Single-nanocrystal spectroscopy

the surface and core states in oxidized SiNc seems to be rather weak and both types of states keep their characteristic.

In the organically passivated SiNc, however, one can draw interesting parallels be-tween the excitation spectra of the colloidal samples and the yellow-emitting molecules formed in the reference sample of UV-irradiated solvent. Even if they share common characteristics, they are inherently different, suggesting that in this case the nanocrys-talline core actively cooperates with the surface states of the capping molecules, with the resulting PL being strongly influenced by both these characteristic (further evidence for this conclusion will be provided in the following section).

3.4. Single-nanocrystal spectroscopy

3.4.1. Single-nanocrystal spectroscopy of the silicon colloid

Single-nanocrystal spectroscopy (1NcS) provides further insight into the optical prop-erties of nanocrystals. So far, 1NcS measurements on SiNc were performed by five laboratories world-wide, for details see Sec. 1.5.

When carrying out 1NcS measurements^[A1], a droplet of nanocrystals dispersed in a liquid is usually used as a sample. We deposited a droplet (1.5µl) of the colloid onto a dove prism; the droplet, however, needed to be strongly diluted (∼1:100) since when the original colloid was used, spatially more or less homogeneous luminescence of the deposited layer was observed under an optical microscope (and under the excitation with an evanescent wave of 458-nm Ar laser). When the diluted colloid is deposited onto a prism, larger agglomerates of nanocrystals form (Fig. 3.16a). Nevertheless, in between them, a lot of single nanocrystals appear, as can be guessed from the image of a slit (Fig. 3.16b), placed in the collection system to define the spatial region, from which the spectra will be acquired. The image of these acquired spectra on the CCD (Fig. 3.16c) with ten times longer acquisition time than that used in the micro-PL photo then illustrates the multitude of light-emitting objects between the agglomerates even better.

An interesting feature of these spectra is the obvious presence of two peaks, which become even better discernible when the spectra are plotted into a graph (Fig. 3.16d).

The spectra can be nicely fitted with a combination of two Gaussian peaks^∗(inset in Fig. 3.16d), which were then used to statistically characterize the measured data.

Altogether, 63 spectra of 1Ncs were observed. All of these spectra consist of two peaks^† whose distance is 150 meV (Fig. 3.16g) and widths are around 100 meV^‡ (Fig. 3.16f). The centers of the two peaks shift between 525–575 and 560–620 nm,

re-∗even though the measured signal is quite high, spectra are still considerably noisy; as a result, a combination of two Lorentzian peaks also fits the measured data very well and it is not possible to determine (based on the fit) which of these two functions describes the measured data better

†a few spectra had actually four peaks with the third one being higher than the second one, these were considered as a combination of spectra of two individual nanocrystals

‡more specifically, (150±14) meV distance and (94±15) and (125±47) meV widths (the arithmetic average and standard deviations), in the second peak, however, the median is 116 meV and its signal is weaker than in the case of the first peak, resulting in noisier data

(a) (b) (c)

(d) (e)

40 80 120 160 200 240 0

10 20 30

numberofpeaks

peaks' widths - FWHM (meV)

(f)

120 140 160 180 0

5 10 15 20

numberofpeaks

1 st

and 2 nd

peaks' distance (meV)

(g)

120 140 160 180 200 0

1 2 3 4

numberofpeaks

2 nd

and 3 rd

peaks' distance (meV)

(h) (i)

Figure 3.16: 1NcS of the colloidal SiNc. PL of one investigated spot under laser excitation (458 nm) as seen through the eyepiece of the optical microscope can be seen in (a) (after^[A1]), the red rectangle denotes the position of the slit. The image of PL from inside the slit is shown in (b), while (c) (after^[A1]) represents the image of spectra acquired from the area of the slit on the CCD. Several chosen smoothed spectra of 1Ncs can be found in (d) (after^[A1]), while the inset shows raw measured data together with a fit with two Gaussian peaks. Such two-Gaussian fits of all 1Nc spectra yielded statistics of peaks’ positions that are compared with macroscopically measured spectra (gray shaded curve) in (e) (after^[A1]), widths in (f) and distances in (g). Several spectra were intense enough to enable us to estimate also the distance of the third peak, shown in (h). Lastly, (i) contains the sum of all the measured single-nanocrystal spectra (black curve) which is again plotted together with macroscopically measured PL.

spectively, while most of the first peak’s centers (62%) are situated between 550–560 nm (Fig. 3.16e).

A part of the 1NcS spectra were intense enough to allow us to discern a third peak.

For example, the non-smoothed spectrum in the inset of Fig. 3.16d exhibits a distinct indication of the third peak, similarly to the smoothed spectra in black and light blue in the same graph. Although only 13 triple-peak spectra were found, the spectral position

3.4. Single-nanocrystal spectroscopy

(a) (b)

Figure 3.17: ReferenceXAsample in the microscopical PL measurement: PL image under the micro-scope (a) and the image on of the spectra inside the slit on the CCD (b).

of the peaks seems equidistant, with the third peak being again∼150 meV^§apart from the second one (Fig. 3.16h).

In contrast to macroscopically measured PL, these microscopically obtained spectra show every sign of 1Nc spectra: widths of the peaks get narrow downto 100 meV^¶ and their centers shift, indicating that the PL comes from 1Ncs of different sizes.

The fact that the distance between the two peaks remains the same even when the spectra are shifted strongly indicates that this double-peak spectrum is emitted by and characteristic of a single nanoobject, the silicon nanocrystal.

Both the 1NcS and macroscopic PL should be compared from the point of view of their spectral position. The positions of the fitted Gaussians overlap very reasonably with macroscopic PL, as is evidenced inFig. 3.16e, which gives a strong confirmation of the conclusion that the same emission process underlies both the micro- and macroscop-ically studied spectra. When all the measured 1Nc spectra are summed up (Fig. 3.16i), however, they do not entirely copy the macroscopic PL spectrum; although the spectral positions of the sum and macroscopic PL match very well, the 1Nc PL spectra keep the double-peak character even after the summation. We can think of two reasons that might lie behind this discrepancy: firstly, 63 spectra might not be a statistically significant number and more spectra may be needed to recreate the whole macroscopic spectrum and secondly, reabsorption might play a role in the macroscopically measured samples, which are more concentrated than the microscopically measured ones.

3.4.2. Reference microscopical measurement of irradiated xylene

Since the reference sample of irradiated xylene exhibits PL in a very similar spectral re-gion to that of our samples, we needed to make sure that the observed single-nanocrystal spectra can really be correlated with nanocrystals. To carry out the reference measure-ment, we prepared a sample of diluted irradiated xylene (XA) that underwent the very same treatment as the Si-coll for the 1NcS measurement and we made one acquisition of the microscopical 1NcS measurement on the colloid, confirming the double-peak structure, on the same day when the reference was measured.

§(154±16) meV

¶this is a typical FWHM for 1NcS spectra measured at room temperature, at low temperatures, however, the FWHM of 1NcS spectra of oxide-passivated SiNc can be as low as a few meVs, see Sec. 1.5.3

The PL of the dilutedXAsample observed under the microscope (Fig. 3.17a) looked similar to the same observation on the Si-coll except for a tendency towards the forma-tion of chain-like structures in the reference samples. The PL spectra acquired from inside the slit (Fig. 3.17b) clearly exhibited no double-peak structure (even though the overall detected PL intensity was slightly weaker than in the previous case), confirming the correlation of the double-peak structure with nanocrystals.

3.4.3. Single-nanocrystal spectra: summary and discussion

The obtained spectra of 1Nc provide further proof that the observed PL comes from nanocrystals since features characteristic of nanocrystals, i.e. band narrowing and spectral shift, were observed. The observed peak’s width (∼100 meV) are in accordance with previous studies by different groups^[51–55,A9] (the summary of 1NcS measurements and the corresponding samples can be found in Tab. 1.5on page 20), and the detected peak’s positions reasonably overlap with the microscopically measured spectrum. The positions of the peak shift in a relatively narrow spectral window, indicating that the nanocrystals in the sample exhibit narrow distribution of sizes (and thus corroboration DLS measurements, see Sec. 3.1.4).

Nevertheless, 1NcS also revealed an unexpected outcome, namely the occurrence of two (or even three) peaks, about 150 meV apart. Although this structure has been observed previously^[51,52], it was attributed to the vibration of surface oxide, and therefore not expected in our organically capped SiNc.

When the double-peak 1Nc spectra by different groups are plotted together in one graph with the x-axis in energy (Fig. 3.18a^[A1]), it becomes obvious that the spectra are very similar except for an energetic shift. In this plot, we included, besides the already mentioned double-peak spectra^[51,52], also a “single-peak” spectrum by English et al.^{[ 55 ]}. On closer inspection, a second peak appears roughly at the same position as that in the “double-peak” spectra. Interestingly, the SiNcs studied by this group were also organically capped^k (see Sec. 1.5.6 for the discussion of the literature data).

It is important to note that the SiNc in Fig. 3.18a were not only prepared by completely different methods, but also substantially vary in their surface chemistries (our samples and samples by English et al.^{[ 55 ]}are organically capped). This comparison throws serious doubt on the “traditional” explanation that the energy splitting of the two peaks is due to silicon-oxide vibration^[51,52], indicating that it is more likely related to the nanocrystalline core, which, in contrast to the surface, may have quite similar properties in all these cases.

Nevertheless, the identification of the energy splitting of∼150 meV with a physical property of a nanocrystal is slightly troublesome since the corresponding energy is quite high. It is very unlikely to be connected with phonons in the nanocrystals as it is far beyond the energy of phonons in bulk silicon (∼60 meV) and new, nanocrystal-related phonons can be expected to be situated even below this energy (see Fig. 1.8).

However, we found a very close resemblance between our 1Nc spectra and those of CdSe-core-ZnS-capped nanocrystals published very recently by Wang et al.^{[ 63 ]}. In that study, the authors claim to have prepared CdSe-core-ZnS-capped nanocrystals with a

kand, moreover, also the PL lifetime shortening into the nanosecond range was observed

3.4. Single-nanocrystal spectroscopy

(a)

Martin et al.^{[ 52 ]} our spectra Mason et al.^{[ 51 ]} English et al.^{[ 55 ]}

(b)

Figure 3.18: Comparison of 1Nc spectra containing two peaks by different groups is shown in (a) (after^[A1], spectra are redrawn from Fig. 1.11b,1.11c, 1.13) , while (b) represents modified scheme of excitation and emission processes in a “typical” nanocrystal.

gradual transition between the core and the surface layers, which was found to result in (i) the decrease of luminescence lifetime (20→4 ns), (ii) the occurrence of a triple-peak structure (peaks are 164 meV apart) in 1Nc spectra and (iii) the absence of blinking.

These three phenomena were explained by the radiative recombination of a trion, i.e.

a quasiparticle consisting of two holes and one electron. During the recombination, the electron-hole pair recombines while a part of its energy is transfered to the second hole, which gets excited onto another quantized hole energy level in the nanocrystal^∗∗. The PL of a trion is made possible due to the suppression of Auger recombination, which occurs due to the special design of the core-surface boundary.

hole level energy difference types D= 2.75 nm D= 4.0 nm heavy/mixed 170 meV 140 meV heavy/heavy 300 meV 300 meV

Table 3.4: Energy distance of quantized hole levels for SiNcs of different diameters (after^[64]).

One can draw surprising par-allels between the measurements by Wang et al.^{[ 63 ]} and our re-sults. We observed a decrease in PL lifetime and 1Nc spectra of or-ganically capped SiNc feature two peaks, although 1Nc spectra of oxidized SiNc (of the same type which was used for the preparation

of the organically capped SiNc) exhibits only one, though slightly asymmetric peak (see Fig. 1.12c). Even if it is far-fetched to hastily draw conclusions about the observation of trions in our system, the idea that the energy splitting is due to quantized hole energy levels in a nanocrystal seems plausible. The difference between the 1NcS spectra of oxidized and organically capped SiNc may then be due to the exclusion (or at least curbing) of the surface trapping of excitons in the organically capped samples. The energetic splitting of the ground state of the organically capped nanocrystalline system would modify the scheme or energy levels proposed in Fig. 3.15b to that in Fig. 3.18b.

Theoretical calculations suggest that the energy difference of quantized hole levels

∗∗such a recombination looks like a “radiative Auger” process

Si atom 4.6×10⁻²⁰ mg atoms in a 3-nm nc 700

3-nm nc 3.45×10⁻¹⁷ mg SiNcs in 2.5 mg 6×10¹⁶

concentration (e.g.

sampleWSiX) 1.2×10^{17 SiNcs}_cm3

(a)

α⁴⁴²_1sedX 1.19 cm⁻¹

α⁴⁴²_xylen 0.747 cm⁻¹

α⁴⁴²_SiNc 0.443 cm⁻¹ σ_SiNc⁴⁴² 10⁻¹⁷ cm² concentration

in 1sedX 4×10^{16 SiNcs}_cm3

(b)

Table 3.5: Estimates of the dose of nanocrystals in (a) and of the concentration of nanocrystals in 1sedXfrom absorption in (b).

in SiNc might be in the hundred-meV range and even the shift of the levels due to varying size of the nanocrystals does not significantly influence the energy difference between the levels^[64] (see Tab. 3.4). This result supports our conclusion since the energy splitting is likely not to be sensitive to the nanocrystal core’s size (compare the shift in PL spectra in Fig. 3.18a).

In document Text práce (8.515Mb) (Stránka 75-80)