A search for super-imposed oscillations to the primordial power spectrum in Planck and SPT-3G 2018 data

This paper analyzes Planck and SPT-3G 2018 data to search for super-imposed oscillations in the primordial power spectrum, finding that combined datasets provide tighter constraints on oscillation amplitudes and reveal improved fits with significant statistical gains compared to individual datasets.

The Gist

Imagine the universe as a giant, cosmic drum that was struck at the very moment of its birth (the Big Bang). When that drum was hit, it didn't just make a single sound; it created a complex, vibrating pattern of ripples that spread out across space. In cosmology, we call this the Primordial Power Spectrum.

For decades, scientists have looked at the "echo" of this drumbeat in the Cosmic Microwave Background (CMB)—the oldest light in the universe. The standard theory says this echo should look like a smooth, predictable curve, much like a perfectly smooth hill. This is the "Power Law" model.

However, some scientists suspect that the drum might have had a few extra bumps or ripples on its surface, creating oscillations (wiggles) in that smooth curve. These wiggles could be the fingerprints of new physics from the very first fraction of a second after the Big Bang.

This paper is a detective story where the authors try to find these hidden wiggles using two different "ears" to listen to the universe: Planck and SPT-3G.

The Two Listeners

1. Planck (The Wide-Angle Lens): This was a satellite that mapped the entire sky. It's like a high-quality camera that took a beautiful, wide photo of the universe. It sees the big picture very well, but it's a bit blurry when you zoom in on the tiny details (high-resolution).
2. SPT-3G (The Zoom Lens): This is a telescope sitting at the South Pole. It can't see the whole sky, but it has incredible zoom. It can see the tiny, high-frequency details that Planck misses. It's like looking at a grain of sand under a microscope.

The Search for the "Wiggles"

The authors tested five different theories about what these wiggles might look like:

• Linear vs. Logarithmic: Are the wiggles spaced out evenly like the rungs on a ladder (linear), or do they get closer together like the notes on a piano scale (logarithmic)?
• Constant vs. Damped: Do the wiggles go on forever with the same strength, or do they fade away (damp) after a certain point?
• Tilted: Does the strength of the wiggles change as you look at different scales?

The Detective Work

The team combined data from both the "Wide-Angle" (Planck) and the "Zoom" (SPT-3G) telescopes. Here is what they found, explained simply:

1. The "Better Fit" Mystery
When they looked at the data, they found that adding these wiggles to their models made the picture fit the data slightly better than the smooth "Power Law" model. It's like trying to fit a puzzle piece: the smooth hill fits okay, but a hill with a specific bump fits just a little bit better.

• The Catch: In science, adding a "bump" (an extra parameter) usually costs you points because it makes the theory more complicated. So, while the wiggles fit the data better, they aren't statistically "proven" yet because the improvement isn't huge enough to overcome the penalty for complexity.

2. The Power of Teamwork
The most exciting part of the paper is what happened when they combined the two telescopes.

• Planck alone said, "I see a wiggle here, but I'm not 100% sure."
• SPT-3G alone said, "I see a wiggle there, but I can't see the whole picture."
• Together: When they combined the data, the "wiggle" became much clearer. The combined data put much tighter constraints on how big these wiggles could be. It's like two people looking at a blurry image from different angles; when they combine their views, the image snaps into focus.

3. The "Gaussian" Surprise
One specific type of wiggle—a "damped" one that fades away like a sound fading into the distance—showed a very strong signal when Planck and SPT-3G were combined. The improvement in the fit was significant (a $\Delta\chi^2$ of about -17.5). This suggests that if there are wiggles, they might be localized (happening in a specific range of sizes) rather than stretching across the whole universe.

Why Does This Matter?

Think of the universe's history as a song. The standard model says it's a simple, smooth melody. These "oscillations" suggest there might be a complex harmony or a specific instrument playing a unique riff that we haven't heard clearly yet.

• If we find them: It would prove that the universe didn't just expand smoothly; something dramatic happened in the first split second (like a "sharp turn" in the universe's expansion or a specific type of quantum field) that left these ripples.
• If we don't: It confirms that the universe is incredibly simple and smooth, which is also a profound discovery.

The Bottom Line

This paper is a "dress rehearsal" for the future. The authors used the best data we have right now (Planck and SPT-3G) to hunt for these cosmic ripples. They found some promising hints, especially when combining the two telescopes, but they need more sensitive data to be sure.

They are essentially saying: "We have a hunch that the universe has a secret pattern. Our current tools (Planck + SPT-3G) have given us a stronger clue than ever before, but we need the next generation of telescopes (like the Simons Observatory) to finally confirm if the music has a hidden rhythm."

Technical Summary

Here is a detailed technical summary of the paper "A search for super-imposed oscillations to the primordial power spectrum in Planck and SPT-3G 2018 data."

1. Problem Statement

The standard $\Lambda$CDM cosmological model assumes a featureless, power-law primordial power spectrum (PPS) for scalar fluctuations. While this model fits Cosmic Microwave Background (CMB) data well, high-precision observations from the Planck satellite have revealed slight deviations ($\Delta\chi^2 \sim -[10-15]$) that suggest the presence of "super-imposed oscillations" in the PPS. These oscillations could arise from early-universe physics such as non-vacuum initial states, axion monodromy, or sharp turns in the inflationary potential.

However, Planck's angular resolution and sensitivity limit the ability to probe these features at high multipoles (small scales). Ground-based experiments like the South Pole Telescope (SPT-3G) offer higher resolution and sensitivity in these regimes. The challenge lies in combining Planck and SPT-3G data to test these oscillatory models without introducing biases from potential tensions between datasets (e.g., the known tension between Planck and ACT DR4) or double-counting information due to overlapping multipoles.

2. Methodology

Models and Templates
The authors investigate five specific templates for super-imposed oscillations in the primordial curvature power spectrum $P(k)$:

1. Global Linear Oscillations: Constant amplitude with a power-law tilt ($n_{lin}$) in $k$.
2. Global Logarithmic Oscillations: Constant amplitude, periodic in $\ln(k)$.
3. Tilted Linear Oscillations: Linear oscillations where amplitude varies as a power law in $k$ (mimicking Effective Field Theory predictions).
4. Damped Linear Oscillations: Linear oscillations modulated by a Gaussian envelope (localized in $k$).
5. Damped Logarithmic Oscillations: Logarithmic oscillations modulated by a Gaussian envelope.

These models introduce 3 to 5 additional parameters beyond the standard 6-parameter $\Lambda$CDM model (amplitude $\alpha$, frequency $\omega$, phase $\phi$, tilt $n_{lin}$, Gaussian center $\mu$, and width $\beta$).

Datasets

• Planck 2018 (P18): Temperature and polarization (TP) likelihoods (binned Plik). Lensing likelihood was excluded to match the feature analysis of Ref. [1].
• SPT-3G 2018: Temperature and polarization auto/cross-correlation data from 90, 150, and 220 GHz channels.
• Combined Dataset (P18+SPT3G): A conservative combination was constructed to avoid covariance issues. The datasets were "glued" such that there is no multipole overlap:
  • Planck provides data for $\ell_{TT} < 2000$, $\ell_{TE} < 1400$, $\ell_{EE} < 1000$.
  • SPT-3G provides data for $\ell_{TT} > 2000$, $\ell_{TE} > 1400$, $\ell_{EE} > 1000$.
  • The optical depth prior ($\tau$) from Planck was applied to the SPT-only analysis to break degeneracies.

Statistical Approach

• Sampler: The authors used PolyChord for nested sampling to handle the multi-modal nature of the posterior distributions and to compute Bayesian evidence.
• Optimization: The BOBYQA minimizer was used to find local minima and best-fit points around the likelihood peaks.
• Priors: Standard priors were used for cosmological parameters. For oscillation parameters, priors were chosen to match previous Planck feature searches, with specific ranges for amplitude, frequency, and Gaussian envelope parameters.

3. Key Contributions

1. First Joint Analysis with SPT-3G: This is the first study to search for super-imposed oscillations using the 2018 SPT-3G public data in conjunction with Planck.
2. Testing Modulated Oscillations: The paper tests Gaussian-modulated (damped) oscillations, including the first application of modulated logarithmic oscillations to current CMB data.
3. Conservative Data Combination: The authors developed a robust method to combine Planck and SPT-3G data without a full covariance matrix by strictly separating multipole ranges, ensuring that improvements in fit are not artifacts of double-counting.
4. EFT Constraints: The study specifically constrains the power-law tilt ($n_{lin}$) of linear oscillations, a parameter relevant to Boundary Effective Field Theory (EFT) models.

4. Key Results

Fit Improvements ($\Delta\chi^2$)
All five models provided an improved fit to the data compared to the standard power-law spectrum, though none were statistically preferred in a Bayesian sense due to the penalty for extra parameters.

• Planck Only: Linear undamped and tilted models showed $\Delta\chi^2 \approx -11.8$; Logarithmic undamped showed $\approx -9.3$.
• SPT-3G Only: Logarithmic oscillations showed a significant improvement ($\Delta\chi^2 \approx -12.0$), while linear oscillations showed a smaller improvement ($\approx -7.0$).
• Combined (P18+SPT3G): The combined dataset yielded the strongest constraints and largest improvements.
  • Logarithmic (Damped): $\Delta\chi^2 \approx -17.5$.
  • Linear (Damped): $\Delta\chi^2 \approx -14.7$.
  • Logarithmic (Undamped): $\Delta\chi^2 \approx -14.3$.

Parameter Constraints

• Amplitude ($\alpha$): The combined dataset provided tighter constraints on the oscillation amplitude than either dataset alone. For example, for linear undamped oscillations, the 95% CL upper limit on $\alpha$ improved from $\sim 0.036$ (Planck) to $\sim 0.031$ (Combined).
• Tilt ($n_{lin}$): For linear oscillations with a power-law amplitude, the combined data constrained the tilt to $n_{lin} = 0.38 \pm 0.30$ (68% CL). Neither Planck nor SPT-3G alone could meaningfully constrain this parameter.
• Frequency ($\omega$): SPT-3G data preferred frequencies around $\log_{10}\omega \sim 1.5$ for logarithmic oscillations, which did not perfectly overlap with Planck's preferred frequencies for constant amplitude models. However, for Gaussian-modulated models, the preferred parameter ranges overlapped, leading to the large $\Delta\chi^2$ in the combined analysis.

Reconstructed Power Spectrum
The reconstructed primordial power spectrum using the combined data showed that SPT-3G data effectively constrains the spectrum at small scales (high $k$), reducing the uncertainty in the reconstructed features compared to Planck alone.

5. Significance and Conclusions

• Validation of SPT-3G: The study confirms that SPT-3G data is highly consistent with Planck in the power-law $\Lambda$CDM regime and provides a powerful new window for testing primordial features at high multipoles ($\ell > 2000$).
• Synergy: The combination of Planck and SPT-3G data is crucial. While individual datasets show improvements over the power law, the combined dataset significantly tightens constraints on oscillation amplitudes and allows for the measurement of the spectral tilt of the oscillation amplitude ($n_{lin}$).
• Future Prospects: The findings suggest that upcoming high-resolution CMB experiments (e.g., Simons Observatory, CMB-S4) will be able to further test these models. Specifically, future polarization data will be essential to distinguish between primordial features and phenomenological effects like the lensing amplitude anomaly ($A_L$), which the modulated linear oscillation template mimics.
• No Strong Evidence Yet: Despite the improved $\chi^2$, the Bayesian evidence does not favor these complex models over the simple power law, indicating that the current "features" in the data may still be statistical fluctuations or systematic effects rather than definitive evidence of new physics.

In summary, this paper establishes a rigorous framework for combining ground-based and satellite CMB data to search for primordial features, demonstrating that while current data shows intriguing hints of oscillations, higher sensitivity and resolution are required to confirm them.