Interfaces in amorphous/crystalline silicon heterojunction solar cells vorgelegt von M. Sc. Mathias Mews geboren in Gera von der Fakultät IV - Elektrotechnik und Informatik der Technischen Universität Berlin zur Erlangung des akademischen Grades Doktor der Naturwissenschaften - Dr. rer. nat. genehmigte Dissertation
Promotionsausschuss: Vorsitzender: Gutachter:
Prof. Prof. Prof. Prof.
Dr. Dr. Dr. Dr.
Christian Boit Bernd Rech Pere Roca i Cabarrocas Bernd Szyszka
Tag der wissenschaftlichen Aussprache: 30.Mai 2016 Berlin 2016
Contents 1 Introduction
1
2 Materials, interfaces and devices
4
2.1
Amorphous silicon . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4
2.1.1
Hydrogen in amorphous silicon . . . . . . . . . . . . . . . . . .
4
2.1.2
Amorphous silicon density of states . . . . . . . . . . . . . . . .
5
2.1.3
Doping in amorphous silicon . . . . . . . . . . . . . . . . . . . .
7
Amorphous/crystalline silicon heterojunctions . . . . . . . . . . . . . .
7
2.2.1
Amorphous/crystalline silicon heterojunction solar cells . . . . .
7
2.2.2
Interface properties . . . . . . . . . . . . . . . . . . . . . . . . .
10
2.3
Recombination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
12
2.4
Solar cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
13
2.2
3 Preparation, spectroscopy and simulation 3.1
3.2
15
Sample preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
15
3.1.1
Wafer pre-treatment . . . . . . . . . . . . . . . . . . . . . . . .
15
3.1.2
Plasma-enhanced chemical vapor deposition . . . . . . . . . . .
15
3.1.3
Hydrogen plasma treatments of a-Si:H . . . . . . . . . . . . . .
17
3.1.4
Liquid silicon precursors . . . . . . . . . . . . . . . . . . . . . .
18
3.1.5
Black silicon . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
18
3.1.6
Solar cell fabrication . . . . . . . . . . . . . . . . . . . . . . . .
21
Spectroscopy and device simulation . . . . . . . . . . . . . . . . . . . .
21
3.2.1
Transient photoconductive decay measurements . . . . . . . . .
21
3.2.2
Fourier transform infrared spectroscopy . . . . . . . . . . . . . .
23
3.2.3
Spectral ellipsometry . . . . . . . . . . . . . . . . . . . . . . . .
24
3.2.4
Photoelectron spectroscopy . . . . . . . . . . . . . . . . . . . .
25
3.2.5
Surface photovoltage . . . . . . . . . . . . . . . . . . . . . . . .
28
3.2.6
Auxiliary measurement methods . . . . . . . . . . . . . . . . . .
29
3.2.7
Device simulation . . . . . . . . . . . . . . . . . . . . . . . . . .
30
i
4 Hydrogen plasma treatments
31
4.1
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
31
4.2
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
32
4.3
Experimental details . . . . . . . . . . . . . . . . . . . . . . . . . . . .
33
4.4
Hydrogen plasma treatments of a-Si:H . . . . . . . . . . . . . . . . . .
34
4.4.1
Carrier lifetime spectroscopy . . . . . . . . . . . . . . . . . . . .
34
4.4.2
Diffusion profiles . . . . . . . . . . . . . . . . . . . . . . . . . .
36
4.4.3
Hydrogen density, mass density and band gap . . . . . . . . . .
38
4.4.4
Valence band spectroscopy . . . . . . . . . . . . . . . . . . . . .
40
4.4.5
Solar cell results . . . . . . . . . . . . . . . . . . . . . . . . . . .
41
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
43
4.5
5 Black silicon texture
44
5.1
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
44
5.2
Nanotextures and (i)a-Si:H passivation . . . . . . . . . . . . . . . . . .
45
5.3
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
47
5.4
Experimental details . . . . . . . . . . . . . . . . . . . . . . . . . . . .
47
5.5
Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . .
48
5.5.1
Reflectivity of black silicon surfaces and solar cells . . . . . . . .
48
5.5.2
Passivation layer optimization . . . . . . . . . . . . . . . . . . .
50
5.5.3
Quantum efficiency measurements . . . . . . . . . . . . . . . . .
53
5.5.4
Parasitic absorption in black silicon nanostructures . . . . . . .
55
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
55
5.6
6 Solution-processed amorphous silicon
58
6.1
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
58
6.2
Promise and challenge of liquid silicon . . . . . . . . . . . . . . . . . .
59
6.3
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
60
6.4
Experimental details . . . . . . . . . . . . . . . . . . . . . . . . . . . .
61
6.5
Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . .
62
6.5.1
From polysilane to amorphous silicon . . . . . . . . . . . . . . .
62
6.5.2
Valence band spectroscopy . . . . . . . . . . . . . . . . . . . . .
64
6.5.3
Liquid processed amorphous silicon passivation layers . . . . . .
65
6.5.4
Oxidation of liquid processed amorphous silicon . . . . . . . . .
67
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
69
6.6
7 Valence band alignment and hole transport 7.1
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii
70 70
7.2 7.3 7.4 7.5
7.6
SHJ valence band offset modification . . . . . . . . . . . Introduction . . . . . . . . . . . . . . . . . . . . . . . . . Experimental and simulation details . . . . . . . . . . . . Results and discussion . . . . . . . . . . . . . . . . . . . 7.5.1 SHJ valence band alignment and passivation . . . 7.5.2 Valence band offset and transport across the SHJ 7.5.3 Valence band stairwell . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
8 Prospects for silicon heterojunction solar cells 8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.2 Carrier selective contacts in SHJ solar cells . . . . . . . . . . . . 8.2.1 Charge selective contacts . . . . . . . . . . . . . . . . . . 8.2.2 ITO/(p)a-Si:H contact . . . . . . . . . . . . . . . . . . . 8.2.3 Hole contact materials . . . . . . . . . . . . . . . . . . . 8.2.4 Oxygen vacancies in tungsten oxide hole collection layers 8.2.5 Electron contact materials . . . . . . . . . . . . . . . . . 8.3 Back-contact back-junction solar cells . . . . . . . . . . . . . . . 8.3.1 state of the art . . . . . . . . . . . . . . . . . . . . . . . 8.3.2 Printable silicon wafer based BC-BJ solar cells . . . . . . 8.4 Crystalline silicon based tandem solar cells . . . . . . . . . . . . 8.4.1 General concept . . . . . . . . . . . . . . . . . . . . . . . 8.4.2 Tandems with perovskites . . . . . . . . . . . . . . . . . 8.4.3 Tandems with InGaP and related materials . . . . . . . 8.4.4 Summary of crystalline silicon based tandem concepts . . 8.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 Conclusions and outlook
. . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . .
71 73 75 76 76 77 79 80
. . . . . . . . . . . . . . . .
83 83 84 84 86 87 89 92 94 94 95 96 96 97 99 99 100 101
10 Appendix 104 10.1 Abbreviations and symbols . . . . . . . . . . . . . . . . . . . . . . . . . 104 10.2 Publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 10.3 Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
iii
Chapter 1 Introduction The first monocrystalline silicon solar cells with a p/n-junction were fabricated in 1954 [1]. Today more than 200 GW photovoltaic capacity are installed worldwide and the annual production capacity is about 40 GW [2]. 92 % of the global PV production and therefore most of the installed capacity are crystalline silicon based modules, with multi-crystalline silicon contributing 56 % and mono-crystalline silicon contributing 46 % of the overall market [3]. The first solar cells fabricated at Bell Labs only had an efficiency of about 6 % and featured a diffused p/n-junction [1] This technology gradually evolved by etching random pyramid structures on the wafer to suppress reflection [4], adding a front surface anti-reflection layer [5] and emitter passivation layer [6] and by applying locally diffused rear contacts to limit recombination [7], finally achieving 25 % efficiency in 1998 [8]. One feature of this technology is the direct contact of the diffused emitter and base contacts to the metal contacts, which limits the achievable output voltage of silicon solar cells with diffused junctions [8]. A different crystalline silicon based approach was presented by Sanyo in 1992 [9]. Instead of relying on a diffused p/n-junction the amorphous/crystalline silicon heterojunction (SHJ) uses doped and hydrogenated amorphous silicon (a-Si:H) layers to form the p/n-junction and the base contact. The use of doped a-Si:H layers instead of diffused dopants offers two advantages. First a passivation layer can be inserted into the junction. In the case of the SHJ this is intrinsic hydrogenated amorphous silicon ((i)a-Si:H), which reduces the defect density in the junction and enables SHJ solar cells to reach open circuit voltages of up to 750 mV [10], in contrast to diffused silicon homojunctions, which are limited to about 700 mV [8]. Secondly a-Si:H depositions are conducted at roughly 200 ◦ C, in contrast to diffused junction formation at about 1000 ◦ C. The SHJ technology offers higher efficiencies [11], better temperature coefficients [12] and lower life-cycle greenhouse gas emissions [13] than the the more wide spread silicon homojunction technology. Also, the fabrication cost per peak power gen1
2
CHAPTER 1. INTRODUCTION
eration of both technologies is comparable [14]. The development of SHJ solar cells was driven by record efficiencies from Sanyo (Panasonic) [9, 10, 15–18] with the latest record power conversion efficiency of 25.6 % [11]. Studies published by research institutes found that the amorphous/crystalline silicon interface needs to be devoid of epitaxial growth [19, 20] and that high hydrogen concentrations at the interface are beneficial [21]. Furthermore it is necessary to add a transparent conductive oxide (TCO) on the doped a-Si:H layers, since their conductivity is not sufficient for lateral carrier transport towards the metal grid fingers [22]. The interface between the commonly used TCO indium tin oxide and the p-doped a-Si:H is a tunnel-recombination junction [23] with non-ideal band alignment, which leads to inherent limitations of the fill factor [24, 25]. One disadvantage of SHJ solar cells is parasitic absorption in a-Si:H and TCO layers [26], which triggered work on rearemitter [27] and back-contact back-junction [28, 29] devices, to decrease the parasitic absorption. The work presented in this thesis relates to many of this points. Chapter 4 discusses a process designed to fabricate epitaxial free (i)a-Si:H passivation layers. Chapter 5 deals with the application of "black" silicon [30] as anti-reflection structure for SHJ solar cells. Chapter 6 discusses the application of liquid silicon precursors [31] for SHJ solar cells. Chapter 7 discusses the band alignment at the SHJ and its influence on hole transport. Chapter 8 discusses alternative contact schemes to overcome the fill factor limitations of SHJ solar cells and potential applications of SHJ in tandem devices [32] to overcome the current record of 25.6 % power conversion efficiency.
Structure of this thesis This thesis contains two chapters, with short summaries of the necessary basics to understand the later experimental results, four chapters detailing these results and a chapter discussing the status and future prospects of the SHJ technology. The four results chapters each are based on an article published in a peer-reviewed scientific journal. Two chapters include additional results, which were not part of the original publication. Chapter 2 gives a brief summary of the materials and device concepts relevant to this thesis and references to more detailed literature. Chapter 3 gives descriptions of the preparation, spectroscopy and device simulation techniques used in this thesis. Chapter 4 is based on a publication about hydrogen plasma treatments of SHJs [33]. A paragraph about related solar cell results is added to the already published study. Chapter 5 details a study on nanotextured substrates for use in SHJ solar cells [34].
3 The chapter starts with a short description of the changes in the approach presented in the preceding chapter and how they relate to this study and is followed by the expanded reprint of the actual study. Chapter 6 deals with the liquid silicon precursor neopentasilane and its application for the preparation of a-Si:H passivation layers [35]. Chapter 7 discusses the relation between the valence band offset at the SHJ and the solar cell transport properties [36]. Chapter 8 discusses future topics in the field of SHJ solar cells, connects the previously mentioned results with these future topics and presents additional results on tungsten oxide hole collection layers and tandem devices consisting of SHJ and perovskite solar cells. The thesis closes with a discussion of the achieved results and an outlook on the future research on silicon heterojunction solar cells.
Chapter 2 Materials, interfaces and devices 2.1
Amorphous silicon
Amorphous silicon [37, 38] is a solid form of silicon, consisting of silicon atoms with sp3 -hybrid orbitals and the corresponding close range order. Silicon-silicon bond lengths and orientations in amorphous silicon are statistically distributed and do not feature long range order. Amorphous silicon typically contains a large number of nanosized voids and unsaturated, so called dangling bonds. For the saturation of these dangling bonds hydrogen is introduced into hydrogenated amorphous silicon (a-Si:H) [39].
2.1.1
Hydrogen in amorphous silicon
Amorphous silicon used for electronic applications is usually hydrogenated, because of the significantly lower density of dangling bonds [40]. Hydrogen in a-Si:H is incorporated in different bonding configurations [39, 41]. Typical configurations are silicon atoms with one, or two bonds being saturated by hydrogen. Furthermore infrared spectroscopy has revealed microvoids filled by 6 or more hydrogen atoms. These are predominantly found in a-Si:H with hydrogen concentrations of about 14 at%, or higher. Additionally so called platelets [39] and molecular hydrogen [42] were reported. Hydrogen influences the optical and structural properties of a-Si:H. On the one hand hydrogen tends to saturate weak silicon-silicon bonds by breaking them and replacing them with silicon-hydrogen bonds. These weak silicon-silicon bonds constitute the band tails [43]. Replacing them by silicon-hydrogen bonds, therefore leads to an increase of the a-Si:H band gap, due to a shift of the valence band away from the Fermi-level [44]. These effects can be assessed using ab-initio-calculations [45, 46], optical spectroscopy [47–51], or photoelectron spectroscopy of the valence band [44] (cf. Chapter 4 and 6). A compilation of literature data about the influence of the a-Si:H 4
5
2.1. AMORPHOUS SILICON
Figure 2.1: Amorphous silicon band gap for layers with different hydrogen density. The graph contains data from the work of Schulze et al. [44], Ross und Jaklik [47], Matsuda et al. [48], Cody et al. [50] and Shiray et al. [51]. The line is a guide to the eye. Squares are samples produced by PECVD, circles mark sputtered samples and triangles are mixed data sets. hydrogen density on its band gap is shown in Figure 2.1. Most pulications report an increase of the band gap of about 10 to 15 meV per at% hydrogen. The only exception being Freeman and Paul [49], who report strongly scattered values and an increase of about 30 meV per at%. On the other hand pronounced inclusion of hydrogen in a-Si:H affects the statistical distribution of bond lengths and orientations, the structural quality of the material [44, 52]. This can be quantified using photoelectron spectroscopy of the valence band and valence band tail [53,54], or photothermal deflection spectroscopy [55]. However no direct relation between the Urbach energy and the hydrogen density could be found.
2.1.2
Amorphous silicon density of states
The density of states (DOS) of a-Si:H features valence and conduction band tails with exponential slopes [37]. The valence band tail follows: Nvt
Ev − E . = N0v exp E0v
(2.1)
6
CHAPTER 2. MATERIALS, INTERFACES AND DEVICES
Figure 2.2: Minority carrier lifetime at an injection level of 1015 /cm3 (τ15 ) of c-Si samples coated with (i)a-Si:H passivation layers and doped a-Si:H layers, as well as resistivity (ρ) of nominally identical doped layers on glass. The p-doping was varied by changing the TMB gas phase ratio (a) and the n-doping by changing the PH3 gas phase ratio. with Nvt the density of states in the valence band, the energetic position of the valence band edge Ev and the Urbach energy E0v of the valence band tail N0v . The band tail stems from the statistical distribution of bond angles and bond lengths in amorphous silicon [43, 56] and is called Urbach tail [53]. Additionally the dangling bonds in a-Si:H are reflected in Gaussian defect densities inside the bandgap. The dangling bonds are amphoteric states and exist in three different states [57]. They can be dangling bonds, which is a single valence electron. Alternatively they can collect either a hole, or an electron, leading to the formation of charged, or floating bonds. The occupation of these positive, neutral, or negatively charged dangling bonds depends on the Fermi-Level and is described by the DefectPool-Model [57]. The defect pool model implies that a post deposition change of the Fermi level in a given a-Si:H layer is bound to increase the defect density in this layer. This happens e.g. during deposition of p-doped a-Si:H layers on passivation layers, leading to a general decrease of the a-Si:H/c-Si interface passivation [58, 59]. Minority carrier lifetime values for samples with doped layers deposited on top of the passivation layer are shown in Figure 2.2. Increasing p-doping leads to decreasing minority carrier lifetimes (cf. section 3.2.1) and the overall passivation level for stacks of intrinsic and p-doped layers is lower, than for n-type. The deposition of a p-doped layer leads to stronger degradation of the intrinsic layers passivation, because intrinsic a-Si:H is
2.2. AMORPHOUS/CRYSTALLINE SILICON HETEROJUNCTIONS
7
actually slightly n-type [60], which results in a stronger shift of the Fermi-level after the deposition of a p-type layer.
2.1.3
Doping in amorphous silicon
Intentional doping of a-Si:H is necessary for device applications. Commonly for chemical vapor deposition this is facilitated by addition of phospine (PH3 ), for n-type doping with phosphorus and diborane (B2 H6 ), or trimethylborane (TMB) for p-type doping with boron [37]. For liquid silicon precursors like cyclopentasilane and other silanes doping is achieved by adding white phosphorus [61], or decaborane [62]. Not intentionally doped a-Si:H is called intrinsic. Intrinsic hydrogenated amorphous silicon ((i)a-Si:H) is very often slightly n-doped due to oxygen dopants [60]. Intrinsic a-Si:H can be deposited on c-Si wafers to passivate dangling bonds at the wafers surface [9], while p-type and n-type layers can be used to form carrier selective contacts. The doping efficiency in a-Si:H is low, because the amorphous network can easily accommodate 3-, or 5-fold coordinated atoms. Electron densities of 1017 /cm3 were reported in the conduction band of n-type a-Si:H grown with a gas phase concentration of phosphine of about 10 % [63]. The carrier density decreases again at higher phosphine concentrations, since a defect rich phosphor-silicon alloy is formed [63]. Furthermore doping increases the density of shallow defect states a) in the band tails and b) as neutral donors. This is worse for boron doped a-Si:H [63]. Overall phosphorous doped a-Si:H is similar to (i)a-Si:H, whereas boron doped a-Si:H has more structural defects and a lower band gap [64]. The reason is, that diborane is broken into BH3 , which features a strong dipole and removes hydrogen from the growing silicon surface [65]. This leads to an increased density of nucleation spots for silicon growth, increases the layer growth rate and thereby reduces the structural quality of the material [64].
2.2 2.2.1
Amorphous/crystalline silicon heterojunctions Amorphous/crystalline silicon heterojunction solar cells
The amorphous/crystalline silicon heterojunction (SHJ) solar cell with a thin intrinsic layer is a wafer based solar cell, which includes a p/n-junction formed by the deposition of e.g. p-doped a-Si:H on n-doped c-Si [9, 66]. Devices of inverse polarity are possible but uncommon, due to lower power conversion efficiencies [67]. Additionally a thin (i)a-Si:H layer is added in between the c-Si and the doped a-Si:H layers. This intrinsic layer passivates the c-Si surface and enables a very low defect density at the junction, which leads to a very high open circuit voltage of up to 750 mV [10]. This open circuit
8
CHAPTER 2. MATERIALS, INTERFACES AND DEVICES
voltage is about 50 mV higher than values reached by silicon homojunction solar cells and comparatively close to the maximum theoretical value of 769 mV [68]. The solar cell structure comprises the already mentioned n-type c-Si wafer of about 100 to 200 µm thickness with about 5 nm thick (i)a-Si:H passivation layers on both sides. For carrier extraction n-doped and p-doped a-Si:H electron and hole extraction layers are deposited with a thickness between 5 and 10 nm. The p-type hole contact is very often called emitter, minority carrier contact, or hole contact. Similarly the n-type a-Si:H base contact layer is called electron contact, majority contact, or back-surfacefield (BSF). Since the p-type a-Si:H has a comparatively low conductivity a transparent conductive metal oxid (TCO) layer, commonly indium-tin-oxid (ITO), is deposited on the front side of the solar cell. This layer doubles as medium for lateral carrier transport and as anti-reflection layer [26]. TCO and a-Si:H layers thicknesses in SHJ solar cells have to be optimized to minimize theirs parasitic light absorption [26]. This is especially critical for the p-type a-Si:H, as this layer is situated between the n-type c-Si and the n-type TCO on the front side. It has to be thick enough and highly doped to enable the formation of a proper p/n-junction [23], but as thin as possible to reduce its parasitic absorption. Alternatively it is possible to build rear-emitter SHJ solar cells in which the (p)a-Si:H/TCO contact is on the back-side of the solar cell. This puts the TCO layer with the lower conductivity [26] and the a-Si:H layer with the higher band gap [64] at the solar cell front-side and reduces parasitic absorption [27]. An additional TCO layer is deposited on the back-side of the solar cell to reduce light absorption by plasmons in the metal back contact [69]. Commonly ITO [69], or aluminum doped zinc-oxide (AZO) [70] are used. Finally the solar cell is finished by deposition of metal contacts. The front side is contacted with a metal grid and the back-side is either contacted with a grid to produce a bifacial solar cell [18], or with a full area metal contact. A sketch of the SHJ solar cell stack is shown in Figure 2.3 a and the band diagramm of the whole device is sketched in Figure 2.3 b. Carrier excitation takes place mainly in the n-type c-Si. Electrons diffuse through the wafer, the backside (i)a-Si:H passivation layer, the (n)a-Si:H electron collector, the TCO and are collected at the metal contact. The only barrier for electron diffusion is the conduction band offset at the SHJ of about 200 meV [71], which is overcome by thermionic emission. Similarly the hole extraction is hindered slightly by the valence band offset at the hole collecting SHJ [23]. Additionally the hole extraction includes a tunnel-recombination contact at the interface between the (p)a-Si:H hole collector and the front side TCO [23]. Investigations of the ITO//p)a-Si:H and AZO/(p)a-Si interface [24, 25, 72] found that the low work function of ITO and AZO induces a depletion layer in the (p)a-Si:H emit-
2.2. AMORPHOUS/CRYSTALLINE SILICON HETEROJUNCTIONS
9
Figure 2.3: (a) Scheme of a monofacial SHJ solar cell with front-side emitter. The light is coming in from the left. (b) Band line-up of an amorphous/crystalline silicon heterojunction solar cell with n-type crystalline silicon absorber. The thickness of the different layers is adjusted to match the stack in section (a) and are not to scale. Furthermore charge excitation is marked in green. The charge carrier flow is sketched and the tunnel-recombination junction at the ITO/(p)a-Si:H interface is marked in red. The Fermi-level (EF ), valence band maximum (VBM) and conduction band minimum (CBM) are shown.
10
CHAPTER 2. MATERIALS, INTERFACES AND DEVICES
ter, which is limiting the current extraction at this junction and thereby the fill factor of SHJ solar cells [24, 25]. Similar issues led to the application of high work function metal oxides [73] like molybdenum oxide (MoOx ) [74] and tungsten oxide (WOx ) [75] in organic solar cells and light-emitting devices. Therefore work on improving the band line-up at the TCO/aSi:H interfaces has started. Battaglia et al. demonstrated that MoOx forms a suitable heterojunction with (n)c-Si [76] and improved the efficiency using a SHJ solar cell with a MoOx hole collection layer [77]. Afterwards many groups started working on silicon based solar cells with MoOx and WOx [78–82] hole collectors.
2.2.2
Interface properties
Passivation of electronic defects Junctions of c-Si feature unsaturated silicon bonds at the interface. A-Si:H has proven effective in passivating these dangling bonds [83], since a huge percentage of those can be passivated by bonds to other silicon atoms [19,20] and because a-Si:H provides sufficient hydrogen to passivate the remaining dangling bonds [21]. A thermal treatment of the SHJ can lead to the reorganization of hydrogen close to the interface [84] and can therefore reduce the defect density at the interface. Two interface properties are regarded as important for good passivation, or low defect densities. First an increased hydrogen concentration at the SHJ to passivate the remaining silicon dangling bonds [21, 85]. Secondly partial epitaxial growth at this interface is detrimental to the surface passivation, because it offers additional recombination sites [19,20]. Therefore atomically sharp interfaces are required.
Band line-up The Fermi-levels in both the n-type c-Si wafer and the (i)a-Si:H passivation layer would be different, if the layers were not in contact. The same holds for the n-type c-Si and the p-type a-Si:H hole contact in a solar cell. This leads to a band bending between the passivation layer and the wafer of about 100 meV [44] and of about 800 to 500 meV across the junction between (p)a-Si:H and the (n)c-Si. The values depend on the doping levels in the respective layers. Sketches illustrating the band line-up at an (i)a-Si:H/(n)c-Si heterojunction and a (p)a-Si:H/(i)a-Si:H/(n)c-Si heterojunction are shown in Figures 2.4 and 2.5, respectively. Analytic techniques, which allow to measure band offset rely either on conductivity measurements [71], or on photoelectron spectroscopy measurements of the density of occupied states [86]. Techniques based on conductivity measurements enable the measurement of both band offsets, depending on the doping level and majority carrier type and the conduction band offset at the a-Si:H/c-Si heterojunction was determined to be about 200 meV [71].
2.2. AMORPHOUS/CRYSTALLINE SILICON HETEROJUNCTIONS
11
Figure 2.4: Band line-up at the heterojunction between (i)a-Si:H and (n)c-Si. The Fermi-level (EF ), conduction band minimum (CBM), valence band maximum (VBM) and the offsets at the respective bands (ΔEx ) are marked. Photoelectron spectroscopy in contrast is limited to the measurement of occupied states and necessitates ultra high vacuum equipment. The valence band offset at the a-Si:H/cSi heterojunction was determined to be about 450 meV for the heterojunction between doped a-Si:H and inversely doped c-Si substrates [86]. The valence band offset between (i)a-Si:H and n-type c-Si was found to be about 200 to 300 meV and varies depending on the hydrogen density in the layer [44]. Additionally the sum of the energetic difference between the band positions and the band bending adds up to the band offsets. The valence band offset ΔEV results from the line-up of the valence bands of a-Si:H and c-Si E c−Si and the band bending φ between them. E a−Si V V − E a−Si + eφ. ΔEV = E c−Si V V
(2.2)
Conduction band offsets are calculated accordingly. Some of these band offsets may be barriers for carrier transport. Especially the hole transport across the a-Si:H/c-Si heterojunction is still a matter of discussion. The band bending directs holes generated in the n-type c-Si towards the p/n-junction. This leads to an increased hole concentration close to the interface. ΔEv here is about 200 to 500 meV [44, 86], but due to their huge concentration and the built-in field of the junction carriers cross the junction. The exact carrier transport mechanism is still in question [23, 87, 88]. Most simulation studies rely on thermionic emission alone [89], but a few experimental [87, 88] and computational [23] works have found indications of tunneling across the interface and tunnel hopping in the a-Si:H layers. Chapter 7 provides an experimental and simulational investigation of this question, as well as a
CHAPTER 2. MATERIALS, INTERFACES AND DEVICES
12
Figure 2.5: Sketch of the band line-up at the heterojunction between a stack of p-doped and intrinsic a-Si:H and n-type c-Si. The possible transport paths for holes from the n-type c-Si absorber into the p-type a-Si:H hole contact are shown. more detailed discussion of the literature.
2.3
Recombination
The absorber material of a solar cell under light excitation features a higher charge carrier density n, than its intrinsic charge carrier density n0 . The excess charge carrier density is reduced by recombination R. Recombination is any process leading to the extinction of an excited charge carrier state. The time constant of recombination processes τ is [90]:
1 Δn 1 = = . τeff R(Δn, n0 ) i τi
(2.3)
Generally an effective charge carrier lifetime τeff is discussed, because different recombination mechanisms τi are affecting the charge carrier density. These different recombination mechanisms are: Auger recombination is similar to the Auger effect, since an excited charge carrier transfers his excess energy to another charge carrier in the electron gas and this energy is then transfered to phonon modes. The excited charge carrier then recombines with a charge carrier of inverse polarity [90]. This is an intrinsic recombination process and cannot be prevented.
13
2.4. SOLAR CELLS
Radiative recombination is the inverse of the photoeffect. Its probability increases with increasing minority charge carrier density. The density of excited minority charge carriers in silicon is comparatively low, due to its indirect band gap and common doping levels. Therefore radiative recombination is less important in silicon than Auger recombination [90].
Shockley-Read-Hall-recombination describes the relaxation of an excited charge carrier state via a defect state within the band gap [91]. Since the defect is situated within the band gap, it can collect and trap e.g. an electron from the conduction band and later emit it into the valence band. This is more effective for defects closer to the mid of the band gap, since those defects can most easily trap and emit charges from both bands. Relevant defects are extrinsic defects like impurity atoms, crystal dislocations, or defects and dangling bonds. Therefore the process is called extrinsic recombination, in contrast to Auger and radiative recombination, which are intrinsic properties of semiconductors [90]. SRH recombination is strong at surfaces, or interfaces, since these feature high dangling bond densities. Additionally those dangling bond defect states are situated roughly in the middle of the band gap and therefore constitute ideal SRH-recombination centers. Effective passivation of the c-Si surface is a cornerstone of the SHJ technology (cf. section 2.2) and is also central to the results discussed in chapters 4,5 and 6. The effective charge carrier lifetime of a semiconductor layer is the combination of all those processes 1 1 1 1 1 = + + + . τeff τrad τAuger τSRH τsurface
(2.4)
with the respective lifetimes for radiative (rad), Auger and SRH recombination.
2.4
Solar cells
Solar cells are diodes under light illumination, in which photo excited charge carrier are separated according to their charge and collected at respective contacts. The photovoltaic output of these devices is described using four parameters. The open circuit voltage VOC is the voltage generated by the illuminated diode at open circuit conditions. The jSC is the current density generated at short circuit conditions. The fill factor (FF) is calculated by FF =
VMPP · jMPP . VOC · jSC
(2.5)
14
CHAPTER 2. MATERIALS, INTERFACES AND DEVICES
Figure 2.6: IV-Curves of three SHJ solar cells with 4.5, 6.5 and 11 nm thick (i)a-Si:H passivation layers at the hole contact. with VMPP and jMPP as the voltage and current density at the maximum power point. Finally the efficiency η of a solar cell is calculated by dividing its power output by the incident light power. The maximum current density for an absorber with a bandgap of 1.1 eV is 46 mA/cm2 [92], for AMG1.5 irradiation with a power density of 1000 W/cm2 . The open circuit voltage of a solar cell is limited by the recombination current density j0 and the limiting intrinsic recombination channel in silicon is Auger recombination, which limits the achievable VOC to about 769 mV [68]. The best achievable fill factor was estimated using empirical expressions [93] and depends strongly on the solar cells VOC . It was estimated to be about 89 % for a VOC of 769 mV [94]. A solar cell with these parameters would have an efficiency of 31.48 % under AMG1.5 irradiation. The best real silicon solar cell ever made reached an efficiency of 25.6 % and is a SHJ solar cell in back-contact back-junction configuration [11]. Figure 2.6 shows the j(V)-curves of three SHJ solar cells with intrinsic amorphous silicon passivation layers of different thicknesses. The passivation layers of these cells have thicknesses of 5.0, 6.5 and 11 nm, fill factors of 77.1, 63.1 and 46.6 % respectively and show how sensitive SHJ solar cells are to the thickness of the intrinsic passivation layer. Commonly this layer has a thickess of 5 nm, but already a thickness of 6.5 nm leads to s strongly degraded fill factor.
Chapter 3 Preparation, spectroscopy and simulation 3.1 3.1.1
Sample preparation Wafer pre-treatment
Wafers were used as bought, some times cut into smaller pieces and subjected to chemical treatments of cleaning and sometimes structured to reduce their reflectivity. Wafers were cleaned using a process established by the Radio Company of America (RCA cleaning) [95]. This step was conducted for all samples. Additionally some samples were exposed to an alkaline KOH solution with the commercial product Alkatex to prepare random pyramid surfaces on the silicon wafer and reduce its reflectivity [96]. This step was omitted in many of the experiments described in this thesis. One reason is that planar untextured wafer surfaces provide a more defined experimental environment. Additionally textured surfaces are not accessible for many measurement techniques like UV photoelectron spectroscopy, or spectral ellipsometry. Also omitting any process step and especially the texturization simplifies the process, leading to better reproducibility, lower experimental errors and a higher process yield.
3.1.2
Plasma-enhanced chemical vapor deposition
Plasma-enhanced chemical vapor deposition (PECVD) is a method for fabrication of thin films. In this thesis a parallel-plate reactor with radio frequencey excitation (13.56 MHz, or 60 MHz) as shown in Figure 3.1 was used. Silane (SiH4 ) is used as silicon precursor, hydrogen may be added to tailor layer properties, diborane (B2 H6 ) and phosphine (PH3 ) are used as dopant sources and carbon dioxide (CO2 ) was added to fabricate silicon oxide layers. The substrate is placed on the grounded and heated 15
16
CHAPTER 3. PREPARATION, SPECTROSCOPY AND SIMULATION
Figure 3.1: Sketch of parallel-plate PECVD reactor.
electrode and the dissociated precursor gases form a layer on the substrate and in the rest of the chamber. Dissociation of the molecules with electron impact allows to grow layers below the thermal dissociation temperature of the precursors. This is especially important for a-Si:H, since hydrogen in a-Si:H gets mobile at about 200◦ C and starts to diffuse out of the layer [58], but thermal dissociation of silane happens at about 400◦ C [58]. This renders thermal chemical vapor deposition of a-Si:H with monosilane precursors effectively impossible. The promise and problem of PECVD is the amount of process parameters, their interactions and their influence on plasma and layer properties. The deposition process of a-Si:H and its alloys was investigated in great detail for the application in amorphous/microcrystalline silicon thin-film solar cells. One big difference between both technologies is the substrate for a-Si:H growth. SHJ solar cells use a wafer substrate, whereas thin film cells were deposited on glass. This leads to different growth processes and changes the connection between plasma properties, or deposition parameters and layer properties [97]. As already discussed (cf. section 2.2) (i)a-Si:H passivation layers need to be grown without unintentional epitaxial growth [19,20] and have to contain high hydrogen densities at the interface [21]. The following summary gives a list of process parameters and their influence on the properties of (i)a-Si:H passivation layers and the SHJ. Increasing the hydrogen precursor gas fraction does not necessarily lead to an increase of the hydrogen density in the layer. The layer’s hydrogen density increases with increased hydrogen precursor gas fraction up to a gas fraction of about 50 %. Higher hydrogen fractions lead to a denser layer and lower hydrogen concentrations [98]. Finally for even higher hydrogen concentrations epitaxial growth starts and the interface passivation decreases drastically [99]. The plasma power density is mainly driving
3.1. SAMPLE PREPARATION
17
the dissociation of the precursors [100] and therefore the deposition rate. Increasing the deposition rate directly reduces the probability of epitaxial growth [19] and leads to layers with higher hydrogen concentrations and more voids. The gas pressure is also directly influencing the deposition rate. A higher gas pressure leads to a highter deposition rate, a lower risk of epitaxial growth and a less dense layer with a higher hydrogen density [101]. The substrate temperature influences the diffusion velocity of molecules on the sample surface and therefore increasing substrate temperatures lead to more rigid Si-Si networks with less hydrogen filled voids, or dihydrides [102]. At some point the temperature is high enough to initiate epitaxial growth and the surface passivation decreases drastically [19]. The substrate surface is influencing the initial growth stages. It has a strong influence on the thickness of the incubation layer [103]. The c-Si surface orientation is important for the growth of a-Si:H on c-Si, as epitaxial growth happens more frequently on (100) than (111) oriented surfaces [99]. The reason is the higher density of open bonds and the therefore higher free energy of the (100) surface. All in all the growth of a-Si:H with PECVD is well understood.
3.1.3
Hydrogen plasma treatments of a-Si:H
As already outlined, hydrogen is important for the electronic and structural properties of a-Si:H. Therefore multiple groups have used hydrogen plasma treatments (HPT) of a-Si:H. One application is layer-by-layer growth of microcrystalline silicon layers [104]. In this process the HPT is the energy source for the crystallization [105]. Also hydrogen atoms break weak silicon-silicon bonds, which are replaced by stronger bonds [106]. Another approach is to hydrogenate a-Si:H with a HPT. This was applied to a-Si:H layers prepared by sputtering [51]. It was found that a short, or soft treatment increases the structural quality of the material and the hydrogen density [51]. In contrast long treatments lead to a decrease of the hydrogen density [107] and crystallization. Additionally a-Si:H can be etched [108] by a HPT. Recently HPTs were applied to a-Si:H, produced by thermal conversion of polysilane and with low initial hydrogen concentrations of only a few percent [109]. A similar process was applied to treat passivation layers for SHJ solar cells [110]. Using a HPT it was possible to increase the passivation of SHJs further, although these layers already contained large hydrogen concentrations. It was found that the HPT mainly increases the density of dihydrides and hydrogen filled voids in these layers.
18
3.1.4
CHAPTER 3. PREPARATION, SPECTROSCOPY AND SIMULATION
Liquid silicon precursors
Silicon is the material of choice for many electronic applications. Still solution processing of silicon has not been investigated in depth. The most common approach is to synthesize small silane molecules like cyclopentasilane [31,61], neopentasilane [111,112], or trisilane [113] and dissolve these using organic solvents like toluene [31], or cyclooctane [114]. A silicon polymer can be formed by ring-opening, or bond breaking due to photo induced polymerization by UV light [31], or by thermal polymerization [112]. Polydihydrosilane, the polymer derived from cyclopentasilane, includes mainly SiH2 groups, whereas the polymer derived from neopentasilane includes about 70 % SiH3 groups, which enable cross linking during the later thermal conversion to amorphous silicon [112]. Furthermore there are numerous differences between both processes, especially regarding yield and separation and recycling of by-products [112]. Still the general process flow, as depicted in Figure 3.2 is similar. These liquid silicon precursors allow for two new approaches regarding silicon processing. First cyclopentasilane and neopentasilane have boiling points of about 150◦ C [31] and therefore allow for chemical vapor deposition of a-Si:H and c-Si with much higher deposition rates [62, 115]. Secondly these materials can be applied by ink-jep printing [31], which potentially enables to fabricate laterally structured silicon devices without the use of complex structuring processes. This is especially interesting for backcontact back-junction solar cells [11], which have the contacts for both polarities on the same side. After layers of polysilanes have been deposited by e.g. spin-coating they have to be converted to a-Si:H. This is done by thermal conversion in nitrogen atmosphere. Unfortunately the a-Si:H network is formed at about 330◦ C [114], whereas hydrogen starts to diffuse out of a-Si:H above 200◦ C [114]. Therefore a-Si:H layers processed in this way contain low hydrogen densities, high defect densities and a high concentration of voids [109, 114]. Hydrogen plasma treatments of these layers can increases their hydrogen density and decrease their void density [109].
3.1.5
Black silicon
Common anti-reflection structures in solar cells are anti-reflection coatings and micrometer sized light scattering structures [26]. A third possibility is a layer with a graded reflective index. Such layers enable effective reflection values below 2 % for the visible part of the spectrum [116]. The different methods producing such structures produce a silicon surface with spike like structures, which lead to gradually increasing density of silicon in air along the light path. The most common ways to fabricate
19
3.1. SAMPLE PREPARATION
neopentasilane
cyclopentasilane
SiH3
H2
Si
SiH2
H2Si
Si
Si
H2
H2
H3Si
Si
SiH3
SiH3 thermal/photo polymerization
polydihydrosilane SiH3
SiH2
SiH2
SiH2
SiH2 SiH2 Si H3Si
SiH2 SiH 2
H3Si SiH2
Si
SiH2
SiH3 n
n coating and thermal conversion / chemical vapor deposition
SiH
Si
Si amorphous silicon
Si
SiH Si
Si
Si
-
SiH Si
Si
HSi
Si
Si
SiH SiH
Si
-
SiH2
Si
Si
HSi
HSi
Si
Si
Figure 3.2: Process flow for the conversion of liquid silicon to amorphous silicon. Small silane precursors like cyclopentasilane, or neopentasilane are polymerized e.g. by irradiation with UV light and dissolved in organic solvents. Afterwards this then liquid precursor is applied to the sample by either coating the sample and thermally converting the layer, or by thermal CVD, and amorphous silicon is formed.
20
CHAPTER 3. PREPARATION, SPECTROSCOPY AND SIMULATION
Figure 3.3: Metal assisted etching of a silicon surface. (a) A metal particle (eg. gold, or silver) is deposited on a silicon surface in a solution of HF and H2 O2 . The hydrogen peroxide in the solution is reduced at the metal surface and transfers a hole into the metal. This hole then diffuses into the silicon and reduces the silicon. (b) The hole diffusion into the silicon leads to oxidation of the silicon surface and this oxide is etched away and SiF6 is formed, while the metal particle is penetrating deeper into the silicon. (c) Holes diffusing from the metal into the silicon lead to etching of silicon along the metal particle’s path. (after a picture from the dissertation of N. Geyer [122])
such structures are metal-catalyzed etching using gold [30, 117], or silver [118, 119] and reactive ion etching [120, 121]. Although these processes lead to different surface morphologies all these structures are called black silicon. In this thesis samples prepared by metal-assisted etching using gold nanoparticles as catalyst were used. This method is sketched in Figure 3.3. In general a solution of HF and H2 O2 is used to etch silicon. Hydrogen peroxide is reduced to oxidize silicon and the resulting silicon oxide is then etched away by HF. The basic idea about metal-assisted etching is that this etching process can be enhanced from etch rates of a few nanometers per hour to a few hundred nanometers per minute by introducing a metal catalyst. This metal must have a higher electron affinity than silicon to be able to collect holes from the solution and then transfer them to the silicon anode (Figure 3.3a). Applicable metals are gold, silver, platinum and also aluminum. During etching the silicon is etched predominantly below the metal catalysts, leading to the formation of tubes, or cones below metal particles. Additionally holes are diffusing from the particle into the material and lead to the formation of porous silicon around the tubes, or cones. The etched shapes depend on the used metal and solution. Gold for example leads to spike shaped structures, since a high concentration of holes is injected into the material and leads to further etching of the tubes. The general etching direction depends on the substrate surface and the catalyst coverage, but tends to follow either the , or the planes [123]. An exemplary formula for the reactions involved in metal-assisted etching is Si + H2 O2 + 6HF → 2H2 O + H2 SiF6 + H2
(3.1)
3.2. SPECTROSCOPY AND DEVICE SIMULATION
21
Detailed reviews on this subject are available in the literature [118, 122].
3.1.6
Solar cell fabrication
The structure of the SHJ solar cell was already discussed in section 2.2.1. Apart from the a-Si:H depositions it is necessary to deposit TCO layers and metal contacts. Furthermore the solar cell area has to be defined, to exclude electrical connection of the active area with not illuminated areas during measurement. About 75 to 80 nm of RF-sputtered ITO constitute the front contact. The thickness is optimized for anti-reflection purposes and the conductivity is optimized to enable sufficient lateral conductivity, while ensuring a high transparency [26]. The front contact is then finished by the deposition of a 1.5 µm thick silver contact grid with a 10 nm thick titanium adhesion layer between silver and ITO. The back contact either consists of a second ITO layer, or an AZO layer. This back side TCO layer is necessary to decouple the light reflected at the backside from plasmons in the metal back contact [26] and to prevent alloying of the aluminum silicon contact [124], which happens at about 150◦ C and removes the passivation of the SHJ [125]. The back contact is finished by thermal evaporation of a 500 nm thick silver contact, which includes an adhesion layer if the TCO is ITO. For AZO this adhesion layer is not necessary. The cell area was defined either by sputtering the front-side ITO through a mask, or by photo lithography and etching of the ITO.
3.2 3.2.1
Spectroscopy and device simulation Transient photoconductive decay measurements
Quasi-steady state photoconductance decay (QSSPC), or transient photo conductance decay (TRPCD) measurements allow to measure the injection dependent carrier lifetime in samples with known electrical parameters and conduction mechanism. Both methods share the same working principle. A flash lamp is used to generate free charge carriers in the sample. The sample is placed above an inductive coil, which is used to measure the time dependent conductivity of the sample. The injected carrier density is then calculated from its conductivity and since the carrier density decays over time, the injection dependent carrier lifetime can be measured. It is possible to calculate the carrier lifetime (τ ) according to equation 3.2, if the light flash is decaying at least a few
22
CHAPTER 3. PREPARATION, SPECTROSCOPY AND SIMULATION
Figure 3.4: Typical injection dependent minority carrier density for an a-Si(n)/aSi(i)/c-Si(n)/a-Si(i)/µc-Si(p)-sample as measured with TRPCD.
orders of magnitude faster than the carrier lifetime. τeff (∆n) =
σ(t) jph (µn + µp )
(3.2)
This is then called TRPCD. Therefore the conductivity of the sample (σ), the mobilities of holes and electrons in the material (µn , µp ) and the excitation current density have to be known. If the decay time of carrier density and light flash are similar, the additional excitation of further carriers due to the light pulse has to be taken into account [126], which leads to the QSSPC technique. Furthermore the excitation current density has to be calibrated to the light intensity. This is facilitated using a reference solar cell. A typical measurement is shown in figure 3.4. The decrease of the carrier lifetime at high injection is due to Auger recombination and the scatter at low injection densities is due to higher noise. A fit of this curve with a semi analytical model allows to estimate the interface defect density and the interface charge at the interface in question [127]. It is thereby possible to discriminate between chemical passivation and field effect passivation [126]. Samples for photoconductive decay measurements are ideally symmetrically processed and must not be convered with metal, to allow for inductive coupling. A WCT-100 setup was used for QSSPC and TRPCD measurements. The excitation carrier density of holes (∆p) and electrons (∆n) measured during photoconductive decay measurements depends on the splitting of the quasi Fermi levels under illumination. This splitting is an upper boundary for the VOC of a solar cell. The VOC potential of a sample can be estimated from a carrier lifetime measurement
23
3.2. SPECTROSCOPY AND DEVICE SIMULATION using
kT ∆n(NA + ∆p) VOC = ln +1 . q n2i !
(3.3)
This value is commonly called implied VOC [126] and it is the maximum achievable VOC for a sample with the corresponding lifetime.
3.2.2
Fourier transform infrared spectroscopy
Infrared spectroscopy is a tool for the characterization of vibrations of chemical bonds [128]. In the context of a-Si:H/c-Si solar cells this includes the Si-H-vibrations [39]. Different Si-H microstructures like monohydrides and dihydrides can be identified using infrared spectroscopy [129]. The transmission T of a sample is calculated by dividing the absorption spectrum of the sample by the absorption of an uncoated reference. Suitable substrates have low absorption in the infrared. Crystalline silicon with a high resistivity is recommendable, as free carriers absorb infrared light. HF-dipped wafer surfaces are terminated by a high concentration of Si-Hx bonds at their surface [130]. A measurable concentration of these bonds remains even weeks after the HF-Dip. Therefore the reference wafer was kept in ambient air for at least 2 months before the measurements. The Fourier-Transformation includes taking an interferogram and converting it to a spectrum of optical absorption versus wavelengths afterwards. This step reduces the noise level. One problem with transmittance infrared measurements on wafers is that c-Si in between the symmetric a-Si:H layers constitutes a resonator [131], which leads to a harmonic component in the spectrum. This harmonic can be reduced by measuring in Brewster angle configuration with polarized light [132]. Furthermore the thickness difference between the reference wafer and the sample leads to the introduction of a second visible harmonic while calculating the transmission. Both harmonics can be cut from the signal with a Fourier filter [133]. Part of a FT-IR measurement before and after the Fourier filter is shown in Figure 3.5. The harmonic with the periodicity of about 5 wavenumbers results from the resonanance in the c-Si. The harmonic with a periodicity of a few hundred wavenumbers results from the superposition of the signals of reference and sample. The hydrogen concentration NH in a-Si:H can be calculated using proportionality factors A following [129] Z α dω (3.4) NH = A · I = A · ω with the wavenumber ω, the intensity I. The stretching vibration with the proportionality factors A2000 = 9 ± 1 × 1019 cm−2 und A2100 = 2, 2 ± 0, 2 × 1019 cm−2 for the modes
24
CHAPTER 3. PREPARATION, SPECTROSCOPY AND SIMULATION
Figure 3.5: FT-IR transmission signal of an (i)a-Si:H/c-Si/(i)a-Si:H sample obtained against a c-Si reference with harmonics. The result of applying the Fourier-filter is shown as a blue line. at 2000 and 2100 cm−1 [129] is commonly used for this analysis. The mode at 1980 to 2010 cm−1 is called low-stretching-mode (LSM) and is attributed to monohydride vibrations [134]. The signal at 2060 to 2160 cm−1 is called High-Stretching-Mode (HSM) and is attributed to dihydride vibrations [132], clusters of mono- and dihydrides [129], or voids filled with dihydrides [41]. The HSM is pronounced in a-Si:H layers of low mass density [41]. In this thesis the signal of dihydrides at the surface was fitted on its own and excluded from the calculations of the hydrogen density in the layer, since the layer thickness is very small. A Bruker IFS125HR spectrometer with an RT-DLa TGA detector, a KBr prism and a spectral resolution of 4 cm−1 was used throughout this thesis.
3.2.3
Spectral ellipsometry
Ellipsometry measures the ellipticity of circularly polarized light after its reflection from a sample surface [135]. The change in polarization upon reflection is described by the two angles ψ and ∆, which are calculated using the tilting angle αt and the ellipticity γ with cos ψ = cos 2α cos 2γ (3.5) tan ∆ = ±
tan 2γ sin 2αt
(3.6)
3.2. SPECTROSCOPY AND DEVICE SIMULATION
25
with the light wavelength λ, the film thickness d and the refractive index n. The extinction k of a sample can be calculated using its absorption α k=
λ α. 4π
(3.7)
Spectral ellipsometry enables to deduce the dielectric function of some materials. The model for the dielectric function of amorphous semiconductors like a-Si:H includes a Tauc fit of the band gap [136] and a Lorentz oscillator to account for absorption of light by the solid body. This model [137, 138] includes the amorphous semiconductors band gap Egap and four fitting parameters for the Lorentz oscillator. Among those the broadening parameter and the amplitude of the oscillator, were connected to the structural disorder in the layer [139]. The combination of the atomic hydrogen density cH as measured with FT-IR and the refractive index as measured with spectral ellipsometry enables to calculate the mass density ρ of a-Si:H [140], with ρ=
3mSi n2∞ − 1 × C 2 H n∞ + 2 4π[2αSi−Si + 1−C (αa−Si−H − 0, 5αSi−Si )] H
(3.8)
and αSi−Si = 1, 87 × 10−24 cm−3 und αa−Si−H = 1, 96 × 10−24 cm−3 for the bonding polarization strength of Si-Si and Si-H bonds, respectively and mSi = 28, 085 u = 4.66 × 10−23 g the atomic mass of silicon and ρa−Si = 2, 287 g/cm3 density of c-Si. In this this thesis spectral ellipsometry between 190 and 850 nm is used and all samples are measured at three angles (50,60,70◦ ). A surface roughness layer of about 0.5 nm thickness was included in the model of each sample. This layer was assumed to be an effective medium [141] consisting of equal parts of air and the topmost layer. An example for the refractive index and extinction of two (i)a-Si:H layers deposited at different substrate temperatures is shown in figure 3.6
3.2.4
Photoelectron spectroscopy
Photoelectron spectroscopy (PES) is a collective term for a number of vacuum based spectroscopy methods. In most cases photons are directed at a sample and excite electrons above the vacuum edge. These electrons are then collected in an energy dispersive analyzer and a spectrum of electron kinetic energies is collected. This measurements are usually conducted in ultra-high vacuum and the excitation is done using ultra-violet light, or x-rays. Three different types of PES have been used in this thesis and are discussed in the following.
26
CHAPTER 3. PREPARATION, SPECTROSCOPY AND SIMULATION
Figure 3.6: Refractive index n and extinction k of two (i)a-Si:H samples. The only difference between their deposition conditions was the deposition temperature. X-ray photoelectron spectroscopy X-ray photoelectron spectroscopy (XPS) is commonly used to analyze the chemical structure of the first few nanometers of a sample. X-ray photons with energies of about a few keV are used to excite core level electrons from the sample and their binding energy is then analyzed [142]. Typical applications are to analyze the fraction of different oxidation states of a given molecule in a sample [143], measure thicknesses of layers of a few nanometer thickness [144], or identify compound fractions [145]. In this thesis Al-Kα (1486.7 eV) and Mg-Kα (1253.7 eV) radiation were used as excitation source. Constant final state yield spectroscopy Constant final state yield spectroscopy (CFSYS) describes a measurement mode during which the energy analyzer is kept at a constant energy, the constant final state and the excitation energy is varied [146]. If the incoming light flux and the reflectivity of the sample are measured, the analyzer signal can be directly related to the sample’s density of states. At HZB a Xenon arc lamp is used for excitation [147, 148]. Its broad spectrum is cut off at about 7.3 eV by the suprasil beam splitter used to monitor the incoming light flux and by LiF windows. This energy range is sufficient to measure the complete a-Si:H valence band. Furthermore an excitation energy of 6 to 7.5 eV leads to a signal originating from the first 5 to 10 nm of the layer, which is less affected by surface states.
27
3.2. SPECTROSCOPY AND DEVICE SIMULATION
The model of a measured a-Si:H density of states (N (E)) consists [147, 149] of the linear conduction (Nc (E)) and valence bands (Nv (E)), the Urbach tails for the valence band Evt − E (3.9) Evt = N0v exp E0v and the conduction band Ect = N0c exp
E − Ect E0c
(3.10)
and a Gaussian defect density in the band gap (E − ED )2 ND (E) = ND exp − . 2 2σD !
(3.11)
An example of a measured valence band density of states of a-Si:H is shown in figure 3.7. The valence band, its exponential tail and the Gaussian defects in the band gap are fitted and shown. The different dangling bond states proposed by the defect pool model (cf. section 2.1.2) cannot be identified.
Figure 3.7: Measured density of occupied states of intrinsic a-Si:H (black). The valence band tail is fitted using an exponential decreasing funtion (blue) and a gaussian defect density (red) was used to fit the occupied danling bond states in the band gap.
UV photoelectron spectroscopy The most common light source for ultra-violet photoelectron spectroscopy (UPS) is the He I line with a excitation energy of 21.22 eV. UPS is very surface sensitive, in this energy range, and only measures a few atom layers. Therefore He-UPS is more
28
CHAPTER 3. PREPARATION, SPECTROSCOPY AND SIMULATION
suited for the analysis of surfaces than of layers. He-UPS was used throughout this thesis to measure the position of valence bands, which cannot be resolved using the 7 eV excitation used for CFSYS (cf. chapter 7) and is typically used to determine work functions of metal oxides from the secondary electron cutoff [142].
3.2.5
Surface photovoltage
Surface photovoltage (SPV) allows to measure different heterojunction properties [150]. It can be used to characterize the band bending (φ) in a heterojunction [44], to quantify a carrier-selective junction’s built-in field [80] and qualitatively measure the passivation of a surface [151]. Furthermore it is possible to measure the defect density at a surface by voltage dependent measurements [152]. The SPV setup at the HZB uses the following principle [153]: The sample’s back side is grounded on one electrode and a thin isolator (e.g. mica) is placed on the coated side of the sample. A second electrode on top of this isolator enables capacity measurements of the sample. Then a light pulse is directed through the top electrode and into the sample. The light wavelength is chosen to prevent strong absorption in possible coatings, but ensure strong absorption in c-Si. At HZB a 905 nm laser is used. The light pulse then leads to a strong increase of the carrier density in the substrate and if the intensity is sufficient, any band bending in the substrate is equalized by the light pulse. Once the generated carrier have been conducted out of the sample, or have recombined, a voltage signal is generated. This voltage signal has to be corrected for the Dember voltage to calculated the band bending in the measured heterojunction. Although SPV is a easy and fast measuring method, which allows to analyze many important heterojunction and surface parameters, there are a few disadvantages to the method. First SPV results are always more qualitative and less quantitative. Figure 3.8 shows band bending values measured on finished solar cells plotted versus the maximum (red) and average (black) open circuit voltages measured on these solar cells. This data shows a systematically higher VOC than band bending. Possible reasons are insufficient carrier excitation within the c-Si, due to insufficient light intensity, reflection at the sample, or parasitic absorption in the layers of the samples. Another more fundamental problem is that the derivations for the calculation of the band bending from the SPV voltage were conducted using Boltzman statistics [153] and for higher band bendings these become increasingly inaccurate as the Fermi-level is closer to the bands during the measurement. Therefore measurements of higher band bendings of some hundred meV are not suitable for quantitative analysis. Still there is a direct correlation between the measured band bending and the final VOC . Therefore band bending measurements can be used for optimizations and process control, but not for
3.2. SPECTROSCOPY AND DEVICE SIMULATION
29
Figure 3.8: Band bending in SHJ solar cells with different emitter (MoOx , (p)a-Si:H) and base contact (Titanium oxide, (n)a-Si:H) materials and the open circuit voltages reached with the same solar cells. The line is a guide to the eye. quantified analysis. Note that SPV measurements are used to measure the band bending between (i)a-Si:H and (n)c-Si in the works of Schulze et al. [44] and in chapters 4 and 7. The error made in those measurements is acceptable since the total band bending at this junction is always smaller than 150 meV and the comparatively small error in the band bending determination is insignificant, especially compared to the much larger error of the PES measurements used to measure Ev .
3.2.6
Auxiliary measurement methods
Some additional measurements methods have been used during this thesis. Only short summaries of those are given, since they are not central to the results of the thesis, or very basic. Secondary ion mass spectroscopy (SIMS) uses sputtering of a sample with neutral ions (e.g. Argon) to ablate part of a sample’s surface and the removed ions are then analyzed using mass spectroscopy [154]. This method is able to analyze sample compositions on a parts per billion range and is commonly used to investigate doping profiles in semiconductors [154]. In the context of SHJ SIMS has proven to be useful for the analysis of hydrogen profiles in the a-Si:H layers [85]. Scanning electron microscopy (SEM) uses secondary electrons excited by an electron beam directed onto a sample to construct an image of a sample surface. Electrons
30
CHAPTER 3. PREPARATION, SPECTROSCOPY AND SIMULATION
impacting the electrically grounded sample excite electrons from the sample, which are then analyzed and the intensity of the secondary electron beam is translated into the brightness of the picture [155]. This method is mainly used to get an impression of structures, or thicknesses of conduction layers. In this work a Hitachi S41000 SEM was used to investigate black silicon structures (cf. chapter 5). Spectral Response S(λ) is the ratio of the short circuit current jSC and the the illumination power density E(λ) at a certain wavelength λ. Commonly the external quantum efficiency EQE is calculated to quantify the wavelength dependent extracted charge carrer density jSC /e in relation to the number of incident photons (E(λ) λ /(¯ hc)) with Plancks constant h ¯ and the speed of light c. EQE =
h jSC (λ) h ¯c ¯c = S(λ) e E(λ)λ eλ
(3.12)
For the comparison of devices with different reflectivity R the internal quatum efficiency IQE is calculated with IQE(λ) = EQE(λ) / (1-R(λ)). Spectral response measurements were conducted using a grating monochromator and a Xenon light source. Additionally a Halogen lamp is used to provide a constant background illumination and carrier density.
3.2.7
Device simulation
The numerical device simulation programm AFORS-HET [89] was used to corroborate transport measurements (cf. chapter 7) and to assess optical loss mechanisms (cf. chapter 5). AFORS-HET was specifically designed to model SHJ solar cell structures as shown in Figure 2.3 and includes all necessary models and functions, while providing an easy to use interface. It allows the user to define semiconductor and metal layers with different electrical properties. Boundary conditions can be metal, semiconductor, or metal-insulator-semiconductor contacts. Electrical calculations are based on the fundamental Poisson’s equation and the continuity equation. Carrier transport is simulated using drift-diffusion, thermionic emission [156] and tunneling through a spike [89]. Unfortunately band to band tunneling and tunnel hopping are not implemented (cf. chapter 7). Optical calculations are based on Lambert-Beer’s law for absorption of light and the reflection of surfaces can either be calculated by transfer matrix calculations using refractive index data for all sub layers, or transmission and reflection files of all layers can be measured and loaded. All calculations are one-dimensional, which means that cell structures like back-contact back-junction solar cells [157], or light trapping structures like random pyramids [96], or black silicon [30] have to be simplified.
Chapter 4 Hydrogen plasma treatments The contents of this chapter were published in Applied Physics Letters 102 (2013) 122106 under the title "Hydrogen plasma treatments for passivation of amorphouscrystalline silicon-heterojunctions on surfaces promoting epitaxy" [33]. Furthermore it includes some results obtained during the authors master’s thesis.
4.1
Abstract
The impact of post-deposition hydrogen plasma treatment (HPT) on passivation in amorphous/crystalline silicon (a-Si:H/c-Si) interfaces is investigated. Combining low temperature a-Si:H deposition and HPT, a high effective charge carrier lifetime > 8 ms is achieved on c-Si(100), which serves as model for surfaces promoting epitaxy of a-Si:H. It is shown that the passivation improvement stems from diffusion of hydrogen atoms to the heterointerface and subsequent dangling bond passivation. Concomitantly, the a-Si:H hydrogen density increases, leading to band gap widening and void formation, while the film disorder is not increased. Thus, HPT allows for a-Si:H band gap and a-Si:H/c-Si band offset engineering.
Author contributions The authors list comprised Mathias Mews, Tim F. Schulze, Nicola Mingirulli and Lars Korte. The author of this thesis was the lead author of the publication. He wrote the article and conceived the experiments. Furthermore he conducted all experiments described therein, apart from the SIMS measurements and chemical cleaning of the wafers. He did the analysis of all measurements. Tim F. Schulze suggested to evaluate the activation energy of the defect reduction upon hydrogen plasma treatments at different 31
32
CHAPTER 4. HYDROGEN PLASMA TREATMENTS
temperature and while discussing this result with the main author proposed to do SIMS measurements. He furthermore contributed to writing the article by writing the first version of the introduction and editing and discussing the article with the other authors. Nicola Mingirulli was involved in the early stages of this project. He suggested to conduct variations of the hydrogen plasma duration and power density, which would eventually lead to the described temperature variation. He read and commented on the manuscript. Lars Korte was involved in discussions of the results during the whole project, he developed the CFSYS fitting routine and he read, commented and edited the manuscript.
Acknowledgment of third parties The SIMS measurements were conducted as a paid service by RTG Mikroanalyse GmbH. Chemical cleaning of the wafers was done by the HZB employee Kerstin Jacob.
Post publication changes to this chapter Section 4.4.5 was added to discuss related solar cell results, which were not part of the original publication. In the headline of Table 4.1 "conduction band" was replaced by "valence band". The notation for surface orientations was replaced with (100). Inserts are now single graphs and some abbreviations were changed to be in line with the rest of this thesis. The previously only mentioned oxygen profile was added to the SIMS measurements in Fig 4.3.
4.2
Introduction
Suppressing charge carrier recombination at surface dangling bonds on crystalline silicon (’passivation’) is an important task in numerous semiconductor devices, including high-efficiency solar cells. Different materials, which provide passivation either by repelling charge carriers from the surface (field-effect passivation), or chemically saturate dangling bonds, have been developed. Undoped amorphous silicon ((i)a-Si:H) provides excellent chemical passivation and favorable properties as a heterojunction contact, which enables to build full-area passivated hetero-p/n-junctions by deposition of doped/undoped a-Si:H stacks. This concept is the basis of the silicon heterojunction solar cell (SHJ-SC), which has been shown to reach conversion efficiencies >23% [18]. The suppression of epitaxial growth during a-Si:H deposition by plasma-enhanced chemical
4.3. EXPERIMENTAL DETAILS
33
vapor deposition (PECVD) is crucial for reaching high open-circuit voltage (VOC ) in SHJ solar cells, which can be challenging for (100)-oriented surfaces [99], or in the presence of morphological features such as random pyramid textures [158]. In these cases, the a-Si:H deposition parameter range is restrained to regimes leading to inferior a-Si:H bulk quality – and thus passivation [19] – by the requirement to suppress epitaxy. Recently it was shown that a HPT step during PECVD of a-Si:H/c-Si structures can lead to a passivation increase [110]. Here we explore post-deposition hydrogen plasma treatment (HPT) as a means to improve passivation by (i)a-Si:H on surfaces promoting epitaxy, taking the example of the Si(100) surface which is technologically important for back-contact back-junction SHJ solar cells [28]. We analyze structural changes induced by HPT of ultra thin (i)a-Si:H passivation layers, and relate them to the ensuing passivation improvement. Using a combination of Fourier-transform infrared spectroscopy (FT-IR), spectroscopic ellipsometry (SE), photoelectron spectroscopy (PES), and deuterium profiling, it is unequivocally shown that the passivation improvement is caused by diffusion of atomic hydrogen towards the heterointerface. Concomitantly, the a-Si:H band gap Egap is widened by increased hydrogen incorporation, and the mass density reduced by the formation of microscopic voids. Contrary to previous reports [110] the bond angle disorder in the a-Si:H network, for which the Urbach energy is a measure, is not increased by the HPT. Thus, post-deposition HPT enables passivation by (i)a-Si:H on surfaces promoting epitaxy and extend the usability of a-Si:H/c-Si heterojunctions to include thin-film- or nanostructured silicon absorber materials, on which the affinity towards epitaxial growth prevents the preparation of a classical a-Si:H heteroemitter. Furthermore HPT induced changes in the band gap and band offsets offer additional degrees of freedom for tailoring the heterointerface.
4.3
Experimental details
In our study, we used 280 µm thick (100)-oriented 1 - 5 Ωcm phosphorus-doped high quality float zone silicon wafers as substrate material. Prior to deposition the wafers were cleaned following the RCA procedure and dipped in diluted hydrofluoric acid (2 min, 1 %) to strip off the native silicon oxide. (i)a-Si:H-layers were deposited with PECVD at 170◦ C using the ’LP’ parameter set [159]. HPTs were done in a conventional parallel plate RF-PECVD (13,56 MHz) at 1 mbar process pressure. The substrate temperature (Tsub ) was varied between 35 and 180 ◦ C. and the RF power density was 60 mW/cm2 . Layer thicknesses and band gaps of a-Si:H layers were determined by SE
34
CHAPTER 4. HYDROGEN PLASMA TREATMENTS
Figure 4.1: Dependence of the free charge carrier lifetime τ15 at an injection level of 1015 cm−3 on the substrate temperature during a post deposition HPT. Filled symbols are values after HPT. Half-filled symbols mark the same samples after an additional thermal annealing. The triangle marks a sample that has not been exposed to a HPT. on a Sentech SE850 (wavelength range 350-2500 nm), followed by fitting of a TaucLorentz model [137] to the data. Minority carrier lifetime (τ ) measurements were conducted using a photoconductance decay (PCD) setup (Sinton Consulting WCT-100). H bonding configurations and total H content were quantified with FT-IR measured on a Bruker IFS 125HR in Brewster angle configuration [132], and the calculation of the hydrogen content of the layers was done using the approach of Langford et al. [129]. For characterization of the valence band tail, reflecting the bond angle disorder in the a-Si:H layers, near-ultraviolet PES (NUVPES) measurements were conducted in the constant-final-state-yield mode (CFSYS) [146]. To prevent surface contamination after deposition the samples have been vacuum transferred from the deposition chamber to the CFSYS analysis chamber. A model density of states (DOS) was fitted to the data to obtain the (i)a-Si:H parameters [86].
4.4 4.4.1
Hydrogen plasma treatments of a-Si:H Carrier lifetime spectroscopy
We begin by analyzing the minority carrier lifetimes measured on c-Si samples symmetrically passivated with 7 nm of (i)a-Si:H, which were exposed to 4 min of HPT at
4.4. HYDROGEN PLASMA TREATMENTS OF A-SI:H
35
Figure 4.2: Arrhenius-plot of the interface defect density Dit on the substrate temperature during a post deposition HPT. The activation energy of the defect density reduction is about 440 meV.
different Tsub after deposition, and subsequent anneal at 200◦ C for 20 min under ambient pressure and N2 flow. As seen in Fig. 4.1, the HPT leads to a dramatic improvement of carrier lifetime, with a monotonic increase of the passivation level with Tsub (filled symbols). The lifetime of an annealed sample without HPT is placed at Tsub = 0◦ C. For Tsub > 140◦ C, HPT enables higher lifetimes than by thermal annealing alone. At Tsub > 170◦ C the lifetime reaches 8 ms even without thermal annealing, which is stateof-the-art passivation for thin (i)a-Si:H layers [160–162]. The distinct dependence of the passivation result on Tsub lends itself to further analysis. The contribution of an eventual interface charge and the chemical passivation were disentangled using a semianalytical model for interface recombination [127], which was fitted to the measured PCD curves. In accordance with previous results on passivation by (i)a-Si:H, we found that the interface fixed charge is small (Qf < −7 × 1010 cm−2 ), and the trend in τ is entirely dominated by the changes in the interface defect density (Dit ). An Arrhenius plot of Dit over 1/Tsub (inset in Fig. 4.2) reveals an activation energy of 0.44 eV for the defect saturation process upon HPT. This value is surprisingly close to the activation energy for in-diffusion of atomic hydrogen from a plasma into thick a-Si:H-films [163]. This suggests the diffusion of atomic hydrogen towards the heterointerface to be the underlying cause of the passivation effect observed in Fig. 4.1.
36
4.4.2
CHAPTER 4. HYDROGEN PLASMA TREATMENTS
Diffusion profiles
To corroborate this hypothesis, we performed deuterium (D) in-diffusion experiments on thin (i)a-Si:H passivation layers in order to correlate the deuterium profiles as measured by secondary-ion-mass spectroscopy (SIMS) with the ensuing passivation effect. To this aim, c-Si wafers were (i)a-Si:H coated on one side and treated with a deuteriumplasma (parameters identical to the HPT step used on all other samples). Afterwards the samples were coated with an additional (i)a-Si:H capping layer at Tsub =90◦ C to prevent oxidation of the deuterated layer and to provide a sacrifical layer for the initial sputtering phase of the SIMS measurement. The SIMS measurements were conducted at RTG Mikroanalyse GmbH with a Cameca IMS-4f using Argon ions at a kinetic energy of 6.5 keV. Hydrogen, deuterium and oxygen profiles of samples treated at different Tsub with a deuterium plasma are shown in Fig. 4.3. The (i)a-Si:H/c-Si-interface is at the origin of the abscissa and the shaded box marks the approximate position of the interface between the deuterated (i)a-Si:H and the capping layer. Both interfaces were assumed to coincide with the peaks in the oxygen profile in figure 4.3c. The hydrogen density in the (i)a-Si:H layer is diminished upon deuterium in-diffusion, as would be expected based on previous results on thick a-Si:H layers [163]. The deuterium density in the (i)a-Si:HD increases with Tsub , as expected from the thermally activated in-diffusion. Furthermore the deuterium profiles show mostly linear decays, which implies that the amount of diffusing atoms is larger than the amount of trap states [164]. Additionally the profiles display deuterium and hydrogen peaks close to the a-Si:H/c-Si-interface, which is typical for epitaxy free a-Si:H/c-Si-interfaces [85, 165]. From Fig. 4.3a, it can be seen that resulting from the temperature dependence of the deuterium diffusion coefficient, the deuterium concentration at the heterointerface monotonically varies with the substrate temperature during deuterium plasma treatment. Assuming that in-diffusion of atomic hydrogen/deuterium promotes the defect saturation at the heterointerface and since hydrogen and deuterium have the same diffusion coefficient [163], it is tempting to relate the defect density as inferred from PCD measurements with the deuterium concentration in the SIMS profiles. Fig. 4.4 shows the minority carrier lifetime of symmetrically passivated c-Si wafers (10 nm (i)a-Si:H on both sides) after a HPT, versus the deuterium density at the (i)a-Si:H/c-Si-interface of the SIMS samples equivalently treated with a deuterium plasma. Although the number of samples is small, a clear correlation is visible which underlines the role of atomic hydrogen diffusion towards the heterointerface in interface passivation.
4.4. HYDROGEN PLASMA TREATMENTS OF A-SI:H
37
Figure 4.3: (a) deuterium (2 H), (b) hydrogen (1 H) and (c) oxygen (O) profiles of (i)aSi:H/(i)a-Si:DH/c-Si-samples treated with a deuterium plasma at different Tsub . The single line shows the position of the a-Si:HD/c-Si-interface and the shaded box marks the approximate position of the a-Si:H/a-Si:HD-interface.
38
CHAPTER 4. HYDROGEN PLASMA TREATMENTS
Figure 4.4: Minority carrier lifetime at an injection level of 1015 /cm3 (τ15 ) in (i)aSi:H/(i)a-Si:DH/c-Si-samples treated with a deuterium plasma at different Tsub . The minority carrier lifetime is plotted against the deuterium density (2 H) at the interface. The (i)a-Si:DH layers are 10 nm thick and the line is a guide to the eye.
4.4.3
Hydrogen density, mass density and band gap
Towards device application of HPT passivation layers, it is important to understand structural changes in the thin films upon HPT. In a previous report, based on the rather indirect measure of the broadening parameter C of the Tauc-Lorentz model for the a-Si:H dielectric function, it was invoked that the disorder in the amorphous network is increased by the HPT [110]. This would influence the stability of the a-Si:H passivation, as the disorder is an important parameter in the metastable defect creation processes in a-Si:H [57]. In the following we analyze the hydrogen microstructure, mass density and valence band tail slope of the HPT treated (i)a-Si:H layers in order to shed light on these suggested structural changes. Fig. 4.5 shows data on the hydrogen bonding, band gap and mass density, derived from SE and FT-IR measurements. The mass density is determined combining the hydrogen density (from FT-IR) and the long-wavelength limit of the dielectric function (from SE) [140]. It is clearly visible from Fig. 4.5 that the overall hydrogen content increases along with the substrate temperature during the HPT, while at the same time the optical band gap is widened, and the mass density steeply decreases. Deconvolution of the Si-H stretching modes (SM) [129] reveals that the low-frequency SM (LSM), usually associated with monohydrides, is decreased, while the high-frequency SM (HSM), commonly assigned to clustered monohydrides, is pronouncedly increased upon HPT. This implies that upon HPT there is a distinct
4.4. HYDROGEN PLASMA TREATMENTS OF A-SI:H
39
Figure 4.5: (a) mass density ρ, (b) hydrogen concentration C and (c) band gap Egap ρ of a-Si:H(i)-layers on c-Si substrates. In (b) black squares mark the total volume hydrogen content, red circles the HSM concentration and blue triangles the LSM concentration.
40
CHAPTER 4. HYDROGEN PLASMA TREATMENTS
Figure 4.6: Mass density ρ of the (i)a-Si:H layers plotted against the overall hydrogen density (Cges ) in the same layers. reorganization of the hydrogen bonding configuration in the thin (i)a-Si:H layer, and not just an increase of total Si-H bonds. In Fig. 4.6 the mass density is plotted versus the total hydrogen content. The slope of the mass density decay upon hydrogen incorporation is compatible with the formation of multi-vacancies like platelets or even (nanosized) voids in the a-Si:H film [39]. In summary, it is found that along with the well-known band gap widening, a loss of a-Si:H film compactness and a reorganization of hydrogen bonding in favor of multi-vacancy-related bonding configuration is accompanying the HPT.
4.4.4
Valence band spectroscopy
A crucial question relates to the development of strain in the Si-Si network, being the source of metastable defect creation [57]. Therefore we analyzed the a-Si:H valence band tail, whose electronic DOS can be measured by PES. We conducted CFSYS measurements of four samples, two of which were exposed to a HPT at 13.56 MHz and 60 MHz, respectively. The other two samples were references, which were deposited at identical conditions as the hydrogen plasma treated samples. The Urbach tail energies, characterizing the slope of the exponentially decaying valence band tail DOS, and the position of the valence band maximum relative to the Fermi level of these samples are compiled in Table 4.1. From Table 4.1 it is obvious that the HPT shifts the valence band maximum away from the Fermi level by 70 to 150 meV, while the Urbach
4.4. HYDROGEN PLASMA TREATMENTS OF A-SI:H
41
Table 4.1: Parameters of the valence band edge before and after HPT, as determined by PES: Urbach energy (Eu ) and valence band edge (Ev − Ef ). sample hydrogen plasma Eu (meV) Ev − Ef (eV) A,B no H-plasma 63 ± 5 1.29 ±0.03 C 60 MHz 70 ± 5 1.45 ± 0.03 D 13.56 MHz 70 ± 5 1.36 ± 0.03 energy of the valence band tails is not affected. The first observation is consistent with the band gap widening found upon HPT [110], which is known to proceed by a downward shift in energy of the valence band edge [44], and is also observed in our SE data (Fig. 4.5). The constant Urbach energy implies that there is no increase in the strain of the silicon network. This is interesting as the incorporation of higher hydrogen contents during PECVD growth usually leads to an increase of strain in aSi:H [44]. However, in the present case, the post-deposition incorporation of hydrogen seems not to be related with an increase of strain. Interestingly, this finding is in accordance with previous studies on hydrogen/deuterium in-diffusion into thick a-Si:H films [166,167]. These authors observed clustering of excess hydrogen upon HPT, while no evidence of increased strain was found. Thus, it can be concluded that concerning structural changes upon HPT, ultrathin a-Si:H passivation layers behave similarly to thick a-Si:H specimen, i.e. display hydrogen clustering manifesting in an increase of the HSM and a loss of compactness along with Egap widening, while the level of strain remains constant. This renders the HPT particularly interesting for solar cell device applications for two reasons: Due to the constant and low Urbach energy hydrogen plasma treated films can be expected to be as stable as untreated ones against creation of metastable defects [51] and the increase of the band gap with the hydrogen content is greater than for hydrogen rich layers directly deposited with PECVD, due to the missing compensating effect of a steeper band tail on the band gap widening [44].
4.4.5
Solar cell results
Solar cells were fabricated using layers exposed to the HPT in the 60 MHz chamber. For these experiments HPT process time variations were conducted on (i)a-Si:H layers and solar cell precursors. Figure 4.7 displays results from a HPT duration variation. Figure 4.7(a) shows the improvement of the carrier lifetime upon HPT. The minority carrier lifetime reaches a saturation after 180 seconds. This reflects the time necessary for significant amounts of hydrogen to reach the SHJ. Figure 4.7(b) shows the increase of the band gap upon HPT, which confirms that also during this process the hydrogen density in the layer was increased. The influence of the increased minority carrier
42
CHAPTER 4. HYDROGEN PLASMA TREATMENTS
Figure 4.7: (a) Minority carrier lifetime at an injection level of 1015 /cm3 (τ15 ) in (i)aSi:H/(n)c-Si samples, (b) band gap (Egap ) of the (i)a-Si:H layers, (c) VOC and (d) fill factor (FF) of SHJ solar cells, with identically processed passivation layers, plotted against the duration of the hydrogen plasma treatment (tHPT ) of the (i)a-Si:H layer. Lines are a guides to the eye.
Table 4.2: Implied IV parameters of (n)c-Si wafers passivated with (i)a-Si:H layers with and without HPT as extracted from TRPCD measurements.. with hydrogen plasma without hydrogen plasma implied VOC [mV] 729.5 680.5 UMPP [mV] 635 578 implied FF [%] 82.2 80.8
lifetime and (i)a-Si:H hydrogen content can be seen in Figure 4.7(c) and (d). From Figure 4.7(c) it is conceivable that the increased minority carrier lifetime translates into an increased solar cell VOC , whereas the comparison of Figure 4.7(b) and (d) shows that the increase of the (i)a-Si:H band gap does not decreases the solar cell FF. This is an interesting result, as the valence band offset at the SHJ constitutes a transport barrier for holes diffusing towards the emitter and could be expected to hinder the carrier transport. A more detailed discussion is found in chapter 7. Furthermore the HPT mainly affects passivation at the SHJ and this passivation is also reflected in the pseudo FF [168, 169]. Therefore it is possible to calculate a maximum possible FF allowed by the passivation of the SHJ from the data of TRPCD measurements. It is only necessary to translate the carrier injections density into a current density and shift it by an assumed photocurrent [170]. The result of this procedure is summarized in Table 4.2. The implied VOC of the sample that underwent a HPT is higher, but also its implied FF is 82.2 %, whereas the implied FF of the standard cell is 80.8 %. It can therefore be concluded that the improved passivation after HPT could even be beneficial to the transport properties of the SHJ solar cell and a possible detrimental influence of the increased band gap and valence band offset is not found.
4.5. CONCLUSION
4.5
43
Conclusion
The HPT process discussed so far opens a new route towards band gap engineering, without severely deteriorating the electronic quality of the layers. Furthermore, the efficient in-diffusion of hydrogen from the plasma in combination with small thicknesses of the passivation layer allows for a dramatic increase of the passivation potential on c-Si surfaces. We achieved τ15 = 11 ms and an implied VOC of 737 mV for a HPT at 170◦ C and thermal annealing of 7 nm a-Si:H. Finally, the possibility of deposition at suboptimal (i)a-Si:H to avoid epitaxy, combined with a post-deposition HPT, enables the application of heteroemitters on surfaces promoting epitaxy and (nano)structured surfaces. Thus, the two-step concept may be a feasible candidate for application of SHJ concepts to nanostructured absorbers, such as for nanowire arrays [171]. In summary, a two-step approach towards highly passivating SHJ emitters, involving the deposition of a non-epitaxial a-Si:H(i)-layer and a HPT, has been investigated. The previously established improvement of the charge carrier lifetime at the (i)a-Si:H/c-Siinterface after a well-adjusted HPT was found to be due to diffusion of hydrogen atoms to the a-Si:H(i)/c-Si-interface and improved chemical passivation due to dangling bond saturation. Furthermore, this approach is highlighted to enable the preparation of state-of-the-art passivation layers on the silicon (100)-surface and surfaces which are prone to epitaxial growth, without introducing defect-creating strain into the amorphous network.
Chapter 5 Black silicon texture The contents of this chapter were published in physica status solidi - rapid research letters 8 (2014) 831-835 under the title "Amorphous/crystalline silicon heterojunction solar cells with black silicon texture" [34].
5.1
Abstract
Excellent passivation of black silicon surfaces by thin amorphous silicon layers deposited with plasma enhanced chemical vapor deposition is demonstrated. Minority charge carrier lifetimes of 1.3 milliseconds, enabling an implied open circuit voltage of 714 mV, were achieved. The influence of amorphous silicon parasitic epitaxial growth and thickness, as well as of the texture depth is investigated. Furthermore quantum efficiency gains for wavelenghts above 600 nm, as compared to random textured solar cells, are demonstrated in 17.2 % efficient amorphous-crystalline silicon-heterojunction solar cells with black silicon texture.
Author contributions The authors list comprised Mathias Mews, Caspar Leendertz and Lars Korte from HZB. Furthermore the work was conducted in cooperation with Michael Algasinger and Svetoslav Koynov of Walter-Schottky-Institut of the Technische Universität München. The author of this thesis was the lead author of the publication. He wrote the article and conceived the experiments. He conducted all experimental work apart from preparing the black silicon texture, REM measurements, chemical wafer cleaning and cell isolation. The analysis of all measurements was done by the main author. Caspar Leendertz was responsible for organizing the project and was involved in continuous discussion with the lead author. Michael Algasinger prepared the black silicon 44
5.2. NANOTEXTURES AND (I)A-SI:H PASSIVATION
45
texture and modified it after discussing initial results with the main author. Thus leading to the texture variation shown in Figure 5.5. Svetoslav Koynov developed the black silicon etching process and was discussing the results with Michael Algasinger. Lars Korte was involved in continuous discussions during the project. All four coauthors read, edited and commented on the manuscript.
Acknowledgment of third parties Chemial cleaning of the wafers and cell isolation by photolithography and etching was done by HZB employee Kerstin Jacob. REM measurements were conducted by the HZB employee Carola Klimm.
Post publication changes to this chapter Section 5.2 was added to explain the adaption of the HPT process developed in the last chapter to the passivation of this silicon nanotextures. Also section 5.5.4 was added to give a more detailed explanation of the parasitic absorption in the nanotexture and the resulting losses in quantum efficiency.
5.2
Nanotextures and (i)a-Si:H passivation
The preceding chapter discussed the post-deposition HPT of (i)a-Si:H passivation layers. The main advantage of this approach is the ability to partially decouple the (i)a-Si:H properties from the deposition process. While working with an ideal surface such as (111) pyramids this is seldom necessary, but on surfaces with higher densities of dangling bonds per surface area like (100), or (110) surfaces and on crystallographically undefined surfaces like black silicon as prepared by metal-catalyzed etching [30,117], or reactive-ion etching [120] this may be of great importance. Unfortunately deposition process conditions which lead to state of the art passivation layers on (111) surfaces, may lead to epitaxial growth on other surfaces [99]. The general approach is to deposit an (i)a-Si:H layer with a low structural quality, which is far off epitaxial growth regions and afterwards improve its structural quality with a HPT, but to not crystallize the (i)a-Si:H layer in the process. Planar (100) surfaces are chosen as a reference surfaces, since they allow ellipsometry measurements and are susceptible to epitaxial growth. Spectral ellipsometry is a suitable tool to monitor this approach, since it is known that the imaginary part of the dielectric funtion (2 ) of (i)a-Si:H/c-Si samples allows to identify parasitic epitaxial growth at the interface, or crystallization of the (i)a-Si:H
46
CHAPTER 5. BLACK SILICON TEXTURE
Figure 5.1: Imaginary part of the dielectric funtion (2 ) of an (i)a-Si:H/c-Si-sample directly after deposition and after a post deposition HPT. The layer was co-deposited on Si-(100) and Si-(111) surfaces and is 5 nm thick.
layer [20]. The amplitude of the Lorentz-Oscillator describing dipole transitions in (i)a-Si:H is related to its structural quality [139]. Figure 5.1 shows the imaginary part of the dielectric function of an (i)a-Si:H/c-Si-sample after the a-Si:H deposition and after an additional HPT. The samples have been deposited at a comparatively low temperature of 135◦ C and with a high power density of 56 mW/cm2 . Therefore they have a comparatively high Urbach energy of 83 meV, which corresponds to high structural disorder and leads to low minority carrier lifetimes (τ15 ) of 1.9 ms on (100) and 2.3 ms on Si-(111). The structural properties are reflected in a low Amplitude and a high width of the Lorentz oscillator of the (i)a-Si:H dipol transition at about 3 to 3.5 eV. The HPT leads to an increase of the Lorentz oscillator’s amplitude and to a decrease of its width. This can be related to an increased structural quality of the (i)a-Si:H layer and an increase of the minority carrier lifetime in these samples, which increase to 7.7 ms on Si-(100) and 5.5 ms on Si-(111). Note that the properties of the dipole transition in the (i)a-Si:H layers of the preceding chapter were not influenced by the same HPT process, as their structural quality was already higher in their as-deposited state. This combination of an (i)a-Si:H passivation layer with a low structural quality and a post deposition hydrogen plasma treatment was applied to black silicon structures and the results are discussed in the remainder of this chapter.
5.3. INTRODUCTION
5.3
47
Introduction
The random, high aspect ratio nanoscale surface texture, which is commonly called black silicon, has excellent anti-reflection and light scattering properties and its successful application on the front surface of solar cells has recently been shown [119,172,173]. The major challenge is the passivation of recombination active defects on the nanostructured and enlarged black silicon surface [121]. Previous works aiming at the application of black silicon in solar cells employ dielectric passivation layers [119, 172, 173]. AlOx layers deposited with atomic layer deposition for passivation of nanotextured silicon fabricated with reactive ion etching resulted in a few milliseconds minority carrier lifetime [174] and 18.7 % solar cell efficiency [173]. However, the alternative route of both passivation and contacting of black silicon with hydrogenated amorphous silicon (aSi:H) has not been investigated. A-Si:H can provide excellent passivation of crystalline silicon surfaces and enables the electrical contacting of silicon wafers at the same time. Furthermore the use of amorphous-crystalline silicon heterojunctions (SHJ) is a proven concept for high-efficiency silicon solar cells [10]. The combination of the excellent optical properties of black silicon with the high quality surface passivation of amorphous silicon becomes even more interesting as the SHJ-technology moves towards thinner wafers in order to realize even higher open circuit voltages [18].
5.4
Experimental details
For the present study 280 µm thick (100) and (111) oriented 1 - 5 Ωcm phosphorusdoped float zone silicon wafers were used as substrate material. Black silicon was fabricated by metal-assisted etching (MAE) of (100) Si wafers, as described in earlier work [30]. Only one sample surface was textured, while the other remained planar. The samples were stripped of their native oxide in HF (5 %). One nanometer of gold was evaporated, leading to isolated gold islands on the samples. Texture etching was carried out in a 1:5:10 solution of HF (50%), H2 O2 (30%) and water at room temperature. For gold removal the samples were treated for 2, or 4 min in iodine and potassium iodine solution (I2 :KI:H2 O = 1:4:40 weight ratio). A SC1 cleaning step was applied to remove nanoporous material [175]. Prior to a-Si:H deposition the wafers were cleaned following the RCA procedure and dipped in diluted hydrofluoric acid (3 min, 1 %) to strip off the surface oxide. A-Si:H layers were deposited using plasma-enhanced chemical vapor deposition (PECVD). Intrinsic and doped layers were deposited at 0.5 mbar, with 60 MHz excitation and a silane flow of 10 sccm. 2 sccm diborane, or phosphine were added to deposit n- and p-type a-Si:H respectively. A hydrogen flow of 5 sccm was applied during the deposi-
48
CHAPTER 5. BLACK SILICON TEXTURE
tion of intrinsic layers. Doped layers were deposited using a plasma power density of 18 mW/cm2 and 56 mW/cm2 were applied for intrinsic layer deposition. The substrate temperature was 130◦ C for p-doped and intrinsic layers, but 195◦ C for n-doped layers. The distance between electrode and substrate was 19 mm for intrinsic layers and 23 mm for doped layer deposition. Intrinsic layers ((i)a-Si:H) were exposed to a hydrogen plasma treatment to decrease the interface defect density [33]. Process parameters of the hydrogen plasma treatments were 0.4 mbar, 130◦ C substrate temperature, a hydrogen flow of 20 sccm, 15 mm electrode to substrate distance and a power density of 112 mW/cm2 . Indium tin oxide (ITO) layers were deposited using reactive radio frequency sputtering from an ITO target and 0.1 % oxygen in the process gas flow. Metal contacts were fabricated by thermal evaporation. Front grids consist of a 10 nm titanium adhesion layer and 1.5 µm of silver. 10 nm of titanium and 500 nm of silver constitute the full area back contacts. A photoconductance decay setup (Sinton Consulting WCT-100) was used for minority carrier lifetime measurements [126], a Hitachi S4100 with cold field emitter for Scanning electron microscopy (SEM) and a Perkin Elmer Lambda 19 spectrometer for optical measurements. The effective reflectivity of the textured surfaces is calculated from 300 to 1200 nm using the ASTM G173 spectrum as irradiation.
5.5 5.5.1
Results and discussion Reflectivity of black silicon surfaces and solar cells
The black silicon surfaces exhibit an effective reflectivity of 3.3 to 4.7 % in the range from 300 to 1200 nm, compared to about 7 % for an alkaline texture. The effective reflectivity value does not change after deposition of a-Si:H layers with deposition times of up to 60 s yielding a thickness of 30 nm, since the amorphous silicon follows the contour of the nanotexture and has a refractive index similar to the one of c-Si. Depositing ITO layers with a nominal thickness of 90 to 95 nm decreases the effective reflectivity to about 2.3 to 2.5 %. Because the refractive index of ITO lies between the refractive indexes of silicon and air. Therefore the grading of the reflective index between air and the silicon wafer [176] already brought about by the black Si texture is further flattened out. SEM images of black silicon covered with a-Si:H (Fig. 5.2a) and with a-Si:H and ITO (Fig. 5.2b) are displayed. The roughly conformal coverage of the nanotexture with ITO is visible. The average texture height is in the range of 500 nm.
5.5. RESULTS AND DISCUSSION
49
Figure 5.2: SEM pictures of black silicon fabricated by 70 s of MAE and covered with 15 nm a-Si:H (a), 15 nm a-Si:H and 95 nm ITO (b) and a-Si:H, ITO and nominally 1.5 µm Ag (c).
50
CHAPTER 5. BLACK SILICON TEXTURE
Figure 5.3: Dependence of the minority charge carrier lifetime τ15 at an injection level of 1015 cm−3 on the plasma power density during intrinsic layer deposition. A postdeposition annealing step (200◦ C,20 min) was applied to these samples. Values for planar (100), planar (111) and black silicon surfaces (70 s MAE) are displayed. Lines are guides to the eye.
5.5.2
Passivation layer optimization
One major challenge for passivation of SHJs is the preparation of passivation layers with high structural quality without parasitic epitaxial growth [99]. This is a severe restriction on (100)-surfaces, but far less problematic on (111) [99]. As metal assisted etching (MAE) progresses preferentially along the direction [118] this could restrict the achievable passivation quality on black silicon. One parameter that allows to change from amorphous to epitaxial silicon growth is the plasma power density [19]. By increasing the power density the deposition rate increases, reducing the structural quality of a-Si:H and thereby preventing epitaxial growth. Therefore the a-Si:H passivation quality on black silicon is explored as a function of the plasma power density. Furthermore the epitaxial growth has to be identified. It is known that thermal annealing at 200 to 300◦ C improves passivation of atomically sharp SHJ-interfaces [160], but leads to a degradation of the minority carrier lifetime for interfaces with partial epitaxial growth [85]. Therefore a post deposition annealing step at 200◦ C for 20 min was applied to discriminate between epitaxy free samples and those with partial epitaxy. Minority carrier lifetime values for planar and nanotextured samples passivated with intrinsic a-Si:H deposited at varying plasma power densities are shown in Fig. 5.3. Intrinsic a-Si:H deposition times of 30 s, corresponding to film thicknesses of about
5.5. RESULTS AND DISCUSSION
51
15 nm, are chosen for planar (111) and (100) wafers, while deposition times of 60 s are chosen for black silicon to account for the higher aspect ratio. The planar back side of the black silicon samples is passivated with 4 nm intrinsic a-Si:H deposited at 56 mW/cm2 . A change in the plasma power density from 36 to 18 mW/cm2 results in a drop of the minority carrier lifetime on (100) Si by more than one order of magnitude. A similar, but less pronounced, lifetime decrease is visible for black silicon samples. In contrast, (111) samples do not experience any drop in lifetime in the depicted parameter range. Therefore it can be assumed that the black silicon surface possesses a similar tendency towards epitaxial growth as (100) silicon surfaces. In a second parameter variation the (i)a-Si:H passivation layer thickness was varied on planar and on black-textured samples. The dependence of minority carrier lifetime on the passivation layer deposition time and thickness is shown in Fig. 5.4a. Due to the high aspect ratio a thicker (i)a-Si:H layer is needed for effective passivation of black silicon. While the passivation of planar wafers with (i)a-Si:H already reaches saturation at an (i)a-Si:H thickness of 10 nm, the black silicon samples are most likely not even fully covered at this thickness and the passivation saturates only at about 15 nm film thickness. Furthermore the maximum lifetime is smaller for the black silicon surfaces, possibly due to the enlarged surface area and thus the higher total number of defects. The difference between (100) and (111) surfaces can be explained as being due to the higher density of dangling bonds on (100) surfaces compared to the (111) surface. The overall higher minority carrier lifetime of the black silicon samples in this second experiment are attributed to more thorough removal of gold residuals by prolonged etching (4 minutes). The thickness value for which the minority charge carrier lifetime saturates can be used to deduce the surface increase due to the nanotexture as compared to a planar surface. Fujiwara and Kondo [177] found that a-Si:H on a planar wafer changes from interface layer structural properties to bulk layer properties at a thickness of about 4 nm. They also observed a saturation of open circuit voltage and thus lifetime at this thickness. The same saturation can be found for the nanotextured surfaces at about 15 nm. This determines the surface increase due to the nanotexture to about a factor of 3.5 to 4 compared to a planar surface and is in agreement with experiments using SEM and atomic force microscopy [175]. The best lifetime value achieved during these experiments is 1.3 ms, enabling 714 mV implied open circuit voltage. This high minority carrier lifetime has been achieved on fully nanotextured surfaces with an effective reflectivity of about 3 % using a passivation layer with 15 nm nominal thickness.
52
CHAPTER 5. BLACK SILICON TEXTURE
Figure 5.4: Minority charge carrier lifetimes τ15 at an injection level of 1015 cm−3 (a) in dependence of the intrinsic passivation layer deposition time and thickness on planar wafers for (100), (111) and black silicon wafers. (b) black silicon with passivation layers and after the deposition of p-doped layers on the black silicon side and n-doped layers on the untextured side. Etch time for the black silicon samples was 70 s. Lines are guides to the eye.
5.5. RESULTS AND DISCUSSION
53
The application of black silicon in a-Si:H/c-Si solar cells necessitates to form a p/njunction and apply a back-surface field using doped a-Si:H layers. In Fig. 5.4b minority carrier lifetimes of black silicon samples with passivation layers and after deposition of n-type layers on the untextured side and p-type layers on the textured side are depicted. As can be seen in Fig. 5.4b, the lifetime values of all samples deteriorate drastically, corresponding to an implied VOC decrease from 715 to 655 mV for the sample with a 15 nm thick passivation layer. This effect is known from planar and alkaline textured SHJ structures and is typically explained by Fermi level shift induced defect generation [58]. Still the degradation is stronger for the passivation layers on black silicon, as the planar reference only degrades from 730 mV to 725 mV implied VOC . Possible reasons are a different physical structure of the intrinsic amorphous silicon due to the growth on the black silicon surface, lateral inhomogeneous a-Si:H thicknesses, or gold residuals from the metal assisted etching process.
5.5.3
Quantum efficiency measurements
Solar cells were fabricated using these samples and the IV characteristics and internal quantum efficiency (IQE) were measured. A strong decrease of the blue response for the black silicon solar cell as compared to a KOH textured reference is visible in Fig. 5.5. One possible reason for decreased blue response is enhanced carrier recombination in a nanoporous region at the black etched surface [178]. A SC1 cleaning step is able to remove these nanoporous regions [175] and was applied during the experiments presented here. Therefore parasitic absorption in nanoporous silicon can be excluded. Another possible reason is parasitic absorption in amorphous silicon layers. To investigate whether this is the reason for the IQE decrease, the internal quantum efficiency of two solar cells with 70 s metal assisted etching time, but different intrinsic a-Si:H deposition times are shown in Fig. 5.5. Although there is a slight decrease of the IQE with increased a-Si:H thickness, it is obvious that this cannot be the reason for the strong blue deficiency of the nanotextured solar cells. A third reason could be enhanced Auger recombination [119], or parasitic absorption in the nanotexture. Therefore a solar cell with a 30 s long black silicon etch was fabricated, leading to a more shallow texture and an effective reflectivity of about 10 %. The IQE of this solar cell is also shown in Fig. 5.5a. For this solar cell only a slight decrease of the blue response is visible. Therefore the IQE decrease appears to be due to parasitic absorption, or enhanced recombination in the nanotexture. A possible reason for this parasitic absorption is that the p-type emitter surrounding the nanotexture
54
CHAPTER 5. BLACK SILICON TEXTURE
Figure 5.5: (a) Internal quantum efficiency (IQE) of solar cells with KOH texture and 19 s intrinsic a-Si:H deposition time for the front side, compared to the IQE of solar cells with two black silicon texture times (70 s and 30 s) and two (i)a-Si:H deposition times (30 s, or 40 s). P-type a-Si:H deposition time was 50 s for the black silicon texture and 15 s for the KOH texture. (b) External quantum efficiency (EQE) of the same devices.
enforces a majority carrier type inversion in the small structures. This could prevent the collection of minority carriers from this region. The solar cell with 30 s MAE time has an efficiency of 17.2 %, a fill factor of 77.4 %, a short circuit current of 34.2 mA and an open circuit voltage of 649 mV. Thereby displaying better output parameters than the solar cell constructed on the 70 s texture, which has only 31.7 mA short circuit current, due to its lower blue response, an open circuit voltage of 604 mV and a fill factor of 72.7 %. Leading to an efficiency of 13.9 %. Compared to a solar cell with random pyramid texture, a short circuit current of 36.2 mA and an efficiency of 20.2 % the 30 s black silicon solar cell has an improved quantum efficiency for all wavelenghts higher than 700 nm. Still the Fermi-level shift induced degradation of the passivation layers decreases the open circuit voltage well below the 707 mV, which were achieved applying the same deposition parameters on a planar (100) wafer.
5.6. CONCLUSION
5.5.4
55
Parasitic absorption in black silicon nanostructures
This section gives a more detailed explanation of the previously mentioned parasitic absorption in the nanotexture of the black silicon solar cells. Figure 5.6a displays IQE curves of four different solar cells with different MAE etch times. Two of these have already been shown in Figure 5.5a. It is discernible that the decrease in blue response is stronger the longer the etch time was. In a first attempt to understand this the IQE of a SHJ solar cell with a measured black silicon reflection was simulated and then modified to give a similar result. The best match was obtained if the incoming light was reduced by parasitic absorption in c-Si. Simulations of SHJ solar cells with parasitically absorbing c-Si layers of different thicknesses in front of the active absorber were conducted and their IQE is plotted in Figure 5.6b. The c-Si in this model was supposed to absorb light but to not contribute to the photocurrent and to not change the charge transfer in the device. The plot reveals, that the receding blue response can indeed be simulated by parasitic absorption in c-Si. Note that the quantum efficiency for long wavelengths is unchanged for all simulated cells, because of the simple optical model without ray tracing and 100 % back reflection. Therefore one could assume that an increased etching time leads to an increase of parasitic absorption by c-Si. To further elucidate on this SEM pictures of the etched silicon surface for the shortest and the longest etch time used on devices are presented in Figure 5.6c and d. The texture depth is increased with increasing etching time. This leads to the assumption that the silicon volume inside the nanotexture is responsible for the measured parasitic absorption. A possible explanation would be an inversion layer in the nanotexture. Similar inversion layers have been reported in SHJ solar cells, close to the p/n-junction [179]. Basically the inversion layer always present in a SHJ solar cells usually just penetrates a few nanometers into the layer. This leads to comparatively small inversion layer with for the shallower textures after 15, or 30 s etching time (cf. Figure 5.6e). In contrast in the deeper nanotextures, a thin needle of silicon is surrounded by a p/n-junction and the inversion layer encompasses most of the nanotexture (cf. Figure 5.6f). Minority carrier generated in this p-type inversion layer do not contribute to the overall current of the device, since the current is generated by the minority carrier generation in the n-type absorber. This parasitic absorption could be reduced by designing an index gradient texture with bigger features.
5.6
Conclusion
It was shown that it is possible to passivate black silicon surfaces with a-Si:H. The passivation layer thickness has to be increased by at least a factor of three compared to
56
CHAPTER 5. BLACK SILICON TEXTURE
Figure 5.6: (a) Measured internal quantum efficiency (IQE) of solar cells with black silicon front side texture and different etching times for the metal-assisted chemical etching. The arrow points towards higher etching times. (b) Simulated IQE of solar cells with black silicon anti-reflection properties, but reduced incident light. The Incoming light was reduced by absorption in a c-Si layer of varying thickness. This layers is modelled to absorb light, but no neither contribute current, nor to affect the electrical properties of the solar cells. The arrow points towards the spectra with thicker c-Si parasitic absorption layers. (c+d) SEM cross sections of silicon surfaces after 15 (c) and 70 s (d). (e+f) Sketch of the suspected p/n-junction in the structures shown in the SEM pictures. The yellow range is the suspected inversion layer in the n-type c-Si.
5.6. CONCLUSION
57
planar surfaces to reach high quality passivation, which concurs with earlier measurements of the planar-to-textured surface ratio [175]. Furthermore investigations indicate, that black silicon surfaces exhibit a similar tendency towards epitaxial growth as (100) surfaces. Black silicon surfaces show great promise for the application in a-Si:H/c-Si solar cells since implied open circuit voltages of about 714 mV have been reached for initial passivation layers, as well as excellent optical properties with reflectivities below 5 %. Amorphous-crystalline silicon heterojunction solar cells with black silicon front surface exhibiting a promising efficiency of 17.2 % and gain in quantum efficiency compared to random pyramid textures for wavelengths above 600 nm have been fabricated. Unfortunately strong parasitic absorption in the high wavelength range of the visible spectrum was measured. This could be explained by parasitic absorption in inversion layers, which are constituting a larger volume in the nanotextured absorbers, due to the embedding of the (n)c-Si spikes in p-doped a-Si:H layers.
Chapter 6 Solution-processed amorphous silicon The contents of this chapter were published in Appl. Phys. Lett. 105 (2014) 122113 under the title "Solution-processed amorphous silicon surface passivation layers" [35].
6.1
Abstract
Amorphous silicon thin films, fabricated by thermal conversion of neopentasilane, were used to passivate crystalline silicon surfaces. The conversion is investigated using X-ray and constant-final-state-yield photoelectron spectroscopy, and minority charge carrier lifetime spectroscopy. Liquid processed amorphous silicon exhibits high Urbach energies from 90 to 120 meV and 200 meV lower optical band gaps than material prepared by plasma enhanced chemical vapor deposition. Applying a hydrogen plasma treatment, a minority charge carrier lifetime of 1.37 ms at an injection level of 1015 /cm3 enabling an implied open circuit voltage of 724 mV was achieved, demonstrating excellent silicon surface passivation.
Author contributions The author list comprises Mathias Mews, Tobias Sontheimer, Lars Korte and Bernd Rech from HZB. The work described in this chapter resulted from a cooperation with Evonik Industries. The coauthors from Evonik Industries are Christoph Mader, Stephan Traut and Odo Wunnicke. The author of this thesis was the lead author of the publication. He wrote the article, conceived the experiments and conducted the photoelectron spectroscopy measurements, hydrogen plasma treatments and carrier lifetime measurements at HZB. Furthermore 58
6.2. PROMISE AND CHALLENGE OF LIQUID SILICON
59
he analyzed these measurements and some of the carrier lifetime measurements conducted at Evonik Industries Facilities. Christoph Mader was responsible for the organization of the activities at Evonik Industries. These included preparation of liquid silicon, spin-coating of the precursor and its conversion, as well as carrier lifetime measurements. He contributed to planning the experiments described in this chapter and did the degradation experiments described in Figure 6.5. Furthermore he read and commented on the manuscript. Stephan Traut was responsible for developing the synthesis of liquid silicon and adjusting the process for this experiments. Tobias Sontheimer organized communication with Evonik and the transfer of samples via parcel. Odo Wunnicke is the responsible project leader at Evonik Industries and was involved in the discussion of the experiments. Lars Korte was involved in discussing the experiments, provided valuable insights to the analysis of the photoelectron spectroscopy data and read and commented the manuscript. Bernd Rech initiated the cooperation and project with Evonik.
Acknowledgment of third parties Preparation of the liquid silicon precursor, spin-coating of layers and their thermal conversion was conducted by Jasmin Lehmkuhl and Tobias Gergs from Evonik Industries. The Institute of Solar Energy Research Hamelin provided samples with AlOx /SiNx stacks on the backside as a paid service.
Post publication changes to this chapter The insert in Figure 6.3 was extracted to Figure 6.4. Furthermore section 6.2 was added to explain how the hydrogen plasma process developed in chapter 4 serves as prerequisite for this chapter and how the approach differs from the one presented in chapter 5.
6.2
Promise and challenge of liquid silicon
The back-junction back-contact solar cell has caused some research interest in recent times, because the possibility to put both contacts on the cell back side enables higher photo currents and therefor higher power conversion efficiencies. The best SHJ with an efficiency of 25.6 % is a back-contact back-junction solar cell [11]. The inherent
60
CHAPTER 6. SOLUTION-PROCESSED AMORPHOUS SILICON
challenge of this technology is the need for precise structuring of electrically active a-Si:H and TCO layers. Research institutes rely on photo lithography [28, 157], laser ablation of sacrificial layers [180], or masked depositions [29]. Still all these processes are tedious and expensive. Printing of electrically active regions without the need for any later structuring could provide a viable production path for these devices. There are two viable approaches. First the use of silicon-containing inks for screen printing [181], as pioneered by DuPont [182] and NanoGram [181]. Secondly the use of a liquid silicon-containing precursor, which can be dissolved in appropriate solvents to be spin-coated, or printed. Cyclopentasilane [31] and Neopentasilane [112] have been investigated for this approach. Unfortunately a-Si:H processed from liquid precursors features some disadvantageous properties for the application in SHJ solar cells. Due to the thermal conversion it generally features very low hydrogen densities [114] and tends to be porous [109]. HPT processes such as the one discussed in chapter 4 were originally used to hydrogenate amorphous silicon [107] and to increase its density [51]. Therefore a HPT could provide a way to fabricate well passivated liquid processed SHJ solar cells.
6.3
Introduction
Today the most widely used semiconductor material is silicon, applied as crystalline wafers and amorphous, microcrystalline, or polycrystalline layers for photovoltaic and electronic applications. In many cases the material preparation involves complex vacuum processes and laterally structured devices have to be realized using costly processes like photolithography. Printing or drop-casting a liquid silicon precursor and converting it afterwards to amorphous or crystalline layers would offer great potential for simplified fabrication of silicon based semiconductor devices. Solution processed intrinsic amorphous silicon (a-Si:H) could find applications as buffer layer in high-efficiency amorphous-crystalline silicon hetero-junction (a-Si:H/c-Si-SHJ) solar cells [9], as capping layer for thin AlOx rear-side passivation layers of p-type c-Si solar cells [183], or as surface passivation in light emitting devices [184]. Other potential applications for solution processed a-Si:H layers are electron and hole collector layers in amorphous-crystalline silicon bipolar transistors [185], and a-Si:H/c-Si solar cells [10] in conventional as well as in back-contact back-junction configuration. First experiments with liquid polysilane precursors such as cyclopentasilane [31], neopentasilane [111] or trisilane [113] were conducted to manufacture a variety of structures and devices. Amorphous silicon (a-Si:H) thin films were applied in solar cells [112,186], light-emitting devices [187] and thin-film transistors [31]. Furthermore crystalline sili-
6.4. EXPERIMENTAL DETAILS
61
con nanowires were used as anode material in lithium batteries [188]. This contribution aims at investigating the conversion from neopentasilane to a-Si:H and evaluating the potential of solution processed a-Si:H layers for surface passivation of c-Si. First the conversion from the polysilane precursor to a-Si:H is investigated using X-ray photoelectron spectroscopy (XPS). Second the electronic properties of spin-coated amorphous silicon are compared to state of the art a-Si:H deposited using plasma-enhanced chemical vapor deposition (PECVD). Finally the minority carrier passivation achieved with spin-coated a-Si:H and polysilane layers is discussed.
6.4
Experimental details
165 µm thick, (100) oriented and damage etched n-type Cz silicon wafers with a resistivity of 3-5 Ωcm were used as substrates for passivation experiments. To exclude any influence of damage to the wafer backside during processing and to provide rear surface passivation, the wafer backsides were coated with 10 nm AlOx and 90 nm SiNx [189]. P-type (100) oriented chemically-mechanically polished 500 µm thick Cz silicon wafers were used as substrates for photoelectron spectroscopy (PES). The liquid silicon layers were prepared following an approach, presented in an earlier publication [111]. Briefly neopentasilane (Si5 H12 ) [190] was oligomerized by thermal treatment and dissolved in toluene and cyclooctane, yielding a 30 wt% solution. Afterwards 50 to 90 nm thin films were prepared using spin-coating and subsequent annealing on a hot plate at temperatures between 250 and 600◦ C for 10 s to 30 s. All process steps were carried out in nitrogen atmosphere to prevent oxidation of the neopentasilane. To further decrease the defect density in the a-Si:H layers and at the a-Si:H/c-Si interface hydrogen plasma post-deposition treatments were applied [33] in a parallel plate plasma enhanced chemical vapor deposition reactor (PECVD). The process pressure was 0.4 mbar, the excitation frequency 60 MHz, the substrate temperature 195 ◦ C, the distance between electrode and substrate 16 mm and the plasma power density 115 mW/cm2 . A Sinton Consulting WCT-100 photoconductance decay measurement setup was used for lifetime measurements [126]. PES was executed using either Mg-Kα excitation for core level spectroscopy or variable photon energy in the near-UV range, from 3.7 to 7.3 eV, to conduct PES in the constant-final-state-yield mode (CFSYS) [146]. A model density of states [149] was fitted to the data to obtain the valence band position and Urbach tail slope of the a-Si:H layers. The samples were transported in nitrogen atmosphere from the preparation setup to the PES setup, to limit the influence of oxidation
62
CHAPTER 6. SOLUTION-PROCESSED AMORPHOUS SILICON
on the PES measurements. Since sample transfer from the transport equipment to the PES setup is not possible without breaching the inert atmosphere, the samples were exposed to air for a short time (< 5 min). Afterwards the samples were sputtered in an argon plasma to remove oxidized regions. The sputtering conditions were optimized to remove only the oxidized surface layer. Core-level spectroscopy using XPS was conducted after sputtering, while CFSYS measurements of the valence band density of states were carried out before sputtering, as the sputtering process alters the valence band properties significantly. Note, that at the low excitation energies used for CFSYS, the information depth is about 5-10 nm and therefore the influence of the surface on the spectrum is small.
6.5 6.5.1
Results and Discussion From polysilane to amorphous silicon
To resolve the chemical conversion from neopentasilane to amorphous silicon its dependence on the treatment temperature was investigated using XPS measurements of 50 nm thick films converted at temperatures ranging from 300 to 600◦ C. The dependence of the Si 2s and O 1s core level binding energies on the conversion temperature is shown in Fig. 6.1a. For temperatures below 400◦ C the binding energy of the peaks increases, due to a higher oxygen fraction in these samples [191]. The Si 2p core levels dependence on conversion temperature will be discussed in the following (see Fig. 6.1b). Note that the spin-orbit splitting of the Si 2p peak is not discernible within the resolution of the used measurement system. The four oxidation states of the silicon 2p0 are distributed between 100 eV and 105 eV [192]. XPS measurements on polysilane molecule films place their Si 2p binding energy at about 102.5 eV [193]. Therefore a core-level PES signal from polysilane would be difficult to distinguish from the different SiOx phases also present in the samples. Following an approach proposed by Hansch et al. [145] to distinguish contributions of SiOx from those of SiNx to the Si 2p core level spectrum, the ratio of the O 1s to the Si 2s core level intensity was compared to the ratio of the higher energy contributions of the Si 2p signal to the Si 2p0 signal. This data is shown in Fig. 6.1b. The ratio of the O 1s to the Si 2s signal is higher than the ratio of Si 2p+1,+2,+3,+4 to the Si 2p0 component. Therefore all contributions to the Si 2p signal between 100 and 105 eV can be attributed to SiOx . Accordingly, no polysilane is detected, i.e. the whole polysilane component of these layers is oxidized during sample transfer. Moreover the total intensity of oxidic components after sputtering increases with de-
6.5. RESULTS AND DISCUSSION
63
Figure 6.1: (a) Dependence of the Si 2s and O 1s binding energy on the conversion temperature. (b) Conversion temperature dependence of the ratio between the silicon oxide and the silicon peak area, as determined from the O 1s and Si 2s peaks and from the components of the Si 2p peak. Samples were transported in nitrogen atmosphere and were Ar sputter-cleaned prior to measurements.
64
CHAPTER 6. SOLUTION-PROCESSED AMORPHOUS SILICON
Figure 6.2: Valence band position EF -EV (a) and valence band tail Urbach energy Eu (b) for different amorphous silicon layers. The values for a PECVD deposited 10 nm thin layer are displayed on the left-hand side. Values for 50 nm thick amorphous silicon layers, prepared by thermal conversion of neopentasilane at three different temperatures for 30 s, are displayed on the right-hand side. creasing conversion temperature from 400◦ C downwards. Therefore the conversion from polysilane to amorphous silicon occurs between 300 to 400◦ C and at 400◦ C the whole precursor is fully converted. Note that the sample, annealed at 500◦ C, still shows a small O 1s signal i.e. some residual oxygen. As this oxygen is not reflected in the sample’s Si 2p signal, it must be due to to a residual surface contamination related to the shorter sputtering time for this sample. To sum up the preceding paragraphs, it was found that the chemical conversion from neopentasilane to amorphous silicon is possible for conversion temperatures of 400◦ C and above.
6.5.2
Valence band spectroscopy
To obtain additional information about the electronic structure of the valence band of these layers, CFSYS measurements were conducted. A PECVD reference layer was transferred to the PES setup through atmosphere, in order to obtain similar surface properties as for the liquid processed a-Si:H. UV-CFSYS can be used to obtain the valence band position of a-Si:H. The valence band positions of the PECVD deposited reference layer and three liquid processed layers are depicted in Fig. 6.2a. The valence band of the thermally converted layers is
6.5. RESULTS AND DISCUSSION
65
closer to the Fermi level than for the PECVD layer and moves even closer with increasing conversion temperature. This is consistent with a lower hydrogen content in these layers and a further decrease of it, with increasing conversion temperature [44]. Moreover a low hydrogen content is supposedly detrimental to the electronic quality of a-Si:H. The second parameter extracted from CFSYS data is the valence tail Urbach energy. It is a measure of electronic and structural disorder in a-Si:H, since the valence band tail consists of Si-Si-binding states with bond lengths and angles deviating from the eqilibrium [57]. In general a low Urbach energy of about 60 meV is considered favorable for application of a-Si:H in electronic devices [44]. The PECVD reference layer exhibits an Urbach energy of about 55 meV (Fig. 6.2b). The liquid processed layers converted at 400◦ C exhibit an Urbach energy of about 125 meV, which decreases with increasing conversion temperature to about 115 meV for conversion at 500◦ C and to about 90 meV for 600◦ C. This result indicates poor structural quality and may be disadvantageous for application as electron contact, hole contact, or as buffer layer in solar cells. In contrast the structural bulk quality is less important for surface passivation layers.
6.5.3
Liquid processed amorphous silicon passivation layers
Passivation of c-Si with a-Si:H relies on two mechanisms: Formation of an atomically sharp SHJ consisting mainly of Si-Si bonds and saturation of the remaining dangling bonds with hydrogen provided during and after the a-Si:H deposition [44]. The neopentasilane precursor transforms into a-Si:H by Si-H bond breaking and formation of Si-Si-bonds. Therefore the hydrogen concentration in the precursor decreases during thermal conversion and is about 10 % at 400◦ C [114]. Thus the conversion temperature has a strong impact on interface passivation, as a higher temperature increases the Si-Si-bond fraction, but at the same time decreases the amount of hydrogen in the layer which is needed to saturate dangling bonds at the SHJ. To investigate the influence of the treatment temperature a series of samples with different treatment temperatures and times was prepared. The minority charge carrier lifetimes at an injection level of 1015 cm−3 (τ15 ) obtained for different treatment conditions with 60 nm thick liquid processed layers are depicted in Fig. 6.3. Note that these samples have an AlOx /SiNx passivation stack on the rear side. AlOx passivation is known to improve by thermal annealing [194]. Nonetheless any influence of this effect can be discounted in the present experiment, as the minority carrier lifetime enabled by the unannealed AlOx /SiNx stack on the used substrates is already in the range of 2.4 ms, thus well above any minority charge carrier lifetime measured in this series.
66
CHAPTER 6. SOLUTION-PROCESSED AMORPHOUS SILICON
Figure 6.3: Dependence of the minority charge carrier lifetime at an injection level of 1015 cm−3 (τ15 ) on the conversion temperature. The passivation layers are 60 nm thick. The vertical line marks the approximate position of the polysilane-to-amorphous silicon transition. Furthermore it was demonstrated that these AlOx /SiNx stacks are stable for firing steps with process temperatures of up to 910◦ C and still allow for minority carrier lifetimes of about 5 ms after firing [189]. Therefore the sample surface passivated by neopentasilane, amorphous silicon, or any intermediate state is limiting and determines the measured effective minority charge carrier lifetime. Three temperature series with different thermal treatment times have been prepared and all show similar tendencies. The highest τ15 is reached with annealing at about 300◦ C. Unfortunately the passivation obtained by application of these layers degrades after a few minutes in air to a few microseconds and cannot be reestablished with thermal annealing. Furthermore the dependence of τ on the minority carrier density for these samples shows, that passivation is due to a field effect, i.e. a fixed charge close to the interface. Minority carrier lifetime curves for samples treated at 300 and 500◦ C are shown in Fig. 6.4. The shape of the curves [127] shows that the main passivation mechanism at 300 and 500◦ C are field effect and chemical passivation, respectively. The field-effect passivation vanishes upon oxidation of the spin-on film, leading to the observed breakdown of surface passivation. The PES measurements show that the formation of a-Si:H starts if the temperature is increased to about 350 and 400◦ C. The passivation quality decreases again for these conversion temperatures. Masuda et al. have shown that the hydrogen density in the
6.5. RESULTS AND DISCUSSION
67
Figure 6.4: Dependence of the minority charge carrier lifetime τ on the injected carrier density ∆n, for two samples treated at 300 and 500◦ C, respectively. layer is strongly decreased, upon increasing the conversion temperature in this range and that the tetrahedral a-Si:H bond structure starts to form at 350◦ C [114]. Therefore τ15 is likely decreased because Si-H bonds break, and weak Si-Si and Si-dangling bonds are generated, leading to increased bulk defect density in the films and thus to an increased defect density at the interface to the c-Si. Investigating the Si-H bond distribution using infrared spectroscopy, or H outdiffusion could support this conclusions, but is difficult due to oxidation of the polysilane and the lack of in-situ measurement possibilities. Detailed accounts on changes in PECVD grown a-Si:H upon changes in the H microstructure can be found in the literature [195]. In this process the dangling bond concentration is influenced both by Si-H-bond breaking and hydrogen outdiffusion, as well as by Si-Si-bond formation. The highest τ15 is reached at 500◦ C, with ≈ 200 µs.
6.5.4
Oxidation of liquid processed amorphous silicon
Apart from a low defect density enabled by the passivation layer, it is also important, to obtain a long-term stable passivation. A-Si:H is known to degrade under illumination [196]. The defect density can be reduced again by thermal annealing at temperatures between 200 and 300◦ C. Thus the degradation is reversible. Another reason for reversible degradation is found for liquid processed a-Si:H. τ15 of samples
68
CHAPTER 6. SOLUTION-PROCESSED AMORPHOUS SILICON
Figure 6.5: Dependence of the minority charge carrier lifetime at an injection level of 1015 cm3 (τ15 ) for samples with 60 nm spin-coated a-Si:H on the air exposure time after film conversion. Values for as deposited and annealed (2 min at 300◦ C) samples are displayed. processed at 500 and 600◦ C degrades with increasing air exposure time of the samples. Furthermore it can be increased above its initial value using a thermal annealing step at 300◦ C for 2 min. This degradation is reversible using air exposure and annealing cycles. Thus the stability of liquid processed a-Si:H is comparable to that of PECVD a-Si:H. The dependence of τ15 on air exposure time is shown in Fig. 6.5. The degradation of τ15 follows the same exponential decrease for as deposited and annealed samples. Therefore the reversible degradation mechanism is likely the same in both cases. Additionally, no degradation of τ15 is measured, if the samples are stored in nitrogen atmosphere and the degradation is the same for storage in the dark, or under daylight illumination. Therefore this degradation is either related to oxygen, or more likely water and furthermore not relevant for device applications, as passivation layers used in devices are encapsulated, or covered by other layers. The best recoverable minority carrier lifetime reached with spin-coated a-Si:H, in the experiments presented up to this point, is only about 200 µs. A possible reason is that polysilane layers converted to a-Si:H are known to contain only a few atomic percent of hydrogen [114], whereas a high atomic hydrogen content at the SHJ is known to be a prerequisite for high quality interface passivation [21]. Thus it could be beneficial for the interface passivation, if the hydrogen content of the thermally converted layers and at the SHJ is increased, e.g. using the diffusion of hydrogen from a hydrogen plasma
6.6. CONCLUSION
69
through the a-Si:H layer to the SHJ [33]. Therefore 90 nm thick liquid processed a-Si:H layers were annealed and afterwards exposed to a HPT. The annealing step increases τ15 from ≈ 300 µs to ≈ 500 µs and the subsequent HPT leads to a further increase to ≈ 1.37 ms. This lifetime value corresponds to an excellent implied open circuit voltage [126] of 724 mV. This shows that very high quality passivation with liquid processable a-Si:H layers is possible after HPT. The high thickness of these films precludes, for the moment, their use as buffer layers in SHJ solar cells.
6.6
Conclusion
In summary, using information from XPS the conversion temperature of neopentasilane to amorphous silicon was identified to be in the range of 300 to 350◦ C. Furthermore it was shown that liquid neopentasilane precursors can be used to prepare high quality a-Si:H passivation layers for state of the art c-Si surface passivation. Currently these a-Si:H layers are not suited for application in a-Si:H/c-Si solar cells due to their high valence band tail urbach energies of about 80 to 120 meV and the high thicknesses needed for passivation. To enable application of the investigated layers in SHJ solar cells it will be crucial to improve their structural quality and increase their hydrogen content, resulting in an improved SHJ interface and thinner passivation layers. Homogeneous deposition of a few nanometer thin layers with spin-coating is challenging, but spray coating has already been applied to deposit liquid silicon precursors [62] for thin-film solar cells and could be used for the deposition of thin passivation layers. Moreover using spray coating could enhance the material quality and enable the usage of multi-chamber systems to conduct spray-coating and HPT, in the same system. Once these challenges are overcome, liquid processed a-Si:H offers great promise for development of solution processable a-Si:H layers for photovoltaic and other electronic applications and may present a first step towards solution processable, printable silicon based solar cell structures.
Chapter 7 Valence band alignment and hole transport The contents of this chapter were published in Applied Physics Letters 107 (2015) 013902 under the title "Valence band alignment and hole transport in amorphous/crystalline silicon heterojunction solar cells" [36].
7.1
Abstract
To investigate the hole transport across amorphous/crystalline silicon heterojunctions, solar cells with varying band offsets were fabricated using amorphous silicon suboxide films. The suboxides enable good passivation if covered by a doped amorphous silicon layer. Increasing valence band offsets yield rising hole transport barriers and reduced device effciencies. Carrier transport by thermal emission is reduced and tunnel hopping through valence band tail states increases for larger barriers. Nevertheless, stacks of films with different band gaps, forming a band offset staircase at the heterojunction could allow the application of these layers in silicon heterojunction solar cells.
Author contributions The author list comprises Mathias Mews, Martin Liebhaber, Lars Korte and Bernd Rech from HZB. The author of this thesis was the lead author of the publication. He wrote the article and conceived the experiments. Furthermore he conducted the amorphous silicon oxide, ITO and metal deposition and developed the amorphous silicon oxide deposition recipe. He conducted all the measurements and the AFORS-HET simulations described in this paper. 70
7.2. SHJ VALENCE BAND OFFSET MODIFICATION
71
Martin Liebhaber was involved with a prior study published with equal contribution authorship of Martin Liebhaber and the author of this thesis [197]. The valance band offset measurements conducted in this previous study are used for the simulations described in the present chapter. Furthermore Martin Liebhaber did spectral ellipsometry measurements for process control after the deposition of the amorphous silicon oxide layers and contributed to the discussions of the simulations. He read and commented on the manuscript. Lars Korte was involved in the discussion of this work and provided insights to the simulations described in this paper. He also read and commented on the manuscript. Bernd Rech commented on the simulations and read the manuscript.
Acknowledgment of third parties Chemical wafer cleaning and isolation of cells by photolithography and etching was conducted by the HZB employees Kerstin Jacob and Mona Wittig.
Post publication changes to this chapter Section 7.2 was added to connect this results with the previous chapters.
7.2
SHJ valence band offset modification
The publications discussed in chapters 4, 5 and 6 dealt mainly with aspects of surface passivation in the context of SHJ solar cells. The study in this chapter focuses instead on the current transport across the SHJ. Two of these earlier chapters also discussed the band alignment at the SHJ in relation to the hydrogen content in (i)a-Si:H layers and its influence on the valence band maximum position. A method to reliably tailor the valence band maximum position in a-Si:H and also ∆Ev at the SHJ would offer interesting possibilities, since the influence of the valence band offset at the SHJ on the carrier transport of this interface was the topic of a number of publications [12, 23, 87], but all publications rely either solely on simulations [23], or on device results with unknown band alignment [12, 87]. Therefore the transport across this interface is not fully understood. A post-deposition HPT could modify ∆Ev at the SHJ, by hydrogen introduction, without severely influencing the structural quality of the material. Figure 7.1 shows the valence band position (EV −EF ), Urbach energy (E0v ) and band gap (Egap ) of (i)a-Si:H layers plotted against the electrode to substrate distance (dEL ) during the HPT the
72
CHAPTER 7. VALENCE BAND ALIGNMENT AND HOLE TRANSPORT
Figure 7.1: (a) valence band position (EV − EF ), (b) Urbach energy (E0v ) and (c) band gap (Egap ) of (i)a-Si:H layers plotted against the electrode to substrate distance (dEL ) during the post-deposition HPT the layers were exposed to. Lines are guides to the eye and the sample plotted at a dEL was not exposed to a HPT.
layers received after their depositions. Panel (a) and (c) illustrate that the variation of the HPT allows to tailor EV − EF and Egap of the layers without a strong influence on E0v (figure 7.1b). A change of the dEL influences the electric field in the process chamber, changes the kinetic energy distribution of the hydrogen atoms in the plasma and increases the ion bombardment of the layer. This obviously decreases the amount of hydrogen entering the layer. While currently no experimental data is available that whould allow to verify these hypotheses, it is clear that variations of the HPT process allow to adjust the hydrogen density in the a-Si:H. This would ideally enable to modify the valence band offset at the SHJ to investigate its influence on the carrier transport across the SHJ. A comprehensive picture of the relation between the a-Si:H Egap and its hydrogen content can be assembled from literature and a compilation of different Egap values measured on a-Si:H layers by different authors using different deposition methods and regimes. It is shown in figure 2.1. Unfortunately changing the hydrogen density in a-Si:H only allows to shift the band gap and thereby the band offset by about 500meV. ˙ Additionally it is unclear if a further optimized HPT would allow to access the bigger range. Furthermore according to simulation studies changes in the order of 1 eV would be necessary to significantly alter the carrier transport across the SHJ [23]. Therefore it has to be concluded that changing the hydrogen density in a-Si:H will most likely not allow to conduct a systematic study
7.3. INTRODUCTION
73
of the relationship between ∆Ev at the SHJ and hole transport. A second approach for the adjustment of ∆Ev at the SHJ is to change the valence band maximum in the a-Si:H by alloying the material. Possible additives to increase the a-Si:H band gap are carbon [198] and oxygen [199]. Attaining sufficient surface passivation with silicon carbide is regarded as challenging [200,201]. In contrast amorphous silicon oxide (a-SiOx :H) is known to enable good passivation of the SHJ [199]. Therefore a study of the band alignment at the (i)a-SiOx :H/(n)c-Si SHJ was conducted by Martin Liebhaber and the author of this thesis [197]. This study focuses on the determination of the ∆Ev for the whole stoichiometry range from a-Si:H to a-SiO2 , while the study discussed in the following deals with the implications of a changing ∆Ev on the current transport across the SHJ.
7.3
Introduction
Silicon wafer based solar cells provide most of today’s global photovoltaic power generation. A number of technologies are competing for future market shares. Among those amorphous/crystalline silicon heterojunction (SHJ) solar cells [66] provide the highest potential efficiency, [10] mainly due to the very low defect density at the SHJ. However parasitic absorption in the amorphous silicon (a-Si:H) front contact and passivation layers is restraining their efficiency. [26] The total current density lost in 5 nm p-type a-Si:H is about 1.2 mA/cm2 and about 0.6 mA/cm2 are lost by absorption in 5 nm intrinsic a-Si:H. [26] Therefore the replacement of the front hole contact and the passivation layer with high band gap and low absorption a-Si:H alloys like amorphous, or microcrystalline silicon oxide (a-SiOx :H) [12, 70, 202, 203], or silicon carbide [201] is an ongoing research topic. This letter focuses on the hole transport across the SHJ between a-Si:H, or a-SiOx :H and crystalline silicon (c-Si). This interface serves as a model system for the application of high band gap layers, for reduction of parasitic absorption in SHJ solar cells. [66] A sketch of the band line-up at the SHJ between a p-type a-Si:H hole contact, an intrinsic a-SiOx :H passivation layer and an n-type c-Si absorber is shown in Fig. 7.2. Holes generated in the (n)c-Si are directed towards the p/n-junction, where they have to overcome the valence band offset (∆EV ) to enter the (i)a-SiOx :H and the (p)a-Si:H layer. Values for ∆EV at the SHJ were reported to be about 200 to 450 meV [33,44,86] and are no obstacle to carrier transport, although the exact carrier mechanism is still subject of discussion. [23, 87, 88] Two possible transport paths exist. First, the holes can overcome the band offset barrier by thermionic emission. Secondly, they can tunnel into the a-Si:H passivation
74
CHAPTER 7. VALENCE BAND ALIGNMENT AND HOLE TRANSPORT
Figure 7.2: Band line-up at the hole contact of SHJ solar cells with a n-type c-Si absorber. Holes generated in the c-Si can traverse the SHJ by thermionic emission over the valence band offset (∆EV ), or by tunnel hopping through the valence band tail states of the intrinsic amorphous silicon(oxide). They then travel in the (p)a-Si:H valence band and are collected at the external contacts. The sketch also shows the Fermi-level (EF ), valence band maximum (VBM), conduction band minimum (CBM) and conduction band offset (∆EC ).
layer [23] and then travel by tunnel hopping [87] in its valence band tail states. [204] Tunnel hopping is expected to become increasingly important for increased ∆EV [23] and decreasing electrical quality of the passivation layer. [204] Fujiwara et al. [199] demonstrated the feasibility of a-SiOx :H passivation and demonstrated that increasing CO2 precursor gas fractions lead to decreasing fill factors (FF). Furthermore, studies on a-SiOx :H passivation layers [203,205,206] and their application in solar cells [202] were conducted. Recently Seif et al. [12] reproduced the results of Fujiwara et al. [199] and employed device simulation to reason that an increasing oxygen content leading to a rise of ∆EV may explain the decreasing FF due to a transport barrier, which impedes thermionic emission of holes across ∆EV . Unfortunately Seif et al. [12] used stacks of a-SiOx :H and a-Si:H and only changed the oxygen content of the top layer. Therefore ∆EV at the SHJ may not have been changed at all. Recently we determined ∆EV at this interface for the full stoichiometry range. [197] This letter aims at combining band line-up measurements with device and passivation results to gain insight about the carrier transport across this junction.
7.4. EXPERIMENTAL AND SIMULATION DETAILS
7.4
75
Experimental and simulation details
To this end, (p)a-Si:H/(i)a-SiOx :H/(n)c-Si/(i,n)a-Si:H solar cells were processed and characterized, as follows: The substrates for solar cell preparation are 280 µm thick, polished phosphorous doped float-zone grown c-Si wafers with (111) surface orientation and a resistivity of 3 Ωcm. Wafers were cleaned using the RCA process and dipped in diluted hydrofloric acid (1 %, 2 min) before layer deposition. A-Si:H and a-SiOx :H layers were deposited using plasma-enhanced chemical vapor deposition. Intrinsic a-SiOx :H was prepared using 60 MHz excitation, a process pressure of 0.5 mbar, a substrate temperature of 175◦ C, an electrode distance of 2 cm, a plasma power density of 56 mW/cm2 , and a total gas flow of 15 sccm. The gas flow consisted of 5 sccm hydrogen, and a total of 10 sccm silane and CO2 . The a-Si:H layers were deposited using 10 sccm of silane and no CO2 . For the a-SiOx :H layers, the CO2 flow was raised in 1 sccm steps up to a value of 5 sccm, and the silane flow was decreased accordingly. For solar cell fabrication wafers were coated with 4 nm intrinsic a-Si:H and 8 nm n-type a-Si:H on the backside and 5 nm intrinsic a-SiOx :H and 8 nm p-type a-Si:H on the front side. The solar cells were then completed by ITO sputtering and metalization with Ti/Ag stacks. The full solar cell process is discussed elsewhere. [34] Photoconductance decay and illumination dependent open circuit voltage (so-called SunsVOC ) measurements [126] were carried out in between various process steps. ∆EV of the a-SiOx :H/c-Si-SHJs were determined using photoelectron spectroscopy in the constant-final-state-yield mode [146] and a procedure described elswhere. [44, 86] The measurement of oxygen content and ∆EV in these layers was discussed in an earlier publication. [197] Numerical device simulation with AFORS-HET [89] was used to discuss the experimental data. The a-SiOx :H layers were simulated using a model for a-Si:H with an electron affinity of 3.724 eV and a defect density consisting of two Gaussian defect densities of about 1014 cm−3 and band tails at the valence and conduction band. The band gap of the pure a-Si:H was assumed to be 1.68 eV. For a-SiOx :H, it was increased to mirror the measured increase in valence band offset, holding the conduction band position constant. The Urbach energy (E0v ) of the valence band tail was adjusted according to measured values [197] and the urbach energy of the conduction band tail (E0c ) was set to two thirds of the valence band E0v . [207]
76
CHAPTER 7. VALENCE BAND ALIGNMENT AND HOLE TRANSPORT
Figure 7.3: Band alignment and interface passivation properties of amorphous silicon suboxide/crystalline silicon heterojunctions (a-SiOx :H/c-Si-HJ) with different oxygen fractions. (a) Valence band offsets (∆Ev ) at the SHJ, as measured in our earlier study. [197] (b) Minority carrier lifetime at an injection level of 1015 cm−2 (τ15 ) for c-Si passivated with a-SiOx :H (black squares) and after subsequent deposition of a p-type a-Si:H layer on top of the a-SiOx :H, measured on the same samples (empty blue circles). (c) a-SiOx :H/c-Si-HJ interface defect density (Dit ) of the same samples as in (b).
7.5 7.5.1
Results and discussion SHJ valence band alignment and passivation
Fig. 7.3a shows the evolution of ∆Ev for the layers used in this study. An increase in the oxygen fraction leads to an increase of ∆Ev . Additionally, Fig. 7.3b displays the evolution of the effective minority carrier lifetime in solar cell precursors with an intrinsic a-Si:H passivation and n-type a-Si:H contact layer on the back side, and an (i)a-SiOx :H layer on the front side. The minority carrier lifetime decreases for increasing oxygen content, but rises drastically in the samples with higher oxygen concentrations after the deposition of an additional p-type a-Si:H layer on top of the (i)a-SiOx :H layers. To further investigate this process the interface defect density was extracted from the injection dependence of the minority carrier lifetime. [127] The results are depicted in Fig. 7.3c. It is obvious that the lifetime gain is based on a decreased interface defect
7.5. RESULTS AND DISCUSSION
77
density (Dit ) upon (p)a-Si:H deposition. Note, that the model applied for the extraction of the Dit assumes symmetrical samples and the samples in the present study were not symmetric. This may lead to an overestimation of the calculated Dit , if the Dit at the investigated interface is of the same order of magnitude as at the sample’s backside. However the error is below 10 % for effective minority carrier lifetimes below 4 ms, since the backside passivation enables effective lifetimes of about 12 ms. The Dit at the SHJ interface is directly related to the Si bonding environment [19] and the hydrogen density [33] at the interface. Since the temperature of the (p)a-Si:H deposition is too low to affect the Si bonding structure, [59] the decrease in Dit is likely due to an increase of the hydrogen density at the SHJ. It is conceivable that the additional hydrogen is provided by the plasma process during the (p)a-Si:H deposition.
7.5.2
Valence band offset and transport across the SHJ
The current-voltage (j(U)) curves of solar cells fabricated with these layer stacks are depicted in Fig. 7.4a. An increase of ∆EV at the hole contact leads to the development of s-shaped j(U) curves. This behavior was predicted on the basis of numerical simulations, [12, 23, 87] but no systematic experimental evidence has been presented so far. Fig. 7.4b depicts the solar cell’s open circuit voltages (VOC ), together with the implied open circuit voltages (iVOC ), [126] as extracted from photoconductance decay measurements prior to ITO depositions. The VOC of the solar cells decreases slowly for increasing ∆EV . Increasing the oxygen fraction in a-SiOx :H layers increases the layer porosity and the density of Dihydrides and hydrogen filled voids. [206] Furthermore hydrogen in Dihydride configuration and in voids is more mobile, than Monohydrides, and is driven out of the layer during ITO sputtering. [208] This could explain why layers with higher oxygen contents degrade stronger during follow up processes than layers with lower oxygen contents and show decreased minority carrier lifetimes after ITO sputtering. Fig. 7.4c displays the dependence of fill factors (FF) and pseudo fill factors (pFF) of the solar cells on ∆EV . The pFF was measured using SunsVOC , a method that induces no external current flow in the device and therefore reflects the maximum FF excluding carrier transport related effects. The FF decreases with increasing ∆EV , whereas the pFF is slightly increased. Consequently the decreasing FF can be related to a transport barrier and effects of e.g. decreased passivation at low injection can be excluded. [24,25,209,210] The reason for the increase of the pFF with decreasing FF is unknown, although similar results have been reported before. [77, 211] One possibility is, that the transport barrier increases the passivation at low injection densities and no
78
CHAPTER 7. VALENCE BAND ALIGNMENT AND HOLE TRANSPORT
Figure 7.4: Influence of ∆EV at the SHJ on (a) the current-voltage (j(U)) characteristics of SHJ solar cells with different a-SiOx :H passivation layers, (b) the open circuit voltage (VOC ) and implied VOC (iVOC ) and (c) the solar cell fill factor (FF), SunsVOC pseudo fill factor (pFF), simulated fill factor (sFF) of the same solar cells.
7.5. RESULTS AND DISCUSSION
79
net current flow. Numerical simulations were used to simulate FF (sFF) values for the same devices, which are also plotted in Fig. 7.4b. Obviously the simulation overestimates the influence of ∆EV on the FF. Since the FF of the solar cells decreases with increasing ∆EV , while the pFF increases, it can be concluded that the increasing ∆EV creates a transport barrier. Furthermore, the overestimation of the FF degradation with increasing ∆EV in the numerical simulation gives further insight into the carrier transport mechanisms. Only thermionic emission is employed to simulate the transport across the SHJ. The assumption of thermionic emission as the main transport mechanism in SHJ solar cells is commonly used in simulation studies. [12, 212–214] However, the results depicted in Fig. 7.4 imply that another transport path becomes prevalent once ∆EV is increasing well above the thermal energy of carriers at 300 K. Indeed, there are a few simulation studies which ascribe a significant importance to tunneling through the interface [23, 88] or tunnel hopping in valence band tail states [87, 215] in the a-Si:H layers. Comparing the results of these studies and our simulations to the experimental data in Fig. 7.4, it is clear that all simulations of the hole transport using only thermionic emission overestimate the detrimental influence of high ∆EV . [88,212,213] In contrast, simulations that include tunnel hopping in a-Si:H, or its alloys show good agreement with our experimental data. [23, 87] Kanevce and Metzger [23] discuss the transport across the interface between n-type a-Si:H and p-type c-Si in detail. Their findings include that a significant fraction of the current flow across the SHJ is delivered via tunneling into and tunnel hopping in the a-Si:H layers. Their simulated j(U) curves for varying band offsets at the minority carrier contact qualitatively match the experimental findings of the present study. Note, that their simulations treat devices of inverse polarity compared to our study. Still it is reasonable to compare these results to the present study, as the respective electron and hole mobilities are of the same order of magnitude and the band offset ranges are comparable. All in all, for minority carrier band offsets above 400 meV, it is mandatory to consider tunneling into and tunnel hopping in the a-Si:H layer to obtain sufficient agreement of experimental data and simulation.
7.5.3
Valence band stairwell
The presented experimental results have important implications for the application of high band gap alloys at the hole contact of SHJ solar cells, [70, 201–203] since all Si alloys reported in literature feature high ∆EV values. [44, 197, 198, 216] This is most
80
CHAPTER 7. VALENCE BAND ALIGNMENT AND HOLE TRANSPORT
unfortunate, since the application of those high band gap alloys is desirable to reduce the parasitic absorption in front side minority carrier contacts. [26] However one way to enable the application of high band gap alloys may be the application of a stack of materials. This stack could comprise an a-Si:H layer with a moderate ∆EV of about 200 to 300 meV and a second layer with a higher band gap. This splits the effective ∆EV between two interfaces and thermionic emission across each single interface is facilitated. Thus, hole transport is enabled across a valence band staircase, which allows carrier transport across a larger ∆EV , which charge carriers could not overcome efficiently in a single step. Additionally, stacks comprising even more layers or graded layers are a possibility. Solar cells comprising hole contact passivation layer stacks of 2 nm a-Si:H and 3 nm a-SiO0.3 :H were fabricated to test this approach. The band offset between a-Si:H and c-Si is 270±50 meV, while ∆EV at the a-SiO0.3 :H/c-Si SHJ is about 585±50 meV. [197] The j(U)-characteristics of such a solar cell is shown in Fig. 7.5a together with a reference. The solar cell in which the holes have to traverse ∆EV in a single step displays a clear transport barrier, as reflected in the s-shaped j(U)-curve and a fill factor of only 63 %. In contrast, there is no transport barrier, albeit a somewhat reduced FF of 70 %, for the solar cell in which ∆EV is divided between two interfaces. The solar cells with an a-Si:H passivation layer have a FF of 78 %. Note that the VOC ’s of the two cells are identical, indicating comparable interface defect densities. Numerical simulations were conducted to further illustrate this concept. Solar cells comprising different ∆EV at the SHJ and devices with the same total ∆EV , but a 2 nm interlayer with a ∆EV of 270 meV were simulated using only thermionic emission as transport mechanism. The FF values from these simulations are shown in Fig. 7.5b. As expected, the application of a material stack does not change the trend of decreasing FF with increasing ∆EV , but increases the FF for a given valence band offset. While the experimental stack will likely feature increased hopping currents, due to the only 3 nm thin oxide layer, it still illustrates the concepts. Furthermore, since the simulations do not include hopping the reason for the improved FFs is most likely more efficient thermionic emission along the layer stack, as compared to the single interface with one large ∆Ev .
7.6
Conclusion
In conclusion it was possible to demonstrate the importance of an appropriate band alignment at the p-n-junction of SHJ solar cells. Specifically, an increased ∆EV at the hole contact of SHJ solar cells gives rise to a transport barrier. Moreover, the increasing
7.6. CONCLUSION
81
Figure 7.5: (a) j(U)-curves of solar cells with a 5 nm thick a-SiO0.3 :H passivation layer and a solar cell with a passivation layer stack consisting of 2 nm a-Si:H and 3 nm aSiO0.3 :H. The ∆EV of the a-Si:H/c-Si interface is about 270±50 meV, whereas ∆EV between the a-SiO0.3 :H and the c-Si is about 585±50 meV. (b) FF extracted from simulated j(U) curves using thermionic emission over an a-SiOx :H/c-Si interface with variable ∆EV and the same simulation for a stack of a-SiOx :H/a-Si:H/c-Si. For the stack, the total ∆EV is plotted on the abscissa. ∆EV for the a-Si:H/c-Si-SHJ in the stack is 270 meV.
82
CHAPTER 7. VALENCE BAND ALIGNMENT AND HOLE TRANSPORT
transport barrier leads to an increasing contribution of tunnel hopping through valence band tail states and tunneling transport and a decreasing contribution of thermionic emission, which in turn decreases device efficiencies. The latter effect is a general problem for high band gap a-Si:H alloys as hole contacts of SHJ solar cells. An approach to mitigate this problem by applying a stack of at least two layers with a step-like increase of ∆EV at the SHJ was presented. The improved FFs of the solar cells with stacked passivation layers highlight a viable approach towards the application of high band gap a-Si:H alloys as hole contact layers in SHJ solar cells. This concept is promising for the combination of a passivation layer with a moderate band gap and a high band gap hole contact.
Chapter 8 Prospects for silicon heterojunction solar cells This chapter includes results on monolithic tandem solar cells with perovskite top cells and silicon heterojunction bottom cells, which were published by Steve Albrecht et al. in Energy Environ. Sci. 9 (2016) 81-88 under the title "Monolithic perovskite/siliconheterojunction tandem solar cells processed at low temperature" are presented. The author of this thesis fabricated the SHJ bottom cells and made the necessary adjustments to their fabrication process, for the inclusion in tandem devices.
8.1
Introduction
The photovoltaic application of amorphous/crystalline silicon heterojunctions (SHJ) was pioneered by Sanyo in 1992 (since 2010 Panasonic) and further developed over the course of the past 24 years [9, 10, 15–18], culminating in a record power conversion efficiency of 25.6 % [11]. The SHJ solar cell with a thin intrinsic layer is a wafer based solar cell, which includes a p/n-junction formed by the deposition of a doped a-Si:H on inversely doped c-Si [9]. Additionally a thin (i)a-Si:H layer is added in between the c-Si and the doped a-Si:H layers. This intrinsic layer passivates the c-Si surface and enables a very low defect density at the junction, which leads to a very high open circuit voltage of up to 750 mV [10]. This open circuit voltage is about 40 mV higher than values reached by silicon homojunction solar cells [8] and comparatively close to the maximum theoretical value of 769 mV [68]. The only other contact enabling similar voltages of about 700 to 725 mV is a contact including a SiOx tunnel contact [217]. Their high open circuit voltages enable SHJ solar cells to reach one of the highest device efficiencies among the crystalline silicon based solar cells [218]. Companies like Kaneka [219], LG [220], Choshu [221] and Sharp [222] have joined Panasonic in 83
84CHAPTER 8. PROSPECTS FOR SILICON HETEROJUNCTION SOLAR CELLS producing, or developing solar modules of this type and equipment providers like Meyer Burger [218] are offering production equipment, or whole turn-key production lines. The goal of the following sections is to outline future research possibilities in the field of SHJ solar cells and to discuss how these relate to the results of this thesis.
8.2
Carrier selective contacts in SHJ solar cells
Chapter 7 dealt with the hole transport across the SHJ. It was found that the valence band offset at the SHJ is significantly impeding the hole transport, if it is larger than 300 meV. This valence band offset is tied to the a-Si:H hole contact’s band gap and increases with increasing band gap [44]. A larger band gap would reduce the parasitic absorption in the hole contact [26], but introduces a transport barrier in the device (cf. chapter 7). This leads to the question, which alternative contact layers for SHJ solar cells are available. The following section will address this issue. Charge selective contacts on c-Si absorbers will be discussed in general. Then the ITO/(p)a-Si:H contact used in SHJ solar cells will be discussed, followed by review of available contact layers for the hole and electron contacts. A case study on the influence of oxygen vacancies in tungsten oxide hole contacts for SHJ solar cells will be presented.
8.2.1
Charge selective contacts
In a very general picture a solar cell can be regarded as an absorber layer sandwiched between two charge selective contacts [223,224]. This is sketched in Figure 8.1. Incident photons are absorbed in an absorber layer and diffuse to the selective contacts. The Fermi-level splitting (∆EF ) in the absorber determines the maximum achievable open circuit voltage of the final solar cell. Each carrier selective contact should enable effective transport of one carrier type and effective blocking of the other. Therefore a good hole contact needs a minimal valence band offset, but a large conduction band offset, whereas the electron contact needs a low conduction band offset and a high valence band offset. One way to obtain such a structure is a double heterojunction solar cell, consisting of three materials with suitable electron affinity (Eea ) and band gap. A suitable band alignment is shown in Figure 8.1. Additional surface passivation layers may be needed, to suppress defect states in the heterojunction. Intrinsic a-Si:H [9] and SiOx tunnel contacts [225] can provide adequate passivation of c-Si surfaces. Figure 8.2 shows simulations of the FF of an n-type c-Si absorber, with (i)a-Si:H layers on both sides and carrier selective contacts. The contacts are Schottky contacts with a variable work function and are defined to block one carrier type and conduct the other.
8.2. CARRIER SELECTIVE CONTACTS IN SHJ SOLAR CELLS
85
Figure 8.1: Sketch of a general solar cell concept. An absorber layer is sandwiched between two contacts for the two polarities. The charge selective contacts each have one big band offset to the absorber to block one carrier type and one small band offset to conduct transport of the other carrier type.
Figure 8.2: Simulation of the FF of a device consisting of an n-type c-Si absorber, passivated by (i)a-Si:H on both sides and with charge selective contacts on either side. The hole contact’s work function (WF) was varied from 6 eV to 4.3 eV, while the electron contact’s WF was kept at 3 eV and the electron contact’s WF was varied from 3 eV to 4.25 eV, while the hole contact’s WF was kept at 6 eV. The electron affinicty (EA) and ionization energy (IE) of c-Si are shown as vertical lines.
86CHAPTER 8. PROSPECTS FOR SILICON HETEROJUNCTION SOLAR CELLS The figure also shows the electron affinity (EA) and ionization energy (IE) of c-Si. The electron contact’s work function was kept at 3 eV, while the hole contact’s work function was varied. The FF is constant as long as the work function of the hole contact is higher than the ionization energy of c-Si. Additionally, it starts to decrease once the work function is lower than the ionization energy and drops drastically once the work function is smaller than 5 eV. The reason for the decreasing FF with decreasing work function at the hole contact is an increasing hole transport barrier, because the contact is below the ionization energy of c-Si. The same simulation was repeated for the electron contact, while the work function of the hole contact was kept at 6 eV. Once the work function of the electron contact is bigger than the electron affinity of c-Si, the FF decreases. The reason for the decreasing FF is an increasing electron transport barrier, once the work function is higher than the electron affinity. Additionally the simulation becomes numerically instable if the work function of the electron contact is higher than 4.25 eV, since the extraction of excess electrons from the absorber layer is no longer possible. Overall this shows that hole extraction from silicon necessitates a contact level below 5.3 eV less than the vacuum energy and electron extraction necessitates a contact level above 4 eV below the vacuum energy.
8.2.2
ITO/(p)a-Si:H contact
After discussing charge carrier selective contacts for SHJ solar cells in general, it is interesting to review the properties of the ITO/(p)a-Si:H contact, which is commonly used in SHJ solar cells. Comparing the band diagram of a SHJ solar cell (cf. Figure 2.3) with the sketch in Figure 8.1 it is obvious that the SHJ solar cell is a solar cell with carrier selective contacts and an intermediate absorber layer. The electron contact appears to be a typical carrier selective contact, consisting of a stack of (n)a-Si:H and ITO, or AZO. The hole contact in contrast features a different band alignment along its (p)a-Si:H/ITO contact. The (p)a-Si:H/ITO-interface within this contact is a tunnel recombination contact. This contact is necessary, since a more conductive material is needed on (p)a-Si:H to offer lateral conductivity [22] and it has to be a recombination contact, because no conductive and transparent material with a suitable electron affinity is widely available. Unfortunately the work function of ITO is only 4.2 to 4.7 eV [226], therefore a significant hole transport barrier is featured at this contact, which is reflected in high series resistances and inherent fill factor limitations [25]. Therefore replacing ITO with high work function metal oxides like WOx [25] has recently started. Additionally this led to experiments in which the ITO/(p)aSi:H stack was replaced with metal oxides like MoOx [76, 78] and WOx [80], thereby
8.2. CARRIER SELECTIVE CONTACTS IN SHJ SOLAR CELLS
87
moving the tunnel recombination contact to the TCO/(i)a-Si:H interface, omitting the p/n-junction. This leads to the question, which suitable materials for carrier selective contacts on silicon are known and could be applied to overcome the inherent fill factor problem of SHJ solar cells with ITO/(p)a-Si:H hole contacts. Also, are there any materials better suited for the back contact of the SHJ than (n)a-Si:H?
8.2.3
Hole contact materials
Holes generated in the c-Si absorber can be collected from the valence band of the (i)a-Si:H passivation layer, or tunnel directly from the c-Si valence band. The electron affinity of a-Si:H is about 3.7 eV and its band gap is in the range of 1.6 to 1.8 eV. This adds up to an ionization energy of 5.3 to 5.5 eV. The electron affinity of c-Si is about 4.05 eV and with its band gap of 1.15 eV this adds up to an ionization energy of 5.2 eV. It is possible to either form a p/n-junction were the holes are collected into the valence band of the contact layer, or a tunnel recombination junction with an n-type semiconductor, or metal were the holes are recombining with electrons from the contact’s conduction band (or highest occupied molecular orbital (HOMO) for an organic semiconducor). For the p/n-junction a p-type semiconductor with an ionization energy smaller than 5.2 eV is needed. Additionally the band gap should be high enough to effectively block electrons from entering the contact layer and to guaranty low parasitic absorption in the layer. For the recombination contact an n-type semiconductor, or a metal with a contact energy level more than 5.5 eV below the vacuum level would be needed. This contact level could be the conduction band of a semiconductor, the Fermi level of a degenerate semiconductor, or metal, or the last unoccupied orbital (LUMO) of an n-type organic semiconductor. Additionally the contact layer should have a high carrier density at the contact, a sufficient lateral and vertical conductivity, but still feature low absorption at and above the absorber band gap. Four material classes contain potential candidates. The first are those metals, which can form Schottky recombination contacts on silicon. The problems of metal Schottky contacts on silicon are plentiful. Firstly the work functions of metals like gold (5.1 to 5.5 eV), platinum (5.1 to 5.9 eV), or palladium (5.2 to 5.6 eV) are only barely high enough. Additionally direct metal to silicon contacts are defect rich and incompatible with common silicon passivation layers. Some metals form silicides with a-Si:H and lead to the degradation of the carrier lifetime [124], some like gold diffuse into the respective passivation layer, or the contact degrades due to oxidation of the metal.
88CHAPTER 8. PROSPECTS FOR SILICON HETEROJUNCTION SOLAR CELLS Also metal contacts of sufficient thickness lead to strong light absorption and have to be optically decoupled from the absorber to reduce absorption by plasmons [69]. Therefore pure metal contacts are not an option for carrier selective contacts in SHJ solar cells. Another possibility would be p-type semiconductor layers. For this a p-type semiconductor with a high band gap, low absorption below its band gap and suitable band alignment would be needed. GaP has a band gap above 2 eV, good electrical quality, low sub band gap absorption and was investigated for this application [227, 228], but efficiencies of only 8 % [229] were reported so far, since the epitaxial growth of GaP is difficult on c-Si and impossible on a-Si:H, or SiO2 . Especially the missing passivation layer, but also contaminations of the c-Si wafer during the heteroepitaxial GaP growth by metal-organic vapor phase epitaxy have so far prevented the preparation of adequate GaP/Si solar cells. Amorphous p-type semiconductors could be grown on a-Si:H, or SiOx passivation layers. Of the known candidates NiO, Cr2 O3 and Co3 O4 [73] only NiO was investigated as a hole collector on n-type silicon [230]. However the solar cells were only measured with non-standard illumination, showed low VOC and FF values and no efficiencies were reported [231]. All in all no other semiconductor has yet shown better results than the (p)a-Si:H hole collector. Although (p)a-Si:H’s inherently low lateral conductivity necessitates a more conductive second layer to enable sufficient lateral conductivity [22]. Different derivatives of a-Si:H like amorphous, or microcrystalline silicon carbide [201], or silicon oxides [70] have been employed. Some of them enable reduced parasitic absorption in the hole conductor [70, 201], but none of them will reduce the band discontinuity at the recombination junction between the hole conductor and the transparent conductive oxide, since introduction of oxygen [197] and carbon [198] leads to an increase of this offset by moving the hole conductor’s valence band further away from the vacuum energy and the crystalline part of microcrystalline silicon features the same electron affinity as crystalline silicon [232]. The third material class are organic semiconductors yielding hybrid solar cells [233]. Some p-type polythiophenes like PEDOT:PSS [234] and P3HT [235] have been incorporated as hole collectors on n-type silicon wafers. Most of these studies are still in the proof of concept stage and limited to a power conversion efficiency in the range of 10 % if the organic semiconductor is used as minority carrier contact. The major problem of this approach is that the organic semiconductor offers no advantage over amorphous silicon, as it also needs a second contact layer for lateral conductivity. On the contrary, these devices create new problems since they feature similar stability issues as all devices using organic semiconductors, which adds a high level of uncertainty to the
8.2. CARRIER SELECTIVE CONTACTS IN SHJ SOLAR CELLS
89
long term stability of potential modules. Devices based on inorganic semiconductors already offer 30 years of almost stable electricity production. Therefore the application of organic semiconductors in silicon based solar cells is unlikely to enter industrial production. The last but most promising material class includes n-type high band gap metal oxide semiconductors. These can be used to fabricate recombination contacts on (n)c-Si. Metal oxides have high band gaps and comparatively low parasitic absorption. Many of them can be n-doped by native, or induced oxygen vacancies [236]. Additionally there is a group of metal oxides with work functions above 5.5 eV [73]. Among them are MoO3 , WO2 and V2 O5 . All three of those are n-doped by oxygen deficiency. Especially MoO3 [237], but also WOx [75, 238] are used as hole collectors in solar cells with organic semiconductor absorber layers. Some studies about using either MoO3 [76–79, 81, 82, 239], or WO3 [80, 240] as hole collectors in SHJ solar cells have already been conducted. Another possible, but rarely studied material is V2 O5 [236]. This approach to the SHJ solar cell has two advantages over the conventional ITO/(p)a-Si:H heterojunction: The replacement of the (p)a-Si:H contact with the far more transparent metal oxide can reduce parasitic absorption in the front contact [26,80], and the higher work function of MoO3 and WO3 has enabled higher real [78] and implied FFs [80] than the ITO/(p)a-Si:H contact.
8.2.4
Oxygen vacancies in tungsten oxide hole collection layers
Tungsten oxide is among the more promising hole contact layers for silicon heterojunction solar cells. So far MoOx hole collection layers have yielded higher fill factors than WOx hole collection layers [80,236]. This was reported for the metals oxides deposited on (i)a-Si:H passivation layers [80] and for deposition directly onto c-Si [236]. In contrast both layers perform equally well, if they are deposited on (p)a-Si:H to adjust the work function at the TCO/(p)a-Si:H interface. The following experiments were conducted to explain why tungsten oxide hole collection layers on silicon tend to yield low solar cell FFs. Chemical analysis of tungsten oxide Tungsten oxide was deposited by thermal evaporation at 920◦ C from a tungsten crucible and analyzed by photoelectron spectroscopy without leaving the ultra-high vacuum system. This allows to investigate the chemical structure and also the work function of the material: Especially the latter parameter, being a surface property and sensitive
90CHAPTER 8. PROSPECTS FOR SILICON HETEROJUNCTION SOLAR CELLS
Figure 8.3: XPS spectra of the W 4f peak of WO2.89 (a) and WO2.97 (b). The measured data (black) was fitted with the sum (red) of the doublets of W6+ (green), W5+ (blue), the signal shifted by oxygen vacancies (brown) and a Shirley background (violet). The residual (orange) is also shown. to adsorption/contamination, is likely to change upon vacuum break. Figure 8.3 shows X-ray photoelectron spectra for the W4f orbital of WOx . The spectra consist of a doublet of 7/2 and 5/2 spin states; each half of the doublet consists of three peaks arising from tungsten with an oxidation number of 6+ (WO3 ), 5+ (W2 O5 ) and a third component, which relates to oxygen vacancies at the surface, whose potential leads to a lateral inhomogeneous barrier for electron extraction with a local difference of 0.5 eV [241, 242]. The intensity of this signal is a measure of the oxygen vacancy density at the surface. Also this measurement allows calculating the stoichiometry of WOx , using the relative peak areas of the respective oxidation states [238]. The stoichiometry of tungsten oxide is known to influence its work function [75]. Figure 8.4a shows the work function of WOx plotted against its stoichiometry. It is visible that WOx with a stoichiometry close to WO3 has a work function of more than 6 eV. It is therefore suitable for application as hole contact layer in SHJ solar cells, or for application in the tunnel recombination contact with a (p)a-Si:H hole contact. The band bending induced by WOx on (i)a-Si:H/c-Si samples was measured using the surface photovoltage method and is plotted in Figure 8.4b against the intensity fraction of the oxygen vacancy related
8.2. CARRIER SELECTIVE CONTACTS IN SHJ SOLAR CELLS
91
Figure 8.4: (a) Work function of tungsten oxide plotted against the stoichiometry of the layer. (b) Band bending induced by tungsten oxide on n-type c-Si plotted against the oxygen vacancy concentration at the tungsten oxide surface. The straight line at 640 meV marks the band bending reached with a (p)a-Si:H layer. signal in the W4f measurements. This intensity fraction is a qualitative measure for the density of oxygen vacancies in WOx [6]. The oxygen vacancies are known to ndope the layer [236], and it is therefore obvious that negative charge supplied by these oxygen vacancies leads to reduced band bending in the c-Si. These oxygen vacancies are predominantly generated if the WOx is deposited onto a-Si:H (Figure 8.4b). An explanation could be hydrogen diffusion from the underlying a-Si:H, since hydrogen is known to reduce WOx [243]. Solar cells results Solar cells using tungsten oxide as a hole contact (instead of (p)a-Si), or as a contact layer to (p)a-Si:H were fabricated. The band bending in the solar cells was measured using surface photovoltage measurements on the active cell area of the finished devices. Plots of the solar cell VOC ’s and FF’s in dependence of the band bending are shown in Figure 8.5. Clearly, a high band bending is correlated to a high VOC and FF. Connecting these results with those from the previous section it can be postulated that a high concentration of oxygen vacancies and therefore donors in WOx leads to an increase of the Fermi-level in WOx . This, in turn, leads to a low band bending in the n-type c-Si absorber. The low band bending leads to poor hole selectivity of the WOx /(n)c-Si-junction and a low VOC and FF of WOx /(i)a-Si:H/(n)c-Si solar cells. This reduction of band bending has no impact, if WOx is deposited on a (p)aSi:H hole contact instead of the (i)a-Si:H passivation layer, since the (p)a-Si:H already induces a sufficient band bending. From these results, in can be concluded that the
92CHAPTER 8. PROSPECTS FOR SILICON HETEROJUNCTION SOLAR CELLS
Figure 8.5: (a) VOC of SHJ solar cells with tungsten oxide contact layer plotted against the band bending measured in the same devices. (b) Fill factor of SHJ solar cells with tungsten oxide contact layers plotted against the band bending measured in the same devices. concentration of oxygen vacancies in WOx has to be strongly reduced to fabricate viable SHJ solar cells with WOx hole contacts, and to realize an improvement of SHJ solar cells with tungsten oxide hole collection layers. Therefore high rate thermal evaporation and high temperature vacuum processes after the WOx deposition should be avoided, since WO3 can be reduced in oxygen devoid environments at 700◦ C and higher temperatures [244]. In contrast deposition methods like reactive sputtering [244], or chemical vapor deposition [245], which allow to adjust the oxygen concentration in the layer should be applied.
8.2.5
Electron contact materials
Electron contacts of SHJ solar cells should collect electrons from the conduction band of c-Si at about 4.05 eV below the vacuum level, or the (i)a-Si:H passivation layer at about 3.7 eV below the vacuum level. Additionally the back contact of a solar cell needs to feature a layer with a refractive index below 2, too enable effective light management at the back side [69]. The contact could consist of a single layer fulfilling both properties. Like a suitable metal oxide [73, 246], or a stack of one layer with the right work function and a second thicker layer to fulfill the optical requirements [247]. Possible materials are inorganic, or organic semiconductors in combination with a more transparent second layer (e.g. the commonly used (n)a-Si:H/ITO stack [25]), Alkali
8.2. CARRIER SELECTIVE CONTACTS IN SHJ SOLAR CELLS
93
metal salts [239, 248], or metal oxides [73, 246] and direct metal contacts. Most alkali metals have work functions between 2 and 3 eV, but these metals are unstable if exposed to moisture and materials with a lower reflective index are needed due to optical requirements of a solar cell back contact. As explained in the previous section organic semiconductors are most likely not an option due to their instability. Thus, the three interesting material classes are n-type semiconductors, low work funtion n-type metal oxides and alkali metal fluorides in crystalline structure. N-doped a-Si:H has the same electron affinity as the intrinsic passivation layer. Additionally, (n)a-Si:H and its derivatives are the only common non-oxide semiconductor material that is compatible with the typical silicon surface passivation materials (i)aSi:H and SiOx . It is still beneficial to include a TCO on top of the (n)a-Si:H to reduce the plasmonic absorption at the metal back contact [69,249]. Multiple groups use AZO as the back side TCO, since the TCO conductivity is less critical at this contact [27], the lower work function of AZO is slightly beneficial [25] and AZO is cheaper than ITO. Other low work function metal oxides suitable for hole extraction from silicon are TiOx [246] and Cs2 CO3 [248]. These contact materials are not widely studied, but have already yielded silicon solar cells with efficiencies of 20.5 % [246] and 14 % [248], respectively. Although both of these devices used the TiOx [250] and the Cs2 CO3 [248] as majority carrier contact and own their output features mainly to their minority carrer contact. Recently alkali metal flourides like LiFx , KFx and CsFx were studied for the electron contact in SHJ solar cells [239]. These materials offer a nearly perfect band alignment, because they have work functions in the range of 2 to 2.9 eV [239], but very high band gaps. Additionally they are mostly transparent in the visible range. The downside of this materials is, that they feature low conductivities and are dissolving in water. Additonally this very thin contact layers could be sensitive to sputter damage, which may prohibit contact stacks of alkali flourides and ITO. The devices reported so far feature alkali flouride contacts with metal overlayers [239], which are optically inferior to contacts with ITO/metal stacks. Still the application of silicon and dopant free charge carrier selective contacts in silicon heterojunction solar cells offers great potential and may lead to substantial process simplifications. The combination of MoOx hole contacts and LiFx electron contacts has already enabled 19.4 % efficiency in silicon based solar cells with (i)a-Si:H passivation layers [239].
94CHAPTER 8. PROSPECTS FOR SILICON HETEROJUNCTION SOLAR CELLS
Figure 8.6: (a) Sketch of an IBC solar cell. The design includes a front side antireflection and passivation SiNx layer (blue), additionally a diffused, or a-Si:H front surface field is placed below the anti-reflection layer (blue shading). The intrinsic aSi:H passivation layer at the back-side is continuous, whereas the n- (red) and p-doped (green) a-Si:H layers are structured. The metal, or metal/TCO contacts on these aSi:H layers are smaller than the respective contact regions to prevent shunts. The doped a-Si:H layers may overlap, since their lateral conductivity is low. (b) Sketch of a top-view of an IBC structure.
8.3 8.3.1
Back-contact back-junction solar cells state of the art
Back-contact back-junction solar cells have not been discussed so far. Therefore this section outlines the state of the art regarding those devices and the next section highlights future research routes and their connection to the work of this thesis. SHJ solar cells in back-contact back-junction (BCBJ) configuration [157] combine the high current potential of BCBJ solar cells with the high-voltage potential of SHJ solar cells. Consequently an efficiency of 25.6 %, the highest efficiency ever achieved with a silicon wafer based solar cell, was demonstrated with a SHJ-BCBJ device [11]. Most solar cells of this type share an interdigitated back contact (IBC) struture [11,28,157,222] as sketched in Figure 8.6. The difference lies in the process used to fabricate the structure. Usually a combination of passivation and anti-reflection layers is used on the solar cell front side. Common examples are stacks of silicon nitride and either silicon oxide, or aluminum oxide [28]. The back side is covered with an (i)a-Si:H passivation layer, which is followed by full-area doped layer depositions and structuring steps, during which the a-Si:H stack is partly removed (eg. by photolithography and etching [28, 157, 222], or laser structuring [180]), or doped layer deposition through masks [29]. These processes need an increased amount of vacuum based deposition processes, aligning and cleaning steps due to the necessary barrier, or protection layers and the removal of layers. In contrast masked depositions reduce the amount of process steps, but suffer from inhomogeneous depositions [29] and also necessitate difficult aligning steps. Overall
8.3. BACK-CONTACT BACK-JUNCTION SOLAR CELLS
95
SHJ-BCBJ solar cells offer the highest efficiency of all silicon wafer based solar cells, but the complex process leads to higher cell fabrication costs than for other technologies [14], because of lower yield, throughput and increased PECVD cost. Potentially IBC-BCBJ SHJ solar cells could still be cheaper produced than conventional SHJ solar cells, since the metallization cost drops dramatically, if the use of silver pastes is omitted and sputtered Ag contacts are used [14].
8.3.2
Printable silicon wafer based BC-BJ solar cells
Printing equipment for either inkjet, screen [251], or stencil printing [66] is standard production equipment in solar cells factories. Inkjet and stencil printing allow for the printing of line widths of less than 100 µm depending on the rhelogical properties of the paste [66]. Additionally the pitch of SHJ-IBC solar cells is in the range of 1 mm [29] and a small overlap between the doped region is of no consequence for the device functionality [252]. Therefore printing of the IBC structure could be a viable option for the production of IBC solar cells and probably also for SHJ-IBC solar cells. For printing of SHJ-IBC solar cells it would be necessary to form a passivation layer either by PECVD, or from a liquid precursor [35]. The minority and majority carrier contact would then be printed using liquid silicon precursors for a-Si:H [31,61,111–113], or metal oxide containing solutions [253, 254]. Liquid silicon precursors like cyclopentasilane [31,61], neopentasilane [111,112], or trisilane [113] are available. A processable liquid is formed by dissolving these in organic solvents like toluene [31], or cyclooctane [114] and forming the silicon based polymer by ring-opening, or bond breaking using photo induced polymerization by UV light [31], or thermal polymerization [112]. Additionally other constituents may be needed for production of a printable liquid silicon with suitable rheological properties. It was demonstrated in chapter 6 that intrinsic a-Si:H passivation layers enabling lifetimes of about 1.5 milliseconds [35] can be prepared following this approach. In a consecutive step, doped layers would be printed on the passivation layer by adding white phosphorus [61], or decaborane [62] to the solution prior to the polymerization. Unfortunately the pyrolithic conversion of these doped layers will lead to hydrogen effusion from the whole stack [114]. Therefore it is necessary to rehydrogenate the a-Si:H layers. Hydrogen plasma processes originally used for the hydrogenation of sputtered amorphous silicon [51] could be applied to rehydrogenate liquid processed a-Si:H thin films [35, 109], as described in chapters 4 and 6. The device would then be finished by printing non-overlapping transparent conductive oxide and metal contacts. Alternatively suitable metal oxides could be printed in the shape of interdigitated con-
96CHAPTER 8. PROSPECTS FOR SILICON HETEROJUNCTION SOLAR CELLS
Figure 8.7: VOC and jSC of solar cells with c-Si [10, 11], InP [255], Cu(In,Ga)Se2 [256], CdTe [257], Perovskite [258], GaInP [259] and a-Si:H [260] absorbers plotted vs. the band gap of the absorber material. tacts on the amorphous silicon passivation layer to form the charge selective contacts, since a number of processes for the solution based fabrication of transparent conductive metal oxides exists [253, 254]. Screen printing of the metal contacts is already the industry standard and would only have to be adjusted to different contact geometries. Overall all necessary processes for the fabrication of printable SHJ solar cells are available, but have not been combined yet. This prospect is getting more exciting, since first silicon heterojunction solar cells with metal oxide hole contact materials like MoOx ,WOx [77, 78, 80, 239] and electron contacts using TiOx [246] have been demonstrated.
8.4 8.4.1
Crystalline silicon based tandem solar cells General concept
The maximum efficiency of any semiconductor solar cell is, among other things, limited by the band gap of its absorber [92]. All photons with an energy lower than the band gap are not absorbed and the band gap also limits the achievable Fermi-level splitting. Solar cells with a high band gap are limited by a low jSC , whereas solar cells with a low band gap are limited by low VOC , since the excess energy of high energy photons is lost to thermalization. To illustrate this Figure 8.7 shows exemplary VOC and jSC values of solar cell materials with different band gaps. A common approach to circumvent
8.4. CRYSTALLINE SILICON BASED TANDEM SOLAR CELLS
97
thermalization and transparency losses is to combine two materials with different band gaps into one device [261, 262]. These devices could be connected in series, or in parallel. For devices in series it is important to optimize the absorption in both layers towards similar current densities in both absorbers. This way the bottom cell looses a lot of current and some voltage, due to lower injection densities, but the higher voltage of the top cell overcompensates for these losses. The optimal tandem solar cell combines two materials with band gaps of 0.94 eV and 1.6 eV and could potentially achieve 46 % power conversion efficiency, whereas the best possible single junction solar cell would have a band gap of 1.34 eV and could achieve 33.6 % efficiency [263]. This estimation is based on the detailed balance theory [92] and does not account for intrinsic, or extrinsic recombination, resistive effects and optical losses and is therefore higher than actual efficiencies. Considering the values presented in Figure 8.7 perovskites, or GaInP solar cells may be suitable top cells for c-Si, or Cu(In,Ga)Se2 bottom cells. Furthermore among the c-Si based devices SHJ solar cells have the highest voltage and are therefore ideal for the application in a tandem device.
8.4.2
Tandems with perovskites
The name perovskite relates to the crystal structure of CaTiO3 but is commonly used as collective name for materials with its crystal structure such as CH3 NH3 PbI3 and related materials. CH3 NH3 PbI3 is a organic-inorganic hybrid material [264], which is used as absorber material in a class of polycrystalline thin film solar cells. It has generated some interest, since the efficiencies rose to the range of 20 % within a few years and because its band gap is in the right range for application as a top cell on c-Si, or Cu(In,Ga)Se2 bottom cells and can be tuned by exchanging the halide [265], or the cation [266]. Additionally CH3 NH3 PbI3 features a sharp absorption edge at its band gap energy [267], leading to low sub band gap absorption. Furthermore the fabrication of a perovskite solar cell is done at temperatures not harmful to either SHJ, or Cu(In,Ga)Se2 bottom cells [32]. These properties make it an interesting candidate for the top cell on a SHJ bottom cell. A number of simulation studies have investigated this tandem cell concept [272, 273]. Additionally very promising first four-terminal tandems with perovskite top cells and 18.2 % on SHJ [270], homojunction silicon solar cells with 17 % [268] and 20.5 % efficiency on Cu(In,Ga)Se2 cells [268, 269] as well as two-terminal tandems with 21.2 % on SHJ bottom cells [32, 271] were prepared. These values are summarized in Figure 8.8a and among those is the first monolithic tandem of a silicon heterojunction and a perovskite top cell with an efficiency of 19.9 %, which was fabricated using SHJ bottom cells fabricated as part of the work conducted in this
98CHAPTER 8. PROSPECTS FOR SILICON HETEROJUNCTION SOLAR CELLS
Figure 8.8: (a) Power conversion efficiency of tandem solar cells with perovskite top cells plotted vs. their publication date. The graph includes four-terminal tandems on Cu(In,Ga)Se2 fabricated at Stanford Univeristy [268] and EMPA [269] (green squares), as well as four-terminal tandems on c-Si by Standford [268] and EPFL [270] (red squares) and monolithic tandems on SHJ bottom cells by EPFL [271] and HZB [32] (red circles). (b) j(U)-characteristics of a monolithic tandem solar cell of a SHJ solar cell and a perovskite top cell, as well as the j(U)-curves of both sub cells. thesis [32]. The j(U)-characteristics of the individual sub-cells and the final tandem device are shown in Figure 8.8b. These first results on tandem solar cells with perovskite top cells and SHJ bottom cells are promising and offer much space for improvement and optimization. However they also face two major problems: The first one is stability. The choice of suitable hole conducting materials is limited [274] and the most common choice spiro-OMeTAD is prone to damage during follow-up processes and long-term degradation [274]. Additionally the perovskite itself tends to degrade by dissociating into its constituents and loses its photovoltaic properties, if exposed to humidity and air [274–276]. This process is intensified by light irradiation [274, 275]. The quest to fully understand and prevent the degradation processes in these layers, or to replace them by stable alternatives is the most important prerequisite to fabricate applicable tandem devices with perovskite top cells and perovskite solar cells in general [274]. Promising improvements of the perovskite stability were enabled by replacing more than 20 % of the iodine with bromine [277], which may be beneficial for tandem applications since Br addition increases the band gap [266] and the stability. Additionally it was found that adding Cs to replace some formamidinium leads to further improved stability [278]. The second equally critical question revolves around the lead ion in CH3 NH3 PbI3 . Distribution of lead containing electronic devices is forbidden in most countries. For
8.4. CRYSTALLINE SILICON BASED TANDEM SOLAR CELLS
99
CH3 NH3 PbI3 the specific problem is that PbI3 is washed out during water driven CH3 NH3 PbI3 decomposition [274]. Therefore it is advisable to find another perovskite material without lead, but with similar photovoltaic properties. All in all SHJ and perovskite tandem devices are an interesting prospect. Their main advantage is the easy application of perovskite layers at moderate temperatures, which already lead to promising results. The critical problems are the stability and the toxicity of the perovskite solar cell. If these problems can be resolved this technology may prove its worth in the future.
8.4.3
Tandems with InGaP and related materials
The group of (Ga,In,Al)(As,P,N) based semiconductors is established for the use of optoelectronic devices like lasers, light emitting devices and multi-junction solar cells [262] with power conversion efficiencies above 40 %. Many studies have worked on replacing the established but very expensive GaInP/GaInAs/Ge triple junctions with tandem, or triple junction solar cells grown on silicon wafers. The growth of III-V materials on silicon is problematic due to different lattice constants and thermal expansion coefficients [279]. The most promising material for direct growth on silicon is GaP, since it is almost lattice matched to silicon (100) surfaces [228]. The best monolithic triple junction solar cell on silicon uses direct wafer bonding to create a 30 % efficienct GaInP/GaAs/Si triple junction [280]. Unfortunately neither direct wafer bonding, nor hetero-epitaxial growth of GaP are compatible with SHJ solar cells, due to the involved temperatures and processing conditions. Another approach would be a four terminal tandem using e.g. a GaInP top cell and a silicon bottom cell. This approach has already yielded 27 % power conversion efficiency [281]. All in all the combination of silicon bottom cells with III-V top cells is a promising approach, that benefits from established technology for both subcells, but suffers from the high cost of (Ga,In,Al)(As,P,N) growth.
8.4.4
Summary of crystalline silicon based tandem concepts
Comparing the two most promising tandem concepts, it is obvious that tandems of silicon and GaAs based materials offer established technology and stable devices at a higher cost, whereas perovskites promise cheaper production and easier processes, but suffer from unresolved degradation and toxicity issues. Potential industrial production of tandems of silicon and (Ga,In)P depends strongly on the cost of (Ga,In)P growth and the more complex module structure of four-terminal tandems. It is an open question,
100CHAPTER 8. PROSPECTS FOR SILICON HETEROJUNCTION SOLAR CELLS if cost reduction will enable this technology to compete on the photovoltaic market. In contrast, tandems of silicon and perovskites offer easier processes and monolithic tandems are comparatively easy to implement. The problem of perovskites are their unresolved stability issues and the use of lead containing absorbers. It may actually be necessary to replace CH3 NH3 PbI3 with a material of similar structure, that does include neither lead, nor organic constituents.
8.5
Conclusion
Overall the silicon heterojunction technology is a mature technology. Still, the technology offers some exciting prospects for research. The redesign of the band alignment at the carrier selective contacts of the silicon heterojunction solar cell could overcome limitations in carrier transport and may offer new opportunities for alternative device structures. The combination of the silicon heterojunction solar cell with the backcontact back-junction approach enabled great efficiencies, but has not entered industrial production. Additionally, this concept may benefit from new contact materials. Finally, the silicon heterojunction solar cell is a potential bottom cell for the combination with high band gap top cells. The combination with perovskite, or (In,Ga)P top cells could lead to higher efficiencies, than demonstrated with SHJ single junction solar cells. However some technological and fundamental problems remain and many interesting prospects lie ahead, which offers opportunities for further research.
Chapter 9 Conclusions and outlook The passivation of the amorphous/crystalline silicon heterojunction (SHJ) and the hole transport across the same are the dominating topics in this thesis. The different chapters focus on: The influence of hydrogen on SHJ passivation and an alternative passivation process for surfaces with higher surface free energy (chapter 4), the application of anti-reflection nanostructures and the current generation in these structures (chapter 5), the application of a novel liquid precursor for SHJ passivation, the investigation of its conversion process and the differences between the resulting layer and PECV deposited a-Si:H layers (chapter 6), the transport across the SHJ and its dependence on the valence band offset (chapter 7) and the future research topics for silicon heterojunction solar cells (chapter 8). Chapter 4 comprised an investigation of a two step process for the fabrication of (i)a-Si:H passivation layers on non-ideal c-Si surfaces. The more common approach of directly depositing the best possible passivation layer in a single step is sensitive to surface properties, such as the higher surface free energy of the (100) surface, and results in higher defect densities at Si-(100), or crystallographically undefined surfaces, than on Si-(111). Using a two step approach allows to deposit structurally worse aSi:H layers first. These layers are less likely to form epitaxial regions on surfaces with higher free surface energies. In a second step the layer is then exposed to a hydrogen plasma treatment, which allows to introduce additional hydrogen into the layer and reduce the structural disorder. The two step approach allows to form an epitaxy free and hydrogen rich interface at the SHJ. Chapter 5 describes the application of the aforementioned method to passivate nanotextured silicon surfaces with (i)a-Si:H. The technological question in this project was to enable epitaxy free (i)a-Si:H growth on the crystallographically undefined nanotexture and then apply a hydrogen plasma, which improves the structural quality of the layer, but does not induce crystallization. Additionally the nanotextured solar cells 101
102
CHAPTER 9. CONCLUSIONS AND OUTLOOK
featured a low quantum efficiency at the shorter wavelengths of the visible spectrum. It was deduced from simulations that this low quantum efficiency is based on parasitic absorption in crystalline silicon and a follow-up experiment with nanotextures of varied height enabled to show that this parasitic absorption happens inside the nanotexture. This was explained by the formation of inversion layers in the n-type c-Si next to the p/n-junction. In a planar junction the inversion layers encompasses a certain small volume, but in the needle shaped nanotexture the inversion layer fills large portions, or maybe the whole of the nanotexture and leads to enhanced parasitic absorption. Scattering in the nanotexture and the resulting prolonged light path may even enhance this effect. The work on liquid silicon precursors for SHJ passivation is another application for the hydrogen plasma developed in chapter 4. In contrast to the previous application, the structural quality of the as-deposited a-Si:H layers was acceptable, as shown by photo electron spectroscopy measurements of the valence band. Amorphous silicon layers prepared from liquid precursors feature low hydrogen densities. Also, these layers tend to be macroscopically porous [109]. Both issues can be solved by a HPT. In addition to achieving a well passivated SHJ with a liquid silicon precursor, the conversion of the liquid silicon precursor was investigated and the differences of these layers to PECV deposited material were discussed. Using minority carrier lifetime spectroscopy and XPS the conversion from polysilane to a-Si:H was found to happen at about 350◦ C. Chapter 7 focuses on the charge transfer across the SHJ. A parameter set for a-SiOx :H layers with stoichiometry from a-Si:H to a-SiO2 was developed. These layers feature higher band gaps and therefore higher valence band offsets to the crystalline silicon. Hence they were incorporated as passivation layers at the p/n-junction of SHJ solar cells to investigate the relation between ∆Ev at and the hole transport across the SHJ. The combination of experimental results and device simulation allows to deduce a significant influence of tunnel hopping in the hole transport across the SHJ, especially in the passivation layer. Finally, chapter 8 discusses future topics for research on the silicon heterojunction solar cells, connects them with results from this thesis and discusses results on tungsten oxide hole conduction layers and tandem cells consisting of perovskite top cells and silicon heterojunction bottom cells.
103
Outlook The silicon heterojunction solar cell technology has evolved for more than 20 years, technological maturity is achieved, mass production is set up and most fundamental questions have been answered. However, alternative charge selective contacts may open new routes towards higher efficiency, or simplified processes, combination of SHJ bottom cells with high band gap top cells may lead to strongly increased efficiencies, and back-contact back-junction devices have shown record efficiencies, but further work on an industrially viable process for their production is necessary [29]. In the foreseeable future, research on SHJ solar cells will focus on two topics. Firstly crystalline silicon passivated with amorphous silicon provides a reference system to test alternative charge selective contact materials. Work on using high work function TCOs like MoOx and WOx [73] as the hole contact [77, 78, 80] has just started and could enable to further improve the fill factor in SHJ solar cells. Furthermore, low work function TCOs [73] like TiO [246, 282], or ZnO and alkali metal flourides [239] could provide better electron contacts. Recent results from Panasonic [10, 11] and Patents from Meyer Burger [247] suggest that this problem was already solved on the industrial level, but the scientific investigation is ongoing. Secondly, SHJ solar cells are ideal bottom cells for future tandem solar cells. There are two suitable material classes for the design of appropriate top cells. On the one hand the GaInP/GaAs system, which already enabled more than 30 % efficiency in triple junction configuration with a silicon bottom cell [280], offers stable solar cells and established technology at a high cost. On the other hand tandems of silicon bottom cells and perovskite top cells have already reached 21.2 % efficiency [271] after only minimal research time and potentially offer very simple processes, but suffer from as yet unsolved degradation issues.
Chapter 10 Appendix 10.1
Abbreviations and symbols
AlOx aluminum oxide of stoichiometry x a-Si:H hydrogenated amorphous silicon a-SiOx :H hydrogenated amorphous silicon (sub)oxide with stoichiometry x a-Si:H/c-Si HJ amorphous/crystalline silicon heterojunction AZO aluminum doped zinc oxide B 2 H6 diborane BC-BJ back-contact back-junction b-Si black silicon BSF back-surface field CBM conduction band minimum CFSYS constant final state yield spectroscopy cH atomic hydrogen density c-Si crystalline silicon ∆Ec conduction band offset ∆EF Fermi level splitting dEL distance between top electrode and substrate ∆Ev valence band offset Dit interface defect density e elementary charge 2 imaginary part of the dielectric function E0c Urbach energy of the conduction band tail E0v Urbach energy of the valence band tail Eea Electron affinity EF Fermi level 104
10.1. ABBREVIATIONS AND SYMBOLS EFe EFh Egap Ev Evac EQE FT-IR HOMO HPT HSM (i)a-Si:H IBC IE ITO IQE jph jSC kb λ LSM LUMO µn,p MoOx n n0 NH (n)a-Si:H (n)c-Si φ (p)a-Si:H PCD PECVD PES pFF PH3 QSSPC R
electron quasi Fermi level hole quasi Fermi level band gap valence band edge vacuum energy external quantum efficiency Fourier transform infrared spectroscopy highest occupied molecular orbital hydrogen plasma treatment high-frequency stretching mode intrinsic hydrogenated amorphous silicon interdigitated back contact ionization energy indium-tin-oxide internal quantum efficiency photo current solar cell short circuit current Boltzmann constant wavelenght low-frequency stretching mode last unoccupied molecular orbital mobility of holes (p), or electrons (n) molybdenum oxide charge carrier density intrinsic charge carrier density hydrogen density n-doped hydrogenated amorphous silicon n-doped crystalline silicon work function p-doped hydrogenated amorphous silicon photoconductive decay (measurements) plasma-enhance chemical vapor deposition photoelectron spectroscopy pseudo fill factor phosphine quasi steady state photo conductance decay Recombination
105
106 ρ σ SE SEM sFF SHJ SIMS SiNx SiOx T Tsub τ τeff TCO TMB TRPCD U UPS VBM VOC WF WOx XPS
CHAPTER 10. APPENDIX mass density conductivity spectral ellipsometry scanning electron microscopy simulated fill factor silicon hetero junction secondary ion mass spectroscopy silicon nitride of stoichiometry x silicon oxide of stoichiometry x temperature substrate temperature (minority) charge carrier lifetime effective (minority) charge carrier lifetime transparent conductive oxide trimethylborane transient photo conductance decay voltage ultra-violet photoelectron spectroscopy valence band maximum solar cell open circuit voltage work function tungsten oxide x-ray photoelectron spectroscopy
10.2. PUBLICATIONS
10.2
107
Publications
peer-reviewed articles Articles 2 to 5 make up chapters 7, 6, 5 and 4 of this thesis respectively and results from articles 1 and 11 are contained in chapter 8. Article 7 is related to the work presented in chapter 7 and was conducted beforehand, whereas article 10 is a study following upon the work presented in chapter 4. The work presented in articles 6 and 9 was conducted in the timeframe of this thesis, but the content of this articles is unrelated to this thesis. Article 8 finally contains results predating the time frame of this thesis. 1. Mathias Mews, Lars Korte, Bernd Rech Oxygen vacancies in tungsten oxide and their influence on tungsten oxide/silicon heterojunction solar cells Solar Energy Materials and Solar Cells (2016) 2. Mathias Mews, Martin Liebhaber, Bernd Rech, Lars Korte Valence band alignment and hole transport in amorphous/crystalline silicon heterojunction solar cells Applied Physics Letters 107 (1), 013102 (2015) 3. Mathias Mews, Christoph Mader, Stephan Traut, Tobias Sontheimer, Odo Wunnicke, Lars Korte, Bernd Rech Solution-processed amorphous silicon surface passivation layers Applied Physics Letters 105 (12), 122113 (2014) 4. Mathias Mews, Caspar Leendertz, Michael Algasinger, Svetoslav Koynov, Lars Korte Amorphous/crystalline silicon heterojunction solar cells with black silicon texture Physica Status Solidi Rapid Research Letters 8 (10), 831-835 (2014) 5. Mathias Mews, Tim F. Schulze, Nicola Mingirulli, Lars Korte Hydrogen plasma treatments for passivation of amorphous-crystalline silicon-heterojunctions on surfaces promoting epitaxy Applied Physics Letters 102 (13), 122106 (2013) 6. Johannes Ziegler, Mathias Mews, Kai Kaufmann, Thomas Schneider, Alexander N. Sprafke, Lars Korte, Ralf B. Wehrspohn Plasma-enhanced atomic-layer-deposited MoOx emitters for silicon heterojunction solar cells Applied Physics A 120 (3), 811-816 (2015) 7. Martin Liebhaber, Mathias Mews, Tim F. Schulze, Lars Korte, Bernd Rech, Klaus Lips Valence band offset in heterojunctions between crystalline silicon and amorphous silicon (sub)oxides (a-SiOx :H, 0
