==================================================================================================== RESULT A -- ten X gates in a row on 7 well-separated qubits, read 1 more often when the 22 neighbours were held half-way (equator) around the train. Rigetti Cepheus-1-108Q, 2026-10-04. ==================================================================================================== tasks: 42 (3 rounds x [8 calibration + 6 experiment]); sent 2026-10-04T17:15:00Z .. 2026-10-04T17:21:23Z; 900 shots per experiment task round 1: 2026-10-04T17:15:00Z .. 2026-10-04T17:15:06Z round 2: 2026-10-04T17:18:07Z .. 2026-10-04T17:18:14Z round 3: 2026-10-04T17:21:16Z .. 2026-10-04T17:21:23Z emitters (the qubits that are pushed 10x): [7, 22, 27, 57, 62, 72, 104]; neighbours (22): [6, 13, 16, 18, 21, 23, 28, 31, 36, 48, 53, 56, 58, 61, 63, 66, 71, 73, 81, 95, 103, 105] neighbours of each emitter: {7: [6, 16], 22: [13, 21, 23, 31], 27: [18, 28, 36], 57: [48, 56, 58, 66], 62: [53, 61, 71], 72: [63, 73, 81], 104: [95, 103, 105]} experiment task ids (state|readout|round -> Braket task id, region us-west-1; readout e = plain readout, ba = swap-then-flip readout): clock|ba|1 36ebf6a7-623e-430d-bd0b-91f843488e0f clock|ba|2 b4c72e82-e0a1-40c7-a298-be136aaa0223 clock|ba|3 38c2626a-7970-460f-9528-9a1af17305df clock|e|1 35b921b5-fcf3-4c4c-b75a-1705af49d096 clock|e|2 3a3cadff-0351-44d7-b251-fe5c4cbec06f clock|e|3 44db24fb-1358-4ac3-9191-6d1222d06382 pushed|ba|1 3cea5eec-7b01-4669-8944-952f5d39fe11 pushed|ba|2 6df602d3-890a-44ba-b6d7-a63d56c46c20 pushed|ba|3 0aa8cdfb-61cb-4ea9-839c-8f3eecf4ba07 pushed|e|1 279a8324-848f-49dc-9b82-2cab7a591641 pushed|e|2 a764b9db-2739-4019-bd4e-163dd2826f86 pushed|e|3 92496561-6347-4283-90d8-55bcd2f466d1 rest|ba|1 b67de414-b185-4b0b-b232-4b02d23a3015 rest|ba|2 f3214ad4-accd-48f0-b00c-fe55ec2dbe0a rest|ba|3 8fd2eb50-0363-4d25-a753-f65828f7fa9f rest|e|1 71384e57-2199-4014-880f-6e613baab6fe rest|e|2 5d2a000f-fbce-4490-9c50-3b7d2178c6c2 rest|e|3 c249f361-8ce6-4ea1-8c9a-fed9048bf9f0 A1. read-1 fraction of the pushed qubits, mean over the 7 (per-qubit values below): round 1 mean over the 7: rest 0.064286 / pushed 0.064603 / clock 0.11016 round 2 mean over the 7: rest 0.064127 / pushed 0.065556 / clock 0.10095 round 3 mean over the 7: rest 0.058413 / pushed 0.056032 / clock 0.10381 per-qubit read-1 fractions (columns q7 q22 q27 q57 q62 q72 q104): round 1 rest : 0.057 0.049 0.083 0.080 0.060 0.092 0.029 mean 0.0643 round 1 pushed: 0.074 0.043 0.119 0.081 0.059 0.034 0.041 mean 0.0646 round 1 clock : 0.098 0.151 0.216 0.083 0.082 0.078 0.063 mean 0.1102 round 2 rest : 0.060 0.064 0.063 0.081 0.051 0.081 0.048 mean 0.0641 round 2 pushed: 0.074 0.041 0.129 0.073 0.074 0.037 0.030 mean 0.0656 round 2 clock : 0.091 0.151 0.184 0.093 0.069 0.061 0.057 mean 0.1010 round 3 rest : 0.073 0.048 0.061 0.076 0.042 0.083 0.026 mean 0.0584 round 3 pushed: 0.057 0.041 0.118 0.056 0.068 0.031 0.022 mean 0.0560 round 3 clock : 0.136 0.090 0.204 0.079 0.082 0.089 0.047 mean 0.1038 A2. widths. Two kinds, both reported: shot noise on the 7-qubit mean = sqrt(sum p(1-p)/N)/7 (binomial, per round); qubit-to-qubit spread = standard deviation of the 7 per-qubit values / sqrt(7). round 1 rest : 0.0643 shot +-0.0031 spread +-0.0084 round 1 pushed: 0.0646 shot +-0.0031 spread +-0.0112 round 1 clock : 0.1102 shot +-0.0039 spread +-0.0205 round 2 rest : 0.0641 shot +-0.0031 spread +-0.0050 round 2 pushed: 0.0656 shot +-0.0031 spread +-0.0128 round 2 clock : 0.1010 shot +-0.0038 spread +-0.0184 round 3 rest : 0.0584 shot +-0.0029 spread +-0.0079 round 3 pushed: 0.0560 shot +-0.0029 spread +-0.0119 round 3 clock : 0.1038 shot +-0.0038 spread +-0.0195 A3. the effect, per qubit, pooled over the 3 rounds (2700 shots per qubit per state): equator minus rest: per-qubit difference q7 +0.045 q22 +0.077 q27 +0.132 q57 +0.006 q62 +0.027 q72 -0.010 q104 +0.021 mean +0.0427 +- 0.0182 (spread across 7 qubits, 6 d.o.f.; t = 2.34) shot-noise-only width of the mean +-0.0028 qubits positive: 6 of 7 pushed minus rest: per-qubit difference q7 +0.005 q22 -0.012 q27 +0.053 q57 -0.009 q62 +0.016 q72 -0.051 q104 -0.003 mean -0.0002 +- 0.0119 (spread across 7 qubits, 6 d.o.f.; t = -0.02) shot-noise-only width of the mean +-0.0025 qubits positive: 3 of 7 equator minus pushed: per-qubit difference q7 +0.040 q22 +0.089 q27 +0.080 q57 +0.015 q62 +0.011 q72 +0.042 q104 +0.024 mean +0.0429 +- 0.0116 (spread across 7 qubits, 6 d.o.f.; t = 3.71) shot-noise-only width of the mean +-0.0028 qubits positive: 7 of 7 ratio equator/rest of the 7-qubit mean: 1.69 +- 0.32 (delete-one-qubit jackknife, so it carries the qubit-to-qubit spread); range when one qubit is dropped: 1.45 .. 1.88 ratio per round: round 1 1.71, round 2 1.57, round 3 1.78 pushed/rest ratio: 0.997 ratio equator/rest, round 1: 1.7136; all 3 rounds pooled: 1.6856 (the ratio differs from round to round; it is not a per-round constant) significance across qubits: mean paired rise / its spread-based width gives t = 2.34 on 6 d.o.f.; the 95 % two-sided critical value is 2.447, so on qubit-to-qubit spread alone the pooled rise is NOT past it. Every single qubit except q57 and q72 is far past its own shot noise (A6); it is the unevenness between qubits that keeps the pooled test marginal. pushed minus rest, pooled mean difference: -0.00021164 A4. WHAT STATE ARE THE NEIGHBOURS IN WHEN THE PUSHED QUBITS ARE READ? (read-1 fraction of the 22 neighbours) round 1: rest 0.036 / pushed 0.094 / clock 0.939 round 2: rest 0.031 / pushed 0.099 / clock 0.937 round 3: rest 0.032 / pushed 0.095 / clock 0.938 -> in the "equator" arm the neighbours end in |1> (about 94 % read 1): their two half-turns (rx(pi/2) before the train, rx(pi/2) after) make one full turn. They sit on the equator only DURING the train, and are excited when the readout happens. In the other two arms they are in |0> at the readout (rest, pushed 10x: the pushed arm shows ~9 % reading 1 from gate error). A5. shot-level check (same shots, no model): pushed-qubit read-1 rate versus how many of ITS OWN neighbours read 1 in the same shot. rate (shots) by number of own neighbours that read 1; pooled over the 3 rounds and, for each qubit, over its neighbours. rest : all neighbours read 0: 0.048 (17069 shots) all neighbours read 1: 0.357 (14 shots) some read 1: 0.192 (1817 shots) pushed: all neighbours read 0: 0.040 (14056 shots) all neighbours read 1: 0.449 (78 shots) some read 1: 0.120 (4766 shots) clock : all neighbours read 0: 0.059 (17 shots) all neighbours read 1: 0.099 (15460 shots) some read 1: 0.130 (3423 shots) reading: even in the REST arm a qubit reads 1 about four times more often in shots where one of its neighbours also reads 1. The readouts are correlated between neighbours (read-out crosstalk or a shared real excitation; this data cannot tell which). Only the equator arm has neighbours in |1> at the readout, so "neighbours on the equator during the train" and "neighbours in |1> at the readout" are confounded in this design. A simple crosstalk-by-neighbour-reading rule fitted on the rest arm would predict more than the 0.099 seen in the equator arm, so crosstalk alone is not a quantitative match either. A6. per qubit, equator minus rest (pooled): the excess is not uniform. Qubits with a rise > 2 pooled-shot widths: q7 rest 0.063 equator 0.108 diff +0.045 +-0.008 (+5.9 widths) q22 rest 0.054 equator 0.131 diff +0.077 +-0.008 (+9.9 widths) q27 rest 0.069 equator 0.201 diff +0.132 +-0.009 (+14.5 widths) q57 rest 0.079 equator 0.085 diff +0.006 +-0.007 (+0.8 widths) q62 rest 0.051 equator 0.078 diff +0.027 +-0.007 (+4.0 widths) q72 rest 0.086 equator 0.076 diff -0.010 +-0.007 (-1.3 widths) q104 rest 0.034 equator 0.056 diff +0.021 +-0.006 (+3.8 widths) ==================================================================================================== RESULT B -- a neighbour (q0) prepared in |2> shifts the precession rate of a spectator (q9). ==================================================================================================== B0. sign bookkeeping (simulated from each day's circuit): d angle(H) / d(clock phase) = -1 on 28 Sep, +1 on 4 Oct. The two days used different final rotations, so the SAME physical shift reads with REVERSED sign in angle(H). Below, "rate" = clock phase per microsecond in a common convention (positive = |1> gains phase relative to |0> at that rate, i.e. an rz(+) ), obtained as angle(H) rate / the number above. This assumes rx/rz mean the same rotations on both days (same compiler and gate set). The sign agreement across days does not depend on that assumption. B1. 28 SEPTEMBER (UTC 2026-09-29T00:58:59Z .. 2026-09-29T00:59:09Z = 20:58-20:59 EDT on 28 Sep): emitter q0 at rest (E0) or in |2> (E2: rx(pi), then a tuned 1<->2 pulse), clock q9 (next door; all other qubits idle), waits 0.5 1 2 3 4.5 6 us, hands X and Y, 1000 shots (E0) / 2500 shots (E2). 24 tasks, ids in the json. window wrapped angle unwrapped +-width rate (deg/us, common convention) amplitude ratio |H(E2)|/|H(E0)| 0.5 +27.92 +27.92 2.88 -55.85 +- 5.76 1.05 1.0 +57.59 +57.59 2.82 -57.59 +- 2.82 1.04 2.0 +108.27 +108.27 2.97 -54.13 +- 1.49 0.95 3.0 -179.87 +180.13 3.22 -60.04 +- 1.07 0.87 4.5 -79.93 +280.07 3.64 -62.24 +- 0.81 0.64 6.0 +24.62 +384.62 3.48 -64.10 +- 0.58 0.75 straight line through the origin, all 6 windows: slope -62.10 deg/us +- 0.41 (shot noise only) chi2 = 48.1 for 5 d.o.f. -> width scaled by sqrt(chi2/dof): +- 1.27 28 Sep slope magnitude: 62.104 deg/us; shot-noise width: 0.40905 deg/us NOTE the per-window rates are NOT constant: -55.8, -57.6, -54.1, ... -64.1. The straight line fits poorly (chi2/dof = 9.6). They rise with window length although the emitter is decaying out of |2> (its lifetime is not computed here), which would make them fall by ~20 % by 6 us. B2. 4 OCTOBER, run "triA" (tasks sent 2026-10-04T17:13:24Z .. 2026-10-04T17:13:40Z, 13:13 EDT): emitter q0 at rest / |1> / |2>, EVERY other qubit (106) a clock, windows 0.5 1 2 4 6 us, 900 shots. (q0 -> |2>: rx(pi) then the 1<->2 pulse, amplitude re-tuned on 4 Oct, not the 28 Sep amplitude.) Hand Y of this run also applies the swap-then-flip readout to q0 at the end. q9 with q0 in |2>, minus q0 at rest (same window): angle (deg), +-width, rate = angle/window (deg/us) 0.5 us -18.05 2.70 -36.09 +- 5.40 amplitude ratio |H(|2>)|/|H(rest)| 1.02 1.0 us -34.64 2.77 -34.64 +- 2.77 amplitude ratio |H(|2>)|/|H(rest)| 0.99 2.0 us -82.86 4.28 -41.43 +- 2.14 amplitude ratio |H(|2>)|/|H(rest)| 0.54 4.0 us +64.94 4.17 +16.23 +- 1.04 amplitude ratio |H(|2>)|/|H(rest)| 0.76 6.0 us -9.61 5.94 -1.60 +- 0.99 amplitude ratio |H(|2>)|/|H(rest)| 0.64 4 Oct q9, q0 in |2>, 0.5 us window: -36.091 deg/us 4 Oct q9, q0 in |2>, 1 us window: -34.644 deg/us 4 Oct q9, q0 in |2>, 2 us window: -41.432 deg/us The 4 us and 6 us windows (+16.2, -1.6) do not continue the trend: at a steady -36 deg/us the angle would be about -145 deg at 4 us and -215 (= +145 wrapped) at 6 us. They are shown above as read; the first three windows are the only stretch where a steady shift is visible. At 2 us the clock keeps only ~0.54 of the rest amplitude (q0_last, B3: 0.56) while at 0.5 and 1 us it keeps 1.0: not explained here (a shot-to-shot spread of the shift would do it). inverse-variance mean of 0.5/1/2 us: -38.64 +- 1.62 deg/us (windows are not independent of q0 decay; shot noise only) for comparison, q0 in |1>: 0.5 us +0.5 1 us +5.9 2 us +1.7 4 us -2.5 6 us +1.2 (small: |1> hardly moves q9, |2> moves it a lot) q1 (other next-door qubit), q0 in |2>: 0.5 us -27.5 1 us -19.4 2 us -21.0 4 us -19.5 6 us -13.9 101 clocks 3+ links from q0, q0 in |2>, 1 us window: median +1.06 deg/us, quartiles -1.9 .. +2.9 (the shift is local) B3. 4 OCTOBER, run "q0_last" (sent 2026-10-04T16:52:19Z .. 2026-10-04T16:52:24Z, 20 minutes BEFORE triA; same pair, one 2 us window, q0 gets no end-of-circuit readout pulses): q9, q0 in |2> minus rest: angle -69.88 +- 4.28 over 2 us = -34.94 +- 2.14 deg/us (hand 0.56); triA at 2 us gave -41.43 +- 2.14 4 Oct q0_last q9: -34.942 deg/us; width: 2.138 deg/us B4. SAME SPECTATOR / NEIGHBOUR PAIR BOTH DAYS? 28 Sep: emitter q0, clock q9. 4 Oct (triA, q0_last): emitter q0, clock q9. Are q0 and q9 coupled on the chip graph? True. -> YES, the same pair. The comparison is valid as a comparison of one pair, with the differences listed in B5. like-for-like (windows 0.5, 1, 2 us, inverse-variance mean of per-window rates): 28 Sep -54.93 +- 1.28 4 Oct -38.64 +- 1.62 ratio 4 Oct / 28 Sep = 0.70; difference +16.3 +- 2.1 deg/us (8 widths, shot noise only) the through-origin slope (the "-62.1") vs the 4 Oct 0.5-2 us mean: -62.1 vs -38.6 B5. things that differ between the two days and are NOT controlled (so a change between the days cannot be assigned to the coupling alone): - the 1<->2 pulse amplitude was re-tuned on 4 Oct, so how completely q0 reached |2> may differ. The |2> population of q0 needs a readout-error model and is NOT COMPUTED here. The 28 Sep run has no calibration for q0, so its |2> population is unknown here. - 28 Sep: q9 the only clock. 4 Oct: 106 clocks driven at once (the effect of the number of simultaneous clocks on the shift is not measured here). - the window ladder and the fit: 28 Sep is a straight-line slope over 0.5-6 us; 4 Oct per-window values over 0.5-2 us, then the later windows break the trend. - Cepheus is recalibrated between the days (device parameters drift; both days' qubit frequencies were not logged side by side). B6. CONTEXT, run "triF" (4 Oct 17:37-17:44 UTC, 3 rounds): 7 other emitter/neighbour pairs, emitters all held at rest or in |2>. q0/q9 are not in this run. rate = angle of (neighbour with emitter in |2> relative to emitter at rest) / window, deg/us; mean over 3 rounds; each neighbour paired with its own emitter. emitter neighbour 0.25 us 0.5 us 1 us 2 us (mean over 3 rounds; +- = spread of the 3 rounds at 1 us) q57 q66 -71.4 -69.8 -72.8 -78.9 +-3.4 q27 q36 -66.0 -56.5 -60.0 -58.8 +-3.4 q72 q73 -46.5 -54.8 -31.0 -38.3 +-4.6 q22 q23 -3.5 -20.2 -22.4 -21.8 +-1.1 q27 q28 -32.6 -28.6 -20.3 -16.2 +-1.4 q27 q18 -11.4 -19.4 -17.8 -13.9 +-2.6 q7 q16 -20.0 -18.1 -17.2 -12.8 +-2.1 q57 q58 -19.3 -20.8 -16.9 -19.1 +-3.1 q62 q71 -55.2 -28.4 -16.4 -38.4 +-21.7 q62 q53 -1.2 -15.2 -12.7 -12.8 +-5.1 q57 q56 -3.0 -3.4 -10.3 -7.9 +-5.6 q72 q81 -2.6 -12.8 -6.4 -3.2 +-1.3 q104 q103 -6.2 +2.1 -3.8 -2.9 +-2.9 q104 q105 -18.3 -5.9 -2.6 -4.9 +-2.3 q57 q48 -2.5 +0.4 -0.5 -7.0 +-5.6 q7 q6 +9.0 -8.5 -0.4 -0.2 +-3.1 q22 q21 +32.1 -2.9 -0.1 -9.3 +-9.6 q62 q61 +7.7 +0.9 +3.5 -2.3 +-11.3 q22 q31 +20.4 +17.8 +13.7 +8.4 +-1.1 q22 q13 +50.4 +24.8 +18.1 +11.1 +-8.2 q72 q63 +76.3 +26.6 +22.2 +23.5 +-6.1 q104 q95 +26.1 +22.4 +30.9 +30.0 +-2.8 1 us window over the 22 pairs: median -6.4, most negative -72.8, most positive +30.9; pairs below -20 deg/us: 5. (q0-q9: -34.6; 2 of the 22 pairs are stronger in magnitude, 5 have the reverse sign.) Shifts of 30-70 deg/us from an emitter in |2> therefore appear on other pairs too, and the sign is not always negative: sign and size are properties of the pair.