Context

The glass spiral staircase by architect Eva Jiřičná has long been admired as a masterclass in lightweight design, a quiet demonstration of how minimal a load-bearing element can become when geometry takes over from mass. While the architectural vision is uniquely hers, translating it into reality required exceptional engineering (a collaborative effort that involved my former colleague, Ing. Karel Košek).

The glass and steel spiral staircase standing in the Café B. Braun interior, with the dark tapering steel newel at its centre and a lattice of thin stainless rods around it
The staircase as built, in the Café B. Braun interior in Prague. The dark tapering blade at the centre is the newel; everything bright around it is the stainless cage. Photo: archiweb.cz (photographer uncredited)

The architectural idea in this article is not mine. What is mine is the engineering that follows from it: my 2016 diploma thesis for the Civil Engineering programme at the Faculty of Civil Engineering, CTU in Prague, which took a staircase of this type and worked it through from first principles — a finite-element model, a full set of code checks, a modal analysis, and a complete drawing set. The point was never to invent an alternative. It was to reverse-engineer the behaviour and understand exactly why the original solution is correct.

Close view of the laminated glass treads carried on a dense lattice of thin stainless steel rods and discs
Treads and cage at close range: laminated glass sitting on a lattice woven from stainless rod between 4 and 20 mm thick. Nothing here is thicker than it needs to be. Photo: archiweb.cz (photographer uncredited)

The Approach and the Model

The thesis asked a narrow but complex question: given the published geometry and extreme slenderness, how does the steel actually behave under real-world conditions?

To answer it I built a finite-element model in RFEM 5.06, mixing beam and shell elements — 932 one-dimensional and 8,001 two-dimensional elements, on a 50 mm mesh locally refined to 5 mm wherever stress concentrates around the tread supports. The same model carried both phases of the assessment: the static checks and the dynamic one.

The model through a full revolution. The three yellow bars at landing level are the built-up beams that anchor the newel into the slab; the green ring at the foot is its line hinge; the thin verticals ending in green pyramids are the hangers. Model: RFEM 5.06, Dlubal Software.

Describing what went into the model is the quickest way to explain the structure. At the centre stands the newel — not a column but a helix cut out of a 300 × 6 mm steel tube, five metres long, in ordinary S235JR, pin-supported at its foot and anchored into the floor slab at landing level through three built-up beams. Around it the balustrade is a stainless space truss: 12 mm balusters that double as spacers between the treads, a twisted 20 mm rail below and a paired 20 mm rail above that also serves as the handrail, a five-row lattice of paired 4 mm rods, and a double 20 mm helix hooping the whole cage, all in grade 1.4301. That cage does not stand on the ground. It hangs from three stainless tie rods anchored into the slab above.

Side elevation of the spiral staircase showing the helical steel newel, the trussed balustrade cage, the glass treads and the tie rods running to the slab above
Side elevation from the thesis drawing set. The tapering blade at the centre (12) is the newel; the vertical lines running off the top are the three tie rods that carry the balustrade cage.

The treads are the only glass in the structure, and they are not cantilevers. Each 30.2 mm laminated pane — a 5 mm sandblasted float wearing layer over two 12 mm heat-strengthened plies, bonded with PVB — is pin-supported at four points on a small underslung king-post truss, clamped to the balusters that pass through it, and tied back to the helix on an L-bracket. Load therefore leaves a tread in two directions at once: inward to the helix and down to the ground, and outward into the hanging cage and up to the slab. The erection sequence in the technical report makes the hierarchy plain — raise the newel, anchor the hangers, weld and hang the balustrade as a spatial truss girder, and only then lay the treads in.

Developed elevation of the staircase, unrolled flat, showing the eighteen treads, the trussed balustrade above them and the king-post trusses beneath each tread
The same structure unrolled flat. Read this way the staircase is a single inclined truss eighteen bays long, with a small king-post truss slung under every tread.

For the static analysis I set up 113 load cases, condensed into 27 combinations at ultimate and serviceability limit states. Alongside the code loads — 3.0 kN/m² distributed and a 2.0 kN concentrated load on the treads, 0.5 kN/m on the handrail — most of those cases exist to simulate the staircase actually being used: one person, two people, three people, at every plausible position on the flight, moving up one behind another. The goal was to map how the force paths shift as the load walks up the spiral, and to compare the result against the visible cross-sections of the realised structure. Four combinations ended up governing everything.

Six of the load cases in sequence: self-weight of the glass wearing layer, 3.0 kN/m² over every tread, 2.0 kN concentrated on every tread, 0.5 kN/m along the handrail, and finally two people low on the flight (treads 1 and 3) and then high on it (treads 15 and 17). On a structure this light, where the load sits matters more than how much of it there is.

What comes back is almost disappointingly small. Under the governing serviceability combination the newel moves 1.8 mm sideways and settles 0.6 mm; the worst tread drops 3.7 mm. The deformation animation below has to magnify those displacements roughly sixtyfold before there is anything for the eye to follow at all — which is, in its way, the most flattering thing that can be said about the design.

The deformed shape under the governing combination, looping. Displacements are magnified about sixty times; at true scale the movement would be invisible at this size. Analysis: RFEM 5.06, Dlubal Software.

The Dynamic Reality

A structure this slender cannot be judged by static forces alone. The second half of the assessment was a modal analysis: the first twenty mode shapes, solved by the Lanczos method, with the mass case set to the self-weight of the steel and the glass.

The first mode at 7.934 Hz: the whole staircase winding and unwinding about its vertical axis. Amplitude is normalised, not physical. Analysis: RF-DYNAM Pro, Dlubal Software.

That first mode is a rotation, and it is worth noticing that the structure's softest direction is the one it is shaped around. The second mode, at 9.837 Hz, is a lateral sway; the third, at 11.051 Hz, is the head of the newel moving on its own. For a staircase the number to beat is roughly 6 Hz, below which footfall starts to feel as though it is being answered by the structure. At 7.9 Hz the design clears it with margin.

Worth being straight about the outcome: on this staircase the dynamics did not turn out to be the binding constraint — the governing checks were all ultimate-limit-state combinations. But that is only knowable once the modal analysis has been run. In glass and steel structures of this weight, dynamic comfort is the check that most often ends up governing, and it is never the one to leave until last.

The Details

Beyond the calculations, the thesis produced a full drawing set: general arrangement, developed elevation, and dimensioned details of the connections that make the structure viable. In a structure with no cladding and no cover plates, the details are the architecture — every node is on permanent display, which is a far harder brief than hiding a bolted gusset behind a lining.

Two junctions carry most of that burden. The first is where the balustrade lattice meets a baluster, drawn as two sections and an elevation through a single node.

Circular detail drawing, section A-A through a balustrade node, with thin rods arriving at 74 degrees onto the paired hoop rails and a machined fitting at the junction
Detail A, section A–A: 8 mm lattice rods arriving at 74° onto the paired 20 mm hoop rails, gathered into one machined fitting.
Circular detail drawing, section B-B through the same balustrade node cut at right angles to the previous section
Detail A, section B–B: the same node cut at right angles, showing how the fitting gathers the rods onto the rail without a visible plate.
Circular detail drawing showing the balustrade node in elevation, with the vertical baluster crossed by hoop rails and lattice rods at 53 degrees
Detail A, elevation: the node seen face on. The 8 mm baluster runs vertically; the paired 20 mm hoops and the 8 mm lattice rods cross it at 53°, at 20 mm centres.

The second junction is the tread itself, and it is where two decisions I still like become visible. The first is that the staircase is explicitly not designed as a fire escape route, which is precisely what buys the glass its freedom. The second is that the whole assembly is worked out so that a single damaged tread can be released and swapped out — glass being a material you should always assume will one day need replacing.

Circular detail drawing, section A-A through the edge of a glass tread held in a steel shoe, with a strut arriving from below at 11 degrees
Detail B, section A–A: the tread edge held in a 40 mm shoe, with the king-post strut arriving from below at 11° and the curved element on the right running back to the newel.
Circular detail drawing, section B-B showing the laminated glass tread build-up hatched and gripped between steel plates over a baluster foot
Detail B, section B–B: the laminated build-up in section — 5 mm over 12 over 12, with its interlayers — gripped between plates 40 mm wide above a 30 mm baluster foot. Releasing this is what lets a broken tread come out.
Circular detail drawing, section C-C showing the tread bracket meeting the hatched wall of the 300 millimetre newel tube through a concentric collar
Detail B, section C–C: where the tread returns to the newel. The hatched arc on the right is the 300 × 6 mm tube; the bracket fans out 55 mm to a concentric ⌀40 / ⌀20 / ⌀16 / ⌀10 fitting.

What I Learned

Two major takeaways stayed with me.

The first is that the helix does almost all of the real work, and it does it in shear rather than in bending. Under the governing combination the peak shear stress in the newel reaches 113.1 MPa against a design shear resistance of 135.7 MPa — a utilisation of 0.83 — while peak principal stress reaches 178 MPa against a yield of 235 MPa. Nothing else in the structure comes close. The visual reading of a spiral staircase is bending: a stack of little cantilevers. The mechanical reading is a twisted blade of ordinary S235 being worked hard in shear, with a very light stainless cage hanging beside it doing the rest.

The second takeaway is a matter of proportion. Peak tensile stress in the glass reaches 9.8 MPa against a characteristic strength of 45 MPa — the treads are barely troubled. And the schedule at the end of the drawing set totals 680.9 kg for the whole staircase: 414.3 kg of laminated glass against 266.6 kg of steel. The glass outweighs everything holding it up by more than half again. That single line in a weight schedule is the entire argument of the design, made without a word of rhetoric.

And underneath both of those, the thing that has stayed with me longest: this project is a textbook example of absolute symbiosis between architecture and structural engineering. The architectural form is the structural form. One cannot exist without the other; there is no decorative envelope hiding the working steel. This is the discipline that Jiřičná's practice is built on, and it can only be truly appreciated by working through the equations, one element at a time.

Recognition

In 2016, this comprehensive model and analysis received the 2nd prize in the Dlubal Software competition for the best diploma thesis carried out using their structural engineering programs.

Reflection

Re-analyzing someone else's design, when that design is a masterpiece, is one of the most useful exercises a young engineer can undertake. You learn far more from dissecting and reconstructing a correct answer than from generating a new, average one. Authorship here sits exactly where it belongs: with Eva Jiřičná and her engineering team. What I took home was a profound education in how structural logic can become architectural poetry.