The most challenging part of making your own cola is getting a proper and stable emulsion of the flavour oils in the concentrate (and beyond, when actually making cola from that). If the emulsion isn’t all there, any of the following could happen:

• Some visible oil drops remain.
• The oil isn’t perfectly dispersed throughout the concentrate, so each bottle you make from it ends up tasting different (and there likely won’t be any flavour left in the last bottles). This happened to me many times - the flavours were way too strong in the first bottles and I ended up changing the recipe based on that.
• The oils gather on the surface of the concentrate after a while. This is basically the worst case.

Traditionally, emulsions are created by using an emulsifier (gum arabic in the recipes mine is based on) and then adding a lot of mechanical energy to the system – the classic is pumping the mixture through a valve at high pressure. This is called homogenization. It disperses the oil into extremely fine drops, making it easier for the emulsifier to attach to the boundaries and keep things stable. And that’s the vastly simplified version – I’ve discovered that emulsions are quite the complex topic.

There are devices suitable for batch sizes typically used in a lab, based on a rotor-stator design: basically rotating stirrers, but part of the stirring end is stationary, so there are high shearing forces in effect which break up the oil drops. Unfortunately, these devices cost upwards of €1500… that’s quite outside my budget for this product. To add insult to injury, I found sources claiming that they don’t even create very good emulsions.

You can get a somewhat decent emulsion with just a powered hand whisk (which is what I did several times), but the flavours don’t disperse well: the oil drops don’t get small enough, so you’ll have weird outliers in each bottle. I added a bit of alcohol in most batches – essential oils tend to dissolve in alcohol – but even that didn’t help too much.

Other stirring approaches I tried:

• Battery-powered milk frother: the flavour mixture gets a little foamy but it’s extremely hard to get even a half-decent emulsion.
• Powered hand whisk: works much better, but you need to whisk for a fairly long time.
• Magnetic stirrer: normally lab equipment, there are cheap consumer devices that do the same. They’re plates containing a rotating (or, more likely, polarity-changing) electric magnet, paired with a PTFE-coated magnetic rod you put in your mixture. Turn the knob and the rod starts spinning. Unfortunately, these things don’t create a lot of shearing force, so this was a waste of money, if a small one.
• Hand whisk attachment for power drill: I didn’t notice a significant difference compared to the powered whisk.
• High speed rotary tool (“Dremel”): I chose an attachment with a lot of surface area arranged in a chaotic way, originally designed for polishing applications or something like that. Unfortunately, this tends to create a ton of bubbles and isn’t really any better in terms of creating the emulsion.
• Sonication: another approach to adding sufficient energy to small batches to create an emulsion: just apply copious amounts of ultrasound. Let’s just say that consumer devices don’t create enough sound pressure to do anything at all.

Emulsifiers 101

So, if I can’t get better devices to make things happen, I figured I should start looking at different emulsifiers. Here’s a brief intro to how they work.

Emulsifiers are a specific type of surfactant (surface active agent, i.e. something that reduces the surface tension between two liquids). Most people will be more familiar with the other big type: detergents, e.g. soap – but we don’t really want those in our food. (As far as I can tell, there is no real difference in how they work and what they do, but they’re intended for different applications: emulsifiers are for mixing things, detergents are for cleaning things.)

Specifically, emulsifiers are made up of molecules with two ends: one is attracted to water (hydrophilic), the other to fat (lipophilic), meaning they can “glue together” water and oil. You can create an emulsion by just stirring vigorously, but normally it will break down if you let it sit, with the oil drops (or water drops, if you have more oil than water) finding their way back together. Emulsifiers slow this process down by a lot.

Balancing – the traditional model

If you look at how to create emulsions, after a while you’ll find the term HLB - hydrophilic-lipophilic balance. Turns out emulsifiers don’t tend to be perfectly balanced – some attract much more easily to oil, some to water. HLB is a scale invented around the 1950s, going from roughly 1 to 18. At the low end, emulsifiers prefer fat, at the high end they prefer water. An emulsifier with a HLB of less than 10 tends to not dissolve well in water; an emulsifier with a HLB of more than 10 tends to not dissolve well in fat. There is no hard boundary, though.

Low HLB emulsifiers are preferrable for emulsions that are made up of mainly fat (W/O = water-in-oil emulsions); high HLB is preferrable for water-based emulsions (O/W). Emulsifiers in the middle range tend to be suitable as a wetting agent, meaning they allow water to stay dispersed in a solid (which I guess typically means solid fat).

The gum arabic used in the original recipes is cited with a HLB of 8, so it’s slightly more soluble in oil than in water. I can say from personal experience that it doesn’t seem to “dissolve” much in either of the two: put in oil, you get oil with a brownish sludge in it; put in water, you get a murky liquid but with some of the powder still recognizable.

In terms of the HLB scale, gum arabic is in the middle area where its main application should be as a wetting agent. That’s… not actually useful for a beverage concentrate.

Balancing – the more complete model

However, the modern take is that HLB is passé. It was originally designed for emulsifiers derived from PEG (polyethylene glycol) and doesn’t really describe other types of emulsifiers as well – and non-ionic surfactants in particular. Guess what? Gum arabic is non-ionic.

The more up-to-date approach is called HLD – hydrophilic-lipophilic difference. (Apparently it’s up-to-date enough that there isn’t even a Wikipedia article on it.) Here, the idea is that you consider not only the emulsifier and the relative quantities of your aqueous and oily components, but also the specific properties of each liquid and any other ingredients (acidity, salinity) as well as the target temperature.

The HLD scale assigns a positive value to a “system” which will tend towards a W/O (water-in-oil) emulsion, and a negative value to a system which will tend towards an O/W emulsion. At zero, the system is perfectly balanced. Emulsions are least stable around zero, so for a stable O/W emulsion (which is what we want in our concentrate), we want a negative HLD. Since our emulsions will be close to 100% water, it’s hard to end up in the twilight zone – at some point, if you reduce the water part, the emulsion will tend to invert and suddenly turn into a W/O emulsion, but this is pretty much impossible in our situation.

I can tell you right now that it’s next to impossible to get all of this information for all of the ingredients in the recipes. How oily are essential oils? Apparently they’re all in the same general ballpark and not the oiliest oils… but without specific values I don’t think it makes sense to really “run the formula”. So, the really important insight here is that our main concern is staying below zero in terms of HLD. That’s easy: we just need an emulsifier that is sufficiently hydrophilic.

First attempt: gum arabic and soy lecithin co-emulsion

Since I had read that emulsions with multiple emulsifiers often work better, I ordered myself some soy lecithin. Its HLB is around 4, so it’s much more lipophilic than gum arabic. I experimented with this before I found out about the details of HLB and HLD, so in hindsight this was a stupid idea, but intuitively it kind of made sense that we want the emulsifier to bind more strongly to the smaller quantity (the oil) so it can stay at its surface to do its thing. I just used half gum arabic and half soy lecithin and hoped for the best.

I started out dissolving the soy lecithin and gum arabic in oil – first impression: the lecithin started flocculating and it took a lot of stirring to get everything back to a smooth consistency again.

When adding the aqueous phase (first some water, then caramel colouring), something surprising (at the time) happened: the consistency changed, turning very syrupy. Things turned back into the expected watery consistency as I kept adding more caramel colouring, though. In hindsight what happened here was probably that it started out as a W/O emulsion which then inverted into an O/W emulsion as the critical mass of water was reached.

Overall, this was a complete failure. Not only did the oil separate quickly, but the emulsifiers flocculated like crazy when adding the concentrate to sparkling water:

Research: low-energy emulsification methods

Back to the drawing board. I looked at many web resources and several scientific papers. Here’s the gist of it: https://onlinelibrary.wiley.com/doi/full/10.1111/1541-4337.12189

Basically, instead of using the traditional “add tons of energy” approach, there are ways to make the emulsions form almost by themselves, i.e. they would form based on “physicochemical processes”.

One method is called “spontaneous emulsification” – for an O/W emulsion, add emulsifier to oil phase. Slowly add mixture to aqueous phase while stirring. Apparently by adding the emulsifier this way, in the final mixing process where the emulsifier rushes towards the water (which it prefers, remember we picked a hydrophilic emulsifier), it thoroughly shakes things up (there’s no proof that this is what happens, but it seems plausible. Researchers tried putting the emulsifier in the aqueous phase instead, and got much larger oil drops as a result). On the science side of things, you can get oil drop sizes with a diameter of less than 50 nm (nanometers) – this qualifies as a “microemulsion”, gaining the interesting property of being thermodynamically stable, i.e. in theory the emulsions doesn’t separate over time. I don’t know if I’m actually achieving that, but suffice to say this has the potential of creating a very fine emulsion.

The pretty much ideal emulsifier for W/O emulsions, according to various papers, is polysorbate 80 (better known by the brand name “Tween 80”). This is a synthetic emulsifier made from polyethoxylated sorbitan and oleic acid – quite a mouthful. Apparently its molecular structure just works well for it to surround drops very well. It’s suitable for food applications, too. About its only downside is that people don’t like synthetic-sounding stuff in their food, and I guess there aren’t any long-term studies on it. The current limit for food applications (obtained by doing animal testing and dividing the discovered limit by 10) is 25 mg per kg of body weight per day.

As per the research, the main downside of low-energy emulsification is that you need a fair bit more emulsifier to make it work. This means that low-energy emulsification can actually become more expensive than buying a proper homogenizer if you create a lot of emulsions, but I figure I won’t manage that in a lifetime.

One other paper I found tested co-emulsions using both polysorbate 80 and soy lecithin at varying concentrations to further improve emulsion properties. I might look into this in the future but I’m a little wary after my complete failure with soy lecithin. According to some sources I found, lecithin doesn’t like acidic conditions, anyway.

Second attempt: spontaneous emulsification with polysorbate 80

I managed to find a single German vendor for polysorbate 80: Würzteufel. I ordered some and made a new batch. It’s a viscous, sticky liquid with a slightly unpleasant fatty smell, but at these concentrations, who cares…

Based on research results, this batch contained about as much emulsifier as oil, so in my 22.5 litre test batches which contain around 5 ml of essential oils, I used 5 ml of polysorbate 80.

As far as I can tell for now, it worked splendidly. Not only was it easy to mix in with the oils, it also created my first ever flavour emulsion without oil drops separating out, and no oily film on top of the final product. I didn’t need to do tons of whisking either, so for the next batch I’m considering just manually stirring with a spoon or something, just to see if that works, too.

A few bottles into this batch, so far the flavour profile is fairly consistent, suggesting that this is a very evenly distributed emulsion. This was my goal, so if this continues until the end of this batch, I’ll be very happy!

I’ll be putting out a new recipe once I’ve spent some time tweaking the flavour in batches that actually have a consistent flavour.

Future possibilities

This is what I might try in later batches:

• Use a 9:1 polysorbate 80 : soy lecithin mixture as suggested by some sources, but I’m not too hopeful about this after what happened with my first experiment with soy lecithin.
• Stir even less… just to make things faster.
• Try with less polysorbate 80. Maybe for our purposes, half of what I used is still more than enough…
• Add a tiny bit of xanthan as an additional emulsion stabilizer.

One more potential issue: Ostwald ripening

I did find several remarks about essential oils (among others) being susceptible to Ostwald ripening. This seems to mean that over time, larger oil drops will tend to grow and smaller oil drops will tend to shrink. Obviously that can lead to the emulsion breaking down completely, given enough time. I’m not sure yet whether this is an issue.

The common wisdom from our science friends is that the effect can be slowed way down by adding 10% of a vegetable oil rich in long-chain triglycerides (e.g. olive oil), or by adding a co-emulsifier. I might look into this if the concentrate doesn’t stay stable long enough for my purposes.