Golden Ratio for Cell Design

Jerry Wan’s post (right) asks whether cell design has a “golden ratio” for aspect ratio. He concludes that there isn’t a universal one, and that’s right. But the reasons are worth working through, because the best shape changes completely depending on where the tabs are and which edge the heat leaves by. A ratio that is ideal for one layout can be the worst choice for another. To show this, we take one representative cell and run it through five tab and cooling layouts. For each one we sweep the aspect ratio from very wide and short to very tall and narrow.

Figure 1 – The five tab and cooling layouts

We use Jerry’s definition: AR = L / W. L is the electrode dimension measured away from the tab edge (from tab to tab when the tabs are on opposite edges), and W is the dimension along the tab edge. Many engineers think in W/L instead, so both are given below.

The short answer. For a 0.06 m² electrode (about 300 × 200 mm) running at 2C for 30 minutes:

LayoutBest W/LPractical W/L windowPeak rise [K]What sets it
1. Top tabs, no cooling≈ 21 – 5 (flat)≈ 51Tab crowding (wide) vs. foil path (tall)
2. Top tabs, bottom-edge coolingAs short as the tabs allow, ≈ 53 – 620 – 30Heat path L vs. tab crowding
3. Opposite tabs, no cooling≈ 1.50.5 – 3 (flat)≈ 48Same as 1, but more uniform
4. Opposite tabs, one side edge cooled≈ 0.3 (trade-off)0.25 – 0.528 – 33Heat path W vs. foil path L
5. Opposite tabs, both side edges cooled≈ 0.3 (trade-off); thermal-only ≈ 0.10.1 – 0.59 – 14Heat path W/2 vs. foil path L

Resistance on its own favours a squat cell (W/L ≈ 1.5 – 2.5), but that bathtub is shallow. The thermal optimum moves by a factor of 50 or more depending on which edge is cooled, and for edge-cooled opposite-tab cells it lands where real long cells actually sit: W ≈ 90 mm, L ≈ 300 – 500 mm.

The cell and the model

The cell is a stack of double-sided electrodes with 15 µm aluminium and 8 µm copper current collectors. It has 4 mAh/cm² per coated face and an area-specific resistance (ASR) of 15 Ω·cm² per face. The in-plane thermal conductivity is 30 W/m·K, in the middle of the range given in The Thermal Circuit of a Cell. Tabs are limited to 80 mm wide. When both tabs share one edge, each is at most 35% of W; when they sit on opposite edges, each can be up to 70% of W. A cooled edge is modelled with a heat transfer coefficient of 1500 W/m²K on the stack edge, and every other surface is adiabatic.

We solve the two foils as 2D resistive sheets connected through the electrode stack. That gives the local current density, and from it the heat generated in the stack, in the foils and at the tab roots. That heat then feeds a transient in-plane conduction model covering a 30-minute 2C event that starts isothermal. The electrode area is held constant throughout, so a wider cell is also a shorter one.

Jerry Wan: Is there a “Golden Ratio” for cell design? Let’s talk Aspect Ratio (AR).

Many engineers overlook a critical design parameter: the Aspect Ratio (AR = Length / Width). It’s not just an arbitrary number. Especially for pouch cells, AR heavily dictates current paths, thermal profiles, and structural stress.

Here is the engineering reality behind AR design:

1. Current Path & Lithium Plating
Tabs are usually drawn from the short side. A higher AR (longer, thinner cell) means a longer current path from tab to the far end. This increases Ohmic drop. During fast charging, the distal areas hit lithium plating potential much earlier, even at conservative rates.

Rule of thumb: For high-rate pouch cells, either keep AR closer to square, or use multi-tab/multi-side designs to shorten the path.

2. Thermal Field & RTE – Round Trip Efficiency
Pouch cells rely on large surfaces for cooling (no metal shell). The tab area is a natural hotspot. As AR increases, the thermal gradient from tab to distal end widens. Long-term cycling leads to uneven aging.

At CATL’s 587 Tech Day on June 2025, they highlighted how polarization (Ohmic, Activation, Concentration) impacts RTE. Excessive length increases internal resistance. CATL found a classic “bathtub curve” for cell width—too wide or too narrow degrades performance. Their data shows 200-300mm is the “Golden Width.” A 300mm cell can reduce heat by over 20% compared to a 400mm cell.

3. Structural Stress & Swelling
This is where pouch cells fundamentally differ from prismatic/cylindrical. Without a metal shell, the large surface is only constrained by the heat-sealed edges. Think of it as a thin plate fixed on four sides. The higher the AR, the larger the unsupported span in the middle, leading to severe center bulging. An AR close to 1 distributes stress evenly and minimizes single-point deformation.

4. Manufacturing Constraints (Al-laminated film & Stacking)
AR is also silently dictated by process limits. Deep stamping of Al-laminated film risks micro-cracks at corners. Extremely wide electrodes make stacking tension and alignment difficult (wrinkles, tab misalignment). Extremely narrow electrodes increase the number of layers and welding steps, killing efficiency.

The Verdict on AR:
For pouch cells, AR typically falls between 1.2 and 2.5. Drone cells might push 2.5-3.0, but this is a compromise—sacrificing electrochemical performance for spatial fit (requiring thinner coatings, lower loading, and heavy thermal management).
There is no universal “golden ratio.” It’s a multi-objective game (energy density vs. fast charging vs. structural integration).

For prismatic wound / jelly-roll cells, empirical limits are: Height < 250mm, Width < 350mm, Thickness < 80mm.

How do you prioritize AR in your current cell designs? Trade-offs for fast charging or space constraints?

This post has been built based on the support of and sponsorship from: illumion, Custom Power, Quarto Technical Services, Eatron Technologies and About:Energy

The electrical bathtub

Two separate foil losses compete, and the bathtub shape comes from the balance between them.

The right-hand wall: the foil path along L. The current in each foil builds from zero at the far edge to the full cell current at the tab. Integrating that linear build-up gives the familiar one-third factor:

Rcell≈ASRA+(R□,Al+R□,Cu)L3W+RtabR_{cell} \approx \frac{ASR}{A} + \left(R_{\square,Al} + R_{\square,Cu}\right)\frac{L}{3W} + R_{tab}

Here R□ = ρ/t is the sheet resistance of each foil: 1.8 mΩ/□ for 15 µm Al and 2.1 mΩ/□ for 8 µm Cu. Dividing by the intrinsic stack resistance ASR/A turns this into a single dimensionless number:

ν2=(R□,Al+R□,Cu)L2ASRpairRcellASR/A≈1+ν23+…\nu^2 = \frac{\left(R_{\square,Al} + R_{\square,Cu}\right) L^2}{ASR_{pair}} \qquad \frac{R_{cell}}{ASR/A} \approx 1 + \frac{\nu^2}{3} + \ldots

ASRpair is the resistance seen by one pair of foils. Each foil is coated on both sides, so this is half the per-face value. ν² tells you whether the foils matter at all. Below about 0.1 they are negligible and the current is uniform. Near 1, the regions close to the tab start doing noticeably more work than the far edge. For this cell, ν² ≈ 5.2 × L² with L in metres: that’s about 0.3 for a 245 mm square cell and 2.5 for a 690 mm long one. Notice that ν² depends on L², not on L/W. We come back to this below.

The left-hand wall: tab constriction. Once a cell is much wider than its tab can be, current has to crowd sideways into the tab, and the foil near the tab root carries a high current density. This loss grows with the ratio of W to tab width. Since the tab width is capped by welding, sealing and busbar practicalities, the loss rises as the cell gets wider and shorter.

cell aspect ratio and the bathtub curve
Figure 2 – The electrical bathtub: foil-path and tab-constriction losses vs aspect ratio

Looking at resistance alone, the minimum for this cell sits at AR ≈ 0.55 (W/L ≈ 1.8), and the bottom is broad: anything from W/L ≈ 1 to 3 is within a few percent. That is a squat cell, and almost nobody builds one for this reason alone, because the penalty for moving away from it is modest and the thermal and packaging gains are large (see below). At that minimum, foils and tabs still add about 16 – 20% to the intrinsic stack resistance, so the tub has a floor well above 1.

Do opposite tabs halve the foil resistance?

You will often see it claimed that opposite tabs halve the foil loss. They don’t. With uniform reaction current, the positive foil carries I(1 − y/L) and the negative foil carries I·y/L. Swap y for L − y and the two integrals become identical, each giving L/3. That is exactly the same total as with both tabs on one edge. The model agrees. For an earlier 2D treatment of how electrode and tab configuration shape the temperature field in a pouch cell, see [1]. At the same aspect ratio the foil-path term is unchanged. The opposite-tab curve in the figure above sits slightly lower only because each tab can be twice as wide when it has an edge to itself.

What opposite tabs really change is uniformity. Every point on the electrode now sees the same total foil path, positive plus negative, so the local current density evens out. At AR = 4 the ratio of maximum to minimum local current density is 1.71 with tabs on one edge and 1.19 with opposite tabs. At AR = 8 it is 2.54 against 1.38. The high-current zone near a single-edge tab is effectively running at a higher local C-rate than the cell as a whole, which is where lithium plating starts first during a fast charge.

Figure 3 – Resistance and current non-uniformity vs aspect ratio (ν² on top axis)

The thermal bathtubs

Three pieces of physics explain almost everything in the thermal results.

1. How much heat. At 2C this stack generates about 55 kW/m³. With nowhere for the heat to go, the temperature rises by

ΔTadiabatic=q˙′′′tρcp≈55000×18002.5×106≈40 K\Delta T_{adiabatic} = \frac{\dot{q}”’ \, t}{\rho c_p} \approx \frac{55\,000 \times 1800}{2.5 \times 10^{6}} \approx 40\ \text{K}

That is the baseline any cooling layout is trying to beat. The foil and tab losses from the electrical bathtub add to it.

2. How far the heat has to travel. For uniform heat generation conducting to a cooled edge, the steady temperature difference grows with the square of the conduction distance D:

ΔT≈q˙′′′D22k∥\Delta T \approx \frac{\dot{q}”’ D^2}{2 k_{\parallel}}

D is L for a cell cooled on the edge opposite its tabs, W for a cell cooled on one side edge, and W/2 when both side edges are cooled. Because the dependence is squared, halving the heat path cuts the temperature difference to a quarter.

3. How far the cooling reaches in time. Heat diffuses in-plane with α = k/ρcp ≈ 1.2 × 10−5 m²/s, so the depth the cooling can “reach” grows as

δ≈αt≈1.2×10−5×1800≈150 mm\delta \approx \sqrt{\alpha \, t} \approx \sqrt{1.2\times10^{-5} \times 1800} \approx 150\ \text{mm}

During a 30-minute fast charge or discharge, any part of the electrode more than about 150 mm from the cooled edge hardly benefits from the cooling at all. This is the strongest argument for long cooled edges and short heat paths.

Uncooled cells follow different rules. With no edge removing heat, surface ΔT comes only from uneven heat generation. That means the tab hot spot: in wide cells, current crowds into the tab; in tall cells, the whole cell current runs through the foil near the tab end.

Figure 4 – Thermal bathtubs: peak rise and surface ΔT vs aspect ratio for each layout

Case by case

1. Tabs on the top edge, no cooling

This case gives a classic bathtub, and it lines up with the electrical one. The lowest peak temperature and the smallest ΔT both occur around W/L ≈ 2 – 2.4. The peak rise stays within 10% of its best value anywhere from W/L ≈ 1 to 5. At the optimum the face is within about 6 K. Taller cells lose that quickly: ΔT reaches 28 K at AR = 2 and 65 K at AR = 4, with a hot band along the tab edge. In a real pack the busbars pull some of that heat out through the tabs, so these figures are an upper bound.

2. Tabs on the top edge, cooling on the bottom edge

This is the worst combination, because the hottest region (the tabs) is as far from the coolant as it can be. At any given shape, surface ΔT is larger than in the uncooled case 1: the cooling is what creates the gradient. The peak temperature keeps falling as L gets shorter, so there is no thermal minimum here. The limit comes from the other wall. Beyond W/L ≈ 5 – 6, tab crowding adds more than 10% to cell resistance, and 600 mm wide electrodes start running into slitting and stacking limits. For this area, that means a target of L ≈ 100 – 150 mm. If the tabs are going to be the hot end anyway, cooling through the tabs is worth a look: Hunt et al. measured about 9% loss of usable capacity with surface cooling on pouch cells, against about 1% with tab cooling [2].

3. Opposite tabs, no cooling

This is the most uniform layout. Surface ΔT stays at or below about 5 K from W/L ≈ 0.4 to 3, and the optimum is around W/L ≈ 1.4 – 1.7. The peak rise of about 48 K is simply the adiabatic heating plus foil losses, because nothing is removing heat. Opposite tabs spread the heat evenly; they don’t take it away.

4. Opposite tabs, one side edge cooled

Now the heat path is W, so the thermal result keeps improving as the cell narrows, with no real minimum anywhere in the range studied (down to W/L ≈ 0.06). Resistance moves the other way: compared with the best achievable, it is 8% higher at AR = 2, 24% higher at AR = 4 and 58% higher at AR = 8. The balance is W/L ≈ 0.25 – 0.5 (AR 2 – 4), with a peak rise of 28 – 33 K and a face ΔT of 22 – 27 K. Cooling one side leaves two and a half to three times the peak rise of cooling both.

5. Opposite tabs, both side edges cooled

This layout has a genuine thermal bathtub. The peak rise falls as W shrinks, because the heat path is W/2. It then levels off at about 9 K for W/L ≈ 0.07 – 0.2 (AR 5 – 16), where the growing foil losses and the tab hot spot cancel out the benefit of the shorter path. But the electrical bathtub bottoms out at W/L ≈ 1.8, so the designer has to choose between them. AR = 2 gives about 14 K with 8% extra resistance; AR = 4 gives about 10 K with 24% extra. Long cells, with L from 500 to 1000 mm, sit at the thermal end of this curve. They recover the electrical penalty with thicker foils and near full-width tabs, which is the same as pushing ν² down. It’s the direction of travel described in Pouch Cell Design and Edge Cooling.

Where real long cells sit

Most opposite-tab pouch cells are around W ≈ 90 mm with L ≈ 300 mm, and some reach L ≈ 500 mm. W is set largely by module and pack height, and a 90 mm edge also leaves room for a tab that spans most of it, which removes the tab-crowding wall almost entirely. Running those geometries through the same model with both long edges cooled, and comparing each with a square cell of the same area:

Cell (W × L)Resistance vs squarePeak risePeak rise, square cellCurrent max/min
90 × 300 mm, 15 µm Al / 8 µm Cu+9%≈ 4.4 K≈ 9.9 K1.07
90 × 500 mm, 15 µm Al / 8 µm Cu+28%≈ 6.5 K≈ 16.5 K1.20
90 × 500 mm, 25 µm Al / 12 µm Cu+18%≈ 5.5 K≈ 15.5 K1.14

Seen this way, the long cell makes sense. It gives up 10 – 30% in foil resistance, some of which comes back with thicker foils, and in return more than halves the peak temperature rise because the heat path is only 45 mm to each cooled edge. At W/L ≈ 0.18 – 0.3 it sits right in the case 4 and 5 trade-off window above. The squat shape that minimises resistance only wins if the cell isn’t edge cooled at all.

Figure 5 – Temperature fields for all five layouts at AR = 2

It’s the size, not just the ratio

Aspect ratio on its own isn’t really the design variable. ν² depends on L² in absolute terms, as does the heat path D². The cooling penetration depth is fixed in millimetres, and so is the maximum tab width. Repeat the same sweep at half and double the electrode area, and the lowest foil-plus-tab overhead you can reach goes from about 6 – 8% (0.03 m²) to 16 – 20% (0.06 m²) to 38 – 45% (0.12 m²). The tub keeps roughly the same position but gets much deeper, so the cost of getting the shape wrong grows quickly as cells get bigger.

That is the context for CATL’s “golden width” remarks at its 587 Ah energy storage technology day in June 2025. Reports of the event say that wider cells raise internal resistance and make the internal current distribution less uniform, which speeds up local ageing [3]. At 500 Ah and above, the bottom of the bathtub is deep enough for that to matter. We couldn’t find a published source for the specific 200 – 300 mm window or the “more than 20% less heat than a 400 mm cell” figure, so treat those as indicative rather than general.

Limits of this analysis

The kinetics are linear and ASR is fixed. In a real cell, resistance falls as temperature rises, which creates positive feedback and makes hot spots worse. Tabs and faces are treated as adiabatic, and only one 2C event is modelled, not steady cycling. The cooled-edge interface matters a lot. For case 5, changing the edge heat transfer coefficient from 1500 to 500 W/m²K raises the best peak rise from about 9 K to 16 K, and increasing it to 5000 W/m²K lowers it to about 6 K. The parameters represent a typical cell, not any specific product, so treat the numbers as trends. For a fully coupled electrochemical-thermal treatment of cell design and cooling, see [4]. Jerry’s mechanical and manufacturing points, including pouch forming, stacking and electrode slitting, sit on top of all this and aren’t modelled here.

Takeaways

  • Resistance alone favours a squat cell (W/L ≈ 1.5 – 2.5), but that bathtub is shallow. Long edge-cooled cells (W ≈ 90 mm, L ≈ 300 – 500 mm) accept a 10 – 30% foil penalty in exchange for less than half the temperature rise. The foil path sets its tall-cell wall and tab crowding sets its wide-cell wall.
  • The thermal optimum is set by the cooling layout: minimise the distance from the hottest point to the cooled edge, and keep that distance within about 150 mm for fast-charge events.
  • Opposite tabs don’t reduce foil resistance, but they make the current distribution much more even, which matters more for plating.
  • Check ν² before worrying about aspect ratio. If it is well below 0.3, shape is mainly a packaging decision. If it approaches 1, shape, tab layout and foil thickness have to be designed together.

References

  1. U.S. Kim, C.B. Shin, C.-S. Kim, “Effect of electrode configuration on the thermal behaviour of a lithium-polymer battery”, Journal of Power Sources (2008). OSTI record
  2. I. Hunt, Y. Zhao, Y. Patel, G.J. Offer, “Surface Cooling Causes Accelerated Degradation Compared to Tab Cooling for Lithium-Ion Pouch Cells”, Journal of The Electrochemical Society 163 (9) A1846–A1852 (2016). Imperial Spiral
  3. 36Kr, “CATL’s 587Ah Energy Storage-Specific Battery Cells: Officially Enter Mass Production” (reporting the 2025 Energy Storage 587 Technology Day, 10 June 2025). 36Kr
  4. “How to Cool Lithium Ion Batteries: Optimising Cell Design using a Thermally Coupled Model”, Journal of The Electrochemical Society (2019), doi:10.1149/2.0501913jes. IOPscience

See also: Pouch Cell Design and Edge Cooling, The Thermal Circuit of a Cell, Thermal Conductivity of the Active Layers, Lithium Plating, Cell Design, Cell Tab Design

Leave a Comment