How do I know if my HPLC flow cell is dirty?

The usual signs are a noisier baseline, lower sample energy than before, and a detector that fails its startup or calibration test. A weak lamp gives the same signs, so clean the cell first. If the energy comes back after cleaning, the lamp wasn't the problem.

What solvent should I use to clean a UV detector flow cell?

Start with your mobile phase. If that doesn't help, use 100% acetonitrile. If the problem stays, Waters recommends 1% formic acid for 30 minutes, then water until the outlet is neutral. Put water between a buffer and pure organic solvent so salt doesn't come out of solution in the cell.

What is the pressure limit of a detector flow cell?

It's low. Waters rates ACQUITY flow cells to 69 bar (1000 psi), and Agilent rates its 1290 DAD cartridge cells to 70 bar. Viscous solvents, high flow rates and anything after the cell raise the pressure, so clean at a modest flow rate.

Why does a new column make the flow cell dirty?

A new column can release contaminants that leave deposits on the cell walls. Waters says to flush a new column to waste, for example 15 minutes or about 10 column volumes, before you connect it to the detector.

How do I tell an air bubble from a dirty flow cell?

A bubble comes and goes, so the noise and spikes change. Dirt stays. Use degassed solvent, have solvent flowing before the lamp turns on, and keep some backpressure after the cell with the regulator or a length of narrow tubing.

Why is my PDA flow cell leaking from inside?

Often because the path after the cell, or inside it, is blocked, so the pressure in the cell rises until the gaskets give. Clear the inlet and outlet first. A Waters 2998 analytical cell can then be rebuilt with kit 700003798 (lenses, retainers and gaskets).

How to clean an HPLC detector flow cell, in the right order

By Dr. Nazmul Alam, analytical chemist · September 19, 2026 · Updated October 4, 2026

The baseline is noisy. The energy reading is low. Maybe the detector fails its startup test in the morning.

The lamp usually gets blamed first. But a dirty flow cell gives the same signs, and cleaning it only costs some solvent and time. A new lamp costs a lot more.

So clean the cell first. If the problem stays, then look at the lamp.

What a dirty cell does

The flow cell is the small chamber where the light passes through your sample. Anything stuck on its walls or windows blocks some of that light. Less light reaches the detector, and the signal gets noisier.

Waters lists the signs of a dirty flow cell as baseline noise, lower sample energy and calibration failure.

Some cells are more sensitive to this. The Waters ACQUITY PDA uses a light-guiding cell. The light travels along the cell by bouncing off the inside wall, so the wall is part of the light path, and a thin film on it costs energy directly. In a conventional cell, the light passes through the middle and mostly doesn’t touch the wall.

Where the dirt comes from

A new column. A new column can release contaminants that leave deposits on the cell walls. Waters says to flush a new column to waste before you connect it to the detector: 15 minutes in the PDA guide, or about 10 column volumes in the operator’s guide. So put the column outlet in a waste beaker first, run the flow, then connect the detector.

Your samples. Matrix, proteins, anything that didn’t fully elute.

Buffer that dried in the cell. This one you can avoid. Flush buffer out with water at the end of the day. The shutdown guide has the order.

The cleaning order

Start with the mildest step. Only go stronger if it didn’t work. This order is from the Waters PDA guide, and the Waters TUV procedure uses the same steps.

flowchart LR
    B["1. Mobile<br/>phase"] -->|"Not better"| C["2. 100%<br/>acetonitrile"]
    C -->|"Not better"| D["3. 1% formic<br/>acid, 30 min"]
    D -->|"Not better"| E["4. System<br/>acid flush"]
    E -->|"Not better"| F["5. Call<br/>service"]

Mobile phase. Flush the cell with your normal mobile phase. Sometimes that’s enough.
100% organic. If there’s no change, flush with pure acetonitrile.
1% formic acid, 30 minutes. Then flush with water until the acid is gone and the outlet is neutral.
System acid flush. Waters has a longer acid flush of the whole system. Don’t run it with a mass spectrometer connected.
Call service. If the cell is still dirty after that, it may need replacing.

Before you start, turn the lamp off, stop the flow and take the column out. Put a union or a short piece of tubing in its place. That’s the Waters procedure.

If you’re going from a buffered mobile phase to acetonitrile, put water in between. Buffer salt meeting pure organic in the cell can leave a deposit.

One disagreement in the manual. The TUV procedure in the operator’s guide says 30 minutes of formic acid on one page and “minimum of 4 hours” at a low flow rate on the next. The PDA guide says 30 minutes. I’d start with 30 minutes, check the energy, and only go longer if it didn’t help.

The pressure limit is low

Waters rates the ACQUITY flow cells to 69 bar (1000 psi). Agilent rates its 1290 DAD cartridge cells to 70 bar. Above that, a cell can break.

That’s low next to the rest of the system, which can run at hundreds of bar. The cell sits after the column, so it only sees the pressure after the column. Things that raise it:

A high flow rate during cleaning
A viscous solvent. Waters says a 1:1 mix of methanol and water runs at about twice the pressure of either one alone
Anything after the detector that adds backpressure

So clean at a modest flow rate, and lower it more with methanol and water.

When the cell leaks from inside

In October 2026 someone posted a Waters 2998 PDA flow cell on Reddit that was leaking from inside the cell assembly. During normal flow, nothing came out of the outlet. They had already flushed it, and reverse flushed it.

If nothing comes out of the outlet, something after the cell, or inside it, is blocked. The pump keeps pushing, the pressure in the cell rises, and the liquid finds the weakest point. In a flow cell that’s the gaskets around the lenses. So the leak is the second fault. The blockage is the first.

flowchart LR
    B["Flow through<br/>the removed cell"] --> C["Check each<br/>tube"]
    C --> D["Clear the<br/>blockage"]
    D --> E["Check for a<br/>cracked lens"]
    E --> F["Fit new<br/>gaskets"]

The 2998 analytical cell can be rebuilt. Waters sells a rebuild kit, part number 700003798, with new lenses, lens retainers, and the two gaskets for the left and right side of the cell body. People in the thread said a rebuild costs much less than a new cell.

They also made the point that decides whether the rebuild lasts: clear the blockage first. If the inlet or outlet tube is still blocked, the new gaskets see the same pressure and fail the same way.

The order:

Take the cell out and flow through it from a pump, off the detector. Watch where the liquid comes out: at the metal fittings, or at the lens housing.
Check the inlet and outlet tubes on their own. Flow should come through each one freely.
Flush the blockage with water first if buffer is a likely cause, then methanol or acetonitrile for organic residue.
Look at the lenses while the cell is open. A cracked lens needs replacing, not only the gaskets.
Only then fit the new gaskets.

Then ask what blocked it. Dried buffer is one likely cause, and the shutdown guide is the prevention. Waters also warns not to connect anything after the cell that can push it over its rating, and a blocked waste line does exactly that.

Air bubbles look like a dirty cell

A bubble in the cell blocks light too, so it also gives low energy and noise. The difference is that bubbles come and go. Dirt stays.

Two things help:

Degassed solvent, flowing, before the lamp turns on. Waters says not to light the lamp on the light-guiding cell when it’s dry or when no solvent is flowing.
A little backpressure after the cell. The detector has a backpressure regulator on its outlet for this. If it was taken off, the PDA guide suggests 30 to 60 cm of 0.23 mm ID tubing on the waste line. To clear bubbles, it asks for at least 17 bar after the cell, and still not more than 69 bar.

The easy way to keep the cell clean: flush buffer out at the end of the day, keep the cell full of clean solvent, and cap the ports if the detector will stand unused.

When it really is the lamp

If the cell is clean, there are no bubbles, and the energy is still low, the lamp is next. The signs and the test are in before you replace the deuterium lamp.

On a Waters 2489 or 2998 the lamp is 201000281, and on a 996 or 2996 PDA it’s WAT052586.

Sources

From vendor documentation:

Waters ACQUITY UPLC PDA and eλ PDA Detector Overview and Maintenance Guide, 715002209 rev A: signs of a dirty cell, the cleaning order, the 69 bar limit, the light-guiding cell, the tubing and pressure for bubbles.
Waters ACQUITY UPLC PDA Detector Getting Started Guide, 71500108703 rev A: flushing a new column before connecting the cell.
Waters ACQUITY UPLC System Operator’s Guide, 71500082502 rev F: the TUV cleaning procedure, the 30 minute and 4 hour figures, conditioning a column to waste, methanol and water viscosity.
Agilent InfinityLab LC Series Specification Compendium, 01200-90062: the 70 bar rating of the 1290 DAD cartridge cells.
Waters kit 700003798, 2998 / Alliance iS PDA analytical flow cell rebuild kit: the kit contents.

Not from vendor documentation, so treat it accordingly:

The leaking 2998 case is from a thread in Reddit’s chromatography community, 1 October 2026. The advice to clear the tubes before fitting new gaskets, to flow through the removed cell to find the leak, to check for a cracked lens, and that a rebuild costs less than a new cell all come from people in that thread. Waters’ own rebuild procedure for this cell needs a customer login, and I haven’t read it.

Checked 3 October 2026. The cleaning steps are from Waters’ ACQUITY manuals. Other detectors have different cells and different ratings, so check your own manual.