The X-ray diffractogram of MAPbBr3/GaAs samples is shown in Fig. 1. The range of 2θ is from 10 to 50° which includes GaAs (002) peak appeared at 31.6°. The other peaks are identified by the database and published literature as a perovskite structure of MAPbBr3 [37,38,39]. Overall, the XRD patterns of the MAPbBr3 films indicate good crystallinity albeit the fabrication of MAPbBr3 was performed under ambient conditions. Figure 1 shows that the MAPbBr3 films are polycrystals with the dominant peaks associated with (100), (200), and (300) at 14.6°, 30.1°, and 45.9° respectively. The good crystal quality may be due to anisotropic crystal formation, that allows MAPbBr3 crystals to grow from GaAs surface vertically. The Miller indices for those peaks are in the same family suggesting identical crystal planes that are perpendicular to the x-axis in Cartesian coordinates with different d-spacing values. In other words, the MAPbBr3 crystal planes are identical i.e., parallel to the surface of the GaAs substrates, because the θ-2θ scan detects every crystal plane that is perpendicular to the 2θ plane (scattering plane). Since the crystal planes are identical, the mean crystallite sizes are theoretically identical. The sizes can be estimated by Scherrer’s equation shown in the following:

$$D = \frac \cos \theta }}$$

(1)

where D is the mean crystallite size, K is the shape factor that usually is equal to 0.9 or 0.94 [40, 41], λ is the X-ray wavelength used in the measurement (1.541 Å), \(\beta_\) is the full width at half maximum of any considering peak at a specified Miller index (hkl), and θ is the diffraction angle. The equation revealed that the mean crystallite size was approximately 150 nm estimated from the full width at half maximum at (100), (200), and (300) peaks. These three peaks are selected because they are dominant. All X-ray diffraction parameters including full width at half maxima and crystallite sizes of every identifiable peak are provided in Table S1 of Supplementary Information. As we see, the MAPbBr3 films are polycrystal films with good crystal quality. The good crystal quality might facilitate our consideration of interface states because we can ignore the states that are contributed by grain boundaries.

The X-ray diffractogram of the MAPbBr3 films deposited on i-GaAs (001) substrates. The diffraction peak at 31.6° is the only peak associated with GaAs. The numbers indicated in parentheses are Miller indices for crystals in real space

The blue data points in Fig. 2a show a room-temperature PL spectrum of the MAPbBr3 layer on i-GaAs (001) excited by the laser diode with an excitation photon wavelength λex = 484 nm. The asymmetric line shape of the PL spectrum indicates that this spectrum consists of several peaks. We consider that there are two dominant peaks. Since the utilization of Gaussian functions to deconvolute the measured PL spectra has been widely conducted [38, 42,43,44,45], two Gaussian functions were used for the deconvolution procedure. Therefore, the following Gaussian function is used.

$$I = \mathop \sum \limits_^ a_ \exp \left( \frac E_ } \right)^ }}^ }}} \right)$$

(2)

where I is the PL intensity, ai is the amplitude of the Gaussian function related to the probability of the recombination, E is the emitted photon energy, Ei is the center of the Gaussian function in energy space, and σi is the PL peak widths. The subscript i is the assigned numbers to recombination processes, i.e. i = 1 and i = 2 are the band-to-band recombination of free photoexcited carriers and the localized-exciton-influenced near-band-edge recombination respectively. The localized-exciton-influenced near-band-edge recombination (hereafter, the excitonic recombination) is usual in MAPbBr3 because MAPbBr3 exhibits high excitonic absorption and binding energy. Photoexcited electrons, excited by the above-bandgap, photons can relax to the CB minimum of MAPbBr3. There, the band-to-band recombination occurs if the electrons directly recombine with holes in the VB. However, the attractive Coulombic force can create binding electron–hole pairs forming excitons. This causes the electrons to recombine with the holes from slightly lower states compared to the CB minimum. Hence, the excitonic recombination exhibits PL peaks at relatively lower energy [43, 45, 46]. Unlike Br vacancy states that tend to appear at the surface and interfaces, the Br vacancy states associate with lower energy, i.e. approximately lower than 2.20 eV (~ 560 nm) [38, 47, 48], compared to the PL peaks corresponding to the excitonic recombination (~ 2.25 eV or ~ 550 nm) [43, 45, 46].

a PL spectrum of the MAPbBr3 layer on i-GaAs (001) under 484-nm excitation. The solid orange curve is the result of fitting two Gaussian functions (the pink and purple curves) to the data. The pink and purple curves correspond to band-to-band recombination (Rec.) and excitonic recombination, respectively. b PL spectrum of MAPbBr3 deposited on glass substrates visualized with absorption spectrum of MAPbBr3. The inset figure is an estimation of the optical bandgap using Tauc plot (colour figure online)

The orange curve in Fig. 2a is the result of fitting the two Gaussian functions to the experimental data. The fitting provides the R-square value of 0.9992 which points to a good fitting. Certainly, adding an additional Gaussian term in the fitting process can improve fitting quality, i.e. the R-square is 0.9999. However, the third Gaussian term (i = 3) added in the fitting appears as a broad peak that cannot be assigned to any possible radiative recombination (see Fig. S4 of Supplementary Information). The board peak may correspond to the recombination via Br vacancy states. However, the peak is too broad providing implausibility to be associated with the Br vacancy states. The pink curve (Fig. 2a) is the first Gaussian function (i = 1) with a peak at about 535 nm (2.31 eV), which is attributed to band-to-band recombination in MAPbBr3 [37, 49, 50]. The difference between the fitted peak position and the various reported values for the Eg of \(}_}\) is within ± 0.05 eV. The second Gaussian function (i = 2) shown by the purple curve can be attributed to the excitonic recombination. Furthermore, we find that both PL peaks are relatively broad, which is due to inhomogeneous broadening. This means the bandgap and also the excitonic states exhibit a large energetic distribution.

Figure 2b shows the PL spectrum (blue dots) and the absorption spectrum (dark red dots) of the MAPbBr3 samples deposited on glass substrates. For the PL measurement, the excitation wavelength was 484 nm. The sharp absorption edge at ~ 2.30 eV indicates that the fabricated MAPbBr3 films exhibit good optical properties, which is consistent with the observation in the X-ray diffractogram shown in Fig. 1. A spike-like excitonic absorption peak appears at around 2.30 eV, which is the same as the reported values elsewhere [51, 52]. The Stokes shift was observed, which is moderate as compared to values in previous publications [45, 46, 50]. The inset figure in Fig. 2b is the Tauc plot for the absorption spectrum. The black dashed line in the inset figure is the linear fitting to the (αhν)2 data. The obtained optical energy bandgap, Eg = 2.29 eV is indicated by a black arrow pointing to the x-axis [49, 53, 54]. As aforementioned, excitons contributed to the absorption and the PL spectra. Therefore, the Tauc plot is only for an approximation of the optical Eg because the excitonic influence might cause an underestimated Eg. Although the Eg can be underestimated, the obtained Eg from the Tauc plot is consistent with the deconvoluted Gaussian function attributed to the band-to-band recombination (within 2.31 ± 0.05 eV).

The data points in Fig. 3 show comparisons of the normalized PL spectra of the MAPbBr3 layer on i-GaAs (001) (Fig. 3a and b) and the MAPbBr3 layer on glass substrates (Fig. 3c and d) excited by the Ti:Sapphire laser at 800 nm (1.55 eV) (Fig. 3a and c) and 900 nm (1.37 eV) (Fig. 3b and d), respectively. The photo of the green emission is provided in Fig. S5 of Supplementary Information. The PL results in Fig. 3 were measured under the same excitation energy density of 27 nJ/cm2. These photon energies are smaller than the Eg of MAPbBr3, but nevertheless the green emission caused by the recombination in MAPbBr3 can be easily confirmed and the line shape of the PL spectrum is similar to that observed for λex = 484 nm. In the PL measurements, we observed unreasonable peak positions and spectral shapes that deviated from the Gaussian function, suggesting poor film uniformity, when the samples were irradiated near the edges or the corners. However, this feature suggesting poor uniformity was hardly observed when the samples were irradiated at the center area because of the relatively higher homogeneous size distribution of MAPbBr3 crystals and better film uniformity. The PL emission of MAPbBr3 is attributed to the two-photon absorption for the MAPbBr3/Glass systems. Differently, the MAPbBr3/GaAs heterointerface can allow the TPU because of the existence of interface states and the possibility of carrier transport through the heterointerface. Therefore, the PL spectra observed in the MAPbBr3/GaAs systems cannot be merely attributed to the two-photon absorption. The 800-nm pulses can directly excite the GaAs substrate (the Eg of GaAs at room temperature is 1.42 eV) [9, 11,12,13,14]. Ideally, the 900-nm pulses cannot directly excite GaAs because the Ti:Sapphire laser generates a 900-nm beam with a bandwidth of ~ 20 nm which does not cover the GaAs energy bandgap corresponding wavelength (~ 870 nm). Moreover, the linear absorption coefficient (α) of GaAs at 900 nm is approximately α900 = 26.6 cm–1 [55, 56]. This suggests that the α900 of GaAs can be neglected. However, the possibility of direct photoexcitation in GaAs by the 900-nm laser beam should not be completely denied because the laser beam can contain slightly shorter wavelength photons (~ 880–900 nm) where the linear absorption coefficients are higher than α900. Similar to Fig. 2, we deconvoluted the PL spectra in Fig. 3 using two Gaussian functions. The results obtained using least squares fitting are shown by the orange curves for the PL data measured under 800-nm photoexcitation (Fig. 3a and c). The light blue curves, in Fig. 3b and d, are the fitting results for the PL spectra measured under 900-nm photoexcitation. The fittings show that the corresponding R-square values larger than 0.995 are good indicators for good fitting results. The fitted curves reveal that the amplitude of the Gaussian functions attributed to band-to-band recombination (the pink curves) for MAPbBr3 deposited on i-GaAs is stronger than that of MAPbBr3 deposited on glass substrates. In contrast, the amplitude of the Gaussian functions attributed to the excitonic recombination (the purple curves) shows weaker for the MAPbBr3/GaAs compared to the MAPbBr3/Glass. This is obvious evidence suggesting that there are interactions between the substrates to MAPbBr3 layers influencing the carrier dynamics or the optical recombination processes. The weak deconvoluted PL peaks associated with the excitonic recombination for the MAPbBr3/GaAs inform that the exciton formation is significantly reduced. The phenomenon can be interpreted as an influence of the interface states at the MAPbBr3/GaAs heterointerface. The photogenerated carriers (either in MAPbBr3 or i-GaAs) can occupy the interface states and recombine non-radiatively which can cause an increasing exciton dissociation rate. Moreover, the relatively high dielectric constant of i-GaAs (εGaAs = 12.9), compared to glass (εGlass = 4.6) [57, 58], allows significant carrier transport in the MAPbBr3/GaAs system, i.e. from GaAs to MAPbBr3 or inverse. The phenomenon can also cause the exciton dissociation. Therefore, the deconvoluted PL peaks for the excitonic recombination are observed weaker for the MAPbBr3/GaAs system than the ones for MAPbBr3/Glass. Based on the discussion, we can conclude that the carrier transport through the MAPbBr3/GaAs heterointerface and the occupation of photoexcited carriers on interface states at the heterointerface is possible. This can be a good merit for the TPU.

a PL spectrum of the MAPbBr3 layer on i-GaAs (001) under 800-nm excitation at energy density of 27 nJ/cm2. The blue points are the experimental data and the solid orange curve is the fitting result. b The PL spectrum for λex = 900 nm. The dark red points are the experimental data and the light blue curve is the fitting result. c PL spectrum of MAPbBr3 deposited on glass measured under 800-nm excitation, and (d) PL measured under 900-nm excitation. In all panels, the pink and purple curves correspond to band-to-band recombination and excitonic recombination, respectively

Considering the ratio of peak intensities of the band-to-band recombination to the excitonic recombination, the ratios are 3.21 and 3.38 for the MAPbBr3/GaAs systems measured under 800-nm and 900-nm photoexcitations respectively (for Fig. 3a and b), and 0.34 and 0.46 for the MAPbBr3/Glass systems measured under 800-nm and 900-nm photoexcitations respectively (for Fig. 3c and d). The ratio data is shown in Table 1 with different substrate types, excitation wavelengths, and excitation energy densities. The data in Table 1 shows that the photoexcited electrons in the MAPbBr3/Glass systems, excited by sub-bandgap photons (λex = 800 nm and λex = 900 nm in our case), tend to relax from the CB of MAPbBr3 to the CB minimum and form localized excitons with holes in the VB regardless of the excitation energy density. Therefore, the ratio is less than unity. In contrast to the results of the MAPbBr3/GaAs system, the ratios appear to be greater than unity which we attribute the results to the influence of interface states at the MAPbBr3/GaAs heterointerface and the significant difference in dielectric constants. However, with increasing excitation energy densities from 27 nJ/cm2 to 55 nJ/cm2, an identical tendency can be observed in both MAPbBr3/GaAs and MAPbBr3/Glass, i.e. the ratio significantly decreases with increasing energy densities. Interestingly, the changes in the ratio seem insignificant for the MAPbBr3/Glass systems (within 0.1). The MAPbBr3/GaAs systems, on the other hand, show much greater changes, i.e. the changes are about one order difference. The drastic changes in the ratio for the MAPbBr3/GaAs systems, as the excitation energy density increases, suggest that the photoexcited carriers experience different mechanisms prior to the recombination. For the MAPbBr3/GaAs systems, the TPU is possible if electrons in GaAs are excited by 800-nm (direct excitation) or 900-nm photons (partially direct or two-photon excitation) and the photoexcited electrons and holes are transported through the MAPbBr3/GaAs heterointerface. In order to achieve the intraband transitions, the photoexcited carriers are required to occupy interface states because the interface states are discrete, and therefore, the interface states can likely function as quantized states which can allow the intraband transitions. At 27 nJ/cm2 for 800-nm and 900-nm photoexcitations, the high ratio values indicate that the excitonic recombination is suppressed. This is due to the fact that the presence of interface states at MAPbBr3/GaAs heterointerface can allow photoexcited carriers to occupy, and later, recombine non-radiatively. Therefore, the excitons simply dissociate at the heterointerface [42]. Hence, it is plausible that the TPU is achieved at the MAPbBr3/GaAs heterointerface when the MAPbBr3/GaAs samples are excited by low energy densities of 800-nm and 900-nm photons. In contrast to the 55 nJ/cm2 case, the drastic decrease in the ratio informs that the excitonic recombination significantly increases. This can be interpreted as follows: The photoexcited carriers excited by low energy densities of 800-nm and 900-nm photons start occupying the interface states. At this stage, the TPU is achieved. With increasing excitation energy densities, the interface states are filled reducing the possibility of the intraband transitions. This means that the TPU is suppressed. In addition, the state-filling effect creates a neutral zone reducing the exciton dissociation [59], and hence, the excitonic recombination increases. Therefore, at high excitation energy densities both for 800-nm and 900-nm photoexcitation conditions, the carrier dynamics switch from the TPU to the two-photon absorption. Due to the neutralization of interface states, the exciton density increases with excitation energy densities where the two-photon absorption dominates the carrier dynamics in the MAPbBr3/GaAs systems. We provide evidence supporting this idea in the next sections.

While the TPU can be achieved in the MAPbBr3/GaAs systems by low excitation energy densities, the TPU is impossible for the MAPbBr3/Glass systems. Therefore, the two-photon absorption is the only carrier dynamics that plays a role in the PL emission of MAPbBr3/Glass. It is known that the two-photon absorption can suppress the excitonic recombination because the photogenerated carriers, by the two-photon absorption, tend to occupy above-band edge states. Therefore, slight changes in the ratios in Table 1 might be due to the reduction of carrier relaxation probability. This suggests that the carrier dynamics, in the MAPbBr3/Glass systems, are independent of the excitation energy densities.

The explanation, in the previous section, is supported by the TRPL data in Fig. 4. Figure 4a and b show the TRPL of MAPbBr3/GaAs samples measured under 800-nm and 400-nm photoexcitations, respectively. In Fig. 4a, 800-nm photons induce the TPU occurring at the MAPbBr3/GaAs heterointerface and the two-photon absorption occurring in the MAPbBr3 layer. Unlike 800-nm photons, 400-nm photons directly excite MAPbBr3 because about 99.9% of 400-nm photons are absorbed in 700-nm-thick MAPbBr3. The detection wavelength range, set by the streak camera, was between 520 and 560 nm which includes the emissions from free carriers and excitons (band-to-band and excitonic recombinations). Since the experimental data show an exponential decay behavior, we used the following fitting function:

$$I = \mathop \sum \limits_^ A_ \exp \left( \frac }}} \right)$$

(3)

where Aj represents the amplitude of each decay component, and τj is the corresponding time constant (carrier lifetime) for recombination. The subscript j corresponds to the recombination processes, i.e. the excitonic recombination (j = 1), the band-to-band recombination (j = 2), and non-radiative recombination (j = 3). All estimated parameters from Eq. (3) are provided in Table 2. The slow component corresponds to the excitonic recombination while the moderate component is for the band-to-band recombination. The fast component is for the non-radiative process. The assignment of the fast decay component to the non-radiative recombination is true only if the non-radiative process depends on the emission intensity. Figure 4a shows the decay profile of MAPbBr3/GaAs samples measured under 800 nm with excitation energy densities u = 27 nJ/cm2 and u = 48 nJ/cm2. The decay profile for u = 27 nJ/cm2 provided a significantly longer τ1 of 15.54 ns compared to the τ1 of 2.73 ns for u = 48 nJ/cm2. As shown in Fig. 3 and Table 1, the excitonic recombination increases with increasing excitation energy densities due to the state-filling effect of interface states at the heterointerface. The neutralization of the interface states suppresses the intraband transition (the TPU), which causes the two-photon absorption to dominate the carrier dynamics in the MAPbBr3/GaAs systems. These mechanisms increase exciton density in the MAPbBr3 layer. According to the exciton recombination model, the carrier lifetime (τ) decreases with increasing carrier density [60]. Hence, the reduction of the carrier lifetime from τ1 = 15.54 to τ1 = 2.73 ns is because the exciton density increases due to the state-filling effect of the interface states that reduces exciton dissociation. The time constant for band-to-band recombination also supports this scenario. Since τ2 changes from τ2 = 0.67 to τ2 = 0.74 ns, the change suggests that the free carrier density decreases, and therefore, the band-to-band recombination. This is true only if the exciton density is improved. Interestingly, the percentage of the exponential decay component for excitonic recombination (j = 1) escalates, from 49.5 to 57.1%, with increasing energy density while the percentage attributed to band-to-band recombination (j = 2) decreases. On the other hand, the percentage for non-radiative recombination (j = 3) increases, i.e. from 0.1 to 0.3% for u = 27 nJ/cm2 and u = 48 nJ/cm2 respectively (See Table S2 in Supplementary Information). The increasing percentages seem insignificant because the changes are only 0.2%. The non-radiative recombination rate, especially the trap-assisted recombination at grain boundaries (GBs) and interfaces depends on the density of states and the carrier density at the GBs and the interfaces. Sherkar et al. reported the state-filling effect at the GBs of methylammonium lead halide and the perovskite interfaces influencing the trap-assisted recombination rate (non-radiative process). In particular, the non-radiative recombination rate changes and gradually decreases when the neutral region, at the GBs and the interfaces, is established because interface states are filled by either electrons or holes [59]. The results of TRPL strongly support the idea of the state-filling effect of the interface states and the dependence of carrier dynamics in MAPbBr3/GaAs systems on excitation energy densities, and thus, the TPU.

The TRPL of the MAPbBr3 layer on the GaAs and glass substrates. The blue and purple points are the TRPL data for λex = 800 nm and 400 nm, respectively. a the TRPL data for MAPbBr3/GaAs measured under 800-nm excitation, and b under 400-nm excitation. The dark and light-colored solid circles represent different excitation energy densities. The dashed dark and pink curves are the results of fitting a triple exponential function to the data (colour figure online)

For Fig. 4b, the 400-nm-excited TRPL of MAPbBr3/GaAs is depicted with different excitation energy densities, i.e. the dark blue data is for u = 90 pJ/cm2 and the light blue data is for u = 450 pJ/cm2. The results qualitatively show almost no difference. However, the fitted parameters indicate that τ1 increases with increasing energy densities from τ1 = 3.86 to τ1 = 13.96 ns. Similarly to τ2, the value changes from τ2 = 0.54 to τ2 = 3.19 ns (See Table 2). The increases in τ1 and τ2 suggest that the exciton dissociation increases with increasing 400-nm photons together with a decrease in band-to-band recombination. Additionally, the percentage of the exponential decay function for excitonic recombination (j = 1) decreases from 51.7 to 26.6% while the percentage of the exponential decay function for band-to-band recombination (j = 2) increases from 48.2 to 57.1%. This suggests that the exciton dissociation increases with increasing excitation energy densities allowing the recombination of free carriers (the band-to-band recombination) to dominantly contribute to the PL decay. In agreement with the percentage of the exponential decay function for non-radiative recombination (j = 3), the percentage increases significantly from 0.01% to 16.3% confirming an increase in the non-radiative recombination rate. The phenomenon can be interpreted by the effect of penetration depth. At u = 90 pJ/cm2, the 400-nm photons excite mostly at the surface of MAPbBr3 generating carriers that can recombine at the surficial part of MAPbBr3. Therefore, the photoexcited carriers are less influenced by the interface states at the MAPbBr3/GaAs heterointerface. In this case, the exciton density is higher because the interaction between the excitons to the interface states is less significant. Therefore, the excitonic recombination dominates the PL decay. With increasing energy densities to u = 450 pJ/cm2, the photoexcited carriers distribute deeper than the surficial part causing the excitons to interact with the interface states. Their interaction with the interface states significantly escalates the excitonic dissociation providing the reduction in the percentage attributed to the excitonic recombination. The interaction between the photoexcited carrier with the interface states also increases the non-radiative processes, and hence, the percentage increases significantly.

In comparison with different excitation wavelengths, when the perovskite layer is excited with 400-nm light, significantly shorter time constants than the ones obtained with 800-nm light were observed. For example, τ1 = 15.54 ns for 800-nm photoexcitation at u = 27 nJ/cm2 and τ1 = 3.86 ns for 400-nm photoexcitation at u = 90 p/cm2. This result can be explained as follows: The 400-nm light is strongly absorbed in the perovskite layer because the photon energy (3.10 eV) is larger than the Eg of MAPbBr3. On the other hand, when the sample is excited with 800-nm light, a nonlinear excitation process occurs, and thus the resulting photocarrier density is significantly lower. Because the excited carriers occupy the states starting from the lowest energy state, the carriers generated with 800-nm light preferentially occupy the excitonic states. The localized excitons tend to have a slower decay than the free carriers that recombine via band-to-band transitions. If the excited-carrier density is increased, e.g. by choosing above-bandgap excitation, the localized states are filled and free carriers are generated in the band. Therefore, band-to-band recombination becomes gradually dominant at higher excitation power densities.

Figure 5 consists of two-dimensional maps visualizing excitation power-dependent normalized PL spectra of MAPbBr3 under 800-nm photoexcitation, 900-nm photoexcitation, and 1000-nm photoexcitation, with different substrates, i.e. Figure 5a–c are the PL spectra for the MAPbBr3/GaAs samples, and Fig. 5d–f are the PL spectra for the MAPbBr3/Glass samples. The figures were interpolated by the nearest neighbor interpolation in order to clarify the features observed in the figures. The two-variable plots are provided in Fig. S6 in Supplementary Information. Peak shifts to higher energy sides were observed in the PL spectra for MAPbBr3/GaAs samples with increasing excitation energy density. The phenomenon was evidently seen in the case of 800-nm and 1000-nm photoexcitations shown in Fig. 5a and c. For Fig. 5b or the PL spectra measured under 900-nm photoexcitation, the PL peak shift was relatively small. This may be due to the complicated carrier dynamics because 900-nm photons possibly induce direct excitation in GaAs due to the small but finite α900. The situation is more complicated as compared to the 800-nm and 1000-nm excitation case because 900-nm photons can also induce the two-photon absorption within both GaAs and MAPbBr3 with an additional possibility of band-to-band excitation within GaAs. The observed blueshift is attributed to the effect of state-filling occurring at the MAPbBr3/GaAs heterointerface. With increasing excitation energy densities, the interface states are gradually occupied favoring the band-to-band recombination of free carriers and of the excitonic recombination. We discussed in section “TRPL profiles of MAPbBr3” that the excitonic recombination increases because of the state-filling effect at the heterointerface, which is in agreement with the data shown in Table 1. Consideringly, the deconvoluted peaks for the band-to-band recombination and the excitonic recombination exhibit slight blueshift with increasing excitation energy densities. This characteristic suggests that the free carriers and the excitons recombine in higher energy states. Therefore, the PL spectra exhibit the blueshift as shown in Fig. 5a–c. The two-dimensional maps for the MAPbBr3/Glass samples are shown in Fig. 5d–f under the excitation wavelength of 800 nm, 900 nm, and 1000 nm, respectively. The peak shifts were not evident in the case of MAPbBr3/Glass systems. An example of the 800-nm photoexcitation case, the deconvoluted exciton-corresponding peak shift from 2.2795 to 2.2796 eV while the band-to-band recombination peak shift from 2.3021 to 2.3032 eV. Therefore, the differences are 0.0001 eV and 0.0011 eV for the deconvoluted peaks attributed to excitonic recombination and the band-to-band recombination respectively. Without the interface states where carrier occupation occurs, the band-filling effects are insignificant, and, therefore, the blueshift is not evident for the MAPbBr3/Glass samples. We also provide comparisons of the deconvoluted peaks in Fig.