© 2013 Jonathan Ochshorn
From the Critique of Milstein Hall introduction: Milstein Hall at Cornell University, designed by Rem Koolhaas and OMA, is an interesting building, in some ways an amazing building, and, by virtually any conceivable objective criterion, a disaster. That something amazing can simultaneously be a disaster is hardly a paradox. In fact, disasters are often amazing, and our amazement often increases proportionally with the range and scope of the disaster.
I will not be criticizing the visual appearance of this building, or making judgments about its subjective, aesthetic merit. I personally find the building interesting, and its underlying formal rationale provocative and compelling. But I am not particularly qualified to render such judgments, and other authorities or connoisseurs of architectural taste may well disagree. What follows, instead, is an objective critique of Milstein Hall, looking at the building in some detail from a series of different points of view, none of which are driven by aesthetic considerations.
From the Nonstructural Failure introduction: "The architect (qua artist) is not so much "help[ing] us along the heroic journey of our own lives" but rather creating, out of thin air, a heroic journey for herself: leaving the world of safe, predictable constructions; proposing buildings that have both the appearance and the reality of danger... and returning in glory from this confrontation with the agents of conformity (whether owners, users, public officials) with the building constructed."
Keeping water out of buildings and controlling heat flow (energy) are basic requirements of building design, if not architectural design. The idea that the latter is endangered by the former is certainly a legitimate fear — given the increasingly perverse interest in understanding architecture as a heroic project (see Mondrian quote, and the discussion of heroic complacency, in the Introduction) — yet the outcome of such an attitude is always disheartening for both the architect and her client.
What follows is not an all-inclusive list of such nonstructural water and thermal control failures. I have not been given official access to such information, so the items that follow are based only on my random observations of the building:
Attachment of stone veneer panels, based on approved shop drawings, differed considerably from contract document details — with unintended consequences for thermal bridging and heat loss. Specifically, the virtually continuous clip angles now used to support the stone veneer interrupt rigid insulation panels, creating a highly conductive pathway (i.e., a thermal bridge, Figure 1) for heat loss or heat gain.
A better strategy, even using long shelf angles, is to detail them so that they "stand off" from the structural slab (Figure 2). In this way, the thermal control layer (insulation) can extend behind the shelf angle, minimizing thermal bridging.
In addition, workers could be seen cutting away at curtain wall elements in an ad hoc manner when stone panels were not fitting properly at the intersection with the exterior wall of Sibley Hall (Figure 3). Such ad hoc in-field modification provides evidence that the complexity of the building was not adequately researched and that neither contract documents nor shop drawings carefully and thoroughly accounted for this complexity.
Bollards were installed above below-grade, heated, spaces in Milstein Hall in order to protect pedestrians from vehicular zones situated directly above these heated spaces. Inexplicably, these bollards are attached, not to the concrete sidewalks on which they appear to sit, but to the structural concrete for the underground portion of Milstein Hall below. Because of this, the bollards interrupt the continuity of both waterproofing and insulation layers that have been installed above the structural concrete slab to which the bollards are attached.1 As a result, there is a risk that any vehicle-bollard collision (Figure 4, left) could dislodge the waterproofing membrane which is flashed onto the surface of the bollard below. Because all the connections are below grade, it would be impossible to know whether any damage has occurred until water leakage, or its many manifestations, appears in the space below.
The discontinuous insulation layer results in thermal bridging, as heat from the spaces below can be conducted directly through the concrete slab and bollards above, which have interrupted all layers of rigid insulation placed over the structural slab. This shows up, quite artistically, as a series of almost perfect circles surrounding each of the bollards after it snows (Figure 4, right).
Seismic joints between Milstein Hall and existing buildings do not seem to be designed to accommodate seismic movement in all possible directions. The detailed cross section of the seismic joint in the Milstein Hall construction documents (Figure 5) shows a flexible connections that spans approximately 5 inches, thereby allowing Milstein Hall to move laterally under seismic loads without literally hitting Sibley or Rand Halls. However, movement of Milstein Hall perpendicular to the axes of seismic induced "pounding" does not appear to be accounted for in the detail; in such a case (which, given the fact that such joints are present in two orthogonal directions, is certain to occur during any seismic event), tearing of the flashing and/or seismic joint material perpendicular to its expansion-contraction axis is likely.
The seismic joint, as built, appears to be different from the detail, in that no curved profile is evident (Figure 6). It is difficult to say what exactly was fabricated and installed, and in what manner it was designed to accommodate movement, if at all.
The seismic joints have also, apparently, been kept free of insulation, not stuffed with batt insulation as shown in the detail. I had emailed the College of Architecture, Art, and Planning's project liaison in November, 2009, remarking that "...the 5-inch space immediately below the curved expansion joint cover is filled with 'batt insulation,' but not otherwise protected against vapor intrusion from the interior space below... It may be that, even without humidifying the Milstein space, there would be high enough relative humidity (generated by building occupants) that such air, working its way up into the insulation, would reach the colder surface of the expansion joint cover and condense, wetting the insulation, and potentially causing other nasty problems during the winter months." The project manager replied in January 2010 that the Milstein Hall project manager "has looked at the issue that you raised and discussed it with team members. It is still a bit on the back burner since we have so many other pressing issues that need to be dealt with immediately. Be assured that we will close the loop with you on this issue." I never found out how or if this issue was resolved, but was told much later that the seismic joints were, in fact, uninsulated.
In the case of the vertical seismic "bellows" that separate the curtain wall of Milstein from the masonry walls of Sibley and Rand Halls, the seal between the two buildings is not particularly rigorous (Figure 7). All of this — the vertical bellows and the horizontal joints — leads to significant thermal bridging around that portion of the perimeter of Milstein Hall where it comes into contact with the adjacent buildings. This thermal weakness is not so much the result of conductive losses due to the low R-value of assemblies without insulative materials, but rather due to the high conductive values of metal joint material and discontinuities in the air barrier at these critical junctures.
The probability of water leaking through joints increases when "critical" seals — those designed to exclude water — are detailed and constructed using only sealants. The causes of sealant failure are too numerous to outline here.2 Numerous instances of leaks, some of which seem to be sealant joint failures, have already arisen in Milstein Hall, through both roofs and curtain walls.
There have been several leaks in the basement of Milstein Hall. Before construction was completed, water appeared in the basement floor adjacent to the lower-level gallery (Figure 8); but even after Milstein Hall was occupied, leaks continued at below-grade connections to Sibley Hall (similar to leaks shown in Figures 9 - 10).
Water is also leaking through the concrete topping slab above Milstein Hall's basement-level gallery. The gallery's entire "storefront" glazing system, facing the lower-level garden, was removed in June 2012, after the building was occupied. A new saw-cut "kerf" was made in the concrete ceiling just inside the upper horizontal mullion, and a new stainless steel flashing section was inserted in order to block this water and direct it back outside. Such a "solution" doesn't really address the main problem, for two reasons: water is still penetrating through the concrete topping slab, and the new stainless steel section creates more thermal bridges (Figure 11).
[Updated Feb. 10, 2016] It turns out that the problems encountered at the gallery window wall were even worse than I had thought. While I have not been able to verify what follows with Cornell officials, it appears that three serious problems were involved: first, the plaza above the lower-level gallery space was designed to be perfectly flat, so that rain water (or melting snow) was not directed to a drain; second, no drains were provided, so that rain water could build up on the plaza level and find other unintended means of entry into the concrete slab, as described above; and third, the two-way slab constituting the plaza itself, and the ceiling of the gallery below, experienced large and unintended deflections, so that rain water would pool towards the center of the slab, taking days to evaporate, long after the rain had stopped.
In other words, the plaza deck design failed because it followed almost none of the rules for unvented roof assembles, as discussed by Joseph Lstiburek in 2011. The following guidelines were not implemented, except as noted: (1) provide drainage below the traffic surface, including a drainage gap above the water control layer (roof membrane); (2) slope the roof membrane to a drain; (3) provide a double-drain, i.e., a second drain from the traffic surface; (4) provide insulation above the roof membrane with a drainage mat above and below — this is the only feature that was actually implemented; (5) slope away from the edge of the deck, towards an interior drain; and (6) connect the water control layer of the deck (i.e., the roof membrane) to the water control layer in the walls of the adjacent existing building (Sibley Hall) — of course, the load-bearing brick wall of Sibley Hall has no water control membrane, so it is important (and difficult) to properly flash the wall-deck intersection. Of these six "rules," only one (providing drainage mats) was properly implemented.
Remarkably, however, it turned out that the unintended deflection of the concrete deck provided a low-point in the slab into which a new drain could be inserted, after the fact. As described in Figure 11a, this drain was cut into the topping slab and connected to a new pipe that runs down through the gallery (actually right through the non-rotating freestanding gallery wall), and from there goes below the gallery floor slab and into a storm sewer pipe. At the same time, the concrete fascia at the edge of the deck was demolished and new flashing and drip edges were installed all along the edge of the deck, above the gallery windows. Although the deck is still deficient, and water still sits on the top of the deck instead of properly draining into the new drain, the new connection of the existing drainage mat below the topping slab to the linear drain inserted into the topping slab seems to have solved the major problem: water that seeps through the topping slab now can drain (through the existing drainage mat under the topping slab and into the new drain pipe) rather than building up and finding its way to the concrete fascia where it used to spill out, over and onto the gallery windows.
Leaks have also occurred on the upper level of Milstein Hall, through the roof (Figure 12). As is common with roof leaks, these appear to be related to defects in the flashing strategy and/or execution where roof membranes meet skylights or masonry walls of existing buildings (Rand Hall, in this case).
In addition to leaking of rain water through the building enclosure, water leaks can manifest themselves in other ways, especially when water is conducted in or through certain masonry materials or mortar. This seems to be the case at the lower level of Milstein Hall, where white powdery material has appeared in the ceiling itself, and especially on the aluminum "storefront" mullions at the west end of Milstein Hall (Figure 13). This may be an example of secondary efflorescence3 — soluble salt deposits left behind as water evaporates after leaking through the concrete slab. Figure 13a and Figure 13b show a line of what may be efflorescence on the concrete ceiling of Milstein Hall. What is most revealing is the water stain that appears immediately to the outside of that line (Figure 13b).
Even though the Milstein Hall dome is covered by the upper-level studio floor, it remains an outdoor space and therefore needs to have a functioning water-control layer. However, waterproofing for this complex doubly-curved slab was not provided in a systematic and thorough way (Figure 14).
The sliding entry door to Milstein Hall is automated by a motion sensor. This is entirely appropriate for entrances that are approached on axis — that is, perpendicular to the door itself. However, the door in Milstein Hall is immediately adjacent to a parallel circulation path that is used by many people who have no intention of entering Milstein Hall. The automated motion sensor triggers the door anyway (Figure 15), leading to heat loss or heat gain, depending on the season, not to mention wear and tear on the motorized mechanism itself.
Skylight curbs were cast in reinforced concrete and interrupt the three layers of rigid insulation on the roof deck under the "green roof" plantings. Before rigid insulation was adhered to these concrete curbs, circles of melted snow could be seen on the roof around the skylights (Figure 16); while the insulation improved the thermal performance, one can see that discontinuities in the thermal control layer (Figure 17) still melt the snow immediately adjacent to the skylight curbs (Figure 18).
Inside Milstein Hall, one can see that snow has melted from much of the skylights, as would be expected, due to the increased heat loss through the glass compared with the rigid insulation under the green roof (Figure 19). From the standpoint of energy consumption, there is a potential trade-off, since daylight within the space is improved, as described on Cornell's innovative design webpage: "Three sizes of skylights are arranged in a radial pattern on the roof with the larger ones at the center and smaller ones toward the perimeter of the building. This creates consistent natural light levels across the entire second floor studio space." An evaluation of the energy-saving benefit of daylighting compared with the energy-losing heat loss through the glass was presumably never done and, if it was done, certainly was never made public. But whatever the cost-benefit outcome of such a calculation (and the potential for energy savings is certainly real), it is rendered moot, since electric lights are almost always on, whether or not adequate daylighting is available (again, see Figure 19).
1 The construction of this thermal bridging at the bollards of Milstein Hall is illustrated in The Construction of Milstein Hall — Part 9 Stone & Soffit , by J. Ochshorn, starting at about 2:55 minutes.
2 See, for example, K. Warseck, "Why Sealant Joints Fail," online, accessed Aug. 15, 2013.
3 An explanation of "secondary efflorescence" by Achim Hering can be found here, accessed March 21, 2015, and illustrated by the image shown below. In his explanation of secondary efflorescence, Hering suggests that "if unchecked, this process will continue to weaken the concrete. In primary efflorescence, one only loses matter that does not get bound in the cement stone. In secondary efflorescence, one loses matter that originally made the concrete strong."
First posted 27 August 2013. Last updated: 10 February 2016