the nanoparticles upon centrifu-gation. Further, the resuspended nanoparticles now exhibited ashort-lived component that was absent in the originalnanoparticles. Table II gives a comparison between Ru(bpy)3

2þ

alone in solution (unassociated) and the short-lived component

FIG. 6. Quenching of mixed nanoparticle/Ru(bpy)32þ solution with various Dt

values with a pre-exponential weighting factor of 2:5 nanoparticles:Ru(bpy)3

2þ. The lines have no physical significance and are shown only forvisualization.

FIG. 7. UV-Vis of an original solution of Ru(bpy)32þ in water and of the

Ru(bpy)32þ solution recovered from the nanoparticles.

TABLE II. Lifetimes of the short-lived Ru(bpy)32þ component in

centrifuged nanoparticles (associated) and pure Ru(bpy)32þ in water.

O2 pressure (atm) Associated Ru(bpy)32þ (ns) Ru(bpy)3

2þ (ns)

0.00 799 6140.21 602 3941.00 476 168

FIG. 5. Stern–Volmer plots of synthetic systems. Ru(bpy)32þ in water and

Ru(dpp)32þ in glycerol. Symbols are actual data points and lines are the

modeled data. The numbers are the RLD Dt integration width for each curve.

FIG. 8. SVQP of Ru(bpy)32þ in water and lifetime associated with Ru(bpy)3

2þ

bound to the nanoparticle. Lines are shown for visualization.

440 Volume 63, Number 4, 2009

that we attribute to the Ru(bpy)32þ associating with the

resuspended nanoparticles. There is a definite increase inlifetime, from 614 ns to 799 ns for the unquenched system, thatwould be typical for unquenched Ru(bpy)3

2þ associated withthe very different polymer environment. Further, this compo-nent is much more poorly quenched, which is consistent withbinding to a protective environment that reduces quenching.Figure 8 shows the decreased quenching of this short-livedcomponent associated with interactions with an oxygen‘‘impermeable’’ polymer as compared with Ru(bpy)3

2þ insolution along with the heterogeneity of the bound material. Itis also possible that some of the quenching is due to re-dissolved Ru(bpy)3

2þ.

A possible mechanism for binding is electrostatic. Thenanoparticles were carboxylated and have a negative chargewhile Ru(bpy)3

2þ is positively charged. Strong electrostaticbinding would be expected under these conditions. We haveseen reduced quenching of ruthenium complexes electrostat-ically bound to anionic vesicles,25 which is consistent with ourcurrent results. We conclude that the failure of thenanoparticle data to match theory is due to association withthe nanoparticles, which invalidates our two-componentmodel.

Phase Shift Model. Rapid lifetime determination is not theonly data treatment method that gives bizarre SVQPs. Singlefrequency phase shift measurements (SFPS) can give verysimilar behavior. By varying the modulation frequencies weobtain SVQPs that also exhibit bimodal quenching as well asanti-quenching. Figure 9 shows Stern–Volmer quenching datafor simulations using s of 1 ls and 10 ls with Ksv values of 10atm�1 and 0 atm�1, respectively, with a ratio of contributions of1:1 and frequencies ranging from 80 kHz to 1.28 MHz.

General Comments. We have demonstrated that uniquebehavior can occur readily even with only two components. Inreality, systems exhibiting heterogeneity generally are morecomplex. Typically, they would consist of more components,and one could get unexpected behavior. Such behavior willmost likely occur if there is a long-lived unquenched or weaklyquenched component such as might arise with micro-crystallization. For example, our nanoparticles had threecomponent decays and showed pathological SVQPs. We havealso seen indications of such behavior in sensor films with

severe crystallization problems. Thus, the phenomena couldappear in a variety of micro-heterogeneous systems.

CONCLUSION

It has been shown here that when given systems that havenot been well defined it is very important to be aware thatbizarre behavior can arise when using fast and convenientapparent lifetime methods with quenchometric measurements.RLD as well as SFPS can both give unusual SVQPs in eventwo-component systems. These behaviors include bimodalquenching curves as well as ‘‘anti-quenching’’ curves. Thesephenomena are exacerbated by a small fraction of longunquenched components.

ACKNOWLEDGMENTS

We thank the National Science Foundation for support through NSF CHE04-100061. B. A. DeGraff is pleased to acknowledge the support of the DreyfusFoundation.

1. J. R. Lakowicz, ‘‘Emerging Biomedical Applications of Time-ResolvedFluorescence Spectroscopy’’, in Topics in Fluorescence Spectroscopy, J. R.Lakowicz, Ed. (Plenum Press, New York, 1994), vol. 4.

2. R. H. Engler, C. Klein, and O. Trinks, Meas. Sci. Technol. 11, 1077(2000).

3. Y. C. Clarke, W. Xu, J. N. Demas, and B. A. DeGraff, Anal. Chem. 72,3468 (2000).

4. E. R. Carraway, J. N. Demas, and J. R. Bacon, Anal. Chem. 63, 337(1991).

5. B. Higgins, B. A. DeGraff, and J. N. Demas, Inorg. Chem. 44, 6662(2005).

6. K. Takato, N. Gokan, and M. Daneko, J. Photochem. Photobiol., A 169,109 (2004).

7. J. T. Hupp, M. H. Keefe, and K. D. Benkstein, Coord. Chem. Rev. 205,201 (1999).

8. H. Rowe, Ph.D. Thesis, University of Virgina, Charlottesville, VA (2004).9. J. R. Bacon and J. N. Demas, Anal. Chem. 59, 2780 (1987).

10. J. L. Li, R. H. Schmehl, and W. S. Xia, Tetrahedron 56, 7045 (2000).11. M. Schaferling, M. Wu, J. Enderlein, H. Bauer, and O. S. Wolfbeis, Appl.

Spectrosc. 57, 1386 (2003).12. T. Q. Ni and L. A. Melton, Appl. Spectrosc. 47, 773 (1993).13. R. J. Woods, S. Scypinski, L. J. C. Love, and H. A. Ashworth, Anal.

Chem. 56, 1395 (1984).14. R. M. Ballew and J. N. Demas, Anal. Chem. 61, 30 (1989).15. R. M. Ballew and J. N. Demas, Anal. Chim. Acta 245, 121 (1991).16. K. K. Sharaman, A. Periasamy, H. Asheworth, J. N. Demas, and N. H.

Snow, Anal. Chem. 71, 947 (1999).17. S. P. Chan, Z. J. Fuller, J. N. Demas, and B. A. DeGraff, Anal. Chem. 73,

4486 (2001).18. P. A. Koen, B. P. Maliwal, and J. R. Lakowicz, US Patent 5,246,867

(1992).19. L. Tolosa, H. Szmacinski, G. Rao, and J. R. Lakowicz, Anal. Biochem.

250, 102 (1997).20. T. K. Stepinac, S. R. Chamot, E. Rungger-Brandle, P. Ferrez, J. Munoz, H.

van den Bergh, C. E. Riva, C. J. Pournaras, and G. A. Wagnieres, Invest.Opthalmol. Vis. Sci. 46, 956 (2005).

21. B. A. DeGraff and J. N. Demas, ‘‘Luminescence-Based Oxygen Sensors’’,Reviews in Fluorescence, C. Geddes and J. R. Lakowicz, Eds. (SpringerScience, New York, 2005), vol. 2, pp. 125–151.

22. C. T. Lin, W. Boettcher, M. Chou, C. Creutz, and N. J. Sutin, J. Am.Chem. Soc. 98, 6536 (1976).

23. J. R. Lakowicz, Principles of Fluorescence Spectroscopy (Plenum Press,New York, 1983), 1st ed., p. 80.

24. H. Xu, S. M. Buck, R. Kopelman, M. A. Philbert, M. Brasuel, E. Monson,C. Behrend, B. Ross, A. Rehemtulla, and Y.-E. Lee Koo, ‘‘FluorescentPEBBLE Nano-sensor and Nanoexplorers for Real-Time Intracellular andBiomedical Applications’’, in Topics in Fluorescence Spectroscopy, PartB, Advanced Concepts in Fluorescence Spectroscopy. MacromolecularSensing, C. Geddes and J. R. Lakowicz, Eds. (Spring Science and BusinessMedia, New York, 2005), vol. 10, Chap. 4, pp. 69–126.

25. A. Jain, W. Xu, J. N. Demas, and B. A. DeGraff, Inorg. Chem. 37, 1876(1998).

FIG. 9. Modeled effect of modulation frequency on SVQPs. Component oneis 1 ls with a Ksv of 10 atm�1 and component two is 10 ls with a Ksv of 0atm�1.

APPLIED SPECTROSCOPY 441