Efficient generation of singlet oxygen due to localized surface plasmon resonance of silver nanoparticles in rose bengal–silver nanoparticle composite films

The effect of silver nanoparticles (AgNP) on the generation of singlet oxygen (1O2*) using rose bengal (RB) as a photosensitizer for photodynamic therapy (PDT) was examined by measuring the phosphorescence spectra of 1O2* from RB-AgNP composite film and RB film on glass plates in CCl4. 1O2* was successfully detected by direct observation of the characteristic 1O2* phosphorescence around 1275 nm in two films. The excitation spectra of 1O2* in the two films were in agreement with the that of RB in film or in solution. The results indicate that the generation of 1O2* is ascribed to a PDT type Ⅱ reaction. The phosphorescence intensity of 1O2* in RB-AgNP composite film was larger than that in RB film. The enhancement of the phosphorescence intensity of 1O2* is most likely attributable to the strong electric fields due to the localized surface plasmon resonance of AgNP aggregates.

The enhancement of photocurrents in indium tin oxide (ITO) electrodes modified with dye [a porphyrin derivative, a phthalocyanine derivative, a tris(bipyridine)ruthenium(II) derivative, or a poly(3-hexylthiophene)] as a photosensitizer, was previously reported, and this result was obtained due to the LSPR-enhanced excitation using AuNPs, AgNPs, or gold nanorods (AuNRs) as NMNPs. [7][8][9][10][11][12][13][14][15][16][17][18][19][20] We also reported that the great enhancement of photocurrents in the composite films of a zinc-porphyrin (ZnP: donor)-viologen (V: acceptor) linked compound and AgNP or AuNP on ITO electrodes was caused by the combination of the photo-induced intramolecular electron transfer from ZnP to V and the LSPR-enhanced excitation of ZnP due to AgNP or AuNP. 13,15) Another interesting result found was the remarkable enhancement of photocurrents due to the combination of LSPR and magnetic field effects (MFEs) in the composite films. 18,19) Photon upconversion based on sensitized triplet-triplet annihilation (PUC-TTA) has attracted attention in the field of next-generation solar cells, such as dye-sensitized, organic, and perovskite solar cells and photocatalysis. [21][22][23][24][25][26] Previously, we found that the efficiency of PUC-TTA using platinum(II) octaethyl-porphyrin as a triplet photosensitizer and 9,10-diphenylanthracene as a fluorescent emitter was increased when using AgNP, AuNP, or triangular silver nanoplate (AgPL) as compared to not using them in solidliquid and solid systems. [27][28][29] Singlet exciton fission (SF) is a process in which a singlet molecular excited state splits into two triplet excited states. [30][31][32][33][34] Hence, the SF is expected to surmount the Shockley-Queisser efficiency limit for a solar cell containing a single optimized semiconductor junction. From this viewpoint, an external quantum efficiency above 100% in organic photovoltaic cells has been reported using SF in several papers. 33,34) We previously found the interesting result that the efficiency of the SF of rubrene (Rub) was enhanced in the presence of AgNP or AgPL with the aid of MFEs. 35,36) Reactive oxygen species (ROS) are chemically reactive molecules containing singlet oxygen ( 1 O 2 *) or superoxide (O 2 · − ). The high reactivity is of major importance in a variety of applications such as photodynamic therapy (PDT), water treatment, and catalytic oxidation. PDT is a therapeutic treatment using light and a photosensitizer, which generates cytotoxic substances by light irradiation. There are two types of reaction mechanisms during PDT, i.e. type I and type II reactions. [37][38][39] The singlet excited state of photosensitizers is generated from the ground state upon photo-irradiation. Then, the long-lived triplet excited state of photosensitizers is generated from the singlet excited state via an intersystem crossing process. The triplet excited state experiences electron-transfer reactions with a biological substrate/photosensitizer to generate radical anions, and further interacts with oxygen to produce oxygenated species such as O 2 · − and OH· radicals (PDT type I reaction). On the other hand, 1 O 2 * as a ROS is generated by triplet-triplet energy transfer from the triplet excited state of the photosensitizer to oxygen. In other words, 1 O 2 * plays a key role in the cytotoxicity of the PDT type II reaction.
The enhancement of 1 O 2 * generation in the PDT type II reaction is expected using NMNPs, similar to the applications for the performance improvement of photofunctional materials using the LSPR of NMNPs as described above. In fact, the enhancements of 1 O 2 * generation due to the LSPRenhanced excitation of NMNPs have been reported in many papers. [40][41][42][43][44][45][46][47] 1 O 2 * generation was investigated by an indirect method using fluorescent dye (1,3-diphenylisobenzofuran; DPBF etc.) in previous studies. [40][41][42][43][44][45][46][47] It is known that electric fields due to the LSPR of NMNP aggregate, as a result of LSPR coupling, are much stronger than those of NMNPs. Therefore, the efficient generation of 1 O 2 * is expected by using NMNP aggregates. However, the effects of the LSPR of NMNP aggregates on 1 O 2 * generation in PDT type II reactions have not been reported yet.
We previously reported the use of electrostatic layer-bylayer adsorption for fabricating multi-structures of AuNPs, AgNPs, and AuNRs in dye-NMNP composite films. [7][8][9][10][11][12][13][14][15][16][17][18][19][20]35) The layer-by-layer method is convenient, and sophisticated equipment such as vacuum systems is not required. It is easy to control the deposition density of charged NMNPs by changing the immersion time of the substrate in a corresponding colloidal solution. In other words, the ratio of NMNP and NMNP aggregate can be controlled by changing the immersion time.
The formation of new photosensitizer-NMNP composite systems requires optimal conditions such as spectral properties of the photosensitizer and NMNP and/or NMNP aggregates, and the photochemical reactions in complicated processes in the PDT type II reaction. Furthermore, many processes in the PDT type II reaction can be altered by NMNP and/or NMNP aggregates. As the sum of many effects of NMNP and/or NMNP aggregates, the 1 O 2 * generation in the PDT type II reaction is modified. However, in the indirect method, it is difficult to elucidate the effects of NMNP and/or NMNP aggregates on many processes in the PDT type II reaction.
The 1 O 2 * generated from the PDT type II reaction can be directly measured by its phosphorescence in the NIR rage. In addition, we can directly measure the phosphorescence of 1 O 2 * at the solid-liquid interface in a dye-NMNP substrate-CCl 4 system with a high degree of sensitivity because the lifetime of the phosphorescence of 1 O 2 * in CCl 4 is much longer compared with that of other solvents such as CHCl 3 , CH 3 CN, water, and so on. 48) The reaction mechanism in dye-NMNP composite films can be clarified by the comparison between the fluorescence and/or the phosphorescence spectra and the excitation spectra at the corresponding peaks of the phosphorescence and/or the fluorescence for 1 O 2 * and/or dye as a photosensitizer.
In this study, we prepared rose bengal (RB)-AgNP composite film and RB film using the layer-by-layer method, and directly measured the phosphorescence of 1 O 2 * at the solid-liquid interface in an RB-AgNP substrate-CCl 4 system. We elucidated the effects of AgNP and AgNP aggregates on many processes in the PDT type II reaction to find a very useful means of improving the efficiency of 1 O 2 *generation using AgNP aggregates. Fig. 1), silver nitrate (AgNO 3 , FUJIFILM Wako Pure Chemical Corporation), trisodium citrate dihydrate (FUJIFILM Wako Pure Chemical Corporation), poly(ethyleneimine) (PEI; Mw = 50,000-1100,000, FUJIFILM Wako Pure Chemical Corporation), poly(sodium-4-styrenesulfonate) (PSS; average Mw = ∼70,000, Sigma-Aldrich), and the other chemicals were used as received. Water was deionized using a Milli-Q system (Millipore).

RB (FUJIFILM Wako Pure Chemical Corporation) (
AgNPs capped with citrate were synthesized as previously reported. 49) Briefly, AgNO 3 (36 mg) was dissolved in H 2 O (200 ml). After refluxing, 4 ml of sodium citrate aqueous solution (10 wt%) was injected into the solution and then refluxed for 30 min to produce a AgNP aqueous colloidal   solution. The resultant AgNPs were capped with citrate ions, producing negatively charged surfaces. The extinction spectrum of AgNP in aqueous colloidal solution was measured [ Fig. 2(a)]. The peak of LSPR due to AgNP was observed at 426 nm. The structure of the AgNP was determined from transmission electron microscope (TEM) images. The average diameter of AgNPs was ca. 50 nm, as determined from a TEM image [ Fig. 2(b)]. The image was measured using a TEM (FEI Titan Themis) after drying the sample under a vacuum for more than one day.
The RB-AgNP and RB substrates were prepared by electrostatic layer-by-layer adsorption on the basis of the electrostatic interaction between the negatively charged AgNP capped with citrate, the negatively charged polymer, the positively charged polymer, and the positively charged glass as described below, as reported previously. [10][11][12][13][14] The RB-AgNP and RB substrates were prepared using the following procedure (Fig. 3).
The glass plates were made hydrophilic via treatment with a mixed solution (1:1) of aqueous hydrogen peroxide (31%) and ammonia (28%). First, the hydrophilic glass was immersed into an aqueous PEI solution (0.6 mM) containing 0.2 M NaCl for 20 min to produce a glass modified with PEI (PEI/glass).
This positively charged glass was then immersed into an aqueous colloidal solution of negatively charged AgNPs for 2 h to immobilize the AgNPs on the positively charged glass by electrostatic adsorption (denoted as AgNP/PEI/glass). We chose 2 h for the immersion time, because the maximum enhancement of photocurrent and/or fluorescence due to the LSPR of AgNP in porphyrin derivative-AgNP composite film occurred at an immersion time of 2 h. 11) The AgNPs on the back side of the AgNP/PEI/glass were wiped off. Next, the AgNP/PEI/glass was immersed into an aqueous PEI solution to produce the AgNP/PEI/glass modified with PEI (PEI/AgNP/PEI/glass), as described above. Then, the PEI/AgNP/PEI/glass was immersed into an aqueous PSS solution to produce the AgNP/PEI/glass modified with PSS (PSS/PEI/AgNP/PEI/glass). In addition, the PSS/PEI/AgNP/ PEI/glass was immersed into an aqueous PEI solution and then immersed into octanethiol for 3 h to produce PEI/PSS/ PEI/AgNP/PEI/glass. Finally, 1 μl of an acetone solution of RB (1 mM) was spin-coated onto the surface of PEI/PSS/PEI/ AgNP/PEI/glass at 1000 rpm for 10 s and then 1500 rpm for 10 s using a photoresist spinner (KYOWARIKEN K-359S-1) (RB/PEI/PSS/PEI/AgNP/PEI/glass: RB-AgNP substrate), as shown in Fig. 3. The layers of PEI/PSS/PEI were added to avoid the nonradiative metal quenching due to AgNP for fluorescence and/or phosphorescence of RB. Similarly, the PEI/glass(ref) and the RB/PEI/glass(ref) (RB substrate) were prepared as a reference (Fig. 3).
The absorption and extinction spectra of the solutions and/ or the substrates were measured by a Shimadzu UV-3150 spectrometer. The fluorescence and phosphorescence spectra of the solutions and/or the substrates were measured by a HORIBA Fluorolog-3-NIR spectrophotometer.

Results and discussion
3.1. Characterization of the RB-AgNP and RB substrates Two absorption peaks due to RB were observed at 525 and 565 nm in acetone solution of RB (Fig. 1). A fluorescence peak due to RB was observed at 582 nm and a fluorescence shoulder was observed at 525 nm in the acetone solution of RB (Fig. 1). Figure 4(a) shows the extinction spectra of the RB-AgNP and RB substrates, PEI/PSS/PEI/AgNP/PEI/glass, and PEI/ glass(ref). In PEI/PSS/PEI/AgNP/PEI/glass, the broad extinction band at approximately 330−500 nm is assignable to the LSPR band of isolated AgNPs or to the band of the transverse oscillation mode of coupled AgNPs. Also, the broad bands at approximately 500−800 nm are assignable to the LSPR band of AgNP aggregates as well as to the band for the longitudinal oscillation mode of coupled AgNPs, as reported previously. [10][11][12][13]18,27,35,50,51) The extinction bands for isolated AgNPs, AgNP aggregates, and RB were observed in the RB-AgNP substrate. As a result of comparison between the RB-AgNP substrate and PEI/PSS/PEI/AgNP/PEI/glass, the broad extinction band at approximately 400−470 nm is assignable to the LSPR band of isolated AgNPs or to the band of the transverse oscillation mode of coupled AgNPs in the RB-AgNP substrate. Also, the broad bands at approximately 500−800 nm are assignable to the LSPR band of AgNP aggregates as well as to the band for the longitudinal oscillation mode of coupled AgNPs in the RB-AgNP substrate. In the RB substrate, two absorption peaks due to RB were observed at 527 and 565 nm in CCl 4 [ Fig. 4(a)]. The two peaks were the same as those in the acetone solution of RB (Fig. 1). On the other hand, two absorption peaks due to the RB in RB-AgNP substrate were observed at 522 and 557 nm in CCl 4 in the extinction spectra of the RB-AgNP substrate. The two absorption peaks were different from those in the RB substrate and acetone solution of RB. The results and the comparison among three extinction spectra in Fig. 4(a) suggest that the extinction bands at 470-600 nm consist of those of RB and AgNP aggregates.
The absorption bands for RB were clearly observed in the extinction difference spectra between RB-AgNP substrate and PEI/PSS/PEI/AgNP/PEI/glass and between RB substrate and PEI/glass(ref). The extinction spectra in Fig. 4(b) suggest that the absorption for RB at 569 nm in RB-AgNP substrate is twice as large as that in RB substrate, and the extinction of AgNP aggregates at 569 nm is 1.6 times larger than that of RB in RB-AgNP substrate. In addition, the absorption spectrum of RB in RB-AgNP substrate is superimposed upon the extinction spectrum of LSPR due to the AgNP aggregates, but not the isolated AgNP, in the CCl 4 [ Fig. 4(b)]. The fluorescence spectrum of RB in acetone solution (Fig. 1) partially overlaps the extinction due to the LSPR band of AgNP aggregates in the RB-AgNP substrate [ Fig. 4(a)]. Therefore, the effect of LSPR due to the AgNP aggregates on the absorption process of RB is expected in RB-AgNP substrate in CCl 4 , similar to the porphyrin derivative-AgNP films, [10][11][12][13]18,27) Rub-AgNP film, 35) and AgPL polymer-composite films 36) in previous papers.

Phosphorescence spectra of 1 O 2 * of the RB-AgNP and the RB substrates
Emission and excitation spectra of the RB-AgNP and the RB substrates in CCl 4 were examined to elucidate the effect of AgNP on the generation of 1 O 2 * in the solid-liquid system. The samples for the emission and excitation spectra are shown in Fig. 5. The peaks of the emission bands of the RB-AgNP and the RB substrates in the NIR region are clearly observed at 1275 nm in CCl 4 (Fig. 6), because the lifetime of the phosphorescence of 1 O 2 * in CCl 4 is much longer compared with that in the other solvents such as CHCl 3 , CH 3 CN, water, and so on. 48) The emission bands are assignable to the phosphorescence of O 2 *. Therefore, it is clear that O 2 * is generated by the excitation at 567 nm upon the RB-AgNP and the RB substrates in CCl 4 . Next, the excitation spectra at 1275 nm for the phosphorescence of 1 O 2 * were measured in the RB-AgNP and RB substrates. The excitation spectra are shown in Fig. 7. The two excitation spectra are in good agreement with that for RB in the acetone solution (Fig. 1) and the substrates [ Fig. 4(a)]. Hence, the generation of 1 O 2 * is responsible for the photoexcitation of RB on glass at the solid-liquid interface of the RB-AgNP and RB substrates in CCl 4 . On the basis of these observations, the generation of 1 O 2 * in the RB-AgNP and RB substrates is explained in terms of the PDT type II reaction.
We compared the phosphorescence spectra for 1 O 2 * between the RB-AgNP and RB substrates to examine the  The Japan Society of Applied Physics by IOP Publishing Ltd phosphorescence intensity for 1 O 2 *. The phosphorescence intensity for 1 O 2 *in the RB-AgNP substrate was 6.6 times larger than that in the RB substrate (Fig. 6). In other words, the relative enhancement of phosphorescence intensity of 1 O 2 * at 1275 nm in the RB-AgNP substrate was estimated to be 6.6 as compared with that in the RB substrate. In addition, the excitation intensity at 1275 nm in the RB-AgNP substrate was 10.4 times larger than that in the RB substrate (Fig. 7). The relative enhancement of excitation intensity of RB at 569 nm in the RB-AgNP substrate was estimated to be 10.4 as compared with that in the RB substrate.
The emission spectra for the RB-AgNP and RB substrates excited at 527 nm in the visible region were measured in CCl 4 [ Figs. 8(a) and 8(b)]. The emission band at 560-600 nm is clearly observed and an emission peak was observed at 580 nm in the RB substrate [ Fig. 8(a)]. The emission band is assignable to the fluorescence for RB, because the emission band and the emission peak are similar to those in the fluorescence spectrum in acetone solution of RB (Fig. 1). A weak emission band at 600-800 nm is observed in Fig. 8(b). The weak emission band is attributable to the phosphorescence for RB. 52,53) The excitation spectra at 580 nm are in good agreement with the absorption spectra for RB in the acetone solution (Fig. 1) and the substrates (Fig. 4).
In contrast, a broad emission spectrum in the wider wavelength range (550-850 nm) was obtained, and the peak of the emission band was observed at 720 nm in the RB-AgNP substrate [ Fig. 8(b)]. The broad emission band above 600 nm is attributable to the phosphorescence for RB. Next, the phosphorescence intensity at 720 nm in the RB-AgNP substrate was 23.9 times larger than that in the RB substrate. The relative enhancement of phosphorescence intensity of RB at 720 nm in the RB-AgNP substrate was estimated to be 23.9 as compared with that in the RB substrate. In addition, the excitation intensity at 720 nm in the RB-AgNP substrate was larger than that in the RB substrate [Figs. 9(a) and 9(b)]. The relative excitation intensity of RB at 527 nm in the RB-AgNP substrate was estimated to be 28.5 as compared with that in the RB substrate [ Fig. 9(b)]. The enhancement of excitation intensity is similar to that of the phosphorescence intensity of RB [ Fig. 8(b)].
In addition, the shoulder of the emission band was observed at 580 nm in the RB-AgNP substrate [ Fig. 8(b)]. The result indicates that the broad emission spectrum consists of both the fluorescence and phosphorescence spectra of RB. In other words, the fluorescence spectrum for RB was superimposed upon the phosphorescence spectrum for RB in the broad emission spectrum [ Figs. 8(a) and 8(b)]. Similar to the effect of AgNP on the phosphorescence intensity of 1 O 2 *, it is suggested that the fluorescence intensity at 580 nm in the RB-AgNP substrate was 11.2 times larger than that in  The Japan Society of Applied Physics by IOP Publishing Ltd the RB substrate. The relative enhancement of fluorescence intensity of RB at 580 nm in the RB-AgNP substrate was estimated to be 11.2 as compared with that in the RB substrate. In addition, the excitation intensity at 580 nm in the RB-AgNP substrate was 12.0 times larger than that in the RB substrate. The relative enhancement of excitation intensity of RB at 580 nm in the RB-AgNP substrate was estimated to be 12.0 as compared with that in the RB substrate. The enhancement of the excitation intensity due to AgNP is similar to that of the fluorescence intensity of RB.
The enhancement of the phosphorescence intensity (23.9 times) of RB is much larger than that (6.6 times) of the phosphorescence intensity of 1 O 2 *. The result indicates that the yield and the lifetime of triplet RB probably decrease in the presence of AgNP, since the phosphorescence spectrum of RB overlaps with the extinction spectrum of LSPR due to AgNP aggregates. As a result, the enhancement (6.6 times) of the phosphorescence intensity of 1 O 2 * is much smaller than that (23.9 times) of the phosphorescence intensity of RB, since the efficiency of triplet-triplet energy transfer from 3 RB* to 3 O 2 decreases in the presence of AgNP. Similarly, the enhancement (6.6 times) of the phosphorescence intensity of 1 O 2 * is smaller than that (11.2 times) of the fluorescence intensity of RB. The result indicates that the yield and the lifetime of singlet RB probably decrease in the presence of AgNP, since the fluorescence spectrum of RB partly overlaps with the extinction spectrum of LSPR due to AgNP aggregates. The difference of enhancement between the phosphorescence and the fluorescence intensity of RB may be caused by nonradiative metal quenching and/or an increase in the radiative rate due to plasmon−exciton coupling of AgNP aggregates, since the phosphorescence spectrum of RB mostly overlaps with the extinction spectrum of LSPR due to AgNP aggregates as compared with the fluorescence spectrum of RB.
The enhancements of phosphorescence intensity for 1 O 2 *, and fluorescence and phosphorescence intensities for RB due to AgNP were estimated to be 6.6, 11.2, and 23.9 times, respectively. A large enhancement of the phosphorescence intensity for RB in the presence of AgNP was observed as compared with that of the phosphorescence intensity for 1 O 2 * or the fluorescence intensity for RB. This interesting result is different from those in the case of isolated core-shell silver nanoparticles. 42) In other words, the effect of AgNP aggregates on the generation of 1 O 2 *, 3 RB*, or 1 RB* is different from that of isolated AgNP.
The absorption for RB at 569 nm in RB-AgNP substrate is twice as large as that in RB substrate, and the extinction of AgNP aggregates at 569 nm is 1.6 times larger than that of RB in RB-AgNP substrate [ Fig. 4(b)], as described above. On the basis of the correction due to these results, the enhancement of phosphorescence intensity for 1 O 2 * in the presence of AgNP was estimated to be 5.7. The enhancement (5.7 times) is slightly smaller than the relative enhancement (6.6) of phosphorescence intensity of 1 O 2 * at 1275 nm (Fig. 6). The enhancement (5.7 times) in the presence of AgNP clearly indicates that 1 O 2 * generation increased in the presence of AgNP. It is possible to directly obtain the enhancement due to AgNP on the phosphorescence intensity of 1 O 2 * in the NIR range using the layer-by-layer method and solid-liquid interface in the RB-AgNP substrate-CCl 4 system. The enhancement of the phosphorescence intensity of 1 O 2 * in the presence of AgNP is most likely attributable to the large electric fields due to the LSPR of AgNP aggregates. In other words, the enhancement of 1 O 2 * generation in the presence of AgNP is caused by the LSPR-enhanced excitation due to the AgNP aggregates (Fig. 10), because the absorption spectrum of RB in RB-AgNP substrate is in good agreement with the extinction spectrum of LSPR due to the AgNP aggregates. The enhancements of phosphorescence intensity for 1 O 2 * and fluorescence and phosphorescence  The Japan Society of Applied Physics by IOP Publishing Ltd intensities for RB due to AgNP are also attributable to the large electric fields due to the LSPR of AgNP aggregates.
The above discussions strongly indicate that it is necessary for the efficient generation of 1 O 2 * to choose an appropriate combination of a photosensitizer and a metal nanostructure including a metal nanoparticle, since the optimum conditions of the effects of the metal nanostructure including AgNP aggregates on the three processes are different.

Conclusions
A large enhancement of the phosphorescence intensity of 1 O 2 * in RB-AgNP composite film was clearly observed compared with that in RB film. Similar enhancements of the fluorescence and phosphorescence intensities of RB in RB-AgNP composite film were observed as compared with those in RB film. On the basis of the comparison of these results, these enhancements are most likely attributable to LSPRenhanced excitation due to the strong electric fields of AgNP aggregates. The present study strongly indicates that the electrostatic layer-by-layer method and solid-liquid interface in the dye (photosensitizer)-NMNP substrate-CCl 4 system provide a very useful means of improving the efficiency of 1 O 2 * generation using strong electric fields due to the LSPR of NMNP aggregates, and also clarify the mechanism of LSPR of NMNP aggregates in PDT type II reactions. Further investigations on the effects of the size and density of NMNPs including AgNP on 1 O 2 * generation in RB-NMNP composite films due to the PDT type II reaction are now in progress.