1 Introduction

Silicon, as a monoatomic crystal, is in general not suitable for optoelectronic applications due to its indirect bandgap and poor light absorption and emission properties. Since the nanostructured silicon has quantum confinement effect, silicon bandgap tuning is possible in different forms of silicon, such as porous silicon [1, 2], silicon nanowires [3–5], silicon nanocrystals (Si NCs) [6–8], and silicon quantum dots (Si QDs) or Si NCs embedded in dielectric [9–12]. It should be explained that a Si NCs with size less than the exciton Bohr radius, which shows quantum size effects are usually called Si QDs [13]. The enhancement of bandgap and indirect-to-direct conversion of QDs of indirect-gap materials are ascribed to quantum confinement effects [14]. Si QDs exhibit strong quantum confinement and therefore their optical, electrical, and photovoltaic properties can be tuned by controlling their size, density, shape, crystalline structure, surrounding matrix, and doping content [15–21].

In terms of photovoltaic market, Si-based solar cells, as they are based on abundant, nontoxic materials and the mature device processing technology, have been identified as the leading photovoltaic technology. Currently, the highest efficiency of ~25.6 % for a crystalline silicon solar cell is obtained [22], which is very close to a recent theoretical limit of 29.43 % [23], so obviously only limited further progress can now be gained. To break through this ideal efficiency limit, the concept of ‘all-Si’ tandem solar cells [19] has been proposed as a way of improving conversion efficiency. Each cell with different energy bandgap in tandem stack can be optimized for a different part of incident light. Theoretical calculations show that efficiencies of tandem solar cells can be improved up to 42.5 and 47.5 % for two- and three-cell tandem stacks, respectively [24].

Several recent reports have suggested that Si quantum dots (QDs) embedded in different dielectric matrix (such as SiO2, Si3N4, or SiC) are a promising candidate for the upper cell in an ‘all-Si’ tandem solar cell [18, 25, 26]. The material’s effective bandgap can be varied by varying the size of Si QDs, due to three-dimensional quantum confinement. An effective bandgap of about 1.7 eV has been demonstrated for 2 nm diameter Si QDs embedded in silica [27]. The tunneling probability of carriers into Si QDs embedded in these host matrices is heavily dependent on the barrier height between the Si QDs and the matrix [27]. SiO2 (bandgap of 9 eV [19]) is a commonly used matrix and Si QDs is generally fabricated in Si-rich silicon oxide films. Another dielectric option is Si3N4 (bandgap of 5.3 eV [19]) matrix. SiC has a much lower bandgap energy (2.36 eV in 3C β-SiC [28]) compared to Si3N4 and SiO2, resulting in the highest expected mobility of carriers for a given QDs size and separation [29]. Consequently there is potential for more conductive devices to be produced using a SiC matrix as compared to Si3N4 or SiO2 matrixes [30]. However, both the polarity and length of the bonds decrease towards those of Si–Si for oxide, nitride, or carbide, and SiC is an isoelectronic covalently bonded material as Si is, thus there is a notably smaller tendency for segregation and precipitation of Si than in SiO2 or Si3N4 matrices [24]. Moreover, quantum confinement is relatively weak for Si QDs embedded in Si3N4 or SiC matrix compared to in SiO2 matrix. In summary, these three materials each have their own advantages and are worthwhile to further study. Mangolini [2], Janz et al. [31], Schnable et al. [32] have provided an exhaustive comparison of these three matrices, the readers can find more detailed information in the above reviews.

A lot of research covering materials deposition and device development has already been carried out in relation to Si QDs films. The aim of this paper is to review the past research works mainly on the experimental aspects of Si QDs films for photovoltaic applications. It covers the formation of Si QDs nanostructures, physical properties of the Si QDs, doping, and the performance of photovoltaic devices. This paper is structured as follows. In Sect. 2, we will present a summary of different synthesis approaches. The methods used to improve the properties (such as size control, density improvements, doping, and defect passivation) of Si QDs films will be discussed in Sect. 3. In Sect. 4, we summarize the photovoltaic studies with Si QDs and propose ideas for further exploration. Conclusions are presented in Sect. 5, where the existing problems and outlook for the future work is given.

2 Fabrication techniques of Si QDs

Commonly, in order to obtain Si QDs embedded in a dielectric matrix, firstly silicon-rich thin films [i.e. silicon-rich silicon oxide (SRO) or silicon-rich silicon nitride (SRN) or silicon-rich silicon oxynitride (SRON) or silicon-rich silicon carbide (SRC)] are deposited, and the as-deposited films are generally amorphous. Subjecting the as-deposited films to a post-deposition thermal treatment may result in the phase separation of the excess Si. During post-deposition annealing, the host matrix tends toward a stoichiometric composition, and the extent of phase separation between the excess Si and a dielectric host (i.e. SiO2 or Si3N4 or SiC) and crystallization degree can be modulated through annealing temperature and time [33–35]. As an example, one can see in Fig. 1 that the crystallization degree increased with the increase of annealing temperature, being demonstrated by the increase of intensity of Si NCs peak along with the significant decrease of amorphous Raman contribution. It is important to note here that care must be taken when analyzing the precise position and full width at half maximum of Si NCs peak as it is affected by quantum confinement as well as residual stress [36, 37]. There are two typical approaches to the fabrication of Si QDs embedded in a dielectric matrix: (1) a Si-rich single layer film and (2) a multilayer (or called superlattice) structure film (e.g. SRO/SiO2 multilayer). In the first approach, the size of Si QDs is controlled through the amount of excess Si in the as-deposited films and the post-annealing parameters [38–40]. In the second approach (for example for SRO/SiO2 multilayer, SiO2 layers act as barrier layers), upon high-temperature annealing, diffusion of the excess Si is restricted by the SiO2 layers, and sizes controlled Si QDs in multilayer structures are formed (see Fig. 2), as discussed further in Sect. 3 below.

Fig. 1
figure 1

Raman spectra of 48 at.% Si SRO film deposited on quartz annealed at 1000, 1100, and 1150 °C for 1 h. Reproduced with permission from [33] © 2005 The American Physical Society

Full size image
Fig. 2
figure 2

Multilayer structure of alternating Si-rich dielectric/stoichiometric dielectric after annealing

Full size image

Various deposition methods are suitable for such fabrication. There are two primary approaches: chemical vapor deposition (CVD), and magnetron sputtering. In addition, Si QDs can also be prepared by other methods such as evaporation, ion implantation, pulsed laser ablation, and direct ion beam deposition. Table 1 summarizes the representative examples of the Si QDs growth parameters of the above mentioned techniques. It should be noted that the fabrication techniques for silicon quantum dots embedded in nitride or carbide matrices are very similar to those for SiO2-embedded Si QDs. The quality of material and the degree of phase separation depends not only on the amount of excess Si and its post-deposition processing, but also on the fabrication method [41]. The authors of [41] have compared different properties caused by fabrication methods [including sputtering, plasma enhanced chemical vapour deposition (PECVD), and ion implantation], and have presented some speculation on the reasons of difference. But still, a comprehensive understanding is missing. Furthermore, a comparison between CVD method and magnetron cosputtering is given in Refs. [42, 43].

Table 1 Deposition parameter summary of Si QDs embedded in a dielectric matrix
Full size table

Several kinds of CVD techniques, such as PECVD [44, 45], inductively coupled plasma CVD (ICPCVD) [46], hot-wire CVD (HWCVD) [47], laser-assisted CVD (LACVD) [48], and catalytic CVD (Cat-CVD) [49], have been used to prepare Si QDs. The necessity to use toxic and explosive gases (such as SiH4, CH4, and H2 etc.) and very time consuming in multilayer structure are the common disadvantages of CVD [50]. Process parameters like precursor gas, gas flow rates, chamber pressure, plasma power, and substrate temperature, etc. can significantly affect the film composition and quality, and the size of silicon crystallites [14, 38, 51–53]. Moreover, the film thickness is commonly a direct function of duration time of deposition [46].

From the “Precursor gas” column of Table 1, SiH4 and NH3 are often used as the precursor gases for depositing SRN, and SiH4 and CH4 are often used for SRC. In the case of SRO, it is noteworthy that N2O gas is used as the oxygen precursor in many works like [44–46, 54, 55] due to lower activation energy [56] and safety reason [57]. However, N is also incorporated into the film, which blue-shifts the optical bandgap and reduces the absorption efficiency [43]. While, the incorporated N leads to the requirement of a higher annealing temperature as compared to a nitrogen-free film due to the decreased mobility of Si in N containing SRO [57]. By suitably adjusting the flow rates of precursor gases, it is possible to make films with different silicon concentrations [52]. For example, for Si QDs in SiNx matrix grown by PECVD, with the increase of gas flow ratio of SiH4 versus NH3, the crystallization is enhanced, and the amount of dots increases, being demonstrated by the increase of X-ray diffraction intensity of nanocrystalline Si QDs peak (Fig. 3) [58]. Using grazing incidence X-ray diffraction (GIXRD), nanocrystalline Si QDs can be evidenced by the peaks corresponding to the (111), (220), and (311) silicon crystal plane [58]. The sub-stoichiometric Si-based dielectric materials, which can be SiOx, SiNx, SiOxNy, or SiCx, undergo phase separation and crystallization during the high-temperature (typically 1100 °C) annealing process, resulting in the formation of Si QDs within the dielectric matrix [44, 52, 59]. However, such high temperatures involve a huge thermal budget, and may not be suitable for device fabrication due either to technical reasons or cost considerations [59]. Recently, progresses for in situ formation of Si QDs in Si-rich silicon nitride matrix [49, 53, 60, 61] and Si-rich silicon carbide matrix [62, 63] by various CVD methods, such as PECVD, Cat-CVD, and ICP-CVD without annealing have been developed. However, the control of the size and the density of Si QDs using these in situ growth methods remains a challenge.

Fig. 3
figure 3

GIXRD results for silicon rich nitride films deposited at different gas flow ratios and annealed at 1000 °C for 1 h. Reproduced with permission from [58] © 2012 Elsevier

Full size image

Magnetron sputtering represents another suitable method for obtaining Si-rich films. Argon is excited and ionized, and high energy argon ions are used to sputter material from a target. The production of Si QDs materials are determined by several control control parameters of sputtering: the substrate temperature [64], sputtering power [65], sputtering rate [66], sputtering duration time [67], and reactive gases (N2, NH3, or O2) flow rates in the case of reactive sputtering [68–71] etc. Generally, silicon-rich thin films are deposited on room temperature substrates [36, 65, 66, 72] or on substrates heated at temperatures between 100 and 500 °C [64, 68, 73–75] and the as-deposited films are amorphous. Sputtering is often employed in conjunction with a high-temperature annealing process to form Si QDs in a dielectric matrix [50, 65, 66]. The most efficient annealing temperature for obtaining Si QDs is 1100 °C [16, 65, 73, 74].

Annealing of Si-rich silicide films or multilayers can be made either by rapid thermal processing/annealing (RTP/RTA) or in conventional tube furnace (CTF) [50, 59]. The annealing is typically performed in a gas atmosphere, e.g. N2 or Ar, and the gas atmosphere can influences the structural characteristics of films [44]. For example, by annealing SRO/SiO2 multilayer in N2, Si QDs with small size and low density are observed in comparison with the case of annealing in Ar when bigger QDs with higher density are obtained [44]. The effect of annealing in N2 on Si NCs formation is often explained in literatures by the suppression of Si diffusion produced by the presence of nitrogen and leading to smaller sizes [76–78]. RTP represents a suitable candidate as a cost- and time-effective technology in current Si-based processing [79]. The major feature of RTP is the rapid heating ramps (up to 100 K/s) and dwell times of seconds to minutes [80]. Hiller et al. [59] pointed that PECVD samples generally contain a significant amount of hydrogen. If the ramps are too fast, too rapid a heat build-up in the film will results in a blistering of the film because of the hydrogen does not effuse timely. Therefore, the heating ramps should be set reasonable. Wan et al. [50] have shown that RTA treatments can lead to a better degree of crystallization on Si nanocrystal and less amorphous Si residual in SRC thin films with respect to the tube-furnace annealing. However, in Ref. [80] no influence of the type of annealing (RTP or CTF) was observed on the formation and size of Si QDs in SiOx/SiO2 multilayers. There may exist a high density of defects at the interface between QDs and matrix, being higher in the RTP annealed multilayers [80]. A H2-passivation is needed to diminish the defects density [14]. Moreover, high-temperature RTP will also create significant stress in the material due to the fast temperature ramping rate and the thermal shock [36, 81]. Wan et al. and Huang et al. [28, 36] have reported that an additional high-temperature tube-furnace annealing could reduce the residual stress markedly.

3 Material performance improvements

3.1 Si QDs size control and density improvements

The growth of Si QDs can be affected by the barrier layer thickness, Si-rich layer thickness, Si excess, annealing condition, and dopant element [20, 52, 57, 84, 98]. For optoelectronic applications, a high level of size and density control of Si QDs is required to ensure efficient carrier transport [100]. This section reviews several past approaches to size- and density-controlled synthesis of Si QDs.

For example for SRO layers, an early method to obtain Si QDs entails deposition of a thick SRO (SiOx, x < 2) monolayer [45, 55, 64]. High-temperature post-annealing causes a phase separation between the two constituent phases, i.e. Si and SiO2, and subsequent thermal crystallization to form Si QDs embedded in a SiO2 matrix [31]. The size of Si QDs can be controlled by the chemical stoichiometry of the SiOx [101], film thickness [45], and annealing parameters [102]. The QDs size can be reduced by reducing the amount of excess Si. Nevertheless, the dot density is usually reduced simultaneously [70]. The low dot density leads to the disadvantage of poor electrical conductivity of the films. Experimental studies have shown that formation of Si QDs with an average size from 4.3 down to 3 nm from a SiOx monolayer film requires a stoichiometry x from 1.17 to 1.63 [103]. High-temperature annealing of single-layer thick films (such as SRO, SRN, and SRC) tends to form Si QDs with a broad size distribution and with weak control over spacing of Si QDs. It is indeed possible to fabricate spatially well-ordered, high-density Si QDs in a singly deposited SiOx layer with appropriate thickness and under favourable annealing conditions [45]. However, this approach is unsuitable to be used as a common approach, because it is difficult to achieve the optimized fabrication parameters.

The growth of Si QDs through the use of multilayer structures is an effective way to control size and spatial locations of Si QDs. High-temperature annealing of amorphous Si/SiO2 superlattices produces Si QDs, and the dot size is controlled at least in the direction perpendicular to the film surface [104]. The major drawback of this Si/SiO2 multilayer approach is that most of the crystallites touch each other, tend to form as bricks, and the resultant layers are more like a polycrystalline silicon film [105]. This severely decreases the luminescence efficiency because the grain boundaries often provide an effective route for nonradiative processes [70].

More promising results may be achieved by replacing the a-Si layer with a SRO layer in SRO/SiO2 multilayers. Zacharias et al. [70] first proposed this new alternative approach to form Si QDs with high density and narrow size distribution. SiOx/SiO2 superlattices were deposited by reactive evaporation onto a suitable substrate. The samples were then annealed at 1100 °C for 1 h under N2 atmosphere. Experimental results have shown that the desired dot size can be controlled by adjusting the thickness of the silicon-rich layer [70]. Figure 4 shows TEM and high-resolution TEM images of Si QDs formed in such a structure. Cho et al. [106] pointed that, under moderate annealing conditions, Si QDs in Si-rich layers as approximately spherical and with a diameter close to the SiOx sublayer thickness of the as-grown superlattices can be formed. Hao et al. [90] reported the dot size of Si QDs embedded in an oxide matrix as SRO/SiO2 multilayers decreases with increases in O/Si ratio of the SRO layer. The density of Si QDs in the SRO layer would be determined by the concentration of excess Si in the layer. It should be noticed that not always the higher O/Si ratio the lower Si QDs density obtained as common expectation. In the work of Kourkoutis et al. [15], it is shown that the Si QDs size, shape and arrangement in SRO/SiO2 multilayers can be governed by adjusting the SRO stoichiometry. They provided a 3-dimensional view of the Si QDs networks by EFTEM tomography. And the results shown that with decreasing Si concentration the QDs average size decreases, the QDs density increases, the size distribution narrows and the shape becomes more spherical [15]. This multilayer approach was adopted by many researchers [46, 107–109] to create superlattices for Si QDs solar cells.

Fig. 4
figure 4

Cross-sectional TEM images of SiO/SiO2 multilayer film: a multilayer structure of as-prepared SiO/SiO2 superlattice. The darker layers represent the SiO layers. b High-resolution TEM of the same film, the crystalline particles are circled for clarity. Reproduced with permission from [70] © 2002 American Institute of Physics

Full size image

Besides, another type of structure, multi-alternating layers of Si/SiO superlattice has been widely studied [97, 110, 111]. For instance, in Li et al.’s work [111], the morphology, electroluminescence (EL), and photoluminescence (PL) in annealed Si/SiO multilayers were reported and compared with single-layered one. Si NCs are formed within both Si and SiO layers after high-temperature annealing at 1100 °C for 1 h. The multilayered approach of Si/SiO is found to promote carrier transport and enhance the EL emission of Si NCs [111]. It is supposed that the Si interlayers act as extra carrier paths for carrier transport [111]. Similarly, multilayer such as Si/SiNx [112], SiNx/Si3N4 [85], SiCx/SiC [113], Si/SiC [114], SiOx/Si3N4 [115], and SiC/SiOx [116] also were researched.

Besides, gradient Si-rich oxide multilayer structure [16] was proposed to achieve super-high density Si QDs thin film formation while preserving dot size controllability for better photovoltaic properties. Recently, we have successfully fabricated gradient Si-rich SiNx multilayer structures [75]. It is proved again that this kind of structure is good at formation of high densities of Si QDs. In our recent work [10], we have proposed a light-filtering rapid thermal processing (LRTP) method, and fabricated high density Si QDs by use of this method. As a result, we have found that the samples treated with LRTP possess higher dot density, PL intensity, and conductivity than the conventional RTP samples. Despite of the obvious advantages of the LRTP method, little is known about the mechanism of growth of Si QDs under LRTP. The solution to this problem probably has to resort to some numerical simulation method such as Monte Carlo simulation.

3.2 Si QDs doping

For Si QDs-based solar cells, highly conductive Si QDs films are necessary [117]. By increasing carrier concentration and decreasing electrical resistivity via doping can be expected to improve the electrical properties of Si QDs films [12, 117, 118]. Moreover, fabrication of p–n junctions by doping with n- and p-type impurities is a key step to form QDs active layers in photovoltaic devices. However, it is proved that the doping of QDs is different from that of the corresponding bulk materials [119–121]. There are several difficulties about the impurity doping in a nanostructured semiconductor [122], including kinetically unfavourable nanocrystal doping caused by self-purification [120], large impurity formation energy [123], and the difficulty in direct quantitative characterization and tracking of the locations of dopants elements with nanometer-resolution [124]. Despite these difficulties, in practice, several researchers have reported successful doping of Si QDs films by several elements including phosphorus [20, 95, 125], boron [73, 126, 127], and antimony [128].

The incorporation of impurity elements in films containing Si QDs has been achieved through several means including sputtering [107], PECVD [21] and ion implantation [129]. In Hao et al.’s experiment [130], the phosphorus doping in the Si QDs superlattices was achieved by cosputtering of Si, SiO2, and P2O5 with a post-deposition anneal. The O/Si ratio and P concentration in the SRO layers were controlled by adjusting the sputtering powers of these three targets. They have found that the resistivity of the sample containing 0.1 at.% P is seven orders of magnitude lower than that of the undoped sample. They have attributed this dramatic decrease in resistivity to effective P-doping in Si QDs. The same authors also reported successful doping of similar layers with boron [117]. Boron doping at a concentration of around 0.5 at.% results in a reduction on the resistivity by six orders of magnitude. They have attributed this decrease in resistivity to the consequence of an increase in mobile carrier concentration. By employing the B-doped Si QDs/SiO2 multilayer structure deposited onto an n-type c-Si wafer, a high energy conversion efficiency of 13.4 % has been achieved by Kim’s group [124].

To achieve the desired properties in the photovoltaic applications, precise control of the position and concentration of dopant atoms at the nanoscale is crucial [124], and adequate analytical tools are essential in order to trace their properties. However, currently, several commonly used characterization techniques including X-ray photoelectron spectroscopy (XPS) and secondary ion mass spectroscopy (SIMS), can’t give exact positioning of dopant atoms. For instance, B has a very low relative sensitivity factor, which makes it difficult to detect by XPS [117]. Figure 5 shows the B1 s peaks of as-deposited and annealed SRO films with boron concentration of ~0.5 at.% in which the peak around 187 eV can be attributed to B–B/B–Si bonding [117]. In the case of SRN films, B was only identified with dopant concentration equal or larger than 1.0 at.% [126]. It should be mentioned that Auger electron spectroscopy is better suitable for the light elements [131] and therefore probably useful for B determination. SIMS was just used to obtain the relative concentration of impurities [124, 132]. In order to overcome the above difficulties, Conibeer et al. [133] suggested that further characterization of the exact positioning of doping elements can be sought from high resolution XPS, scanning transmission electron microscopy, atom probe tomography, and electron paramagnetic resonance. Besides, several characterization techniques based on synchrotron radiation would be useful to attempt in the future, such as soft X-ray absorption spectroscopy, X-ray absorption near-edge structure, which have been employed to probe the surface chemistry of Si nanostructures [134].

Fig. 5
figure 5

XPS B1 s spectra of as-deposited and annealed boron-doped SRO film (O/Si = 0.7) with boron concentration of ~0.5 at.%. Reproduced with permission from [117] © 2008 Elsevier

Full size image

In addition, the effects of some doping elements on growth of Si QDs are also worth noting. Hao et al. [135] found that incorporation of B suppressed Si crystallization and has little influence on the quantum dot size [135]. Phosphorus-doping could accelerate the phase separation in SRO layers and also the crystallization of Si NCs [136]. The presence of P atoms improves phase separation in Si-rich SiNx films and thus Si crystallization rate [20]. Yoon [137] has studied the effects of the addition of Ni on the growth of Si NCs by thermal annealing of Ni-implanted SiOx films with different content of Ni. It is established that the Ni impurity strongly effect the formation of Si NCs. Moderate Ni concentrations have been found to accelerate the processes of the formation of Si NCs, but the use of high Ni contents leads to the formation of NiSi2 NCs rather than Si NCs. The influence of Sn on the formation of Si NC in an amorphous SiOx (x ≈ 1.15) thin film is studied by Voitovych et al. [138]. It is found that the Sn dopant accelerates the processes of crystallization of a-Si clusters and the crystallization temperature could be lowered by 200 °C when adding Sn. Si NCs formed in the Sn-doped samples are smaller in size than those formed in the undoped films, while the volume fraction of the crystalline phase in the Sn-doped films is higher than that in the films free from Sn.

3.3 Defect passivation

Due to the high surface-to-volume ratio presented by the Si QDs size, issues related to the surface of Si QDs and the interface between Si QDs and dielectric matrix come to the fore [82]. Moreover, Si dangling bonds would increase proportionately as the film becomes more Si-rich [139]. As pointed out in Ref. [80], the present challenge is how to minimize the interface defect density which governs both optical [140, 141] and electrical [142] properties. By undertaking passivation step to reduce the high defect density in Si QDs films, it is believed that the performance of Si QDs solar cells can be significantly improved.

There is a strong likelihood that dangling bond, Pb, defects will occur at nanostructure/matrix material interfaces in general [56]. Zacharias et al. [143] summarized different Pb defects in the Si NCs/SiO2 system. Hydrogen passivation is regarded as an effective method for the passivation of electronically active Pb defects [143]. Many authors have studied the effect on the luminescence of hydrogen incorporation into Si QDs embedded in a dielectric matrix by annealing in a hydrogen-containing gaseous environment such as pure H2 gas [144, 145] or mixed H2/N2 gas [146, 147] or mixed H2/Ar gas [108, 148]. In addition, passivation treatments can be accomplished using atomic H plasma instead of hydrogen molecules in some experiments [149, 150].

It has been proposed that hydrogen passivation is related to the treatment temperature [108, 151]. The amount of hydrogen entering into the Si QDs film increases with treatment temperature increase. Meanwhile, dehydrogenation of hydrogen atoms terminating the dangling bonds becomes more prominent during a rise of temperature [151]. The steady-state passivation level is determined by the thermodynamic balance between defect passivation and depassivation [149]. The treatment temperature of 400 °C is considered as optimal in Ref. [151]. For Si-NCs/SiO2 multilayer thin films, passivation can also be performed by thermal oxidation [152] or exposing the films directly to the atomic oxygen plasma [153].

As for SRC/SiC multilayer, the host matrix consists of amorphous SiC and crystalline SiC phases after annealing, it is possible to contain a high defect density; a hydrogen passivation can be carried out in order to prepare an electronically high-quality film with less recombination-active defects [150]. In Ref. [154], it is shown that the as-annealed SixC1−x/SiC superlattices exhibited no PL but once the films were subjected to hydrogen passivation, which decreased the defect density as detected by electron spin resonance, PL emission was observed. Kořínek et al. [150] studied the effect of hydrogen passivation on the PL properties of the B-doped Si-NC/SiC multilayers. They found that hydrogen passivation leads to an additional increase in PL intensities. Ding et al. [155] reported the generation of additional defect states in SiC/SiOx hetero-superlattice (HSL) during the post deposition annealing. These additional defects are likely to result in a poor device performance of the HSL structure, and thus have to be passivated prior to application of HSL as a solar cell absorber. In their later work [116], the effectiveness of defect passivation methods such as hydrogen plasma in a PECVD reactor, hydrogen dissociation catalysis in a HWCVD reactor, and forming gas annealing (FGA) is examined. They suggest that FGA is the most promising method for the HSL sample passivation. In Di et al.’s work [108], the impact of forming gas (Ar:H2) anneal on the single junction Si QDs solar cells’ electrical and photovoltaic properties has been studied. After annealing in forming gas, the reduction of series resistance is largely (see Fig. 6) and this behavior is attributed to the hydrogen passivation at the interfaces of the Si nanocrystals [108].

Fig. 6
figure 6

Contact (Rc) and sheet (Rsheet) resistances as a function of annealing conditions. Reproduced with permission from [108] © The Authors 2010

Full size image

4 Photovoltaic devices

Si QDs have also already been incorporated in solar cells, as will be discussed below. Recently, photovoltaic devices with different structures, such as p-type Si QDs/n-type c-Si heterojunction, n-type Si QDs/p-type c-Si heterojunction, and p–i–n diodes have been successfully fabricated [124, 132, 151]. However, there have been few reports [156] on experimental realization of Si QDs-based ‘all-Si’ tandem solar cells mainly due to the difficulty in charge transfer between the QDs and the formation of tunnel junctions. In this section we summarize the photovoltaic studies with Si QDs and propose ideas for further exploration.

Table 2 summarizes the photovoltaic parameters of the representative examples of Si QDs photovoltaic devices with QDs as active material. Several characteristics can be drawn from Table 2 as follows: (a) The open circuit voltage obtained from these Si QDs solar cells is much lower than that expected for an absorber layer with a bandgap larger than that of c-Si. (b) The performance of Si QDs solar cells shown in Table 2 is limited by the relatively low short-circuit current density, which can be attributed to the poor carrier transport properties of Si QDs layers. (c) The active cell area was too small to learn the lateral conductivity and charge collection. (d) For photovoltaic properties, the absorption contribution from the Si QDs cannot be established because of the Si substrate absorbs many of the incident light and contributes to the photovoltaic effect. (e) The most efficient Si QDs cells prepared so far was a p-type Si QDs/n-type c-Si heterojunction solar cell with B concentration of 4 × 1020 atoms cm−3, which shows an efficiency of 13.4 % [124]. This corresponds to a short circuit current of 33.7 mA cm−2 (AM l.5), an open circuit voltage of 525 mV, and a fill factor FF of 78.5 %. Although these cell parameters are not comparable to crystalline silicon solar cell efficiencies, they can be viewed as a positive step towards realization of an ‘all-Si’ tandem cell based on Si QDs materials, as they demonstrate electrical conduction in Si QDs layers.

Table 2 Photovoltaic performance of some of the Si QDs solar cells produced in the laboratory by employing different device designs
Full size table

Usually, these Si QDs solar cells are fabricated on Si wafers. For example, Song et al. [157] reported that p-type Si QDs/n-type c-Si heterojunction devices showed an open circuit voltage of 463 mV, a short circuit current density of 19 mA cm−2, a fill factor of 53 %, and an energy conversion efficiency of 4.66 %. The internal quantum efficiency, external quantum efficiency measurements showed a higher blue response than that of a conventional c-Si single junction solar cell and was due to increasing the bandgap of Si-NC:SiC window layer [157]. However, in this device, the Si substrate masks the effect of the Si QDs layer in this device due to its superior electronic properties. Photovoltaic characterization of Si QDs films without any influence from Si substrate become a challenge. There are a few researches fabricating Si QDs photovoltaic device on an insulating substrate [25, 108, 158]. Such photovoltaic devices have been used for realizations of Si QDs solar cells without electrical wafer influence. Perez-Wurfl et al. [158] have fabricated p–i–n diodes onto quartz substrates with a total active area of 0.12 cm2. A schematic of such structure is shown in Fig. 7. As shown in the SIMS data, significant interdiffusion of B and P was occurred in the i-layer after high-temperature annealing and the structure was changed from p–i–n to p–n structure. The resulting p–n structure shows photovoltaic behavior, with an open circuit voltage of 373 mV. The same structure device with a total active area of 2.2 mm2 was re-fabricated in the same authors’ later study [25]. This device yielded 492 mV open-circuit voltage and 0.02 mA cm−2 short-circuit current density. The authors believe that the measured short-circuit current density is limited by a large series resistance of 28 kΩ cm2.

Fig. 7
figure 7

Schematic of the p–i–n structure photovoltaic device and SIMS profiles of as-deposited (solid line) and annealed (dotted line) p–i–n structure. Reproduced with permission from [158] © 2009 American Institute of Physics

Full size image

Fabrication a substrate-free photovoltaic device is another way to characterize the photovoltaic performance of Si QDs films without any influence from Si substrate. Löper et al. [9] have developed a membrane-based p–i–n device structure in which the substrate is locally removed by chemical etching and then the rest of substrate be encapsulated after solid-phase crystallization of the Si QDs. This device yielded 282 mV open-circuit voltage and 0.339 mA cm−2 short-circuit current density. The apparent weakness of this cell lies in the short-circuit current density; the experimental short-circuit current JSC is only 0.068 idealized maximum photogenerated current Jgen, indicating highly recombinative defects in the depletion region [9]. Although the photovoltaic performance is limited by charge carrier recombination, the measured values are directly related to the Si QDs layer, without being affected by the Si substrate [9].

In the fabrication of photovoltaic devices using Si QDs embedded in its silicide matrix, it is still a difficult task to improve current transport and to reduce carrier recombination [9, 107, 125]. Light trapping techniques will be needed for the development of Si QDs solar cells as the cell absorber layer is very thin. Further device optimization such as optimization of Si QDs growth parameters (film deposition conditions, interface engineering, material passivation, etc.), optimization of the composition of the cell’s absorber material, surface plasmons or texturing, and back-surface field are needed to improve the device performance.

5 Conclusions and outlook

We have provided an overview of the selected recent progress with respect to photovoltaic applications of Si QDs films. It has been seen that Si QDs films can be produced by different growth methods. There have been several strategies used to fabricate high density, size-controlled Si QDs with narrow size distribution, including using multilayer structure. With regard to solar cell applications, moderate doping and defects passivation are of high importance to enable better overall material performance. Photovoltaic devices based on Si QDs films have been fabricated by by several groups. This review work shows that Si QDs films has the potential of producing cheap solar cells, which have the possibility to overcome the Shockley−Queisser limit of conventional crystalline Si solar cells.

Although many efforts have been made on improving the conversion efficiency of Si QDs solar cells, it still remains lower than that for conventional crystalline Si solar cells. As far as we know, the best efficiency so far for a Si QDs cell is 13.4 % [124], which is about half of the best efficiency obtained for conventional crystalline Si solar cells (25.6 %) [22]. To further improve the cell performance, the majority of research regarding Si QDs solar cells focuses on three main aspects: (1) developing efficient methods to synthesize size-controllable Si QDs with high dot density, (2) good conductive properties, and (3) judicious and careful device optimization. But problems remain with precise control of doping elements, understanding doping mechanism, improving current transport and increasing Voc. The future research (such as the materials growth, defects passivation, device design, electrode contacts, efficiency of light trapping, etc.) of Si QDs highly relies on the understanding and precise control of doping elements and electronic transport mechanism. Though the current conversion efficiencies reported for the Si QDs solar cells are lower than the conventional crystalline Si solar cells, it is highly expected that further studies at the Si QDs solar cells will lead to a breakthrough that will enable us to realize high-efficiency photovoltaic devices in the near future.