Saturday, 21 November 2009

Painting Into A Corner ~ Part 2 - Exploring J3

Zero colours continue to be the subject of interest - and confusion - online. In my previous post on this subject I included a "J3" sample which was actually a facsimile of the 2-6 swatch included in the 8609 joint Army/Navy standard of February 1945. To be precise 2-6 was the colour standard which superceded J3. In respect of J3 itself there is still some doubt concerning the precise appearance of this colour.

In a two-part article at in August 2004 leading artifact collector and Japanese aircraft colour researcher James F Lansdale discussed the IJN paint colour standard J3. In doing so he drew attention to Reisen metal fragments submitted by Robert C Mikesh for a detailed examination by the staff of the Conservation Analytical Laboratory (CAL) of the Smithsonian Institute in 1992. What the report stated is as follows (my emphasis):-

"Based on the information gathered so far, it is likely that the original color of these fragments is not very different from the present color; the present color is slightly more yellow than the original, due to yellowed binder, but, whilst this could cause a shift from white to tan, it would not cause a shift from grey to olive green."

"Not very different" does not imply no difference at all and in fact CAL note the yellowing of the binder and the probable shift from a white to tan this might cause in a hypothetical situation.

However, the report goes further than this:-

"The colorimetric data suggest that the olive and tan paints may have yellowed; the FTIR data identify the binders as ones which yellow over time, and with exposure to sun and heat."

"It is possible to imagine a scenario in which the large fragment might have had a grey color in place of the olive color, and the degradation of the nitrocellulose (which releases nitric acid) caused the pigments to change color. However, it seems extremely unlikely that the small fragment, which has very little nitrocellulose in the binder, would have changed to almost the same color."

"For an accurate determination of the original color, pigment identification would be necessary; this requires the destruction of some of the sample, and would be best done by commercial paint analysis labs."

A copy of the relevant page from the actual report is shown here (Fig.1 above) to demonstrate that the quotations have not been taken out of context.

This report appears to be inconclusive in determining the actual degree of colour shift but more significantly it does not rule it out - in fact it confirms a probable shift. In addition the broad terms "grey" and "olive" used in the report when discussing the probability of colour shift are ambiguous given the subtle colour space occupied by the paint.

The yellowing confirmed by the report may be significant when it comes to the shift of a warm grey to an olive/khaki grey. A "warm" grey is already on the peripheral of this colour space and only a small amount of yellow tinting is required to shift it towards a more olive/khaki appearing grey, especially in visual perception. This yellow tinting, implicit in the original colour, is exacerbated by the age and heat induced yellowing or ambering of the binder over time. It is paradoxical too. UV exposure and other environmental factors (especially heat and moisture) will degrade the upper surface of the paint film, breaking down the molecular structure and breaking up the constituent pigments, resulting in the chalky surface familiar from heavily weathered artifacts. The yellowing in the binder is "bleached" away in exactly the same manner as "yellowed" decals may be corrected by UV exposure. However those areas of the surface protected from UV exposure will still be subject to age and heat induced changes and will gradually show an increased yellowing and darkening over time. Incrementally this shift may be small but in terms of precise colour definition it is significant. When examining an artifact where both conditions of surface are present or where the "chalked" surface has been rubbed back to reveal an underlying level of "original" paint it is necessary to bear in mind that both appearances of colour represent two probable shifts not related to the original appearance of the paint - the one being the degraded paint surface resulting from UV and environmental exposure and the other the yellowed, darkened surface resulting from age and heat induced changes. Therefore the original appearance of the paint surface at the time of application may have been somewhere between the two current appearances suggested by the metal paint example illustrated in Fig.3 above.

Fabric Samples

Throughout the artifact research much has been made of the fact that the fabric surfaces were (are) apparently painted in a neutral grey, sometimes identified as J3, whilst the metal surfaces were painted in the "olive grey" paint. So far there has been no explanation as to why this was done other than an acknowledgement that the painting processes for fabric and metal surfaces were different. That this difference in appearance was contemporary is confirmed by photographs which do appear to show lighter toned flying control surfaces.

It appears also that the yellow (or amber) appearance of the metal paint (a probable consequence of both binder and pigment) was apparent almost immediately after application. Ichiro Hasegawa, in his article 'The Basic Paint of Zero-Sen', comments on this impression:-

"I used to visit the base (Oppama); Zero-sens were grey-green, presumably built by Nakajima. The color was different from that of 97-sen types often seen in Army airfields in Ibaragi. The former were painted in a lighter shade of (grey) green, a little yellow or beige."

The Yokosuka 0266 report reinforces this impression, describing the current Zero fighters as being painted J3 grey slightly tinted towards amber.

Over time UV exposure arrested and reversed this impression making the paint appear more blue-grey. It is probable that the J3 paint used on the Zero was deliberately tinted towards amber as the result of a specific pigment or additive, such as yellow oxide, which would have resulted in both the ambering and a slightly greenish caste to the paint. The variable purity of yellow oxide (hydrated iron oxide) could explain some of the differences seen on Zero artifacts. Over time and on protected surfaces the yellowing and darkening binder would have increased the perception of an olive or khaki colour. The shift from a warm, slightly amber dove grey to the more brownish "olive grey" is a small one in terms of the adjacent colour spaces occupied. The bright, grey-green hues promoted in Japan are much more difficult to reconcile with the physical evidence. The difficulty the Japanese were experiencing in achieving a stable grey paint for the Zero has already been commented on by a leading Japanese researcher.

In another posting about the fabric samples from Reisen s/n 2266 & s/n 5289 Mr Lansdale commented that the colours of these samples were both matched to Munsell 10Y 5/1 or "close to FS 16357", (Fig.3 above).

From a precise colour measurement perspective FS 16357 is not close to 10 Y 5/1. The actual DE2000 difference calculation is 11.5 (where a calculation of 2.0 or less is required to produce a close match). Actually FS 14201 is a closer value, with a calculation of 5.40 (and that particular FS value has featured before in the study of paint colours on Zero artifacts). However, as any calculation more than 2.0 equates inversely proportionately to a close match it may be said that it is not possible to provide a meaningful FS equivalent for 10 Y 5/1 and that the DE2000 calculations for both 16357 and 14201 render them arguably superfluous to any scientific study of these particular fabric samples.

The closest match in a commercial colour standard (other than JPMA which uses Munsell directly) is RAL* 7003 'Moosgrau' (Moss Grey), which is a very close match indeed at only 1.08. In the schematic of rendered chips (Fig.4 above) the sRGB values are shown within each chip and the differences may be noted (especially how far the values of 16357 are from the Munsell chip). The DE2000 calculation from 10 Y 5/1 is given in brackets on each chip. The value 10 Y 5/1 is not obviously J3 either (see below) which rather discourages the idea that the metal airframes were painted an "olive grey" whilst the fabric flying surfaces were painted in the "more neutral" grey of J3. The conclusion, at least from these particular samples, is that the fabric surfaces were painted in a similar hue to the main airframe, a warm, slightly brown (or yellow) looking grey but that the application process or the type of paint used resulted in a lighter toned appearance.

* RAL = Reichsausschuss für Lieferbedingungen" - (Committee of the German Reich for Terms and Conditions of Sale) - from 1925

J3 Described

In addition to the 2-6 chip (and the extant Kariki 117 colour standard sample chip for J3), a chip of the colour was also included in the Yokosuka 0266 report on Zero camouflage schemes, both of which are reproduced at Fig.2 above.

Regarding the actual 2-6 chip in Document 8609, Donald W. Thorpe has matched it to the Munsell value 5 GY 6/1. This 8609 colour was reported to represent a rationalisation of the Army Hai Ryoku Shoku (Ash Green Colour) and the Navy J3 Hai Iro (Ash Colour). The closest FS match to 5 GY 6/1 is 16307 with a DE2000 difference calculation of 2.07 - so it is not exact but it is quite close. However, referencing this value to the Reisen fabric samples above, 5 GY 6/1 is quite a distance from 10 Y 5/1 at 9.63.

Referring to the 2-6 chip the Japanese researcher "Summer" described it as follows:

"It has a faint yellow colour and is somewhere between CN-55 and CN-60 (JPMA). It is approximately ash green colour, identical to J3".

The CN-55 to CN-60 comparison equates approximately to Munsell N5.5 and N6. These are neutral greys (consisting of black and white only) and it is apparent that the comparison is approximate and does not allow for the "faint yellow" impression. The closest FS match to the median of Summer's comparison is 17178 at 2.51 (a metallic colour in FS 595b so useless for direct visual comparison), but closer is RAL 9022 Perlhellgrau (Pearl Light Grey) at 1.60.

In respect of the 117 J3 chip Summer describes it, identically to 2-6, as being somewhere between N5.5 and N6 but with a faint yellowish caste. He notes it as almost identical to the 2-6 chip reproduced in the Gakken book.

Japanese researcher Watanabe-san has not, AFAIK, revealed a comparison for the 117 J3 or the 8609 2-6.

Mr Thorpe's comparison of 5 GY 6/1 appears closer than the approximate neutral greys cited by Summer but the precise appearance of J3 remains problematic. In fact I think those neutral greys may be something of a red herring and the FS/RAL comparisons superfluous to the study. In the chips a faint yellow caste to 5 GY 6/1 is apparent, especially when compared directly to the pure neutral grey of N 5.5-N 6, which is entirely consistent with the historical perception of the ash or ash grey colour (see below). The yellow caste has the effect of making the grey appear slightly greenish to most visual perception. These comparisons are shown in the schematic of rendered chips at Fig.4 above.

When the median value between the two extremes of the colour values of the artifacts (the "chalked" grey and the "olive grey") is computer modelled the resulting colour value is close to Munsell 5 GY 6/2. Now there's a funny thing . . .

An Ash Distraction

The colour description "ash" or "ash grey" is, according to Methuen, one of the oldest in existence. The Munsell identification of this colour as 1B2 brings to mind immediately Summer's description of the J3 chip as having a faint yellow caste. Methuen mention the yellow caste associated with the colour and the adjacent 1B3 would certainly not look out of place on a Zero. The yellow hue, combined with black and white, create the perception of greenishness to the colour, which is largely illusory (although at this stage the use of a direct green tinting additive to the paint cannot be ruled out.).

Painting A Model

In modelling terms it is understandable to take the extant appearance of the artifacts and replicate them exactly. On the other hand it seems some modellers still prefer the traditional whiteish-grey "chalked" colour. In Japan the image of a bright, pale grey-green (the classic hairyokushoku) appears to be favoured (and this perception should be considered in light of the above - especially the 5 GY 6/1 value). The reality was perhaps somewhere in the merging of all three perceptions but I have yet to see a Zero model precisely replicated in the dull, warm dove grey suggested by the Aichi D3A1 painting, not least perhaps because it is so difficult to arrive at this colour without it shifting too much towards one of the other interpretations. Therefore one tends to see models which appear to be too brown, too grey or too green - and often too bright.

The starting point, in terms of pure primary pigmentation, is black and white. To this is added yellow ochre (not yellow) to the degree of "warming" the grey but not turning it too brown - a delicate operation. The choice of white - whether "cold" or "warm" - also affects the final appearance. The choice of yellow ochre will also affect the degree of "greenish" caste, if any, in the final result which should always appear to be slightly more grey than brown. The advantage of using primary colours is in avoiding the odd colour shifts produced by using ready-mixed "authentic" hobby paints or other subtle hues which have untypical and/or tinted pigments. Have fun with this and don't sweat it. When the wet colour looks right make sure the ratio is recorded and then wait for it to dry on a test sample. If this still looks right you are in business. If not you need to calculate the shift from wet to dry, plan accordingly and begin again. If the final result has a very low reflectivity, is difficult to determine in hue from warm grey to brown to green under various illuminants, could be described as "grey slightly towards amber" or "grey mouse colour", suggests a convincing military paint colour from the 1940's and has lustre you can feel justly pleased with your achievement. If it just looks brown you better start again!

Note: The footnote to the caption of the second illustration in the first plate above refers to the translation by Ryutaro Nambu in the article ‘Out of Ameiro Cloud into Hai-Ryokushoku Sky’ by Yoshihito Kurosu. Inevitably there are variations in wording to this version.

Acknowledgement - Since first publication this post has been corrected with valuable input from James F Lansdale.

Image credit: James F Lansdale via; rendered colour chips ©2009 'Straggler'


ILJ said...

Thanks for your excellent summary of a very interesting issue.

Here is my attempt at a realistic model color for the A6M2N (1942 - Kiska Island):

The model was photographed under tungsten illumination with a custom white balance correction. Undiluted Tamiya XF-76 was used for the metal surfaces and XF-12 IJN grey for the fabric surfaces. I believe the result takes into account the desaturating and lightening effects of aerial perspective on the color evidence from the relic aircraft fragments and from contemporary B&W photographs.

What do you think?

Ken Glass said...

Hello Ian,

Two good looking A6M2-N builds.

Ken Glass

Straggler 脱走兵 said...

Hi Ian

I think your website deserves to be better known! What a superb themed collection. I should like to link with permission.

XF-76 is a reasonable choice as it seems to represent an attempt to determine the original colour from the known extremes as related here and is close to the original pigments when mixed. Bear in mind the colour shift over time is probably yellower + darker and the green may be illusory. I'm currently leaning towards duller greys with less green but I realise that may seem counter-intuitive to recent trends.

Kind regards

Paul Flint said...

Thank You for continuing this series...