SiliconPV 2025, 15 h In e na ional Con e ence on C ys alline Silicon Pho o ol aicse
Sus ainable Manu ac u ing o Te awa Sola Ene gy
TIB-OP will se DOI wi h TIBdoi
© Au ho s. This wo k is licensed unde a C ea i e Commons A ibu ion 4.0 In e na ional License
Submi ed: 1970-01-01
Recycling Silicon Ke Was e and Qua z Po Sc ap o
Silicon P oduc ion Via Ca bo he mic Reduc ion
Bi gi Ryningen1, Be hane Da sene Dimd1, Naga ajan Somi Ganesan1,
P˚
al Te lie1, I ene B ags ad1, Ma in Bellmann1, Roa Jensen1,
A id Inge Sø ik2, and To inn K ogs ad2
1SINTEF Indus y, Depa men o Sus ainable Ene gy Technology, T ondheim, No way
2No he n Silicon AS, No way
*Co espondence: Be hane Da sene Dimd, be hane.dimd@sin e .no
Abs ac : The ise in pho o ol aic (PV) was e ac oss all s ages o he PV alue chain
has c ea ed a need o e icien ecycling solu ions o ein oduce hese ma e ials in o
use. In esponse, he EU p ojec (ICARUS) is explo ing se e al ecycling ou es o
key was e s eams, including silicon ke , qua z (used c ucibles and po sc ap), and
g aphi e ( u nace insula ion), o silicon p oduc ion. This pape p esen s he de elop-
men o a ca bo he mic ecycling ou e ha uses was e c ucibles, po sc ap, and ke as
aw ma e ials o silicon p oduc ion. Silicon was p oduced in a pilo -scale subme ged
a c u nace, and i s pu i y was analyzed using ICP-MS o quan i y ace impu i ies and
o e all pu i y. The esul s showed ha he silicon ha is p oduced om was e ma e-
ials closely ma ches he cha ac e is ics o me allu gical-g ade silicon. These indings
demons a e he echnical easibili y o was e ma e ials in silicon p oduc ion, wi h sig-
ni ican implica ions o indus ial p ac ices, sus ainabili y e o s, and u u e policies
aimed a p omo ing ci cula economies in high-demand sec o s like silicon p oduc ion.
Keywo ds: Silicon Ke Was e, C ucible Sc ap, Recycling, Ca bo he mic Reduc ion,
PV Silicon
1 In oduc ion
As pho o ol aic (PV) ins alla ions con inue o g ow, he esul ing was e s eams om
PV alue chain a e also inc easing. Silicon ke , g aphi e, and silica was es om his
indus y a e la gely land illed as ine ma e ials, wi h only a small po ion being eused.
This leads o low e-sou ce use e iciency and a signi ican en i onmen al impac , e-
qui ing u gen need o mo e e ec i e ecycling p ac ices in silicon p oduc ion. Silicon
ke and c ucible/po sc ap ep esen signi ican po ions o he PV indus y was e, and
hei quan i ies a e expec ed o ise as demand o silicon con inues o inc ease, mak-
ing hei ecycling impo an . Acco ding o p ojec ions based on s udies [1] and [2],
he es ima ed global silicon ke was e gene a ion om PV manu ac u ing p ocess in
2023 is app oxima ely 387 kMT (kilo me ic ons). This includes con ibu ions om wi e
sawing, as well as ke gene a ed du ing c opping, squa ing, cham e ing, and g inding
B. Ryningen e al. |SiliconPV 2025
p ocesses. Simila ly, he global c ucible and po sc ap was e is es ima ed a 88.8 kMT
and 9.88 kMT, espec i ely o he same yea .
In gene al, a ious s udies in he li e a u e ha e p ima ily ocused on e ining he la ge
olumes o silicon ke gene a ed du ing di e en s ages o silicon wa e p oduc ion [3]
and e-cycling hese ma e ials h ough mel ing [4]. Less a en ion has been gi en o
al e na i e ecycling echniques such as ca bo he mic educ ion. This pape add esses
his esea ch gap by explo ing he echnical easibili y o using was e silicon ke was e
and c ucible sc ap as aw ma e ials o silicon p oduc ion in ca bo he mic educ ion
p ocesses in a subme ged a c u nace (SAF). The s udy aims o e alua e he po en ial
o hese was e ma e ials o p oduce high-pu i y silicon and assess hei sui abili y o
indus ial-scale applica ions, con ibu ing o bo h esou ce e iciency and sus ainabili y
in silicon p oduc ion o PV and o he applica ions.
The emainde o his pape is o ganized as ollows. Sec ion 2 desc ibes he ma e-
ials and me hods, including he esea ch con ex , cha ac e iza ion o was e ma e ials,
p epa a ion o sel - educing b ique es h ough b ique ing, and ope a ion o he sub-
me ged a c u nace (SAF). Sec ion 3 p esen s he esul s o silicon p oduc ion using
he SAF, de ailing bo h he quan i y o silicon p oduced and i s pu i y analysis. Fi-
nally, Sec ion 4 p o ides concluding ema ks along wi h ecommenda ions o u u e
esea ch.
2 Ma e ials and Me hods
This sec ion p o ides a b ie o e iew o he esea ch, desc ibes he was e ma e ials
and hei cha ac e iza ion o silicon p oduc ion, and de ails he cha ge p epa a ion and
expe imen al se up.
2.1 Resea ch Con ex
The esea ch p esen ed in his pape is pa o he ICARUS p ojec , unded unde
he Eu o-pean Union’s Ho izon 2020 esea ch and inno a ion p og amme [5]. The
ICARUS p ojec aims o demons a e modula p ocessing solu ions a an indus ial
scale, a ge ing he eco e y o up o 95% o high- alue aw ma e ials gene a ed du ing
silicon ingo and wa e manu ac u ing. Speci ically, his wo k belongs o a wo k package
which aimed o ein oduce silicon ke , g aphi e, and silica was e back in o he silicon
alue chain h ough ca bo he mic educ ion. I is pa o ou expe imen al campaigns
conduc ed in a pilo -scale SAF.
The wo k conduc ed in his pape can be summa ized as shown in Figu e 1. I p i-
ma ily consis s o ou s ages. The i s s age in ol es he collec ion and p e ea men
o aw ma e ials, including silicon ke and silicon po sc ap. I includes p ocesses such
as washing, d ying, g inding and cha ac e iza ion o he ma e ials. The second s age is
agglome a ion, in which hese p e ea ed aw ma e ials a e b ique ed in o sel - educing
b ique es wi h op imal s oichiome y o ca bo he mic educ ion. The hi d s age con-
sis s o semi-pilo -scale ca bo he mic educ ion expe imen s o p oduce silicon in a
SAF. Finally, a ious es s a e conduc ed o analyse he pu i y o he p oduced silicon.
2.2 Cha ac e iza ion o Was e Ma e ials
The was e ma e ials used in his s udy we e cha ac e ized using LECO (Combus ion
Analyse s) and ICP-MS (Induc i ely Coupled Plasma Mass Spec ome y) analysis o
he ke , and XRD (X- ay Di ac ion) o he po sc ap. LECO analysis o he ke sam-
B. Ryningen e al. |SiliconPV 2025
Figu e 1. Ca bo he mic educ ion o was e ma e ials o silicon p oduc ion.
ples showed an a e age ca bon con en o 1.21 w .% wi h a s anda d de ia ion o 0.08,
and an oxygen con en o 4.44 w .% wi h a s anda d de ia ion o 0.05. Simila ly, ICP-
MS analysis was pe o med o measu e he impu i y con en in he silicon ke samples.
The esul s indica e ha he main dopan impu i ies a e phospho us (P) and gallium
(Ga), while calcium (Ca) is he me allic con aminan wi h he highes concen a ion. On
he o he hand, he po sc ap used in his s udy has an a e age pa icle size o less
han 200 µm. I p ima ily con ains SiO2 om he c ucible, along wi h some silicon. XRD
analysis was pe o med o de e mine he p ecise le els o Si and SiO2in a sample o
powde ed po sc ap. The analysis showed h ee main phases in he sample: Si (16%),
SiO2c is obali e (2%), and amo phous SiO2(82%).
2.3 Agglome a ion o Cha ging Ma e ial (B ique ing P ocess)
B ique ing is used as an agglome a ion me hod o ein oduce he was e ma e ials o
ca bo he mic educ ion in a SAF. Fo his pu pose, qua z sou ce ma e ials, such as
pu e qua z, silicon ke and po sc ap, and ca bon sou ce ma e ial, such as pu e ca -
bon black, a e used. These ma e ials a e agglome a ed in o sel - educing b ique es
using b ique ing machine. Th ee ypes o b ique es we e made o he expe imen :
Type I (pu e qua z and ca bon black wi h binde and wa e ), Type II (po sc ap, ca -
bon black, binde , and wa e ), and Type III (silicon ke wi h binde and wa e ). The
b ique es we e p oduced using a DMSMAC b ique ing machine, Figu e 2. The g een
b ique es we e hea - ea ed o ensu e ha hey possess su icien mechanical s eng h
and low mois u e con en , enabling hem o wi hs and he igo s o cha ging in o he
SAF. This hea ea men included bo h d ying in a hea ing cabine a 105 ◦C o 12–24
hou s and bu ning a 200 ◦C o ano he 12–24 hou s.
Figu e 2. B ique ing p ocess. DMSMAC b ique ing machine and sample b ique es.
B. Ryningen e al. |SiliconPV 2025
2.4 Subme ged A c Fu nace Ope a ion
Silicon is p ima ily p oduced h ough ca bo he mic educ ion in a subme ged a c u -
nace. The p ocess in ol es mixing qua z wi h a ca bon-con aining ma e ial and hen
hea ing he mix u e o high empe a u es. Du ing his p ocess, ca bon educes sili-
con dioxide o p oduce mol en silicon, which is subsequen ly cooled and solidi ied in o
silicon blocks. Fo his wo k, an in-house-buil pilo -scale SAF u nace was used o con-
duc wo expe imen al campaigns (on sepa a e days, one yea apa ). Each campaign
consis ed o wo pilo expe imen s, o aling ou expe imen s. These a e:
• Pilo 1: Type I b ique es and qua z lumps as cha ge.
• Pilo 2: Type I b ique es, qua z lumps, and po sc ap lumps as cha ge.
• Pilo 3: Type II b ique es and po sc ap lumps as cha ge.
• Pilo 4: Type II & III b ique es and po sc ap lumps as cha ge.
The expe imen s aim o in es iga e he use o sel - educing b ique es made om
was e ma e ials in SAF u nace o silicon p oduc ion. The expe imen al campaigns
we e un o e a pe iod o h ee days and h ee nigh s. These expe imen al campaigns
included se e al key ac i i ies including p ehea ing he u nace, ope a ing he u nace,
p epa ing he cha ge ma e- ial, conduc ing odding and apping ope a ions. Figu e 3
shows some o he ac i i ies in hese campaigns.
Figu e 3. Main subme ged a c u nace ope a ions in he campaign.
In SAF ope a ion, elec ical powe is supplied h ough ca bon elec odes subme ged
di ec ly in o he cha ge ma e ial. Elec ic cu en passes h ough he cha ge, gene a -
ing in ense hea due o esis ance, leading o ex emely high empe a u es wi hin he
u nace ha d i e he ca bo he mic educ ion eac ion. In no mal ope a ion, he expe -
imen s we e un in cu en con ol mode, and he p ocess ope a o a ies he cu en
a ound 3 kA o achie e a 140–160 kW load. The ol age is measu ed and con olled
by cu en , he elec ode posi ion, and bu den esis ance.
3 Resul and Discussion
This sec ion p esen s he esul s om wo expe imen al campaigns, comp ising a
o al o ou pilo expe imen s. The objec i e o hese campaigns was o p oduce
me allu gical-g ade silicon using sel - educing b ique es made om pu e and was e
ma e ials, speci ically silicon ke and silicon po sc ap h ough ca bo he mic educ ion
in SAF.
B. Ryningen e al. |SiliconPV 2025
3.1 Campaign 1 (Pilo s 1 and 2)
In his SAF campaign, conduc ed in June 2023, app oxima ely 60 kg o silicon was p o-
duced ac oss six sepa a e apping ope a ions du ing he wo pilo expe imen s. Silicon
was p oduced using sel - educing b ique es made om bo h i gin and was e ma e-
ials. Addi ionally, pu e qua z lumps we e used in Pilo 1, and po sc ap lumps we e
used in Pilo 2 as addi ional cha ge ma e ials. The pu i y o he p oduced silicon was
assessed by analysing six samples om each apping ope a ion using ICP-MS anal-
ysis, which quan i ies ace impu i ies o de e mine he o e all pu i y o he sample.
Table 1 summa izes he esul s o his analysis. F om his able, i can be concluded
ha he p oduced silicon mee s he pu i y equi emen s o me allu gical-g ade silicon
[6]. Addi ionally, he pu i y is consis en ac oss he a ious samples.
Table 1. Pu i y o he apped silicon measu ed by ICP-MS in campaign 1 in bo h pilo s.
Sample 1 2 3 4 5 6
Pilo 1 & 2
(%)
97.89 98.27 98.47 98.39 98.34 98.79
De ailed esul s om he ICP-MS analysis, quan i ying ace impu i ies concen a ion
le els, a e illus a ed in Figu e 4 o dopan s and Figu e 5 o me allic impu i ies. The
esul s show he p esence o dopan impu i ies such as bo on (B), phospho us (P),
and gallium (Ga). Phospho us was ound o be he mos p e alen dopan and ha
i s concen a ion le el in all he apping is sligh ly abo e he h eshold o me allu gical
g ade silicon [6]. The concen a ions o bo h B and Ga a e well wi hin he h eshold o
me allu gical g ade silicon. Simila ly, he ICP-MS analysis also showed he p esence
o me allic impu i ies such as aluminium (Al), i on (Fe), calcium (Ca), i anium (Ti),
ch omium (C ), nickel (Ni), coppe (Cu), and manganese (Mn) in he sample silicon
(Figu e 5). The concen a ion o Al exceeds he h eshold o me allu gical g ade silicon
in he i s ou apping bu alls wi hin he limi s in he las wo ap-ping. Fo Fe, only
he i s , hi d, and las apping ha ing concen a ions wi hin he limi . Calcium exceeds
he h eshold only in he i s apping. In con as , Ti impu i y concen a ions a e abo e
he h eshold ac oss all apping. The impu i y concen a ions o C , Ni, and Cu a e
all wi hin accep able limi s. Despi e he abo e a ia ions, i can be concluded ha he
concen a ions o mos impu i ies a e gene ally wi hin accep able limi s o me allu gical
g ade silicon, showing he ma e ial’s sui abili y o u he applica ions [6].
Figu e 4. Dopan impu i ies in he apped silicon samples in campaign 1.
3.2 Campaign 2 (Pilo s 3 and 4)
In his SAF campaign, conduc ed in May 2024, app oxima ely 32 kg o silicon was
p oduced ac oss six sepa a e apping ope a ions. Un o una ely, only Pilo 4 p oduced
B. Ryningen e al. |SiliconPV 2025
Figu e 5. Me allic impu i ies in he apped silicon samples in Campaign 1.
silicon du ing his campaign. The eason why Pilo 3 did no p oduce silicon emains
unclea , and u he analysis is equi ed o iden i y he unde lying cause. Silicon was
p oduced using sel - educing Type II and Type III b ique es, along wi h po sc ap lumps
as cha ge ma e ials. Like he p e ious campaign, he pu i y o he silicon p oduced was
assessed using ICP-MS analysis o six samples om each apping ope a ion. Table
2 summa izes he esul s o his analysis, indica ing ha he p oduced silicon mee s
he pu i y equi emen s o me allu gical-g ade silicon [6] and has sligh ly highe pu i y
compa ed o silicon om he p e ious campaign.
Table 2. Pu i y o he apped silicon measu ed by ICP-MS in campaign 2 pilo 4.
Sample 1 2 3 4 5 6
Pilo 4 (%) 98.43 98.16 98.01 98.39 98.47 99.05
Simila o he p e ious campaign, de ailed ICP-MS analysis esul s, quan i ying ace
impu i y concen a ions, a e p esen ed in Figu e 6 o dopan s and Figu e 7 o me al-
lic impu i ies. The esul s indica e he p esence o dopan impu i ies such as bo on
(B), phospho us (P), and gallium (Ga). Phospho us (P) was again ound o be he
mos p e alen dopan , al -hough i s concen a ion was signi ican ly lowe compa ed
o he p e ious campaign. Howe e , P le els emain sligh ly abo e he h eshold o
me allu gical-g ade silicon. Simila ly, bo on concen a ions we e signi ican ly lowe and
wi hin he accep able h eshold. In con as , highe concen a ions o gallium impu i ies
we e de ec ed.
Figu e 6. Dopan impu i ies in he apped silicon samples in Pilo 4.
The me allic impu i ies de ec ed include aluminium (Al), i on (Fe), calcium (Ca), i a-
nium (Ti), ch omium (C ), nickel (Ni), coppe (Cu), and manganese (Mn) (Figu e 7).
The concen a ion o Al in hese samples is signi ican ly lowe han in he samples om
he p e ious campaign. The i s ou apping ha e Al concen a ions ha a e sligh ly
abo e he h eshold o me allu gical g ade silicon, while he las wo apping ha e Al
concen a ions well wi hin he h eshold limi . Fo Fe, all apping ha e concen a ions
wi hin he accep able limi . Howe e , he Ca and Ti impu i y concen a ions a e highe
B. Ryningen e al. |SiliconPV 2025
in hese samples compa ed o he p e ious campaign and a e well abo e he h esh-
old. Simila ly, he C and Ni impu i y con en s a e high and exceed he h eshold limi s
o me allu gical g ade silicon. On he o he hand, he concen a ions o Cu and Mn
a e well wi hin he accep able limi s. All in all, he silicon p oduced in his campaign
has a pu i y and cha ac e is ics compa able o me allu gical g ade silicon [6]. This is
a signi ican inding because 70% o he ma e ials used in his campaign came om
was e po sc ap and was e silicon ke , unlike he p e ious campaign.
Figu e 7. Me allic impu i ies in he apped silicon samples in Pilo 4.
4 Conclusion
This s udy p esen ed he wo k conduc ed in he wo expe imen al campaigns comp is-
ing ou pilo -scale expe imen s which aimed a p oducing me allu gical-g ade silicon
using sel - educing b ique es made om bo h pu e and was e ma e ials (silicon ke
and po sc ap) h ough ca bo he mic educ ion in a subme ged a c u nace. O e all,
hese esul s highligh he echnical easibili y o using sel - educing b ique es om
was e ma e ials o p oducing me allu gical-g ade silicon, he eby suppo ing sus ain-
able p ac ices wi hin silicon manu ac u ing. The silicon pu i y ha is achie ed using
was e ma e ials has signi ican implica ions o educing he en i onmen al oo p in
and cos o silicon p oduc ion. Addi ionally, he use o was e silicon sou ces such as
ke and po sc ap o e a p omising pa hway o esou ce eco e y and was e minimiza-
ion. Fu u e wo k should ocus on u he op imizing ecycling p ocesses, pa icula ly
add essing dopan impu i ies and me allic con aminan s. Addi ional e o s should also
add ess ope a ional challenges, such as he p oduc ion inconsis encies obse ed be-
ween pilo expe imen s. Fu he mo e, he demons a ed echnical easibili y should
be complemen ed by comp ehensi e assessmen s o economic iabili y and en i on-
men al impac s. Only his holis ic app oach will pa e he way o mo e sus ainable and
cos -e ec i e silicon p oduc ion p ac ices.
Da a a ailabili y s a emen
The da a ha suppo he indings o his s udy can be a ailable om he co esponding
au ho upon easonable eques .
Au ho con ibu ions
Bi gi Ryningen: Concep ualiza ion, Fo mal analysis, Me hodology, Supe ision, W i -
ing – o iginal d a , and W i ing – e iew & edi ing; Be hane Da sene Dimd: Con-
cep ualiza ion, Fo mal analysis, Me hodology, Da a cu a ion, Visualiza ion, W i ing –
o iginal d a , and W i ing – e iew & edi ing; Naga ajan Somi Ganesan: Fo mal anal-
ysis and W i ing – o iginal d a ; P˚
al Te lie: Me hodology and W i ing – o iginal d a ;
B. Ryningen e al. |SiliconPV 2025
I ene B ags ad: Me hodology and Da a cu a ion; Ma in Bellmann: Concep ualiza ion,
Me hodology, Supe ision, and W i ing – e iew & edi ing; Roa Jensen: Concep ual-
iza ion, Me hodology, and Supe ision; A id Inge Sø ik: Me hodology and Supe i-
sion; To inn K ogs ad: Me hodology and Supe ision.
Compe ing in e es s
The au ho s decla e ha hey ha e no compe ing in e es s.
Funding
This wo k is pa o ICARUS p ojec unded by Ho izon 2020 esea ch and inno a ion
p o-g am unde g an ag eemen No 958365.
Re e ences
[1] P. B ailo sky, K. Baumann, M. Held, e al., “Insigh s in o ci cula ma e ial and was e lows
om c-si p indus y,” EPJ Pho o ol aics, ol. 14, p. 5, 2023. DOI:10.1051/epjp /2022029.
[2] “P manu ac u ing in eu ope: Unde s anding he alue chain o a success ul indus ial pol-
icy,” ETIP PV Indus y Wo king G oup Whi e Pape , Eu opean Technology and Inno a ion
Pla o m o Pho o ol aics, Tech. Rep., 2023.
[3] H. Yang, I. Liu, C. Liu, H. Hsu, and C. Lan, “Recycling and euse o ke -loss silicon om
diamond wi e sawing o pho o ol aic indus y,” Was e Managemen , ol. 84, pp. 204–210,
2019. DOI:10.1016/j.wasman.2018.11.045.
[4] J. Kong, P. Xing, Y. Liu, e al., “An economical app oach o he ecycling o high-pu i y
silicon om diamond-wi e saw ke slu y was e,” Silicon, ol. 11, no. 1, pp. 367–376, 2019.
DOI:10.1007/s12633-018-9889-x.
[5] ICARUS H2020 P ojec . “P ocessing and e ining o seconda y aw ma e ials om silicon
pho o ol aic manu ac u ing.” (2021), [Online]. A ailable: h ps://www.ica us.eu.com/
( isi ed on 08/12/2025).
[6] WikiChip. “Me allu gical-g ade silicon (mgs).” (2018), [Online]. A ailable: h ps : / / en .
wikichip . o g / wiki / me allu gical - g ade _ silicon # google _ igne e ( isi ed on
08/12/2025).
