Seismic Considerations for Guastavino Ceiling, Vault, And

Home , Ceiling, Guastavino tile, Vault (architecture)

Seismic Considerations for Guastavino Ceiling, Vault, and Dome Construction Author(s): Doug Robertson Source: APT Bulletin: The Journal of Preservation Technology, Vol. 30, No. 4, Preserving Historic Guastavino Tile Ceilings, Domes, and Vaults (1999), pp. 51-58 Published by: Association for Preservation Technology International (APT) Stable URL: https://www.jstor.org/stable/1504710 Accessed: 29-08-2018 15:42 UTC

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at https://about.jstor.org/terms

Association for Preservation Technology International (APT) is collaborating with JSTOR to digitize, preserve and extend access to APT Bulletin: The Journal of Preservation Technology

This content downloaded from 24.4.7.238 on Wed, 29 Aug 2018 15:42:09 UTC All use subject to https://about.jstor.org/terms Seismic Considerations for Guastavino Ceiling, Vault, and Dome Construction

DOUG ROBERTSON

The seismic and program improve- Guastavino construction is found pre- in past earthquakes has been seldom and ments to the Hearst Memorial dominantly in turn-of-the-century archi- poorly documented. tecture across the eastern United States, The findings presented in this article Mining Building on the University with fewer examples as you travel fur- are based on investigations and evalua- of California, Berkeley campus, ther from New York City, the home of tions related to the seismic behavior of will be the first project in which the Guastavino Company. There are various examples of masonry vaulted Guastavino construction is deliber- very few examples of Guastavino con- and domed construction and, in particu- struction in California, where frequent lar, field investigations and testing of the ately strengthened for improved earthquakes pose a risk to this brittle vaulted ceiling in the Hearst Memorial seismic performance. architectural system. Within the San Mining Building (Fig. 1).1 Francisco Bay region there are three known examples of the Guastavinos' Historical Perspective work: the San Francisco Stock Exchange, Grace Cathedral in San Francisco, and To place the seismic vulnerability of the Hearst Memorial Mining Building Guastavino construction in perspective, on the University of California, Berkeley it is useful to first consider the seismic campus. performance of other forms of unrein- Similar forms of unreinforced ma- forced masonry construction in past sonry vaulted and domed construction earthquakes. Following is an overview are prevalent in older architecture in the of the seismic performance of masonry United States and abroad. Investigation buildings, the behavior of conventional of some of these buildings and literature brick vaulted and domed masonry searches reveal relatively little about the construction, and knowledge of the seismic behavior of this type of con- seismic behavior of Guastavino. struction. Little analysis and testing of vaulted and domed masonry construc- Unreinforced masonry. The majority of tion has been done, and its performance earthquake engineers and researchers consider unreinforced masonry buildings to be the most hazardous form of building construction. Over many years, earthquake after earthquake has reaf- firmed this view. The poor seismic performance of this type of building is due to many factors, including the brittle nature of the materials, the non-homo- geneous manner in which the materials jii are used, deficient workmanship, and .... in design and detailing that inadequately consider the effects of earthquakes on I I I . "!I " building construction. The "gluing" together of masonry pieces with mortar and the strong yet ...... A:, i?i?...... brittle nature of masonry materials make this type of construction subject to potentially sudden and catastrophic Fig. 1. Hearst Memorial Mining Building. Courtesy of Bancroft Library, University of California, failures under dynamic earthquake Berkeley. forces. For this reason, "unreinforced"

This content downloaded from 24.4.7.238 on Wed, 29 Aug 2018 15:42:09 UTC All use subject to https://about.jstor.org/terms 52 APT BULLETIN

60 regularly spaced steel wide-flange beams Ductile Concrete Frame, Low Rise (Fig. 3). There are three primary reasons that 50 Wood Frame, Low-Rise this jack vault system is vulnerable to earthquake damage. The most common = URM Bearing Wall, deficiency is that the walls are not posi- tively connected to these relatively rigid vaulted floor, ceiling, and roof systems. In an earthquake, the walls pull away from E a Non-ductile o Low Rise Concrete, Frame, , the vaults, leading to collapse of either 20 Low Rise the walls, the vaults, or both. The walls parallel with the steel I-beams are partic- URM Infill" Frame, . -o.,"- 0000"is 11,00 ularly susceptible to this type of damage 10 --LowRise . -0'0 since, unlike the walls that support the o ...... jZ s . - Steel . Frame,- Low Rise steel beams, there are no beams to provide a nominal tie to the walls. These 6 7 8 9 10 vaults also often have a shallow radius, Modified Mercalli Intensity (MMI) which make them prone to earthquake damage when the supporting, unbraced, Fig. 2. Damage projections for various building types. Illustration by author. steel I-beams spread laterally. Lastly, these vaults often support loose, heavy fill materials, such as dirt, ash, or rubble, masonry construction is no longer vaulted which in an earthquake add to theiror domed roof or ceiling con- permitted struction by modernthen becomes U.S. building difficult to iden- inertial forces and their vulnerability. codes tify, in zones In his speechand at the 1893 World's Fair theof moderateinterest in to its high behavior is of seismic risk. little on "cohesive construction,"concern Rafael compared to the general Fig. 2,structural taken Guastavino referred to the Persians as from collapse, ATC-132, concern shows for rescuing the relative potential the "fathers of the cohesiveperformance mode of survivors, andof masonry the urgency to buildings rebuild. compared to other lateral construction."3 Ironically, it is in this building systems. Performance is mea- One regiontype of the world where several exam- of system employed by the sured in terms of damage projections Guastavinos ples of failure of this system have been that is prone to earthquake accessed damage as documented. In 1990,a widespread dam-percentagewas used extensively of building through- replacement out many age occurred in a Richter magnitude cost. 7.3 European Performance cities andwas elsewhere considered in the earthquake in atnorthwest world. Iran that killeddifferent This levelssystem, of ground-sometimes shaking referred intensity an estimated 35,000 to 50,000 people to as using "jack the vault," Modified used bricks Mercalli or clay Intensity and damaged about 100,000 buildings.tile to form(MMI) repetitive scale. vaults These spanning estimates Measured peak ground acceleration between were developed the bottom by flange of polling a substantial number of earthquake experts with a systematic method- ology. (1tASTxIvIN(O The RIB AND DOME SYSTEVMchart suggests that masonry performs poorly compared to all other , a.. m .-, v R o ~ TT ~_ House r51 btemz -.*&.- materials and systems. . . ;NGo L,,,,2N GO2D, ,ToI3L, t 5, 8T Masonry vaults and domes. Relatively few records exist of past performance of N1---,--, vaulted and domed masonry construction in earthquakes. There are a number of possible reasons for the limited historical record. This type of construction is often found in regions of the world with relatively low seismicity or in third- world countries c Lio CT AB. where earthquake damage is poorly documented. Probably the most important reason is that failure of SCALE 1ONE FOOT this type of system -i-Rt M@i C.. is thought, in many instances, to ,,, , ,. *lead ,,z,._ to more significant damage or collapse of the unreinforced Fig. 3. Jack vault. Courtesy of the Guastavino/Collins Archive, Drawings and Archives, Avery Archi- masonry tectural and Fine Arts Library, Columbia University. structure. Damage to the

This content downloaded from 24.4.7.238 on Wed, 29 Aug 2018 15:42:09 UTC All use subject to https://about.jstor.org/terms SEISMIC CONSIDERATIONS FOR GUASTAVINO CONSTRUCTION 53

varied from 0.02g to 0.65g in the areas The strength of Guastavino under surrounding the epicenter. Many build- gravity loading is generally undisputed. ings with this type of vaulted masonry/I- ENTRANT CAORNER George Collins pointed out one impor- beam system collapsed. These collapses NAVE tant difference between the timbrel vault were reportedly precipitated in most and the more conventional stone-ma- instances by the walls pulling away from sonry vault. He noted that the vault "is the vaults.4 very thin, consisting of little more than a In March 1997 another earthquake surface, and derives its rigidity not from in northwest Iran, with a magnitude of massiveness or thickness but rather from 5.5, caused widespread damage, killing its particular geometric form." 10 This 965 people and injuring more than statement explains precisely why the Fig. 4. Plan view of the Basilica of St. Francis 2,600. None of the recently engineered of Assisi. (Gray areas indicate collapsed Guastavinos' many works have per- steel or concrete buildings experienced zones.) Illustration by author. formed so well over the past century. any noticeable damage. However, many However, it must be emphasized that of the buildings with this repetitive brick Guastavino construction derives its vaulted roof system collapsed.5 substantial strength not just from its are also often constructed in countries The most widely recognized and vaulted form but also from its geometric where the quality of construction may notable example of earthquake damage orientation relative to the direction of not be well controlled. Though these of a vaulted or domed structure was the loading. While Guastavino construction failed vaults may not have been consis- damage to the ceiling in the Basilica of has performed admirably over the past tent with the quality and construction St. Francis of Assisi in September 1997. century under the forces of gravity, its methods used in Guastavino construc- Accelerometers indicated ground accel- performance when subjected to the tion, their failures are still useful in erations of 0.2g or greater. Subsequent added lateral or vertical loads from an beginning to understand the potential engineering analysis indicated that the earthquake has not yet been adequately seismic risks and vulnerabilities of capacity of these vaults would be reached tested. vaulted and domed construction. at a lateral acceleration of about 0.2g.6 Damage was certainly partly Guastavino construction. There is little The Guastavino System attributable to the increased mass of information available on the seismic loose fill above the vaults and also the Although masonry construction has performance of Guastavino construc- building's geometry. The ceiling vault proven very vulnerable to earthquake tion, yet there are a few indicators that collapse was in an area referred to by damage, the Guastavino system gener- poor performance could occur. At Grace earthquake engineers as a "reentrant ally offers a number of advances over Cathedral in San Francisco, the Guasta- corner," where a significant change in more conventional stone and brick vino ceilings experienced some damage the plan of the building occurs at an vaulted and domed construction. Rafael in the 1989 Loma Prieta earthquake.7 inward corner, in this instance, at the Guastavino was a master of design as The cathedral is founded on a rock site transition between the basilica's nave well as construction, often implementing and is about 55 miles from the earth- and transept (Fig. 4). This condition is details of construction that added quake epicenter. The level of shaking recognized by engineers to be particu- strength and redundancy helpful in was therefore relatively low, with peak larly vulnerable to earthquake damage. reducing the effects of earthquake ground acceleration of about 0.15g. The failure of these particular penden- forces. Photos and drawings indicate Yet, even at this low acceleration, some tives may have been avoided or their that he generally detailed and provided cracking was reported along the ridge of performance significantly improved by well-anchored and rigid boundary con- four pendentives. This cracking, visible adding cross building ties at this loca- ditions, which in an earthquake help from the top, caused fragments of mor- tion. This type of building, with its tall prevent the vaults from spreading later- tar to fall to the floor below. Strengthen- and massive exterior walls, is particu- ally and losing support. Although these ing using Gunite placed from the back- larly vulnerable to earthquakes. The details were likely implemented to im- side of the ceiling had been proposed in lateral forces produced by an earth- prove the gravity-loading behavior of the 1950s or early 1960s, but this work quake cause the walls to pull away from the vaults and domes, they may also was never carried out.8'9 the vaulted/domed ceilings, leaving the improve their seismic performance. Another example of Guastavino roof and ceiling without support. The Guastavinos also used multiple construction is found in the Hearst Available documentation seems to courses of tile with overlapping mortar Memorial Mining Building. The build- indicate that all of these examples of joints. Though this adds more mass to ing site is located roughly 60 miles from vault collapse involved vaults construc- the system, the overlapping joints can the Loma Prieta earthquake epicenter ted of brick and not clay tile. Brick provide improved strength. In order to and experienced an estimated accelera- vaults have been observed to normally achieve Guastavino's "cohesive con- tion of about 0.1g. The building was include a single course of brick com- struction," the bond between courses of completely undamaged except for a pared to the multi-course technique used tile and mortar is essential, particularly couple of fallen Guastavino tiles. in Guastavino construction. These vaults under earthquake loading. Guastavino

This content downloaded from 24.4.7.238 on Wed, 29 Aug 2018 15:42:09 UTC All use subject to https://about.jstor.org/terms 54 APT BULLETIN

tant to its seismic performance, not only c- 0 .2 2.00 20 the boundary connection of the Guas- S- San Francisco San Frandsoo tavino itself but also the strength and = 1.5 rigidity of the surrounding support S1.50 - New York City structure. S1.00 S1.0 C The one characteristic relevant to all 0o.s w 0.5 Guastavino work is the vault span-to- 8. 0.50 S0.00 S0.0 I I I thickness ratio. The thickness, which 0.0 1.0 2.0 3.0 0.0 1.0 20 3.0 depends on the number of tile and mor- uC Period (Seconds) Cu Period (Seconds) tar courses, will influence the rigidity and strength of the system. In order to Fig. 5. Earthquake response spectra Fig. 6.for Earthquake New response spectra for Santake advantage of this multi-layer shell, York and San Francisco. Illustration Francisco by author. and Hearst. Illustration by author. it is important that the courses of tile and mortar have sufficient strength and bond to transfer stresses. Failure to recognized the importance ofthe thislikely bond,fault rupture mechanism maintainand this monolithic behavior can depth of rupture, and the type and depth noting the "danger of sliding in the lead to significantly greater seismic of soil all play a part in determining the horizontal joints, only in case strong stresses and deformation, making the probable earthquake response. cements are not used." 11 system much more susceptible to failure. Some of the buildings that contain The Guastavinos also frequently The specific, and often unique, details of Guastavino construction are themselves added extra tiles to form stiffening ribs each Guastavino application will lead to susceptible to earthquake damage if left at a repetitive spacing or at changes in variable seismic behavior. Therefore, the unstrengthened. It may make little sense geometry. Above the vaults they added Guastavino system and details in each to strengthen the Guastavino construc- supplemental steel members to brace building must be evaluated individually. tion, only to have the supporting struc- vault ribs and boundaries and to provide Two examples of earthquake re- ture severely damaged in an earthquake. cross ties to prevent the supports from sponse spectra, developed using the However, Guastavino is often found in spreading. They also used cement mor- NEHRP Guidelines for the Seismic high-occupancy public spaces, where the tar except at the joints of the face tile, Rehabilitation of Buildings,12 are shown life and safety of building occupants may where plaster-of-paris mortar was gener- in Fig. 5. One spectrum is applicable to be threatened more by falling tiles then ally used for quick set time. New York City and the other to North- The implementation of these by damage details to the primary structure. ern California for the maximum consid- The magnitude of earthquake forces should generally help to improve the ered earthquake (MCE) on a Class B seismic behavior of Guastavino imposed con- on any building system is firstrock soil site. Since no reduction factor dependent on the rigidity of the global struction. However, Guastavino's system "R" is permitted by NHERP for unrein- remains very brittle and potentially building structure. More rigid structures hazardous when subjected to will dynamic experience higher accelerations and earthquake forces. more flexible structures lower accelerations, due to their period of response.

Guastavino Seismic Evaluation Secondly, the forces imposed on a Guas- tavino system are dependent on its proximity and lateral rigidity relative to other To determine whether seismic strength- ;rs ening should be implemented as a part horizontal roof and floor diaphragm MM!: of the preservation strategy on a specific construction. For instance, a rigid ...... project, a number of risk factors should concrete roof located directly above the be considered. These include the level Guastavinoof system may attract much of seismic hazard associated with the build- the lateral force, reducing the loads on ing site, the capability of the primary the Guastavino elements, whereas a building structure to resist seismic flexible wood roof diaphragm in the ...... forces, the level of force that can be same proximity will take relatively little transferred through the building struc- load compared to the more rigid Guas- ture to the Guastavino system, the tavino. In cases where the Guastavino boundary conditions of the Guastavino system serves as the sole roof or floor system, and other specific details of thediaphragm system, it may provide the particular installation. only lateral load path for resisting The seismic hazard at any given seismic forces. building site depends on the region's As in the jack vault system, the Fig. 7. Hearst Memorial Mining Building geotectonics and the site soil characteris- boundary conditions of Guastavino Guastavino ceiling. Courtesy of Bancroft construction are generally very impor- tics. The sites proximity to active faults, Library, University of California, Berkeley.

This content downloaded from 24.4.7.238 on Wed, 29 Aug 2018 15:42:09 UTC All use subject to https://about.jstor.org/terms SEISMIC CONSIDERATIONS FOR GUASTAVINO CONSTRUCTION 55

forced masonry construction such 300- as Seismic design criteria. In Fig. 6, the Guastavino, design forces can be deter- site-specific spectrum for the Hearst site mined directly from these spectra based is compared with the spectrum for a m200, 2 on building period. Most of the build- more typical northern California rock ings that incorporate the Guastavino 10Do -I E~~g Cbg9r~ site (NEHRP soil Class B). The Hearst 2O -p system rely on a masonry or concrete -8 hd , o - -lr 2 W 8 spectrum has higher accelerations than shear-wall lateral system, which leads to the typical spectrum because Hearst is -- S rur'thawr CoIh PN (M iC"lb btw R bs) a relatively short building period. There- Sl rolgthonha ColhrN At bs) considered a "near fault" site. More V t00 fore, in most instances the seismic re- Vault Leigth (inhrhs) specifically, the is building within 800 sponse of Guastavino falls at the plateau feet of the active Hayward fault. of the spectrum where the acceleration Fig. 8. South is side vault: hoop stress at back ofWithout base isolation, the high the greatest. face tiles. Courtesy of William Kreysler and spectral acceleration represented by the The plateau of the New York spec- Associates. Hearst spectrum plateau would be used trum reaches a peak horizontal ground for seismic design. Base isolation length- acceleration of 0.4g or 40% of the force ens the building period, represented by of gravity. Though this does not seem of Material Science and Mineral Engi- the declining portion of the spectrum, particularly high compared to the Cali- thus reducing the building base shear neering. fornia spectrum, which reaches 1.5g, In 1978 the first seismic evaluation of from about 1.8g at the spectrum plateau to 0.25g at a design period of three sec- these smaller earthquake forces should campus buildings classified Hearst's not be dismissed too hastily. The Guas- onds. Although isolation reduces the seismic safety as "very poor." In recent tavino ceiling in the Grace Cathedral horizontal acceleration, it does not years, its outdated facilities have also experienced some cracking at accelera- reduce the vertical acceleration associ- fallen behind its evolving high-tech tions of only about 0.15g in the Loma ated with proximity to a fault. teaching and research mission. In 1993 Prieta earthquake, and partial collapse The seismic design criteria for the preliminary studies were undertaken to to the Basilica of St. Francis of Assisi Hearst Guastavino ceiling was thus consider seismic and program improve- occurred at an estimated acceleration of driven by two primary factors: first, base ments for the building, design was offi- 0.2g. isolation significantly reduces the lateral cially begun in 1996, and construction is acceleration, and second, isolation does now underway. The building is to be Preservation and Seismic Protection placed on a system of base isolation nothing to reduce the near-fault vertical acceleration. The design vertical acceler- bearings, which includes 134 high Do preservation and seismic-risk-reduc- ation considered in analyses was 2.5g. damping rubber bearings and 24 fluid tion goals conflict? Many preservation These forces were derived from site- viscous dampers (12 in each direction), purists, even in California, believe that specific time history analyses for the all placed below the existing first-floor the often-invasive nature of seismic building that were used for design of the level. The base isolation system will strengthening is entirely inconsistent isolation system. greatly reduce the lateral acceleration of with the goals of historic preservation. the building in an earthquake. However, this perspective may be slowly changing. There are no assurances that monumental unreinforced masonry " , " . . .. buildings, left unstrengthened, will be safe from damage or destruction when the next major earthquake occurs. One -~:~-- .1. .. 2;..? ::r ?. - , - Z~ need only look at the 1994 Northridge or the 1995 Kobe earthquake devasta- tion to understand the potential for ]FRP ? RI B5S ,FFLNN4 F 1. 0.-,, economic and human loss when such an L (MAX509VEV TO FRP 5TIFFEN.N.- .. .-.. -.I I RIB 40 4 OPT. O.1. . I eI~tN& "BAK OF rTI', event occurs. M4AX) BONPED TO BACK C TILE '--RP AN-E BRACE CEILIFY;

"ZP AW YLRP AOLE The Hearst Memorial Mining Building -0 .....VC TO A. ; " ' . ",: . - - '.' 'FRP 5TIr NN.N6 R1I Background. The Hearst Memorial ?? ?id7,? FRP VIAPIRA&-fl- Mining Building, located on the Univer- ARD1~ll FEA sity of California, Berkeley campus, was designed in the Beaux-Arts style by r. campus architect John Galen Howard .. :"" E-.{ '. AE and was dedicated in 1907. Constructed as a mining and mineral engineering building, it now houses the Department Fig. 9. Section of vaulted ceiling strengthening. Courtesy of William Kreysler and Associates.

This content downloaded from 24.4.7.238 on Wed, 29 Aug 2018 15:42:09 UTC All use subject to https://about.jstor.org/terms 56 APT BULLETIN

sion tests suggested that the bond between layers was generally negligible. Of the thirty tests, twenty showed negligible EPOXY MORTAR ADHESIVE TOEBACK) (2 C.. . . .EL S, BACK bond, while the remaining ten tests had widely varying tension capacities rang- ing from 13 to 99 psi. The failure plane AND HIH L 5 IS AG1 PTA OVER BACK OF CAILIN) between tile and mortar appeared ran- dom, with about half of the cores failing at the face tile to mortar plane and the . li ...... other half occurring at the mortar to backing tile interface. An initial concern about this testing method was that the coring might cause FACE1/4' TILEDIAM. CONNECTION:,M HOLE THROU6H CENTER OF SM H MORTAR vibration or stresses that might initiate ADHESIVE.TiLE INJECTED SPACING TO BENTH DETERMINED POLYURETHANE SEALANT/ FACE TE MORTAR SIMILAR AT PENDENTVES. 6" x 12" FACE TILES failure, leading to unreliable test results.

6" x 12" BACKING TILES A similar core sampling approach for (JUNINe sOND) testing shear capacity in brick walls was abandoned more than ten years ago for Fig. 10. Detail of cross section at vault stiffening rib. Courtesy of William thisKreysler reason. andHowever, Associates. the coring reportedly was very smooth, without vibration, and it is believed that the manner and load at which the cores failed were Seismic evaluation. In 1993 the first layer sandwiched between two layers of largely due to the weak bond between investigation of the Guastavino ceiling clay tile, a rough corrugated face tile layers of tile and mortar. Other evidence included a sounding study of the area laid in a herringbone lay-up, and a appears to support this conclusion. above the galleries (Fig. 7). This tech- smooth tile backing layer placed in a Inspection of the cores following testing nique has become the commonly ac- running-bond lay-up. As discussed generally revealed a clean break at the cepted method for evaluating the adhe- previously, this bond between alternat- tile to mortar interface, with no tile or sion between the face tiles and the ing layers of tile and mortar is essential mortar remnants adhered to the oppos- backing mortar/tile substrate. A wooden to transfer lateral stresses, which under ing half of the core sample. Material mallet was used to detect a hollow ring normal gravity load are lower and less testing of the tile and mortar revealed sound, which was assumed to indicate important to the ceiling stability. To reasonably high strengths, confirming marginal or compromised bond between assess the bond strength, direct tension that test failures were not attributable to the face tile and the mortar substrate. tests were conducted. The testing was material failure. Mortar and tile com- Several qualifiers regarding the accuracy accomplished from the backside of the pression strengths ranged from 4,396 to of the sounding were provided in the vaults and pendentives after locating 5,871 psi and 3,017 to 6,552 psi, re- testing report.13 The report indicates face tiles with expected good adhesion spectively. At one location the mortar that the "Interpretation of the sounds is by sounding from the underside. Testing was carefully removed around one tile somewhat subjective " and also points was performed by drilling 3-inch-diame- that had been first sounded to confirm out that "Tiles were generally sounded ter cores from the backside, penetrating reasonable bond. However, the tile fell in only one area (usually near the center only to the back face of the face tile. A from the ceiling under its own weight of the tile)" and that indications some- steel plate connected to a threaded rod once the surrounding mortar was re- times vary between different parts of the was then bonded to the back of the core. moved. Although these results do not tile. This study indicated only about 5% Next, a tension load was applied to the provide conclusive evidence, they do of the tiles have questionable bond to rod to determine the tension capacity of cast some doubt on the reliability of the substrate. Therefore, it was assumed the core and to locate the failure plane. sounding for determining the conditions in early design studies that the bond This was considered the least invasive of face tile and backing mortar bond. between layers of tile and mortar was testing method that would provide a One can conceive how a tile with a well- generally sufficient to maintain the measure of actual bond strength without filled collar joint and tight edge joints system's monolithic behavior under impacting large areas of tile or the ap- could produce a solid sound, as if well seismic forces. pearance of the ceiling. bonded, while actually having no bond Last year, a second investigation was The results of this testing were unex- whatsoever. This technique also provides carried out to identify boundary condi- pected and conflicted with the assumed no assurances of the condition of bond tions, determine material properties, and conditions based on the earlier sounding between subsequent layers of tile and confirm by more explicit testing tech- study and descriptions of the "tena- mortar. niques the bond strength between the cious" nature of the mortar bond be- The weak bond between tile and three-layers of Guastavino construction. tween layers of tile as described by mortar courses could be attributable to These layers include a single mortar George Collins.10 The majority of ten- a number of factors. During construc-

This content downloaded from 24.4.7.238 on Wed, 29 Aug 2018 15:42:09 UTC All use subject to https://about.jstor.org/terms SEISMIC CONSIDERATIONS FOR GUASTAVINO CONSTRUCTION 57

tion, the masons may have failed to wet EXISTING STEEL I-BEAMS the in-place tile or mortar prior to plac- ing the subsequent course. As the next course was placed, the surface moisture EXISTING STEEL ROD BOLTED TO I-BEAM ABOVE AND EMBEDDED may have been absorbed by the new dry IN TILE TALL BELOH materials, drawings the moisture away from the bond interface and leading to a weakened joint. This is a common failing in masonry construction even today. Another explanation may be that the masons were unfamiliar with the materials and dry climate of California, since most of their prior work had been done in eastern states.

Finite element analysis. Based on the TISA CELINE 2ITHSQU material properties derived from testing EPOXY ADHCSIVE, TYPICAL and other information gathered from BAND ( PANEL E TO-.. field investigations, finite element analy- OUASTAVINO TILE CEILINCS7 ses were carried out first to consider the potential seismic vulnerabilities of the FRP DIAPHRAW6 M PANELS existing unstrengthened system and then to evaluate various strengthening alternatives. Two basic models were developed for the longest span vault condition. The first included mortar joints 5# URETHANE FOAM (CLASS I FIRE RETARDANT) replicating the herringbone face tile OVER STEEL MESH pattern. With this model, the expected benefits of this pattern were considered and the mortar joint stresses evaluated. This model was first used to evaluate the stresses in the existing system. A second PEN VE HITH model was then created to evaluate FOAM various strengthening alternatives. To simplify the analyses and perform these evaluations more efficiently, this model excluded the mortar joints. Finally, once the strengthening approach was selected, 0UASTAVINO TILE the strengthening elements were added HALL (BACKING TILES TILES ARE 12" SCUARE to the original, more detailed model, HOLLOhl C;ORE- TILES) and the new system was analyzed to verify its effectiveness. Fig. 11. Section at pedentive. Courtesy of William Kreysler Associates. Load testing by Rafael Guastavino himself showed the tremendous load- carrying capability of his system under vertical loading conditions. However, vaulted his ceiling behaved in a flexural ineffective. Once the face-tile mortar testing generally included a vault with mode, with compression in the face tiles joints were removed from the model, the boundary conditions at a single coinci- at the upper end and tension at the bond stresses between face tile and dent elevation, whereas many of his lower end of the vault consistent with backing mortar greatly exceeded its projects incorporated much more com- the hoop stresses depicted in Fig. 8. The minimal capacity. For subsequent analy- plex geometry. The original assumption, critical load case considered in analyses ses, it was assumed that the face tile that the vaulted ceiling in Hearst would included a downward acceleration of would not contribute to the system behave under vertical loads as one ex- 2.5g (seismic plus gravity) plus a lateral strength but would be supported by the pects of a properly designed arch, pro-load of 0.3g. remaining strengthened system. ducing uniform compression and out- Under these loading conditions, The existing ceiling analyses revealed ward thrust at the base and no tensile analyses indicated that the weak plaster- high hoop, through-thickness, and shear stresses, was incorrect. Analysis revealed of-paris face tile mortar joints would be stresses in various areas throughout the that under downward vertical loads the significantly overstressed, rendering it system. The level of stress would be

This content downloaded from 24.4.7.238 on Wed, 29 Aug 2018 15:42:09 UTC All use subject to https://about.jstor.org/terms 58 APT BULLETIN

excessive in most masonry construction composite diaphragms, rather than DOUG ribs, ROBERTSON is a structural engi- neer and an associate with the San Francisco but particularly in the Hearst ceiling, will be placed horizontally in each pen- engineering firm of Rutherford and Chekene. because of the very weak bond between dentive (Fig. 11) to provide the neces-For more than 15 years he has participated in tile and mortar courses. sary stiffness and strength. Alternative the investigation, assessment, and seismic A three-pronged strengthening methods ap- for anchoring the face tiles strengthening design of historic buildings. proach was developed jointly by wereRuther- studied, but the high stresses that ford and Chekene, the project structural would be developed in using these Notes rigid engineering firm, and William Kreysler connection methods were found to be and Associates, a firm specializing untenable. in the 1. Preserving Historic Guastavino Tile Ceilings, design and fabrication/construction A ofcomparison between the maximum Vaults, and Domes," Conference at Columbia University, sponsored by the New York Land- composite materials. The first step hoop is to stresses in the existing Guastavino marks Conservancy, February 6, 1999. This reduce the ceiling deformation and system thus and the strengthened Guastavino article expands on the information presented by the global stresses to acceptable levels;system is shown in Fig. 8. For the the verti- author at the conference. second, provide a back-up system cal to axis, negative values represent 2. ATC-13com- (Applied Technology Council), provide support should significant pression dete- and positive value tension Earthquake Damage Evaluation Data for rioration occur during long duration stresses. The horizontal axis follows California the (Redwood City, Calif.: 1985). earthquake shaking; and third, posi- length of the south vault as depicted 3. Rafael in Guastavino, "Cohesive Construction, tively anchor the face tiles to prevent Fig. 9 on the right. Its Past, Its Present, Its Future," The American Architect and Building News 41 (922: 26, individual tiles from falling and to retain August 1893): 125-129 the vault curvature (Fig. 9). Conclusion To reduce overall deformations and 4. M. Mehrain, "Reconnaissance Report on the Northern Iran Earthquake of June 21, 1990," Guastavino vaulted and domed con- stresses in the vaults, epoxy-fiberglass National Center for Earthquake Engineering double-channel ribs will be bonded struction with may eventually prove to pro- Research, Report NCEER-90-0017, October 4, 1990. epoxy to the back of the vaults at vide a 4-ft.superior earthquake performance spacing. Second, a backing system, compared with more conventional 5. EEIR (Earthquake Engineering Research consisting of rigid steel wire mesh unreinforced em- brick and stone masonry Institute) Newsletter 31 (7: July 1997). bedded in sprayed-on urethane foamvaults and(3 domes. However, considering 6. Croci, Giorgio,"The Basilica of St. Francis of lbs/ft3), will be placed over the top the rigidof and brittle nature of the materi- Assisi After the September 1997 Earthquake," Journal of the International Association of vaults. This lightweight and rigid als foam and the prolific use of Guastavino Bridge and Structural Engineering (IABSE), SEI and steel mesh system spans between construction in vastly differing applica- 8 (1: February 1998). composite ribs and supports the tions Guas- and conditions, seismic deficiencies 7. Doug Robertson's phone conversation with tavino system by adhering firmly must to theexist in some buildings. Grace Cathedral archivist, January 1999. back of the vaults. Lastly, the face Althoughtiles the seismic hazards in the 8. Avery library drawings, 1999. will be connected with regularly eastern spaced U.S. are considerably less than in 9. Michael Lampen. Phone conversation 1/4-inch-diameter urethane elastomer California, a degree of seismic risk between Doug Robertson and Grace Cathedral pins, injected through a hole drilled remains in for this unique architectural archivist, January 1999. the center of tiles. Fig. 10 shows motif.a de- Future renovation projects of 10. George R. Collins, "The Transfer of Thin tailed section of the strengthening buildings sys- that include this important Masonry Vaulting from Spain to America," tem. The spacing of pins remains historic to be resource should not discount the Journal of the Society of Architectural Histori- determined, based on the magnitude potential of loss of the system in an earth- ans 37 (October, 1968): 176-201. stresses in each area. Rather than quake.resist- The seismic vulnerably of Guas- 11. Peter B. Wright, "The Works of Rafael ing the expected high hoop and tavino work should be given appropri- Guastavino, Part II, What is Cohesive Construc- tion?" Brickbuilder 10 (May 1901): 100-102. through-thickness stresses, these atepins and are due consideration together with designed to elongate slightly under other in- project program and renovation 12. FEMA 273, "NEHRP Handbook for the Seismic Rehabilitation of Buildings," Federal plane stresses, so that high stresses goals. do Even though planning for a major Emergency Management Agency, Washington not fail the connection. Connecting earthquake all may not be economically D.C., October 1997. tiles is considered unnecessary, since feasible the on many projects, architects and 13. Schwein/Christensen Laboratories, Inc. vault curvature is essentially maintained, engineers should seriously consider the "Masonry Testing Hearst Memorial Mining and the unconnected tiles remain impactcon- of smaller earthquakes that Building Seismic Renovation,"April 30, 1993. fined by connected tiles. Before a occur tile with greater frequency. This study could fall, a majority of the surrounding indicates that some Guastavino systems mortar would need to be lost, which will perform is poorly in a major earth- considered unlikely. The urethane quake. pins However, we do not know with are injected through the foam backing any certainty how these relatively deli- and over the steel mesh system to cate pro- thin shells will perform in smaller vide positive anchorage. The penden- events with accelerations of 0.1, 0.2, or tives will be strengthened in a similar 0.3g, especially as the form and details manner, except that three fiberglass are unique to each installation.

This content downloaded from 24.4.7.238 on Wed, 29 Aug 2018 15:42:09 UTC All use subject to https://about.jstor.org/terms