Amorphous and Microcrystalline Silicon Thin-Film Solar Cell

1. Theory

• Definition: Hydrogenated amorphous silicon (a-Si:H) is a disordered silicon network with about 10 atomic % of hydrogen (Wikipedia) and an optical gap of about 1.7 eV, which Intal and Ebong call direct; Fink gives 1.6 to 1.9 eV depending on disorder and hydrogen. Bond lengths and angles vary by about ±10 % (Fink). The bare network has about 10${}^{20}$ cm${}^{-3}$ dangling bonds (0.2 % of the atoms, computed), and hydrogen passivation lowers the density by orders of magnitude.
• Microcrystalline silicon: Hydrogenated microcrystalline silicon (µc-Si:H, called nc-Si in the efficiency tables) is a mixture of crystallites and amorphous tissue with an indirect gap of about 1.1 eV and 2 to 5 % hydrogen. The best absorber lies near the transition to amorphous growth, at a Raman crystallinity of 60 to 70 % (Fink). The Raman peaks are near 520 (crystalline), 510 (defective) and 480 cm${}^{-1}$ (amorphous) (Droz et al.).
• Cell structure: Both materials are used in p-i-n cells, with p and n layers of 20 to 40 nm and undoped absorbers of about 300 nm (a-Si:H) and 1300 nm (µc-Si:H) in the dissertation of Fink. The front contact is textured ZnO:Al, sputtered above 250 °C and etched in dilute HCl for about 80 s, with a sheet resistance below 20 $\Omega$/sq as target (4.4 $\Omega$/sq measured on Gen5 glass) and an average transmission of 83.7 %; the back contact is ZnO/Ag for tandems and ZnO/Al for single junctions (Applied Materials).
• Deposition: Plasma-enhanced chemical vapour deposition from SiH${}_4$ at 30 to 300 °C and 13.56 MHz (Wikipedia), 200 to 300 °C in Intal and Ebong, and 70 to 130 MHz very-high-frequency plasmas for microcrystalline layers (Droz et al.); the growth rate in Fink is about 0.5 nm/s. A triode reactor with a remote electrode and grid lowers the ion flux and allows intrinsic layers above 500 nm without worse degradation (Intal and Ebong). Deposition near 75 °C permits plastic and paper substrates (Wikipedia).
• Records: Green et al. list a-Si 10.2 ± 0.3 % (1.001 cm${}^2$, 896 mV, 16.36 mA/cm${}^2$, 69.8 %), nc-Si 11.9 ± 0.3 % (1.044 cm${}^2$, 550 mV, 29.72 mA/cm${}^2$, 75.0 %), an a-Si/nc-Si tandem 12.7 ± 0.4 % (1.000 cm${}^2$, 1.342 V, 13.45 mA/cm${}^2$, 70.2 %) and an a-Si/nc-Si/nc-Si triple 14.0 ± 0.4 % (1.045 cm${}^2$, 1.922 V, 9.94 mA/cm${}^2$, 73.4 %), all measured at AIST between 2014 and 2017, the tandem and triple being stabilised by 1000 h at 1 sun and 50 °C. Modules: TEL Solar a-Si/nc-Si 12.3 ± 0.3 % (14322 cm${}^2$ total area, 280.1 V, 0.902 A, 69.9 %) and a CSG minimodule of 10.5 ± 0.3 % (94.0 cm${}^2$, 0.492 V per cell, 29.7 mA/cm${}^2$, 72.1 %).
• Micromorph tandem: The a-Si:H top cell over a µc-Si:H bottom cell is called micromorph; Fischer et al. (1996, from the DOI) report a µc-Si:H cell of 3.6 µm with more than 25 mA/cm${}^2$ and 7.7 % stable, and a tandem of 13 % initial and 10 % degraded.
• Market: Intal and Ebong give a 2024 market of 2.26 billion dollars and state that Xunlight, Masdar PV and OptiSolar left the field after 2015.

2. Working Principle

• Collection by drift: The diffusion length in the undoped network is below 0.5 µm and 1 to 2 µm in passivated intrinsic layers (Intal and Ebong), so the doped layers set up a field across the undoped layer and the carriers are swept out by drift. The collection length $\mu \tau V_{bi}/d$ limits the thickness to $\sqrt{\mu \tau V_{bi}}$, which is 0.3 to 1.0 µm for the printed range of $\mu \tau$ (computed, built-in voltage of 1 V assumed) and fits the absorber of 300 nm of Fink.
• Light-induced degradation: Staebler and Wronski found in 1977 that the conductivity of a-Si:H falls under light. Electron-hole recombination breaks weak bonds and the number of dangling bonds rises; the hydrogen-collision model (Branz, 1999) lets two recombination events free mobile hydrogen that forms two dangling bonds and a metastable pair elsewhere, and the atomistic simulation of Abtew and Drabold forms such pairs only in the light-excited runs. Most of the loss is in the fill factor (from above 0.7 to about 0.6), heating above 150 °C reverses it, and nanocrystalline silicon degrades less (Wikipedia). Thin layers in multijunction cells, nanocrystalline seed layers and tuned hydrogen dilution reduce it (Wikipedia; Intal and Ebong).
• Light trapping: The textured ZnO:Al has a haze rising from nearly 0 to above 80 % with etch time and gains most of its quantum efficiency above 700 nm (Applied Materials). The ideal crater depth and width are about the wavelength, a randomising texture can lengthen the path by $4n^2\approx 50$, and craters with opening angles below 115 ± 10° lower the Raman crystallinity (Fink). The Raman intensity at 785 nm follows the external quantum efficiency (Ledinsky et al., abstract). A ZnO layer between the cells reflects the 550 to 700 nm band back into the a-Si:H cell, with a theoretical gain of 1.3 points (SPIE).
• Voltage and current: The a-Si cell has 896 mV, 64 % of the radiative limit at 1.7 eV, and the nc-Si cell 550 mV, 65 % at 1.1 eV (computed); their currents are 73 and 67 % of the limits (computed). Droz et al. note a maximum of about 900 mV for a-Si cells and correlate the voltage of µc-Si cells with the Raman crystallinity (R = 0.89).
• Tandem: The voltages of the stacked cells add and the current is set by the weaker cell. An ideal split of the spectrum between 1.75 and 1.12 eV gives about 21 and 23 mA/cm${}^2$ (computed), against 13.45 mA/cm${}^2$ for the certified tandem.

3. Applications

• Large-area glass modules: TEL Solar a-Si/nc-Si modules of 12.3 % and the Gen5 line of Applied Materials with 156 W (aperture efficiency 11.5 %, 2010) (Green et al.; Applied Materials).
• Lightweight and building-integrated: Kaneka a-Si:H/µc-Si:H modules of 13 to 14 % below 4 kg/m${}^2$ for building integration and truck roofs, and roll-to-roll lines of Xunlight (25 MW) and Anwell (40 MW) (Intal and Ebong; Wikipedia).
• Detectors and contacts: a-Si:H is used for detectors and optoelectronic devices and as the passivating layer of heterojunction solar cells (Stuckelberger et al.); photodiode arrays for imaging have pixels below 1 mm (Wikipedia). The heterojunction use is treated on the HIT page.
• Tandem partners: Intal and Ebong name hybrid a-Si:H/organic and a-Si:H/perovskite tandems and a stabilized target of 18 %.

Absorption, Gap and Defects

Name	Formula	Related Formulae	Variables/Constants
Absorption edge and optical gap	${\left(\alpha h\nu \right)}^{1/2}=B\left(h\nu -E_g\right)$	$\alpha \left(h\nu \right)={\alpha }_0\exp \left(\frac{h\nu }{E_U}\right)$	$\alpha :absorptioncoefficient$ $h\nu :photonenergy$ $B:edgeconstant$ $E_g:optical(Tauc)gap$ $E_U:Urbachenergyofthebandtail$ $\begin{array}{r}{\alpha }_0:prefactor,bothformsarebackground\\ knowledgeandnoTaucfitorUrbachenergy\\ wasfoundinthesources,gaps:1.7eV\\ (IntalandEbong),1.6to1.9eVdepending\\ ondisorderandhydrogen(Fink),1.85to\\ 1.90eVforthetopcellofatriple\\ junction(IntalandEbong),andfor\mu \\ c\text{-}Si:Habout1.1eV(Fink)or1.10to1.30\\ eV(IntalandEbong)\end{array}$ $\begin{array}{r}(computed:1.7eVcorrespondstoa\\ wavelengthof729nmand1.1eVto1127\\ nm,sotheamorphouscellcannotusethe\\ near-infraredthatthemicrocrystalline\\ cellabsorbs)\end{array}$
Single-pass absorption and light-trapping limit	$A=1-e^{-\alpha d}$	$\Delta L_{max}\approx 4n^2$ $d_{inf}\approx \frac{1}{2\alpha }$	$A:absorbedfraction$ $d:layerthickness$ $\begin{array}{r}\Delta L_{max}:path-lengthenhancementofa\\ randomisingtexture(Yablonovitchlimit)\end{array}$ $n:refractiveindex$ $\begin{array}{r}{d}_{inf}:informationdepthoftheRaman\\ signal(1/edepthoftheback-scattered\\ signal,whichdecaysas{e}^{-2\alpha z}),IntalandEbong\\ give\alpha >10^{5}c{m}^{-1}\\ acrossmostofthevisiblerangeandmore\\ than90\%absorptionin300to500nm,\\ Finkuses4n^2\approx 50forn\approx 3.6at800nmand\\ d_{inf}=56and125nmat532nmand625and\\ 833nmat647nmfora-Si:Hand\mu c-Si:H\end{array}$ $\begin{array}{r}(computed:\alpha =10^{5}c{m}^{-1}gives95.0\%at\\ 0.3\mu \mathrm{m}and99.3\%at0.5\mu \mathrm{m};4n^2=51.8\\ forn=3.6;\alpha =10^{4}c{m}^{-1}(assumed)absorbs\\ 73\%inonepassthrough1.3\mu \mathrm{m};with\\ d_{inf}=1/(2\alpha )Fin{k}^{\prime }sprinted8\times 10^{3}and\\ 6\times 10^3\\ c{m}^{-1}giveexactlyhis625and833nmat647\\ nm,andhis56and125nmat532nmimply\\ \alpha =8.9\times 10^{4}and4.0\times 10^{4}c{m}^{-1};theearlier\\ form0.3/\alpha didnotreproducehisnumbers,\\ auditcorrection)\end{array}$
Dangling-bond fraction	$f_d=\frac{N_d}{N_{Si}}$	$N_d\left(\mathrm{bare}\right)\approx 10^{20}\mathrm{c}{\mathrm{m}}^{-3},N_{Si}\approx 4.9\times 10^{22}\mathrm{c}{\mathrm{m}}^{-3}$	$\begin{array}{r}{f}_d:fractionofsiliconatomswitha\\ danglingbond\end{array}$ $N_d:dangling-bonddensity$ $\begin{array}{r}{N}_{Si}:atomicdensity,Finkgives10^{20}c{m}^{-3}\\ (0.2\%oftheatoms)and10^{16}to10^{18}c{m}^{-3}\\ afterpassivation,IntalandEbonggive\\ 10^{15}to10^{16}c{m}^{-3}afterpassivationat200to\\ 300°C\end{array}$ $\begin{array}{r}(computed:10^{20}/4.9\times 10^{22}=0.204\%;10^{15},\\ 10^{16}and10^{18}c{m}^{-3}correspondto2.0\times 10^{-8},\\ 2.0\times 10^{-7}and2.0\times 10^{-5}oftheatoms)\end{array}$
Raman crystallinity	$X_c=\frac{I_c}{I_c+y{I}_a}$	$I_c\left(520\mathrm{c}{\mathrm{m}}^{-1}\right),I_{dc}\left(510\mathrm{c}{\mathrm{m}}^{-1}\right),I_a\left(480\mathrm{c}{\mathrm{m}}^{-1}\right)$	$X_c:Ramancrystallinity$ $\begin{array}{r}{I}_c,I_a:integratedintensitiesofthe\\ crystallineandamorphouspeaks\end{array}$ $I_{dc}:defectivecrystallinepeak$ $\begin{array}{r}y:ratiooftheRamancross-sections,Droz\\ etal.sety=1(literaturevaluesfrom\\ 0.88downto0.1),peakwidthsabout10,\\ 26and53c{m}^{-1},Finkgivesthebest\\ microcrystallinematerialat{X}_c\approx 0.6to\\ 0.7,andDrozetal.correlate{V}_{oc}withthe\\ crystallinitywithR=0.89,thesignof\\ thetrend(alowervoltageforahigher\\ crystallinity)isbackgroundknowledge\end{array}$ $\begin{array}{r}(computed:fory=1and{X}_c=0.65the\\ intensityratio{I}_c/I_{a}is1.86;theexample\\ cellofDrozetal.has{X}_c=0.28to0.52\\ forthefourspectra)\end{array}$

Junction, Degradation and Tandem

Name	Formula	Related Formulae	Variables/Constants
Drift collection length and thickness limit	$l_c=\mu \tau F=\mu \tau \frac{V_{bi}}{d}$	$d_{max}=\sqrt{\mu \tau V_{bi}}$	$l_c:collectionlength$ $\mu \tau :mobility-lifetimeproduct$ $F:fieldintheundopedlayer$ $V_{bi}:built-involtage$ $\begin{array}{r}d:layerthickness,collectionneeds{l}_c>d,\\ IntalandEbonggive\mu \tau =10^{-9}to10^{-8}c{m}^2/V\\ inthesentenceonthebarenetworkand\\ diffusionlengthsof1to2\mu \mathrm{m}for\\ passivatedintrinsiclayers,Finkuses300\\ nmfora-Si:H,theuseofthis\mu \tau fora\\ devicelayerandtheuniformfieldaremy\\ assumptions\end{array}$ $\begin{array}{r}(computedwith{V}_{bi}=1V(assumed):d_{max}=\\ 0.32\mu \mathrm{m}for10^{-9}c{m}^2/Vand1.0\mu \mathrm{m}for10^{-8}\\ c{m}^2/V;alayerof0.3\mu \mathrm{m}hasF=33kV/cm\\ and{l}_c=0.33and3.3\mu \mathrm{m},alayerof1\mu \mathrm{m}\\ hasF=10kV/cmand{l}_c=0.1and1.0\mu \mathrm{m})\end{array}$
Light-induced degradation	$D=1-\frac{{\eta }_{stab}}{{\eta }_{init}}$	${\eta }_{stab}={\eta }_{init}\left(1-D\right)$	$D:relativeloss$ $\begin{array}{r}{\eta }_{init},{\eta }_{stab}:efficiencybeforeandafter\\ lightsoaking,Wikipediagives10to30\%\\ insixmonthsforcellsand10to15\%\\ formodules,AbtewandDraboldandIntal\\ andEbonggive15to20\%,Fischeretal.\\ giveamicromorphcellof13\%initial\\ and10\%degraded,thefillfactorfalls\\ fromabove0.7toabout0.6(Wikipedia)\end{array}$ $\begin{array}{r}(computed:13\to 10\%givesD=23.1\%;\\ Greenetal.markthetandemandtriple\\ entriesasstabilisedby1000hat1sun\\ and50°C,so12.7and14.0\%arevalues\\ afterlightsoaking(footnotelofa\\ tool-readsummary);afillfactorof\\ 0.7\to 0.6aloneisalossof14.3\%)\end{array}$
Detailed-balance limit and voltage ratio	$r_V=\frac{V_{oc}}{V_{oc,SQ}}$	$\Delta V=\frac{E_g}{q}-V_{oc},\frac{J_{sc}}{J_{sc,max}}$	$r_V:fractionoftheradiativevoltage$ $V_{oc,SQ}:voltageatthedetailed-balancelimit$ $\Delta V:deficit$ $\begin{array}{r}{J}_{sc,max}:photocurrentabovethegapunder\\ AM1.5G,themethodistheoneoftheSolar\\ CellFundamentalspage,thecellsarethe\\ a-Siandnc-SientriesofGreenetal.and\\ theirgapsof1.7and1.1eVareassumed\\ fromthesources\end{array}$ $\begin{array}{r}(computedat298.15K,bothfaces:1.70eV\\ gives{V}_{oc,SQ}=1.402V,J_{sc,max}=22.44mA/c{m}^2\\ and28.6\%;1.10eVgives0.842V,44.23\\ and32.2\%;a-Si(896mV,16.36)has{r}_V=\\ 0.639,\Delta V=0.804Vand{J}_{sc}/J_{sc,max}=0.729;\\ nc-Si(550mV,29.72)has{r}_V=0.654,\Delta V\\ =0.550Vand0.672;at1.75and1.85eV\\ thea-Sideficitis0.854and0.954V)\end{array}$
Series tandem: voltage, current and spectral split	$V_{oc}={\sum }_i{V}_{oc,i},J_{sc}={\min }_i{J}_{sc,i}$	$J_{top}=q{\int }_{E_t}^{\infty }\Phi dE,J_{bot}=q{\int }_{E_b}^{E_t}\Phi dE$	$V_{oc,i},J_{sc,i}:voltageandcurrentofjunctioni$ $\Phi :AM1.5Gphotonfluxperenergy$ $\begin{array}{r}{E}_t,E_b:gapsofthetopandbottomcell,\\ thesplitistheidealonewithout\\ reflection,parasiticabsorptionor\\ transmissionofthetopcell,thegapsare\\ assumed,IntalandEbonglistsub-cell\\ currentsof4to5,6to7and10to12\\ mA/c{m}^{2}foratriple,theSPIEarticle\\ gives10.7and12.5mA/c{m}^{2}foratandem\end{array}$ $\begin{array}{r}(computed:1.75/1.12eVgives21.02and\\ 22.79mA/c{m}^2,1.70/1.10eVgives22.44and\\ 21.79;1.85/1.6/1.1eVgives18.25,7.15\\ and18.83;themeasuredtandemcurrentof\\ 13.45mA/c{m}^{2}is64\%ofthesmallerideal\\ currentandthetriplecurrentof9.94\\ exceedsthe7.15ofa1.6eVmiddlecell;\\ thevoltageofthetandem1.342Vis92.8\\ \%of0.896+0.550Vandthetriple1.922V\\ is96.3\%of0.896+2\times 0.550V,butthese\\ aredifferentdevices)\end{array}$

Efficiency Checks

Name	Formula	Related Formulae	Variables/Constants
Efficiency from the reported parameters	$\eta =\frac{V_{oc}{J}_{sc}FF}{P_{in}},P_{in}=100mW/c{\mathrm{m}}^2$	${\eta }_{check}=\frac{V_{oc}{J}_{sc}FF}{1000}$ $\Delta \eta ={\eta }_{printed}-{\eta }_{check}$	$\begin{array}{r}\eta :efficiency,V_{oc}inmV,J_{sc}inmA/c{m}^2,FF\\ asafraction({\eta }_{check}isinpercent:the\\ 1000turnsmVintoV,and1mW/c{m}^2\\ is1percentof{P}_{in}),theentriesarefromthe\\ efficiencytables(version64)ofGreenet\\ al.andfromtheSPIEarticle\end{array}$ $\begin{array}{r}(computed:a-Si896,16.36,69.8\%give\\ 10.23\%(10.2);nc-Si550,29.72,75.0\%\\ give12.26\%against11.9printed,a\\ differenceof0.36;thevaluewouldbe\\ restoredbyFF=72.8\%or{V}_{oc}=534mV;\\ tandem1342,13.45,70.2\%give12.67\%\\ (12.7);triple1922,9.94,73.4\%give\\ 14.02\%(14.0);CSG492,29.7,72.1\%\\ give10.54\%(10.5);SPIEtandem1.41V,\\ 12.1,0.65gives11.09\%(11.1)andwith\\ 10.7mA/c{m}^{2}9.81\%)\end{array}$
Module efficiency from voltage, current and area	${\eta }_{mod}=\frac{V_{oc}{I}_{sc}FF}{GA}$	$P_{max}=V_{oc}{I}_{sc}FF$	${\eta }_{mod}:moduleefficiency$ $V_{oc},I_{sc}:modulevoltageandcurrent$ $FF:fillfactor$ $G:irradianceof1000W/m^2$ $\begin{array}{r}A:area,theTELSolarentryisfromGreen\\ etal.(totalarea),theGen5moduleof\\ 156Wand11.5\%apertureefficiencyis\\ fromtheAppliedMaterialsarticle(which\\ printsasubstrateof1\times 1.3m^2)\end{array}$ $\begin{array}{r}(computed:280.1V,0.902A,69.9\%on\\ 14322c{m}^{2}give176.6Wand12.33\%\\ (12.3\pm 0.3);156Won1.3m^{2}gives12.0\%,\\ whichisabovetheaperturevalueof11.5\\ \%andsonotpossibleforanaperture\\ smallerthanthesubstrate;156Won1.43\\ m^2(1.1\times 1.3m^2,assumed)gives10.9\%,\\ and11.5\%needsanapertureareaof\\ 1.357m^2)\end{array}$

Important Questions

Ques: Is the 14.0 % triple junction all-amorphous or a-Si/nc-Si/nc-Si?

Intal and Ebong call 14.0 % the highest independently confirmed all-a-Si:H triple-junction efficiency ‘as of 2025’, reached from 13.6 % in 2015 by triode PECVD on the middle and bottom cells, but they cite a 2016 paper for it. Green et al. list 14.0 ± 0.4 % for an a-Si/nc-Si/nc-Si cell measured at AIST in May 2016 (1.922 V, 9.94 $mA/c{\mathrm{m}}^2$, 73.4 %, 1.045 $\mathrm{c}{\mathrm{m}}^2$), with a product of 14.02 % (computed). The voltage supports the table: after the 0.896 V of the a-Si cell, 1.026 V remains for two junctions, or 0.513 V each (computed), close to the 0.550 V of the nc-Si cell, whereas three amorphous junctions would give about 2.7 V (0.9 V each, assumed). The cell is more likely a-Si/nc-Si/nc-Si and the year is 2016, not 2025. Reference: Green, Dunlop, Yoshita, Kopidakis, Bothe, Siefer, Hinken, Rauer, Hohl-Ebinger and Hao, ‘Solar cell efficiency tables (Version 64)’, Prog. Photovolt. 32(7), 425-441 (2024); Intal and Ebong, ‘Thin-Film Solar Photovoltaics: Trends and Future Directions’, arXiv:2508.05589 (2025).

Ques: Does light soaking cost 10 to 30 %, 15 to 20 % or 23 %?

Wikipedia gives 10 to 30 % in the first six months for cells and 10 to 15 % for commercial modules, Abtew and Drabold quote 15 to 20 %, and Intal and Ebong give 15 to 20 % within the first 1000 h. Fischer et al. report a micromorph tandem falling from 13 % to 10 %, which is 23 % (computed). The ranges overlap near 15 to 20 %. Wikipedia says that most of the loss is in the fill factor, from above 0.7 to about 0.6, which alone is 14.3 % (computed), and that thin layers in multijunction cells lose less. The 23 % belongs to an early tandem (the year 1996 is taken from the DOI), and 10 to 30 % covers many designs. For a normal single-junction cell 15 to 20 % is the more likely loss. Reference: Wikipedia, ‘Staebler-Wronski effect’; Abtew and Drabold, ‘Atomistic simulation of light-induced changes in hydrogenated amorphous silicon’, arXiv:cond-mat/0510818 (year not stated, about 2005 to 2006); Intal and Ebong, ‘Thin-Film Solar Photovoltaics: Trends and Future Directions’, arXiv:2508.05589 (2025); Fischer et al., ‘The micromorph solar cell: extending a-Si:H technology towards thin film crystalline silicon’ (IEEE PVSC conference report, year taken as 1996 from the DOI).

Ques: Is the defect density after passivation 10^15 to 10^16 or 10^16 to 10^18 cm^-3?

Intal and Ebong give 10${}^{15}$ to 10${}^{16}$ $\mathrm{c}{\mathrm{m}}^{-3}$ after hydrogen passivation at 200 to 300 °C, whereas Fink gives 10${}^{16}$ to 10${}^{18}$ $\mathrm{c}{\mathrm{m}}^{-3}$, starting from 10${}^{20}$ $\mathrm{c}{\mathrm{m}}^{-3}$ for the bare network, which is 0.204 % of the atoms (computed). Wikipedia says only ‘several orders of magnitude’. The ranges touch at 10${}^{16}$ $\mathrm{c}{\mathrm{m}}^{-3}$. Fink describes passivated material in general (my reading), while Intal and Ebong write about device-grade intrinsic layers that reach diffusion lengths of 1 to 2 µm, so the two statements need not conflict. The upper range of Fink corresponds to 2.0 × 10${}^{-5}$ of the atoms (computed). For layers used in record cells the lower range is more likely, but neither source names the measurement method, so the values cannot be compared directly. Reference: Intal and Ebong, ‘Thin-Film Solar Photovoltaics: Trends and Future Directions’, arXiv:2508.05589 (2025); Fink, ‘Betrachtung der Kristallinitaetsentwicklung in mikrokristallinem Duennschicht-Silizium mit in-situ Raman-Spektroskopie’, dissertation, RWTH Aachen (2017), Schriften des Forschungszentrums Juelich, Energie & Umwelt 405 (2018) (German; excerpt to Sec. 2.3.1); Wikipedia, ‘Amorphous silicon’.

Ques: Is the gap of a-Si:H 1.7, 1.6 to 1.9 or 1.85 to 1.90 eV?

Intal and Ebong give a direct optical gap of 1.7 eV for a-Si:H, Fink gives 1.6 to 1.9 eV depending on structural disorder and hydrogen content, and the triple-junction table of Intal and Ebong uses 1.85 to 1.90 eV for the top cell. For the bottom cell Fink gives about 1.1 eV for µc-Si:H and the table 1.10 to 1.30 eV. The values describe different films, because more hydrogen widens the gap, so they do not conflict. The choice changes the voltage analysis: the a-Si cell of 0.896 V has $\Delta V$ = 0.804, 0.854 and 0.954 V at 1.70, 1.75 and 1.85 eV, and $r_V$ = 0.639 at 1.70 eV (computed). For a single-junction record cell 1.7 eV is the more likely gap. The sources do not say whether the quoted gaps are Tauc or other definitions. Reference: Intal and Ebong, ‘Thin-Film Solar Photovoltaics: Trends and Future Directions’, arXiv:2508.05589 (2025); Fink, ‘Betrachtung der Kristallinitaetsentwicklung in mikrokristallinem Duennschicht-Silizium mit in-situ Raman-Spektroskopie’, dissertation, RWTH Aachen (2017), Schriften des Forschungszentrums Juelich, Energie & Umwelt 405 (2018) (German; excerpt to Sec. 2.3.1); Green, Dunlop, Yoshita, Kopidakis, Bothe, Siefer, Hinken, Rauer, Hohl-Ebinger and Hao, ‘Solar cell efficiency tables (Version 64)’, Prog. Photovolt. 32(7), 425-441 (2024).

Ques: What is the efficiency of thin-film silicon modules: 8, 9 to 10, 12.3 or 13 to 14 %?

Intal and Ebong print 9 to 10 % for single-junction a-Si modules (Sec. 2.1), below 9 % (Sec. 4), near 8 % (Sec. 5) and a practical cap near 9 % (introduction), and 13 to 14 % for Kaneka tandem modules. Green et al. list 12.3 ± 0.3 % for a TEL Solar tandem module of 14322 $\mathrm{c}{\mathrm{m}}^2$ total area, which the product confirms as 12.33 % (computed); the 13 to 14 % is not in the extracts of the tables. Applied Materials reported 156 W and 11.5 % aperture efficiency on Gen5 glass. Their printed substrate of 1 × 1.3 m${}^2$ gives 12.0 % (computed), above the aperture value, which is not possible, whereas 1.1 × 1.3 m${}^2$ (assumed) gives 10.9 %. The 8 to 10 % is for single junctions and 12.3 % is the certified tandem value. Reference: Intal and Ebong, ‘Thin-Film Solar Photovoltaics: Trends and Future Directions’, arXiv:2508.05589 (2025); Green, Dunlop, Yoshita, Kopidakis, Bothe, Siefer, Hinken, Rauer, Hohl-Ebinger and Hao, ‘Solar cell efficiency tables (Version 64)’, Prog. Photovolt. 32(7), 425-441 (2024); Obermeyer, Severin, Kress, Klein, Schmidt and Wieder (Applied Materials), ‘Textured ZnO enables high-efficiency silicon solar modules’, Renewable Energy World (26 July 2010).

Ques: How thick is the µc-Si:H absorber: 1.3, 2 to 2.5 or 3.6 µm?

Fink uses absorbers of about 300 nm (a-Si:H) and 1300 nm (µc-Si:H), Droz et al. grow µc-Si:H i-layers of 2.0 to 2.5 µm, and Fischer et al. used 3.6 µm for a 7.7 % stable cell with more than 25 $mA/c{\mathrm{m}}^2$. The nc-Si cell of Green et al. gives 29.72 $mA/c{\mathrm{m}}^2$ on 1.044 $\mathrm{c}{\mathrm{m}}^2$, 67 % of the limit of 44.23 $mA/c{\mathrm{m}}^2$ at 1.1 eV (computed), but its thickness is not printed. A single pass through 1.3 µm absorbs 73 % for $\alpha$ = 10${}^4$ $\mathrm{c}{\mathrm{m}}^{-1}$ (assumed, computed), so a thicker cell or a texture is needed for the infrared. The values belong to different groups and years. For a modern cell 1.3 to 2.5 µm is the more likely range, since 3.6 µm is from the first micromorph paper, and the thickness of the record is unknown. Reference: Fink, ‘Betrachtung der Kristallinitaetsentwicklung in mikrokristallinem Duennschicht-Silizium mit in-situ Raman-Spektroskopie’, dissertation, RWTH Aachen (2017), Schriften des Forschungszentrums Juelich, Energie & Umwelt 405 (2018) (German; excerpt to Sec. 2.3.1); Droz, Vallat-Sauvain, Bailat, Feitknecht, Meier and Shah, ‘Relationship between Raman crystallinity and open-circuit voltage in microcrystalline silicon solar cells’, Sol. Energy Mater. Sol. Cells 81(1), 61-71 (2004); Fischer et al., ‘The micromorph solar cell: extending a-Si:H technology towards thin film crystalline silicon’ (IEEE PVSC conference report, year taken as 1996 from the DOI); Green, Dunlop, Yoshita, Kopidakis, Bothe, Siefer, Hinken, Rauer, Hohl-Ebinger and Hao, ‘Solar cell efficiency tables (Version 64)’, Prog. Photovolt. 32(7), 425-441 (2024).

Ques: Which tandem current is right: 12.1, 10.7 or 12.5 mA/cm^2?

The SPIE article gives a tandem with a scattering front and no intermediate reflector at 12.1 $mA/c{\mathrm{m}}^2$, a fill factor of 0.65 and 11.1 %, and measured junction currents of 10.7 (a-Si:H) and 12.5 $mA/c{\mathrm{m}}^2$ (µc-Si). In a series stack the current cannot exceed that of the lower junction, so 12.1 against 10.7 is a conflict. The top cell has 0.91 V, and I read the second value of 0.5 V, printed as the top cell again, as the bottom cell. The sum of 1.41 V gives 1.41 × 12.1 × 0.65 = 11.09 % (computed), whereas 10.7 $mA/c{\mathrm{m}}^2$ would give 9.81 %. The efficiency therefore fits 12.1 $mA/c{\mathrm{m}}^2$, and the junction values more likely come from another sample or from the quantum-efficiency curves; the article does not say which. The reflector’s theoretical 12.4 % is a gain of 1.3 points. Reference: SPIE Newsroom, ‘Tandem solar cells with integrated intermediate reflectors’ (authors and date not stated on the page).

Ques: Does the nc-Si entry of the efficiency tables add up?

Green et al. list 11.9 ± 0.3 % for the nc-Si cell of AIST on 1.044 $\mathrm{c}{\mathrm{m}}^2$ with 550 mV, 29.72 $mA/c{\mathrm{m}}^2$ and 75.0 %, but the product is 12.26 % (computed), 0.36 points more and outside the stated uncertainty. The other thin-film silicon entries agree within 0.04 points: the a-Si cell gives 10.23 % (10.2), the tandem 12.67 % (12.7), the triple 14.02 % (14.0) and the CSG minimodule 10.54 % (10.5) (computed). A fill factor of 72.8 % or a voltage of 534 mV would restore 11.9 % (computed). Possible causes are a misprint in one of the three numbers, a different area for the current, or rounding, but the extract gives no explanation. The efficiency is the certified quantity and is more likely correct, so the voltage, current and fill factor should be checked in the printed table before use. Reference: Green, Dunlop, Yoshita, Kopidakis, Bothe, Siefer, Hinken, Rauer, Hohl-Ebinger and Hao, ‘Solar cell efficiency tables (Version 64)’, Prog. Photovolt. 32(7), 425-441 (2024).

Ques: Is the 2.26 billion dollar market a-Si or a-Si:H/µc-Si tandem?

Table 3 of Intal and Ebong labels a 2024 market of 2.26 billion dollars as a-Si, while Section 5 attributes the same figure to a-Si:H/µc-Si tandems. The authors also state that CdTe, CIGS and a-Si together peaked at 15 % of the market in 2010 and fell to 5.5 % in 2024, and that Xunlight, Masdar PV and OptiSolar left the field after 2015. With the CIGS market of 2.8 billion dollars at a share of 1.1 % (used on the CIGS page), the total is about 255 billion dollars and 5.5 % is 14 billion dollars, which leaves 8.9 billion for CdTe (computed). The sum is plausible but does not fix the label. The table entry, a-Si, is slightly more likely because it is the more specific statement; both figures are unreferenced and are only a rough guide. Reference: Intal and Ebong, ‘Thin-Film Solar Photovoltaics: Trends and Future Directions’, arXiv:2508.05589 (2025).

Ques: Do the sub-cell currents of the triple junction agree with the measurement?

Table 1 of Intal and Ebong lists sub-cell currents of 4 to 5 $mA/c{\mathrm{m}}^2$ for the top a-Si:H cell (1.85 to 1.90 eV), 6 to 7 for a middle a-SiGe:H cell with about 20 % germanium (1.55 to 1.65 eV) and 10 to 12 for the bottom cell (1.10 to 1.30 eV). In a series stack the three currents are equal, and the triple of Green et al. gives 9.94 $mA/c{\mathrm{m}}^2$. An ideal split of the AM1.5G spectrum with gaps of 1.85, 1.6 and 1.1 eV gives 18.25, 7.15 and 18.83 $mA/c{\mathrm{m}}^2$ (computed), so the table does not match either. The measured 9.94 $mA/c{\mathrm{m}}^2$ exceeds the 7.15 of a 1.6 eV middle cell (computed), so the table describes another design than the certified cell, and its values should not be used. Reference: Intal and Ebong, ‘Thin-Film Solar Photovoltaics: Trends and Future Directions’, arXiv:2508.05589 (2025); Green, Dunlop, Yoshita, Kopidakis, Bothe, Siefer, Hinken, Rauer, Hohl-Ebinger and Hao, ‘Solar cell efficiency tables (Version 64)’, Prog. Photovolt. 32(7), 425-441 (2024).

Related pages in this vault: Copper Indium Gallium Selenide (CIGS) Thin-Film Solar Cell, Silicon Heterojunction (HIT) Solar Cell, Kesterite (CZTS and CZTSSe) Thin-Film Solar Cell, Cadmium Telluride (CdTe) Thin-Film Solar Cell, Solar Cell Fundamentals, Solar Cells and Modules (Crystalline Silicon and Thin Film).