Constructing the projective operators for the uncertainty rela- tionstions

BSField 2

8.11 Constructing the projective operators for the uncertainty rela- tionstions

In the presence of transmission losses{η, η⁰}and imbalanced beamsplitter ratios{α, α⁰}in the verification interferometers shown in Fig. 8.9, the projectors no longer correspond to the pure state descriptionsMˆi =

|WiihWi|in Eq. 8.2 (ref.³⁸, chapter 7). Using the standard technique for loss propagations and beamsplitter transformations¹⁹³, the original projectors|WiihWi|become mixed states of the following form,

|W_iihW_i| 7→(1−q₁⁽ⁱ⁾)|0000ih0000|+q⁽ⁱ⁾₁ |Π_iihΠ_i|, (8.21) whereq⁽ⁱ⁾₁ gives the probability of a successful projective measurement in mode ifor an entangled state

|Πii,|0000ih0000|is the vacuum state, and|ΠiihΠi|is a pure state containing a single-photon shared among four optical modes (ref.³⁸, chapter 7). In the case of a conditional measurement (i.e., post-selecting the cases where we find a single-excitation among the four outputs of the cascaded beamsplitters in Fig. 8.1c),|ΠiihΠi| describes the projective measurement for the output modei,Mˆi =|ΠiihΠi|. Unlike the original projectors

|Wiiin Eq. 8.2, these projectors|Πiimay not be orthogonal, but they span the single-photon subspaceρˆ₁of the physical stateρˆ_W.

Figure 8.9:A simplified setup for the verification protocol (sum uncertainty). The setup includes differ- ential transmission efficiencies{η, η⁰}(blue) and imbalanced beam-splitter ratios{α, α⁰}(red). The wiggly dashed arrows correspond to the auxiliary output modes which are traced over for loss propagation.

Generally, any imbalances (whether they are due to differential losses or beamsplitter ratios) in the verifi- cation interferometers cause reductions of the overlaps|hWi|Πii|²between projectors|Wiiand|Πii, thereby making the protocol less sensitive to entanglement inρˆW (ref.³⁸, section 7.6.1). In practice, the corrected bounds∆^(K)_b always decrease towards smaller two-photon componentycfrom the ideal lossless and balanced case, as shown for our experimental parameters{η, η⁰}and{α, α⁰}in Fig. 8.4. Furthermore, the uncertainties in the measurement of{η, η⁰}and{α, α⁰}cause an uncertainty in the determination of the bounds∆^(K)_b .

Fig. 8.9 depicts the setup for our verification protocol indicating the losses{η, η⁰}and beamsplitter ra- tios{α, α⁰}of the interferometers. Experimental parameters and their uncertainties for{η, η⁰}and{α, α⁰} are shown in Table 8.1. In our data analysis, we infer the photon statistics of modes{1⁰,· · ·,4⁰}inρˆ_out at the outputs of the verification interferometers from the measured photodetection statistics at detectors {D1,· · · , D4}. Thus, we exclude the losses corresponding to the output paths of the verification interferom- eters from our analysis. The small imbalances between the terms{α, α⁰}and{η, η⁰}in Table 8.1 contribute to the small correction of the theoretical bounds∆^(K)_b from the ideal projectors|Wiito non-ideal projectors

|Πii. To understand the small corrections of∆^(K)_b from|Wiito|Πiifor our parameters, we investigate the effect of{η, η⁰} and{α, α⁰}on the bound∆⁽³⁾_b for states containing at most tripartite entanglement for a fixed two-photon contaminationy_c = 0.035, corresponding to the lowest measuredy_cin our experiment.

Table 8.1:Experimental parameters and their uncertainties for beamsplitter ratios{α, α⁰}and trans- mission efficiencies{η, η⁰}of the verification interferometers. The systematic uncertainties (δκ) of{κ}

are fractionally(δκ/κ) = 5%forκ∈ {α, α⁰, η, η⁰}. Note thatα12'α34'α₁₄⁰ 'α⁰₂₃. The absolute differ- ences in the pairs of transmission efficiencies ({η1, η2},{η3, η4},{η₁⁰, η⁰₄},{η⁰₂, η⁰₃}) influence the correction to∆^(K)_b .

α12 α34 α⁰₂₃ α⁰₁₄ η1 η2 η3 η4 η₁⁰ η⁰₄ η⁰₂ η⁰₃ 0.5 0.53 0.52 0.53 0.57 0.57 0.52 0.56 0.67 0.66 0.62 0.66

a b

c

Figure 8.10:The effect of imbalances and losses to the determination of∆⁽³⁾_b .Scanning the boundary∆⁽³⁾_b for states containing at most three-mode entanglement as a function ofa, the beamsplitter ratioα⁰₂₃(shown as a black line), andb, the transmission efficiencyη⁰₃(shown as a black line) atyc= 0.035, which corresponds to the lowest two-photon contamination measured in Fig. 8.4. The measured∆atyc = (3.5±0.9)×10⁻² is shown as a filled circle, with a horizontal error indicating the systematic uncertainty in estimatinga,α⁰₂₃ andb,η⁰₃, respectively. The vertical error is the statistical uncertainty for the measured ∆. c, Histogram H(∆⁽³⁾_b )of the three-mode boundary∆⁽³⁾_b by repeating the calculations from randomly drawn sets of the transmission efficiencies{η, η⁰}and beamsplitter ratios{α, α⁰}atyc = 0.035. The histogram is fitted to a Gaussian function (shown as a black line) with(1/e)half-widthδ∆⁽³⁾_b = 0.018. The uncertaintyδ∆⁽³⁾ is determined by the joint distribution of{α, α⁰}and{η, η⁰}. Here, we assume independent normal distributions for the individual parameters in{α, α⁰}and{η, η⁰}. (Inset) Confidence level in the violation of the inequality

∆ ≥ ∆⁽³⁾_b for the three-mode bound ∆⁽³⁾_b . Experimentally measured ∆ are shown as filled circles, and the black line indicates the the three-mode bound∆⁽³⁾_b , along with its uncertainty for yc = 0.035. The large suppression of∆from∆⁽³⁾_b compared to the uncertaintyδ∆⁽³⁾_b for the bound affirms the unambiguous detection of genuine four mode entanglement.

Figs. 8.10a and 8.10b illustrate the processes of reductions in the three-mode boundary ∆⁽³⁾_b atyc = 0.035, due to a, imbalanced beamsplitter ratio (α₂₃⁰ ) and b, differential loss (η₃⁰), while leaving all other parameters in{α, α⁰} and{η, η⁰}fixed (Table 8.1). In particular, the correction ranges of∆⁽³⁾_b due to the

individual uncertainties of α⁰₂₃ andη⁰₃ are small compared to the measured ∼ 20σ (standard deviation) suppression of the sum uncertainty∆aty_c= 0.035relative to∆⁽³⁾_b (shown as a filled circle in Fig. 8.10).

Finally, we discuss our analysis of the uncertainty δ∆⁽³⁾_b in the bound∆⁽³⁾_b (Fig. 8.10c) due to the systematic uncertainties of all the parameters in{α, α⁰}and{η, η⁰}. We construct the histogramHof∆⁽³⁾_b by iterating the calculation of∆⁽³⁾_b with randomly drawn sets of{α, α⁰, η, η⁰}. Here, the parameters{α, α⁰}and {η, η⁰}are assumed to follow independent normal distributions, with their means and systematic uncertainties shown in Table 8.1. By fitting the histogram with a Gaussian distribution, we infer an uncertaintyδ∆⁽³⁾_b = 0.018and the centerh∆⁽³⁾_b i= 0.25for the boundary∆⁽³⁾_b . These values should be compared to the measured

∆ = (5.6±1.1)×10⁻²aty_c= (3.5±0.9)×10⁻². As depicted in the inset of Fig. 8.10c, our measurement yields∼ 9σsuppression of the uncertainty bound, reflecting the high confidence level in the violation of the bound∆⁽³⁾_b . Our experiment, therefore, unambiguously verifies the presence of four-mode entanglement with the imbalances({α, α⁰},{η, η⁰})in the verification interferometers.

Chapter 9

Entanglement of spin waves among four quantum memories

This chapter is largely based on ref.³³. Reference³³refers to the then current literature in2010at the time of publication.