Entanglement by Path Identity

Mario Krenn,1, 2, Armin Hochrainer,1, 2 Mayukh Lahiri,1, 2 and Anton Zeilinger1, 2,

1
Vienna Center for Quantum Science & Technology (VCQ), Faculty of Physics,
University of Vienna, Boltzmanngasse 5, 1090 Vienna, Austria.
2
Institute for Quantum Optics and Quantum Information (IQOQI),
Austrian Academy of Sciences, Boltzmanngasse 3, 1090 Vienna, Austria.
Quantum entanglement is one of the most prominent features of quantum mechanics and forms
the basis of quantum information technologies. Here we present a novel method for the creation of
quantum entanglement in multipartite and high-dimensional systems. The two ingredients are 1)
superposition of photon pairs with different origins and 2) aligning photons such that their paths
are identical. We explain the experimentally feasible creation of various classes of multiphoton
entanglement encoded in polarization as well as in high-dimensional Hilbert spaces starting only
from non-entangled photon pairs. For two photons, arbitrary high-dimensional entanglement can be
arXiv:1610.00642v3 [quant-ph] 9 Feb 2017

created. The idea of generating entanglement by path identity could also apply to other quantum
entities than photons. We discovered the technique by analyzing the output of a computer algorithm.
This shows that computer designed quantum experiments can be inspirations for new techniques.

In 1991 Zou, Wang and Mandel reported an ex- a c d b

d c
periment where they induce coherence between two
photonic beams without interacting with any of them a
[1, 2]. They used two SPDC (spontaneous para-
2
metric down-conversion) crystals, where one photon 2
pair is in a superposition of being created in crys-
a
tal 1 and crystal 2 which can be described as
1
|i = 12 (|ai|bi + |ci|di). The striking idea (origi-
nally proposed by Zhe-Yu Ou) was to overlap one of 1 ,
the paths from each crystal (Figure 1), which can be
written as |bi = |di. This method removes the which- Figure 1. a: The simplest example which uses the overlap-
crystal information of the final photon in path d. In ping modes has been discussed first in [1]. The two crys-
contrast to a quantum eraser, the information here is tals (grey squares) can produce one pair of photons (blue
lines), either in the first or in the second crystal with the
not erased by postselection. Instead, all photons ar-
pump beam depicted with black lines. If the two processes
rive in the same output irrespectively in which crystal are coherent and the photons have same the frequency and
they are created. The resulting state can be written polarization, it is not known in which crystal the photons
as |i = 12 (|ai + |ci) |di: One photon is in path d, are created. In that case, the resulting photon pair is in
and its partner photon is in a superposition of be- the state |i = 12 (|ai + |ci) |di. b: A simple sketch of
ing in path a and c. There has been some follow-up the same experiment. For simplicity, we will use this more
work in recent years in the areas of quantum spec- abstract representation of physical experiments in the rest
of the manuscript.
troscopy [3, 4], quantum imaging [5, 6], studies of
complementarity [710], optical polarization [11] and
in microwave superconducting cavities [12]. However, few special cases [17, 18]. Furthermore, we present
this striking idea has not been investigated in the con- for the first time a method to create arbitrary high-
text of quantum entanglement generation yet. dimensional two-photon entangled states, for exam-
Here we show that by superposing photon pairs cre- ple, arbitrary high-dimensional Bell states in orbital
ated in different crystals, and overlapping the pho- angular momentum (OAM) or frequency of photons.
tons paths, one can generate very flexible experiments Multi-photon entanglement in Polarization First,
producing various types of entanglement, both in the we consider 4-photon polarization entanglement (Fig-
multiphoton and the high-dimensional regime. We ure 2a). Crystal 1 and 2 can produce horizontally
start by presenting different schemes to produce var- polarized pairs while crystal 3 and 4 can produce ver-
ious multiphoton polarization-entangled states such tically polarized ones. The crystals are pumped co-
as Greenberger-Horne-Zeilinger (GHZ) states [13] and herently and the pump power is adjusted such that
W states [14, 15], and contrast them with traditional we can neglect the cases where more than two pairs
methods of creating these states [16]. The method are created. The idea is that four-photon coincidences
is then generalized to high-dimensional multiphoton (i.e. one photon in each of the four paths) can only
entangled states (such as a 4-particle 3-dimensional happen either when the two pairs come from crystal
GHZ state), which so far can only be produced in a 1 & 2 or crystal 3 & 4. No other event produces four-
 2

a a b b a b c d
a b c d e f ,
, ,
4 7
, , , 6
c 4 5
d 6 ,
5
,
3
, ,
, , ,
3 4
, , 1 2 3
1 2
, ,
Figure 2. Multiphoton entanglement with polarization. a: 1 2
In four crystals, two photon-pairs are produced. Crystal 1
and 2 produce horizontally polarized photons, while crys- Figure 3. W-states such as |i =
tal 3 and 4 produce vertically polarized ones. Four-photon 1
(|V HHHi + |HV HHi + |HHV Hi + |HHHV i) repre-
2
coincidences can only happen when crystal 1 and 2 fire sent a different type of multi-photon entanglement. While
together or when crystal 3 and 4 fires. This leads to a 4- the GHZ state is considered as the most non-classical
particle GHZ-state |i = 12 (|H, H, H, Hi + |V, V, V, V i). state, a W-state is the most robust entangled state
b: Entangled states with more numbers of particles can because the loss of one particle leaves an entangled state.
be created in an analogous way here a n-photon GHZ Here, four-folds can only happen if crystal 1 & 2 produces
state |i = 12 (|H, H, H, ...i + |V, V, V, ...i) is shown. both a pair of photons, or crystal 3 & 4, or crystal 5 &
6 or 6 & 7. Interestingly, in this setup, one photon from
crystal 1 can stimulate an emission in crystal 3. However,
photon coincidences. For example, if the pairs are this will not lead to four-fold coincidences, as there is no
produced in crystal 1 & 3, there will be two photons photon in path b. Thus, this event can be neglected.
in path c, but none in path b. The resulting four-
photon state can be written as (see Appendix for a
detailed calculation) The new scheme does not need entangled photons
 2 to start with. Furthermore removing of the which-
|i = |Ha , Hc i + |Hb , Hd i + |Va , Vb i + |Vc , Vd i , crystal-information using a PBS is not necessary, as it
1   has never been created in the first place. Our method
4-fold
|i |Ha , Hb , Hc , Hd i + |Va , Vb , Vc , Vd i , does not use cascaded down-conversion, as it has been
2 shown in recent articles producing multiphoton polar-
(1)
ization entangled states [25, 26]. In our examples,
where H and V stand for horizontal and vertical polar- stimulated emission does not happen (which would
ization, respectivly, and the subscript stands for the introduce noise in the entangled state), because the
photons path. The final result is a 4-photon GHZ- input modes into the crystal are orthogonal to the
state. A realistic diagram of the experimental setup output modes (for instance, having different polariza-
as well as discussion about requirements for tempo- tion). In other cases, such as for the generation of
ral coherence and indistinguishability (applying the the 4 photon W-state (figure 3), stimulated emission
methods from [19] to the four-photon case) can be can happen but its contributions to the four-photon
found in the Appendix. In an analogous way, by in- coincidences are negligible.
creasing the number of crystals and the pump power, Multi-photon entanglement in higher dimensions
entangled states with more photons can be created. The principle can be generalized to produce high-
In figure 2b the scheme for creating a n-photon GHZ dimensional multiphoton entangled states [27, 28].
state is shown. High-dimensional entanglement has been investigated
In contrast to our new method, the traditional mainly in the two-photon case [2932], with two re-
way of creating 4-photon entangled states requires cent exceptions which investigated three-dimensional
two crystals each producing a pair of polarization entanglement with three photons [17], and teleporta-
entangled photons. One photon from each crys- tion of two degrees-of-freedom of a single photon [33].
tal goes to a polarizing beam splitter (PBS), which Figure 4a shows our proposal for an experiment creat-
removes the which-crystal information. Triggering ing a 3-dimensional 4-party GHZ-state, starting from
on events where all 4 detectors click, a 4-particle crystals which create separable photon pairs. There
GHZ |i = 12 (|H, H, H, Hi + |V, V, V, V i) state is are 3 layers of 2 crystals i.e. 6 crystals that are
created [20]. With that traditional method, GHZ- pumped coherently, and photon pairs are created in
entanglement with 8 photons [21, 22] and very re- two of them (because the pump power is set to such
cently, up to 10 photons have been created [23, 24]. a level that higher-order emissions can be neglected).
 3

a a b c d b a b the OAM of the photon), and the subscript denotes

the photons path. In the same way as before, by
0,0 0,0 0,0 neglecting cases where more than two photon pairs are
5 6 4 produced, one finds that a 4-fold coincidence event in
+1 +1 +1 +1 detectors a, b, c, d can only be created either if crystal
1 & 2 fire together or crystal 3 & 4 or crystal 5 s& 6.
0,0
This leads to
0,0
3
0,0
3 4 4-fold 1 
|i |2a , 2b , 2c , 2d i + |1a , 1b , 1c , 1d i
+1 +1 +1 +1 3
0,0 
2 +|0a , 0b , 0c , 0d i . (3)
0,0 0,0
1 2 Our idea can further be generalized to cases of more
0,0
1 than 4 photons. As an example, a scheme for 6
photons entangled in 5 dimensions is shown in the
Figure 4. Multiphoton-Entanglement with high-
Appendix. In general, adding additional columns
dimensional degrees of freedom (in this example: or- (and increasing the pump power) increases the pho-
bital angular momentum). a: The setup produces two ton number n, while adding additional layers increases
photon-pairs. Four-Photon coincidences can only oc- the dimensionality of entanglement d. That allows for
cur when crystal 1 & 2 fire together, or crystal 3 & the creation of arbitrary n-photon states entangled in
4 or crystal 5 & 6. The produced photon pairs are d = n 1 dimensions dimensions (for even n, and
all in the lowest mode (such as OAM=0). After each d = n for odd n when one photon is used as a trig-
layer of crystals, a hologram increases the OAM of the
photons (depicted as red line). After the third layer,
ger). Furthermore, symmetries of the states can be ex-
photons from crystal 1 & 2 have OAM=2 and pho- ploited which lead to a vast number of available entan-
tons from the middle layer have OAM=1, which leads gled states (such as asymmetrically entangled states,
to a four-particle three-dimensional entangled GHZ-state which exist only when n and d are both larger than
(|i = 13 (|0, 0, 0, 0i + |1, 1, 1, 1i + |2, 2, 2, 2i)). b: The two [27]). A detailed analysis of which states can be
same technique can also be applied to two-particle states produced in this way is given in the in Appendix.
to produce general high-dimensional states. All crystals The efficiency, E, of state generation is the prob-
produce pairs of Gaussian photons. The red lines in- ability of getting a desired state from all n-fold pho-
dicate OAM holograms, the green lines indicate phase
shifters. For example, if the holograms all are OAM=1
ton terms. The GHZ state has an efficiency of E =
d
and phases are ignored, then the final output state is n/2 , other states have higher efficiencies (details
2 )
( nd
|i = 21 (|0, 0i + |1, 1i + |2, 2i + |3, 3i). However, if one in Appendix). The efficiency of this method and
changes phases and the OAM in photon b, arbitrary 4-
dimensional states can be created for example all 16
the commonly used technique for polarization GHZ
four-dimensional Bell states. By increasing the number of states are the same [23]. The expected efficiency of
crystals, more dimensions can be added. a 3-dimensional 3-photon GHZ state (the only high-
dimensional GHZ state where the experimental imple-
mentation is known) with our new technique is signif-
Each photon from the first layer (crystals 1 and 2) icantly higher than the known technique [18].
passes through two mode shifters, which in total shift Two-photon arbitrary high-dimensional entangle-
its mode by +2. In the case of orbital angular mo- ment Finally, we show that the same technique can
mentum (OAM) of photons [34], mode shifters are be applied to generate arbitrary high-dimensional en-
holograms which add one unit of OAM to the pho- tangled two-photon states, starting again with only
ton (an analogous method could be done with dis- separable (non-entangeld) photon pairs. As shown in
crete frequency or time bins [35, 36]). Photons from figure 4b, four crystals are set up in sequence (only
the second layer (crystal 3 and 4) pass through one one photon pair is produced) and their output modes
mode shifter, while the photons created in the upper- are overlapped. Between each crystal, one adds arbi-
most layer (crystal 5 and 6) stay in their initial mode. trary phase shifters and mode shifters. That allows
The resulting 4-photon state can be described in the for adjusting every individual term in the superposi-
following way (see Appendix for details): tion independently. For example, with all phases set
 to = 2 and all mode shifters being +1, the setup
|i = |2a , 2d i + |2b , 2c i + |1a , 1c i + |1b , 1d i creates |i = 12 (|0, 0i + i|1, 1i |2, 2i i|3, 3i). The
2 dimension can be increased by increasing the number
+|0a , 0b i + |0c , 0d i , (2) of layers (crystals); the minimum number of layers for
creating a d-dimensional entangled state is d.
where 0, 1 and 2 stand for the mode number (such as Traditional methods for producing high-
 4

dimensional entanglement exploit the entanglement in [19] for two-photon states) in general for multi-
produced directly in a crystal. Such methods can only photon experiments; Generalizing the creation of en-
produce very restricted type of states. Furthermore, tanglement by path identity (or more generally, iden-
those states are never maximally entangled and have tity of some degree of freedom) to other quantum en-
low rates of production. Our technique overcomes tities, e.g. microwave superconducting cavities [12],
these restrictions and can produce arbitrary high- atomic systems [43, 44], trapped ions [45], supercon-
dimensionally entangled two-photon states. We can ducting circuits [46].
also tune the amount of entanglement in the following Finally, we discovered this technique by analyzing
ways: 1) by adjusting the pump laser power between the output of a computer algorithm which designs new
different crystals, we can produce non-maximally quantum optical experiments [18]. From there, we
entangled pure states; 2) by pumping the crystals generalized the idea (see Appendix). It shows that
with pumps that are not fully coherent to each other, automated designs of quantum optical experiments
we can produce entangled mixed states. by algorithms can not only produce specific quantum
The number of photon pairs created does not de- states or transformations, but can also be a source of
pend on the number of crystals in the experimental inspiration for new techniques which can further be
setup. For example in figure 4b, even though there are investigated by human scientists.
4 crystals, only one photon pairs are created. There-
fore the expected two-photon rate is of the same order
as in a conventional single-crystal source. Moreover,
our method requires only separable photon pairs to ACKNOWLEGDEMENTS
begin with. Therefore, for OAM of photons, the pro-
duction rates can be significantly higher than the rate
The authors thank Melvin for investigating 100 mil-
achievable with a traditional method (where higher-
lion quantum optical experiments. We thank Manuel
dimensional entanglement created directly in the crys-
Erhard and Thomas Scheidl for helpful discussions.
tal). This is because it is substantially easier to create
We also thank Dominik Leitner for providing com-
photon pairs in zero-order (Gaussian spatial mode)
putation resources. This work was supported by the
than in higher-order modes.
Austrian Academy of Sciences (OAW), by the Eu-
Interestingly, the simplest special case of the tech-
ropean Research Council (SIQS Grant No. 600645
nique presented here is a commonly used source of
EU-FP7-ICT) and the Austrian Science Fund (FWF)
two-photon polarization-entanglement. The so-called
with SFB F40 (FOQUS) and FWF project CoQuS
cross-crystal source uses two crystal after each other,
No. W1210-N16.
where the first one can create a horizontally polarized
photon pair, the second one creates vertically polar-
ized photon pairs [37]. Pumping both crystal at the
same time and producing one pair of photons, one can
create a |i = 12 (|H, Hi + |V, V i) state. That tech-
[email protected]

nique can now be seen as a special case of a much [email protected]
broader technique to produce highly flexible high- [1] LJ Wang, XY Zou, and L Mandel. Induced coher-
ence without induced emission. Physical Review A,
dimensional multiphoton states in various degrees of
44(7):4614, 1991.
freedom, by exploiting superposition of photon-pair [2] XY Zou, LJ Wang, and Leonard Mandel. Induced
origins and overlapping of paths of photons. coherence and indistinguishability in optical interfer-
To conclude, we investigate new types of photonic ence. Physical review letters, 67(3):318, 1991.
entanglement generation by combining two methods: [3] SP Kulik, GA Maslennikov, SP Merkulova, AN Penin,
First, photon pairs which originate from different crys- LK Radchenko, and VN Krasheninnikov. Two-photon
tals are coherently superposed. Second, we align pho- interference in the presence of absorption. Journal
of Experimental and Theoretical Physics, 98(1):3138,
ton paths to manipulate the structure of the entan-
2004.
gled state. It allows for the generation of very general [4] Dmitry A Kalashnikov, Anna V Paterova, Sergei P
quantum entangled states, for high-dimensional and Kulik, and Leonid A Krivitsky. Infrared spectroscopy
multi-photonic systems. Our method can be favorably with visible light. Nature Photonics, 2016.
suited for photonic quantum computation schemes [5] Gabriela Barreto Lemos, Victoria Borish, Garrett D
particularly in miniaturized compact devices. Cole, Sven Ramelow, Radek Lapkiewicz, and Anton
Topics for future research involve: Exploring the Zeilinger. Quantum imaging with undetected pho-
tons. Nature, 512(7515):409412, 2014.
relation to generating entanglement by propagation, [6] Mayukh Lahiri, Radek Lapkiewicz, Gabriela Barreto
detection and post selection [38], by using the indis- Lemos, and Anton Zeilinger. Theory of quantum
tinguishability [3941] or by using linear optics [42]; imaging with undetected photons. Physical Review
Treating the temporal coherence (such as investigated A, 92(1):013832, 2015.
 5

[7] TJ Herzog, JG Rarity, H Weinfurter, and A Zeilinger. imental generation of an eight-photon greenberger
Frustrated two-photon creation via interference. hornezeilinger state. Nature communications, 2:546,
Physical review letters, 72(5):629, 1994. 2011.
[8] Thomas J Herzog, Paul G Kwiat, Harald Weinfurter, [23] Xi-Lin Wang, Luo-Kan Chen, Wei Li, H-L Huang,
and Anton Zeilinger. Complementarity and the quan- Chang Liu, Chao Chen, Y-H Luo, Z-E Su, Dian Wu,
tum eraser. Physical Review Letters, 75(17):3034, Z-D Li, et al. Experimental ten-photon entanglement.
1995. Physical Review Letters, 117(21):210502, 2016.
[9] A Heuer, R Menzel, and PW Milonni. Induced coher- [24] Luo-Kan Chen, Zheng-Da Li, Xing-Can Yao, Miao
ence, vacuum fields, and complementarity in bipho- Huang, Wei Li, He Lu, Xiao Yuan, Yan-Bao Zhang,
ton generation. Physical review letters, 114(5):053601, Xiao Jiang, Cheng-Zhi Peng, et al. Observation of
2015. ten-photon entanglement using thin bib 3 o 6 crystals.
[10] A Heuer, R Menzel, and PW Milonni. Complementar- Optica, 4(1):7783, 2017.
ity in biphoton generation with stimulated or induced [25] Hannes Hubel, Deny R Hamel, Alessandro Fedrizzi,
coherence. Physical Review A, 92(3):033834, 2015. Sven Ramelow, Kevin J Resch, and Thomas Jen-
[11] Mayukh Lahiri, Armin Hochrainer, Radek Lap- newein. Direct generation of photon triplets using
kiewicz, Gabriela Barreto Lemos, and Anton cascaded photon-pair sources. Nature, 466(7306):601
Zeilinger. Partial polarization by quantum distin- 603, 2010.
guishability. arXiv preprint arXiv:1510.04192, 2015. [26] Deny R Hamel, Lynden K Shalm, Hannes Hubel,
[12] Pasi Lahteenmaki, Gheorghe Sorin Paraoanu, Juha Aaron J Miller, Francesco Marsili, Varun B Verma,
Hassel, and Pertti J Hakonen. Coherence and multi- Richard P Mirin, Sae Woo Nam, Kevin J Resch,
mode correlations from vacuum fluctuations in a mi- and Thomas Jennewein. Direct generation of three-
crowave superconducting cavity. Nature Communica- photon polarization entanglement. Nature Photonics,
tions, 7:12548, 2016. 8(10):801807, 2014.
[13] Daniel M Greenberger, Michael A Horne, and Anton [27] Marcus Huber and Julio I de Vicente. Structure of
Zeilinger. Going beyond bells theorem. In Bells theo- multidimensional entanglement in multipartite sys-
rem, quantum theory and conceptions of the universe, tems. Physical review letters, 110(3):030501, 2013.
pages 6972. Springer, 1989. [28] Jay Lawrence. Rotational covariance and greenberger-
[14] A Zeilinger, MA Horne, and DM Greenberger. Nasa horne-zeilinger theorems for three or more particles
conf. publ. no. 3135. 1997. of any dimension. Physical Review A, 89(1):012105,
[15] Mohamed Bourennane, Manfred Eibl, Christian 2014.
Kurtsiefer, Sascha Gaertner, Harald Weinfurter, Ot- [29] Adetunmise C Dada, Jonathan Leach, Gerald S
fried Guhne, Philipp Hyllus, Dagmar Bru, Maciej Buller, Miles J Padgett, and Erika Andersson. Exper-
Lewenstein, and Anna Sanpera. Experimental detec- imental high-dimensional two-photon entanglement
tion of multipartite entanglement using witness oper- and violations of generalized bell inequalities. Nature
ators. Physical review letters, 92(8):087902, 2004. Physics, 7(9):677680, 2011.
[16] Jian-Wei Pan, Zeng-Bing Chen, Chao-Yang Lu, [30] J Romero, D Giovannini, S Franke-Arnold, SM Bar-
Harald Weinfurter, Anton Zeilinger, and Marek nett, and MJ Padgett. Increasing the dimension in
Zukowski. Multiphoton entanglement and interfer- high-dimensional two-photon orbital angular momen-
ometry. Reviews of Modern Physics, 84(2):777, 2012. tum entanglement. Physical Review A, 86(1):012334,
[17] Mehul Malik, Manuel Erhard, Marcus Huber, Mario 2012.
Krenn, Robert Fickler, and Anton Zeilinger. Multi- [31] Mario Krenn, Marcus Huber, Robert Fickler, Radek
photon entanglement in high dimensions. Nature Pho- Lapkiewicz, Sven Ramelow, and Anton Zeilinger.
tonics, 10(4):248252, 2016. Generation and confirmation of a (100 100)-
[18] Mario Krenn, Mehul Malik, Robert Fickler, Radek dimensional entangled quantum system. Proceedings
Lapkiewicz, and Anton Zeilinger. Automated search of the National Academy of Sciences, 111(17):6243
for new quantum experiments. Physical review letters, 6247, 2014.
116(9):090405, 2016. [32] Yingwen Zhang, Filippus S Roux, Thomas Kon-
[19] Anand Kumar Jha, Malcolm N OSullivan, Kam rad, Megan Agnew, Jonathan Leach, and Andrew
Wai Clifford Chan, and Robert W Boyd. Temporal Forbes. Engineering two-photon high-dimensional
coherence and indistinguishability in two-photon in- states through quantum interference. Science ad-
terference effects. Physical Review A, 77(2):021801, vances, 2(2):e1501165, 2016.
2008. [33] Xi-Lin Wang, Xin-Dong Cai, Zu-En Su, Ming-Cheng
[20] Jian-Wei Pan, Matthew Daniell, Sara Gasparoni, Gre- Chen, Dian Wu, Li Li, Nai-Le Liu, Chao-Yang Lu,
gor Weihs, and Anton Zeilinger. Experimental demon- and Jian-Wei Pan. Quantum teleportation of mul-
stration of four-photon entanglement and high-fidelity tiple degrees of freedom of a single photon. Nature,
teleportation. Physical Review Letters, 86(20):4435, 518(7540):516519, 2015.
2001. [34] Mario Krenn, Mehul Malik, Manuel Erhard, and An-
[21] Xing-Can Yao, Tian-Xiong Wang, Ping Xu, He Lu, ton Zeilinger. Orbital angular momentum of photons
Ge-Sheng Pan, Xiao-Hui Bao, Cheng-Zhi Peng, Chao- and the entanglement of laguerregaussian modes.
Yang Lu, Yu-Ao Chen, and Jian-Wei Pan. Observa- Phil. Trans. R. Soc. A, 375(2087):20150442, 2017.
tion of eight-photon entanglement. Nature Photonics, [35] Daniel Richart, Yvo Fischer, and Harald Wein-
6(4):225228, 2012. furter. Experimental implementation of higher dimen-
[22] Yun-Feng Huang, Bi-Heng Liu, Liang Peng, Yu-Hu Li, sional timeenergy entanglement. Applied Physics B,
Li Li, Chuan-Feng Li, and Guang-Can Guo. Exper- 106(3):543550, 2012.
 6

[36] Anthony Martin, Thiago Guerreiro, Alexey Tiranov, a b c d

Sebastien Designolle, Florian Frowis, Nicolas Brun-
ner, Marcus Huber, and Nicolas Gisin. Quantify-
ing photonic high-dimensional entanglement. arXiv
3 4
preprint arXiv:1701.03269, 2017.
[37] Paul G Kwiat, Edo Waks, Andrew G White, Ian lp3 lp4
Appelbaum, and Philippe H Eberhard. Ultrabright
source of polarization-entangled photons. Physical
Review A, 60(2):R773, 1999.
[38] C Thiel, J Von Zanthier, Thierry Bastin, E Solano,
and GS Agarwal. Generation of symmetric dicke
states of remote qubits with linear optics. Physical SPDC crystal
review letters, 99(19):193602, 2007. 1 2
[39] N Killoran, M Cramer, and MB Plenio. Extracting trombone
entanglement from identical particles. Physical review
letters, 112(15):150501, 2014. lp1 lp2 detector
[40] Toshihiko Sasaki, Tsubasa Ichikawa, and Izumi Tsut- LASER
beam splitter
sui. Entanglement of indistinguishable particles. beam stopper
Physical Review A, 83(1):012113, 2011. mirror
[41] Malte C Tichy, Florian Mintert, and Andreas Buch-
leitner. Limits to multipartite entanglement genera-
tion with bosons and fermions. Physical Review A, Figure 5. A realistic diagram of the experiment described
87(2):022319, 2013. in the main text, which produces a 4-photon GHZ-state
[42] Piotr Migdal, Javier Rodrguez-Laguna, Michal Osz- in polarization. The pump length `p1 and `p2 start at the
maniec, and Maciej Lewenstein. Multiphoton states BS and end at crystals 1 and 2. The lengths `p3 and `p4
related via linear optics. Physical Review A, start at crystals 1 and 2, and end at crystals 3 and 4. The
89(6):062329, 2014. lengths for the down-conversion photons `i start at the
[43] Michael Keller, Mateusz Kotyrba, Florian Le- crystals 1 and 2, and end at crystals 3 and 4.
upold, Mandip Singh, Maximilian Ebner, and Anton
Zeilinger. Bose-einstein condensate of metastable he-
lium for quantum correlation experiments. Physical achieved, in analogy to the cross-crystal two-photon
Review A, 90(6):063607, 2014. source, with type-I SPDC and a diagonally polarized
[44] Raphael Lopes, Almazbek Imanaliev, Alain Aspect, pump beam.
Marc Cheneau, Denis Boiron, and Christoph I West-
In order to make sure that the states are produced
brook. Atomic hong-ou-mandel experiment. Nature,
520(7545):6668, 2015. in a coherent superposition, path lengths need to be
[45] Rainer Blatt and David Wineland. Entangled states chosen appropriately. In the example above (Fig-
of trapped atomic ions. Nature, 453(7198):10081015, ure 5), it needs to be fulfilled that one cannot dis-
2008. tinguish whether the 4-fold coincidence count came
[46] R Barends, J Kelly, A Megrant, A Veitia, D Sank, from crystal 1 and 2, or whether it came from crys-
E Jeffrey, TC White, J Mutus, AG Fowler, B Camp- tal 3 and 4. For that, the path lengths `1 , `2 , `3 and
bell, et al. Superconducting quantum circuits at the
`4 need to be equal within the coherence length of
surface code threshold for fault tolerance. Nature,
508(7497):500503, 2014. the down-converted photons. Furthermore, the pump
laser paths need to be matched such that `p1 = `p2
and `p3 = `p4 within the coherence time of the laser.
APPENDIX I. REALISTIC DIAGRAM OF THE The final requirement is that the path lengths of the
POLARIZATION GHZ SETUP AND laser between the two crystals must be matched with
COHERENCE TIME REQUIREMENTS the path length of the down-converted photons, such
that `p3 = `1 within the coherence time of the laser.
A possible implementation of the scheme described It can be summarized as
in figure 2a , which produces a 4-photon GHZ state
in polarization, is shown in more detail here. A laser |`i `j |  cSP DC , (4a)
(depicted in blue), which is splitted at a 50/50 beam |`p1 `p2 |  cpump , |`p3 `p4 |  cpump , (4b)
splitter arrive at the same time at crystal 1 and 2. |`p3 `i |  cpump , |`p4 `i |  cpump , (4c)
For that, the pathlength between the beam splitter
and the crystals 1 and 2 (written as `p1 and `p2 ) need where cSP DC and cpump are the coherence length
are matched with a standard trombone system. of the down-converted photons and the pump laser,
Crystals 1 and 2 can both produce photon pairs respectively. We find these restrictions on the tempo-
with horizontal polarization, while crystal 3 and 4 can ral coherence and indistinguishability by applying the
both produce vertically polarized photon pairs with results that were obtained for the two-photon case by
non-colinear phase-matching conditions. This can be Jha et al. [19]. These are similar requirements as for
 7

standard 4-photon GHZ sources based overlapping the reflectivity R:

photons at a polarizing beam splitter. Furthermore,
Ta (aa,P ) = Ta aa,P + Ra a0,P (7)
also the requirements for the matching of frequencies
is analogous to the standard 4-photon GHZ source us- where a0,P is an empty loss mode, and |R|2 +|T |2 = 1.
ing an PBS. Considering misalignment in all four arms, which we
For other polarization experiments, very similar re- denote as T(a,b,..,n) (x) = Ta (Tb (..(Tn (x)..)), we find
quirements are necessary. In experiments involving  
VV VV HH HH
spatial modes (such as in Figure 4 or Figure 6 in the |i = Uc,d Ua,b T(a,b,c,d) Ub,d Ua,c |vaci
Appendix), it is further important to eliminate spa- 
tially dependent phase shifts acquired on propagation = g 2 aa,V ab,V ac,V ad,V
of pump and down-conversion photons. This can be 
done with using standard 4-f lens systems. + Ta Tb Tc Td aa,H ab,H ac,H ad,H |vaci
We show in Section II that if there are misalign- g2 
ments between the overlapping paths, it does not re- + Ta Tc a2 2 2
a,H ac,H + Tb Td ab,H ad,H
2
2
duce the coherence between the different terms but 
changes the relative amplitudes between them. This + a2 2 2 2
a,V ab,V + ac,V ad,V |vaci
is because misaligned beams do not arrive at the de- + (less than four photons) + O(g 3 ) (8)
tectors and consequently do not lead to a four-fold
coincidence count. Very similar conditions apply for If Ti = 1, the 4-photon terms of equation (8) are rep-
all other experimental setups presented in the main resented by the first line of equation (1).
text. Terms that do not lead to four photons in the spatial
modes a, b, c, d are discarded by the four-fold coinci-
dence detection, and higher order terms in g are dis-
APPENDIX II. DETAILED CALCULATION OF carded as their probability is small for g  1. There-
THE QUANTUM STATES AND THE EFFECT fore we find that the resulting state can be written
OF MISALIGNMENT as
|Va , Vb , Vc , Vd i + Ta Tb Tc Td |Ha , Hb , Hc , Hd i
Here we perform the detailed calculation of the 4- |i = p
1 + Ta2 Tb2 Tc2 Td2
photon GHZ experiment. The down-conversion pro-
(9)
cess can be described as a series expansion in the form
of That means that the coherence between the state does
  not decrease, but the state is not equally distributed
P1 P2
Ua,b = 1 + g aa,P1 ab,P2 aa,P1 ab,P2 thus it is not maximally entangled. However, reducing
g2  2 the laser power before crystal 3 and 4 (thus changing
+ aa,P1 ab,P2 aa,P1 ab,P2 + O(g 3 ) the value of g for the crystal 3 and 4) can compen-
2
(5) sate for the drop in entanglement. Conversely, a fil-
ter in the down-converted photons could be used to
where aa,P and aa,P are creation and annihilation op- compensate for lower pump power. If Ti = 1, equa-
erators for a photon in the mode a and with polar- tion (9) is the same as the second line of equation
ization P , respectively, and g is proportional to the (1). In the following examples, for simplicity we set
SPDC rate and the pump power. For simplicity, we Ta = Tb = .. = T .
restrict ourselves to single-mode analysis. In the 4- For a six-photon GHZ state in polarization, an anal-
photon GHZ setup, four crystals are used, therefore ogous calculation with
the state can be expressed as VV VV VV

HH HH HH

|i = Ua,b Uc,d Ue,f T(a,b,c,d,e,f ) Ua,c Ub,e Ud,f |vaci,
V V V V HH HH
|i = Uc,d Ua,b Ub,d Ua,c |vaci, (6) (10)
where |vaci is the vacuum state. In addition, we want leads to the six-photon GHZ state
to consider the effect of misalignment between the |Va , Vb , Vc , Vd , Ve , Vf i + T 6 |Ha , Hb , Hc , Hd , He , Hf i
modes from the different crystals. As introduced in |i = .
1 + T 12
[2], if the down-converted modes are perfectly over- (11)
lapped, they can be treated as one mode. In the ex-
treme case of complete misalignment, the photons do Similarly, for the W-state, the experiment can be de-
not reach the detector, which can be described by in- scribed by
serting a beam stop between the crystals. For the HV HH
|i =Ua,b Uc,d T(a,b,c,d)
intermediate cases, we can describe the situation us- V H HV HH HV HH


ing a variable beam splitter with transmissivity T and Ua,b Ua,c Ub,d Ua,d Uc,b |vaci (12)
 8

which leads to (accounting for misalignment between Similarly as in examples above, misalignment reduces
each crystal, and using four-fold coincidences) to the entanglement by unweighting the state. How-
ever, that effect can be compensated by adjusting the
1 4
|i = T |Ha , Hb , Hc , Vd i + T 4 |Ha , Hb , Vc , Hd i pump power before each crystal. Therefore maximally
N  entangled, arbitrary, high-dimensional entangled two-
+ T 2 |Ha , Vb , Hc , Hd i + |Va , Hb , Hc , Hd i , (13) photon states can be created.
where N is a normalization constant. In this exper-
iment, we expect that the noise from higher-order APPENDIX III. CONSTRUCTION OF
terms is slightly higher because of induced emission GENERAL EXPERIMENTS
due to the fact that the input states are the same
as the states produced in the crystals (in contrast to Two-Photon Case For two photons, arbitrary
other examples). The different coefficients in front of high-dimensional quantum states can be created. If
the terms can be compensated again by adjusting the d crystals are pumped coherently as shown in Figure
power of the pump laser. 4b in the main text, the resulting state is in a super-
The examples considering a high-dimensional de- position of being created in either of the crystals. As
gree of freedom (such as orbital angular momentum the down-converted photons can be manipulated be-
(OAM) of photons) use mode shifters in addition to tween each crystal, the mode-number and phase can
down-conversion crystals (figure 4a in the main text), be adjusted for each individual term of the complete
which can be defined as state.
 
Sa aa,` = aa,`+1 (14)
a a b c d e f b
and S(a,b,..,n) (x) = Sa (Sb (..(Sn (x)..)). Then, the state 0,0 0,0 0,0
in the four-photon three-dimensional experiment can a b c T
+1
be described as
0,0
|i = Ua,b Uc,d T(a,b,c,d) 7



0,0 0,0 0,0
S(a,b,c,d) Ua,c Ub,d S(a,b,c,d) Ua,d Ub,c |vaci (15) +2 +1 0,0
+1 6
where U produces the same polarization, and zero- 0,0
order OAM modes. A completely analogous calcula- 0,0 0,0 0,0 5
+1
tion as above leads to -1
+1
1 -1
|i = |0a , 0b , 0c , 0d i + T 4 |1a , 1b , 1c , 1d i 0,0 0,0
N
3 4
0,0 0,0
 0,0
4
+ T |2a , 2b , 2c , 2d i , (16) +2
+1 +1
0,0 0,0
where again the different coefficients, which occur due 1 2
to imperfect overlapping of the modes, can be compen-
0,0 0,0 0,0
sated by adjusting the pump power in crystal 5 and
6.
Finally, the experiment for the two-photon high- Figure 6. a: The experimental configuration for a 6-
dimensional entangled state allows for adjustable particle 5-dimensional GHZ state. b: An experiment
phases, which can be written as which creates an asymmetricly entangled quantum state,
with a Schmidt-Rank-Vector of (4,2,2). One photon (in
Pa, aa,` = exp i aa,`


(17) detector T) is used to trigger the three-photon state in
(a,b,c).
For example, with four crystals in a row with mode
shifters constantly adding +1, and phase shifters
changing the phase by /2, we can write Multi-Photon Case We analyse the case where n-
photon states are created in n different paths. That
|i = Ua,b Ta S(a,b) Pb, 2 Ua,b S(a,b) Pb, 2 Ua,b requires at least c = n2 crystals to fire simultaneously,
S(a,b) Pb, 2 Ua,b S(a,b) Pb, 2 Ua,b





|vaci. (18) and together emit photons in n different paths. In
the example 6a, n = 6 and c = 3 (six photons are
This leads to created with three crystals - which is represented in
1  every row of the setup). There are d = n 1 different
|i = |0a , 0b i + iT |1a , 1b i T |2a , 2b i iT |3a , 3b i ways to arrange c crystals such that they produce n-
N
(19) fold coincidence counts. In that example, d = 5 and
 9

a n=6-fold coincidence count can be created by c = 3 with and `i,j Z. Furthermore, one can use filters
crystals emitting in six paths (ab-cd-ef, ac-be-df, ad- both in the pump and down-converted photon
bf-ce, ae-bd-cf, af-bc-de). That means, arbitrary (n- paths, such that ci,j C with |ci,j | 1. The
W-state (presented in figure 3 of the main text)
a b c d is a special case of figure 7. A high-dimensional
a b
expample is shown in figure 6b. It shows a setup
a for creating n=3 photon entangled state (with
b c d
one photon acting as a trigger |0T i) with d = 4
5 terms in an asymmetric configuration (|i =
0,0 0,0 6
1
5 6 2 |0T i (|0a 0b 0c i + |1a 0b 1c i + |2a 1b 0c i + |3a 1b 1c i)).
The state can be quantified by the Schmidt-Rank
Vector (4,2,2) [27]. A variaty of similar states can be
written down in that form.
0,0 0,0
3 4 3 4

APPENDIX IV. EFFICIENCY OF STATE

0,0 0,0 CREATION
1 2

The method is probabilistically (not every gener-

1 2 ated n-photon state leads to a valid n-fold photon
state with one photon in each of the n paths). The
effciency E is the number of valid n-fold events di-
Figure 7. a: Experiment for an arbitrary 4-photon 3- vided by the number of possible n-photon events. For
dimensional GHZ state b: A generalisation of GHZ state. d-dimensional n-particle GHZ states, there are d com-
Every crystal can be replaced by a row of crystals each pro- binations that lead to a valid n-fold, and for c crystals
ducing arbitrary high-dimensional 2-photon states. The there are cn/2 possible combinations leading to a n-
red lines indicate an arbitrary mode shifter (such as holo-
photon states. For GHZ-states, c = nd 2 . This leads
grams for OAM). d
to an effciency of E = nd n/2 .
( 2 )
1)-dimensionally entangled n-photon GHZ state can
be created. For the case of n=4, that can be written
as APPENDIX V. HOW WE FOUND THE
TECHNIQUE
1
|1 i = |`1 , `2 , `3 , `4 i+
N The computer algorithm described in [18] produced
c1,1 |`5 , `6 , `7 , `8 i+ setups for experimentally feasible high-dimensional
c1,2 |`9 , `10 , `11 , `12 i


(20) multi-photon states. We analyzed a result of an un-
usually high-dimensional entangled 3-photon state, as
where `i Z and ci,j {0, ei } (which is depicted in it was not expected that such a state can be cre-
Figure 7a). ated, given the restricted set of available elements of
Each crystal could now be replaced by a row of crys- the algorithm (figure 8a). One restriction was that
tals (such as Figure 4b in the main text), where each each setup starts with two SPDC crystals, which can
row produces an arbitrary 2-photon d-dimensional en- produce each 3-dimensional 2-photon entanglement.
tangled states. That leads to the following possible However, in that particularly unusual solution, the al-
state gorithm used three types of available SPDC process in
a superposition (crystal 1 and crystal fires twice, and
1 crystal 1 and 2 fire once together). Upon understand-
|2 i =
N ing that technique, we gave the algorithm explicitly
the possibility to use several crystals in superposition
X
X

c1,i |`1,i , `2,i ia,b c2,i |`3,i , `4,i ic,d +
i i and overlapping their paths. We further restricted
X
X
other available elements, in order to understand how
c3,i |`5,i , `6,i ia,c c4,i |`7,i , `8,i ib,d +
powerful this technique is. The algorithm was able
i i
X
X
 to find a high-dimensional entangled state using only
c5,i |`9,i , `10,i ia,d c6,i |`11,i , `12,i ib,c (21) crystals and mode shifters (figure 8b). We further re-
i i stricted it to the simplest case of 2-dimensional polar-
 10

a
ization, were it also found solutions (figure 8c). From
SRV=(10,6,6)
b c there on, the we extracted the principle idea of the
SRV=(6,4,4) 4-party 2-dim GHZ technique and subsequently generalized it.
a b
a b a b
BS c
,
d 4 4
BS

c d c d
BS
,
-1 -1 3 +2 3
Ref
+2 +4 DP +3 +3 +4
, ,
1 2 1 2 1 2

Figure 8. a: We investigated an experiment with an un-

expectedly large Schmidt-Rank Vector (10,6,6) [27]. Each
crystal produces a three-dimensional two-photon entan-
gled state, and |Ta i stands for the trigger in mode a. The
experiment works because the crystals are producing four-
photons in a coherent superposition of one SPDC event
in each of the crystals simultaneously, and two SPDC
events in each of the crystals. b: By explicitly allow-
ing superpositions of crystals and enableing overlapping
of their output paths (but removing any other interac-
tion such as beam splitters), the program was still able to
find high-dimensional multi-photon experimental setups.
c: Restricting the program further to allow only for two-
dimensional polarisation and no trigger, it still found a
solution. From that solution we understood the very basic
idea, which we were able to generalize further.