Transient acceleration events in LISA Pathfinder data: Properties and possible physical origin

We present an in depth analysis of the transient events, or glitches, detected at a rate of about one per day in the differential acceleration data of LISA Pathfinder. We show that these glitches fall in two rather distinct categories: fast transients in the interferometric motion readout on one side, and true force transient events on the other. The former are fast and rare in ordinary conditions. The second may last from seconds to hours and constitute the majority of the glitches. We present an analysis of the physical and statistical properties of both categories, including a cross-analysis with other time series like magnetic fields, temperature, and other dynamical variables. Based on these analyses we discuss the possible sources of the force glitches and identify the most likely, among which the outgassing environment surrounding the test-masses stands out. We discuss the impact of these findings on the LISA design and operation, and some risk mitigation measures, including experimental studies that may be conducted on the ground, aimed at clarifying some of the questions left open by our analysis.

Figure 1

Rendering of the LISA Technology Package. The rendering shows the two test-masses hosted inside their respective electrode-housings (some of the electrodes are not represented), and the vacuum chambers enclosing both test-masses and electrode housings. The picture also shows the high stability optical bench hosting all interferometric readouts, and many other features of the instrument, launch lock, UV-light based test-mass neutralizer, etc. that are not relevant here. [Copyright: $\mathrm{ESA}/\mathrm{ATG}$ medialab].

Figure 2

Left: example of an impulse carrying glitch. The picture shows: (green) the native data after low-pass filtering and background subtraction; (orange) the fitting template; (gray) the residual after subtraction of the template. Right: example of a fast, low impulse glitch. The figure shows data after the 0.5 Hz low-pass filtering.

Figure 3

Amplitude spectral density (ASD) $S_{\mathrm{\Delta }g}^{1/2}(f)$ of quasistationary noise, vs the frequency $f$, of the best noise run of February 2017. Green data points: ASD of glitch-free stretches only. Orange data points: ASD of residuals after glitch removal. Errors represent 68% confidence intervals.

Figure 4

Histogram of the waiting time $\mathrm{\Delta }T$ for ordinary runs (orange) and cold runs (blue). Straight lines, and associated shadowed areas, represent, respectively, the Bayesian fit to an exponential distribution, and the corresponding 68% confidence interval. For the cold runs, the fit is limited to data with $\mathrm{\Delta }T\le 0.2\mathrm{d}$. The rates from the Bayesian estimation for ordinary runs, and cold runs with $\mathrm{\Delta }T\le 0.2\mathrm{d}$, are, respectively, $\lambda =0.96_{-0.09}^{+0.11}{\mathrm{d}}^{-1}$ and $\lambda =32_{-2}^{+2}{\mathrm{d}}^{-1}$.

Figure 5

Glitch occurrence rate $\lambda$, during ordinary runs, vs time $t$ from launch, throughout the entire mission. Points are calculated by grouping glitches observed during runs the start times of which differ by less than a month. Vertical errors bars are Bayesian estimates assuming exponential distribution, and corresponds to 68.3% ($1\sigma$) likelihood (see Appendix pp2). Horizontal error bars correspond to the total duration of the considered epoch. The dashed line, and the associated gray shaded area, represent, respectively, the mean rate and its error from the Bayesian estimate. The blue shaded area indicates the epoch of cold runs.

Figure 6

Glitch occurrence rate $\lambda$ (left scale), and LTP bay temperature (right scale), during cold runs, as a function of time since the beginning of cooldown. Small temperature changes before day 12 were the result of adjustments of heaters settings aimed at stabilizing the system behavior. Reheating to ordinary conditions started at day 12.

Figure 7

Absolute impulse per unit mass $|\mathrm{\Delta }v|$, and duration $\mathrm{\Delta }$, of impulse carrying glitches fitted to templates in Eqs. (3)–(5). Upper left panel: the 81 positive impulse glitches observed during ordinary runs. Upper right panel: the 17 negative impulse glitches observed during ordinary runs. Lower left panel: the 306 positive impulse glitches observed during cold runs. Lower right panel: the 28 negative impulse glitches observed during cold runs. For reference, the gray dashed line represents, for any given duration, the amplitude of a glitch of the kind in Eq. (4) that would have $\mathrm{SNR}=3$. In the two upper panels, SNR is calculated for the lowest noise ordinary run of February 2017. In the lower panels, the line is calculated for the sensitivity of the cold runs. For a detailed analysis of the glitch SNR, see Fig. 9.

Figure 8

Absolute impulse per unit mass $|\mathrm{\Delta }v|$ (top), and duration $\mathrm{\Delta }$ (bottom) as a function of time from launch for the glitches of Fig. 7. Note that the apparent clustering corresponds to the different measurement runs. The blue shaded area corresponds to the epoch of cold runs.

Figure 9

Histogram of the SNR for all glitches of Fig. 7. The lowest observed value is $\mathrm{SNR}\sim 3$. The probability density refers to the logarithm of SNR.

Figure 10

Histogram of peak amplitude $|\mathrm{\Delta }{g}_{\max }|$ for all glitches of Fig. 7. The probability density is for the logarithm of the amplitude.

Figure 11

(upper) Histogram of the waiting time $\mathrm{\Delta }T$ for fast, low impulse glitches in cold runs. (lower) Histogram of $(\mathrm{\Delta }{v}_{\text{glitch}}^2/\mathrm{\Delta }{v}_{\mathrm{rms}}^2)$ for low impulse glitches. The orange line is the properly normalized distribution for a chi-square with one degree of freedom.

Figure 12

$-\mathrm{\Delta }{x}_{\mathrm{OMS}}(t)$ (black), $-\mathrm{\Delta }{x}_{\mathrm{GRS}}(t)$ (orange), and their difference (green) for an impulse-carrying glitch ($\mathrm{\Delta }v=22.1\mathrm{pm}/\mathrm{s}$, $\mathrm{\Delta }=7.82\mathrm{s}$). The negative signs on the first two data series have been used to show that the residuals in the difference are dominated by the large noise in $\mathrm{\Delta }{x}_{\mathrm{GRS}}(t)$. Also shown (dashed line) is the signal one would observe in $\mathrm{\Delta }{x}_{\mathrm{OMS}}(t)-\mathrm{\Delta }{x}_{\mathrm{GRS}}(t)$, if the glitch were due to a spurious signal in $n_{\mathrm{OMS}}(t)$.

Figure 13

An example of the $\mathrm{\Delta }{\mathrm{\Omega }}_{\varphi }(\delta t)$ series for one of the very few glitches with significant associated torque. The value in $\delta t=0$ is the result of the fit for native data. The arrows indicate the interval excluded for the sake of error evaluation.

Figure 14

Absolute lever arm $|r_{\varphi }|$, as defined in Eq. (10), as a function of absolute impulse $|\mathrm{\Delta }v|$ (left) and duration $\mathrm{\Delta }$ (right), for impulse-carrying glitches of ordinary runs. The black data points represent the glitches for which the lever arm has been found to be significantly different from zero. The green horizontal segments represent the upper bound for $|r_{\varphi }|$ for glitches for which $|\mathrm{\Delta }{\mathrm{\Omega }}_{\varphi }|\le \delta {\mathrm{\Omega }}_{\varphi }$. For 9 of the 98 impulse-carrying glitches of the ordinary runs, we were not able to perform the analysis on the $\mathrm{\Delta }{\gamma }_{\varphi }(t)$ data series. The gray horizontal line refers to the 11 mm “electrical” reference, as described in text.

Figure 15

Schematic representation of the GRS, horizontal section. The yellow square is the Au-Pt test mass. The orange hollow square represents the electrode housing, whose mid section carries four symmetric holes, one of which is the input port for the laser beam (in red). The green part is the section of a titanium structure that holds the various elements together, while in gray is the specially shaped tungsten gravitational balance mass. The brown circle is the section of the vacuum chamber, while the cyan rectangle represent the optical window to transmit the laser beam.

Figure 16

Blue line and left scale: pressure in one of the two GRS during cooldown, after about 24 h pumping at about $115°\mathrm{C}$ for vacuum preparation (bakeout) on ground. Red line and right scale: temperature in the test facility. Vacuum preparations were performed by inserting the GRS, with its venting valve open, inside a wider vacuum chamber. The pressure and temperature shown in the figure are those of this wider chamber. The time origin is set at the end of the bakeout phase. (Data courtesy OHB Italia)