Ag nanoplatelets as efficient photosensitizers for TiO₂ nanorods

Noble metal nanomaterials have attracted significant interest due to their catalytic, electronic, and optical properties.^1,2 An important application of these noble metal nanostructures is their usage as photosensitizers in photocatalytic and photovoltaic devices.^3–5 One such example is TiO₂ nanoparticles decorated with plasmonic noble metal nanostructures, which has been a system of high interest for developing the next generation of photo-sensitized nanodevices.^6–9 

The primary reasons that noble metals and alloys are recognized as superior options for photosensitizers include their high light absorption efficiency and their potential for efficient injection of excited electrons into the semiconductor. The localized surface plasmon resonances (SPRs) generated in metallic nanostructures can be tuned by altering their size and shape^2,10,11 and used to improve the efficiency of absorption at desired wavelengths and for stronger coupling to the semiconductor. With these advantages in mind, it is desirable to quantitatively characterize the photoelectron excitation and injection processes of the metal photosensitizers so as to maximize the number of excited electrons pumped into the semiconductor conduction band (CB).^6,12,13

The interactions at the interface of the metal photosensitizer and semiconductor particle play a critical role in affecting the efficiency of photosensitized functions.^6,14–16 As such, chemical and physical modifications of metal nanoparticle surfaces have been pursued as a means of improving photovoltaic and photocatalytic conversion efficiencies.¹⁵ One modification is related to the sensitizer nanoparticle size and shape. It was reported that the photon-to-electron conversion efficiency of Au nanoparticles on the TiO₂ surface scaled inversely with the metal particle size,¹⁷ presumably due to the size-dependent increase in the surface-to-volume ratio of the sensitizer. With a higher surface area, the metal SPR couples stronger with the semiconductor, which increases the injection efficiency. For example, Bach and co-workers showed that a 13.3% efficiency was determined for 5 nm Au particles, whereas the efficiency decreased to 3.3% for 40 nm particles.¹⁷ Similarly, Lian and co-workers reported that the quantum efficiency of plasmonic hot-electron transfer from Ag nanoparticles to anatase nanoporous TiO₂ films scaled inversely with particle size.¹⁸ 

As an alternative approach, it has been demonstrated that disk-shaped metal “nanoplatelets” exhibit enhanced photocatalytic efficiencies, presumably due to their increased surface area for interactions with the semiconductor surface.^19,20 As an example, gold nanoplatelets on graphene films displayed higher hydrogen production rates from water splitting under simulated sunlight irradiation.¹⁹ Similarly, we recently examined the photocatalysis efficiency of Ag nanoplatelets and nanoparticles as photosensitizers in the photodegradation reaction of methylene blue under LED illumination.²¹ Despite the fact that the excitation cross section for the nanoparticles was larger than that of the nanoplatelets, both systems exhibited comparable catalytic efficiencies under 470 and 590 nm illumination. Conversely, under 740 nm illumination, the nanoplatelets displayed a threefold increase in catalytic efficacy.²¹ Altogether, the size dependence of the nanoparticles and the comparison between nanoparticles and nanoplatelets highlight the influence of surface area of the metallic photosensitizer on the overall photosensitized function.

To provide a mechanistic understanding of the enhanced photocatalytic efficiency of Ag nanoplatelets as a photosensitizer, the dynamics of the electrons driving these processes at the metal/semiconductor interface need to be studied. Transient IR absorption spectroscopy has been shown to be effective in characterizing electron dynamics in semiconductor light-harvesting devices.^6,9,22 Briefly, hot electrons are initially generated in the photosensitizer through photoexcitation. Electrons can be injected into the CB of an adjacent semiconductor through different pathways, including plasmon-induced hot-electron transfer (PIHET)⁹ and plasmon-induced interfacial charge-transfer transition (PICTT).⁶ Electrons in the semiconductor CB can then be probed by their intra-band transitions in the mid-IR region.²² The rise in the transient IR absorption signal is due to the appearance of electrons in the semiconductor CB and can therefore be used for deducing the injection rate of excited electrons from the metal sensitizer into the semiconductor.^6,9,23

In this work, we examine (on an ultrafast timescale) the photo-induced electron injection from Ag nanoplatelets to TiO₂ through transient IR absorption. The deduced injection rate is then compared against previously reported values for different photosensitizer systems, including metals and dyes. Finally, we discuss the mechanisms of the injection of photoexcited electrons that drive the distinct photocatalytic behavior for Ag nanoplatelets and spherical metal nanoparticles on TiO₂ surfaces.

TiO₂ NRs were fabricated using the sol–gel synthesis procedures reproduced from Liu et al.²⁴ Ag–TiO₂ nanocomposites were prepared using a modified method from Su et al.²⁵ The details for TiO₂ NRs and Ag–TiO₂ nanocomposites synthesis can be found in our previous publication.²¹ Both TiO₂ NRs and Ag/TiO₂ nanocomposites were pressed into 100 µm thin slices for the transient IR absorption measurement.

The experimental setup was described previously.^21,26,27 Briefly, a regeneratively amplified Ti:Sapphire laser system (Quantronix) operating at 800 nm and 1 kHz repetition rate was used for measurements. The 800 nm output from the regenerative amplifier was split into two parts to generate pump and probe pulses. A smaller portion (40%) of the output was frequency doubled in a BBO crystal to generate pump pulses at 400 nm (200 fs FWHM, 3 µJ/pulse). The pump pulses were chopped by a New Focus Model 3501 Chopper operating at 500 Hz. The rest (60%) of the output was used to pump an IR optical parametric amplifier (Quantronix, Palitra-FS) to generate the 1900 cm⁻¹ mid-IR pulse (200 fs FWHM, 1.5 µJ/pulse). The IR signal was detected by a liquid-nitrogen cooled HgCdTe detector (J15D14-M204B-S01M-60-D31316, Judson Technologies) and then sent to a lock-in amplifier (SR830 DSP, Standard Research Systems). The digitized output was recorded by a home-made LabVIEW program. The instrument response function (IRF) of our system was experimentally determined by fitting light scattered from a solvent-only sample and was shown to be a Gaussian function with a FWHM of 300 ± 10 fs.²⁶ 

TEM images of the Ag–TiO₂ nanocomposites were obtained using a JEOL 2100F (JEOL USA, Inc.) microscope operating at 200 keV. Samples were prepared via the nanoparticle suspension technique by dispersing the sample powders in ethanol and casting on holey carbon support films (TED PELLA, INC.), which was then dried overnight prior to analysis.

TEM images of the TiO₂ NRs following Ag nanoplatelet deposition are shown in Fig. 1. The average length and diameter of the TiO₂ NRs were measured to be 10 ± 5 µm and 688 ± 32 nm, respectively. The geometry of the Ag nanoplatelets is best described as that of an equilateral triangle. As shown in Fig. 1, our sample consisted of a broad distribution of nanoplatelets with different side lengths but a uniform thickness of 15 ± 5 nm. Image analysis was performed in order to extract the distribution of nanoplatelet side lengths present in the sample (Fig. 2). The distribution is well described by a log-normal distribution centered at 24.3 nm, with a FWHM spanning from roughly 10–50 nm.

To compare the surface area of nanoparticles vs nanoplatelets, we calculated the surface area of these nano-objects as a function of volume.²¹ As depicted in Fig. 3(a), it is found that the surface area of a nanoplatelet (red line) is generally more than 35% higher than that of a nanoparticle (black line) of the same volume. In addition, because of the shape, nanoplatelets have a much larger surface area in direct contact with the semiconductor substrate than spherical nanoparticles. Both factors are in favor of stronger coupling and therefore more efficient injection of photoexcited electrons from nanoplatelets than from nanoparticles.

The UV−vis absorption spectra of the bare TiO₂ nanorods, the Ag nanoplatelet covered TiO₂ nanorods, and the difference between the two samples (i.e., corresponding to the Ag nanoplatelets alone) are shown in Fig. 4. A wide SPR band centered at 450 nm is observed for the Ag–TiO₂ sample, which broadens to the near-IR region. As noted previously, SPR peak broadening is evidence for strong metal–semiconductor coupling at the interface as described by chemical interface damping.^6,28 However, the width of the UV–Vis absorption (Fig. 4) is also invariably influenced by the distribution of Ag nanoplatelet sizes present in the sample (Fig. 2). In order to estimate the contribution of metal–semiconductor coupling on the absorption broadening (i.e., deduced as homogeneous broadening), we fit the difference spectrum (Fig. 4 inset) to a Voigt function. It was revealed that the band was primarily Lorentzian in shape (i.e., Lorentzian to Gaussian ratio of roughly 7), suggesting that metal–semiconductor coupling is influencing the width of the absorption band.

A representative transient IR absorption trace of the Ag–TiO₂ nanocomposite is shown in Fig. 5 (red circles). No transient IR absorption was detected for untreated TiO₂ nanorods (blue circles). This comparison shows that the 400 nm pump does not directly excite electrons of the semiconductor, which is consistent with the UV–vis spectrum (Fig. 4). Additionally, IR spectra taken from colloidal Ag platelets in aqueous solution revealed no resonances near 1900 cm⁻¹. This observation is consistent with the previously reported transient IR absorption of Ag nanoparticle systems for which no IR absorption was observed for hot electrons in Ag.¹⁸ 

Overall, two critical factors determine the amount of photoexcited electrons that are injected into the semiconductor CB from the photosensitizer: (1) the number of electrons in the sensitizer that are excited above the CB band edge and (2) the injection efficiency. The latter is determined as the ratio of the injection rate over the total decay, including electron relaxation within the photosensitizer and the electron injection rate into the semiconductor. In our transient absorption experiments, the IR pulse is only sensitive to electrons in the TiO₂ conduction band. Although the hot electron relaxation rate within Ag is not known, this rate primarily affects the magnitude of the transient IR signal. The electron injection rate can be directly deduced from the rise of the transient signal.⁹ 

The transient absorption signal shown in Fig. 5 was analyzed using a nonlinear least squares fitting based upon a phenomenological model similar to the one used by Anderson and Lian,²² in which electron injection was modeled as an exponential rise convoluted with the IRF. As noted earlier, the IRF of our system was experimentally shown to consist of a Gaussian function with a FWHM of 300 ± 10 fs.²⁵ In order to account for the uncertainty of the IRF, the fit analysis was repeated for three separate instances, in which the FWHM of the IRF was set as either 290, 300, or 310 fs. The resulting rise times for these three fits were deduced as 11.8 ± 0.6, 12.8 ± 0.7, and 14.7 ± 0.8 fs, respectively. The chi-squared values for each of the fits remained reasonably constant at χ² = 0.11. Consequently, despite the comparatively broad (±10 fs) uncertainty in the measured IRF, this had only a relatively minor influence on the deduced value of the rise time. The average electron injection time (and uncertainty) was then determined as follows: Each of the three fit deduced rise times was expanded into two values corresponding to the deduced value ± the uncertainty of the fit, yielding a series of six rise times (i.e., 11.2, 12.4, 12.1, 13.5, 13.9, and 15.5 fs). The electron injection time (and corresponding uncertainty) was then determined as the average and standard deviation of this set of rise times. Based upon this analysis, the electron injection time for our Ag nanoplatelets to TiO₂ NRs was determined to be 13.1 ± 1.5 fs.

The photocatalytic efficiency of the nanoplatelets examined here has previously been shown to be superior to nanoparticles whose diameters range from 40 to 70 nm.²¹ We have shown that the electron injection time from Ag nanoplatelets to TiO₂ is comparable to those from 2.5 nm wide nanoparticles. A reasonable explanation for this effect is that the electron injection time from the nanoplatelets is much faster than those of nanoparticles roughly 50 nm in diameter, which results in a much higher efficiency of producing hot electrons in the semiconductor CB to induce catalytic reactions.

To compare electron injection efficiency from different metal sensitizers to TiO₂ semiconductors, we have tabulated the fastest reported injection time constants for Ag in Table I.^18,29 Of significance, electron injection from Au to TiO₂ has been consistently reported as ∼50 fs or longer^9,30,31 or roughly four times slower compared to electron injection from Ag to TiO₂. We note that the Au disk injection time constant is comparable to nanoparticles of much smaller volume (roughly 140 times smaller), which further highlights the likely importance of the sensitizer surface area in contact with the semiconductor. A faster injection time means that a larger fraction of excited electrons may inject into the semiconductor in competition with the relaxation processes within the metal nanoparticle. This comparison clearly indicates that Ag is far more efficient than Au at injecting hot electrons into TiO₂.

To further compare these two metals, we need to examine the respective optical transitions for generating excited electrons that are sufficiently energetic for injection into the semiconductor. Hot electrons generated in the metal with sufficient energies above the Schottky Barrier (E_SB), between the metal and semiconductor interface, transfer into the adjacent semiconductor. The highest E_SB between Ag and TiO₂ is around 0.7 eV,⁴ while this barrier for Au/TiO₂ is reported to be around 1 eV.³² Hot electrons in Ag and Au are generated through intra-band and inter-band transitions, respectively. The highest energy of the Ag d-band was reported to be 3.7 eV below the Fermi level.⁴ As the pump photon energy in our experiment was only 3.1 eV, the hot electrons generated in Ag are all from transitions within the s band. The most energetic electrons are therefore estimated to be 3.1 eV above the Fermi Energy of Ag. In contrast, the d band of Au lies roughly 2.3 eV below its Fermi level.⁴ In the studies referred here, photons excited the 2.3 eV SPR of Au.⁴ Consequently, the hot electrons generated in Au are primarily from the d-to-s inter-band transition and their energy is close to the Fermi energy of Au and therefore below the conduction band energy of TiO₂. Examination of the absorption cross sections for the respective optical transitions in the two noble metal nanoparticles reveals that for excitations in the energy regions high enough to induce electron injection, Ag is at least one order of magnitude higher than Au [see, for example, Fig. 1(c) of Ref. 33 comparing 5 nm particles of Ag and Au].³³ Therefore, visible-light-induced inter-band transitions limit the production of highly energetic electrons in Au,³⁴ making it a less efficient hot-electron injector material compared to Ag.

The PIHET mechanism was previously introduced to describe electron injection from photosensitizers to semiconductors as two weakly coupled systems.³⁵ For stronger coupling between the metal and semiconductor (e.g., CdSe–Au NR),⁶ the PICTT mechanism has been proposed as a more efficient model. In the strong coupling limit, the excited states of the metal–semiconductor complex consist of strong semiconductor CB character. Photoexcitation can therefore be viewed as direct creation of a hole in the metal and the electron in the semiconductor CB. In Lian’s previous report, both PIHET and PICTT contributed to the charge carrier transfer efficiency, which increases as the Ag nanoparticle size decreased.¹⁸ Based on their measurements, the PIHET effect can be neglected when the Ag nanoparticle size is larger than 8 nm. On average, the length of our Ag nanoplatelets ranges from 10 to 50 nm (centered near 25 nm, Fig. 2). Even though this size is above the 8 nm limit, the surface area for the platelet shape in ratio to the bulk is far closer to a smaller nanoparticle. Figure 3(b) shows that the surface to bulk ratios of the Ag platelets in our sample are comparable to spherical nanoparticles whose diameters are smaller than the lengths of the nanoplatelets. Moreover, this relative difference increases as the nanoplates become wider (i.e., a 10 nm wide nanoplate has a surface to bulk ratio comparable to a 7.3 nm nanoparticle, whereas a 50 nm wide nanoplate is comparable to a 22 nm nanoparticle). Still, a major difference in terms of interacting with the TiO₂ NR between the platelet and the spherical nanoparticle is the amount of surface area in direct contact with the semiconductor. In the case of the platelet, one side of the flat surface may be in direct contact, whereas the spherical particle has only a point contact with the semiconductor. Although it is difficult to quantify the effect of the amount of surface area in direct contact, considering the injection time constants, it appears that the combination of the large surface to bulk ratio and the large surface area in direct contact with the semiconductor make platelets just as efficient at electron injection as small nanoparticles. Consequently, the Ag/TiO₂ system examined here is likely dominated by the PICTT mechanism in the strong coupling limit. This conjecture is also supported by the observation of the broad SPR band shown in Fig. 4, which indicates the influence of strong coupling in broadening the absorption bandwidth.

Finally, we compare Ag nanostructures with non-metal sensitizers. Beyond metals, dye molecules have been extensively used as photosensitizers. In general, molar extinction coefficients for dyes tend to be many orders of magnitude smaller compared to metal nanoparticles, ∼10⁴–10⁵ vs 10⁷–10⁹ M⁻¹ cm⁻¹, respectively.^36,37 Prior reports have measured (through transient mid-IR absorption spectroscopy) electron injection time constants (from dyes to TiO₂) spanning 50–500 fs,^26,38,39 which is more than an order of magnitude slower compared to the electron injection time constant reported here. Consequently, noble metal nanoparticles appear to be far more efficient photosensitizers compared to dye molecules.

In summary, we demonstrate that Ag nanoplatelets are a preferred photosensitizer for providing excited electrons to TiO₂ nanoparticles/rods for photocatalytic and photovoltaic functions. It has previously been shown that TiO₂ nanorods display higher photocatalytic efficiency with Ag nanoplatelets as the photosensitizer when compared against spherical metal nanoparticles of much smaller sizes,²¹ although it was also shown that smaller photosensitizers show faster electron injection time¹⁸ and higher photocatalytic efficiency.¹⁷ Here, we report the injection lifetime of electrons photoexcited by 400 nm light in Ag nanoplatelets to adjacent TiO₂ nanorods measured by femtosecond transient IR absorption spectroscopy. The injection lifetime of 13.1 ± 1.5 fs is comparable to those measured for much smaller Ag nanoparticles. This super-fast injection rate is a manifestation of the strong coupling between Ag and the TiO₂ conduction band, which also resulted in the broadening of the Ag platelet SPR band. The strong coupling can be understood from the fact that the geometry of the platelets affords a much larger surface to bulk ratio and a larger surface area in direct contact with the semiconductor. This injection lifetime is roughly four times faster than previously reported electron injection lifetimes for similar systems using Au as the photosensitizer and an order of magnitude faster than those using dye molecules as photosensitizers. The fast injection rate combined with the fact that Ag is a stronger absorber than other known photosensitizers suggests that Ag nanoplates are superior photosensitizers for generating electrons in the semiconductor conduction band for photocatalytic and photovoltaic applications.

This research was funded by the Department of Army Basic Research Program through the Edgewood Chemical and Biological Center, U.S. Army Research Office (Grant No. W911NF-15-2-0052).

The authors have no conflicts to disclose.

The data that support the findings of this study are available from the corresponding author upon reasonable request.