This is a sel -a chi ed e sion o an o iginal a icle. This e sion
may di e om he o iginal in pagina ion and ypog aphic de ails.
Au ho (s):
Ti le:
Yea :
Ve sion:
Copy igh :
Righ s:
Righ s u l:
Please ci e he o iginal e sion:
CC BY 4.0
h ps://c ea i ecommons.o g/licenses/by/4.0/
So wa e a chi ec u e o quan um compu ing sys ems : a sys ema ic e iew
© 2023 The Au ho s. Published by Else ie Inc.
Published e sion
Khan, A i Ali; Ahmad, Aakash; Waseem, Muhammad; Liang, Peng; Fahmideh,
Mahdi; Mikkonen, Tommi; Ab ahamsson, Pekka
Khan, A. A., Ahmad, A., Waseem, M., Liang, P., Fahmideh, M., Mikkonen, T., & Ab ahamsson, P.
(2023). So wa e a chi ec u e o quan um compu ing sys ems : a sys ema ic e iew. Jou nal o
Sys ems and So wa e, 201, A icle 111682. h ps://doi.o g/10.1016/j.jss.2023.111682
2023
The Jou nal o Sys ems & So wa e 201 (2023) 111682
Con en s lis s a ailable a ScienceDi ec
The Jou nal o Sys ems & So wa e
jou nal homepage: www.else ie .com/loca e/jss
So wa e a chi ec u e o quan um compu ing sys ems — A sys ema ic
e iew✩
A i Ali Khana,∗,Aakash Ahmadb,Muhammad Waseemc,Peng Liangc,Mahdi Fahmidehd,
Tommi Mikkonene,Pekka Ab ahamssone
aM3S Empi ical So wa e Enginee ing Resea ch Uni , Uni e si y o Oulu, 90014 Oulu, Finland
bSchool o Compu ing and Communica ions, Lancas e Uni e si y, Leipzig, Ge many
cSchool o Compu e Science, Wuhan Uni e si y, Wuhan, China
dSchool o Business a Uni e si y o Sou he n Queensland, Queensland, Aus alia
eFacul y o In o ma ion Technology and Communica ion Sciences, Tampe e Uni e si y, 33014 Tampe e, Finland
a icle in o
A icle his o y:
Recei ed 16 July 2022
Recei ed in e ised o m 10 Ma ch 2023
Accep ed 20 Ma ch 2023
A ailable online 24 Ma ch 2023
Keywo ds:
Quan um compu ing
Quan um so wa e enginee ing
Quan um so wa e a chi ec u e
Sys ema ic li e a u e e iew
abs ac
Quan um compu ing sys ems ely on he p inciples o quan um mechanics o pe o m a mul i ude o
compu a ionally challenging asks mo e e icien ly han hei classical coun e pa s. The a chi ec u e o
so wa e-in ensi e sys ems can empowe a chi ec s who can le e age a chi ec u e-cen ic p ocesses,
p ac ices, desc ip ion languages o model, de elop, and e ol e quan um compu ing so wa e (quan um
so wa e o sho ) a highe abs ac ion le els. We conduc ed a Sys ema ic Li e a u e Re iew (SLR)
o in es iga e (i) a chi ec u al p ocess, (ii) modelling no a ions, (iii) a chi ec u e design pa e ns,
(i ) ool suppo , and (i ) challenging ac o s o quan um so wa e a chi ec u e. Resul s o he SLR
indica e ha quan um so wa e ep esen s a new gen e o so wa e-in ensi e sys ems; howe e ,
exis ing p ocesses and no a ions can be ailo ed o de i e he a chi ec ing ac i i ies and de elop
modelling languages o quan um so wa e. Quan um bi s (Qubi s) mapped o Quan um ga es (Quga es)
can be ep esen ed as a chi ec u al componen s and connec o s ha implemen quan um so wa e.
Tool-chains can inco po a e eusable knowledge and human oles (e.g., quan um domain enginee s,
quan um code de elope s) o au oma e and cus omise he a chi ec u al p ocess. Resul s o his SLR
can acili a e esea che s and p ac i ione s o de elop new hypo heses o be es ed, de i e e e ence
a chi ec u es, and le e age a chi ec u e-cen ic p inciples and p ac ices o enginee eme ging and nex
gene a ions o quan um so wa e.
©2023 The Au ho s. Published by Else ie Inc. This is an open access a icle unde he CC BY license
(h p://c ea i ecommons.o g/licenses/by/4.0/).
1. In oduc ion
Quan um compu ing elies on quan um mechanics, a disci-
pline mo e amilia and cen e o a en ion o physicis s a he
han compu e scien is s o so wa e enginee s (Zhao,2020;
Deu sch,1985;Di ac,1981). Howe e , in ecen yea s, wi h an
eme gence o quan um algo i hms and Quan um P og amming
Languages (QPL), so wa e p og amme s ha e been able o exploi
he heo y and p inciple o quan um mechanics o p ocess in o -
ma ion and pe o m speci ic compu a ion asks as e han classi-
cal compu ing sys ems (Chong e al.,2017;Ying,2016). Compa ed
o classical algo i hms o compu a ion, quan um algo i hms ha e
✩Edi o : P o . Neil E ns .
∗Co esponding au ho .
E-mail add esses: [email p o ec ed] (A.A. Khan), [email p o ec ed]
(A. Ahmad), [email p o ec ed] (M. Waseem), [email p o ec ed]
(P. Liang), [email p o ec ed] (M. Fahmideh), [email p o ec ed]
(T. Mikkonen), [email p o ec ed] (P. Ab ahamsson).
he po en ial o sol e a se o compu a ionally challenging p ob-
lems such as na u e-inspi ed compu ing, inancial modelling,
and ad anced enc yp ion wi h inc eased e iciency (Mon ana o,
2016;G imsley e al.,2019;K üge and Maue e ,2020). Quan-
um compu ing a ibu es (e.g., Qubi s, supe posi ion, en angle-
men , and in e e ence) lie a he hea o quan um in o ma ion
p ocessing (Zeilinge ,1999;Gay,2006). Quan um p og amming
languages ha implemen quan um algo i hms enable quan um
sup emacy in compu ing ha is lacking in adi ional compu ing
sys ems. Mon ana o (2016), Gay (2006), So ge (2008). One class
o such p oblems ela e o in o ma ion and compu a ion science
ha equi es la ge amoun s o pa allel p ocessing (Nguyen e al.,
2022) o ackling challenges, such as op imisa ion, enc yp ion,
big da a analy ics, and machine lea ning (Biamon e e al.,2017;
Reben os e al.,2014). O he se o p oblems ela e o e i-
cien and accu a e simula ion o quan um sys ems in na u al
sciences, such as physics (G imsley e al.,2019), chemis y (McA -
dle e al.,2020), ma hema ics (K üge and Maue e ,2020), and
challenges ela ing o hei applica ions (S epney e al.,2005;
h ps://doi.o g/10.1016/j.jss.2023.111682
0164-1212/©2023 The Au ho s. Published by Else ie Inc. This is an open access a icle unde he CC BY license (h p://c ea i ecommons.o g/licenses/by/4.0/).
A.A. Khan, A. Ahmad, M. Waseem e al. The Jou nal o Sys ems & So wa e 201 (2023) 111682
Childs e al.,2018;Mosca,2018). Howe e , QPL and hei un-
de lying algo i hms ocus on compu a ion and implemen a ion
de ails o p oduce execu able speci ica ions, bu lack an o e all
global iew o he so wa e sys ems unde design. Sou ce code
based implemen a ion de ails unde mine a chi ec u al iew(s) as
sys em bluep in , ha can comp omise he quali y and unc ion-
ali y o end p oduc , i.e., quan um so wa e (Zhao,2020;Moguel
e al.,2020;Pia ini e al.,2021). Technology gian s a e scaling up
hei inancial and s a egic in es men s in quan um compu ing
pla o ms, mo e speci ically quan um p og amming languages
such as Q# om Mic oso , Qiski om IBM, and Ci q om Google,
howe e ; quan um so wa e enginee ing and de elopmen is s ill
in i s in ancy (Mic oso ,2021;Behe a e al.,2019;Cou land,
2017). Some ecen esea ch s udies also indica e ha quan um
so wa e p ojec s ha o e look design p inciples o p ima ily
ocus on quan um sou ce code implemen a ions, o en lead o
aul y implemen a ions and bugs in quan um so wa e (Zhao
e al.,2021;Campos and Sou o,2021).
So wa e a chi ec u e as desc ibed in he ISO/IEC 42010 s an-
da d p o ides a global iew o so wa e-in ensi e sys ems,
ep esen ing hei blue-p in , by abs ac ing complex implemen-
a ion de ails wi h a chi ec u al componen s and connec o s
(Anon,2022;Ho meis e e al.,2007;Li e al.,2013). So wa e
de elope s and a chi ec s ha e success ully used a chi ec u al
desc ip ions and speci ica ions o design, de elop, alida e, and
e ol e so wa e-in ensi e sys em a highe -le el o abs ac ions
while main aining sys em unc ionali y and quali y (Mala ol a
e al.,2012;Di F ancesco e al.,2019). A chi ec u al models ha e
been exploi ed o design, de elop, and alida e eme ging gene -
a ions o so wa e-in ensi e sys ems including bu no limi ed
o he in e ne o hings, blockchain applica ions, and a i i-
cially in elligen sys ems (Al eshidi and Ahmad,2019;Fahmideh
e al.,2021a;Xu e al.,2017;Fahmideh e al.,2021b;G ae and
Geo gie ski,2021). Quan um So wa e A chi ec u e (QSA), as a
new gen e o So wa e A chi ec u es (SA), can p o ide a chi ec-
u al desc ip ions (i.e., componen s, connec o s, and con igu a-
ions) o design and de elop quan um so wa e, while abs ac ing
complex and implemen a ion speci ic asks (Ho meis e e al.,
2007;Ga cia e al.,2021). Speci ically, a chi ec u al componen s
can ep esen modules o sou ce code while a chi ec u al con-
nec o s speci y in e ac ions be ween modules o ep esen he
s uc u e and beha iou o a sys em (Ga cia e al.,2021). T ans-
o ma ion om abs ac high-le el models (i.e., design a i ac s)
o low-le el execu able speci ica ions (i.e., sou ce code a i ac s)
can be enabled ia model-d i en a chi ec ing o quan um so -
wa e (Moin e al.,2021;Pé ez-Cas illo e al.,2021). Howe e ,
QSA as an eme ging discipline emains an unde -explo ed a ea
by he cu en gene a ion o designe s and a chi ec s who ind
hemsel es less p epa ed o ackle he challenges ela ed o
QSA in he de elopmen li e-cycle o quan um so wa e (Mag-
nani,2022;Shepa d,2021;Na eh,2021). Despi e a ple ho a o
published esea ch in ecen yea s ha ocuses on enginee ing
and a chi ec ing quan um so wa e, he e do no exis any e i-
dence, i.e., empi ical s udy o da a-d i en analysis o consolida e
a collec i e impac o exis ing esea ch on a chi ec ing quan um
so wa e (Ga cia e al.,2021;Moin e al.,2021;Pé ez-Cas illo
e al.,2021;Magnani,2022).
Sys ema ic Li e a u e Re iews (SLRs) ely on E idence-based
So wa e Enginee ing (EBSE) app oach o iden i y, classi y, com-
pa e, and syn hesise published esea ch as an e idence o em-
pi ically in es iga e he opic unde in es iga ion (Di F ancesco
e al.,2019;Ki chenham and Cha e s,2007). Recen ly, a numbe
o SLRs and e iew based s udies ha e been conduc ed o in es-
iga e he applica ion o So wa e Enginee ing (SE) o quan um
compu ing sys ems, howe e ; he e is no e o o e iew he
s a e-o - he-a on a chi ec ing quan um so wa e (Zhao,2020;
Moguel e al.,2020;Pia ini e al.,2021;Gill e al.,2022). The e-
o e, he objec i e o his e iew is o complemen SE based
s udies and speci ically ocus on iden i ica ion, classi ica ion, and
syn hesis o he published esea ch on he ole ha so wa e a chi-
ec u e plays in de eloping quan um compu ing sys ems. We aim
o in es iga e he co e concep s, unde pinning undamen als o
so wa e a chi ec u al aspec s, o en o e looked in SE ocused
s udies, by ou lining a numbe o Resea ch Ques ions (RQs). These
RQs ocus on (i) a chi ec u al p ocess (uni ying a chi ec ing ac i -
i ies), (ii) modelling no a ions (a chi ec u al ep esen a ion), (iii)
pa e ns and design decisions ( eusable knowledge and bes p ac-
ices), (i ) ool suppo (enabling au oma ion and cus omisa ion)
and ( ) eme ging challenges o quan um so wa e a chi ec u es.
These RQs a e mo i a ed by academic esea ch and indus ial
s udies on so wa e a chi ec u e ha highligh he needs o
p ocess-cen ic a chi ec ing, whe e a p ocess ac s as an umb ella
o suppo a ious a chi ec u al aspec s (Ho meis e e al.,2007;
Mala ol a e al.,2012). Mo eo e , in quan um so wa e enginee -
ing li ecycle (Dey e al.,2020), du ing sys em design, a chi ec u al
aspec s such as so wa e modelling, pa e ns, ools, and human
oles a e as undamen al o a chi ec u e-cen ic enginee ing o
quan um so wa e (Pia ini e al.,2021). Resul s and indings o
his SLR complemen exis ing su eys on Quan um So wa e Engi-
nee ing (QSE) and can p o ide ounda ions o u he seconda y
s udies ha can explo e a chi ec u al p inciples and p ac ices o
design and de elop quan um so wa e.
The esul s o his SLR indica e ha al hough quan um so -
wa e ep esen s a new gene a ion o so wa e applica ions,
ounda ions o quan um so wa e a chi ec u es a e g ounded
in a chi ec u al p ocesses and a chi ec ing ac i i ies o classi-
cal sys ems (e.g., objec , se ice, o componen -based) (K üge
and Maue e ,2020;Mala ol a e al.,2012;Di F ancesco e al.,
2019;DiAdamo e al.,2021). Quan um-speci ic ea u es in ol ing
Qubi s (e.g., quan um en anglemen and quan um supe posi ion)
elabo a ed la e , do equi e ailo ed a chi ec u al p ocesses and
modelling no a ions, such as exploi ing he Uni ied Modelling
Language (UML) o e ec i ely add ess he challenges o he quan-
um age a chi ec u es (Pé ez-Cas illo e al.,2021). Speci ically,
exis ing p ocesses and no a ions need cus omisa ion o enable co-
design o quan um sys ems ha can enable he mapping be ween
Quga es and Qubi s o so wa e a chi ec u al componen s and
connec o s. Tool-chain o suppo quan um a chi ec ing p ocess
can acili a e sys em and so wa e a chi ec s o achie e au oma-
ion and inco po a e human decision suppo while designing and
implemen ing quan um so wa e. The esul s o he SLR can be
bene icial o :
(i) Resea che s who a e in e es ed in unde s anding heo y
and p inciples o a chi ec u e-in ensi e de elopmen , es-
ablishing new hypo heses o be es ed, and de eloping
e e ence a chi ec u es and solu ions o quan um so -
wa e.
(ii) P ac i ione s who would like o unde s and he a chi ec -
ing ac i i ies, pa e ns as eusable knowledge, exis ing and
equi ed ool chain, and he ex en o which he academic
esea ch can be le e aged o de elop indus y scale solu-
ions o quan um so wa e.
The es o he pape is o ganised as ollows: Sec ion 2p esen s
he con ex and backg ound o his esea ch s udy. Sec ion 3
de ails he esea ch me hodology o conduc he s udy. Sec-
ions 4–5p esen he esul o he s udy. Sec ion 6discusses
he co e inding and implica ions o he s udy esul s. Sec ion 7
elabo a ed on h ea s o he alidi y o he esea ch. Sec ion 8
e iews and p o ides compa a i e analysis o he mos ele an
esea ch s udies. Sec ion 9concludes he s udy wi h a discussion
o po en ial u u e esea ch.
2
A.A. Khan, A. Ahmad, M. Waseem e al. The Jou nal o Sys ems & So wa e 201 (2023) 111682
2. Con ex : A chi ec ing so wa e o quan um compu ing
This sec ion con ex ualises quan um compu ing sys ems in
e ms o hei building blocks, i.e., (a) quan um ha dwa e, (b)
quan um so wa e, and (c) quan um so wa e a chi ec u e as
shown in Fig. 1. Mo e speci ically, Fig. 1 p o ides a isual e e -
ence ha co ela es he Qubi s and Quga es o quan um sou ce
code, ep esen ing design and implemen a ion phase o QSE li e-
cycle. So wa e a chi ec u al componen s and connec o s p o ide
a blue-p in o implemen he quan um sou ce code. We use he
illus a ions in Fig. 1, elabo a ed below, o in oduce undamen al
concep s and e minologies ha will be used h oughou he
pape .
2.1. Quan um compu ing sys ems
To gain s a egic ad an ages o quan um in o ma ion p o-
cessing, echnology gian s, such as IBM, Google, Mic oso and
go e nmen al o ganisa ions a e hea ily in es ing in he esea ch
and de elopmen o quan um sys ems (Mic oso ,2021;Behe a
e al.,2019;Cou land,2017;Goled,2021). F om he sys em’s
enginee ing pe spec i e, as shown in Fig. 1, undamen al o quan-
um compu ing ha dwa e is he concep o Qubi (quan um bi )
ha ep esen s he mos undamen al uni o quan um in o -
ma ion p ocessing (Zeilinge ,1999;Gay,2006). Con a y o he
classical bi (bina y digi ) ha is exp essed as [1, 0] in digi-
al compu ing sys ems, a Qubi ep esen s a wo-s a e quan um
compu e and hese wo s a es a e speci ied as |0⟩and |1⟩. The
combina ions o bi s ep esen low o digi al in o ma ion ha
al e s he s a e o bina y logic ga es (on: 0 o : 1) o make digi al
sys ems wo k. Analogous o he bina y ga es, quan um ga e (a.k.a.
he quan um logic) ep esen s he building blocks o a quan um
ci cui and ansi s i s s a e ia Qubi (Gay,2006) as in Eq. (1).
A Qubi can be in a s a e |0⟩ = [1
0]and |1⟩ = [0
1]o (unlike a
classical bi ) in a linea combina ion o bo h s a es.
|0⟩ = [1
0]+ |1⟩ = [0
1](1)
In Fig. 1(a), we illus a e and elabo a e on he dis inc ion
be ween a Bi and Qubi . A Bi is like a ga e in an elec onic
ci cui ha can be ei he on o o , whe eas a Qubi uses he
unique p ope ies o quan um mechanics o p o ide a uni ha
can be one o ze o- o any hing in be ween. The bi can ake
a alue o ‘0’ o ‘1’ as ei he ‘O ’ o ‘On’ wi h 100% p obabili y
(le ). A qubi can be in a s a e o |0⟩o |1 o in a supe posi-
ion s a e wi h 50% |0⟩and 50% |1⟩, supe posi ion s a e (le ).
Two Qubi s a e in an en angled s a e ( igh ) - en angled qubi s
a e linked such ha by looking (i.e., measu ing) one o hese
wo, will e eal he s a e o o he Qubi . Fu he de ails abou
Qubi and Quga e in he con ex o ope a ionalising he QC sys-
ems can be ound in Zhao (2020), Zeilinge (1999). Like he
classical compu ing sys ems, con olling he Qubi s ha manip-
ula e Quga es, he e is a need o quan um so wa e sys ems and
applica ions ha can exploi bene i s o quan um in o ma ion
p ocessing by ope a ionalising quan um compu e s. Fo example,
QuNe Sim (DiAdamo e al.,2021) is a Py hon so wa e ame-
wo k ha is capable o managing quan um ci cui s o simula e
p ocessing and ansmission o quan um in o ma ion ia quan-
um ne wo ks. Fig. 1 shows ha in o de o enable quan um
so wa e applica ions o u ilise quan um ha dwa e, he e is a
need o quan um code compile s ha can ansla e high-le el
compu a ional ins uc ions in o machine ansla ed code o con-
ol quan um ha dwa e (Chong e al.,2017;Suni a e al.,2021).
As a ypical example o such compila ion a e he solu ions by
p oposed by Ying (2016) and, K üge and Maue e (2020), which
ecei e he compiled code ha can be execu ed o simula ed
on quan um pla o ms o enable quan um p ocessing o op i-
mising solu ions ega ding uns uc u ed da a sea ching, pa allel
p ocessing, and na u e inspi ed compu ing. In ecen yea s, a
ple ho a o esea ch and de elopmen has eme ged ha ocused
on quan um algo i hms and p og amming languages o add ess
he abo e-men ioned compu a ional challenges e ec i ely and
e icien ly (Suni a e al.,2021). Quan um algo i hms ha e he
po en ial o p o ide compu a ion e iciency o so wa e enginee -
ing p oblems in a eas including bu no limi ed o da a mining,
machine lea ning, and c yp og aphy ha do no scale op imally
on non-quan um compu ing pla o ms (Mi anskyy e al.,2022).
Despi e he signi icance o quan um p og amming languages o
p oduce execu able speci ica ions o quan um ha dwa e; he e is
a need o o e all enginee ing li ecycle(s) ha goes beyond le el
o sou ce code o speci y, execu e, alida e, and e ol e so wa e-
in ensi e sys em based on equi ed unc ionali y and desi ed
quali y (Moguel e al.,2020;Pia ini e al.,2021).
2.2. So wa e Enginee ing (SE) o quan um compu ing
So wa e enginee ing, as de ined in he ISO/IEC/IEEE 90003:
2018 s anda d aims o apply enginee ing p inciples and p ac-
ices o design, de elop, alida e, deploy, and e ol e so wa e-
in ensi e sys ems e ec i ely (ISO/IEC/IE.E.E. 90003:2018,2021).
In ecen yea s, SE ocused esea ch and de elopmen s a ed o
ackle, such as quan um so wa e models, hei algo i hmic spec-
i ica ions, and simula ed e alua ions o le e age bene i s o quan-
um ha dwa e o quan um in o ma ion p ocessing (Mon ana o,
2016;G imsley e al.,2019;Reben os e al.,2014;Childs e al.,
2018;S o e e al.,2006). Mo e speci ically, so wa e enginee s
can le e age SE p ac ices and pa e ns by ollowing so wa e
p ocess(es) ha comp ises o a mul i ude o enginee ing ac i i ies
including bu no limi ed o equi emen s enginee ing, design,
implemen a ion, e alua ion, and deploymen , as shown in Fig. 1.
SE ac i i ies adop ed om quan um and classical so wa e en-
ginee ing concep s a e used o ep esen a simpli ied iew o
quan um SE p ocess (see Fig. 1 (b)) (Zhao,2020;Do man and
Thaye ,1997). Such gene alised p ocess can be ailo ed (adding,
emo ing, and/o cus omising any ac i i ies) as pe he con ex o
sys em de elopmen .
Quan um compu ing sys ems a e in a phase o con inuous
e olu ion and consequen ly quan um SE ep esen s a new gen-
e a ion o so wa e-in ensi e enginee ing ac i i ies o de elop
applica ions ha can con ol he unde lying ha dwa e (Pia ini
e al.,2021). In ecen yea s, esea ch communi ies on so wa e
enginee ing and so wa e a chi ec u e ha e ocused on es ab-
lishing dedica ed o ums, i.e., con e ences, wo kshops and alike
o ums in an a emp o se he agenda(s), s eamline eme gen
challenges, and p opose communi y wide ini ia i es o enginee
and a chi ec quan um so wa e (Ab eu e al.,2021a;Ba zen
e al.,2021). These QSE ocused esea ch communi ies in end
o ga he esea che s and p ac i ione s and p o ide a o um o
collabo a e and explo e he possibili ies o exploi exis ing so -
wa e enginee ing me hods ha can be applied o quan um e a
compu ing and so wa e sys ems (Ab eu e al.,2021b;Ali e al.,
2022). The Quan um Flagship ep esen s a p ime example o
suppo sus ainable esea ch and de elopmen o consolida ing
and expanding scien i ic leade ship, achie ing excellence and
inno a ion in quan um compu ing echnologies (Anon,2022a).
QSE p ocess may in ol e an addi ional challenge o managing
hyb id applica ions and algo i hms. A hyb id applica ion and i s
unde lying implemen a ion in ol e spli ing he o e all applica-
ion in o classical modules (p e/pos -p ocessing) and quan um
modules (quan um compu a ion) e e ed o as he quan um-
classic spli (Wede e al.,2022), as shown in Fig. 1(b). Resea ch
3
A.A. Khan, A. Ahmad, M. Waseem e al. The Jou nal o Sys ems & So wa e 201 (2023) 111682
Fig. 1. A simpli ied iew o quan um compu ing sys ems ((a) Quan um ha dwa e, (b) Quan um so wa e, (c) Quan um so wa e a chi ec u e).
on he quan um-classic spli is gaining a en ion wi h an aim o
de elop QSE p ocess(es) ha enable quan um so wa e designe s
and de elope s o enginee hyb id applica ions by applying he
quan um-classic spli pa e n (Pé ez-Cas illo and Pia ini,2022).
In addi ion o he needs o inno a i e echnologies and p o-
cesses, p inciple, and p ac ices ha speci ically ackle challenges
o quan um so wa e modelling and a chi ec ing, coding, and
simula ion, exis ing classical SE p ocesses can be cus omised
o enginee and de elop quan um so wa e (S o e e al.,2006;
Baczewski e al.,2017). Fo example, he concep o a chi ec u al
modelling as a gene ic a chi ec ing ac i i y, can be cus omised
wi h ini ia i es like quan um UML p o ile, exploi ing he UML
ac i i y diag ams ha could help model pa allel compu ing o
quan um sea ch algo i hms (Pé ez-Cas illo e al.,2021). UML
p o iles o quan um sys ems enable so wa e designe s o c ea e
mul iple iews as di e en pe spec i es o sys em unde design.
Fo example, he designe can u ilise he ac i i y diag am o
design quan um ci cui s (Pé ez-Cas illo e al.,2021) o u ilise use
case, sequence, o deploymen diag ams o design he in e ac-
ion, con ol low, and con igu a ion iews o classical-quan um
so wa e (Pé ez-Cas illo and Pia ini,2022). Simila ly, exis ing
equi emen s enginee ing p ocess can be ailo ed o suppo
equi emen s o quan um (i.e., quan um en anglemen ) ha is
missing in he exis ing models. In SE p ocess(es), a chi ec ing ep-
esen s a pi o al ac i i y ha accumula es sys em equi emen s
as a model hus leading o so wa e implemen a ion, alida ion,
and e olu ion while main aining a global iew o he sys em
and managing a chi ec u al ade-o s (Mala ol a e al.,2012;
Di F ancesco e al.,2019).
2.3. A chi ec u e o quan um so wa e
A chi ec u e o so wa e in ensi e sys ems, as desc ibed in he
ISO/IEC/IEEE 42010:2022 s anda d, aims o abs ac complex and
implemen a ion speci ic de ails o ep esen sys em bluep in in
an implemen a ion and echnology neu al way (Anon,2022).
4

A.A. Khan, A. Ahmad, M. Waseem e al. The Jou nal o Sys ems & So wa e 201 (2023) 111682
Fig. 2. An o e iew o he esea ch me hodology o SLR.
Empi ically-g ounded academic esea ch and indus ial s udies
on a chi ec ing so wa e-in ensi e sys ems ha e highligh ed ha
he e is no uni ied iew o ep esen so wa e a chi ec u es
(Ho meis e e al.,2007;Mala ol a e al.,2012;Med ido ic and
Taylo ,2000). Di e en a chi ec u al iews (also e e ed o as
a chi ec u al models o ep esen a ions) can also be a ibu ed
o a mul i ude o modelling app oaches suppo ed ia UML,
ADL, and g aph models ha allow so wa e p ac i ione s o c e-
a e cus omised a chi ec u al iew(s) ha i s hei con ex in a
speci ic a chi ec ing ac i i y (Pé ez-Cas illo and Pia ini,2022;
Med ido ic e al.,2002). Fo example, conside ing he 4 +1 a -
chi ec u al iew (Ho meis e e al.,2007), equi emen s enginee s
may be mo e in e es ed in he in e ac ion model(s) exp essed
as g aphs o UML use case diag ams ha cap u e a chi ec u ally
signi ican equi emen s ( unc ionali y and quali y o sys em). In
compa ison, so wa e de elope s and quali y enginee s/ es e s
a e mo e likely o u ilise he componen and connec o models
ha ep esen modules o sou ce code and hei in e ac ions, and
un ime iew ha models sys em execu ion as UML sequence
diag ams. As pe he 4 +1 a chi ec u al iew, in his s udy, we
ha e mainly elied on he componen and connec o a chi ec u e
model (Fig. 1) ha ep esen s so wa e in e ms o compu a-
ions and da a s o es. Howe e , du ing a chi ec u al e iew and
syn hesis, he componen and connec o a chi ec u al models
alone a e no su icien and he e o o consolida e a singu-
la o uni ied iew ha suppo s a ious a chi ec u al ac i i ies
may be imp ac ical. Once exp essed, some a chi ec u al models,
i.e., model d i en a chi ec u e can help o gene a e he necessa y
skele on o lib a ies o sou ce code in a (semi-) au oma ed way
using model-d i en enginee ing (Moin e al.,2021). In ecen
yea s, a chi ec u al models and no a ions ha e p o en o be
success ul o design and de elop so wa e in ensi e sys ems by
enabling eusabili y (pa e ns and s yles), e ol abili y (a chi ec-
u al econ igu a ions), and elas ici y (au o-scaling) (Ho meis e
e al.,2007;Li e al.,2013). Fig. 1(c) illus a es a pa ial a chi-
ec u al iew o a quan um algo i hm o ac o ise in ege s ha is
modelled as UML componen diag am (Pé ez-Cas illo e al.,2021).
The a chi ec u al iew abs ac s he sou ce code le el de ails o
p esen design decisions in e ms o componen s (Sho _Fac o ,
Sho _O de ) ha coo dina e ia a connec o (ge Fac o s) o in e-
ge ac o isa ion. A chi ec u e in i sel ep esen s non-execu able
speci ica ions o he quan um sea ch sys em, howe e ; he ap-
plica ion o model-d i en enginee ing can help a chi ec s and
designe s o de i e sou ce code di ec ly om a chi ec u e models.
In he o e all iew o Fig. 1, we can conclude ha in quan um
compu ing sys ems, so wa e a chi ec u e ep esen s a blue-p in
o de elop so wa e sys ems and applica ions ha manipula e
quan um ha dwa e. Quan um so wa e p ojec s p ima ily ocused
on p oducing quan um sou ce code while o e looking quan um
so wa e design a e o en p one o bugs and un ul illed equi e-
men s (Campos and Sou o,2021). The ole o so wa e a chi ec-
u e in quan um SE is pi o al o de elop he equi emen s, which
lead o so wa e designing, coding, alida ion, and deploymen ,
all acili a ed using a chi ec u al no a ions. So wa e a chi ec u e
o quan um compu ing sys ems (quan um so wa e a chi ec-
u e) can empowe he ole o so wa e enginee s and de elope s
o c ea e models ha ac as basis o sys em implemen a ion.
Based on he a chi ec u al models, model d i en enginee ing
and de elopmen can be exploi ed o he au oma ed gene a ion
o quan um sou ce code (code modules and hei in e ac ions)
om he co esponding quan um so wa e a chi ec u e (based
on a chi ec u al componen s and hei connec o s) (Ying,2016;
Moin e al.,2021;Shepa d,2021).
3. Resea ch me hodology
We ollowed EBSE app oach o conduc his esea ch (Piza d
e al.,2021). As pa o ou esea ch me hodology, we adop ed
he Sys ema ic Li e a u e Re iew (SLR) app oach o iden i y, anal-
yse, and in es iga e he a ailable li e a u e based on he ou -
lined esea ch ques ions. Speci ically, SLR ollows he p inciple
o e idence-based so wa e enginee ing app oach o adop a ig-
o ous p ocess o conduc ing he e iew based on well-de ined
p o ocol o ex ac , analyse, and epo he esul s (Ki chenham
e al.,2004). SLR p o ides ‘‘a means o e alua ing and in e p e ing
all a ailable esea ch ele an o a pa icula esea ch ques ion, opic
a ea, o phenomenon o in e es ’’,keele2007guidelines. We ollowed
he guidelines p o ided by Ki chenham and Cha e s o conduc
his SLR (Ki chenham and Cha e s,2007), which consis s o h ee
co e s eps, i.e., planning, conduc ing, and epo ing he e iew as
illus a ed in Fig. 2.
Each s ep o SLR, as illus a ed in Fig. 2, is elabo a ed below.
While pe o ming he li e a y e iews and seconda y s udies in
he con ex o so wa e enginee ing esea ch, he e is an ongo-
ing deba e abou conduc ing he Mul i ocal Li e a u e Re iews
(MLRs) – including g ey li e a u e – ins ead o SLRs o as
e ol ing a eas like quan um compu ing and quan um so wa e
enginee ing (Ga ousi e al.,2019). We p e e ed he SLR, based
on he guidelines in Ki chenham and Cha e s (2007), o only
e iew pee - e iewed published esea ch as seconda y s udies on
quan um so wa e a chi ec ing. Non-pee e iewed s udies and
g ey li e a u e a e also discussed o discuss he esul s o p ima y
s udies, howe e ; such s udies and li e a u e a e complemen a y
and a e no included in he lis o p ima y s udies o SLR.
5
A.A. Khan, A. Ahmad, M. Waseem e al. The Jou nal o Sys ems & So wa e 201 (2023) 111682
Table 1
Resea ch ques ions o his SLR.
A: Demog aphic de ails o published esea ch
# Resea ch ques ion Ra ionale
RQ1.1 Wha a e he ypes and he equency o
publica ions on quan um so wa e
a chi ec u e?
This RQ aims o pinpoin he ypes o publica ions (e.g., jou nal a icles,
con e ence p oceedings) and highligh he equency o publica ions (numbe
o publica ions pe yea ). The RQ p o ides an unde s anding o he esea ch
p og ess (i.e., ype and equency published esea ch o e he yea s) wi h
espec o he opic unde in es iga ion.
RQ1.2 Wha a e he esea ch ypes and epo ed
con ibu ions in published s udies on quan um
so wa e a chi ec u e?
Types o esea ch (i.e., solu ion ype, e alua ion ype) and esea ch
con ibu ions help us o unde s and he di e si y o published esea ch,
solu ions o add ess he p oblems, empi ical ounda ions, and heo e ical
p inciples as he a ailable e idence in he SLR.
RQ1.3 Wha a e he applica ion domains o which
he p oposed a chi ec u al solu ions can be
applied?
Applica ion domain e e s o he a eas (e.g., ne wo k secu i y, sys em
enginee ing) o which a chi ec u al solu ions can be applied o add ess
speci ic challenges. A classi ica ion o applica ion domains help us unde s and
he ex en o which a chi ec u al solu ions add ess so wa e design challenges
pe aining o di e en a eas.
A chi ec u al solu ions o quan um so wa e and eme ging challenges
# Resea ch ques ion Ra ionale
RQ2.1 A e he e any a chi ec u al p ocesses o
quan um so wa e?
A chi ec u al p ocess include a numbe o a chi ec ing ac i i ies o p o ide a
s ep-wise and inc emen al app oach o de elop a chi ec u al solu ions. By
in es iga ing he a chi ec u al p ocess and i s unde lying a chi ec ing
ac i i ies, we can unde s and a chi ec u al analysis, syn hesis, and e alua ion
o p oposed solu ions.
RQ2.2 Wha modelling no a ions ha e been used o
ep esen quan um so wa e a chi ec u al
solu ions?
Modelling no a ions isually depic he de ail sequence o a chi ec ing
ac i i ies and show he ela ions be ween he nume ous uni s o he so wa e
sys em. Answe o his RQ will gi e an unde s anding o exis ing g aphical
no a ion used o speci y quan um so wa e a chi ec u e.
RQ2.3 Wha pa e ns exis o quan um so wa e
a chi ec u es?
Pa e ns ep esen eusable knowledge and bes p ac ices o design and
implemen so wa e solu ions. The answe o his RQ will help o in es iga e
he pa e ns which e eal eusable (a chi ec u al) knowledge and bes
p ac ices o a chi ec quan um so wa e sys ems.
RQ2.4 A e he e any ools and/o amewo ks o
suppo au oma ion and cus omisa ion o
a chi ec u al solu ions o quan um so wa e?
To s udy he a ailable ools and amewo k suppo ha can enable
au oma ion and cus omisa ion (i.e., use decision suppo ) o he a chi ec u al
p ocess and i s ac i i ies. We aim o u he analyse ools ha complemen
he a chi ec u al solu ions wi h hei au oma ion and cus omisa ion.
RQ2.5 Wha challenges ha e been epo ed o
quan um so wa e a chi ec u e?
Va ious challenges could impac he p ocess o de eloping quan um so wa e
a chi ec u e. Analysing he challenges will pinpoin he issues and ac o s ha
impac a chi ec u al solu ions o quan um so wa e.
3.1. Planning he e iew
As he ini ial s ep, he planning phase s a s wi h de elop-
ing he esea ch ques ions ha encapsula e he key esea ch
objec i es o he SLR.
3.1.1. S ep 1: Speci y esea ch ques ions
We ou line he Resea ch Ques ions (RQs) o in es iga e mul i-
ace ed in o ma ion including demog aphy, a chi ec u al ac i i ies,
a chi ec u al modelling no a ions, a chi ec u al design pa e ns, ools
and amewo ks, and challenges. The RQs o in es iga e he men-
ioned mul i- ace ed in o ma ion a e ou lined and he de ails
along a ionale o each RQ is p o ided in Table 1. Answe o
he epo ed RQs helps us documen he SLR esul s desc ibed in
subsequen sec ions o his pape .
3.1.2. S ep 2: Iden i y da a sou ces
In sys ema ic e iews and mapping s udies, Elec onic Da a
Sou ces (EDS) allow an au oma ed sea ch, based on p ede ined
and o en cus omised sea ch s ing(s), o iden i y he ele an
li e a u e on a opic unde in es iga ion (Chen e al.,2010). A
numbe o empi ical s udies ha e in es iga ed me hods o con-
duc ing sys ema ic sea ches along wi h pu ing o wa d a lis o
EDS ha can help selec li e a u e e icien ly while minimising
he po en ial bias and isk o missing ele an da a (Mou ão e al.,
2017). Based on he ecommenda ions o adop ing a sys ema ic
sea ch p ocess and selec ing he mos ele an da a sou ce, we se-
lec ed i e EDS o an au oma ed sea ch (Zhang e al.,2011). These
EDS include ACM Digi al Lib a y, IEEE Xplo e, Science Di ec ,
Sp inge Link, and Wiley Online Lib a y ha ep esen p ominen
sou ces o sea ch li e a u e on compu ing in gene al and so wa e
enginee ing and so wa e a chi ec u e esea ch in pa icula . The
lis o EDS ha we selec ed is no an exhaus i e, no does i
gua an ee o co e all possible exis ing li e a u e, howe e , p io
empi ically-based s udies on SLRs ha e highligh ed hese i e
elec onic sou ces as necessa ily su icien and app op ia e o
iden i y he ele an li e a u e (Chen e al.,2010;Zhang e al.,
2011).
3.1.3. S ep 3: Fo mula e sea ch s a egy
The i s h ee au ho s analysed he RQs o iden i y he key
e ms o keywo ds. Mo eo e , all he au ho s we e in i ed o
pa icipa e in he g oup mee ing o inalise he key e ms. The aim
o epo ing he key esea ch e ms is o de elop he sea ch s ing
and explo e he selec ed digi al lib a ies using ha s ing. Finally,
he au ho s ag eed o conside he ollowing sea ch s ing o he
da a sea ch: (So wa e) AND (A chi ec u e OR Design OR F amewo k
OR Pa e n) AND (Quan um)
The key e ms a e conca ena ed using he ‘‘OR’’ and ‘‘AND’’
boolean ope a o s o de elop he abo e-gi en sea ch s ing. The
decision o inalise he sea ch s ing was based on a pilo sea ch
o ele an li e a u e on IEEE eXplo e and Google Schola . In he
pilo sea ch, we aimed a iden i ying he i les o exis ing s udies
and a ious synonyms used o e e o so wa e a chi ec u e in
he quan um compu ing con ex . Fo example, we obse ed ha
use o key e m ’model’ as a synonym o a chi ec u e yielded a
6
A.A. Khan, A. Ahmad, M. Waseem e al. The Jou nal o Sys ems & So wa e 201 (2023) 111682
Table 2
Inclusion and exclusion c i e ia.
Code Inclusion c i e ia Code Exclusion c i e ia
Inc1 S udies ha speci ically ocus on so wa e
a chi ec u e s udies in quan um compu ing
domain
Excl1 Exclude g ey li e a u e and duplica e s udies.
Inc2 Pee - e iewed published esea ch (e.g.,
con e ence p oceedings, jou nal a icles,
wo kshop/symposium pape s
Excl2 I mul iple s udies a e published in he same
p ojec , hen conside he one wi h maximum
con ibu ion.
Inc3 Pee - e iewed s udies a ailable in ull- ex . Excl3 Exclude s udies ha do no model o desc ibe
s uc u e and/o beha iou o quan um so wa e.
Inc4 Repo ed in English language. Excl4 Exclude he s udies ha do no discuss any o
he so wa e a chi ec u al aspec s as ou lined in
he RQs (e.g., p ocess, pa e ns, no a ions, ools)
signi ican ly la ge bu i ele an numbe o s udies ha discuss
so wa e p ocess models ( ocused on QSE a he han QSA). Based
on he consensus o he esea che s, we omi ed he key e m
’model’ o a oid an exhaus i e sea ch space. Mo eo e , based on
he pilo sea ching phase, we included key e m ’ amewo k’ ha
did iden i y some ele an s udies. The main goal o he inal
sea ch s ing was o iden i y he mos ele an li e a u e as much
as possible while a oiding po en ially i ele an s udies ha can
exhaus manual scanning o i les, keywo ds, and abs ac o
s udy selec ion. The eplica ion package based on he gi en sea ch
s ing is p o ided in Khan e al. (2022a).
3.1.4. S ep 4: De ine inclusion and exclusion c i e ia
Based on he guidelines by Ki chenham and Cha e s (2007)
o including o excluding he iden i ied s udies, we ou lined
he inclusion and exclusion c i e ia in Table 2. By ollowing he
c i e ia any i ele an , edundan , o non-English s udies we e
excluded. S udy inclusion and exclusion was ollowed by a quali y
assessmen s ep o assess he quali y o each included s udy and
elimina e any s udy ha did no sa is y he quali a i e assessmen
c i e ia (see Sec ion 3.2.2). The inclusion and exclusion c i e ia
il e s he sea ch indings e u ned by he sea ch s ing. The key
poin s o he c i e ia we e de eloped by he i s h ee au ho s
based on Ki chenham and Cha e s (2007). Table 2 p o ides he
c i e ia o he inclusion and exclusion o he li e a u e o e iew
along wi h he codes (Incl 1–4: as he inclusion c i e ia and Excl
1–4: as he exclusion c i e ia). We discuss he de ails in Table 2
la e o elabo a e he selec ion o p ima y s udies o be included
in he SLR.
3.2. Conduc ing he e iew
The second phase o he SLR p ocess is conduc ing he e iew,
which is based on he p o ocol de ined in he i s phase, i.e., plan-
ning he e iew (See Fig. 2). Following a e he key s eps in ol ed
in his phase:
3.2.1. S ep-1: Selec p ima y s udies
P ima y s udies sea ch p ocess s a ed wi h explo ing he
selec ed digi al eposi o ies using he sea ch s ing discussed
in Sec ion 3.1.3. The sea ch p ocess was ini ia ed on 30 h Sep em-
be 2021 and ended on 9 h Oc obe 2021. Ini ially, he sea ch
s ing e u ned a o al o 8,406 s udies, which a e u he il e ed
by he i s h ee au ho s based on he s udies i les, keywo ds,
and abs ac s agains he inclusion and exclusion c i e ia (see
Fig. 3). The second phase sc eening e u ned a o al o 589 s udies.
The hi d phase o inclusion and exclusion sc eening was pe -
o med based on he ull- ex e iew o he s udies, whe e 32
p ima y s udies we e inally selec ed (see Fig. 3). Addi ionally,
he ou h and i h au ho s we e in i ed o con i m he sea ch
indings and lis o selec ed s udies.
Fo example, we used he ad anced sea ch op ion o IEEE
Xplo e (‘Sea ch Te m’) o execu e he sea ch s ing o iden-
i y published s udies (in ‘Full Tex & Me aDa a). The sea ch
yielded a o al o 32115 s udies, majo i y o which ocused on
quan um sys ems in gene al and quan um ha dwa e in pa ic-
ula . While ying o elimina e an exhaus i e lis o i ele an
s udies, we in e changed he sea ch pa ame e ( om ‘in Full
Tex & Me aDa a’ o ‘in Abs ac ’) and ound 397 s udies ha
missed some ele an s udies ha we e disco e ed be o e he
sea ch pa ame e in e change. The e o e, we decided o manually
scan h ough he 32115 s udies a e we applied u he digi al
lib a y-speci ic il e ing o elimina e sea ch esul s classi ied un-
de ‘S anda ds’,‘Books’ and o he alike ca ego ies o ge a o al
o 1751 candida e s udies om IEEE eXplo e. Based on a simila
app oach, o en digi al lib a y-speci ic il e ing, we ex ac ed and
iden i ied he candida e s udies o p oceed wi h hei sc eening,
inclusion/exclusion, and quali a i e assessmen , as in Fig. 3.
Mo eo e , he backwa d snowballing app oach was used o
manually sea ch he e e ences lis o he selec ed 32 p ima y
s udies o iden i y addi ional s udies ha migh ha e been missed
du ing he sea ch s ing-based e iew p ocess (Wohlin,2014).
The backwa d snowballing e en ually e u ned wo mo e s ud-
ies ha explici ly ul illed he inclusion and exclusion c i e ia.
The snowballing p ocess was mainly pe o med by he i s and
second au ho s. Addi ionally, he hi d and ou h au ho s we e
in i ed o mu ually e i y he indings epo ed by he i s and
second au ho s. To include s udies ha discuss so wa e a chi-
ec u e, we speci ically looked o a chi ec u al models (g aphical
no a ions, e.g., UML diag ams) o a chi ec u al speci ica ions (de-
sc ip i e no a ions, e.g., ADLs) ha ep esen he s uc u e o
beha iou o so wa e sys em (Med ido ic and Taylo ,2000;
Med ido ic e al.,2002). Finally, (32+2) s udies a e sho lis ed
(see Fig. 3) o e iew, analyse and add ess he esea ch ques-
ions based on hei indings. The selec ed p ima y s udies lis
is p o ided in Appendix (Table 11). Fu he mo e, We included
se e al non-pee - e iewed s udies a ailable on he a Xi open-
access eposi o y (Zhao,2020;Nguyen e al.,2022;Fahmideh
e al.,2021a,b;G ae and Geo gie ski,2021;Moin e al.,2021;
Pé ez-Cas illo e al.,2021;Dey e al.,2020;Mi anskyy e al.,2022;
Wang e al.,2022;Paulo and de Cama go,2021;Khan e al.,
2022b) o complemen he s udy’s o e all indings. Howe e , we
did no include hem in he p ima y s udies lis ( Appendix –
Table 11) as pe he guidelines o SLR (Ki chenham e al.,2004)
app oach. In addi ion o ollowing he guidelines o he SLRs o
li e a u e inclusion (Ki chenham e al.,2004), ou decision was
also mo i a ed by he ac ha p ep in s a e o en subjec o
changes o e ime, wi h se e al e sions ha ing he same i le bu
di e ing con en . Gi en he as -paced esea ch ields o quan um
so wa e enginee ing/a chi ec u e, p ep in s may con ain e o s
o changes ha could comp omise he eliabili y o ou SLR’s
esul s and h ea en i s in e nal alidi y. The e o e, we excluded
p ep in s om ou lis o selec ed p ima y s udies o minimise he
7
A.A. Khan, A. Ahmad, M. Waseem e al. The Jou nal o Sys ems & So wa e 201 (2023) 111682
Fig. 3. S udies selec ion p ocess.
men ioned isk. Fo ins ance, one p ep in changed (Abbo e al.,
2018) i s con en ou imes wi hin wo yea s, highligh ing he
need o cau ion when inco po a ing p ep in s in o a sys ema ic
e iew.
In addi ion o ou adop ed app oach o au oma ed sea ch in
elec onic da abases and backwa d snowballing o iden i y he
ele an s udies, se e al o he app oaches could be used. Some
o hese app oaches include bu a e no limi ed o sea ching
indi idual publica ion enues (e.g., con e ence p oceedings, jou -
nal olumes), esea ch g oup publica ions, and o wa d snow-
balling (Feliza do e al.,2016). Speci ically, o wa d snowballing
– sea ching o s udies ha ci e he s udies con ained in he seed
se – is ound o be mo e use ul in upda ing o ex ending an
al eady conduc ed seconda y s udy bu is s ill p one o missing
ele an li e a u e. Jalali and Wohlin (2012) in es iga es he ap-
plica ion o snowballing app oaches in SLRs and sugges s ha
simila i y in iden i ied li e a u e is expec ed o inc ease i bo h
he backwa d and o wa d snowballing a e pe o med since he
o e lap in he included pape s would be g ea e . This in luenced
ou decision o a oid o wa d snowballing, howe e ; u u e ex-
ensions o his SLR can bene i om o wa d snowballing wi h
an upda ed seed lis o s udies (Feliza do e al.,2016).
3.2.2. S ep-2: Pe o m quali y assessmen (QAs)
The quali y o he selec ed s udies is e alua ed based on he
quali y assessmen c i e ia ha aim o emo e he esea ch bias
and e alua e he deg ee o signi icance and comple eness o
he selec ed s udies (Ki chenham and Cha e s,2007). The qual-
i y assessmen guidelines p o ided by Ki chenham and Cha e s
a e ollowed o de elop he assessmen c i e ia (see Table 3)
(Ki chenham and Cha e s,2007). The c i e ia consis o i e
assessmen ques ions, and each selec ed p ima y s udy assessed
agains hese ques ions (QAs1-QAs5). Assigned sco e (1) i he
p ima y s udy explici ly add essed he QAs ques ions and (0.5)
poin s i he ques ions a e pa ially add essed. Simila ly, s udies
wi h no e idence o conside ing he assessmen ques ions a e
gi en 0 poin . The inal quali y assessmen sco e o each p ima y
s udy is he sum o he sco e assigned agains each QAs ques ion.
The i s au ho applied he assessmen c i e ia and he esul s
we e u he independen ly e i ied by second and hi d au ho s.
We include hose s udies in he inal lis which had accumula i e
QAs sco e g ea e han o equal o 1.5 (Waseem e al.,2020). The
accumula i e inal sco e o each p ima y s udy agains he QAs
ques ions is gi en in Appendix (Table 11).
3.2.3. S ep-3: Pe o m da a ex ac ion
We de ined a se o da a ex ac ion i ems (see Table 4) o
add ess he RQs o mula ed in Sec ion 3.1.1. Da a i ems a e he
pa icula ypes o da a ex ac ed om each selec ed p ima y
s udy ha di ec ly map o he s udy RQs. The i s au ho pe -
o med he pilo da a ex ac ion p ocess o en s udies o e al-
ua e he eliabili y o he ex ac ed da a i ems. The second and
hi d au ho assessed he pilo s udy indings, and based on hei
sugges ions, he i s au ho e ised he da a ex ac ion i ems.
The o mal da a ex ac ion p ocess was pe o med by he i s
h ee au ho s by equally dis ibu ing he o al numbe o selec ed
p ima y s udies, and he s udies dis ibu ion was done based on
he au ho s’ esea ch expe ise and in e es . The gene al (demo-
g aphic) de ails o each selec ed p ima y s udy we e ex ac ed
agains he da a i ems (DI1-DI4), and he es (DI5-DI13) a e
speci ic o he s udy RQs.
We inally conduc ed he Cohen’s Kappa es o check in e -
pe sonal bias in he p ima y s udies selec ion (Sec ion 3.2.1),
quali y assessmen (Sec ion 3.2.2), and da a ex ac ion
(Sec ion 3.2.3) phases. Mainly, he i s h ee au ho s we e in-
ol ed in he s udies selec ion, quali y assessmen , and da a
ex ac ion p ocess. To emo e he in e -pe sonal bias o he
men ioned phases o he SLR p ocess, we in i ed he emaining
au ho s and me ged hem ac oss wo di e en g oups (au ho s 4–
5, au ho s 6–7). They we e asked o andomly selec a se o en
p ima y s udies and sequen ially pe o m he s udies selec ion,
quali y assessmen , and da a ex ac ion p ocess as pe o med by
au ho s 1–3. E en ually, he Cohen’s Kappa es was pe o med
o measu e he ag eemen le el and iden i y he signi ican di -
e ences ac oss he men ioned phases be ween all he h ee
g oups o au ho s (au ho s 1–3, au ho s 4–5, au ho s 6–7). The
Cohen’s Kappa es is widely adop ed in EBSE esea ch (Pé ez
e al.,2020). Cohen’s kappa coe icien (k) is he p opo ion o
chance-expec ed disag eemen s which do no occu , o al e na i ely,
i is he p opo ion o ag eemen a e chance he ag eemen is e-
mo ed om conside a ion,cohen1960coe icien . The (k) coe icien
measu es he le el o ag eemen be ween a g oup o a e s ha
e alua e N-objec s in o (c) mu ually exclusi e ca ego ies (Cohen,
1960). The ag eemen le el be ween he a e s equals chance
ag eemen when Cohen’s kappa coe icien alue (k)=0. The le el
o ag eemen is posi i e when (k) is g ea e han he chance
ag eemen and nega i e i i is less han i . The pe ec ag eemen
occu s be ween a g oup o a e s when he alue o k anges om
8
A.A. Khan, A. Ahmad, M. Waseem e al. The Jou nal o Sys ems & So wa e 201 (2023) 111682
Table 5
Summa y iew o modelling no a ions, modelling a i ac s, and li ecycle suppo . (AR = A chi ec u al Requi emen s, AD = A chi ec u al
Design, AI = A chi ec u al Implemen a ion, AE = A chi ec u al E alua ion, AT= A chi ec u al Deploymen ).
S udy ID Modelling no a ion Modelling a i ac P ocess suppo
AR AD AI AE AD
S1 Box and a ows Componen diag am ✓
S2 G aph-based model S a e g aph ✓ ✓
S4 UML Class diag am ✓ ✓ ✓
S5 G aph-based model P ocess low model ✓ ✓
S6 Box and a ows S a e ansi ion diag am ✓ ✓
S7 G aph-based model S a e g aph ✓
S9 UML S a e ansi ion diag am ✓ ✓
S14 Box and a ows Componen diag am ✓ ✓
S21 Box and a ows
G aph-based model
S a e g aph ✓ ✓
S22 Box and a ows Componen diag am ✓ ✓
S25 G aph-based model S a e g aph ✓ ✓
S27 G aph-based model P ocess low model ✓ ✓ ✓
S28 Box and a ows
G aph-based model
S a e g aph ✓ ✓
S31 G aph-based model P ocess low model ✓
S32 UML (Q-UML) Class diag am
Sequence diag am
✓ ✓
S33 Box and a ows Componen diag am ✓ ✓ ✓
Key Findings o RQ2.1
Finding 7: An a chi ec u e design endea ou o he
quan um so wa e equi es an a chi ec ing p ocess o
inco po a e a numbe o a chi ec ing ac i i ies. Exis ing
a chi ec u al p ocess can be le e aged o suppo i e
a chi ec ing ac i i ies o quan um so wa e namely (i)
a chi ec u al equi emen s, (ii) a chi ec u al modelling, (iii)
a chi ec u al implemen a ion, (i ) a chi ec u al alida ion,
and ( ) a chi ec u al deploymen .
Finding 8: Quan um speci ic equi emen s such as mod-
elling Qubi s o Quga es and co-design o quan um ha d-
wa e and so wa e equi es domain speci ic modelling
and ans o ma ion o be suppo ed by a chi ec u al
p ocess ac i i ies.
5.2. A chi ec u al modelling no a ions (RQ2.2)
We now answe RQ2.2 ha in es iga es he modelling
no a ions, ep esen ing a mul i ude o g aphical models o de-
sc ip i e no a ions o speci y, documen , o ep esen he a chi-
ec u al models. F om a chi ec u al p ocess pe spec i e (RQ2.1),
he e ms modelling no a ion, modelling language, and a chi ec-
u al language a e i ually synonymous and o en used in e -
changeably all e e ing o same concep o a chi ec u al
ep esen a ion ei he g aphically o ex ually (Mala ol a e al.,
2012;Pé ez-Cas illo e al.,2021). Fo example, o suppo quan-
um modelling languages o speci ying QSAs, Ca los e al. [S32]
ha e de eloped Q-UML - an ex ension o classical UML (Uni ied
Modelling Language) – o suppo s uc u al and beha iou al
ep esen a ion o quan um sea ch algo i hms (Med ido ic e al.,
2002). Speci ically, conside ing he (co-) design and implemen-
a ion challenges o QSAs, he ole o a chi ec u al modelling
becomes pi o al o p o ide a so wa e blue-p in model ha ac s
as a b idge be ween a chi ec u al equi emen s and hei imple-
men a ions, as in Fig. 7. A chi ec u al models essen ially becomes
he d i ing a i ac in he con ex o model-d i en a chi ec ing,
whe e a chi ec u al models and model ans o ma ion can be
exploi ed o model-based implemen a ion and alida ion o he
sys em (Moin e al.,2021). To sys ema ically classi y, analyse and
compa e a chi ec u al modelling o desc ip ion languages, some
amewo ks ha e been de eloped ha p o ide a c i e ia-d i en
analysis o a chi ec u al modelling (Mala ol a e al.,2012;Med i-
do ic and Taylo ,2000). These e alua ion c i e ia can be gene ally
classi ied in o h ee main ypes, each ype explo ing he ole o
modelling no a ions o suppo (i) a chi ec u al speci ica ions
(e.g., a chi ec u al ep esen a ion, a chi ec u al s uc u e, syn ax,
and seman ics and, analysing s a ic and dynamic na u e o he a -
chi ec u es), (ii) quali y a ibu es (e.g., ex ension, cus omisa ion,
in e ope abili y o he no a ions), (iii) a chi ec u al p ocess (a -
chi ec u al equi emen s, implemen a ion, alida ion). The ocus
o his RQ is a chi ec u al ep esen a ion, no quali y a ibu es
o modelling no a ions, he e o e, we mainly ocus on aspec s o
a chi ec u al ep esen a ion and suppo o a chi ec u al p ocess
(Fig. 7) wi h he help o Table 5.Table 5 ac s as a s uc u ed
ca alogue o summa ise he ollowing in o ma ion o answe his
ques ion.
A ailable e idence e lec s he published esea ch, ha p o-
ides de ails o he modelling no a ion o QSAs [S32].
Modelling no a ion ep esen s a speci ic me hod o echnique
ha is being used o ep esen he model o QSA. Fo example,
he Q-UML solu ion p o ided by Ca los e al. [S32] is an ex ension
o he UML o s uc u al and beha iou al ep esen a ion o he
QSA. In addi ion o he ex ensions o al eady exis ing modelling
no a ions (i.e., QSA speci ic ailo ing), con en ional no a ions such
as g aph-based models o box and a ow s uc u es ha e been ex-
ploi ed o speci y he s uc u e and seman ics o QSAs [S16][S27].
Fo example, Killo an e al. [S27] exploi s g aph-based models o
ep esen modules o code o implemen he quan um so wa e.
Speci ically, in g aph-based modelling he modules o sou ce code
a e ep esen ed as g aph nodes (compu a ional elemen s and
da a s o es), whe eas g aph edges ep esen he in e connec ion
he code modules. This means ha T ansac ionCommi module
(node_1) ans e s con ol o Upda e T ansac ionReco d module
(node_2) ia commi connec o (edge_A) in a chi ec u al g aph
o quan um so wa e.
Modelling a i ac ep esen s a speci ic a i ac (i.e., isual
diag am, model) o ep esen an ins ance o he a chi ec u al
model. Fo example, Ca los e al. [S32] used UML class diag am
is being used o ep esen he s uc u e, whe eas UML sequence
diag ams a e used o ep esen he beha iou o he quan um
sea ch algo i hm.
A chi ec u al p ocess suppo needs modelling no a ion (and
i s unde lying a i ac s) o suppo speci ic ac i i ies in he a -
chi ec u al p ocess om RQ2.1. Fo example, Q-UML p esen s
class and sequence diag ams o (i) model equi emen s and (ii)
15

A.A. Khan, A. Ahmad, M. Waseem e al. The Jou nal o Sys ems & So wa e 201 (2023) 111682
speci y s uc u al ep esen a ion and execu ion low o he quan-
um sea ch design. The p oposed solu ion Q-UML does no p o-
ide suppo o o he a chi ec ing ac i i ies such as a chi ec u al
implemen a ion o e alua ion.
Table 5 summa ises he co e indings o RQ2.2 o s eamline
mos adop ed modelling no a ions, he a i ac s being used o
model he QSAs, and hei impac s on a chi ec u al p ocess. We
can conclude ha mos p ominen modelling no a ions can be
b oadly classi ied in o h ee main ypes as UML p o iles and
ex ensions such as [S4, S9, S32] (3 s udies), g aph-based models
including [S2, S5, S7, S21, S25, S27, S28, S31] (8 s udies), and box
and a ow no a ions including [S1, S6, S14, S16, S19, S21, S22, S28,
S33] (9 s udies). Some o he mos used s a e ansi ion diag ams,
s a e g aph, and p ocess low models diag am. In he con ex o
a chi ec u al p ocess suppo , exis ing modelling no a ions a e
p ima ily ocused on suppo ing a chi ec u al equi emen s [S5,
S9, S16, S21, S32] (05 s udies), design [S1, S2, S4, S5, S6, S7,
S9, S16, S19, S21, S22, S25, S27, S28, S31, S32, S33] (17 s udies)
and implemen a ion phases [S2, S4, S14, S19, S27, S28, S33] (7
s udies), whe eas he e is much less suppo o li e-cycle ac i i-
ies like a chi ec u al e alua ion [S14, S22, S27] (3 s udies) and
deploymen [S4, S6, S25, S33] (4 s udies). Modelling no a ions
a e undamen al o he c ea ion o a chi ec u al design models
ha p o ide ounda ions o a chi ec u al implemen a ion (Med-
ido ic e al.,2002). In he con ex o his esea ch, models can
acili a e o he a chi ec u al aspec s including bu no limi ed o
design decisions (pa e ns and s yles ha p omo e euse) and
ools ha suppo cus omisa ion, human decision suppo , and
au oma ion, de ailed in subsequen sec ions o his pape .
Key Findings o RQ2.2
Finding 9: Modelling no a ions o speci y quan um so -
wa e a chi ec u es p ima ily ely on box and a ow
no a ions (ha ing componen diag ams) and g aph-based
models (ha ing s a e g aph) o ep esen he s uc u es
and beha iou o quan um so wa e unde design. Unlike
con en ional so wa e a chi ec u es ha mos ly exploi
UML no a ions (o en conside ed as a de ac o app oach
o so wa e design), he e is much less e idence on
UML-based modelling quan um so wa e a chi ec u es
Finding 10: I appea s ha he e is a need o a chi ec-
u al desc ip ion languages and UML p o iles ha can be
help ul o le e age exis ing ools, amewo ks, and a -
chi ec u al knowledge o empowe he ole o designe s
and a chi ec s o model, de elop, and e ol e quan um
so wa e based on e-usabili y and (semi-) au oma ion.
5.3. A chi ec u e design pa e ns (RQ2.3)
To answe RQ2.3, we iden i ied a o al o six quan um so -
wa e a chi ec u e pa e ns discussed in (n=17, 50%) s udies.
In design o a chi ec u al con ex , pa e ns ep esen eusable
design knowledge, e e ed o as bes p ac ices and concen a ed
wisdom o designe s o add ess ecu ing challenges o so wa e
de elopmen . Fo example, o add ess he challenges o sys-
em s uc u ing and deploymen he laye ed a chi ec u e pa e n
helps a chi ec s o o ganise so wa e-in ensi e sys ems and ap-
plica ions in o a ious laye s, each dedica ed o di e en conce ns
such as da a managemen , use in e acing and compu a ions [S1,
S18]. A collec ion o pa e ns o mally o in o mally o ganised
in o a sequence, esul s in a chi ec u al pa e n languages (Ley-
mann,2019). The ocus o his s udy is indi idual pa e ns a he
han pa e n languages. The se o iden i ied quan um so wa e
a chi ec u e pa e ns is p esen ed in Table 6. The mos ecu ing
Table 6
Quan um so wa e a chi ec u e design pa e ns.
Pa e n name S udy IDs
Laye ed pa e n S3, S5, S9, S14, S18, S26, S28, S29
Pipe and il e pa e n S2, S20, S21, S27, S31
Composi e design pa e n S4
P o o ype design pa e n S24
Recu si e con ainmen S9
Two-qubi ga e pa e n S20
design pa e ns discussed in he 18 p ima y s udies a e laye ed (n
= 8, 24%) and pipe and il e a chi ec u e (n = 5, 15%) pa e ns. The
o he pa e ns ha ing low equency o occu ence a e (compos-
i e design, p o o ype design, ecu si e con ainmen and wo-qubi
ga e). In he ollowing ex , we b ie ly desc ibe he example o
alaye ed pa e n o he gene al-pu pose mic oa chi ec u e o
quan um so wa e [S3]. Gene ally, he laye ed pa e n a chi ec u e
o quan um so wa e mainly consis s o se e al p ope ies ha we
also need o es ima e. These p ope ies include app op ia e in-
s uc ion leng h, pipeline dep h ( o pa allel quan um ga es), and
mul iple con ol channels pe single ins uc ion. These p ope ies
help o cons uc he basic blocks o quan um so wa e, such as
he iming con ol uni and he mic ocode ins uc ion se o he
o e all sys em. Acco ding o ou esul s, he second mos e-
quen ly epo ed pa e n used o designing quan um so wa e is
pipe and il e . Killo an e al. [S27] p oposed an open-sou ce quan-
um p og amming a chi ec u e (i.e., S awbe y Fields) based on
pipe and il e pa e ns. The elemen s o he p oposed a chi ec-
u e a e o ganised as he on -end and he back-end. The on -
end laye consis s o in e ac i e se e , applica ion, ield API, and
quan um p og amming language componen s, and he back-end
componen s include a quan um p ocesso and simula o . Bo h
laye s communica e h ough he compile engine. Ou esul s in-
dica e ha he pa e ns o quan um so wa e a e simila o o he
ypes o so wa e (e.g., monoli hic based a chi ec u e, se ices-
o ien ed based a chi ec u e, mic ose ices-based a chi ec u e).
Howe e , hese pa e ns deal wi h a se ies o ins uc ions ha
need o be execu ed on quan um p ocesso s.
Key Findings o RQ2.3
Finding 11:Laye ed and pipe and il e pa e ns a e
iden i ied as he mos ecu ing quan um so wa e a chi-
ec u e pa e ns. Howe e , hese a e gene ic o classical
pa e ns ha can be used o design any so wa e sys-
em. To his end, u he esea ch e o s a e equi ed o
explo e and p opose new pa e ns o pa icula ly ocus
on quan um compu ing a ibu es (e.g. supe posi ion and
quan um en anglemen ) and acili a e he a chi ec u e o
quan um so wa e sys ems.
5.4. A chi ec u e ools and amewo ks (RQ2.4)
RQ2.4 is de eloped o iden i y ools and amewo ks used
o suppo he a chi ec ing ac i i ies discussed in Sec ion 5.1.
We explo ed he selec ed p ima y s udies and no iced ha only
(n=11, 32%) s udies discussed a chi ec u al ools and ame-
wo ks (see Table 7). The ools and amewo ks p o ide semi- o
ully au oma ed solu ions o pe o m a chi ec ing ac i i ies. Tools
b oadly e e o so wa e solu ions ha au oma e, enhance, o
cus omise p ocess ac i i ies. On he o he hand, a amewo k is a
se o ools used o pe o m a bunch o ac i i ies, e.g., designing,
implemen a ion, and documen a ion. Each iden i ied ool and
amewo k is in e p e ed based on he ollowing i e c i e ia (Saj-
jad e al.,2018), as lis ed in Table 7). Sou ce ype e e s o he
16
A.A. Khan, A. Ahmad, M. Waseem e al. The Jou nal o Sys ems & So wa e 201 (2023) 111682
Table 7
Lis o iden i ied ools.
Tool/F amewo k Sou ce ype Inpu ins uc ions Ou pu Au oma ion le el E alua ion S udy
XACC (eX eme-scale Accele a o ) CS HL QSC FA EX S4
Link laye CS QI SF FA EX S6
Au o E/E amewo k OS MV SF SA IM S9
eQASM CS QI QA SA EX S14
JKQ ( ool se ) OS HL SF FA EX S16
Kwan OS MV SF FA EX S17
JKQ DDSIM OS HL SF FA EX S19
QuNe Sim OS HL SF SA IM S25
S awbe y ields OS HL SF FA EX S27
qco OS HL SF FA EX S28
GH-QPL CS HL QSC SA IM S33
ype as open sou ce (OS) o close sou ce (CS). In open sou ce,
he copy igh holde s g an he use pe missions o s udy, use o
upda e he ool, amewo k o sys em. Inpu ins uc ions a e he
ins uc ions p o ided o execu e he logic. The ins uc ion ypes
a e ca ego ised as high-le el (HL), quan um ins uc ion (QI), and
ma hema ical a iables (MV). Ou pu a e he ype o pos exe-
cu ion indings and ca ego ised as quan um sou ce code (QSC),
quan um algo i hm (QA), and simula ion indings (SF). Au oma-
ion le el e e s o he au oma ion le el o he ool o amewo k.
Au oma ion could be ully-au oma ed (FA), semi-au oma ed (SA),
o non-au oma ed (NA). E alua ion e e s o he pe o mance
assessmen o a pa icula ool and amewo k. E alua ion could
be explici (EX) o implici (IM). Implici means ha ool o
amewo k is pa ially e alua ed o ew o he componen s a e
empi ically assessed.
The esul s gi en in Table 7 e eal ha (n=7, 64%) ools and
amewo ks a e open sou ce (OS). Simila ly, (n=7, 64%) ools and
amewo ks accep inpu code in high-le el (HL) p og amming
o ma (i.e ins uc ions ha a e mo e o less independen o a
speci ic ype o compu e ). Mo eo e , (n=8, 73%) ools and
amewo ks simula e he high-le el inpu ins uc ions and gi e
he ou pu based on he simula ion indings (SF). We u he no-
iced ha (n=7, 64%) ools and amewo ks a e ully-au oma ed
(FA) and (n=8, 73%) a e explici ly (EX) e alua ed based on hei
pe o mance. The isualisa ion and summa y o he esul s on
ool suppo a e p o ided in Fig. 8 and Table 8.
Finally, he iden i ied ools and amewo ks a e classi ied wi h
espec o hei con ibu ion ac oss he a chi ec u al p ocess ac-
i i ies epo ed in Sec ion 5.1. Thema ic analysis app oach dis-
cussed in Sec ion 3.2.4 is ollowed o ca ego ise he iden i ied
ools and amewo ks and p esen he oolchain. I should be
no ed ha a speci ic ool o amewo k migh con ibu e o mo e
han one a chi ec ing ac i i ies and we conside hem ac oss
mul iple ac i i ies (see Table 8).
The co e a chi ec ing ac i i ies wi h espec o he ools and
amewo ks suppo a e subsequen ly discussed:
A chi ec u al equi emen s: We explo ed he selec ed p ima y
s udies and iden i ied a single amewo k ha ocuses on a chi-
ec u al equi emen s (see Table 8) [S9]. Lan e al. [S9], p oposed a
quan um compu ing based a chi ec u al amewo k o minimise
he gap be ween he unc ional domains and mee he equi e-
men s o he open elec ical and elec onic au omo i e embedded
sys ems. A chi ec u al equi emen s is a less ocus ac i i y wi h
espec o ools and amewo ks and he eason migh be ha
quan um so wa e a chi ec u e ield is in he e olu ion phase and
s ill he a chi ec u al equi emen s ac i i ies do no ha e ool
based au oma ion and cus omisa ion suppo .
A chi ec u al implemen a ion: We iden i ied ha a o al o six
ools and amewo ks con ibu ed o he a chi ec u al implemen-
a ion ac i i y (see Table 8). Mo e na ow, hese ools and ame-
wo ks explici ly ocus on he code compila ion and design o code
ans o ma ion sub-ac i i ies (see Fig. 8). The powe o quan um
compu e could only be ealised by implemen ing quan um algo-
i hms o con ol he ha dwa e de ices, imp o e he pe o mance
and e i y he quan um a ibu es (Magnani,2022). The e o e,
esea che s and p ac i ione s a e ushing o de elop s a egies,
ools, amewo ks and guidelines o implemen algo i hms in
a simple and e icien way. Fo example, XACC (eX eme-scale
ACCele a o ) p o ides in e aces o enhance hyb id compila ion
o p og ammes de eloped bo h in quan um and classical p o-
g amming languages [S4]. XACC p og amming amewo k is de-
signed in a manne ha i is en i ely independen o selec ed
language, compu a ional model and ha dwa e. The implemen a-
ion ools and amewo ks ins an ly assis in ealising he eal-
wo ld compu a ion bene i s o quan um compu e s and inc ease
i s applica ion ac oss a ious indus ial domains.
A chi ec u al modelling: We no iced ha only wo a chi ec u al
modelling ools and amewo ks a e de eloped, which explic-
i ly add ess design model and a chi ec u e model sub-ac i i ies
[S27, S28] (see Fig. 8). Modelling ac i i ies pe o med o de elop
he o e all a chi ec u e, which ac s as a bluep in o he im-
plemen a ion. The quan um so wa e enginee ing ield is s ill
unde eloped, and i is impo an o c ea e high-le el modelling
abs ac ions o classical so wa e enginee s o unde s and and
model he quan um p og ammes. Fo example, S awbe y Fields
is an open sou ce a chi ec u al amewo k de eloped o design
and op imise he so wa e sys ems o pho onic quan um com-
pu e s [S27]. S awbe y Fields has buil -in engine o con e
he code de eloped in domain speci ic p og amming language
(blackbi d) and un using he pho onic quan um compu e s.
A chi ec u al deploymen : Finally, we no iced ha only one
amewo k ocuses on deploymen ac i i ies i.e., link laye [S6].
I is de eloped o quan um communica ion ha imp o es he
en anglemen a ibu es be ween quan um compu e s in o obus
and well de ined se ices. Addi ionally, s a egies o ne wo k
scheduling a e de eloped o e alua e he p o ocol pe o mance
wi h espec o di e en use cases. A chi ec u al deploymen is
a sligh ly less ocused ac i i y and in nea e m he ools o
au oma e he deploymen ac i i ies will be demanding need.
Key Findings o RQ2.4
Finding 12: The iden i ied ools and amewo ks a e
ca ego ise based on he i e co e a ibu es namely (i)
sou ce ype, (ii) inpu s, (iii) ou pu s, (i ) au oma ion, and
( ) e alua ion le el.
Finding 13: The iden i ied ools and amewo ks a e
mapped ac oss he a chi ec ing ac i i ies and p esen ed
as a oolchain (see Fig. 8). A chi ec u al implemen a ion
is iden i ied as he mos common ac i i y wi h espec
o ools and amewo ks. We no iced ha six ools and
amewo ks (n=6, 55%) a e de eloped o au oma e and
cus omise he a chi ec u al implemen a ion ac i i ies.
17
A.A. Khan, A. Ahmad, M. Waseem e al. The Jou nal o Sys ems & So wa e 201 (2023) 111682
Fig. 8. Tool suppo o a chi ec ing ac i i ies.
Table 8
Summa y iew o ools and amewo ks ac oss a chi ec ing ac i i ies (AR = A chi ec u al Requi emen s, AM = A chi ec u al Modelling,
AI = A chi ec u al Implemen a ion, AV = A chi ec u al Valida ion, AD= A chi ec u al Deploymen ).
S udy ID Tool name Tool ocus P ocess suppo
AR AM AI AV AD
S4 xACC Code compila ion ✓
S9 Au o.E/E F amewo k Requi emen s ✓
S27 S awbe y ields Domain modelling ✓
S28 qCOR Design ✓ ✓
S14 eQASIM P og amme low and execu ion ✓
S16 JKQ Code compila ion ✓ ✓
S19 JKQ DDSIM Simula ion, compila ion ✓
S33 GH-QPL T ansla ion and compila ion ✓
S6 LinkLaye Quan um communica ion ✓
S17 Kwan Simula ion ✓
S19 JKQ DDSIM Simula ion ✓
S25 QuNe Sim Simula ion ✓
5.5. A chi ec u e challenges
The selec ed p ima y s udies a e explo ed o iden i y he key
challenges o quan um so wa e a chi ec u e (RQ2.5). We ound
ha only (n=16, 47%) p ima y s udies epo ed he a chi ec u e
challenging ac o s. The iden i ied challenges a e u he clas-
si ied ac oss ou co e hemes: quan um da a ansmission and
secu i y,p ocess-cen ic a chi ec ing,a chi ec u al ools and ech-
nological suppo , and a chi ec ing knowledge and expe ise. The
hema ic analysis app oach discussed in Sec ion 3.2.4 is ollowed
o sys ema ically iden i y he mos common hemes o he chal-
lenging ac o s (see Fig. 9). Fo ine-g ained analysis, he main
hemes (co e ca ego ies) and sub- hemes (challenging ac o s) a e
p esen ed in Fig. 9 and explici ly discussed below:
5.5.1. Quan um da a ansmission and secu i y
This heme co e s he challenging ac o s ela ed o he secu-
i y o ne wo k a chi ec u e de eloped o quan um da a ans-
mission. We iden i ied a o al o ou sub- hemes (challenging
ac o s) ela ed o he secu i y o quan um ne wo k a chi ec u e
(see Fig. 9). The iden i ied challenging ac o s a e ho oughly
discussed as ollow:
Quan um key dis ibu ion (QKD): The quan um key dis ibu ion
(QKD) app oach is used o de elop he ul a-secu e ne wo k o
quan um da a ansmission [S1]. QKD in ol es sending he en-
c yp ed da a and dec yp ion keys o e quan um ne wo k in qubi
s a e. Howe e , he exis ing QKD sys ems a e designed o wo k
on he single link quan um ne wo k and becomes challenging o
ope a e ac oss mul iple ne wo ks whe e he sys em design and
18
A.A. Khan, A. Ahmad, M. Waseem e al. The Jou nal o Sys ems & So wa e 201 (2023) 111682
Fig. 9. Thema ic classi ica ion o iden i ied challenges.
p o ocols ge mo e complex [S1,S2]. I is e iden ha he e is a
s ong need o QKD a chi ec u e ha could deploy ac oss mul iple
ne wo ks o ansmi ing secu e quan um da a.
Quan um communica ion a chi ec u e: A chi ec ing a quan-
um ne wo k is challenging wi h espec o communica ion pe -
spec i es. Quan um ne wo k a chi ec u e is dis inc o classical
because o quan um a ibu es including supe posi ion, en angle-
men , and quan um measu emen [S33]. These a ibu es b ings
signi ican cons ain s o design he quan um communica ion
a chi ec u e. In classical communica ion, he da a bi s used o
con ey he message. In con as , he qubi s a e used o ansmi
he da a o e quan um communica ion channel, howe e ; de-
eloping a quan um communica ion a chi ec u e needs a majo
pa adigm shi o conside he cha ac e is ics o quan um me-
chanics [S33]. The open-sou ce communi y should join he e o s
o design and ab ica e he quan um communica ion a chi ec u e
models and in e aces.
Quan um elepo a ion s a egies: Techniques used o ans-
e quan um in o ma ion be ween sende and ecei e is called
quan um elepo a ion. Telepo a ion in science ic ion e e s o
ans e a physical objec om loca ion A o B; howe e , in quan-
um compu ing i is used o ans e he Qubi s. I has pi o al ole
in he con inuing p og ess o quan um communica ion, and quan-
um ne wo ks. Howe e , elepo a ion is a majo challenge in
p esen day quan um compu ing science because o lack o ele-
po a ion p o ocols, s a egies and echniques. Qubi s ansmis-
sion ac oss mul iple nodes and compu a ion in he cloud domain
is only possible by using he quan um elepo a ion s a egies
[S33]. The e is a s ong need o elepo a ion p o ocols and
s a egies ha could eshape he quan um elepo a ion p ocess.
Quan um c yp og aphy: P ac ically, quan um c yp og aphy is
in i s in ancy because o da a ansmission a es and p ocess-
ing limi a ions. These issues a e complica ed and challenging
o ackle as he high-quali y single pho ons o long-dis ance
equi ed low ansmission loss a es. I inc eases he echnolog-
ical cos o quan um c yp og aphy as compa ed o he classical.
Simila ly, de eloping a sha ing in as uc u e o secu e da a en-
c yp ion and dec yp ion is a signi ican challenge o quan um
c yp og aphy [S33]. The e ec i e enc yp ion and dec yp ion solu-
ion is possible by in oducing he in e media e node be ween he
sende and ecei e . P esen ly, ackling quan um c yp og aphy
challenges is complex, and wo ld-leading echnology gian s a e
acing o p opose e ec i e solu ions.
5.5.2. P ocess-cen ic a chi ec ing
This heme is de eloped o ca ego ise he key challenging
ac o s (sub- hemes) ha could impac he design p ocess o
quan um so wa e a chi ec u e. Following is he de ail desc ip-
ion o each selec ed challenge ha co e s he p ocess-cen ic
heme.
A chi ec u al design models: The e is a lack o models o de-
signing quan um so wa e a chi ec u es. The exis ing models a e
simpli ied ex ended e sions o classical modelling app oaches
and do no explici ly co e he quan um p ope ies including
supe posi ion, in e e ence, and en anglemen [S7, S24]. The un-
a ailabili y o pa icula quan um so wa e design models make i
ha d o design he sys em a chi ec u e. The expec a ions o con-
side quan um compu ing as al e na i e o classical inc eased ex-
ponen ially [S7, S24]. Consequen ly, i becomes impo an o p o-
pose igo ous design models in ad ance o a chi ec ing quan um
so wa e sys ems.
A chi ec u al pa e n selec ion: A chi ec u al pa e n is a com-
mon and eusable solu ion o gene ally occu ing a chi ec u al
p oblems. Selec ing an app op ia e a chi ec u e pa e n o a spe-
ci ic quan um p oblem is a challenging ea . The mul i-c i e ia de-
cision making (MCDM) model could be a bes solu ion o choose
a igh pa e n o igh p oblem [S26]. MCDM model p o ides a
pla o m o ackle he commonly occu ed quan um a chi ec u al
p oblems.
Designing scalable quan um so wa e a chi ec u e: Scalable sys-
ems e e o he in o ma ion p ocessing concep whe e a com-
plex sys em could be de eloped using he basic building blocks. In
quan um a chi ec u al scalabili y, he qubi s p ope ies imp o e
o emains consis en when hey a e ex ended ac oss mul i-
qubi s sys ems [S14, S33]. Howe e , a chi ec u al scalabili y also
needs o conside he Qubi s ope a ions wi h speci ic iming, in
ime ins uc ions e ching and p ocessing o ensu e ha desi ed
ope a ions a e accu a ely pe o med [S14]. I is ha d o li e up
he eal-wo ld p omises and sup emacy o quan um compu e s
wi hou a chi ec u al scalabili y [S33].
5.5.3. A chi ec u al ools and echnologies
The ools and echnologies heme is de eloped o classi y
he challenges ela ed o he echnical suppo o a chi ec ing
ac i i ies. In-dep h discussion o hese challenges is p o ided as
ollows:
Noisy Componen s: Cons uc ing a scalable quan um compu e
is challenging due o en i onmen al in e ac ion noise ha could
19
A.A. Khan, A. Ahmad, M. Waseem e al. The Jou nal o Sys ems & So wa e 201 (2023) 111682
des oy i s highly agile componen s [S12]. En i onmen al in e -
ac ion noise gene a ed because o con ol de ices and hea , which
can se iously dis u b he qubi s supe posi ion s a e and cause
compu a ional e o s. The obus s a is ical and ma hema ical
models o es ima e he noise impac can signi ican ly imp o e he
compu a ion p ocess and p o ec he supe posi ion s a e [S12].
Noisy In e media e-Scale Quan um (NISQ) compu e s: I will
ake decades o esea ch o ealise he aul - ole an quan um
compu e o sol ing he wide ange eal-wo ld p oblems [S15].
Howe e , he concep o noisy in e media e-scale quan um
(NISQ) compu e al eady exis s, which con ains i y o a ew
hund ed Qubi s bu is no sma enough o con inuously pe o m
aul - ee compu a ions [S15, S21]. The e m noisy is used be-
cause he p esen day quan um p ocesso s a e no sophis ica ed
enough o cope wi h he en i onmen al impac s, which cause o
lose he quan um cohe ence. Expe imen al in e es is expec ed
and demanded in designing quan um so wa e and ha dwa e
a chi ec u es o p ocess and execu e a la ge numbe o e o -
ee Qubi s. T ansi ion o quan um compu ing o mo e speci ically
adop ing he quan um ha dwa e and quan um compu ing pla -
o ms equi es inancial in es men s as well as human skills o
manage quan um esou ces. The PISQ (Pe ec In e media e Scale
Quan um) enables he de elopmen o new so wa e applica-
ions by de eloping algo i hms and e alua ing hem on quan um
simula o s ha can be execu ed on exis ing compu ing pla -
o ms (Be els e al.,2021). Solu ions like PISQ may no be long
e m solu ions o suppo quan um so wa e, bu such solu ions
allow esea ch and de elopmen o quan um logic ia simula ions
ha can be deployed and execu ed on non-quan um compu ing
pla o ms.
Lack o compu e -aided design (CAD) ools: Compu e -aided
design ools enable he de elopmen , change, and op imisa ion
o he a chi ec u e design p ocess. These ools a e signi ican ly
impo an o de eloping nanoscale quan um so wa e a chi-
ec u es [S19]. Resea ch o au oma e and op imise he design
app oaches o quan um so wa e sys ems is boos ing; howe e ,
he e is a conside able coo dina ion gap be ween he CAD and
quan um compu ing communi y [S19]. Consequen ly, a ious
p oposed CAD ools a e ailed o achie e he co e a chi ec u al
objec i es.
Simula ing quan um ne wo ks a chi ec u e: The quan um in-
e ne is de ined o ansmi quan um da a, which is a ne wo k
a chi ec u e o mul iple de ices and so wa e ools. The concep
o a quan um in e ne is s ill no in p ac ice, and de elopmen
e o s a e being made o shape i p ac ically. To analyse ne wo k
p o ocols, i is impo an o assess hei signi icance using di -
e en simula ion ools [S25]. Howe e , limi ed s udies discussed
such ools o e alua ing quan um ne wo k p o ocols and he e
is a s ong need o ad anced simula ion ools.
A chi ec u al p og amming languages: Quan um a chi ec u al
p og amming language should p o ide all he equi ed abs ac-
ions bo h o quan um physicis s and algo i hm designe s. The
exis ing languages a e no ich enough o conside o u u e
high-numbe Qubi s algo i hms [S29]. They a e s ill unp edic able
o complex quan um p oblems. In he u u e, he a chi ec u al
languages should suppo high-le el abs ac ions o de eloping
and deploying ad ance algo i hms based on quan um supe po-
si ion and en anglemen . Quan um p og amming languages and
amewo ks p o ided by echnology gian s (e.g., Qiski by IBM,
and Q# by Google) enable so wa e de elope s o implemen QSA
as quan um sou ce code ha can be execu ed o implemen ed
on quan um compu ing pla o ms. Howe e , he esul s o an
explo a o y s udy show ha (i) mined quan um sou ce code
eposi o ies a ailable on Gi Hub and (ii) in e iewed quan um
code de elope s sugges s ha beyond he indus y led p ojec s,
adop ion and applicabili y o quan um p og amming in de el-
ope s’ communi y is s ill limi ed (De S e ano e al.,2022). The
s udy also highligh s ha he cu en gene a ion o so wa e de-
elope s, while implemen ing quan um code, ace a mul i ude o
challenges ha ange om quan um p og amme comp ehension
o sou ce code analysis, manipula ion, and es ing (Wang e al.,
2022).
Lack o simula ion ools: The lack o simula ion ools is consid-
e ed a majo ba ie o quan um so wa e a chi ec u e esea ch.
The need o simula ion ools escala es o la ge-scale p ac ical
and eliable measu emen s [S29]. Gene ally, he a chi ec s a e
in e es ed in knowing how as he a chi ec u e wo ks o a
speci ic applica ion, which ypes o ope a ions i can pe o m, and
wha would be he eliabili y le el o i s esul s? These ques-
ions could possibly be answe ed by p oposing pa icula simula-
o s o quan um so wa e a chi ec u e [S29]. OpenQL p o ides
a quan um p og amming language and i s associa ed quan um
compile o de elop and execu e quan um sou ce code. OpenQL
also p oduces quan um assembly code ha is echnology in-
dependen and can be simula ed using QX Quan um Compu e
Simula o (Khammassi e al.,2021).
5.5.4. A chi ec ing knowledge and expe ise
Designing a eal-wo ld quan um so wa e sys em equi e ade-
qua e knowledge and expe ise, which play majo oles o ealise
he quan um so wa e design and de elopmen ac i i ies. This
heme is de eloped o o ganise he co e challenges ela ed o
quan um so wa e knowledge and expe ise. Following is he
de ail discussion o he iden i ied challenges (sub- hemes).
Lack o expe ienced wo k o ce: Building a wo k o ce o de-
signing a so wa e sys em is subs an ially a majo challenge in
quan um compu ing domain. The skills needed o de elop a clas-
sical compu ing sys em a e di e en o quan um [S30]. The e is
a need o speci ic p o essional expe ise (i.e., human oles in
a chi ec u e-cen ic de elopmen p ocess) such as quan um so -
wa e a chi ec s, quan um code de elope s, and quan um domain
enginee s (Khan e al.,2022b). The echnical eam should unde -
s and physics o cha ac e ise he quan um p ope ies o so wa e
sys ems. Such expe ise du ing he quan um so wa e design and
a chi ec ing phase can en ich he a chi ec ing ac i i ies o be e
mee quan um-speci ic equi emen s o he so wa e. Designing
and a chi ec ing quan um so wa e is adically a di e en con-
cep , and i demands skill ul quan um echnical and manage ial
wo k o ce [S30].
Lack o a chi ec u al knowledge: The esea ch ield o unde -
s and quan um mechanics and in eg a e i in compu ing domain
by designing quan um so wa e a chi ec u e is a om being
ma u e. Va ious a chi ec u al solu ions a e p oposed o de elop
a quan um so wa e sys em; howe e , i equi e deep knowledge
o heo y, echnology, and unde s anding o selec and implemen
a sui able solu ion based on he a chi ec u al p oblem [S31].
I is impo an o educa e he quan um so wa e enginee ing
communi y o eshape he a chi ec u e p ocesses, ac i i ies and
p ac ices [S31].
Key Findings o RQ2.5
Finding 14: Following ou co e hemes o he iden i ied
challenges a e de eloped: quan um da a ansmission and
secu i y, p ocess-cen ic a chi ec ing, a chi ec u al ools and
echnologies, and a chi ec ing knowledge.
Finding 15: We obse ed ha mos o he (n=6, 40%)
challenges a e ela ed o he a chi ec u al ools and ech-
nologies heme. The exis ing ools and echnologies a e
no a ad ance le el o ackle he a chi ec u al p oblems
and i cause a ious challenges. This is inline wi h he
inding o de elop a so wa e enginee ing communi y
ha ocuses on de ising ad ance le el ools and ech-
nologies o managing quan um so wa e a chi ec u e
challenges (Zhao,2020).
20

A.A. Khan, A. Ahmad, M. Waseem e al. The Jou nal o Sys ems & So wa e 201 (2023) 111682
Table 9
A Summa y o he key indings o SLR.
Demog aphy o Published Resea ch (RQ-1.1- RQ1.3)
RQ-1.1 –F equency and ypes o publica ions
F equency: Yea s o publica ions = 2004 o 2021 wi h he mos numbe o publica ions
om 2018 – 2021 (21, 62%)
Types: Jou nal a icles (15, 44%), Con e ence p oceedings (13, 38%), Wo kshop pape (4, 12%),
Symposium pape (2, 6%)
RQ-1.2 –Types o published esea ch
Pe sonal expe ience pape s, Philosophical pape s, Opinion pape s,
Valida ion esea ch, P oposal o solu ion, P oposal o solu ion and alida ion esea ch
RQ-1.3 –Applica ion domains o esea ch
Sys ems and ha dwa e enginee ing (20, 59%), So wa e enginee ing (5, 15%),
Sou ce code compila ion (4, 12%),
Ne wo k secu i y (4, 12%), Sma sys ems (1, 3%)
A chi ec u al solu ions o quan um so wa e (RQ-2.1- RQ-2.5)
RQ-2.1 –A chi ec u al p ocess o quan um so wa e
–A chi ec u al equi emen s
–A chi ec u al modelling
–A chi ec u al implemen a ion
–A chi ec u al alida ion
–A chi ec u al deploymen
RQ-2.2 –A chi ec u al modelling no a ions
–G aph-based models
–Box and a ow
–UML
–UML (Q-UML)
Q-2.3 –A chi ec u al pa e ns
Laye ed pa e n, Pipe and il e pa e n, Composi e design pa e n, P o o ype design pa e n,
Recu si e con ainmen , Two-qubi ga e pa e ns
Q-2.4 –A chi ec u al ools and amewo ks
XACC, Link laye , Au o E/E amewo k, S awbe y ields, qCOR, eQASM, JKQ, Kwan ,
JKQ DDSIM,QuNe Sim, GH-QPL
RQ-2.5 –Eme ging challenges o quan um so wa e a chi ec u es
–P ocess-cen ed a chi ec ing
–A chi ec ing knowledge and expe ise
–Quan um da a ansmission and secu i y
–A chi ec u al ools and echnologies
6. Key indings and implica ions o he SLR
We now summa ise he co e indings o he SLR – discussing
key esul s as answe s o all RQs – ha highligh he s a e-o -
esea ch on a chi ec ing quan um so wa e in Sec ion 6.1. We also
discuss he implica ions o he SLR on u u e academic esea ch
in Sec ion 6.2 and i s signi icance along wi h he po en ial ele-
ance o his SLR o indus ial solu ions ha add ess challenges
o quan um so wa e a chi ec ing in Sec ion 6.3.
6.1. Summa y o key indings
A conclusi e summa y o each RQ is p esen ed in Table 9 ha
s uc u es he gene al demog aphic de ails o published esea ch,
answe ing RQ-1.1 o RQ-1.3 and a chi ec u al solu ions o quan-
um so wa e answe ing RQ-2.1 o RQ-2.5. Table 9 can be looked
up o iden i y he co e inding co esponding o a speci ic RQ
quickly. Fo example, a summa y o he answe o RQ-1.3 high-
ligh s ha a chi ec u al solu ions o quan um so wa e can be
applied o se e al domains such as sys ems and ha dwa e engi-
nee ing, so wa e enginee ing, sou ce code compila ion, ne wo k
secu i y, and sma sys ems. Since he yea 2018, a compa a-
i e g ow h in esea ch on QSE and mo e speci ically quan um
so wa e design, a chi ec u e, and implemen a ion can be a -
ibu ed o a numbe o ac o s. Ou s udy iden i ies h ee such
ac o s including (i) a numbe o pionee ing su eys on quan-
um so wa e enginee ing and de elopmen (Zhao,2020;Pia ini
e al.,2021;Gill e al.,2022;Ali e al.,2022) (ii) communi y-
wide ini ia i es wi h dedica ed wo kshops and con e ences o
quan um so wa e (Moguel e al.,2020;Ab eu e al.,2021a;
Ba zen e al.,2021,2022), along wi h he eme gence o quan um
p og amming models and languages (Zhao e al.,2021;Suni a
e al.,2021;De S e ano e al.,2022;Khammassi e al.,2021).
Mo eo e , beyond academic esea ch, he ecen ly g owing in e -
es o exploi quan um compu ing and echnologies in IT indus y
is based on apid ad ances in quan um ha dwa e and quan um
p og amming languages ha suppo QSE ini ia i es in e ms
o de eloping quan um so wa e sys ems and applica ions (Mi-
c oso ,2021;Behe a e al.,2019;Cou land,2017). I is i al
o men ion ha he launch o he Quan um Flagship p ojec in
2018 ( unded by he Eu opean Commission) e lec s egional and
global ambi ions o os e esea ch and de elopmen on quan-
um compu ing echnologies (Anon,2022a). Simila ly, he s udies
[S31, S34] p esen solu ions ha enable so wa e designe s and
a chi ec s o design and implemen quan um so wa e using a -
chi ec u al componen s and connec o s ha can be mapped o
sou ce code modules and in e ac ion be ween he models. Simi-
la ly, Table 9 highligh s he key indings o RQ-2.3 ha o model
and ep esen quan um so wa e a chi ec u es he mos p omi-
nen a chi ec u al no a ions a e g aph-based modules, box and
a ow s uc u es, and Uni ied Modelling Language. Fo example,
he s udy [S32] p esen s a quan um-speci ic UML named Q-UML
ha exploi s class and sequence diag ams o ep esen he be-
ha iou and s uc u e o quan um so wa e sys ems. The de ails
in Table 9 a e sel -explana o y and ocus on summa ising he co e
indings ha ha e al eady been discussed in Sec ions 4–5.
6.2. Resea ch implica ions
(i) Resea ch ypes based analysis is pe o med o unde s and
he ypes o esea ch conduc ed by he selec ed p ima y
21
A.A. Khan, A. Ahmad, M. Waseem e al. The Jou nal o Sys ems & So wa e 201 (2023) 111682
s udies (see Sec ion 4.2). Howe e , we ound ha none o
he s udies conduc ed e alua ion esea ch o assess a pa -
icula p oblem o solu ion. Quan um so wa e a chi ec u e
is an eme ging esea ch a ea and no e alua ion esea ch
s udies conduc ed o assess he con ibu ions p omised
by he a ailable a chi ec u al solu ions. I is a signi ican
esea ch gap, and we encou age he esea che s o ocus
on e alua ion esea ch o app aise he eal-wo ld signi i-
cance o quan um so wa e sys ems as well as he exis ing
ele an a chi ec u al p oblems.
(ii) Mos o he esea ch s udies we e conduc ed ac oss i e ap-
plica ion a eas (see Fig. 6(b)); howe e , we we e no able o
ind enough e idence ela ed o o he impo an a eas, like
model-d i en quan um so wa e a chi ec u e (MDQSA), quan-
um AI so wa e a chi ec u e, and quan um so wa e a chi-
ec u e applica ions o he indus ial p oblems (Ali e al.,
2022;Be els e al.,2021). The possible eason o lack o
esea ch in he men ioned a eas migh be ha quan um
so wa e a chi ec u e is a no el esea ch a ea and mos o
he s udies ocused on p oposing a chi ec u al solu ions o
quan um ha dwa e sys ems (see Fig. 6(b)). The e o e, we
encou age he esea ch communi y o pu mo e ocus on
he ollowing a eas: (1) Model d i en quan um so wa e
a chi ec u e (MDQSA) o manage complexi y, achie e high
le el euse and educe he de elopmen e o s (Ab eu
e al.,2021b). (2) Quan um AI so wa e a chi ec u e o
imp o e s a e-o - he-a and p opose solu ions o ope a e
beyond he classical compe encies (G ae and Geo gie ski,
2021). (3) Boos indus ial awa eness ela ed o quan um
so wa e a chi ec u e and de elop a chi ec u al solu ions
o deal wi h complex indus ial p oblems (Anon,2022a).
(iii) Conce ning he domain p oblem, we ho oughly in es i-
ga ed he challenging ac o s o quan um so wa e a chi-
ec u e (see Sec ion 5.5) and mapped hese ac o s ac oss
di e en majo hemes. Thema ic mapping p o ides a con-
cep ual amewo k o unde s and he b oad pic u e o he
iden i ied challenges and ba ie s o quan um so wa e a -
chi ec u e (Med ido ic e al.,2002).
In conclusion, his s udy p o ides quick access o he body o
knowledge based on quan um so wa e a chi ec u e li e a u e.
6.3. Indus ial implica ions
(i) We sys ema ically in es iga ed, analysed, and mapped he
exis ing ools and amewo ks ac oss he a chi ec ing ac-
i i ies (see Sec ion 5.4). A mapping be ween a chi ec ing
ac i i ies and co esponding ool suppo can guide p ac-
i ione s in exploi ing he a ailable ool suppo (enabling
au oma ion) o pe o m a speci ic a chi ec ing ac i i y. Fo
example, as shown in Table 8 i a p ac i ione wan s o con-
duc a chi ec u al alida ion, he/she can u ilise he QuNe -
Sim on implemen ed a chi ec u e o simula e quan um
in o ma ion p ocessing on a quan um ne wo k [S25]. In
gene al, he esul s o his SLR can acili a e he p ac i ion-
e s o ge an o e iew and analyse he ex en o which
a chi ec ing ac i i ies, pa e ns and exis ing ool suppo
ha enable semi-au oma ion can be le e aged o de elop
indus ial-scale solu ions o quan um so wa e.
(ii) We p oposed an a chi ec ing p ocess, which consis s o a
sequen ial lis o ac i i ies, ac ions, and e en s o de elop
a scalable quan um so wa e a chi ec u e (see Sec ion 5.1).
The p oposed p ocess ac s as a bluep in o p ac i ione s
o unde s and he inpu s, wo k low, and ou pu s o he
quan um so wa e a chi ec ing p ocess (see Sec ion 5.1).
(iii) Thema ic classi ica ion o iden i ied challenges (see Fig. 9)
p o ides an o e iew o po en ial ba ie s ha need o
conside by p ac i ione s be o e ini ia ing he a chi ec ing
ac i i ies (De S e ano e al.,2022).
(i ) Se e al s udies (n=11, 32%) discussed a chi ec u al ools
and amewo ks (see Sec ion 5.4). We de eloped a oolchain
o he iden i ied ools and amewo ks based on hei
con ibu ion ac oss he a chi ec ing ac i i ies (see Fig. 8).
I will assis he p ac i ione s o selec a sui able ool o
amewo k wi h espec o a speci ic a chi ec ing ac i i y.
Howe e , he e is s ill a need o indus ial e o s o
de elop mo e ad anced ools o manage he unexplo ed
a chi ec ing ac i i ies (Khammassi e al.,2021;Mi anskyy
e al.,2022).
The quan um so wa e a chi ec u e is a new and unexplo ed
esea ch a ea. Academic esea che s and indus ial p ac i ione s
wo king in quan um so wa e a chi ec u e domain a e in i ed o
con ibu e by sha ing hei expe iences. I will alle ia e he gap
be ween academic esea ch and indus ial p ac ices.
7. Th ea s o alidi y
Va ious h ea s could impac he alidi y o his s udy. How-
e e , we adop ed he SLR guidelines p oposed by Ki chenham
and Cha e s o alle ia e hese h ea s (Ki chenham and Cha e s,
2007). The po en ial h ea s a e analysed based on he co e ou
ypes o alidi y h ea s: in e nal alidi y, ex e nal alidi y, con-
s uc alidi y, and conclusion alidi y (Wohlin e al.,2012;Zhou
e al.,2016).
7.1. In e nal alidi y
The ex en o which ce ain ac o s a ec he esul s and
analysis o he ex ac ed da a is called in e nal alidi y. Th ea s o
he in e nal alidi y o his s udy could happen in he ollowing
SLR phases:
Sea ch s a egy:. I migh be possible ha ele an p ima y s ud-
ies a e missed du ing he sea ch p ocess because o he sea ch
s ings and he o e lap ac oss he selec ed s udies due o he
snowballing app oach, as highligh ed by Jalali and Wohlin Jalali
and Wohlin (2012). Howe e , we explici ly de ined he sea ch
s a egy in Sec ion 3.1.3. The i s h ee au ho s ex ac ed he
sea ch e ms based on hei unde s anding o RQs, which we e
u he e ined by all he au ho s in consen mee ings. Mo e-
o e , he sea ch e ms we e used o de elop he sea ch s ing,
which was i e a i ely de eloped by all he au ho s. I should
be no ed ha he au ho s ha e ex ensi e esea ch expe ience
in conduc ing SLR based s udies in he so wa e enginee ing
domain.
S udies selec ion and quali y assessmen :. The inclusion and exclu-
sion c i e ia a e de ined in Sec ion 3.1.4 and used o il e he
sea ch esul s and selec he mos ele an s udies. The i s h ee
au ho s join ly pa icipa ed in he s udies selec ion p ocess. Fu -
he mo e, he i s au ho e alua ed he quali y o each selec ed
s udy agains he assessmen c i e ia de ined in Sec ion 3.2.2. The
second and hi d au ho s independen ly e i ied he assessmen
esul s o a oid pe sonal bias.
Da a ex ac ion:. Pe sonal bias is a undamen al da a ex ac ion
h ea in SLR s udies. We mi iga e his h ea by de ining he da a
ex ac ion o m (see Table 4) o consis en ly ex ac he ele an
da a. The i s h ee au ho s ini ially ex ac ed he da a; howe e ,
he o he co-au ho s pa icipa ed in he discussion mee ings o
emo e any doub and e i y he da a as sugges ed by Wohlin
e al. (2012).
22
A.A. Khan, A. Ahmad, M. Waseem e al. The Jou nal o Sys ems & So wa e 201 (2023) 111682
Da a syn hesis:. Inaccu a e da a classi ica ion and mapping migh
cause subjec i e in e p e a ion bias. Howe e , his h ea has
been alle ia ed by ollowing hema ic classi ica ion guidelines
p o ided by B aun and Cla ke (B aun and Cla ke,2006). Mo e-
o e , quan i a i e and quali a i e me hods a e used o analyse he
collec ed da a. The bias in he da a syn hesis p ocess could im-
pac he da a in e p e a ion p ocess. This h ea has been lessen
by using he well-es ablished desc ip i e s a is ical app oaches
o analyse he quan i a i e da a and hema ic mapping o he
quali a i e da a.
7.2. Ex e nal alidi y
Ex e nal alidi y e e s o he deg ee o which he s udy ind-
ings could be gene alised. We do no claim he gene alisabili y
o his s udy, howe e ; we ied o maximise i by p o iding an
explici o e iew o quan um so wa e a chi ec u e and logically
se ing he collec ed da a, esul s, analysis, and conclusions in
he s udy domain. We ollowed he igo ous p o ocol-based SLR
app oach o a ain ex e nal alidi y. Mo eo e , we ollowed he
guidelines p o ided by Chen e al. (2010) o sea ch and selec he
mos app op ia e digi al eposi o ies and a ge he ele an pee -
e iewed s udies. Me hodological de ails (Sec ion 3, and Fig. 2),
SLR p o ocol, and da a ex ac ion mechanism can suppo he
iden i ica ion and syn hesis o new s udies and mo e RQs o
ex end his esea ch and minimise he h ea o ex e nal alidi y.
7.3. Cons uc alidi y
A ele an cons uc alidi y could be ‘‘da a i ems’’ since we
as he esea ches obse ed, decided, and pick up he ex ag-
men s o con en om he iden i ied s udies. Pe haps, his da a
ex ac ion migh no ha e been co ec ly pe o med due o di -
e en easons. Fo ins ance, inapp op ia e sea ch s a egies could
cause h ea s like e u ning a se o i ele an s udies o missing
he ele an a icles. We ied o mi iga e hese h ea s by ol-
lowing ope a ion measu es, e.g., conduc ing g oup mee ings o
inalise he sea ch s ing, de eloping s udies inclusion and ex-
clusion c i e ia, pe o ming s udies quali y assessmen , and using
da a ex ac ion o m o emo e in e pe sonal bias. Addi ionally,
he sea ch s ing is cus omised acco ding o he peculia i ies o
he selec ed da abases o iden i y he mos ele an s udies.
7.4. Conclusion alidi y
Conclusion alidi y e e s o he deg ee o which he s udy
conclusions a e c edible o easonable. In his SLR, he selec ion
c i e ia was s ic so only quali y s udies (a clea objec i e and
e alua ion) we e selec ed o he analysis in his pape (Ki chen-
ham and Cha e s,2007). Addi ionally, b ains o ming sessions
a e conduc ed by he au ho s o discuss he s udy indings and
d aw he co ec conclusions. I was acknowledged ha beyond
he scope o cu en SLR, u u e e o s may be needed o e ol e
he esul s and conclusions, i newly published esea ch is o be
in es iga ed, ex ending he indings o his SLR.
8. Rela ed wo k
To he bes o ou knowledge, his wo k is he i s comp e-
hensi e sys ema ic li e a u e e iew on he esea ch o quan um
so wa e a chi ec u e, including a chi ec ing ac i i ies, modelling
no a ions, design pa e ns, ools and amewo ks, and a chi ec-
u al challenges. This sec ion discussed he ela ed wo k ha
co e s di e en aspec s o quan um so wa e enginee ing (Zhao,
2020;Gill e al.,2022;Suni a e al.,2021;Paulo and de Cama go,
2021;Ga cía e al.,2022).
Zhao (2020) conduc ed a classical su ey o co e co e quan-
um so wa e enginee ing li e-cycle ac i i ies. Zhao summa ised
ha he quan um so wa e de elopmen concep eme ges om
quan um p og amming languages, and i is conside ed synony-
mous o quan um p og amming (Zhao,2020). Howe e , he e
is a signi ican need o comple e so wa e enginee ing disci-
pline o quan um so wa e de elopmen . This su ey ex ensi ely
discussed he echnological suppo o quan um so wa e de-
elopmen li e-cycle phases, including equi emen s enginee -
ing, design, implemen a ion, es ing, and main enance. The s udy
indings e eal ha hese a eas (phases) a e apidly g owing;
howe e , hey a e s ill a om being ma u e.
Gill e al. (2022) conduc ed a comp ehensi e li e a u e su ey
o p o ide in-dep h obse a ions o quan um compu ing concep s
and discuss he open challenges expe ienced by he quan um
compu ing communi y. A lis o axonomies a e p oposed o p o-
ide concep ual unde s anding o selec ing he a ailable quan um
compu ing echniques and de e mining he op imal s a egies o
u ilise he classical supe compu ing in as uc u e. I is because,
he exis ing quan um compu e s a e s ill no s ong enough o
eplace he supe compu e s. Quan um compu e s a e coping wi h
he scaling-up challenge o quan um qubi s. I is s ill no ce ain
when exac ly quan um compu e s will eplace he classical; how-
e e , i is expec ed ha many exci ing imp o emen s will happen
in he nex decade.
Suni a e al. (2021) conduc ed a sys ema ic e iew ha su -
eys a ailable quan um p og amming language (QPLs) o
o e iew he s a e-o - he-a in he con ex o compu e p o-
g amming o de eloping quan um-in ensi e so wa e sys ems.
The s udy o mula es a numbe o RQs o in es iga e a ious
aspec s o QPL such as ypes o p og amming languages, ecen
ends in he de elopmen o QPLs, along wi h academic and
indus ial p og ess on he de elopmen and adop ion o QPLs.
The su ey also highligh s ha well-cu a ed design/a chi ec u e
o quan um so wa e impac s he selec ion o QPLs o quan um
p og amming. This SLR is also comple ed wi h a ecen ly con-
duc ed mixed me hod esea ch (De S e ano e al.,2022) (mining
Gi Hub eposi o ies and de elope su ey) o in es iga e he
s a e-o -p ac ice on QPLs in he con ex o QSE.
Paulo and de Cama go (2021) ecen ly pe o med a sys ema ic
mapping s udy, e iewing 24 s udies, o analyse he exis ing
esea ch on quan um so wa e de elopmen in ega d o QSE. The
ocus o he sys ema ic mapping is o unde s and he p ominen
p og amming in as uc u es, di e ences be ween he de elop-
men o classical o quan um-in ensi e so wa e, and he applica-
ion domain o quan um so wa e sys ems. The au ho s highligh
ha in he las decade he a ailabili y o ools, echnologies, and
p og amming in as uc u es ha e gi en impe us o academic on
quan um so wa e enginee ing.
Ga cía e al. (2022) conduc ed an SLR o explo e di e en
ypes o algo i hms de eloped o quan um machine lea ning
and i s applica ions. The s udy indings e eal ha he a ious
con en ional/classical algo i hms a e used o machine lea ning
solu ions in he quan um domain e.g., suppo ec o machine
and supe ised machines lea ning k-nea es neighbou s (KNN)
model. The classical algo i hms a e mainly used o image classi-
ica ion p oblems. In b oad, he implica ions o quan um machine
lea ning a e p omising, howe e , achie ing he ull-scale bene i s
o quan um machine lea ning is s ill a om being ma u e.
The la ge-scale implica ions o quan um machine lea ning algo-
i hms a e s ill exceedingly challenging because o quali y, speed
and scalabili y issues. I equi es massi e imp o emen s in he
exis ing QC in as uc u e o ackle complex indus ial p oblems.
23
A.A. Khan, A. Ahmad, M. Waseem e al. The Jou nal o Sys ems & So wa e 201 (2023) 111682
Table 10
A compa ison o esul s be ween his sys ema ic e iew and he exis ing seconda y s udies.
This e iew esul s Exis ing seconda y s udies
Zhao (2020)Gill e al. (2022)Suni a e al. (2021)Paulo and
de Cama go (2021)
Ga cía e al. (2022)
P o ocol based SLR e iew X X ✓(+) ✓(+) ✓(+)
Demog aphic de ail X X ✓(+) X X
Quan um compu ing basics ✓(*) ✓(*) ✓(+) ✓(+) ✓(+)
Quan um so wa e enginee ing ✓(*) ✓(+) X ✓(+) ✓(+)
Quan um so wa e a chi ec u e ✓(+) X (+) ✓(+) X X
A chi ec u e modelling no a ions X X X ✓(+) X
Quan um so wa e design pa e ns ✓(+) X X X X
A chi ec u e ools and amewo ks X X ✓(+) X X
Challenges X X X X X
No e: (✓: included, X: No included, *: Ex ensi e discussion, +: Simple o e iew).
8.1. Compa a i e analysis
The compa a i e analysis o ou wo k wi h he exis ing ela ed
s udies is shown in Table 10. The esul s e eal ha ou indings
a e signi ican ly dis inc o he exis ing ela ed wo k s udies.
Fo ins ance, su icien numbe o p ima y s udies a e published,
howe e ; only one seconda y s udy (Ga cía e al.,2022) pa ially
ollowed he o mal p o ocol-based SLR app oach o conduc he
e iew s udy (Ki chenham and Cha e s,2007). The SLR guide-
lines de eloped by Ki chenham and Cha e s a e widely adop ed
o conduc sys ema ic li e a u e e iews in so wa e enginee -
ing (Ki chenham and Cha e s,2007). Simila ly, we epo ed he
demog aphic de ails o each selec ed p ima y s udy, including
publica ion ype, equency, esea ch ypes, con ibu ion, and ap-
plica ion domains (RQ1), which a e no conside ed in he ela ed
wo k seconda y s udies.
Mo eo e , we p o ided a comp ehensi e o e iew o quan um
so wa e a chi ec u e, howe e Zhao (2020), and Suni a e al.
(2021) p o ided in oduc o y le el de ails o quan um so wa e
a chi ec u e. The subsequen compa ison is made based on a chi-
ec ing ac i i ies and modelling no a ions, which a e igno ed in
he ela ed s udies (see Table 10). We de eloped RQ2.1 and RQ2.2
o espec i ely de ine and discuss he key ac i i ies o quan um
so wa e a chi ec u e and modelling no a ions. Simila ly, Zhao
(2020) and Paulo and de Cama go (2021) p o ided a simple
o e iew o quan um so wa e design. Howe e , we explici ly
co e and discuss he exis ing design pa e ns (RQ2.3) used o
ackle he commonly occu ed quan um so wa e a chi ec u al
p oblems.
Addi ionally, no discussion o quan um so wa e ools and
amewo ks is p o ided in he ela ed seconda y s udies. We
explici ly explo ed he selec ed p ima y s udies o iden i y he
ools and amewo ks ha suppo a ious a chi ec ing ac i i ies
(RQ2.4). Finally, we epo ed quan um so wa e a chi ec u e chal-
lenges and p o ided hei hema ic classi ica ion map (RQ2.5).
Howe e , he exis ing ela ed s udies do no p o ide any de ails
o abs ac le el discussion o quan um so wa e a chi ec u e
challenging ac o s (see Table 10).
9. Conclusions
Quan um so wa e a chi ec u e -design and implemen a ion
bluep in o quan um so wa e - ep esen s a new gen e o so -
wa e a chi ec u es o add ess compu a ion-speci ic challenges
oo ed in quan um compu ing. Wi h a g owing momen um o
he adop ion o quan um age sys ems, indus ial ini ia i es o
echnology gian s (e.g., Google, Mic oso , IBM) and academic
esea ch ha e ocused on exploi ing a chi ec u al solu ions o
de elop quan um so wa e ha manages and manipula es quan-
um ha dwa e. This SLR ocused on in es iga ing pee - e iewed
published esea ch ha s eamlines he ole o so wa e a chi-
ec u es in designing, implemen ing, alida ing, and deploying
quan um so wa e. We e iewed a o al o 34 quali a i ely se-
lec ed s udies o conduc his SLR by answe ing a o al o 08 RQs
o a ine-g ained p esen a ion o he esul s.
Resul s p esen s ha mos o ou e iewed s udies (n = 21,
i.e., 62% app ox.) ha e been published in he las ou yea s (2018–
2021). Majo i y o he published esea ch ypes (i.e., p oposal
o solu ion and alida ion esea ch (n = 20, 59%) indica e ha
quan um so wa e a chi ec u e is in i s in ancy, apidly e ol -
ing by bo owing concep s om classical so wa e a chi ec u es
o add ess quan um speci ic challenges. Quan um-speci ic chal-
lenges include bu a e no limi ed o quan um sys ems co-design
and mapping Qubi s/Quga es o a chi ec u al componen s and
connec o s ha can be e ec i ely add essed by de i ing a p ocess
o a chi ec ing quan um so wa e. To suppo he a chi ec u al
p ocess, modelling no a ions need o build on es ablished oun-
da ions o UML p o iles and a chi ec u al desc ip ion languages
o (semi-) o mal speci ica ion o quan um so wa e a chi ec-
u es. The SLR iden i ied a o al o i e a chi ec ing ac i i ies, six
a chi ec u al pa e ns ha p omo e euse, 11 ools and ame-
wo ks ha can au oma e and cus omise he p ocess o quan um
so wa e a chi ec ing. While in es iga ing he a chi ec u al chal-
lenges, we iden i ied a o al o 15 eme ging challenging ac o s,
classi ied ac oss 04 di e en ca ego ies, o esol e eme ging is-
sues pe aining o a chi ec u al solu ions o quan um so wa e.
The implica ions o his SLR a e o :
(i) The esea che s in e es ed in ocusing on quan um so -
wa e a chi ec u e and willing o ill he open esea ch gaps
discussed in he s udy indings.
(ii) Facili a ing he knowledge ans e o p ac i ione s ega d-
ing quan um so wa e a chi ec u e applica ion domains,
a chi ec ing ac i i ies, modelling no a ions, design pa e ns,
ools and amewo ks, and challenges.
We in i e p ac i ione s o s ep o wa d o ocus mo e on
missing applica ion domains, design a chi ec u e desc ip ion lan-
guages, de elop ools and amewo ks o au oma e he less o-
cused a chi ec ing ac i i ies, and p opose solu ions o ackle he
challenging ac o s. We plan o conduc an empi ical s udy o
mine he code hos ing and ques ions and answe public pla -
o ms o know p ac i ione s’ pe cep ions ega ding he quan um
so wa e a chi ec u e. We inally plan o compa e he esul s o
he empi ical s udy and his SLR o iden i y he gap be ween he
esea ch and p ac ice ega ding quan um so wa e a chi ec u e.
24