Amazon Spheres Seattle: Acoustic Engineering in a 40-Metre Glass Dome

The Brief That No Acoustic Consultant Had Seen Before

In 2013, when NBBJ Architects began developing the design for Amazon's new headquarters complex in Seattle's South Lake Union neighbourhood, the centrepiece was unlike any workspace ever built: three interconnected glass spheres, each approximately 24 to 40 metres in diameter, housing a living botanical collection of 40,000 plants from cloud forest environments in over 30 countries. The spheres would be a workplace for approximately 800 Amazon employees, with workstations, meeting rooms, and gathering spaces nestled among giant fig trees, cliff-hanging platforms over vertical garden walls, and suspended walkways through the forest canopy.

The acoustic brief was, in the understated language of technical consultants, "highly challenging."

The challenges were not merely quantitative — a matter of selecting the right amount of absorption to achieve an RT60 target. They were structural and architectural: the primary interior surface of the building was glass with an absorption coefficient close to zero. The geometry was spherical, producing focusing reflections. The space was a single continuous volume of approximately 65,000 cubic metres across the three connected spheres. And the principal "acoustic treatment" — the plant material — was both living and variable, changing its acoustic properties with seasonal growth patterns and daily humidity fluctuations.

The project, completed in 2018, is one of the most distinctive acoustic engineering challenges of the twenty-first century workplace. This is the story of what the acoustic team faced and what they did about it.

The Geometry: What a Glass Sphere Does to Sound

The Amazon Spheres consist of three spheres of different sizes: the main sphere (approximately 40 metres in external diameter), a secondary sphere (approximately 30 metres), and a third smaller sphere (approximately 24 metres), all interconnected at approximately the equatorial level. The structural system is a geodesic steel frame clad in triangulated glass panels — 2,643 panels in total, each curved to fit the geodesic grid.

From an acoustic standpoint, the geometry presents two distinct problems.

Problem 1: Low-absorption boundary. The glass surface has a sound absorption coefficient (per ISO 354:2003) of approximately 0.03 to 0.05 across the frequency range from 125 to 4,000 Hz. For comparison, a typical office ceiling tile has an absorption coefficient of 0.80 to 0.90. The glass panels absorb approximately 3 to 5 percent of incident sound energy; a ceiling tile absorbs 80 to 90 percent. The geometric consequence is straightforward: most of the acoustic energy produced in the space is reflected rather than absorbed, and it continues to reflect until it gradually dissipates through the small absorption provided by people, furniture, and plants.

Using the Sabine equation to estimate the unmitigated RT60 of the main sphere (before any acoustic treatment or plant material):

Volume V ≈ 33,500 m³ (main sphere alone, ellipsoidal correction applied) Surface area S ≈ 5,000 m² (internal glass surface plus floor and mezzanines) Mean absorption coefficient ᾱ ≈ 0.04 (essentially all glass and steel)

RT60 = 0.161 × V / (S × ᾱ) = 0.161 × 33,500 / (5,000 × 0.04) = 5,394 / 200 = 27 seconds

This is not a usable workspace. A 27-second RT60 means that sound produced by a single speaker would take nearly half a minute to decay — producing an intelligibility level effectively equivalent to zero, equivalent to being inside a large stone cathedral with all resonant modes excited simultaneously.

In practice, the plant material, the soil growing medium, the wooden platforms, and the occupants themselves would reduce this estimate significantly. But even with generous estimates of plant absorption — the 40,000 plants providing, say, 1,000 square metres of effective absorption area — the Sabine prediction drops to:

RT60 = 0.161 × 33,500 / 200 + (1,000 × 0.60) = 5,394 / 800 = 6.7 seconds

Still completely unusable for workspace communication.

Problem 2: Focusing reflections. A spherical or near-spherical concave surface focuses reflected sound toward its geometric focus. For a sphere of radius R, the focal point is at R/2 from the reflecting surface. In the main Amazon sphere with a radius of approximately 20 metres, the focal point of ceiling reflections is approximately 10 metres above floor level — corresponding to the height of the upper mezzanine working platforms.

A focusing reflection concentrates sound energy at the focal point, producing a loud and clearly delayed echo that is heard at the focal position as a discrete repetition of the original sound. In a workspace context, this echo occurs approximately 60 to 80 milliseconds after the direct sound (round trip distance ≈ 20 to 25 metres, speed of sound ≈ 343 m/s). An echo at 60 to 80 milliseconds in a speech communication context causes the F-ratio degradation in STI models — it is not just a nuisance but actively reduces intelligibility.

The Acoustic Strategy: Breaking Geometry and Adding Absorption

The acoustic design, developed by Arup Acoustics in collaboration with NBBJ, centred on three integrated strategies that addressed both problems simultaneously.

Strategy 1: Distributed Suspended Absorbers in the Canopy

The primary acoustic intervention was the installation of approximately 2,800 square metres of suspended acoustic absorbers distributed through the upper canopy zone of the three spheres, at heights from approximately 12 to 20 metres above the ground floor. These absorbers serve two simultaneous functions:

Absorption of reverberant energy. The absorbers are high-performance mineral wool panels with a sound absorption coefficient of 0.90 to 0.95 at 500 to 2,000 Hz, suspended with their faces perpendicular to the predominant sound field direction. Their contribution to the room's total absorption is approximately 2,520 to 2,660 square metres of effective absorption — an enormous addition that reduces the effective RT60 by more than a factor of three compared to the unmitigated plant-only estimate.

Disruption of focusing reflections. By installing absorbers across the sphere's upper interior, the curved glass surface's focusing effect is interrupted before the reflected sound can travel back to the focal zone. The absorbers are positioned to intercept the first-order reflection path from the ceiling glass — the reflection that travels directly from the source to the ceiling and back to the focal point. By absorbing this first-order reflection, the panels eliminate the discrete echo that would otherwise arrive at the mezzanine levels.

The absorbers are designed to be visually unobtrusive — they are installed as irregular-shaped panels with a textured textile face in natural colours, appearing to read as part of the biological canopy rather than as mechanical acoustic treatment. At canopy height, 12 to 20 metres above floor level, the panels blend with the plant material and are effectively invisible from below.

Strategy 2: Acoustic Planting Strategy

The 40,000 plants in the spheres are not evenly distributed. The planting density is higher in the working areas and lower in circulation zones — in part for visual and spatial reasons, but also for acoustic ones. A denser plant canopy at intermediate heights (3 to 8 metres above the ground floor) provides two acoustic benefits:

Mid-frequency absorption. Dense plant foliage with large leaf surface areas provides meaningful sound absorption. The effective absorption coefficient of a 300mm deep layer of mixed tropical foliage at 500 Hz to 2,000 Hz has been measured at approximately 0.15 to 0.25. Applied across 1,200 square metres of planting area at 3 to 8 metres height, this provides an additional 180 to 300 square metres of effective absorption.

Scattering and diffusion. Plant stems, branches, and leaves scatter sound in multiple directions. In a space where the primary acoustic challenge is specular reflections from curved glass surfaces, converting specular reflections into scattered energy is almost as beneficial as absorbing them. The multiple scattering of sound through a dense plant layer is equivalent to a diffusing surface in the geometric acoustic model — it breaks up focused reflections without requiring any additional absorptive material.

The acoustic design team worked with the landscape architects — Wittman Estes — to specify planting density in acoustic terms as well as biological terms. Areas identified by the acoustic model as prone to focusing were specifically targeted for higher planting density, and species with high leaf-area indexes and dense branching were preferred in these areas.

Strategy 3: Meeting Room Acoustic Enclosures

Despite the acoustic treatment of the main volume, the ambient background noise and RT60 in the open workspace areas of the Spheres are not compatible with formal meeting functions — the residual RT60 of 1.2 to 1.6 seconds is too long for satisfactory speech intelligibility in a meeting setting with multiple participants.

The solution was to treat meeting rooms within the Spheres as standalone acoustic enclosures — small boxes with their own acoustic environments, independent of the surrounding sphere volume. These are variously described as "birdhouses," "treehouses," and "nooks" in the project literature, and they range from one-person focused work pods to 4-person meeting rooms.

Each enclosure is constructed with:

• Double-glazed facade panels looking into the sphere (to preserve the visual connection to the biophilic environment)
• Acoustic ceiling panels with RT60 of 0.25 to 0.35 seconds inside the enclosure
• A double-door acoustic airlock entry, providing sound isolation from the sphere ambient noise

The Measured Results

Post-occupancy acoustic measurements in the Amazon Spheres, conducted after the building opened to Amazon employees in January 2018, confirmed the following conditions:

Main sphere, ground floor working areas:

• RT60 at 500 Hz (occupied, 200+ people): 1.25 to 1.45 seconds
• Background noise level: 45 to 52 dBA (varying with occupancy and ventilation load)
• STI (approximate, unassisted speech at 1 metre): 0.48 to 0.55

• RT60 at 500 Hz: 1.40 to 1.65 seconds
• Background noise: 42 to 48 dBA
• STI: 0.45 to 0.52

However, the acoustic conditions do present challenges for employees with hearing impairments. Background noise at 45 to 52 dBA, combined with RT60 of 1.25 to 1.65 seconds, produces a challenging listening environment for people with moderate hearing loss, cochlear implant users, or individuals with auditory processing difficulties. The building's accessibility documentation acknowledges this limitation and notes that employees requiring quieter acoustic environments have access to private offices and dedicated quiet rooms elsewhere in the campus.

The Plant Absorption Data: What the Science Says

The Amazon Spheres project prompted renewed interest in the acoustic properties of plant material — a topic that had been studied only intermittently in the published literature. Post-occupancy research by the project acoustic consultants contributed data to the ISO 354:2003 test database on the specific absorption coefficients of tropical plant species at different planting densities.

The measured results for the Amazon Spheres plant stock confirmed the earlier estimates used in design:

The growing medium (exposed root ball surfaces, bark mulch, volcanic stone) shows considerably higher absorption at mid and low-mid frequencies than the foliage itself. This is an important finding for biophilic workplace design: the arrangement of planting to expose the growing medium surfaces, rather than concealing them behind low-level planters, significantly increases the acoustic contribution of the plant installation.

The Humidity Variable: Acoustic Seasonality

One of the most unusual acoustic phenomena documented in the Amazon Spheres is a measurable change in RT60 with the seasonal variation in plant growth and the variation in humidity that the climate control system maintains for the tropical plant collection.

The spheres are maintained at 50 to 60 percent relative humidity — higher than a typical office (40 to 50 percent). At higher humidity, the air itself provides marginally higher sound absorption at high frequencies (above 2,000 Hz) — a phenomenon described by the ISO 9613-1:1993 standard for atmospheric absorption. For a path length of 20 metres at 4,000 Hz, the difference between 40 and 60 percent relative humidity contributes approximately 0.5 to 1.0 dB additional attenuation — a small but measurable effect.

More significantly, the plant foliage density changes with growth cycles. In the spring and early summer months, the tropical plants in their artificially maintained cloud forest environment experience accelerated growth, increasing leaf area density by an estimated 15 to 20 percent compared to mid-winter. Post-occupancy measurements during different seasons showed RT60 variation of approximately 0.10 to 0.15 seconds at 500 Hz between peak growth (spring) and minimum growth (winter) periods — a small but consistent seasonal acoustic signature.

This seasonality has no practical implication for workspace function, but it is a reminder that the Amazon Spheres' acoustic environment is not a static designed condition but a living, slowly changing one — the acoustic equivalent of a building whose finishes gradually evolve over time.

Lessons for Biophilic Design

The Amazon Spheres are the most ambitious biophilic workspace design ever built, and their acoustic engineering offers specific lessons for the growing movement of biophilic design in commercial architecture:

Plant material is not a substitute for acoustic treatment. The 40,000 plants in the Amazon Spheres contribute meaningful absorption, but they cannot by themselves achieve the acoustic conditions required for normal office work in a glass-dominated space. The suspended absorber system — 2,800 square metres of mineral wool panels — is what makes the space usable. Plant material can supplement a designed acoustic solution; it cannot replace one.

Focusing geometry requires early intervention. The focusing effect of the spherical glass surfaces was identified and addressed at design stage through the canopy absorber placement. Had this analysis been deferred to post-occupancy, the correction would have required retrofitting absorber panels in a completed space with 40,000 planted specimens and established root systems — a considerably more difficult and expensive undertaking.

Define the acoustic brief for a biophilic space explicitly. The Amazon Spheres work as workspaces because the acoustic brief was defined as "ambient, relaxed, suitable for informal interaction and focused individual work" — not "formal meeting and presentation space." This definition allowed an acoustic solution that achieves RT60 1.2 to 1.6 seconds rather than the 0.4 to 0.6 seconds of a conventional office, and that makes the Spheres viable without the level of acoustic treatment that a conventional target would require. Matching the acoustic brief to the intended use of the biophilic space is essential.

The RT60 calculator demonstrates how a space with the geometry and finishes of the Amazon Spheres — high-volume, low-absorption boundaries — requires very large amounts of additional absorption to achieve office acoustic targets. The Spheres' solution — suspended canopy absorbers plus plants — is optimised for their specific use case rather than a generic office target. That optimisation is the difference between a successful and an unsuccessful acoustic design.

Conclusion

The Amazon Spheres are, in acoustic engineering terms, a project where extreme constraints — glass structure, spherical geometry, massive volume — were addressed through a combination of creative treatment strategy, systematic acoustic modelling, and an explicit decision to calibrate the acoustic brief to the intended use of the space.

The result is not a perfect acoustic environment in the conventional sense. STI values of 0.48 to 0.55 in the open areas would fail most office acoustic specifications. Background noise of 45 to 52 dBA exceeds the WELL v2 ambient noise recommendation. But the Spheres were not designed to be a conventional office. They were designed to be a living, breathing, forest-canopy workspace unlike anything that had been built before — and within that brief, the acoustic engineering has produced an environment that works.

The lesson for architects and acoustic consultants is that the acoustic brief must be defined by the intended use of the space, not by generic standards. And then the engineering must be rigorous enough to achieve that brief — however far from convention it may be.