Open Access
Add to BibSonomyAbstract
Recombination processes of photoexcited charge carriers in as-grown micro/nanocrystalline diamond films containing various amounts of non-diamond carbon phase were studied by time resolved photoluminescence spectroscopy. The photoluminescence was found to be strongly sensitive to diamond morphology and concentration of non-diamond carbon phase. We found two photoluminescence components differing in decay rates (nano/microsecond time scales) and in response to ambient air pressure, temperature and excitation light parameters. Experimental data form the basis for interpretation of recombination paths of photoexcited charge carriers in micro/nanocrystalline diamond films in dependence on their morphology and non-diamond carbon phase content.
© 2014 Optical Society of America
1. Introduction
Nanocrystalline diamond (NCD) and microcrystalline diamond (MCD) are unique materials unifying properties of bulk diamond with modifications caused by the presence of the diamond grain boundaries and by the non-diamond carbon phase located between the grains. This combination of bulk and nano-scale properties makes MCD/NCD perspective for a wide spectrum of applications in chemical [1–3] and biological sensing [4–6] as well as in optical and opto-electronical devices [7–9]. However, the complex structure and nano-character of MCD/NCD gives rise to a complicated system of sub-gap energetic states [10]. This makes rather difficult to describe processes such as transport and recombination of charge carriers that are essential for optimization of preparation methods and further extension of MCD/NCD possible applications. To elucidate these characteristics a lot of effort has been made during last two decades [11–14]. It has been shown that major effect on the properties of NDC has the sp2 carbon phase localized between the grains, and the composition and concentration of non-diamond carbon phase in general, even if the ratio of non-diamond to diamond carbon phase is usually low [15,16]. Just the differences in non-diamond phase composition and its interaction with the sp3 carbon on the diamond grain interface could be the reason of fundamental differences in the published results on MCD/NCD by many groups. Although various models have been proposed, proper interpretation of the influence of sp2 on MCD/NCD subsurface processes and interaction between the diamond and non-diamond carbon phase is still missing [12].
We address this issue and present a comprehensive study of recombination processes of photoexcited charge carriers in MCD/NCD sub-gap states. In this work, we aim to clarify the origin of recombination processes of excited carriers in MCD/NCD not only in terms of sub-gap energy states, but also in morphological structures. For this purpose, the research was carried out on a series of well-defined samples with different structure and ratio of non-diamond phase. Raman spectroscopy and SEM were used to monitor the structure and composition of diamond films. Properties of our samples were studied by time resolved photoluminescence (PL) on a broad picosecond-to-millisecond scale under various conditions (vacuum, low temperature). This method has been successfully used for investigation of electronic recombination processes in solid state physics for a long time, nevertheless only few papers on polycrystalline diamond [13, 17] have been published.
2. Samples and Experimental
Mirror-polished n-type (phosphorus doped) (100) oriented Si substrates (10 × 10 mm2) were used for the diamond thin film deposition. First, the samples were cleaned by ultrasonication in acetone for 10 min and in pure isopropyl alcohol for 1 minute. Immediately afterwards, the substrates were mechanically seeded, applying ultrasonic agitation in water based diamond powder suspension (diamond powder from Sigma-Aldrich with an average grains size of 10 nm). The typical seeds density after such process is in range up to 1011 cm2 [18]. Diamond films were grown in a modified hot filament chemical vapor deposition (HFCVD) system [19] from a CH4/H2 gas mixture at the total gas pressure of 3 kPa and substrate temperature in the range of 600 - 700 °C. The samples labeled as A, B and C were deposited at various CH4 concentration, i.e. 1, 5 and 10% of CH4, respectively. The diamond film surfaces of the prepared samples remained as-grown. Finally, to obtain transparent diamond membranes the middle parts of the silicon substrates were etched by NHA solution (HF + HNO3 + CH3COOH – 50:3:8) through a piceine mask. Preparation conditions and basic characterization of all samples are summarized in Table 1.The quality factor fq, which reflects the ratio of diamond and non-diamond carbon phase was calculated using the formula [20]:
where SD characterizes the area under the D-peak in Raman spectra and ΣNDSND describes areas under all other bands of the Raman spectrum. Ideal monocrystalline diamond sample would have fq value of 100. Thickness of the prepared diamond films was determined from cross-section SEM measurements and verified by evaluation of the interference fringes in the reflectance spectra measured in visible and near infrared region with use of commercial software for modeling the optical properties of thin films (FilmWizard).
Table 1. Preparation conditions, thickness and quality of samples A, B and C
In the time resolved PL measurements femtosecond light pulses of the Ti-sapphire femtosecond laser (Tsunami 3960, Newport/Spectra-Physics) with the regenerative amplifier Spitfire Pro-F1KXP (Newport/Spectra-Physics) were used for excitation. The second (400 nm) and fourth (200 nm) harmonics of the amplifier output was used. Pulses at 325 nm were generated by parametric amplifier TOPAS (Newport/Spectra-Physics). The parameters of femtosecond pulses were: repetition rate 1 kHz, pulse duration 100 fs, wavelength 400, 325 and 200 nm and fluence 1.3, 0.3 and 0.04 mJ/cm2, respectively. It was found that the threshold of photoinduced changes of samples was wavelength dependent so that was not possible to find a common magnitude of pump fluence for all excitation wavelengths leading to a sufficient PL signal and not causing the modification of samples. That is why the magnitude of the pump fluence was chosen as the maximum value not exceeding the photo-modification threshold [21]. The time decay of PL was measured by the streak camera C5680 (Hamamatsu) with a spectral sensitivity range of 200 – 900 nm in the time range of 0.1 ns – 1 ms. We used a cryostat chamber with closed helium circulation system and a cascade of oil and turbomolecular pump for the low pressure (pressure approx. 10−3 Pa) and low temperature (to 15K) measurements.
Morphology of the samples was characterized by scanning electron microscope (Zeiss FE SEM Leo 1 550) and the sample composition was studied by Raman spectroscopy (Renishaw In Via Reflex) with the excitation wavelength of 442 nm.
3. Results
3.1. SEM and Raman spectroscopy
Structural characterization and Raman spectra of all three samples are presented in Fig. 1.It is obvious that the samples substantially differ in morphology. Sample A, labeled as microcrystalline diamond (MCD), contains large (~500 nm) randomly oriented diamond grains. Diamond film of sample B has a columnar structure where the height is given by the thickness of the film (7.5 μm). The pores between columns are clearly apparent. Sample C, labeled as nanocrystalline diamond (NCD), reveals a fine structure made of small (≤ 10 nm) nanocrystals. Above mentioned morphologies agree well with the measured Raman spectra. The spectra of all the samples contain a narrow peak at 1 332 cm−1 (D-peak related to sp3 bonds) despite various CH4 concentration during the deposition. Of course, in comparison with sample A, the characteristic D-peak is strongly reduced in intensity in the spectra of the samples B and C whereas the domination of D-band (1350 cm−1) and G-band (1530 cm−1) features is evident. This effect is characteristic for fine-crystalline/amorphous carbon phases. The signal originating from G- and D- band is usually associated with carbon in sp2 phase in any structural formation and situated in ring structural compositions, respectively [16]. Narrow band at 1 150 cm−1 (samples B and C) is related to trans-polyacetylene-like fragments, and is observed with other peaks/bands at 1 332, 1 350 and 1 530 cm−1 in the spectra of a NCD films [22]. Fairly high signal from sp2 phase (G-band) observed in the Raman spectra of sample C is caused by using very high CH4 concentration during the CVD deposition along with a limited etching rate of sp2 bonded carbon by atomic hydrogen. Despite these effects it is evident that higher methane concentration in gas mixture during the sample preparation leads to a larger fraction of non-diamond carbon phase and smaller diamond grain sizes due to higher renucleation.
Fig. 1 Cross-section SEM images and corresponding Raman spectra of the diamond films deposited at various methane concentrations (1, 5 and 10% for the samples A, B and C, respectively).
3.2. Time-resolved photoluminescence at room ambient conditions
The PL decays of all samples are shown on log-log scales in Fig. 2.We display the PL dynamics at 460 nm for all three samples at room temperature and atmospheric pressure in Fig. 2(a). PL was excited by femtosecond laser pulses at 325 nm with energy fluence 0.3 mJ/cm2. Each decay curve is composed of the data measured piecewise to ensure the best time resolution that is limited by the streak camera response on each time scale. The PL dynamics of the samples are different to each other. The PL signal of sample A is strong in sub/nanosecond time interval and contains an additional slow decay stretching up to 1 ms as shown in Fig. 2(b). The PL signals of other two samples are weaker and much faster and their microsecond signals were at (sample B) or below (sample C) the resolution of the streak camera. The PL decay is non-exponential for all samples. We have found that it can be reproduced well by the stretch-exponential (c.exp[-(t/𝜏)δ]) and power-law (c.t-n) functions forthe nanosecond (all samples) and microsecond (sample A) time scales, respectively. The latter appears as a straight-line in the log-log plot (Fig. 2(b)).
Fig. 2 a) PL decay of samples A, B and C at 460 nm on the sub and nanosecond time scales (nanosecond PL component) b) PL decay of sample A at 550 nm on the microsecond scale (microsecond PL component). Excitation laser pulse parameters: 325 nm, 0.3 mJ/cm2. Decays were measured piecewise and for clarity are presented on log-log scale. Dashed curves - streak camera response function (resolution limit). Blue curves - fits of stretch-exponential and power-law functions.
Time integrated PL spectra of our samples are displayed in Fig. 3(a). The PL intensities follow the trend discussed in connection with Fig. 2(a) and the spectra differ also in shape. Sample C shows a broad spectral band from approx. 360 to 700 nm (3.44 – 1.77 eV) with maximum at 450 nm (2.75 eV). The spectrum of sample B is located in similar interval. Only the peak is red shifted to approx. 505 nm (2.46 eV). PL spectrum of sample A is about 45 nm red-shifted and approx. 30 times more intense than that of sample B. The maximum is located at 550 nm (2.26 eV). A weak spectrum modulation of sample A is caused by light interference in the diamond membrane as was verified by transmittance measurement (not shown here). The time evolution of PL spectrum of sample A is presented in Fig. 3(b). The spectra were obtained by integrating the signal over the successive time intervals ≤ 0.1 ns, 1-10 ns and 0.1-1 ms. Spectra in Fig. 3(b) clearly show a gradual spectral shift towards the longer wavelengths with increasing time. We can see a good agreement between the spectra of samples B and C (Fig. 3(a)) and time resolved spectra of sample A (Fig. 3(b)) in corresponding time windows. PL of sample C decays very fast and can be detected only during the initial sub nanosecond time interval which results in its spectrum corresponding well to that of sample A in the same time window. This analysis shows that although the PL nanosecond component of sample A is stronger than those of the other samples, its time integrated spectrum is dominated by the microsecond component. This example clearly demonstrates benefits of time resolved PL in wide time interval, when simple comparison of standard PL for our samples could lead to rather confusing interpretations.
Fig. 3 a) Time integrated spectra of samples A, B and C. b) Time resolved spectra of sample A. For clarity were selected PL spectra multiplied by appropriate constant. Parameters of excitation laser pulses: 325 nm, 0.3 mJ/cm2.
We measured also the PL dynamics at different wavelengths within the PL spectrum of sample A which has the most intense PL. We have found that the nanosecond PL decay does not depend on the wavelength in the interval from approx. 400 to 660 nm as shown in Fig. 4(a).The PL decays measured at the lower wavelengths (≤ 400 nm) showed intensity increase in time interval ≤ 0.1 ns that could be caused by the presence of additional radiative recombination mechanism with time constant under the camera resolution. On the other hand, the microsecond PL decay was spectrally dependent in the whole spectral interval being faster at shorter wavelengths. This decay can be fitted well by the power-law decay function for all wavelengths (see Fig. 2(b) for the data fit at 550 nm). The obtained values of the power-law coefficient n in dependence on PL photon energy are shown in Fig. 4(b). The spectral dependence of the coefficient can be approximated very well by an exponential function n ~exp (E/E0) with E0 = 0.90 eV. This spectral dependence is nearly the same as that found in ultrafast initial decay in NCD membranes [13, Fig. 5].
Fig. 4 a) Spectral dependence of sample A PL decay (nanosecond component). b) Spectral dependence of power-law coefficient n of sample A (microsecond component). Parameters of excitation laser pulses: 325 nm, 0.3 mJ/cm2.
Fig. 5 a) Temperature dependence of sample A PL decay in nanosecond time interval measured at 500 nm (maximal intensity). b) Temperature dependence of sample A PL decay in microsecond time interval measured at 550 nm. Parameters of excitation laser pulses: 325 nm, 0.3 mJ/cm2. For clarity specific PL spectra were divided by appropriate constant.
3.3. Time-resolved photoluminescence - temperature and pressure dependence
To obtain complete information about carrier dynamics the temperature dependence of PL decay on sample A was studied. The PL decay was measured in the interval of 15 K – 300 K. PL was excited by femtosecond pulses at the wavelength of 325 nm. The results are shown in Fig. 5 on the log-log scale. We have observed that the nanosecond decay is slowing down with cooling the sample from room temperature down to approx. 240 K (Fig. 5(a)). The dynamics are not changing with further cooling. The microsecond part of PL decay maintains its power-law shape with nearly unchanged coefficient n in the whole interval of temperatures (Fig. 5b).
The time integrated PL signals of the nanosecond and microsecond PL parts display different temperature behavior (Fig. 6). The PL signal integrated over the initial time interval to 20 ns decreases strongly with decreasing temperature down to 200 K, which correlates with the temperature dependence of PL dynamics. On the other hand the signal integrated from 0.05 to 1 ms changes only weakly in the whole range of temperatures.
Fig. 6 Temperature dependence of time integrated PL spectra of nanosecond and microsecond components of sample A at 500 and 550 nm, respectively. Parameters of excitation laser pulses: 325 nm, 0.3 mJ/cm2. For clarity the microsecond component was rescaled to match the nanosecond component at low temperatures.
We carried out also the PL measurements under low pressure of about 0.001 Pa. We observed apparently no effect of ambient pressure on the PL decay shapes for all samples. The only PL property that was affected by the pressure was the time integrated intensity as shown in Fig. 7. While the magnitude of the microsecond PL component increased only by units of percent (Fig. 7(b)), the amplitude of the nanosecond PL component almost doubled for sample A after the pressure reduction (Fig. 7(a)). On the other hand, only small or even no PL intensity change was observed in the case of sample C.
Fig. 7 a) Influence of air pressure to time integrated spectra of nanosecond PL component of the samples A and C. b) Influence of air pressure to time integrated spectra of microsecond PL component of sample A. Label “vacuum” corresponds to the pressure 0.001 Pa. Parameters of excitation laser pulses: 325 nm, 0.3 mJ/cm2.
3.4. Time-resolved photoluminescence - excitation wavelength and fluence dependence
We have also investigated the behavior of PL in dependence on the excitation wavelength. In Fig. 8 we compare the PL spectra of the nanosecond and the microsecond PL of sample A under different excitation wavelengths, namely 200 nm (6.2 eV), 325 nm (3.8 eV), and 400 nm (3.1 eV). The spectrum of the nanosecond PL component shown in Fig. 8(a) was obtained by integration of the spectrally resolved PL signal over the initial time interval to 20 ns. In Fig. 8(b) we display the spectrum of the microsecond PL component that was integrated over the time interval from 0.05 to 1 ms. In these measurements, the excitation intensity was smaller for shorter wavelengths (about 30 times) and the absorption of sample increases for shorter wavelengths. That is why one cannot correctly compare the PL scaling for different wavelengths of excitation. However, there is a striking difference between the behaviors of both PL parts. The spectrum of the nanosecond component decreases in a magnitude and shifts towards shorter wavelengths with decreasing excitation wavelength whereas the spectrum of the microsecond PL component increases in the magnitude and neither changes its spectral position nor PL decay. The comparison of spectra in Fig. 8(a) and Fig. 8(b) indicates unambiguously different origin of both PL components.
Fig. 8 a) Time integrated spectra of nanosecond PL component of sample A (time interval to 20 ns). b) Time integrated spectra of microsecond PL component of sample A (time interval from 0.05 to 1 ms). Excitation wavelengths were 400 nm, 325 nm and 200 nm with fluence of 1.3 mJ/cm2, 0.3 mJ/cm2 and 0.04 mJ/cm2, respectively.
To confirm this conclusion we have also studied the influence of the excitation fluence on PL for above and below the diamond band gap energy (approx. 5.5 eV). We used excitation wavelengths 325 and 200 nm with the pulse fluencies 0.05 – 0.62 mJ/cm2 and 0.001 – 0.1 mJ/cm2, respectively. Unlike in the previous experiments, the nanosecond PL component is integrated over the time interval from 1 to 20 ns in order to ensure equal detection conditions on all excitation fluencies. Results for sample A are shown in Fig. 9.We observed that the PL decay and the associated shape of the time integrated spectrum of nanosecond component was not affected by the fluence changes at 325 nm (Fig. 9(a), inset), and the time integrated PL intensity was increasing linearly with the fluence (Fig. 9(a)). The nanosecond part of the measured PL under 200 nm excitation wavelength was rather weak and the measurement of fluence dependency was not possible. The microsecond PL component scaled also linearly with the pump fluence under the excitation at 325 nm while it exhibited saturation for fluencies above approximately 0.02 mJ/cm2 in the case of 200 nm excitation. Themicrosecond decay rate was also affected by excitation fluence speeding up for a stronger pumping. However, the shape of the decay was still of the power-law type for both excitations at 200 nm and 325 nm. The power-law coefficient n was thus found to be a function of excitation fluence as shown in the inset of Fig. 9(b). The spectral shape of time integrated spectra of microsecond PL component was also independent of excitation fluence at both excitation wavelengths (data are not presented).
Fig. 9 a) Dependence of pulse fluence on PL intensity of sample A nanosecond component at 500 nm. PL was excited at 325 nm. Inset: time integrated PL spectrum of sample A nanosecond component excited by fluence of 0.62 mJ/cm2 and 0.05 mJ/cm2 (normalized). b) Dependence of excitation intensity on PL intensity of sample A microsecond component at 550 nm. PL was excited at 325 and 200 nm. Inset: fluence dependence of the power-law coefficient n.
4. Discussion
Our results indicate that the behavior of initial nanosecond PL of studied MCD/NCD samples differs from that observed in microsecond time region. Therefore the underlying microscopic origin and/or energy states seem to be different. The overall PL decay is nonexponential. The nanosecond decay can be described well by the stretched-exponential function (Fig. 2(a)) and the microsecond decay can be fitted by the power-law decay function (Fig. 2(b)). The PL evolution described by stretched-exponential function is usually interpreted as a multiexponential decay and is being observed when one PL source is changing its lifetime in time [23] or when there are similar sources with different lifetimes [24]. The physical meaning of stretch-exponential coefficient 𝜏 is then the mean decay time. The power-law decay is well-known from the dynamics of carriers in disordered systems (amorphous or nanocrystalline materials) where the carriers separation occurs [25, 26]. The carrier survival probability in a site due to the trapping process into randomly distributed traps can be indeed described by the power-law function, see [27]. The power-law coefficient n should scale with the separation of the traps, i.e. with the relevant density of states. This means that the values of n for PL decay curves at various photon energies reflect the energy dependence of number of the states. This indicates that the sites responsible for microsecond PL component are exponentially distributed in energy (see Fig. 4(b)) with density increasing with increasing energy (“exponential tail”). The time integrated spectra of nanosecond and microsecond components do not differ substantially (Fig. 3), but the red spectral time evolution is clearly apparent in accord with nanosecond PL decay at shorter wavelengths. The maximum photon energy of nanosecond PL (about 3.4 eV) corresponds to the energetic distance between π-π* states of carbon sp2 phase as measured on NCD by photothermal deflection spectroscopy [11, 22, 28]. Moreover, the time coefficient 𝜏 of sample A (~0.19 ns) agrees well with the results published by Plakhotnik et al. [15] that interpreted PL with lifetime ≥ 0.5 ns as a radiative recombination in non-diamond carbon content around diamond grains. The presence of additional radiative recombination mechanism with fast time constant (≤ 100 ps) in nanosecond PL at short wavelengths can be also interpreted as over band gap recombination between π-π* states of carbon sp2 phase (Fig. 4a).
The process related to the nanosecond decay is thermally activated (see Fig. 6). The thermal activation leads to releasing of the trapped carriers which can in turn radiatively recombine as indicated by an intensity increase and decay acceleration of the nanosecond PL component for higher temperatures (Fig. 5(a)). Very weak temperature dependence of the microsecond power-law-like decay suggests a negligible role of thermal activation (higher energy barriers or another process) between the carrier sites (Fig. 5(b)). The negligible thermal effect, power-law shape of PL decay realized also over 1 milisecond and coefficient n < 1 indicate that the tunneling of trapped carriers could be the recombination mechanism responsible for microsecond PL [29, 30]. Based on this assumption the fluence dependence of power-law coefficient presented in the inset of Fig. 9b can be interpreted as follows. Increasing pulse fluence results in a higher occupation of trap sites and consequently to an increased overlap of wavefunctions of trapped carriers. This would cause an acceleration of tunneling process that is for power-law decaying PL expressed by increasing a magnitude of coefficient n. Different change of time integrated PL intensity of both PL components in response to the change of excitation wavelength indicates a different way of filling the energy states corresponding to the appropriate PL components (Figs. 8(a), and 8(b). The energy states related to the microsecond component seem to be accessible well by photocarriers from bulk diamond created during over band gap excitation (a large increase in intensity between 325 nm and 200 nm that exceeds the bulk diamond band gap) that is in the contrast to results observed for nanosecond PL component. The energy sites responsible for microsecond PL are also shielded from the ambient atmosphere (Fig. 7(b)) or are not sensitive to the influence of adsorbates [13, 31]. For these reasons these states can be likely situated on the surface of diamond grains or directly inside the grains. The latter option, however, seems unlikely according to the results presented in Fig. 9(b), i.e. the PL intensity saturation of microsecond PL with relevant n value about 0.5 for the excitation at 200 nm. If we interpret the PL intensity saturation in terms of trap site filling, the traps are getting full while the overlap of carriers wavefunctions is not as high as in the case of excitation at 325 nm (n ~0.7 and no PL intensity saturation). This implies that not all trap sites are easily accessible for the carriers from the bulk diamond and therefore will be unlikely located inside the diamond grains. Despite all the differences we expect that the origin of states responsible for microsecond PL component should be similar to the states connected with the nanosecond PL due to the similar PL spectrum (Fig. 3).
4.1 Structural and energy states model for high fq diamond
Based on these results and reported data on polycrystalline diamond and amorphous carbon we introduce a simplified structural model and a scheme of energy states of microcrystalline diamond with high fq (Fig. 10). Our three studied samples were prepared at various CH4 concentrations in order to obtain MCD/NCD films of different fraction of non-diamond carbon phase and different morphology as can be seen in Fig. 1.
Fig. 10 a) Schematic model of NCD structure for high fq. b) Simplified model of energy states in microcrystalline diamond for high fq.
For small concentration of CH4 the sample contains large diamond grains and small amount of non-diamond matter. The latter is supposed to consist of the sp2 clusters [32], structures of graphitic origin localized at the grain surface and other unspecified form of carbon in non-diamond phase filling the space between diamond grains (Fig. 10(a)). Similar structure of NCD was also proposed by Dzurňák et al. [21]. The localization of carbon in sp2 phase responsible for nanosecond PL in differently sized clusters (various lifetimes of PL) is in good agreement with the stretch-exponential shape of the decay.
The scheme of energy states is composed of the bulk diamond band of diamond grains (volume states), diamond Σ states that are generally accepted as a diamond tail states caused by carbon atoms in sp3 phase deformed on grain surface, the exponentially decaying trap states on diamond surface (microsecond PL) and separated sp2 carbon phase located in clusters presented by π-π* states that we assume should be responsible for nanosecond PL (Fig. 10(b)). In fact, the NCD is a composite material and as discussed the interaction of sp2 and diamond grains is questionable so we have spatially separated the states linked to sp2 carbon phase from bulk diamond. In the scheme we have neglected the presence of unspecified form of carbon in non-diamond filling the space between diamond grains, because it should not be source of intense PL [32]. We are aware that this carbon will absorb incident light and therefore might play role in filling of energetic states in sp2 clusters and in the charge transport between the clusters, but this effect is clearly not apparent from any measurement we have made. Unlike the other authors we are not expecting that the π - π* states would have only Gaussian shape of density of states (DOS) [28]. We assume that the density of states continue increasing after crossing the gap about 3.4 eV (Fig. 10(b) – Full blue curve). This energetic structure is in good agreement with results gained by photothermal deflection spectroscopy [28] and photoconductivity measurements [33] on NCD. These measurements commonly fit perfectly to Gaussian density of π states till the value about 3 eV, but the absorption or photoconductivity signal never decrease when the energy of incident light cross the π states gap. This discontinuity is usually explained by the presence of diamond Σ – Σ* tail states transition at higher energies that compensate the decrease of π DOS. However these states should be spatially separated from sp2 carbon and therefore the intensive filling of sp2 π states by electrons and holes from Σ tail that would be necessary for interpretation of Fig. 8(a) is based on our previous discussion rather questionable. Moreover we did not observe any PL signal in the interval of wavelengths from 350 nm to diamond band gap (3.4 – 5.5 eV) after excitation at 200 nm so we cannot confirm an influence of sigma states to presented phenomena. Moreover, the density of π states presented in our model could easily interpret the lack of PL in UV region (fast relaxation of photoexcited charge carriers in the region of continuous density of states to gap energies and tail) as well as properties of nanosecond PL part excited by different excitation wavelengths. Laser pulses at 400 nm (3.1 eV) can excite charge carriers only to tail of π states where they are well localized what results in PL maximum around 510 nm. Excitation at 325 nm (3.8 eV) has enough energy to excite carriers over the π state band gap. The carriers are free after excitation, but are limited by the size of sp2 clusters as observed in recent conductivity measurements [34]. However these carriers have a higher probability to be spatially separated that will cause accelerating of PL decay. This effect is further amplified when the excitation at 200 nm is used. We interpreted already our previous results on picosecond PL decay in nanocrystalline diamond in terms of the spatial separation of the electron and hole of a photoexcited electron-hole pair by trapping of one of the carrier by a surface trap [14].
4.2 Structural and energy states model for low fq diamond
According to this discussion and our results, we propose following structural model and model of energy states also for samples with lower fq (Fig. 11). A higher CH4 concentration in the gas mixture during the preparation results in an increase of fraction of non-diamond carbon phase and in a decrease of diamond grain size, i.e. the growth of NCD (Fig. 1 – sample C). Both effects could cause increasing size of sp2 clusters, higher probability of their mutual penetration and a further distortion of clusters by other forms of non-diamond phase (Fig. 11(a)). It is also evident that the creation of specific defects that are sources of microsecond PL requires the CVD process at low methane concentration (Fig. 2(b) and Table 1). This could be interpreted in the term of specific preparation conditions for their formation or higher covering of grain surface by sp2 clusters that would limit the creation of mentioned defects on the diamond grain surface. Considerable decrease of microsecond PL at samples B and C in contrast to A could be also linked to its recombination mechanism. Tunneling effect is significantly dependent on a distance of the closest carriers and therefore the amount of trap size could scale greatly with the total PL intensity.
Fig. 11 a) Schematic model of NCD structure for low fq. b) Simplified model of energy states in nano-crystalline diamond for low fq.
Although the concentration of sp2 is increasing (decrease of fq) the intensity of nanosecond PL decreases (Fig. 11(b)) and its decay gets faster: 𝜏 ~0.19 and 0.05 for sample A and B, respectively. (Figs. 1, 2(a) and 3(a)). This phenomenon can be explained by the increased number of non-radiative decay channels in the low quality samples. An important role in the PL acceleration would play the change of sp2 cluster shape and their mutual connection with decreasing quality of NCD that would decrease localization of trapped charge carriers and allow their easier spatial separation. Also the increasing concentration of non-diamond phase in NCD could cause shielding of sp2 clusters from ambient (Fig. 7(a)) that would result in insensitivity to change of air pressure of nanosecond PL in contrast to the microcrystalline diamond, i.e. the sample with high fq (Fig. 7(a)).
In our model, we have not included explicitly the role of color centers the presence of which in the samples cannot be excluded. However, our experimental results do not seem to reflect the known features of dopant-related centers (spectral shape, lifetime, dependence to experimental parameters).
5. Conclusion
MCD/NCD diamond films containing various amount of non-diamond phase were studied by time resolved PL, Raman spectroscopy and SEM. We have presented that increase of CH4 in gas mixture during preparation cause increase of non-diamond phase and decrease of diamond grains. We have showed that PL of microcrystalline diamond (sample A) with low concentration of non-diamond carbon phase is composed of two independent PL components. PL decay of these components was observed in sub nanosecond and microsecond time scales and followed the stretch-exponential and power-law dependence, respectively. Properties of both PL components were studied at different stages of decay, with a wide range of excitation wavelengths and fluencies, at ambient conditions, as well as at low pressure and temperatures. We have also demonstrated that increasing concentration of non-diamond carbon phase and structural changes in MCD/NCD are causing a quenching of microsecond PL and acceleration of nanosecond PL decay. Characterization of PL decay by power-law and stretched-exponential laws and establishing dependence of relevant coefficients on experimental parameters made it possible to map the MCD/NCD carrier dynamics in a fairly complex parameter space. Based on these results we have proposed the structural model and model of energetic states for diamond films with high and low quality factor fq. We interpret the nanosecond PL in terms of radiative recombination between π – π* states of sp2 carbon phase localized in clusters between diamond grains. The microsecond PL is most probable caused by tunneling between sites of non-diamond carbon origin on the diamond grain surface. The change of preparation conditions leading to lower quality factor of NCD is preventing the formation of the tunneling sites and causing structural changes of sp2 clusters that result in easier spatial separation of photo-generated carriers in sp2. Presented comprehensive model of structural and energy states of MCD/NCD describing well the influence of non-diamond carbon phase on photoluminescence and recombination properties of diamond contributes to understanding a manifold of often apparently contradictory published results of measurements of MCD/NCD prepared by different preparation methods.
Acknowledgments
Authors would like to acknowledge the kind assistance of M. Michalka for SEM and B. Rezek for theoretical consultations about presented phenomenon. This work was supported by Charles University in Prague grant SVV-2013-265306, by GAUK 151910/2012, GAČR projects no. 13-12386S, P108/12/G108, and Slovak Research and Development Agency under the contract No. APVV-0365-12.
References and links
1. J. de Sanoit, E. Vanhove, P. Mailley, and P. Bergonzo, “Electrochemical diamond sensors for TNT detection in water,” Electrochim. Acta 54(24), 5688–5693 (2009). [CrossRef]
2. S. Stehlík, T. Ižák, A. Kromka, B. Dolenský, M. Havlík, and B. Rezek, “Sensitivity of Diamond-Capped Impedance Transducer to Tröger’s Base Derivative,” ACS Appl. Mater. Interfaces 4(8), 3860–3865 (2012). [CrossRef] [PubMed]
3. M. Davydova, M. Stuchlík, B. Rezek, K. Larsson, and A. Kromka, “Sensing of phosgene by a porous-like nanocrystalline diamond layer with buried metallic electrodes,” Sens. Actuators B Chem. 188, 675–680 (2013). [CrossRef]
4. B. Rezek, M. Krátká, A. Kromka, and M. Kalbáčová, “Effects of protein inter-layers on cell-diamond FET characteristics,” Biosens. Bioelectron. 26(4), 1307–1312 (2010). [CrossRef] [PubMed]
5. M. Krátká, A. Kromka, E. Ukraintsev, M. Ledinský, A. Brož, M. Kalbáčová, and B. Rezek, “Function of thin film nanocrystalline diamond-protein SGFET independent of grain size,” Sens. Actuators B Chem. 166-167, 239–245 (2012). [CrossRef]
6. A. Härtl, E. Schmich, J. A. Garrido, J. Hernando, S. C. Catharino, S. Walter, P. Feulner, A. Kromka, D. Steinmüller, and M. Stutzmann, “Protein-modified nanocrystalline diamond thin films for biosensor applications,” Nat. Mater. 3(10), 736–742 (2004). [CrossRef] [PubMed]
7. J. Čermák, B. Rezek, P. Hubík, J. J. Mareš, A. Kromka, and A. Fejfar, “Photo-conductivity and Hall mobility of holes at polypyrrole-diamond interface,” Diamond Related Materials 19(2-3), 174–177 (2010). [CrossRef]
8. V. Prajzler, M. Varga, P. Nekvindová, Z. Remeš, and A. Kromka, “Design and investigation of properties of nanocrystalline diamond optical planar waveguides,” Opt. Express 21(7), 8417–8425 (2013). [CrossRef] [PubMed]
9. L. Ondič, O. Babchenko, M. Varga, A. Kromka, J. Ctyroký, and I. Pelant, “Diamond photonic crystal slab: Leaky modes and modified photoluminescence emission of surface-deposited quantum dots,” Sci Rep 2, 914 (2012). [CrossRef] [PubMed]
10. A. M. Zaitsev, Optical Properties of Diamond (Springer, 2001), Chap.5.
11. M. Vaněček, J. Rosa, M. Nesládek, and L. M. Stals, “Evidence for the principal defect states in CVD diamond films: Optical study,” Diamond Related Materials 5(9), 952–957 (1996). [CrossRef]
12. P. Achatz, J. A. Garrido, M. Stutzmann, O. A. Williams, D. M. Gruen, A. Kromka, and D. Steinmüller, “Optical properties of nanocrystalline diamond thin films,” Appl. Phys. Lett. 88(10), 101908 (2006). [CrossRef]
13. B. Dzurňák, F. Trojánek, J. Preclíková, A. Kromka, B. Rezek, and P. Malý, “Ultrafast photoluminescence spectroscopy of H- and O-terminated nanocrystalline diamond films,” Diamond Related Materials 20(8), 1155–1159 (2011). [CrossRef]
14. P. Němec, J. Preclíková, A. Kromka, B. Rezek, and P. Malý, “Ultrafast dynamics of photoexcited charge carriers in nanocrystalline diamond,” Appl. Phys. Lett. 93(8), 083102 (2008). [CrossRef]
15. B. R. Smith, D. Gruber, and T. Plakhotnik, “The effects of surface oxidation on luminescence of nano diamonds,” Diamond Related Materials 19(4), 314–318 (2010). [CrossRef]
16. T. Ikeda, K. Teii, C. Casiraghi, J. Robertson, and A. C. Ferrari, “Effect of the sp2 carbon phase on n-type conduction in nanodiamond films,” J. Appl. Phys. 104(7), 073720 (2008). [CrossRef]
17. J. Preclíková, F. Trojánek, B. Dzurňák, P. Malý, A. Kromka, and B. Rezek, “Light-assisted adsorption processes in nanocrystalline diamond membranes studied by femtosecond laser spectroscopy,” Diamond Related Materials 19(7-9), 918–922 (2010). [CrossRef]
18. M. Varga, T. Ižák, A. Kromka, M. Veselý, K. Hruška, and M. Michalka, “Study of diamond film nucleation by ultrasonic seeding in different solutions,” Cent. Eur. J. Phys. 10(1), 218–224 (2012). [CrossRef]
19. T. Ižák, M. Marton, M. Varga, M. Vojs, M. Veselý, R. Redhammer, and M. Michalka, “Bias enhanced nucleation of diamond thin films in a modified HFCVD reactor,” Vacuum 84(1), 49–52 (2009). [CrossRef]
20. W. Fortunato, A. J. Chiquito, J. C. Galzerani, and J. R. Moro, “Crystalline quality and phase purity of CVD diamond films studied by Raman spectroscopy,” J. Mater. Sci. 42(17), 7331–7336 (2007). [CrossRef]
21. B. Dzurňák, F. Trojánek, J. Preclíková, A. Kromka, B. Rezek, and P. Malý, “Subgap photoluminescence spectroscopy of nanocrystalline diamond films,” Diamond Related Materials 18(5-8), 776–778 (2009). [CrossRef]
22. M. Varga, Z. Remeš, O. Babchenko, and A. Kromka, “Optical study of defects in nano-diamond films grown in linear antenna microwave plasma CVD from H2/CH4/CO2 gas mixture,” Phys. Status Solidi, B Basic Res. 249(12), 2635–2639 (2012). [CrossRef]
23. K. Dunn, J. Derr, T. Johnston, M. Chaker, and F. Rosei, “Multiexponential photoluminescence decay of blinking nanocrystal ensembles,” Phys. Rev. B 80(3), 035330 (2009). [CrossRef]
24. L. Pavesi and M. Ceschini, “Stretched-exponential decay of the luminescence in porous silicon,” Phys. Rev. B Condens. Matter 48(23), 17625–17628 (1993). [CrossRef] [PubMed]
25. J. Bisquert, “Interpretation of a fractional diffusion equation with nonconserved probability density in terms of experimental systems with trapping or recombination,” Phys. Rev. E Stat. Nonlin. Soft Matter Phys. 72(1), 011109 (2005). [CrossRef] [PubMed]
26. P. H. Sher, J. M. Smith, P. A. Dalgarno, R. J. Warburton, X. Chen, P. J. Dobson, S. M. Daniels, N. L. Pickett, and P. O’Brien, “Power law carrier dynamics in semiconductor nanocrystals at nanosecond timescales,” Appl. Phys. Lett. 92(10), 101111 (2008). [CrossRef]
27. J. Klafter and M. F. Shlesinger, “On the relationship among three theories of relaxation in disordered systems,” Proc. Natl. Acad. Sci. U.S.A. 83(4), 848–851 (1986). [CrossRef] [PubMed]
28. M. Nesládek, K. Meykens, L. M. Stals, M. Vaněček, and J. Rosa, “Origin of characteristic subgap optical absorption in CVD diamond films,” Phys. Rev. B Condens. Matter 54(8), 5552–5561 (1996). [CrossRef] [PubMed]
29. D. J. Huntley, “An explanation of the power-law decay of luminescence,” J. Phys. Condens. Matter 18(4), 1359–1365 (2006). [CrossRef]
30. A. K. Jonscher and A. de Polignac, “The time dependence of luminescence in solids,” J. Phys. C Solid State Phys. 17(35), 6493–6519 (1984). [CrossRef]
31. J. Preclíková, F. Trojánek, A. Kromka, B. Rezek, B. Dzurňák, and P. Malý, “Ultrafast photoluminescence of nanocrystalline diamond films,” Phys. Status Solidi., A Appl. Mater. Sci. 205(9), 2154–2157 (2008). [CrossRef]
32. J. Robertson, “Gap states in diamond-like amorphous carbon,” Philos. Mag. B 76(3), 335–350 (1997). [CrossRef]
33. Z. Remeš, T. Ižák, A. Kromka, and M. Vaněček, “High optical quality nanocrystalline diamond with reduced non-diamond contamination,” Diamond Related Materials 19(5-6), 453–456 (2010). [CrossRef]
34. P. Hubík, J. J. Mareš, H. Kozák, A. Kromka, B. Rezek, J. Krištofík, and D. Kindl, “Transport properties of hydrogen-terminated nanocrystalline diamond films,” Diamond Related Materials 24, 63–68 (2012). [CrossRef]
